From c10302efe569bfd646b4c22df29577a4595b4580 Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Tue, 31 Jul 2012 16:23:59 +0200 Subject: s390/bpf,jit: BPF Just In Time compiler for s390 The s390 implementation of the JIT compiler for packet filter speedup. Reviewed-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kbuild b/arch/s390/Kbuild index 9858476..cc45d25 100644 --- a/arch/s390/Kbuild +++ b/arch/s390/Kbuild @@ -5,3 +5,4 @@ obj-$(CONFIG_CRYPTO_HW) += crypto/ obj-$(CONFIG_S390_HYPFS_FS) += hypfs/ obj-$(CONFIG_APPLDATA_BASE) += appldata/ obj-$(CONFIG_MATHEMU) += math-emu/ +obj-y += net/ diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 107610e..a7088dc 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -84,6 +84,7 @@ config S390 select HAVE_KERNEL_XZ select HAVE_ARCH_MUTEX_CPU_RELAX select HAVE_ARCH_JUMP_LABEL if !MARCH_G5 + select HAVE_BPF_JIT if 64BIT && PACK_STACK select ARCH_SAVE_PAGE_KEYS if HIBERNATION select ARCH_HAS_ATOMIC64_DEC_IF_POSITIVE select HAVE_MEMBLOCK diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h index 11e4e32..d4477ba 100644 --- a/arch/s390/include/asm/processor.h +++ b/arch/s390/include/asm/processor.h @@ -140,6 +140,7 @@ extern int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags); extern unsigned long thread_saved_pc(struct task_struct *t); extern void show_code(struct pt_regs *regs); +extern void print_fn_code(unsigned char *code, unsigned long len); unsigned long get_wchan(struct task_struct *p); #define task_pt_regs(tsk) ((struct pt_regs *) \ diff --git a/arch/s390/kernel/dis.c b/arch/s390/kernel/dis.c index 619c5d3..84fd7e9 100644 --- a/arch/s390/kernel/dis.c +++ b/arch/s390/kernel/dis.c @@ -1601,3 +1601,26 @@ void show_code(struct pt_regs *regs) } printk("\n"); } + +void print_fn_code(unsigned char *code, unsigned long len) +{ + char buffer[64], *ptr; + int opsize, i; + + while (len) { + ptr = buffer; + opsize = insn_length(*code); + ptr += sprintf(ptr, "%p: ", code); + for (i = 0; i < opsize; i++) + ptr += sprintf(ptr, "%02x", code[i]); + *ptr++ = '\t'; + if (i < 4) + *ptr++ = '\t'; + ptr += print_insn(ptr, code, (unsigned long) code); + *ptr++ = '\n'; + *ptr++ = 0; + printk(buffer); + code += opsize; + len -= opsize; + } +} diff --git a/arch/s390/net/Makefile b/arch/s390/net/Makefile new file mode 100644 index 0000000..90568c3 --- /dev/null +++ b/arch/s390/net/Makefile @@ -0,0 +1,4 @@ +# +# Arch-specific network modules +# +obj-$(CONFIG_BPF_JIT) += bpf_jit.o bpf_jit_comp.o diff --git a/arch/s390/net/bpf_jit.S b/arch/s390/net/bpf_jit.S new file mode 100644 index 0000000..7e45d13 --- /dev/null +++ b/arch/s390/net/bpf_jit.S @@ -0,0 +1,130 @@ +/* + * BPF Jit compiler for s390, help functions. + * + * Copyright IBM Corp. 2012 + * + * Author(s): Martin Schwidefsky + */ +#include + +/* + * Calling convention: + * registers %r2, %r6-%r8, %r10-%r11, %r13, %r15 are call saved + * %r2: skb pointer + * %r3: offset parameter + * %r5: BPF A accumulator + * %r8: return address + * %r9: save register for skb pointer + * %r10: skb->data + * %r11: skb->len - skb->data_len (headlen) + * %r12: BPF X accumulator + * + * skb_copy_bits takes 4 parameters: + * %r2 = skb pointer + * %r3 = offset into skb data + * %r4 = length to copy + * %r5 = pointer to temp buffer + */ +#define SKBDATA %r8 + + /* A = *(u32 *) (skb->data+K+X) */ +ENTRY(sk_load_word_ind) + ar %r3,%r12 # offset += X + bmr %r8 # < 0 -> return with cc + + /* A = *(u32 *) (skb->data+K) */ +ENTRY(sk_load_word) + llgfr %r1,%r3 # extend offset + ahi %r3,4 # offset + 4 + clr %r11,%r3 # hlen <= offset + 4 ? + jl sk_load_word_slow + l %r5,0(%r1,%r10) # get word from skb + xr %r1,%r1 # set cc to zero + br %r8 + +sk_load_word_slow: + lgr %r9,%r2 # save %r2 + lhi %r4,4 # 4 bytes + la %r5,160(%r15) # pointer to temp buffer + brasl %r14,skb_copy_bits # get data from skb + l %r5,160(%r15) # load result from temp buffer + ltgr %r2,%r2 # set cc to (%r2 != 0) + lgr %r2,%r9 # restore %r2 + br %r8 + + /* A = *(u16 *) (skb->data+K+X) */ +ENTRY(sk_load_half_ind) + ar %r3,%r12 # offset += X + bmr %r8 # < 0 -> return with cc + + /* A = *(u16 *) (skb->data+K) */ +ENTRY(sk_load_half) + llgfr %r1,%r3 # extend offset + ahi %r3,2 # offset + 2 + clr %r11,%r3 # hlen <= offset + 2 ? + jl sk_load_half_slow + llgh %r5,0(%r1,%r10) # get half from skb + xr %r1,%r1 # set cc to zero + br %r8 + +sk_load_half_slow: + lgr %r9,%r2 # save %r2 + lhi %r4,2 # 2 bytes + la %r5,162(%r15) # pointer to temp buffer + brasl %r14,skb_copy_bits # get data from skb + xc 160(2,%r15),160(%r15) + l %r5,160(%r15) # load result from temp buffer + ltgr %r2,%r2 # set cc to (%r2 != 0) + lgr %r2,%r9 # restore %r2 + br %r8 + + /* A = *(u8 *) (skb->data+K+X) */ +ENTRY(sk_load_byte_ind) + ar %r3,%r12 # offset += X + bmr %r8 # < 0 -> return with cc + + /* A = *(u8 *) (skb->data+K) */ +ENTRY(sk_load_byte) + llgfr %r1,%r3 # extend offset + clr %r11,%r3 # hlen < offset ? + jle sk_load_byte_slow + lhi %r5,0 + ic %r5,0(%r1,%r10) # get byte from skb + xr %r1,%r1 # set cc to zero + br %r8 + +sk_load_byte_slow: + lgr %r9,%r2 # save %r2 + lhi %r4,1 # 1 bytes + la %r5,163(%r15) # pointer to temp buffer + brasl %r14,skb_copy_bits # get data from skb + xc 160(3,%r15),160(%r15) + l %r5,160(%r15) # load result from temp buffer + ltgr %r2,%r2 # set cc to (%r2 != 0) + lgr %r2,%r9 # restore %r2 + br %r8 + + /* A = (*(u8 *)(skb->data+K) & 0xf) << 2 */ +ENTRY(sk_load_byte_msh) + llgfr %r1,%r3 # extend offset + clr %r11,%r3 # hlen < offset ? + jle sk_load_byte_slow + lhi %r12,0 + ic %r12,0(%r1,%r10) # get byte from skb + nill %r12,0x0f + sll %r12,2 + xr %r1,%r1 # set cc to zero + br %r8 + +sk_load_byte_msh_slow: + lgr %r9,%r2 # save %r2 + lhi %r4,2 # 2 bytes + la %r5,162(%r15) # pointer to temp buffer + brasl %r14,skb_copy_bits # get data from skb + xc 160(3,%r15),160(%r15) + l %r12,160(%r15) # load result from temp buffer + nill %r12,0x0f + sll %r12,2 + ltgr %r2,%r2 # set cc to (%r2 != 0) + lgr %r2,%r9 # restore %r2 + br %r8 diff --git a/arch/s390/net/bpf_jit_comp.c b/arch/s390/net/bpf_jit_comp.c new file mode 100644 index 0000000..b07aa3d --- /dev/null +++ b/arch/s390/net/bpf_jit_comp.c @@ -0,0 +1,738 @@ +/* + * BPF Jit compiler for s390. + * + * Copyright IBM Corp. 2012 + * + * Author(s): Martin Schwidefsky + */ +#include +#include +#include +#include +#include + +/* + * Conventions: + * %r2 = skb pointer + * %r3 = offset parameter + * %r4 = scratch register / length parameter + * %r5 = BPF A accumulator + * %r8 = return address + * %r9 = save register for skb pointer + * %r10 = skb->data + * %r11 = skb->len - skb->data_len (headlen) + * %r12 = BPF X accumulator + * %r13 = literal pool pointer + * 0(%r15) - 63(%r15) scratch memory array with BPF_MEMWORDS + */ +int bpf_jit_enable __read_mostly; + +/* + * assembly code in arch/x86/net/bpf_jit.S + */ +extern u8 sk_load_word[], sk_load_half[], sk_load_byte[], sk_load_byte_msh[]; +extern u8 sk_load_word_ind[], sk_load_half_ind[], sk_load_byte_ind[]; + +struct bpf_jit { + unsigned int seen; + u8 *start; + u8 *prg; + u8 *mid; + u8 *lit; + u8 *end; + u8 *base_ip; + u8 *ret0_ip; + u8 *exit_ip; + unsigned int off_load_word; + unsigned int off_load_half; + unsigned int off_load_byte; + unsigned int off_load_bmsh; + unsigned int off_load_iword; + unsigned int off_load_ihalf; + unsigned int off_load_ibyte; +}; + +#define BPF_SIZE_MAX 4096 /* Max size for program */ + +#define SEEN_DATAREF 1 /* might call external helpers */ +#define SEEN_XREG 2 /* ebx is used */ +#define SEEN_MEM 4 /* use mem[] for temporary storage */ +#define SEEN_RET0 8 /* pc_ret0 points to a valid return 0 */ +#define SEEN_LITERAL 16 /* code uses literals */ +#define SEEN_LOAD_WORD 32 /* code uses sk_load_word */ +#define SEEN_LOAD_HALF 64 /* code uses sk_load_half */ +#define SEEN_LOAD_BYTE 128 /* code uses sk_load_byte */ +#define SEEN_LOAD_BMSH 256 /* code uses sk_load_byte_msh */ +#define SEEN_LOAD_IWORD 512 /* code uses sk_load_word_ind */ +#define SEEN_LOAD_IHALF 1024 /* code uses sk_load_half_ind */ +#define SEEN_LOAD_IBYTE 2048 /* code uses sk_load_byte_ind */ + +#define EMIT2(op) \ +({ \ + if (jit->prg + 2 <= jit->mid) \ + *(u16 *) jit->prg = op; \ + jit->prg += 2; \ +}) + +#define EMIT4(op) \ +({ \ + if (jit->prg + 4 <= jit->mid) \ + *(u32 *) jit->prg = op; \ + jit->prg += 4; \ +}) + +#define EMIT4_DISP(op, disp) \ +({ \ + unsigned int __disp = (disp) & 0xfff; \ + EMIT4(op | __disp); \ +}) + +#define EMIT4_IMM(op, imm) \ +({ \ + unsigned int __imm = (imm) & 0xffff; \ + EMIT4(op | __imm); \ +}) + +#define EMIT4_PCREL(op, pcrel) \ +({ \ + long __pcrel = ((pcrel) >> 1) & 0xffff; \ + EMIT4(op | __pcrel); \ +}) + +#define EMIT6(op1, op2) \ +({ \ + if (jit->prg + 6 <= jit->mid) { \ + *(u32 *) jit->prg = op1; \ + *(u16 *) (jit->prg + 4) = op2; \ + } \ + jit->prg += 6; \ +}) + +#define EMIT6_DISP(op1, op2, disp) \ +({ \ + unsigned int __disp = (disp) & 0xfff; \ + EMIT6(op1 | __disp, op2); \ +}) + +#define EMIT_CONST(val) \ +({ \ + unsigned int ret; \ + ret = (unsigned int) (jit->lit - jit->base_ip); \ + jit->seen |= SEEN_LITERAL; \ + if (jit->lit + 4 <= jit->end) \ + *(u32 *) jit->lit = val; \ + jit->lit += 4; \ + ret; \ +}) + +#define EMIT_FN_CONST(bit, fn) \ +({ \ + unsigned int ret; \ + ret = (unsigned int) (jit->lit - jit->base_ip); \ + if (jit->seen & bit) { \ + jit->seen |= SEEN_LITERAL; \ + if (jit->lit + 8 <= jit->end) \ + *(void **) jit->lit = fn; \ + jit->lit += 8; \ + } \ + ret; \ +}) + +static void bpf_jit_prologue(struct bpf_jit *jit) +{ + /* Save registers and create stack frame if necessary */ + if (jit->seen & SEEN_DATAREF) { + /* stmg %r8,%r15,88(%r15) */ + EMIT6(0xeb8ff058, 0x0024); + /* lgr %r14,%r15 */ + EMIT4(0xb90400ef); + /* ahi %r15, */ + EMIT4_IMM(0xa7fa0000, (jit->seen & SEEN_MEM) ? -112 : -80); + /* stg %r14,152(%r15) */ + EMIT6(0xe3e0f098, 0x0024); + } else if ((jit->seen & SEEN_XREG) && (jit->seen & SEEN_LITERAL)) + /* stmg %r12,%r13,120(%r15) */ + EMIT6(0xebcdf078, 0x0024); + else if (jit->seen & SEEN_XREG) + /* stg %r12,120(%r15) */ + EMIT6(0xe3c0f078, 0x0024); + else if (jit->seen & SEEN_LITERAL) + /* stg %r13,128(%r15) */ + EMIT6(0xe3d0f080, 0x0024); + + /* Setup literal pool */ + if (jit->seen & SEEN_LITERAL) { + /* basr %r13,0 */ + EMIT2(0x0dd0); + jit->base_ip = jit->prg; + } + jit->off_load_word = EMIT_FN_CONST(SEEN_LOAD_WORD, sk_load_word); + jit->off_load_half = EMIT_FN_CONST(SEEN_LOAD_HALF, sk_load_half); + jit->off_load_byte = EMIT_FN_CONST(SEEN_LOAD_BYTE, sk_load_byte); + jit->off_load_bmsh = EMIT_FN_CONST(SEEN_LOAD_BMSH, sk_load_byte_msh); + jit->off_load_iword = EMIT_FN_CONST(SEEN_LOAD_IWORD, sk_load_word_ind); + jit->off_load_ihalf = EMIT_FN_CONST(SEEN_LOAD_IHALF, sk_load_half_ind); + jit->off_load_ibyte = EMIT_FN_CONST(SEEN_LOAD_IBYTE, sk_load_byte_ind); + + /* Filter needs to access skb data */ + if (jit->seen & SEEN_DATAREF) { + /* l %r11,(%r2) */ + EMIT4_DISP(0x58b02000, offsetof(struct sk_buff, len)); + /* s %r11,(%r2) */ + EMIT4_DISP(0x5bb02000, offsetof(struct sk_buff, data_len)); + /* lg %r10,(%r2) */ + EMIT6_DISP(0xe3a02000, 0x0004, + offsetof(struct sk_buff, data)); + } +} + +static void bpf_jit_epilogue(struct bpf_jit *jit) +{ + /* Return 0 */ + if (jit->seen & SEEN_RET0) { + jit->ret0_ip = jit->prg; + /* lghi %r2,0 */ + EMIT4(0xa7290000); + } + jit->exit_ip = jit->prg; + /* Restore registers */ + if (jit->seen & SEEN_DATAREF) + /* lmg %r8,%r15,(%r15) */ + EMIT6_DISP(0xeb8ff000, 0x0004, + (jit->seen & SEEN_MEM) ? 200 : 168); + else if ((jit->seen & SEEN_XREG) && (jit->seen & SEEN_LITERAL)) + /* lmg %r12,%r13,120(%r15) */ + EMIT6(0xebcdf078, 0x0004); + else if (jit->seen & SEEN_XREG) + /* lg %r12,120(%r15) */ + EMIT6(0xe3c0f078, 0x0004); + else if (jit->seen & SEEN_LITERAL) + /* lg %r13,128(%r15) */ + EMIT6(0xe3d0f080, 0x0004); + /* br %r14 */ + EMIT2(0x07fe); +} + +/* + * make sure we dont leak kernel information to user + */ +static void bpf_jit_noleaks(struct bpf_jit *jit, struct sock_filter *filter) +{ + /* Clear temporary memory if (seen & SEEN_MEM) */ + if (jit->seen & SEEN_MEM) + /* xc 0(64,%r15),0(%r15) */ + EMIT6(0xd73ff000, 0xf000); + /* Clear X if (seen & SEEN_XREG) */ + if (jit->seen & SEEN_XREG) + /* lhi %r12,0 */ + EMIT4(0xa7c80000); + /* Clear A if the first register does not set it. */ + switch (filter[0].code) { + case BPF_S_LD_W_ABS: + case BPF_S_LD_H_ABS: + case BPF_S_LD_B_ABS: + case BPF_S_LD_W_LEN: + case BPF_S_LD_W_IND: + case BPF_S_LD_H_IND: + case BPF_S_LD_B_IND: + case BPF_S_LDX_B_MSH: + case BPF_S_LD_IMM: + case BPF_S_LD_MEM: + case BPF_S_MISC_TXA: + case BPF_S_ANC_PROTOCOL: + case BPF_S_ANC_PKTTYPE: + case BPF_S_ANC_IFINDEX: + case BPF_S_ANC_MARK: + case BPF_S_ANC_QUEUE: + case BPF_S_ANC_HATYPE: + case BPF_S_ANC_RXHASH: + case BPF_S_ANC_CPU: + case BPF_S_RET_K: + /* first instruction sets A register */ + break; + default: /* A = 0 */ + /* lhi %r5,0 */ + EMIT4(0xa7580000); + } +} + +static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, + unsigned int *addrs, int i, int last) +{ + unsigned int K; + int offset; + unsigned int mask; + + K = filter->k; + switch (filter->code) { + case BPF_S_ALU_ADD_X: /* A += X */ + jit->seen |= SEEN_XREG; + /* ar %r5,%r12 */ + EMIT2(0x1a5c); + break; + case BPF_S_ALU_ADD_K: /* A += K */ + if (!K) + break; + if (K <= 16383) + /* ahi %r5, */ + EMIT4_IMM(0xa75a0000, K); + else + /* a %r5,(%r13) */ + EMIT4_DISP(0x5a50d000, EMIT_CONST(K)); + break; + case BPF_S_ALU_SUB_X: /* A -= X */ + jit->seen |= SEEN_XREG; + /* sr %r5,%r12 */ + EMIT2(0x1b5c); + break; + case BPF_S_ALU_SUB_K: /* A -= K */ + if (!K) + break; + if (K <= 16384) + /* ahi %r5,-K */ + EMIT4_IMM(0xa75a0000, -K); + else + /* s %r5,(%r13) */ + EMIT4_DISP(0x5b50d000, EMIT_CONST(K)); + break; + case BPF_S_ALU_MUL_X: /* A *= X */ + jit->seen |= SEEN_XREG; + /* msr %r5,%r12 */ + EMIT4(0xb252005c); + break; + case BPF_S_ALU_MUL_K: /* A *= K */ + if (K <= 16383) + /* mhi %r5,K */ + EMIT4_IMM(0xa75c0000, K); + else + /* ms %r5,(%r13) */ + EMIT4_DISP(0x7150d000, EMIT_CONST(K)); + break; + case BPF_S_ALU_DIV_X: /* A /= X */ + jit->seen |= SEEN_XREG | SEEN_RET0; + /* ltr %r12,%r12 */ + EMIT2(0x12cc); + /* jz */ + EMIT4_PCREL(0xa7840000, (jit->ret0_ip - jit->prg)); + /* lhi %r4,0 */ + EMIT4(0xa7480000); + /* dr %r4,%r12 */ + EMIT2(0x1d4c); + break; + case BPF_S_ALU_DIV_K: /* A = reciprocal_divide(A, K) */ + /* m %r4,(%r13) */ + EMIT4_DISP(0x5c40d000, EMIT_CONST(K)); + /* lr %r5,%r4 */ + EMIT2(0x1854); + break; + case BPF_S_ALU_AND_X: /* A &= X */ + jit->seen |= SEEN_XREG; + /* nr %r5,%r12 */ + EMIT2(0x145c); + break; + case BPF_S_ALU_AND_K: /* A &= K */ + /* n %r5,(%r13) */ + EMIT4_DISP(0x5450d000, EMIT_CONST(K)); + break; + case BPF_S_ALU_OR_X: /* A |= X */ + jit->seen |= SEEN_XREG; + /* or %r5,%r12 */ + EMIT2(0x165c); + break; + case BPF_S_ALU_OR_K: /* A |= K */ + /* o %r5,(%r13) */ + EMIT4_DISP(0x5650d000, EMIT_CONST(K)); + break; + case BPF_S_ALU_LSH_X: /* A <<= X; */ + jit->seen |= SEEN_XREG; + /* sll %r5,0(%r12) */ + EMIT4(0x8950c000); + break; + case BPF_S_ALU_LSH_K: /* A <<= K */ + if (K == 0) + break; + /* sll %r5,K */ + EMIT4_DISP(0x89500000, K); + break; + case BPF_S_ALU_RSH_X: /* A >>= X; */ + jit->seen |= SEEN_XREG; + /* srl %r5,0(%r12) */ + EMIT4(0x8850c000); + break; + case BPF_S_ALU_RSH_K: /* A >>= K; */ + if (K == 0) + break; + /* srl %r5,K */ + EMIT4_DISP(0x88500000, K); + break; + case BPF_S_ALU_NEG: /* A = -A */ + /* lnr %r5,%r5 */ + EMIT2(0x1155); + break; + case BPF_S_JMP_JA: /* ip += K */ + offset = addrs[i + K] + jit->start - jit->prg; + EMIT4_PCREL(0xa7f40000, offset); + break; + case BPF_S_JMP_JGT_K: /* ip += (A > K) ? jt : jf */ + mask = 0x200000; /* jh */ + goto kbranch; + case BPF_S_JMP_JGE_K: /* ip += (A >= K) ? jt : jf */ + mask = 0xa00000; /* jhe */ + goto kbranch; + case BPF_S_JMP_JEQ_K: /* ip += (A == K) ? jt : jf */ + mask = 0x800000; /* je */ +kbranch: /* Emit compare if the branch targets are different */ + if (filter->jt != filter->jf) { + if (K <= 16383) + /* chi %r5, */ + EMIT4_IMM(0xa75e0000, K); + else + /* c %r5,(%r13) */ + EMIT4_DISP(0x5950d000, EMIT_CONST(K)); + } +branch: if (filter->jt == filter->jf) { + if (filter->jt == 0) + break; + /* j */ + offset = addrs[i + filter->jt] + jit->start - jit->prg; + EMIT4_PCREL(0xa7f40000, offset); + break; + } + if (filter->jt != 0) { + /* brc , */ + offset = addrs[i + filter->jt] + jit->start - jit->prg; + EMIT4_PCREL(0xa7040000 | mask, offset); + } + if (filter->jf != 0) { + /* brc , */ + offset = addrs[i + filter->jf] + jit->start - jit->prg; + EMIT4_PCREL(0xa7040000 | (mask ^ 0xf00000), offset); + } + break; + case BPF_S_JMP_JSET_K: /* ip += (A & K) ? jt : jf */ + mask = 0x700000; /* jnz */ + /* Emit test if the branch targets are different */ + if (filter->jt != filter->jf) { + if (K > 65535) { + /* lr %r4,%r5 */ + EMIT2(0x1845); + /* n %r4,(%r13) */ + EMIT4_DISP(0x5440d000, EMIT_CONST(K)); + } else + /* tmll %r5,K */ + EMIT4_IMM(0xa7510000, K); + } + goto branch; + case BPF_S_JMP_JGT_X: /* ip += (A > X) ? jt : jf */ + mask = 0x200000; /* jh */ + goto xbranch; + case BPF_S_JMP_JGE_X: /* ip += (A >= X) ? jt : jf */ + mask = 0xa00000; /* jhe */ + goto xbranch; + case BPF_S_JMP_JEQ_X: /* ip += (A == X) ? jt : jf */ + mask = 0x800000; /* je */ +xbranch: /* Emit compare if the branch targets are different */ + if (filter->jt != filter->jf) { + jit->seen |= SEEN_XREG; + /* cr %r5,%r12 */ + EMIT2(0x195c); + } + goto branch; + case BPF_S_JMP_JSET_X: /* ip += (A & X) ? jt : jf */ + mask = 0x700000; /* jnz */ + /* Emit test if the branch targets are different */ + if (filter->jt != filter->jf) { + jit->seen |= SEEN_XREG; + /* lr %r4,%r5 */ + EMIT2(0x1845); + /* nr %r4,%r12 */ + EMIT2(0x144c); + } + goto branch; + case BPF_S_LD_W_ABS: /* A = *(u32 *) (skb->data+K) */ + jit->seen |= SEEN_DATAREF | SEEN_RET0 | SEEN_LOAD_WORD; + offset = jit->off_load_word; + goto load_abs; + case BPF_S_LD_H_ABS: /* A = *(u16 *) (skb->data+K) */ + jit->seen |= SEEN_DATAREF | SEEN_RET0 | SEEN_LOAD_HALF; + offset = jit->off_load_half; + goto load_abs; + case BPF_S_LD_B_ABS: /* A = *(u8 *) (skb->data+K) */ + jit->seen |= SEEN_DATAREF | SEEN_RET0 | SEEN_LOAD_BYTE; + offset = jit->off_load_byte; +load_abs: if ((int) K < 0) + goto out; +call_fn: /* lg %r1,(%r13) */ + EMIT6_DISP(0xe310d000, 0x0004, offset); + /* l %r3,(%r13) */ + EMIT4_DISP(0x5830d000, EMIT_CONST(K)); + /* basr %r8,%r1 */ + EMIT2(0x0d81); + /* jnz */ + EMIT4_PCREL(0xa7740000, (jit->ret0_ip - jit->prg)); + break; + case BPF_S_LD_W_IND: /* A = *(u32 *) (skb->data+K+X) */ + jit->seen |= SEEN_DATAREF | SEEN_RET0 | SEEN_LOAD_IWORD; + offset = jit->off_load_iword; + goto call_fn; + case BPF_S_LD_H_IND: /* A = *(u16 *) (skb->data+K+X) */ + jit->seen |= SEEN_DATAREF | SEEN_RET0 | SEEN_LOAD_IHALF; + offset = jit->off_load_ihalf; + goto call_fn; + case BPF_S_LD_B_IND: /* A = *(u8 *) (skb->data+K+X) */ + jit->seen |= SEEN_DATAREF | SEEN_RET0 | SEEN_LOAD_IBYTE; + offset = jit->off_load_ibyte; + goto call_fn; + case BPF_S_LDX_B_MSH: + /* X = (*(u8 *)(skb->data+K) & 0xf) << 2 */ + jit->seen |= SEEN_RET0; + if ((int) K < 0) { + /* j */ + EMIT4_PCREL(0xa7f40000, (jit->ret0_ip - jit->prg)); + break; + } + jit->seen |= SEEN_DATAREF | SEEN_LOAD_BMSH; + offset = jit->off_load_bmsh; + goto call_fn; + case BPF_S_LD_W_LEN: /* A = skb->len; */ + BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, len) != 4); + /* l %r5,(%r2) */ + EMIT4_DISP(0x58502000, offsetof(struct sk_buff, len)); + break; + case BPF_S_LDX_W_LEN: /* X = skb->len; */ + jit->seen |= SEEN_XREG; + /* l %r12,(%r2) */ + EMIT4_DISP(0x58c02000, offsetof(struct sk_buff, len)); + break; + case BPF_S_LD_IMM: /* A = K */ + if (K <= 16383) + /* lhi %r5,K */ + EMIT4_IMM(0xa7580000, K); + else + /* l %r5,(%r13) */ + EMIT4_DISP(0x5850d000, EMIT_CONST(K)); + break; + case BPF_S_LDX_IMM: /* X = K */ + jit->seen |= SEEN_XREG; + if (K <= 16383) + /* lhi %r12, */ + EMIT4_IMM(0xa7c80000, K); + else + /* l %r12,(%r13) */ + EMIT4_DISP(0x58c0d000, EMIT_CONST(K)); + break; + case BPF_S_LD_MEM: /* A = mem[K] */ + jit->seen |= SEEN_MEM; + /* l %r5,(%r15) */ + EMIT4_DISP(0x5850f000, + (jit->seen & SEEN_DATAREF) ? 160 + K*4 : K*4); + break; + case BPF_S_LDX_MEM: /* X = mem[K] */ + jit->seen |= SEEN_XREG | SEEN_MEM; + /* l %r12,(%r15) */ + EMIT4_DISP(0x58c0f000, + (jit->seen & SEEN_DATAREF) ? 160 + K*4 : K*4); + break; + case BPF_S_MISC_TAX: /* X = A */ + jit->seen |= SEEN_XREG; + /* lr %r12,%r5 */ + EMIT2(0x18c5); + break; + case BPF_S_MISC_TXA: /* A = X */ + jit->seen |= SEEN_XREG; + /* lr %r5,%r12 */ + EMIT2(0x185c); + break; + case BPF_S_RET_K: + if (K == 0) { + jit->seen |= SEEN_RET0; + if (last) + break; + /* j */ + EMIT4_PCREL(0xa7f40000, jit->ret0_ip - jit->prg); + } else { + if (K <= 16383) + /* lghi %r2,K */ + EMIT4_IMM(0xa7290000, K); + else + /* llgf %r2,(%r13) */ + EMIT6_DISP(0xe320d000, 0x0016, EMIT_CONST(K)); + /* j */ + if (last && !(jit->seen & SEEN_RET0)) + break; + EMIT4_PCREL(0xa7f40000, jit->exit_ip - jit->prg); + } + break; + case BPF_S_RET_A: + /* llgfr %r2,%r5 */ + EMIT4(0xb9160025); + /* j */ + EMIT4_PCREL(0xa7f40000, jit->exit_ip - jit->prg); + break; + case BPF_S_ST: /* mem[K] = A */ + jit->seen |= SEEN_MEM; + /* st %r5,(%r15) */ + EMIT4_DISP(0x5050f000, + (jit->seen & SEEN_DATAREF) ? 160 + K*4 : K*4); + break; + case BPF_S_STX: /* mem[K] = X : mov %ebx,off8(%rbp) */ + jit->seen |= SEEN_XREG | SEEN_MEM; + /* st %r12,(%r15) */ + EMIT4_DISP(0x50c0f000, + (jit->seen & SEEN_DATAREF) ? 160 + K*4 : K*4); + break; + case BPF_S_ANC_PROTOCOL: /* A = ntohs(skb->protocol); */ + BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, protocol) != 2); + /* lhi %r5,0 */ + EMIT4(0xa7580000); + /* icm %r5,3,(%r2) */ + EMIT4_DISP(0xbf532000, offsetof(struct sk_buff, protocol)); + break; + case BPF_S_ANC_IFINDEX: /* if (!skb->dev) return 0; + * A = skb->dev->ifindex */ + BUILD_BUG_ON(FIELD_SIZEOF(struct net_device, ifindex) != 4); + jit->seen |= SEEN_RET0; + /* lg %r1,(%r2) */ + EMIT6_DISP(0xe3102000, 0x0004, offsetof(struct sk_buff, dev)); + /* ltgr %r1,%r1 */ + EMIT4(0xb9020011); + /* jz */ + EMIT4_PCREL(0xa7840000, jit->ret0_ip - jit->prg); + /* l %r5,(%r1) */ + EMIT4_DISP(0x58501000, offsetof(struct net_device, ifindex)); + break; + case BPF_S_ANC_MARK: /* A = skb->mark */ + BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, mark) != 4); + /* l %r5,(%r2) */ + EMIT4_DISP(0x58502000, offsetof(struct sk_buff, mark)); + break; + case BPF_S_ANC_QUEUE: /* A = skb->queue_mapping */ + BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, queue_mapping) != 2); + /* lhi %r5,0 */ + EMIT4(0xa7580000); + /* icm %r5,3,(%r2) */ + EMIT4_DISP(0xbf532000, offsetof(struct sk_buff, queue_mapping)); + break; + case BPF_S_ANC_HATYPE: /* if (!skb->dev) return 0; + * A = skb->dev->type */ + BUILD_BUG_ON(FIELD_SIZEOF(struct net_device, type) != 2); + jit->seen |= SEEN_RET0; + /* lg %r1,(%r2) */ + EMIT6_DISP(0xe3102000, 0x0004, offsetof(struct sk_buff, dev)); + /* ltgr %r1,%r1 */ + EMIT4(0xb9020011); + /* jz */ + EMIT4_PCREL(0xa7840000, jit->ret0_ip - jit->prg); + /* lhi %r5,0 */ + EMIT4(0xa7580000); + /* icm %r5,3,(%r1) */ + EMIT4_DISP(0xbf531000, offsetof(struct net_device, type)); + break; + case BPF_S_ANC_RXHASH: /* A = skb->rxhash */ + BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, rxhash) != 4); + /* l %r5,(%r2) */ + EMIT4_DISP(0x58502000, offsetof(struct sk_buff, rxhash)); + break; + case BPF_S_ANC_CPU: /* A = smp_processor_id() */ +#ifdef CONFIG_SMP + /* l %r5, */ + EMIT4_DISP(0x58500000, offsetof(struct _lowcore, cpu_nr)); +#else + /* lhi %r5,0 */ + EMIT4(0xa7580000); +#endif + break; + default: /* too complex, give up */ + goto out; + } + addrs[i] = jit->prg - jit->start; + return 0; +out: + return -1; +} + +void bpf_jit_compile(struct sk_filter *fp) +{ + unsigned long size, prg_len, lit_len; + struct bpf_jit jit, cjit; + unsigned int *addrs; + int pass, i; + + if (!bpf_jit_enable) + return; + addrs = kmalloc(fp->len * sizeof(*addrs), GFP_KERNEL); + if (addrs == NULL) + return; + memset(addrs, 0, fp->len * sizeof(*addrs)); + memset(&jit, 0, sizeof(cjit)); + memset(&cjit, 0, sizeof(cjit)); + + for (pass = 0; pass < 10; pass++) { + jit.prg = jit.start; + jit.lit = jit.mid; + + bpf_jit_prologue(&jit); + bpf_jit_noleaks(&jit, fp->insns); + for (i = 0; i < fp->len; i++) { + if (bpf_jit_insn(&jit, fp->insns + i, addrs, i, + i == fp->len - 1)) + goto out; + } + bpf_jit_epilogue(&jit); + if (jit.start) { + WARN_ON(jit.prg > cjit.prg || jit.lit > cjit.lit); + if (memcmp(&jit, &cjit, sizeof(jit)) == 0) + break; + } else if (jit.prg == cjit.prg && jit.lit == cjit.lit) { + prg_len = jit.prg - jit.start; + lit_len = jit.lit - jit.mid; + size = max_t(unsigned long, prg_len + lit_len, + sizeof(struct work_struct)); + if (size >= BPF_SIZE_MAX) + goto out; + jit.start = module_alloc(size); + if (!jit.start) + goto out; + jit.prg = jit.mid = jit.start + prg_len; + jit.lit = jit.end = jit.start + prg_len + lit_len; + jit.base_ip += (unsigned long) jit.start; + jit.exit_ip += (unsigned long) jit.start; + jit.ret0_ip += (unsigned long) jit.start; + } + cjit = jit; + } + if (bpf_jit_enable > 1) { + pr_err("flen=%d proglen=%lu pass=%d image=%p\n", + fp->len, jit.end - jit.start, pass, jit.start); + if (jit.start) { + printk(KERN_ERR "JIT code:\n"); + print_fn_code(jit.start, jit.mid - jit.start); + print_hex_dump(KERN_ERR, "JIT literals:\n", + DUMP_PREFIX_ADDRESS, 16, 1, + jit.mid, jit.end - jit.mid, false); + } + } + if (jit.start) + fp->bpf_func = (void *) jit.start; +out: + kfree(addrs); +} + +static void jit_free_defer(struct work_struct *arg) +{ + module_free(NULL, arg); +} + +/* run from softirq, we must use a work_struct to call + * module_free() from process context + */ +void bpf_jit_free(struct sk_filter *fp) +{ + struct work_struct *work; + + if (fp->bpf_func == sk_run_filter) + return; + work = (struct work_struct *)fp->bpf_func; + INIT_WORK(work, jit_free_defer); + schedule_work(work); +} -- cgit v0.10.2 From 68d9884dbc4215b6693c108eb35a02bd78f7956e Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 28 Aug 2012 15:36:14 +0200 Subject: s390/bpf,jit: improve code generation Make use of new immediate instructions that came with the extended immediate and general instruction extension facilities. Reviewed-by: Martin Schwidefsky Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/net/bpf_jit_comp.c b/arch/s390/net/bpf_jit_comp.c index b07aa3d..f57f0c3 100644 --- a/arch/s390/net/bpf_jit_comp.c +++ b/arch/s390/net/bpf_jit_comp.c @@ -10,6 +10,7 @@ #include #include #include +#include /* * Conventions: @@ -114,6 +115,12 @@ struct bpf_jit { EMIT6(op1 | __disp, op2); \ }) +#define EMIT6_IMM(op, imm) \ +({ \ + unsigned int __imm = (imm); \ + EMIT6(op | (__imm >> 16), __imm & 0xffff); \ +}) + #define EMIT_CONST(val) \ ({ \ unsigned int ret; \ @@ -276,6 +283,9 @@ static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, if (K <= 16383) /* ahi %r5, */ EMIT4_IMM(0xa75a0000, K); + else if (test_facility(21)) + /* alfi %r5, */ + EMIT6_IMM(0xc25b0000, K); else /* a %r5,(%r13) */ EMIT4_DISP(0x5a50d000, EMIT_CONST(K)); @@ -291,6 +301,9 @@ static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, if (K <= 16384) /* ahi %r5,-K */ EMIT4_IMM(0xa75a0000, -K); + else if (test_facility(21)) + /* alfi %r5,-K */ + EMIT6_IMM(0xc25b0000, -K); else /* s %r5,(%r13) */ EMIT4_DISP(0x5b50d000, EMIT_CONST(K)); @@ -304,6 +317,9 @@ static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, if (K <= 16383) /* mhi %r5,K */ EMIT4_IMM(0xa75c0000, K); + else if (test_facility(34)) + /* msfi %r5, */ + EMIT6_IMM(0xc2510000, K); else /* ms %r5,(%r13) */ EMIT4_DISP(0x7150d000, EMIT_CONST(K)); @@ -331,8 +347,12 @@ static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, EMIT2(0x145c); break; case BPF_S_ALU_AND_K: /* A &= K */ - /* n %r5,(%r13) */ - EMIT4_DISP(0x5450d000, EMIT_CONST(K)); + if (test_facility(21)) + /* nilf %r5, */ + EMIT6_IMM(0xc05b0000, K); + else + /* n %r5,(%r13) */ + EMIT4_DISP(0x5450d000, EMIT_CONST(K)); break; case BPF_S_ALU_OR_X: /* A |= X */ jit->seen |= SEEN_XREG; @@ -340,8 +360,12 @@ static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, EMIT2(0x165c); break; case BPF_S_ALU_OR_K: /* A |= K */ - /* o %r5,(%r13) */ - EMIT4_DISP(0x5650d000, EMIT_CONST(K)); + if (test_facility(21)) + /* oilf %r5, */ + EMIT6_IMM(0xc05d0000, K); + else + /* o %r5,(%r13) */ + EMIT4_DISP(0x5650d000, EMIT_CONST(K)); break; case BPF_S_ALU_LSH_X: /* A <<= X; */ jit->seen |= SEEN_XREG; @@ -386,6 +410,9 @@ kbranch: /* Emit compare if the branch targets are different */ if (K <= 16383) /* chi %r5, */ EMIT4_IMM(0xa75e0000, K); + else if (test_facility(21)) + /* clfi %r5, */ + EMIT6_IMM(0xc25f0000, K); else /* c %r5,(%r13) */ EMIT4_DISP(0x5950d000, EMIT_CONST(K)); @@ -508,6 +535,9 @@ call_fn: /* lg %r1,(%r13) */ if (K <= 16383) /* lhi %r5,K */ EMIT4_IMM(0xa7580000, K); + else if (test_facility(21)) + /* llilf %r5, */ + EMIT6_IMM(0xc05f0000, K); else /* l %r5,(%r13) */ EMIT4_DISP(0x5850d000, EMIT_CONST(K)); @@ -517,6 +547,9 @@ call_fn: /* lg %r1,(%r13) */ if (K <= 16383) /* lhi %r12, */ EMIT4_IMM(0xa7c80000, K); + else if (test_facility(21)) + /* llilf %r12, */ + EMIT6_IMM(0xc0cf0000, K); else /* l %r12,(%r13) */ EMIT4_DISP(0x58c0d000, EMIT_CONST(K)); -- cgit v0.10.2 From 535c611ddd3eb076f96579131e30bc5dd02a3b1c Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 14 Aug 2012 13:20:20 +0200 Subject: s390/string: provide asm lib functions for memcpy and memcmp Our memcpy and memcmp variants were implemented by calling the corresponding gcc builtin variants. However gcc is free to replace a call to __builtin_memcmp with a call to memcmp which, when called, will result in an endless recursion within memcmp. So let's provide asm variants and also fix the variants that are used for uncompressing the kernel image. In addition remove all other occurences of builtin function calls. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/boot/compressed/misc.c b/arch/s390/boot/compressed/misc.c index 465eca7..c4c6a1c 100644 --- a/arch/s390/boot/compressed/misc.c +++ b/arch/s390/boot/compressed/misc.c @@ -71,34 +71,37 @@ void *memset(void *s, int c, size_t n) { char *xs; - if (c == 0) - return __builtin_memset(s, 0, n); - - xs = (char *) s; - if (n > 0) - do { - *xs++ = c; - } while (--n > 0); + xs = s; + while (n--) + *xs++ = c; return s; } -void *memcpy(void *__dest, __const void *__src, size_t __n) +void *memcpy(void *dest, const void *src, size_t n) { - return __builtin_memcpy(__dest, __src, __n); + const char *s = src; + char *d = dest; + + while (n--) + *d++ = *s++; + return dest; } -void *memmove(void *__dest, __const void *__src, size_t __n) +void *memmove(void *dest, const void *src, size_t n) { - char *d; - const char *s; - - if (__dest <= __src) - return __builtin_memcpy(__dest, __src, __n); - d = __dest + __n; - s = __src + __n; - while (__n--) - *--d = *--s; - return __dest; + const char *s = src; + char *d = dest; + + if (d <= s) { + while (n--) + *d++ = *s++; + } else { + d += n; + s += n; + while (n--) + *--d = *--s; + } + return dest; } static void error(char *x) diff --git a/arch/s390/include/asm/string.h b/arch/s390/include/asm/string.h index 1bd1352..7e2dcd7 100644 --- a/arch/s390/include/asm/string.h +++ b/arch/s390/include/asm/string.h @@ -96,7 +96,6 @@ static inline char *strcat(char *dst, const char *src) static inline char *strcpy(char *dst, const char *src) { -#if __GNUC__ < 4 register int r0 asm("0") = 0; char *ret = dst; @@ -106,14 +105,10 @@ static inline char *strcpy(char *dst, const char *src) : "+&a" (dst), "+&a" (src) : "d" (r0) : "cc", "memory"); return ret; -#else - return __builtin_strcpy(dst, src); -#endif } static inline size_t strlen(const char *s) { -#if __GNUC__ < 4 register unsigned long r0 asm("0") = 0; const char *tmp = s; @@ -122,9 +117,6 @@ static inline size_t strlen(const char *s) " jo 0b" : "+d" (r0), "+a" (tmp) : : "cc"); return r0 - (unsigned long) s; -#else - return __builtin_strlen(s); -#endif } static inline size_t strnlen(const char * s, size_t n) diff --git a/arch/s390/kernel/s390_ksyms.c b/arch/s390/kernel/s390_ksyms.c index 57b5366..9bdbcef 100644 --- a/arch/s390/kernel/s390_ksyms.c +++ b/arch/s390/kernel/s390_ksyms.c @@ -8,3 +8,5 @@ EXPORT_SYMBOL(_mcount); #if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE) EXPORT_SYMBOL(sie64a); #endif +EXPORT_SYMBOL(memcpy); +EXPORT_SYMBOL(memset); diff --git a/arch/s390/lib/Makefile b/arch/s390/lib/Makefile index 761ab8b..6ab0d0b 100644 --- a/arch/s390/lib/Makefile +++ b/arch/s390/lib/Makefile @@ -4,6 +4,7 @@ lib-y += delay.o string.o uaccess_std.o uaccess_pt.o obj-y += usercopy.o -obj-$(CONFIG_32BIT) += div64.o qrnnd.o ucmpdi2.o +obj-$(CONFIG_32BIT) += div64.o qrnnd.o ucmpdi2.o mem32.o +obj-$(CONFIG_64BIT) += mem64.o lib-$(CONFIG_64BIT) += uaccess_mvcos.o lib-$(CONFIG_SMP) += spinlock.o diff --git a/arch/s390/lib/mem32.S b/arch/s390/lib/mem32.S new file mode 100644 index 0000000..14ca924 --- /dev/null +++ b/arch/s390/lib/mem32.S @@ -0,0 +1,92 @@ +/* + * String handling functions. + * + * Copyright IBM Corp. 2012 + */ + +#include + +/* + * memset implementation + * + * This code corresponds to the C construct below. We do distinguish + * between clearing (c == 0) and setting a memory array (c != 0) simply + * because nearly all memset invocations in the kernel clear memory and + * the xc instruction is preferred in such cases. + * + * void *memset(void *s, int c, size_t n) + * { + * if (likely(c == 0)) + * return __builtin_memset(s, 0, n); + * return __builtin_memset(s, c, n); + * } + */ +ENTRY(memset) + basr %r5,%r0 +.Lmemset_base: + ltr %r4,%r4 + bzr %r14 + ltr %r3,%r3 + jnz .Lmemset_fill + ahi %r4,-1 + lr %r3,%r4 + srl %r3,8 + ltr %r3,%r3 + lr %r1,%r2 + je .Lmemset_clear_rest +.Lmemset_clear_loop: + xc 0(256,%r1),0(%r1) + la %r1,256(%r1) + brct %r3,.Lmemset_clear_loop +.Lmemset_clear_rest: + ex %r4,.Lmemset_xc-.Lmemset_base(%r5) + br %r14 +.Lmemset_fill: + stc %r3,0(%r2) + chi %r4,1 + lr %r1,%r2 + ber %r14 + ahi %r4,-2 + lr %r3,%r4 + srl %r3,8 + ltr %r3,%r3 + je .Lmemset_fill_rest +.Lmemset_fill_loop: + mvc 1(256,%r1),0(%r1) + la %r1,256(%r1) + brct %r3,.Lmemset_fill_loop +.Lmemset_fill_rest: + ex %r4,.Lmemset_mvc-.Lmemset_base(%r5) + br %r14 +.Lmemset_xc: + xc 0(1,%r1),0(%r1) +.Lmemset_mvc: + mvc 1(1,%r1),0(%r1) + +/* + * memcpy implementation + * + * void *memcpy(void *dest, const void *src, size_t n) + */ +ENTRY(memcpy) + basr %r5,%r0 +.Lmemcpy_base: + ltr %r4,%r4 + bzr %r14 + ahi %r4,-1 + lr %r0,%r4 + srl %r0,8 + ltr %r0,%r0 + lr %r1,%r2 + jnz .Lmemcpy_loop +.Lmemcpy_rest: + ex %r4,.Lmemcpy_mvc-.Lmemcpy_base(%r5) + br %r14 +.Lmemcpy_loop: + mvc 0(256,%r1),0(%r3) + la %r1,256(%r1) + la %r3,256(%r3) + brct %r0,.Lmemcpy_loop + j .Lmemcpy_rest +.Lmemcpy_mvc: + mvc 0(1,%r1),0(%r3) diff --git a/arch/s390/lib/mem64.S b/arch/s390/lib/mem64.S new file mode 100644 index 0000000..c6d553e --- /dev/null +++ b/arch/s390/lib/mem64.S @@ -0,0 +1,88 @@ +/* + * String handling functions. + * + * Copyright IBM Corp. 2012 + */ + +#include + +/* + * memset implementation + * + * This code corresponds to the C construct below. We do distinguish + * between clearing (c == 0) and setting a memory array (c != 0) simply + * because nearly all memset invocations in the kernel clear memory and + * the xc instruction is preferred in such cases. + * + * void *memset(void *s, int c, size_t n) + * { + * if (likely(c == 0)) + * return __builtin_memset(s, 0, n); + * return __builtin_memset(s, c, n); + * } + */ +ENTRY(memset) + ltgr %r4,%r4 + bzr %r14 + ltgr %r3,%r3 + jnz .Lmemset_fill + aghi %r4,-1 + srlg %r3,%r4,8 + ltgr %r3,%r3 + lgr %r1,%r2 + jz .Lmemset_clear_rest +.Lmemset_clear_loop: + xc 0(256,%r1),0(%r1) + la %r1,256(%r1) + brctg %r3,.Lmemset_clear_loop +.Lmemset_clear_rest: + larl %r3,.Lmemset_xc + ex %r4,0(%r3) + br %r14 +.Lmemset_fill: + stc %r3,0(%r2) + cghi %r4,1 + lgr %r1,%r2 + ber %r14 + aghi %r4,-2 + srlg %r3,%r4,8 + ltgr %r3,%r3 + jz .Lmemset_fill_rest +.Lmemset_fill_loop: + mvc 1(256,%r1),0(%r1) + la %r1,256(%r1) + brctg %r3,.Lmemset_fill_loop +.Lmemset_fill_rest: + larl %r3,.Lmemset_mvc + ex %r4,0(%r3) + br %r14 +.Lmemset_xc: + xc 0(1,%r1),0(%r1) +.Lmemset_mvc: + mvc 1(1,%r1),0(%r1) + +/* + * memcpy implementation + * + * void *memcpy(void *dest, const void *src, size_t n) + */ +ENTRY(memcpy) + ltgr %r4,%r4 + bzr %r14 + aghi %r4,-1 + srlg %r5,%r4,8 + ltgr %r5,%r5 + lgr %r1,%r2 + jnz .Lmemcpy_loop +.Lmemcpy_rest: + larl %r5,.Lmemcpy_mvc + ex %r4,0(%r5) + br %r14 +.Lmemcpy_loop: + mvc 0(256,%r1),0(%r3) + la %r1,256(%r1) + la %r3,256(%r3) + brctg %r5,.Lmemcpy_loop + j .Lmemcpy_rest +.Lmemcpy_mvc: + mvc 0(1,%r1),0(%r3) diff --git a/arch/s390/lib/string.c b/arch/s390/lib/string.c index 846ec64..b647d5f 100644 --- a/arch/s390/lib/string.c +++ b/arch/s390/lib/string.c @@ -43,11 +43,7 @@ static inline char *__strnend(const char *s, size_t n) */ size_t strlen(const char *s) { -#if __GNUC__ < 4 return __strend(s) - s; -#else - return __builtin_strlen(s); -#endif } EXPORT_SYMBOL(strlen); @@ -73,7 +69,6 @@ EXPORT_SYMBOL(strnlen); */ char *strcpy(char *dest, const char *src) { -#if __GNUC__ < 4 register int r0 asm("0") = 0; char *ret = dest; @@ -82,9 +77,6 @@ char *strcpy(char *dest, const char *src) : "+&a" (dest), "+&a" (src) : "d" (r0) : "cc", "memory" ); return ret; -#else - return __builtin_strcpy(dest, src); -#endif } EXPORT_SYMBOL(strcpy); @@ -106,7 +98,7 @@ size_t strlcpy(char *dest, const char *src, size_t size) if (size) { size_t len = (ret >= size) ? size-1 : ret; dest[len] = '\0'; - __builtin_memcpy(dest, src, len); + memcpy(dest, src, len); } return ret; } @@ -124,8 +116,8 @@ EXPORT_SYMBOL(strlcpy); char *strncpy(char *dest, const char *src, size_t n) { size_t len = __strnend(src, n) - src; - __builtin_memset(dest + len, 0, n - len); - __builtin_memcpy(dest, src, len); + memset(dest + len, 0, n - len); + memcpy(dest, src, len); return dest; } EXPORT_SYMBOL(strncpy); @@ -171,7 +163,7 @@ size_t strlcat(char *dest, const char *src, size_t n) if (len >= n) len = n - 1; dest[len] = '\0'; - __builtin_memcpy(dest, src, len); + memcpy(dest, src, len); } return res; } @@ -194,7 +186,7 @@ char *strncat(char *dest, const char *src, size_t n) char *p = __strend(dest); p[len] = '\0'; - __builtin_memcpy(p, src, len); + memcpy(p, src, len); return dest; } EXPORT_SYMBOL(strncat); @@ -348,41 +340,3 @@ void *memscan(void *s, int c, size_t n) return (void *) ret; } EXPORT_SYMBOL(memscan); - -/** - * memcpy - Copy one area of memory to another - * @dest: Where to copy to - * @src: Where to copy from - * @n: The size of the area. - * - * returns a pointer to @dest - */ -void *memcpy(void *dest, const void *src, size_t n) -{ - return __builtin_memcpy(dest, src, n); -} -EXPORT_SYMBOL(memcpy); - -/** - * memset - Fill a region of memory with the given value - * @s: Pointer to the start of the area. - * @c: The byte to fill the area with - * @n: The size of the area. - * - * returns a pointer to @s - */ -void *memset(void *s, int c, size_t n) -{ - char *xs; - - if (c == 0) - return __builtin_memset(s, 0, n); - - xs = (char *) s; - if (n > 0) - do { - *xs++ = c; - } while (--n > 0); - return s; -} -EXPORT_SYMBOL(memset); -- cgit v0.10.2 From 648609e3f24599a5ad8d53df2ec13fbc37810bcc Mon Sep 17 00:00:00 2001 From: Gerald Schaefer Date: Tue, 21 Aug 2012 12:36:34 +0200 Subject: s390: enable large page support with CONFIG_DEBUG_PAGEALLOC So far, large page support was completely disabled with CONFIG_DEBUG_PAGEALLOC, although it would be sufficient if only the large page kernel mapping was disabled. This patch enables large page support with CONFIG_DEBUG_PAGEALLOC, while it prevents the large kernel mapping in that case. Signed-off-by: Gerald Schaefer Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c index 83c3271..e484079 100644 --- a/arch/s390/kernel/early.c +++ b/arch/s390/kernel/early.c @@ -267,12 +267,10 @@ static noinline __init void setup_facility_list(void) static noinline __init void setup_hpage(void) { -#ifndef CONFIG_DEBUG_PAGEALLOC if (!test_facility(2) || !test_facility(8)) return; S390_lowcore.machine_flags |= MACHINE_FLAG_HPAGE; __ctl_set_bit(0, 23); -#endif } static __init void detect_mvpg(void) diff --git a/arch/s390/mm/vmem.c b/arch/s390/mm/vmem.c index 6f896e7..c22abf9 100644 --- a/arch/s390/mm/vmem.c +++ b/arch/s390/mm/vmem.c @@ -107,7 +107,7 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro) pte = mk_pte_phys(address, __pgprot(ro ? _PAGE_RO : 0)); pm_dir = pmd_offset(pu_dir, address); -#ifdef CONFIG_64BIT +#if defined(CONFIG_64BIT) && !defined(CONFIG_DEBUG_PAGEALLOC) if (MACHINE_HAS_HPAGE && !(address & ~HPAGE_MASK) && (address + HPAGE_SIZE <= start + size) && (address >= HPAGE_SIZE)) { -- cgit v0.10.2 From 881730ad365130f64b5c70c40904b04eb3b79de3 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 23 Aug 2012 16:31:13 +0200 Subject: s390/cache: expose cpu cache topology via sysfs Expose cpu cache topology via sysfs. The created sysfs directory structure is compatible to what x86, ia64 and powerpc have. On s390 we expose only information about cpu caches which are private to a cpu via sysfs . Caches which are shared between cpus do not have a sysfs representation. The reason for that is that the file "shared_cpu_map" is mandatory and only if running under LPAR it is possible to tell which cpus share which cache. Second level hypervisors however do not and cannot expose that information to guests. In order to have a consistent view we made the choice to always only expose information about private cpu caches via sysfs. Example for a z196 cpu (cpu1 in /sys/devices/cpu): cpu1/cache/index0/size -- 64K cpu1/cache/index0/type -- Data cpu1/cache/index0/level -- 1 cpu1/cache/index0/number_of_sets -- 64 cpu1/cache/index0/shared_cpu_map -- 00000000,00000002 cpu1/cache/index0/shared_cpu_list -- 1 cpu1/cache/index0/coherency_line_size -- 256 cpu1/cache/index0/ways_of_associativity -- 4 cpu1/cache/index1/size -- 128K cpu1/cache/index1/type -- Instruction cpu1/cache/index1/level -- 1 cpu1/cache/index1/number_of_sets -- 64 cpu1/cache/index1/shared_cpu_map -- 00000000,00000002 cpu1/cache/index1/shared_cpu_list -- 1 cpu1/cache/index1/coherency_line_size -- 256 cpu1/cache/index1/ways_of_associativity -- 8 cpu1/cache/index2/size -- 1536K cpu1/cache/index2/type -- Unified cpu1/cache/index2/level -- 2 cpu1/cache/index2/number_of_sets -- 512 cpu1/cache/index2/shared_cpu_map -- 00000000,00000002 cpu1/cache/index2/shared_cpu_list -- 1 cpu1/cache/index2/coherency_line_size -- 256 cpu1/cache/index2/ways_of_associativity -- 12 Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile index 9733b3f..86b8247 100644 --- a/arch/s390/kernel/Makefile +++ b/arch/s390/kernel/Makefile @@ -49,6 +49,7 @@ obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o obj-$(CONFIG_FTRACE_SYSCALLS) += ftrace.o obj-$(CONFIG_CRASH_DUMP) += crash_dump.o obj-$(CONFIG_PERF_EVENTS) += perf_event.o perf_cpum_cf.o +obj-$(CONFIG_64BIT) += cache.o # Kexec part S390_KEXEC_OBJS := machine_kexec.o crash.o diff --git a/arch/s390/kernel/cache.c b/arch/s390/kernel/cache.c new file mode 100644 index 0000000..5e20bab --- /dev/null +++ b/arch/s390/kernel/cache.c @@ -0,0 +1,359 @@ +/* + * Extract CPU cache information and expose them via sysfs. + * + * Copyright IBM Corp. 2012 + * Author(s): Heiko Carstens + */ + +#include +#include +#include +#include +#include +#include + +struct cache { + unsigned long size; + unsigned int line_size; + unsigned int associativity; + unsigned int nr_sets; + int level; + int type; + struct list_head list; +}; + +struct cache_dir { + struct kobject *kobj; + struct cache_index_dir *index; +}; + +struct cache_index_dir { + struct kobject kobj; + int cpu; + struct cache *cache; + struct cache_index_dir *next; +}; + +enum { + CACHE_SCOPE_NOTEXISTS, + CACHE_SCOPE_PRIVATE, + CACHE_SCOPE_SHARED, + CACHE_SCOPE_RESERVED, +}; + +enum { + CACHE_TYPE_SEPARATE, + CACHE_TYPE_DATA, + CACHE_TYPE_INSTRUCTION, + CACHE_TYPE_UNIFIED, +}; + +enum { + EXTRACT_TOPOLOGY, + EXTRACT_LINE_SIZE, + EXTRACT_SIZE, + EXTRACT_ASSOCIATIVITY, +}; + +enum { + CACHE_TI_UNIFIED = 0, + CACHE_TI_INSTRUCTION = 0, + CACHE_TI_DATA, +}; + +struct cache_info { + unsigned char : 4; + unsigned char scope : 2; + unsigned char type : 2; +}; + +#define CACHE_MAX_LEVEL 8 + +union cache_topology { + struct cache_info ci[CACHE_MAX_LEVEL]; + unsigned long long raw; +}; + +static const char * const cache_type_string[] = { + "Data", + "Instruction", + "Unified", +}; + +static struct cache_dir *cache_dir_cpu[NR_CPUS]; +static LIST_HEAD(cache_list); + +static inline unsigned long ecag(int ai, int li, int ti) +{ + unsigned long cmd, val; + + cmd = ai << 4 | li << 1 | ti; + asm volatile(".insn rsy,0xeb000000004c,%0,0,0(%1)" /* ecag */ + : "=d" (val) : "a" (cmd)); + return val; +} + +static int __init cache_add(int level, int type) +{ + struct cache *cache; + int ti; + + cache = kzalloc(sizeof(*cache), GFP_KERNEL); + if (!cache) + return -ENOMEM; + ti = type == CACHE_TYPE_DATA ? CACHE_TI_DATA : CACHE_TI_UNIFIED; + cache->size = ecag(EXTRACT_SIZE, level, ti); + cache->line_size = ecag(EXTRACT_LINE_SIZE, level, ti); + cache->associativity = ecag(EXTRACT_ASSOCIATIVITY, level, ti); + cache->nr_sets = cache->size / cache->associativity; + cache->nr_sets /= cache->line_size; + cache->level = level + 1; + cache->type = type; + list_add_tail(&cache->list, &cache_list); + return 0; +} + +static void __init cache_build_info(void) +{ + struct cache *cache, *next; + union cache_topology ct; + int level, rc; + + ct.raw = ecag(EXTRACT_TOPOLOGY, 0, 0); + for (level = 0; level < CACHE_MAX_LEVEL; level++) { + switch (ct.ci[level].scope) { + case CACHE_SCOPE_NOTEXISTS: + case CACHE_SCOPE_RESERVED: + case CACHE_SCOPE_SHARED: + return; + case CACHE_SCOPE_PRIVATE: + break; + } + if (ct.ci[level].type == CACHE_TYPE_SEPARATE) { + rc = cache_add(level, CACHE_TYPE_DATA); + rc |= cache_add(level, CACHE_TYPE_INSTRUCTION); + } else { + rc = cache_add(level, ct.ci[level].type); + } + if (rc) + goto error; + } + return; +error: + list_for_each_entry_safe(cache, next, &cache_list, list) { + list_del(&cache->list); + kfree(cache); + } +} + +static struct cache_dir *__cpuinit cache_create_cache_dir(int cpu) +{ + struct cache_dir *cache_dir; + struct kobject *kobj = NULL; + struct device *dev; + + dev = get_cpu_device(cpu); + if (!dev) + goto out; + kobj = kobject_create_and_add("cache", &dev->kobj); + if (!kobj) + goto out; + cache_dir = kzalloc(sizeof(*cache_dir), GFP_KERNEL); + if (!cache_dir) + goto out; + cache_dir->kobj = kobj; + cache_dir_cpu[cpu] = cache_dir; + return cache_dir; +out: + kobject_put(kobj); + return NULL; +} + +static struct cache_index_dir *kobj_to_cache_index_dir(struct kobject *kobj) +{ + return container_of(kobj, struct cache_index_dir, kobj); +} + +static void cache_index_release(struct kobject *kobj) +{ + struct cache_index_dir *index; + + index = kobj_to_cache_index_dir(kobj); + kfree(index); +} + +static ssize_t cache_index_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct kobj_attribute *kobj_attr; + + kobj_attr = container_of(attr, struct kobj_attribute, attr); + return kobj_attr->show(kobj, kobj_attr, buf); +} + +#define DEFINE_CACHE_ATTR(_name, _format, _value) \ +static ssize_t cache_##_name##_show(struct kobject *kobj, \ + struct kobj_attribute *attr, \ + char *buf) \ +{ \ + struct cache_index_dir *index; \ + \ + index = kobj_to_cache_index_dir(kobj); \ + return sprintf(buf, _format, _value); \ +} \ +static struct kobj_attribute cache_##_name##_attr = \ + __ATTR(_name, 0444, cache_##_name##_show, NULL); + +DEFINE_CACHE_ATTR(size, "%luK\n", index->cache->size >> 10); +DEFINE_CACHE_ATTR(coherency_line_size, "%u\n", index->cache->line_size); +DEFINE_CACHE_ATTR(number_of_sets, "%u\n", index->cache->nr_sets); +DEFINE_CACHE_ATTR(ways_of_associativity, "%u\n", index->cache->associativity); +DEFINE_CACHE_ATTR(type, "%s\n", cache_type_string[index->cache->type - 1]); +DEFINE_CACHE_ATTR(level, "%d\n", index->cache->level); + +static ssize_t shared_cpu_map_func(struct kobject *kobj, int type, char *buf) +{ + struct cache_index_dir *index; + int len; + + index = kobj_to_cache_index_dir(kobj); + len = type ? + cpulist_scnprintf(buf, PAGE_SIZE - 2, cpumask_of(index->cpu)) : + cpumask_scnprintf(buf, PAGE_SIZE - 2, cpumask_of(index->cpu)); + len += sprintf(&buf[len], "\n"); + return len; +} + +static ssize_t shared_cpu_map_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + return shared_cpu_map_func(kobj, 0, buf); +} +static struct kobj_attribute cache_shared_cpu_map_attr = + __ATTR(shared_cpu_map, 0444, shared_cpu_map_show, NULL); + +static ssize_t shared_cpu_list_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + return shared_cpu_map_func(kobj, 1, buf); +} +static struct kobj_attribute cache_shared_cpu_list_attr = + __ATTR(shared_cpu_list, 0444, shared_cpu_list_show, NULL); + +static struct attribute *cache_index_default_attrs[] = { + &cache_type_attr.attr, + &cache_size_attr.attr, + &cache_number_of_sets_attr.attr, + &cache_ways_of_associativity_attr.attr, + &cache_level_attr.attr, + &cache_coherency_line_size_attr.attr, + &cache_shared_cpu_map_attr.attr, + &cache_shared_cpu_list_attr.attr, + NULL, +}; + +static const struct sysfs_ops cache_index_ops = { + .show = cache_index_show, +}; + +static struct kobj_type cache_index_type = { + .sysfs_ops = &cache_index_ops, + .release = cache_index_release, + .default_attrs = cache_index_default_attrs, +}; + +static int __cpuinit cache_create_index_dir(struct cache_dir *cache_dir, + struct cache *cache, int index, + int cpu) +{ + struct cache_index_dir *index_dir; + int rc; + + index_dir = kzalloc(sizeof(*index_dir), GFP_KERNEL); + if (!index_dir) + return -ENOMEM; + index_dir->cache = cache; + index_dir->cpu = cpu; + rc = kobject_init_and_add(&index_dir->kobj, &cache_index_type, + cache_dir->kobj, "index%d", index); + if (rc) + goto out; + index_dir->next = cache_dir->index; + cache_dir->index = index_dir; + return 0; +out: + kfree(index_dir); + return rc; +} + +static int __cpuinit cache_add_cpu(int cpu) +{ + struct cache_dir *cache_dir; + struct cache *cache; + int rc, index = 0; + + if (list_empty(&cache_list)) + return 0; + cache_dir = cache_create_cache_dir(cpu); + if (!cache_dir) + return -ENOMEM; + list_for_each_entry(cache, &cache_list, list) { + rc = cache_create_index_dir(cache_dir, cache, index, cpu); + if (rc) + return rc; + index++; + } + return 0; +} + +static void __cpuinit cache_remove_cpu(int cpu) +{ + struct cache_index_dir *index, *next; + struct cache_dir *cache_dir; + + cache_dir = cache_dir_cpu[cpu]; + if (!cache_dir) + return; + index = cache_dir->index; + while (index) { + next = index->next; + kobject_put(&index->kobj); + index = next; + } + kobject_put(cache_dir->kobj); + kfree(cache_dir); + cache_dir_cpu[cpu] = NULL; +} + +static int __cpuinit cache_hotplug(struct notifier_block *nfb, + unsigned long action, void *hcpu) +{ + int cpu = (long)hcpu; + int rc = 0; + + switch (action & ~CPU_TASKS_FROZEN) { + case CPU_ONLINE: + rc = cache_add_cpu(cpu); + if (rc) + cache_remove_cpu(cpu); + break; + case CPU_DEAD: + cache_remove_cpu(cpu); + break; + } + return rc ? NOTIFY_BAD : NOTIFY_OK; +} + +static int __init cache_init(void) +{ + int cpu; + + if (!test_facility(34)) + return 0; + cache_build_info(); + for_each_online_cpu(cpu) + cache_add_cpu(cpu); + hotcpu_notifier(cache_hotplug, 0); + return 0; +} +device_initcall(cache_init); -- cgit v0.10.2 From 843c48fd0fc8a63324d1a30255ac2750fa564937 Mon Sep 17 00:00:00 2001 From: Jan Glauber Date: Mon, 27 Aug 2012 10:55:18 +0200 Subject: s390/kconfig: split the s390 base menu Split the base menu into more descriptive categories like processor, memory, etc. and move virtualization related entries to the Virtualization menu. Also select KEXEC unconditionally. Signed-off-by: Jan Glauber Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index a7088dc..16506b2 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -55,6 +55,12 @@ config VIRT_CPU_ACCOUNTING config ARCH_SUPPORTS_DEBUG_PAGEALLOC def_bool y +config KEXEC + def_bool y + +config AUDIT_ARCH + def_bool y + config S390 def_bool y select USE_GENERIC_SMP_HELPERS if SMP @@ -134,9 +140,79 @@ source "init/Kconfig" source "kernel/Kconfig.freezer" -menu "Base setup" +menu "Processor type and features" + +config HAVE_MARCH_Z900_FEATURES + def_bool n + +config HAVE_MARCH_Z990_FEATURES + def_bool n + select HAVE_MARCH_Z900_FEATURES + +config HAVE_MARCH_Z9_109_FEATURES + def_bool n + select HAVE_MARCH_Z990_FEATURES + +config HAVE_MARCH_Z10_FEATURES + def_bool n + select HAVE_MARCH_Z9_109_FEATURES + +config HAVE_MARCH_Z196_FEATURES + def_bool n + select HAVE_MARCH_Z10_FEATURES + +choice + prompt "Processor type" + default MARCH_G5 + +config MARCH_G5 + bool "System/390 model G5 and G6" + depends on !64BIT + help + Select this to build a 31 bit kernel that works + on all ESA/390 and z/Architecture machines. + +config MARCH_Z900 + bool "IBM zSeries model z800 and z900" + select HAVE_MARCH_Z900_FEATURES if 64BIT + help + Select this to enable optimizations for model z800/z900 (2064 and + 2066 series). This will enable some optimizations that are not + available on older ESA/390 (31 Bit) only CPUs. + +config MARCH_Z990 + bool "IBM zSeries model z890 and z990" + select HAVE_MARCH_Z990_FEATURES if 64BIT + help + Select this to enable optimizations for model z890/z990 (2084 and + 2086 series). The kernel will be slightly faster but will not work + on older machines. + +config MARCH_Z9_109 + bool "IBM System z9" + select HAVE_MARCH_Z9_109_FEATURES if 64BIT + help + Select this to enable optimizations for IBM System z9 (2094 and + 2096 series). The kernel will be slightly faster but will not work + on older machines. + +config MARCH_Z10 + bool "IBM System z10" + select HAVE_MARCH_Z10_FEATURES if 64BIT + help + Select this to enable optimizations for IBM System z10 (2097 and + 2098 series). The kernel will be slightly faster but will not work + on older machines. + +config MARCH_Z196 + bool "IBM zEnterprise 114 and 196" + select HAVE_MARCH_Z196_FEATURES if 64BIT + help + Select this to enable optimizations for IBM zEnterprise 114 and 196 + (2818 and 2817 series). The kernel will be slightly faster but will + not work on older machines. -comment "Processor type and features" +endchoice config 64BIT def_bool y @@ -148,6 +224,24 @@ config 64BIT config 32BIT def_bool y if !64BIT +config COMPAT + def_bool y + prompt "Kernel support for 31 bit emulation" + depends on 64BIT + select COMPAT_BINFMT_ELF if BINFMT_ELF + select ARCH_WANT_OLD_COMPAT_IPC + help + Select this option if you want to enable your system kernel to + handle system-calls from ELF binaries for 31 bit ESA. This option + (and some other stuff like libraries and such) is needed for + executing 31 bit applications. It is safe to say "Y". + +config SYSVIPC_COMPAT + def_bool y if COMPAT && SYSVIPC + +config KEYS_COMPAT + def_bool y if COMPAT && KEYS + config SMP def_bool y prompt "Symmetric multi-processing support" @@ -203,6 +297,8 @@ config SCHED_BOOK Book scheduler support improves the CPU scheduler's decision making when dealing with machines that have several books. +source kernel/Kconfig.preempt + config MATHEMU def_bool y prompt "IEEE FPU emulation" @@ -212,100 +308,35 @@ config MATHEMU on older ESA/390 machines. Say Y unless you know your machine doesn't need this. -config COMPAT - def_bool y - prompt "Kernel support for 31 bit emulation" - depends on 64BIT - select COMPAT_BINFMT_ELF if BINFMT_ELF - select ARCH_WANT_OLD_COMPAT_IPC - help - Select this option if you want to enable your system kernel to - handle system-calls from ELF binaries for 31 bit ESA. This option - (and some other stuff like libraries and such) is needed for - executing 31 bit applications. It is safe to say "Y". +source kernel/Kconfig.hz -config SYSVIPC_COMPAT - def_bool y if COMPAT && SYSVIPC +endmenu -config KEYS_COMPAT - def_bool y if COMPAT && KEYS +menu "Memory setup" -config AUDIT_ARCH +config ARCH_SPARSEMEM_ENABLE def_bool y + select SPARSEMEM_VMEMMAP_ENABLE + select SPARSEMEM_VMEMMAP + select SPARSEMEM_STATIC if !64BIT -config HAVE_MARCH_Z900_FEATURES - def_bool n - -config HAVE_MARCH_Z990_FEATURES - def_bool n - select HAVE_MARCH_Z900_FEATURES - -config HAVE_MARCH_Z9_109_FEATURES - def_bool n - select HAVE_MARCH_Z990_FEATURES - -config HAVE_MARCH_Z10_FEATURES - def_bool n - select HAVE_MARCH_Z9_109_FEATURES - -config HAVE_MARCH_Z196_FEATURES - def_bool n - select HAVE_MARCH_Z10_FEATURES - -comment "Code generation options" - -choice - prompt "Processor type" - default MARCH_G5 - -config MARCH_G5 - bool "System/390 model G5 and G6" - depends on !64BIT - help - Select this to build a 31 bit kernel that works - on all ESA/390 and z/Architecture machines. - -config MARCH_Z900 - bool "IBM zSeries model z800 and z900" - select HAVE_MARCH_Z900_FEATURES if 64BIT - help - Select this to enable optimizations for model z800/z900 (2064 and - 2066 series). This will enable some optimizations that are not - available on older ESA/390 (31 Bit) only CPUs. +config ARCH_SPARSEMEM_DEFAULT + def_bool y -config MARCH_Z990 - bool "IBM zSeries model z890 and z990" - select HAVE_MARCH_Z990_FEATURES if 64BIT - help - Select this to enable optimizations for model z890/z990 (2084 and - 2086 series). The kernel will be slightly faster but will not work - on older machines. +config ARCH_SELECT_MEMORY_MODEL + def_bool y -config MARCH_Z9_109 - bool "IBM System z9" - select HAVE_MARCH_Z9_109_FEATURES if 64BIT - help - Select this to enable optimizations for IBM System z9 (2094 and - 2096 series). The kernel will be slightly faster but will not work - on older machines. +config ARCH_ENABLE_MEMORY_HOTPLUG + def_bool y if SPARSEMEM -config MARCH_Z10 - bool "IBM System z10" - select HAVE_MARCH_Z10_FEATURES if 64BIT - help - Select this to enable optimizations for IBM System z10 (2097 and - 2098 series). The kernel will be slightly faster but will not work - on older machines. +config ARCH_ENABLE_MEMORY_HOTREMOVE + def_bool y -config MARCH_Z196 - bool "IBM zEnterprise 114 and 196" - select HAVE_MARCH_Z196_FEATURES if 64BIT - help - Select this to enable optimizations for IBM zEnterprise 114 and 196 - (2818 and 2817 series). The kernel will be slightly faster but will - not work on older machines. +config FORCE_MAX_ZONEORDER + int + default "9" -endchoice +source "mm/Kconfig" config PACK_STACK def_bool y @@ -369,34 +400,9 @@ config WARN_DYNAMIC_STACK Say N if you are unsure. -comment "Kernel preemption" - -source "kernel/Kconfig.preempt" - -config ARCH_SPARSEMEM_ENABLE - def_bool y - select SPARSEMEM_VMEMMAP_ENABLE - select SPARSEMEM_VMEMMAP - select SPARSEMEM_STATIC if !64BIT - -config ARCH_SPARSEMEM_DEFAULT - def_bool y - -config ARCH_SELECT_MEMORY_MODEL - def_bool y - -config ARCH_ENABLE_MEMORY_HOTPLUG - def_bool y if SPARSEMEM - -config ARCH_ENABLE_MEMORY_HOTREMOVE - def_bool y - -config ARCH_HIBERNATION_POSSIBLE - def_bool y if 64BIT - -source "mm/Kconfig" +endmenu -comment "I/O subsystem configuration" +menu "I/O subsystem" config QDIO def_tristate y @@ -427,13 +433,84 @@ config CHSC_SCH If unsure, say N. -comment "Misc" +endmenu + +menu "Dump support" + +config CRASH_DUMP + bool "kernel crash dumps" + depends on 64BIT && SMP + select KEXEC + help + Generate crash dump after being started by kexec. + Crash dump kernels are loaded in the main kernel with kexec-tools + into a specially reserved region and then later executed after + a crash by kdump/kexec. + For more details see Documentation/kdump/kdump.txt + +config ZFCPDUMP + def_bool n + prompt "zfcpdump support" + select SMP + help + Select this option if you want to build an zfcpdump enabled kernel. + Refer to for more details on this. + +endmenu + +menu "Executable file formats / Emulations" source "fs/Kconfig.binfmt" -config FORCE_MAX_ZONEORDER - int - default "9" +config SECCOMP + def_bool y + prompt "Enable seccomp to safely compute untrusted bytecode" + depends on PROC_FS + help + This kernel feature is useful for number crunching applications + that may need to compute untrusted bytecode during their + execution. By using pipes or other transports made available to + the process as file descriptors supporting the read/write + syscalls, it's possible to isolate those applications in + their own address space using seccomp. Once seccomp is + enabled via /proc//seccomp, it cannot be disabled + and the task is only allowed to execute a few safe syscalls + defined by each seccomp mode. + + If unsure, say Y. + +endmenu + +menu "Power Management" + +config ARCH_HIBERNATION_POSSIBLE + def_bool y if 64BIT + +source "kernel/power/Kconfig" + +endmenu + +source "net/Kconfig" + +config PCMCIA + def_bool n + +config CCW + def_bool y + +source "drivers/Kconfig" + +source "fs/Kconfig" + +source "arch/s390/Kconfig.debug" + +source "security/Kconfig" + +source "crypto/Kconfig" + +source "lib/Kconfig" + +menu "Virtualization" config PFAULT def_bool y @@ -545,8 +622,6 @@ config APPLDATA_NET_SUM This can also be compiled as a module, which will be called appldata_net_sum.o. -source kernel/Kconfig.hz - config S390_HYPFS_FS def_bool y prompt "s390 hypervisor file system support" @@ -555,32 +630,7 @@ config S390_HYPFS_FS This is a virtual file system intended to provide accounting information in an s390 hypervisor environment. -config KEXEC - def_bool n - prompt "kexec system call" - help - kexec is a system call that implements the ability to shutdown your - current kernel, and to start another kernel. It is like a reboot - but is independent of hardware/microcode support. - -config CRASH_DUMP - bool "kernel crash dumps" - depends on 64BIT && SMP - select KEXEC - help - Generate crash dump after being started by kexec. - Crash dump kernels are loaded in the main kernel with kexec-tools - into a specially reserved region and then later executed after - a crash by kdump/kexec. - For more details see Documentation/kdump/kdump.txt - -config ZFCPDUMP - def_bool n - prompt "zfcpdump support" - select SMP - help - Select this option if you want to build an zfcpdump enabled kernel. - Refer to for more details on this. +source "arch/s390/kvm/Kconfig" config S390_GUEST def_bool y @@ -596,49 +646,4 @@ config S390_GUEST virtio transport. If KVM is detected, the virtio console will be the default console. -config SECCOMP - def_bool y - prompt "Enable seccomp to safely compute untrusted bytecode" - depends on PROC_FS - help - This kernel feature is useful for number crunching applications - that may need to compute untrusted bytecode during their - execution. By using pipes or other transports made available to - the process as file descriptors supporting the read/write - syscalls, it's possible to isolate those applications in - their own address space using seccomp. Once seccomp is - enabled via /proc//seccomp, it cannot be disabled - and the task is only allowed to execute a few safe syscalls - defined by each seccomp mode. - - If unsure, say Y. - -endmenu - -menu "Power Management" - -source "kernel/power/Kconfig" - endmenu - -source "net/Kconfig" - -config PCMCIA - def_bool n - -config CCW - def_bool y - -source "drivers/Kconfig" - -source "fs/Kconfig" - -source "arch/s390/Kconfig.debug" - -source "security/Kconfig" - -source "crypto/Kconfig" - -source "lib/Kconfig" - -source "arch/s390/kvm/Kconfig" diff --git a/arch/s390/kvm/Kconfig b/arch/s390/kvm/Kconfig index 78eb984..9b04a32 100644 --- a/arch/s390/kvm/Kconfig +++ b/arch/s390/kvm/Kconfig @@ -5,7 +5,7 @@ source "virt/kvm/Kconfig" menuconfig VIRTUALIZATION def_bool y - prompt "Virtualization" + prompt "KVM" ---help--- Say Y here to get to see options for using your Linux host to run other operating systems inside virtual machines (guests). -- cgit v0.10.2 From bdd1fc27cde35e092497ce8b2ff862bd3e8d27aa Mon Sep 17 00:00:00 2001 From: Cornelia Huck Date: Mon, 27 Aug 2012 12:56:48 +0200 Subject: s390/kvm: Improve Kconfig help text. CONFIG_S390_GUEST determines whether virtio guest drivers are built, and the virtio console is not enforced as default. Signed-off-by: Cornelia Huck Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 16506b2..443fac9 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -634,16 +634,17 @@ source "arch/s390/kvm/Kconfig" config S390_GUEST def_bool y - prompt "s390 guest support for KVM (EXPERIMENTAL)" + prompt "s390 support for virtio devices (EXPERIMENTAL)" depends on 64BIT && EXPERIMENTAL select VIRTUALIZATION select VIRTIO select VIRTIO_RING select VIRTIO_CONSOLE help + Enabling this option adds support for virtio based paravirtual device + drivers on s390. + Select this option if you want to run the kernel as a guest under - the KVM hypervisor. This will add detection for KVM as well as a - virtio transport. If KVM is detected, the virtio console will be - the default console. + the KVM hypervisor. endmenu -- cgit v0.10.2 From 7755d6b2c05bea032bf2470d771ad9b56e26015d Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 27 Aug 2012 15:14:47 +0200 Subject: s390/cpu hotplug: use hotcpu_notifier() instead of register_cpu_notifier() Saves a couple of lines of code. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index 720fda1..823c5bf 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -972,10 +972,6 @@ static int __cpuinit smp_cpu_notify(struct notifier_block *self, return notifier_from_errno(err); } -static struct notifier_block __cpuinitdata smp_cpu_nb = { - .notifier_call = smp_cpu_notify, -}; - static int __devinit smp_add_present_cpu(int cpu) { struct cpu *c = &pcpu_devices[cpu].cpu; @@ -1050,7 +1046,7 @@ static int __init s390_smp_init(void) { int cpu, rc; - register_cpu_notifier(&smp_cpu_nb); + hotcpu_notifier(smp_cpu_notify, 0); #ifdef CONFIG_HOTPLUG_CPU rc = device_create_file(cpu_subsys.dev_root, &dev_attr_rescan); if (rc) -- cgit v0.10.2 From 0d0e471b46d4ba1de4617d04cf071c6ae4a5df76 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 27 Aug 2012 15:38:19 +0200 Subject: s390/smp: fix smp_find_processor_id() argument mismatch For SMP and !SMP smp_find_processor_id() either takes a u16 or an unsigned int argument. Fix this so both versions take a u16. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/smp.h b/arch/s390/include/asm/smp.h index ce26ac3..da7dcd7 100644 --- a/arch/s390/include/asm/smp.h +++ b/arch/s390/include/asm/smp.h @@ -43,7 +43,7 @@ static inline void smp_call_online_cpu(void (*func)(void *), void *data) func(data); } -static inline int smp_find_processor_id(int address) { return 0; } +static inline int smp_find_processor_id(u16 address) { return 0; } static inline int smp_store_status(int cpu) { return 0; } static inline int smp_vcpu_scheduled(int cpu) { return 1; } static inline void smp_yield_cpu(int cpu) { } -- cgit v0.10.2 From 1c725922dd8274b6e0b370ad33cfaef9013fd70b Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 27 Aug 2012 15:43:49 +0200 Subject: s390/cpu hotplug: mask out CPU_TASKS_FROZEN in cu hotplug notifiers Unify all our cpu hotplug notifiers to mask out the CPU_TASKS_FROZEN bit, so we don't have to add all the *_FROZEN variant cases to the notifiers. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index 823c5bf..87be102 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -959,13 +959,11 @@ static int __cpuinit smp_cpu_notify(struct notifier_block *self, struct device *s = &c->dev; int err = 0; - switch (action) { + switch (action & ~CPU_TASKS_FROZEN) { case CPU_ONLINE: - case CPU_ONLINE_FROZEN: err = sysfs_create_group(&s->kobj, &cpu_online_attr_group); break; case CPU_DEAD: - case CPU_DEAD_FROZEN: sysfs_remove_group(&s->kobj, &cpu_online_attr_group); break; } diff --git a/arch/s390/kernel/vtime.c b/arch/s390/kernel/vtime.c index 4fc97b4..d100c92 100644 --- a/arch/s390/kernel/vtime.c +++ b/arch/s390/kernel/vtime.c @@ -378,9 +378,8 @@ static int __cpuinit s390_nohz_notify(struct notifier_block *self, long cpu = (long) hcpu; idle = &per_cpu(s390_idle, cpu); - switch (action) { + switch (action & ~CPU_TASKS_FROZEN) { case CPU_DYING: - case CPU_DYING_FROZEN: idle->nohz_delay = 0; default: break; diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c index 6c013f5..12c887b 100644 --- a/arch/s390/mm/fault.c +++ b/arch/s390/mm/fault.c @@ -628,9 +628,8 @@ static int __cpuinit pfault_cpu_notify(struct notifier_block *self, struct thread_struct *thread, *next; struct task_struct *tsk; - switch (action) { + switch (action & ~CPU_TASKS_FROZEN) { case CPU_DEAD: - case CPU_DEAD_FROZEN: spin_lock_irq(&pfault_lock); list_for_each_entry_safe(thread, next, &pfault_list, list) { thread->pfault_wait = 0; -- cgit v0.10.2 From 48a8ca03f8fd49a4d0c0c8843d4f5a7008dc2656 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 27 Aug 2012 15:50:29 +0200 Subject: s390/kexec: move machine_crash_shutdown() to machine_kexec.c machine_crash_shutdown() was the only function in crash.c. So move the function and delete one file. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile index 86b8247..9a111b6 100644 --- a/arch/s390/kernel/Makefile +++ b/arch/s390/kernel/Makefile @@ -52,7 +52,7 @@ obj-$(CONFIG_PERF_EVENTS) += perf_event.o perf_cpum_cf.o obj-$(CONFIG_64BIT) += cache.o # Kexec part -S390_KEXEC_OBJS := machine_kexec.o crash.o +S390_KEXEC_OBJS := machine_kexec.o S390_KEXEC_OBJS += $(if $(CONFIG_64BIT),relocate_kernel64.o,relocate_kernel.o) obj-$(CONFIG_KEXEC) += $(S390_KEXEC_OBJS) diff --git a/arch/s390/kernel/crash.c b/arch/s390/kernel/crash.c deleted file mode 100644 index 3819153..0000000 --- a/arch/s390/kernel/crash.c +++ /dev/null @@ -1,14 +0,0 @@ -/* - * Copyright IBM Corp. 2005 - * - * Author(s): Heiko Carstens - * - */ - -#include -#include -#include - -void machine_crash_shutdown(struct pt_regs *regs) -{ -} diff --git a/arch/s390/kernel/machine_kexec.c b/arch/s390/kernel/machine_kexec.c index 493304b..0d5063e0 100644 --- a/arch/s390/kernel/machine_kexec.c +++ b/arch/s390/kernel/machine_kexec.c @@ -191,6 +191,10 @@ void machine_shutdown(void) { } +void machine_crash_shutdown(struct pt_regs *regs) +{ +} + /* * Do normal kexec */ -- cgit v0.10.2 From 75014550516b147e5f530b84c71496341e036d6f Mon Sep 17 00:00:00 2001 From: Holger Dengler Date: Tue, 28 Aug 2012 16:41:50 +0200 Subject: s390/ap: configuration information exploitation Query AP configuration information. Improve performance of AP bus scans by skipping AP device probing, if the AP deviec is not configured. Reviewed-by: Ingo Tuchscherer Signed-off-by: Holger Dengler Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index ae258a4..047c732 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -1,5 +1,5 @@ /* - * Copyright IBM Corp. 2006 + * Copyright IBM Corp. 2006, 2012 * Author(s): Cornelia Huck * Martin Schwidefsky * Ralph Wuerthner @@ -62,13 +62,14 @@ static void ap_interrupt_handler(void *unused1, void *unused2); static void ap_reset(struct ap_device *ap_dev); static void ap_config_timeout(unsigned long ptr); static int ap_select_domain(void); +static void ap_query_configuration(void); /* * Module description. */ MODULE_AUTHOR("IBM Corporation"); -MODULE_DESCRIPTION("Adjunct Processor Bus driver, " - "Copyright IBM Corp. 2006"); +MODULE_DESCRIPTION("Adjunct Processor Bus driver, " \ + "Copyright IBM Corp. 2006, 2012"); MODULE_LICENSE("GPL"); /* @@ -84,6 +85,7 @@ module_param_named(poll_thread, ap_thread_flag, int, 0000); MODULE_PARM_DESC(poll_thread, "Turn on/off poll thread, default is 0 (off)."); static struct device *ap_root_device = NULL; +static struct ap_config_info *ap_configuration; static DEFINE_SPINLOCK(ap_device_list_lock); static LIST_HEAD(ap_device_list); @@ -158,6 +160,17 @@ static int ap_interrupts_available(void) } /** + * ap_configuration_available(): Test if AP configuration + * information is available. + * + * Returns 1 if AP configuration information is available. + */ +static int ap_configuration_available(void) +{ + return test_facility(2) && test_facility(12); +} + +/** * ap_test_queue(): Test adjunct processor queue. * @qid: The AP queue number * @queue_depth: Pointer to queue depth value @@ -242,6 +255,26 @@ __ap_query_functions(ap_qid_t qid, unsigned int *functions) } #endif +#ifdef CONFIG_64BIT +static inline int __ap_query_configuration(struct ap_config_info *config) +{ + register unsigned long reg0 asm ("0") = 0x04000000UL; + register unsigned long reg1 asm ("1") = -EINVAL; + register unsigned char *reg2 asm ("2") = (unsigned char *)config; + + asm volatile( + ".long 0xb2af0000\n" /* PQAP(QCI) */ + "0: la %1,0\n" + "1:\n" + EX_TABLE(0b, 1b) + : "+d" (reg0), "+d" (reg1), "+d" (reg2) + : + : "cc"); + + return reg1; +} +#endif + /** * ap_query_functions(): Query supported functions. * @qid: The AP queue number @@ -305,8 +338,8 @@ int ap_4096_commands_available(ap_qid_t qid) if (ap_query_functions(qid, &functions)) return 0; - return test_ap_facility(functions, 1) && - test_ap_facility(functions, 2); + return ap_test_bit(&functions, 1) && + ap_test_bit(&functions, 2); } EXPORT_SYMBOL(ap_4096_commands_available); @@ -772,6 +805,7 @@ static int ap_bus_resume(struct device *dev) ap_suspend_flag = 0; if (!ap_interrupts_available()) ap_interrupt_indicator = NULL; + ap_query_configuration(); if (!user_set_domain) { ap_domain_index = -1; ap_select_domain(); @@ -997,6 +1031,65 @@ static struct bus_attribute *const ap_bus_attrs[] = { NULL, }; +static inline int ap_test_config(unsigned int *field, unsigned int nr) +{ + if (nr > 0xFFu) + return 0; + return ap_test_bit((field + (nr >> 5)), (nr & 0x1f)); +} + +/* + * ap_test_config_card_id(): Test, whether an AP card ID is configured. + * @id AP card ID + * + * Returns 0 if the card is not configured + * 1 if the card is configured or + * if the configuration information is not available + */ +static inline int ap_test_config_card_id(unsigned int id) +{ + if (!ap_configuration) + return 1; + return ap_test_config(ap_configuration->apm, id); +} + +/* + * ap_test_config_domain(): Test, whether an AP usage domain is configured. + * @domain AP usage domain ID + * + * Returns 0 if the usage domain is not configured + * 1 if the usage domain is configured or + * if the configuration information is not available + */ +static inline int ap_test_config_domain(unsigned int domain) +{ + if (!ap_configuration) + return 1; + return ap_test_config(ap_configuration->aqm, domain); +} + +/** + * ap_query_configuration(): Query AP configuration information. + * + * Query information of installed cards and configured domains from AP. + */ +static void ap_query_configuration(void) +{ +#ifdef CONFIG_64BIT + if (ap_configuration_available()) { + if (!ap_configuration) + ap_configuration = + kzalloc(sizeof(struct ap_config_info), + GFP_KERNEL); + if (ap_configuration) + __ap_query_configuration(ap_configuration); + } else + ap_configuration = NULL; +#else + ap_configuration = NULL; +#endif +} + /** * ap_select_domain(): Select an AP domain. * @@ -1005,6 +1098,7 @@ static struct bus_attribute *const ap_bus_attrs[] = { static int ap_select_domain(void) { int queue_depth, device_type, count, max_count, best_domain; + ap_qid_t qid; int rc, i, j; /* @@ -1018,9 +1112,13 @@ static int ap_select_domain(void) best_domain = -1; max_count = 0; for (i = 0; i < AP_DOMAINS; i++) { + if (!ap_test_config_domain(i)) + continue; count = 0; for (j = 0; j < AP_DEVICES; j++) { - ap_qid_t qid = AP_MKQID(j, i); + if (!ap_test_config_card_id(j)) + continue; + qid = AP_MKQID(j, i); rc = ap_query_queue(qid, &queue_depth, &device_type); if (rc) continue; @@ -1169,6 +1267,7 @@ static void ap_scan_bus(struct work_struct *unused) unsigned int device_functions; int rc, i; + ap_query_configuration(); if (ap_select_domain() != 0) return; for (i = 0; i < AP_DEVICES; i++) { @@ -1176,7 +1275,10 @@ static void ap_scan_bus(struct work_struct *unused) dev = bus_find_device(&ap_bus_type, NULL, (void *)(unsigned long)qid, __ap_scan_bus); - rc = ap_query_queue(qid, &queue_depth, &device_type); + if (ap_test_config_card_id(i)) + rc = ap_query_queue(qid, &queue_depth, &device_type); + else + rc = -ENODEV; if (dev) { if (rc == -EBUSY) { set_current_state(TASK_UNINTERRUPTIBLE); @@ -1227,9 +1329,9 @@ static void ap_scan_bus(struct work_struct *unused) kfree(ap_dev); continue; } - if (test_ap_facility(device_functions, 3)) + if (ap_test_bit(&device_functions, 3)) ap_dev->device_type = AP_DEVICE_TYPE_CEX3C; - else if (test_ap_facility(device_functions, 4)) + else if (ap_test_bit(&device_functions, 4)) ap_dev->device_type = AP_DEVICE_TYPE_CEX3A; else { kfree(ap_dev); @@ -1785,6 +1887,7 @@ int __init ap_module_init(void) goto out_root; } + ap_query_configuration(); if (ap_select_domain() == 0) ap_scan_bus(NULL); diff --git a/drivers/s390/crypto/ap_bus.h b/drivers/s390/crypto/ap_bus.h index 52d6199..5018f66 100644 --- a/drivers/s390/crypto/ap_bus.h +++ b/drivers/s390/crypto/ap_bus.h @@ -1,5 +1,5 @@ /* - * Copyright IBM Corp. 2006 + * Copyright IBM Corp. 2006, 2012 * Author(s): Cornelia Huck * Martin Schwidefsky * Ralph Wuerthner @@ -83,13 +83,12 @@ int ap_queue_status_invalid_test(struct ap_queue_status *status) return !(memcmp(status, &invalid, sizeof(struct ap_queue_status))); } -#define MAX_AP_FACILITY 31 - -static inline int test_ap_facility(unsigned int function, unsigned int nr) +#define AP_MAX_BITS 31 +static inline int ap_test_bit(unsigned int *ptr, unsigned int nr) { - if (nr > MAX_AP_FACILITY) + if (nr > AP_MAX_BITS) return 0; - return function & (unsigned int)(0x80000000 >> nr); + return (*ptr & (0x80000000u >> nr)) != 0; } #define AP_RESPONSE_NORMAL 0x00 @@ -183,6 +182,17 @@ struct ap_message { struct ap_message *); }; +struct ap_config_info { + unsigned int special_command:1; + unsigned int ap_extended:1; + unsigned char reserved1:6; + unsigned char reserved2[15]; + unsigned int apm[8]; /* AP ID mask */ + unsigned int aqm[8]; /* AP queue mask */ + unsigned int adm[8]; /* AP domain mask */ + unsigned char reserved4[16]; +} __packed; + #define AP_DEVICE(dt) \ .dev_type=(dt), \ .match_flags=AP_DEVICE_ID_MATCH_DEVICE_TYPE, -- cgit v0.10.2 From b26bd9413c8359bd9f5c9bd1b789a10ebd2bb484 Mon Sep 17 00:00:00 2001 From: Holger Dengler Date: Tue, 28 Aug 2012 16:43:48 +0200 Subject: s390/ap: Add functiton facility information as AP device attribute. Add the function facility information as new ap_device and sysfs attribute. Also make the number of requests in device queue and in device driver queue accessible in sysfs. Reviewed-by: Ingo Tuchscherer Signed-off-by: Holger Dengler Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index 047c732..f0b53ee 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -325,25 +325,6 @@ static int ap_query_functions(ap_qid_t qid, unsigned int *functions) } /** - * ap_4096_commands_availablen(): Check for availability of 4096 bit RSA - * support. - * @qid: The AP queue number - * - * Returns 1 if 4096 bit RSA keys are support fo the AP, returns 0 if not. - */ -int ap_4096_commands_available(ap_qid_t qid) -{ - unsigned int functions; - - if (ap_query_functions(qid, &functions)) - return 0; - - return ap_test_bit(&functions, 1) && - ap_test_bit(&functions, 2); -} -EXPORT_SYMBOL(ap_4096_commands_available); - -/** * ap_queue_enable_interruption(): Enable interruption on an AP. * @qid: The AP queue number * @ind: the notification indicator byte @@ -690,6 +671,34 @@ static ssize_t ap_request_count_show(struct device *dev, static DEVICE_ATTR(request_count, 0444, ap_request_count_show, NULL); +static ssize_t ap_requestq_count_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ap_device *ap_dev = to_ap_dev(dev); + int rc; + + spin_lock_bh(&ap_dev->lock); + rc = snprintf(buf, PAGE_SIZE, "%d\n", ap_dev->requestq_count); + spin_unlock_bh(&ap_dev->lock); + return rc; +} + +static DEVICE_ATTR(requestq_count, 0444, ap_requestq_count_show, NULL); + +static ssize_t ap_pendingq_count_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ap_device *ap_dev = to_ap_dev(dev); + int rc; + + spin_lock_bh(&ap_dev->lock); + rc = snprintf(buf, PAGE_SIZE, "%d\n", ap_dev->pendingq_count); + spin_unlock_bh(&ap_dev->lock); + return rc; +} + +static DEVICE_ATTR(pendingq_count, 0444, ap_pendingq_count_show, NULL); + static ssize_t ap_modalias_show(struct device *dev, struct device_attribute *attr, char *buf) { @@ -698,11 +707,23 @@ static ssize_t ap_modalias_show(struct device *dev, static DEVICE_ATTR(modalias, 0444, ap_modalias_show, NULL); +static ssize_t ap_functions_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ap_device *ap_dev = to_ap_dev(dev); + return snprintf(buf, PAGE_SIZE, "0x%08X\n", ap_dev->functions); +} + +static DEVICE_ATTR(ap_functions, 0444, ap_functions_show, NULL); + static struct attribute *ap_dev_attrs[] = { &dev_attr_hwtype.attr, &dev_attr_depth.attr, &dev_attr_request_count.attr, + &dev_attr_requestq_count.attr, + &dev_attr_pendingq_count.attr, &dev_attr_modalias.attr, + &dev_attr_ap_functions.attr, NULL }; static struct attribute_group ap_dev_attr_group = { @@ -1329,6 +1350,7 @@ static void ap_scan_bus(struct work_struct *unused) kfree(ap_dev); continue; } + ap_dev->functions = device_functions; if (ap_test_bit(&device_functions, 3)) ap_dev->device_type = AP_DEVICE_TYPE_CEX3C; else if (ap_test_bit(&device_functions, 4)) diff --git a/drivers/s390/crypto/ap_bus.h b/drivers/s390/crypto/ap_bus.h index 5018f66..e49bcf4 100644 --- a/drivers/s390/crypto/ap_bus.h +++ b/drivers/s390/crypto/ap_bus.h @@ -118,6 +118,14 @@ static inline int ap_test_bit(unsigned int *ptr, unsigned int nr) #define AP_DEVICE_TYPE_CEX3C 9 /* + * Known function facilities + */ +#define AP_FUNC_MEX4K 1 +#define AP_FUNC_CRT4K 2 +#define AP_FUNC_COPRO 3 +#define AP_FUNC_ACCEL 4 + +/* * AP reset flag states */ #define AP_RESET_IGNORE 0 /* request timeout will be ignored */ @@ -150,6 +158,7 @@ struct ap_device { ap_qid_t qid; /* AP queue id. */ int queue_depth; /* AP queue depth.*/ int device_type; /* AP device type. */ + unsigned int functions; /* AP device function bitfield. */ int unregistered; /* marks AP device as unregistered */ struct timer_list timeout; /* Timer for request timeouts. */ int reset; /* Reset required after req. timeout. */ @@ -225,6 +234,4 @@ void ap_flush_queue(struct ap_device *ap_dev); int ap_module_init(void); void ap_module_exit(void); -int ap_4096_commands_available(ap_qid_t qid); - #endif /* _AP_BUS_H_ */ diff --git a/drivers/s390/crypto/zcrypt_cex2a.c b/drivers/s390/crypto/zcrypt_cex2a.c index 744c668..8f5f70b 100644 --- a/drivers/s390/crypto/zcrypt_cex2a.c +++ b/drivers/s390/crypto/zcrypt_cex2a.c @@ -449,7 +449,8 @@ static int zcrypt_cex2a_probe(struct ap_device *ap_dev) zdev->min_mod_size = CEX2A_MIN_MOD_SIZE; zdev->max_mod_size = CEX2A_MAX_MOD_SIZE; zdev->max_exp_bit_length = CEX2A_MAX_MOD_SIZE; - if (ap_4096_commands_available(ap_dev->qid)) { + if (ap_test_bit(&ap_dev->functions, AP_FUNC_MEX4K) && + ap_test_bit(&ap_dev->functions, AP_FUNC_CRT4K)) { zdev->max_mod_size = CEX3A_MAX_MOD_SIZE; zdev->max_exp_bit_length = CEX3A_MAX_MOD_SIZE; } -- cgit v0.10.2 From 5e55a488c87aa6cc269b04bfec99e835f03b2c2d Mon Sep 17 00:00:00 2001 From: Holger Dengler Date: Tue, 28 Aug 2012 16:45:36 +0200 Subject: s390/zcrypt: Separate msgtype implementation from card modules. Msgtype implementations are now separated from card specific modules and can be dynamically registered. Existing msgtype implementations are restructured in modules. Signed-off-by: Holger Dengler Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/crypto/Makefile b/drivers/s390/crypto/Makefile index af3c7f1..eca0bfa 100644 --- a/drivers/s390/crypto/Makefile +++ b/drivers/s390/crypto/Makefile @@ -5,3 +5,4 @@ ap-objs := ap_bus.o obj-$(CONFIG_ZCRYPT) += ap.o zcrypt_api.o zcrypt_pcicc.o zcrypt_pcixcc.o obj-$(CONFIG_ZCRYPT) += zcrypt_pcica.o zcrypt_cex2a.o +obj-$(CONFIG_ZCRYPT) += zcrypt_msgtype6.o zcrypt_msgtype50.o diff --git a/drivers/s390/crypto/zcrypt_api.c b/drivers/s390/crypto/zcrypt_api.c index 2f94132..f1f026e 100644 --- a/drivers/s390/crypto/zcrypt_api.c +++ b/drivers/s390/crypto/zcrypt_api.c @@ -1,7 +1,7 @@ /* * zcrypt 2.1.0 * - * Copyright IBM Corp. 2001, 2006 + * Copyright IBM Corp. 2001, 2012 * Author(s): Robert Burroughs * Eric Rossman (edrossma@us.ibm.com) * Cornelia Huck @@ -9,6 +9,7 @@ * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) * Major cleanup & driver split: Martin Schwidefsky * Ralph Wuerthner + * MSGTYPE restruct: Holger Dengler * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -44,8 +45,8 @@ * Module description. */ MODULE_AUTHOR("IBM Corporation"); -MODULE_DESCRIPTION("Cryptographic Coprocessor interface, " - "Copyright IBM Corp. 2001, 2006"); +MODULE_DESCRIPTION("Cryptographic Coprocessor interface, " \ + "Copyright IBM Corp. 2001, 2012"); MODULE_LICENSE("GPL"); static DEFINE_SPINLOCK(zcrypt_device_lock); @@ -56,6 +57,9 @@ static atomic_t zcrypt_open_count = ATOMIC_INIT(0); static int zcrypt_rng_device_add(void); static void zcrypt_rng_device_remove(void); +static DEFINE_SPINLOCK(zcrypt_ops_list_lock); +static LIST_HEAD(zcrypt_ops_list); + /* * Device attributes common for all crypto devices. */ @@ -215,6 +219,8 @@ int zcrypt_device_register(struct zcrypt_device *zdev) { int rc; + if (!zdev->ops) + return -ENODEV; rc = sysfs_create_group(&zdev->ap_dev->device.kobj, &zcrypt_device_attr_group); if (rc) @@ -269,6 +275,67 @@ void zcrypt_device_unregister(struct zcrypt_device *zdev) } EXPORT_SYMBOL(zcrypt_device_unregister); +void zcrypt_msgtype_register(struct zcrypt_ops *zops) +{ + if (zops->owner) { + spin_lock_bh(&zcrypt_ops_list_lock); + list_add_tail(&zops->list, &zcrypt_ops_list); + spin_unlock_bh(&zcrypt_ops_list_lock); + } +} +EXPORT_SYMBOL(zcrypt_msgtype_register); + +void zcrypt_msgtype_unregister(struct zcrypt_ops *zops) +{ + spin_lock_bh(&zcrypt_ops_list_lock); + list_del_init(&zops->list); + spin_unlock_bh(&zcrypt_ops_list_lock); +} +EXPORT_SYMBOL(zcrypt_msgtype_unregister); + +static inline +struct zcrypt_ops *__ops_lookup(unsigned char *name, int variant) +{ + struct zcrypt_ops *zops; + int found = 0; + + spin_lock_bh(&zcrypt_ops_list_lock); + list_for_each_entry(zops, &zcrypt_ops_list, list) { + if ((zops->variant == variant) && + (!strncmp(zops->owner->name, name, MODULE_NAME_LEN))) { + found = 1; + break; + } + } + spin_unlock_bh(&zcrypt_ops_list_lock); + + if (!found) + return NULL; + return zops; +} + +struct zcrypt_ops *zcrypt_msgtype_request(unsigned char *name, int variant) +{ + struct zcrypt_ops *zops = NULL; + + zops = __ops_lookup(name, variant); + if (!zops) { + request_module(name); + zops = __ops_lookup(name, variant); + } + if ((!zops) || (!try_module_get(zops->owner))) + return NULL; + return zops; +} +EXPORT_SYMBOL(zcrypt_msgtype_request); + +void zcrypt_msgtype_release(struct zcrypt_ops *zops) +{ + if (zops) + module_put(zops->owner); +} +EXPORT_SYMBOL(zcrypt_msgtype_release); + /** * zcrypt_read (): Not supported beyond zcrypt 1.3.1. * diff --git a/drivers/s390/crypto/zcrypt_api.h b/drivers/s390/crypto/zcrypt_api.h index 7a32c4b..02b2d35 100644 --- a/drivers/s390/crypto/zcrypt_api.h +++ b/drivers/s390/crypto/zcrypt_api.h @@ -1,7 +1,7 @@ /* * zcrypt 2.1.0 * - * Copyright IBM Corp. 2001, 2006 + * Copyright IBM Corp. 2001, 2012 * Author(s): Robert Burroughs * Eric Rossman (edrossma@us.ibm.com) * Cornelia Huck @@ -9,6 +9,7 @@ * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) * Major cleanup & driver split: Martin Schwidefsky * Ralph Wuerthner + * MSGTYPE restruct: Holger Dengler * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -87,6 +88,9 @@ struct zcrypt_ops { struct ica_rsa_modexpo_crt *); long (*send_cprb)(struct zcrypt_device *, struct ica_xcRB *); long (*rng)(struct zcrypt_device *, char *); + struct list_head list; /* zcrypt ops list. */ + struct module *owner; + int variant; }; struct zcrypt_device { @@ -116,6 +120,10 @@ void zcrypt_device_get(struct zcrypt_device *); int zcrypt_device_put(struct zcrypt_device *); int zcrypt_device_register(struct zcrypt_device *); void zcrypt_device_unregister(struct zcrypt_device *); +void zcrypt_msgtype_register(struct zcrypt_ops *); +void zcrypt_msgtype_unregister(struct zcrypt_ops *); +struct zcrypt_ops *zcrypt_msgtype_request(unsigned char *, int); +void zcrypt_msgtype_release(struct zcrypt_ops *); int zcrypt_api_init(void); void zcrypt_api_exit(void); diff --git a/drivers/s390/crypto/zcrypt_cex2a.c b/drivers/s390/crypto/zcrypt_cex2a.c index 8f5f70b..1e849d6 100644 --- a/drivers/s390/crypto/zcrypt_cex2a.c +++ b/drivers/s390/crypto/zcrypt_cex2a.c @@ -1,13 +1,14 @@ /* * zcrypt 2.1.0 * - * Copyright IBM Corp. 2001, 2006 + * Copyright IBM Corp. 2001, 2012 * Author(s): Robert Burroughs * Eric Rossman (edrossma@us.ibm.com) * * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) * Major cleanup & driver split: Martin Schwidefsky * Ralph Wuerthner + * MSGTYPE restruct: Holger Dengler * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -35,6 +36,7 @@ #include "zcrypt_api.h" #include "zcrypt_error.h" #include "zcrypt_cex2a.h" +#include "zcrypt_msgtype50.h" #define CEX2A_MIN_MOD_SIZE 1 /* 8 bits */ #define CEX2A_MAX_MOD_SIZE 256 /* 2048 bits */ @@ -63,14 +65,12 @@ static struct ap_device_id zcrypt_cex2a_ids[] = { MODULE_DEVICE_TABLE(ap, zcrypt_cex2a_ids); MODULE_AUTHOR("IBM Corporation"); -MODULE_DESCRIPTION("CEX2A Cryptographic Coprocessor device driver, " - "Copyright IBM Corp. 2001, 2006"); +MODULE_DESCRIPTION("CEX2A Cryptographic Coprocessor device driver, " \ + "Copyright IBM Corp. 2001, 2012"); MODULE_LICENSE("GPL"); static int zcrypt_cex2a_probe(struct ap_device *ap_dev); static void zcrypt_cex2a_remove(struct ap_device *ap_dev); -static void zcrypt_cex2a_receive(struct ap_device *, struct ap_message *, - struct ap_message *); static struct ap_driver zcrypt_cex2a_driver = { .probe = zcrypt_cex2a_probe, @@ -80,344 +80,6 @@ static struct ap_driver zcrypt_cex2a_driver = { }; /** - * Convert a ICAMEX message to a type50 MEX message. - * - * @zdev: crypto device pointer - * @zreq: crypto request pointer - * @mex: pointer to user input data - * - * Returns 0 on success or -EFAULT. - */ -static int ICAMEX_msg_to_type50MEX_msg(struct zcrypt_device *zdev, - struct ap_message *ap_msg, - struct ica_rsa_modexpo *mex) -{ - unsigned char *mod, *exp, *inp; - int mod_len; - - mod_len = mex->inputdatalength; - - if (mod_len <= 128) { - struct type50_meb1_msg *meb1 = ap_msg->message; - memset(meb1, 0, sizeof(*meb1)); - ap_msg->length = sizeof(*meb1); - meb1->header.msg_type_code = TYPE50_TYPE_CODE; - meb1->header.msg_len = sizeof(*meb1); - meb1->keyblock_type = TYPE50_MEB1_FMT; - mod = meb1->modulus + sizeof(meb1->modulus) - mod_len; - exp = meb1->exponent + sizeof(meb1->exponent) - mod_len; - inp = meb1->message + sizeof(meb1->message) - mod_len; - } else if (mod_len <= 256) { - struct type50_meb2_msg *meb2 = ap_msg->message; - memset(meb2, 0, sizeof(*meb2)); - ap_msg->length = sizeof(*meb2); - meb2->header.msg_type_code = TYPE50_TYPE_CODE; - meb2->header.msg_len = sizeof(*meb2); - meb2->keyblock_type = TYPE50_MEB2_FMT; - mod = meb2->modulus + sizeof(meb2->modulus) - mod_len; - exp = meb2->exponent + sizeof(meb2->exponent) - mod_len; - inp = meb2->message + sizeof(meb2->message) - mod_len; - } else { - /* mod_len > 256 = 4096 bit RSA Key */ - struct type50_meb3_msg *meb3 = ap_msg->message; - memset(meb3, 0, sizeof(*meb3)); - ap_msg->length = sizeof(*meb3); - meb3->header.msg_type_code = TYPE50_TYPE_CODE; - meb3->header.msg_len = sizeof(*meb3); - meb3->keyblock_type = TYPE50_MEB3_FMT; - mod = meb3->modulus + sizeof(meb3->modulus) - mod_len; - exp = meb3->exponent + sizeof(meb3->exponent) - mod_len; - inp = meb3->message + sizeof(meb3->message) - mod_len; - } - - if (copy_from_user(mod, mex->n_modulus, mod_len) || - copy_from_user(exp, mex->b_key, mod_len) || - copy_from_user(inp, mex->inputdata, mod_len)) - return -EFAULT; - return 0; -} - -/** - * Convert a ICACRT message to a type50 CRT message. - * - * @zdev: crypto device pointer - * @zreq: crypto request pointer - * @crt: pointer to user input data - * - * Returns 0 on success or -EFAULT. - */ -static int ICACRT_msg_to_type50CRT_msg(struct zcrypt_device *zdev, - struct ap_message *ap_msg, - struct ica_rsa_modexpo_crt *crt) -{ - int mod_len, short_len, long_len, long_offset, limit; - unsigned char *p, *q, *dp, *dq, *u, *inp; - - mod_len = crt->inputdatalength; - short_len = mod_len / 2; - long_len = mod_len / 2 + 8; - - /* - * CEX2A cannot handle p, dp, or U > 128 bytes. - * If we have one of these, we need to do extra checking. - * For CEX3A the limit is 256 bytes. - */ - if (zdev->max_mod_size == CEX3A_MAX_MOD_SIZE) - limit = 256; - else - limit = 128; - - if (long_len > limit) { - /* - * zcrypt_rsa_crt already checked for the leading - * zeroes of np_prime, bp_key and u_mult_inc. - */ - long_offset = long_len - limit; - long_len = limit; - } else - long_offset = 0; - - /* - * Instead of doing extra work for p, dp, U > 64 bytes, we'll just use - * the larger message structure. - */ - if (long_len <= 64) { - struct type50_crb1_msg *crb1 = ap_msg->message; - memset(crb1, 0, sizeof(*crb1)); - ap_msg->length = sizeof(*crb1); - crb1->header.msg_type_code = TYPE50_TYPE_CODE; - crb1->header.msg_len = sizeof(*crb1); - crb1->keyblock_type = TYPE50_CRB1_FMT; - p = crb1->p + sizeof(crb1->p) - long_len; - q = crb1->q + sizeof(crb1->q) - short_len; - dp = crb1->dp + sizeof(crb1->dp) - long_len; - dq = crb1->dq + sizeof(crb1->dq) - short_len; - u = crb1->u + sizeof(crb1->u) - long_len; - inp = crb1->message + sizeof(crb1->message) - mod_len; - } else if (long_len <= 128) { - struct type50_crb2_msg *crb2 = ap_msg->message; - memset(crb2, 0, sizeof(*crb2)); - ap_msg->length = sizeof(*crb2); - crb2->header.msg_type_code = TYPE50_TYPE_CODE; - crb2->header.msg_len = sizeof(*crb2); - crb2->keyblock_type = TYPE50_CRB2_FMT; - p = crb2->p + sizeof(crb2->p) - long_len; - q = crb2->q + sizeof(crb2->q) - short_len; - dp = crb2->dp + sizeof(crb2->dp) - long_len; - dq = crb2->dq + sizeof(crb2->dq) - short_len; - u = crb2->u + sizeof(crb2->u) - long_len; - inp = crb2->message + sizeof(crb2->message) - mod_len; - } else { - /* long_len >= 256 */ - struct type50_crb3_msg *crb3 = ap_msg->message; - memset(crb3, 0, sizeof(*crb3)); - ap_msg->length = sizeof(*crb3); - crb3->header.msg_type_code = TYPE50_TYPE_CODE; - crb3->header.msg_len = sizeof(*crb3); - crb3->keyblock_type = TYPE50_CRB3_FMT; - p = crb3->p + sizeof(crb3->p) - long_len; - q = crb3->q + sizeof(crb3->q) - short_len; - dp = crb3->dp + sizeof(crb3->dp) - long_len; - dq = crb3->dq + sizeof(crb3->dq) - short_len; - u = crb3->u + sizeof(crb3->u) - long_len; - inp = crb3->message + sizeof(crb3->message) - mod_len; - } - - if (copy_from_user(p, crt->np_prime + long_offset, long_len) || - copy_from_user(q, crt->nq_prime, short_len) || - copy_from_user(dp, crt->bp_key + long_offset, long_len) || - copy_from_user(dq, crt->bq_key, short_len) || - copy_from_user(u, crt->u_mult_inv + long_offset, long_len) || - copy_from_user(inp, crt->inputdata, mod_len)) - return -EFAULT; - - return 0; -} - -/** - * Copy results from a type 80 reply message back to user space. - * - * @zdev: crypto device pointer - * @reply: reply AP message. - * @data: pointer to user output data - * @length: size of user output data - * - * Returns 0 on success or -EFAULT. - */ -static int convert_type80(struct zcrypt_device *zdev, - struct ap_message *reply, - char __user *outputdata, - unsigned int outputdatalength) -{ - struct type80_hdr *t80h = reply->message; - unsigned char *data; - - if (t80h->len < sizeof(*t80h) + outputdatalength) { - /* The result is too short, the CEX2A card may not do that.. */ - zdev->online = 0; - return -EAGAIN; /* repeat the request on a different device. */ - } - if (zdev->user_space_type == ZCRYPT_CEX2A) - BUG_ON(t80h->len > CEX2A_MAX_RESPONSE_SIZE); - else - BUG_ON(t80h->len > CEX3A_MAX_RESPONSE_SIZE); - data = reply->message + t80h->len - outputdatalength; - if (copy_to_user(outputdata, data, outputdatalength)) - return -EFAULT; - return 0; -} - -static int convert_response(struct zcrypt_device *zdev, - struct ap_message *reply, - char __user *outputdata, - unsigned int outputdatalength) -{ - /* Response type byte is the second byte in the response. */ - switch (((unsigned char *) reply->message)[1]) { - case TYPE82_RSP_CODE: - case TYPE88_RSP_CODE: - return convert_error(zdev, reply); - case TYPE80_RSP_CODE: - return convert_type80(zdev, reply, - outputdata, outputdatalength); - default: /* Unknown response type, this should NEVER EVER happen */ - zdev->online = 0; - return -EAGAIN; /* repeat the request on a different device. */ - } -} - -/** - * This function is called from the AP bus code after a crypto request - * "msg" has finished with the reply message "reply". - * It is called from tasklet context. - * @ap_dev: pointer to the AP device - * @msg: pointer to the AP message - * @reply: pointer to the AP reply message - */ -static void zcrypt_cex2a_receive(struct ap_device *ap_dev, - struct ap_message *msg, - struct ap_message *reply) -{ - static struct error_hdr error_reply = { - .type = TYPE82_RSP_CODE, - .reply_code = REP82_ERROR_MACHINE_FAILURE, - }; - struct type80_hdr *t80h; - int length; - - /* Copy the reply message to the request message buffer. */ - if (IS_ERR(reply)) { - memcpy(msg->message, &error_reply, sizeof(error_reply)); - goto out; - } - t80h = reply->message; - if (t80h->type == TYPE80_RSP_CODE) { - if (ap_dev->device_type == AP_DEVICE_TYPE_CEX2A) - length = min(CEX2A_MAX_RESPONSE_SIZE, (int) t80h->len); - else - length = min(CEX3A_MAX_RESPONSE_SIZE, (int) t80h->len); - memcpy(msg->message, reply->message, length); - } else - memcpy(msg->message, reply->message, sizeof error_reply); -out: - complete((struct completion *) msg->private); -} - -static atomic_t zcrypt_step = ATOMIC_INIT(0); - -/** - * The request distributor calls this function if it picked the CEX2A - * device to handle a modexpo request. - * @zdev: pointer to zcrypt_device structure that identifies the - * CEX2A device to the request distributor - * @mex: pointer to the modexpo request buffer - */ -static long zcrypt_cex2a_modexpo(struct zcrypt_device *zdev, - struct ica_rsa_modexpo *mex) -{ - struct ap_message ap_msg; - struct completion work; - int rc; - - ap_init_message(&ap_msg); - if (zdev->user_space_type == ZCRYPT_CEX2A) - ap_msg.message = kmalloc(CEX2A_MAX_MESSAGE_SIZE, GFP_KERNEL); - else - ap_msg.message = kmalloc(CEX3A_MAX_MESSAGE_SIZE, GFP_KERNEL); - if (!ap_msg.message) - return -ENOMEM; - ap_msg.receive = zcrypt_cex2a_receive; - ap_msg.psmid = (((unsigned long long) current->pid) << 32) + - atomic_inc_return(&zcrypt_step); - ap_msg.private = &work; - rc = ICAMEX_msg_to_type50MEX_msg(zdev, &ap_msg, mex); - if (rc) - goto out_free; - init_completion(&work); - ap_queue_message(zdev->ap_dev, &ap_msg); - rc = wait_for_completion_interruptible(&work); - if (rc == 0) - rc = convert_response(zdev, &ap_msg, mex->outputdata, - mex->outputdatalength); - else - /* Signal pending. */ - ap_cancel_message(zdev->ap_dev, &ap_msg); -out_free: - kfree(ap_msg.message); - return rc; -} - -/** - * The request distributor calls this function if it picked the CEX2A - * device to handle a modexpo_crt request. - * @zdev: pointer to zcrypt_device structure that identifies the - * CEX2A device to the request distributor - * @crt: pointer to the modexpoc_crt request buffer - */ -static long zcrypt_cex2a_modexpo_crt(struct zcrypt_device *zdev, - struct ica_rsa_modexpo_crt *crt) -{ - struct ap_message ap_msg; - struct completion work; - int rc; - - ap_init_message(&ap_msg); - if (zdev->user_space_type == ZCRYPT_CEX2A) - ap_msg.message = kmalloc(CEX2A_MAX_MESSAGE_SIZE, GFP_KERNEL); - else - ap_msg.message = kmalloc(CEX3A_MAX_MESSAGE_SIZE, GFP_KERNEL); - if (!ap_msg.message) - return -ENOMEM; - ap_msg.receive = zcrypt_cex2a_receive; - ap_msg.psmid = (((unsigned long long) current->pid) << 32) + - atomic_inc_return(&zcrypt_step); - ap_msg.private = &work; - rc = ICACRT_msg_to_type50CRT_msg(zdev, &ap_msg, crt); - if (rc) - goto out_free; - init_completion(&work); - ap_queue_message(zdev->ap_dev, &ap_msg); - rc = wait_for_completion_interruptible(&work); - if (rc == 0) - rc = convert_response(zdev, &ap_msg, crt->outputdata, - crt->outputdatalength); - else - /* Signal pending. */ - ap_cancel_message(zdev->ap_dev, &ap_msg); -out_free: - kfree(ap_msg.message); - return rc; -} - -/** - * The crypto operations for a CEX2A card. - */ -static struct zcrypt_ops zcrypt_cex2a_ops = { - .rsa_modexpo = zcrypt_cex2a_modexpo, - .rsa_modexpo_crt = zcrypt_cex2a_modexpo_crt, -}; - -/** * Probe function for CEX2A cards. It always accepts the AP device * since the bus_match already checked the hardware type. * @ap_dev: pointer to the AP device. @@ -458,16 +120,18 @@ static int zcrypt_cex2a_probe(struct ap_device *ap_dev) zdev->speed_rating = CEX3A_SPEED_RATING; break; } - if (zdev != NULL) { - zdev->ap_dev = ap_dev; - zdev->ops = &zcrypt_cex2a_ops; - zdev->online = 1; - ap_dev->reply = &zdev->reply; - ap_dev->private = zdev; - rc = zcrypt_device_register(zdev); - } + if (!zdev) + return -ENODEV; + zdev->ops = zcrypt_msgtype_request(MSGTYPE50_NAME, + MSGTYPE50_VARIANT_DEFAULT); + zdev->ap_dev = ap_dev; + zdev->online = 1; + ap_dev->reply = &zdev->reply; + ap_dev->private = zdev; + rc = zcrypt_device_register(zdev); if (rc) { ap_dev->private = NULL; + zcrypt_msgtype_release(zdev->ops); zcrypt_device_free(zdev); } return rc; @@ -480,8 +144,10 @@ static int zcrypt_cex2a_probe(struct ap_device *ap_dev) static void zcrypt_cex2a_remove(struct ap_device *ap_dev) { struct zcrypt_device *zdev = ap_dev->private; + struct zcrypt_ops *zops = zdev->ops; zcrypt_device_unregister(zdev); + zcrypt_msgtype_release(zops); } int __init zcrypt_cex2a_init(void) diff --git a/drivers/s390/crypto/zcrypt_msgtype50.c b/drivers/s390/crypto/zcrypt_msgtype50.c new file mode 100644 index 0000000..035b6dc --- /dev/null +++ b/drivers/s390/crypto/zcrypt_msgtype50.c @@ -0,0 +1,531 @@ +/* + * zcrypt 2.1.0 + * + * Copyright IBM Corp. 2001, 2012 + * Author(s): Robert Burroughs + * Eric Rossman (edrossma@us.ibm.com) + * + * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) + * Major cleanup & driver split: Martin Schwidefsky + * Ralph Wuerthner + * MSGTYPE restruct: Holger Dengler + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2, or (at your option) + * any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#include +#include +#include +#include +#include +#include + +#include "ap_bus.h" +#include "zcrypt_api.h" +#include "zcrypt_error.h" +#include "zcrypt_msgtype50.h" + +#define CEX3A_MAX_MOD_SIZE 512 /* 4096 bits */ + +#define CEX2A_MAX_RESPONSE_SIZE 0x110 /* max outputdatalength + type80_hdr */ + +#define CEX3A_MAX_RESPONSE_SIZE 0x210 /* 512 bit modulus + * (max outputdatalength) + + * type80_hdr*/ + +MODULE_AUTHOR("IBM Corporation"); +MODULE_DESCRIPTION("Cryptographic Accelerator (message type 50), " \ + "Copyright IBM Corp. 2001, 2012"); +MODULE_LICENSE("GPL"); + +static void zcrypt_cex2a_receive(struct ap_device *, struct ap_message *, + struct ap_message *); + +/** + * The type 50 message family is associated with a CEX2A card. + * + * The four members of the family are described below. + * + * Note that all unsigned char arrays are right-justified and left-padded + * with zeroes. + * + * Note that all reserved fields must be zeroes. + */ +struct type50_hdr { + unsigned char reserved1; + unsigned char msg_type_code; /* 0x50 */ + unsigned short msg_len; + unsigned char reserved2; + unsigned char ignored; + unsigned short reserved3; +} __packed; + +#define TYPE50_TYPE_CODE 0x50 + +#define TYPE50_MEB1_FMT 0x0001 +#define TYPE50_MEB2_FMT 0x0002 +#define TYPE50_MEB3_FMT 0x0003 +#define TYPE50_CRB1_FMT 0x0011 +#define TYPE50_CRB2_FMT 0x0012 +#define TYPE50_CRB3_FMT 0x0013 + +/* Mod-Exp, with a small modulus */ +struct type50_meb1_msg { + struct type50_hdr header; + unsigned short keyblock_type; /* 0x0001 */ + unsigned char reserved[6]; + unsigned char exponent[128]; + unsigned char modulus[128]; + unsigned char message[128]; +} __packed; + +/* Mod-Exp, with a large modulus */ +struct type50_meb2_msg { + struct type50_hdr header; + unsigned short keyblock_type; /* 0x0002 */ + unsigned char reserved[6]; + unsigned char exponent[256]; + unsigned char modulus[256]; + unsigned char message[256]; +} __packed; + +/* Mod-Exp, with a larger modulus */ +struct type50_meb3_msg { + struct type50_hdr header; + unsigned short keyblock_type; /* 0x0003 */ + unsigned char reserved[6]; + unsigned char exponent[512]; + unsigned char modulus[512]; + unsigned char message[512]; +} __packed; + +/* CRT, with a small modulus */ +struct type50_crb1_msg { + struct type50_hdr header; + unsigned short keyblock_type; /* 0x0011 */ + unsigned char reserved[6]; + unsigned char p[64]; + unsigned char q[64]; + unsigned char dp[64]; + unsigned char dq[64]; + unsigned char u[64]; + unsigned char message[128]; +} __packed; + +/* CRT, with a large modulus */ +struct type50_crb2_msg { + struct type50_hdr header; + unsigned short keyblock_type; /* 0x0012 */ + unsigned char reserved[6]; + unsigned char p[128]; + unsigned char q[128]; + unsigned char dp[128]; + unsigned char dq[128]; + unsigned char u[128]; + unsigned char message[256]; +} __packed; + +/* CRT, with a larger modulus */ +struct type50_crb3_msg { + struct type50_hdr header; + unsigned short keyblock_type; /* 0x0013 */ + unsigned char reserved[6]; + unsigned char p[256]; + unsigned char q[256]; + unsigned char dp[256]; + unsigned char dq[256]; + unsigned char u[256]; + unsigned char message[512]; +} __packed; + +/** + * The type 80 response family is associated with a CEX2A card. + * + * Note that all unsigned char arrays are right-justified and left-padded + * with zeroes. + * + * Note that all reserved fields must be zeroes. + */ + +#define TYPE80_RSP_CODE 0x80 + +struct type80_hdr { + unsigned char reserved1; + unsigned char type; /* 0x80 */ + unsigned short len; + unsigned char code; /* 0x00 */ + unsigned char reserved2[3]; + unsigned char reserved3[8]; +} __packed; + +/** + * Convert a ICAMEX message to a type50 MEX message. + * + * @zdev: crypto device pointer + * @zreq: crypto request pointer + * @mex: pointer to user input data + * + * Returns 0 on success or -EFAULT. + */ +static int ICAMEX_msg_to_type50MEX_msg(struct zcrypt_device *zdev, + struct ap_message *ap_msg, + struct ica_rsa_modexpo *mex) +{ + unsigned char *mod, *exp, *inp; + int mod_len; + + mod_len = mex->inputdatalength; + + if (mod_len <= 128) { + struct type50_meb1_msg *meb1 = ap_msg->message; + memset(meb1, 0, sizeof(*meb1)); + ap_msg->length = sizeof(*meb1); + meb1->header.msg_type_code = TYPE50_TYPE_CODE; + meb1->header.msg_len = sizeof(*meb1); + meb1->keyblock_type = TYPE50_MEB1_FMT; + mod = meb1->modulus + sizeof(meb1->modulus) - mod_len; + exp = meb1->exponent + sizeof(meb1->exponent) - mod_len; + inp = meb1->message + sizeof(meb1->message) - mod_len; + } else if (mod_len <= 256) { + struct type50_meb2_msg *meb2 = ap_msg->message; + memset(meb2, 0, sizeof(*meb2)); + ap_msg->length = sizeof(*meb2); + meb2->header.msg_type_code = TYPE50_TYPE_CODE; + meb2->header.msg_len = sizeof(*meb2); + meb2->keyblock_type = TYPE50_MEB2_FMT; + mod = meb2->modulus + sizeof(meb2->modulus) - mod_len; + exp = meb2->exponent + sizeof(meb2->exponent) - mod_len; + inp = meb2->message + sizeof(meb2->message) - mod_len; + } else { + /* mod_len > 256 = 4096 bit RSA Key */ + struct type50_meb3_msg *meb3 = ap_msg->message; + memset(meb3, 0, sizeof(*meb3)); + ap_msg->length = sizeof(*meb3); + meb3->header.msg_type_code = TYPE50_TYPE_CODE; + meb3->header.msg_len = sizeof(*meb3); + meb3->keyblock_type = TYPE50_MEB3_FMT; + mod = meb3->modulus + sizeof(meb3->modulus) - mod_len; + exp = meb3->exponent + sizeof(meb3->exponent) - mod_len; + inp = meb3->message + sizeof(meb3->message) - mod_len; + } + + if (copy_from_user(mod, mex->n_modulus, mod_len) || + copy_from_user(exp, mex->b_key, mod_len) || + copy_from_user(inp, mex->inputdata, mod_len)) + return -EFAULT; + return 0; +} + +/** + * Convert a ICACRT message to a type50 CRT message. + * + * @zdev: crypto device pointer + * @zreq: crypto request pointer + * @crt: pointer to user input data + * + * Returns 0 on success or -EFAULT. + */ +static int ICACRT_msg_to_type50CRT_msg(struct zcrypt_device *zdev, + struct ap_message *ap_msg, + struct ica_rsa_modexpo_crt *crt) +{ + int mod_len, short_len, long_len, long_offset, limit; + unsigned char *p, *q, *dp, *dq, *u, *inp; + + mod_len = crt->inputdatalength; + short_len = mod_len / 2; + long_len = mod_len / 2 + 8; + + /* + * CEX2A cannot handle p, dp, or U > 128 bytes. + * If we have one of these, we need to do extra checking. + * For CEX3A the limit is 256 bytes. + */ + if (zdev->max_mod_size == CEX3A_MAX_MOD_SIZE) + limit = 256; + else + limit = 128; + + if (long_len > limit) { + /* + * zcrypt_rsa_crt already checked for the leading + * zeroes of np_prime, bp_key and u_mult_inc. + */ + long_offset = long_len - limit; + long_len = limit; + } else + long_offset = 0; + + /* + * Instead of doing extra work for p, dp, U > 64 bytes, we'll just use + * the larger message structure. + */ + if (long_len <= 64) { + struct type50_crb1_msg *crb1 = ap_msg->message; + memset(crb1, 0, sizeof(*crb1)); + ap_msg->length = sizeof(*crb1); + crb1->header.msg_type_code = TYPE50_TYPE_CODE; + crb1->header.msg_len = sizeof(*crb1); + crb1->keyblock_type = TYPE50_CRB1_FMT; + p = crb1->p + sizeof(crb1->p) - long_len; + q = crb1->q + sizeof(crb1->q) - short_len; + dp = crb1->dp + sizeof(crb1->dp) - long_len; + dq = crb1->dq + sizeof(crb1->dq) - short_len; + u = crb1->u + sizeof(crb1->u) - long_len; + inp = crb1->message + sizeof(crb1->message) - mod_len; + } else if (long_len <= 128) { + struct type50_crb2_msg *crb2 = ap_msg->message; + memset(crb2, 0, sizeof(*crb2)); + ap_msg->length = sizeof(*crb2); + crb2->header.msg_type_code = TYPE50_TYPE_CODE; + crb2->header.msg_len = sizeof(*crb2); + crb2->keyblock_type = TYPE50_CRB2_FMT; + p = crb2->p + sizeof(crb2->p) - long_len; + q = crb2->q + sizeof(crb2->q) - short_len; + dp = crb2->dp + sizeof(crb2->dp) - long_len; + dq = crb2->dq + sizeof(crb2->dq) - short_len; + u = crb2->u + sizeof(crb2->u) - long_len; + inp = crb2->message + sizeof(crb2->message) - mod_len; + } else { + /* long_len >= 256 */ + struct type50_crb3_msg *crb3 = ap_msg->message; + memset(crb3, 0, sizeof(*crb3)); + ap_msg->length = sizeof(*crb3); + crb3->header.msg_type_code = TYPE50_TYPE_CODE; + crb3->header.msg_len = sizeof(*crb3); + crb3->keyblock_type = TYPE50_CRB3_FMT; + p = crb3->p + sizeof(crb3->p) - long_len; + q = crb3->q + sizeof(crb3->q) - short_len; + dp = crb3->dp + sizeof(crb3->dp) - long_len; + dq = crb3->dq + sizeof(crb3->dq) - short_len; + u = crb3->u + sizeof(crb3->u) - long_len; + inp = crb3->message + sizeof(crb3->message) - mod_len; + } + + if (copy_from_user(p, crt->np_prime + long_offset, long_len) || + copy_from_user(q, crt->nq_prime, short_len) || + copy_from_user(dp, crt->bp_key + long_offset, long_len) || + copy_from_user(dq, crt->bq_key, short_len) || + copy_from_user(u, crt->u_mult_inv + long_offset, long_len) || + copy_from_user(inp, crt->inputdata, mod_len)) + return -EFAULT; + + return 0; +} + +/** + * Copy results from a type 80 reply message back to user space. + * + * @zdev: crypto device pointer + * @reply: reply AP message. + * @data: pointer to user output data + * @length: size of user output data + * + * Returns 0 on success or -EFAULT. + */ +static int convert_type80(struct zcrypt_device *zdev, + struct ap_message *reply, + char __user *outputdata, + unsigned int outputdatalength) +{ + struct type80_hdr *t80h = reply->message; + unsigned char *data; + + if (t80h->len < sizeof(*t80h) + outputdatalength) { + /* The result is too short, the CEX2A card may not do that.. */ + zdev->online = 0; + return -EAGAIN; /* repeat the request on a different device. */ + } + if (zdev->user_space_type == ZCRYPT_CEX2A) + BUG_ON(t80h->len > CEX2A_MAX_RESPONSE_SIZE); + else + BUG_ON(t80h->len > CEX3A_MAX_RESPONSE_SIZE); + data = reply->message + t80h->len - outputdatalength; + if (copy_to_user(outputdata, data, outputdatalength)) + return -EFAULT; + return 0; +} + +static int convert_response(struct zcrypt_device *zdev, + struct ap_message *reply, + char __user *outputdata, + unsigned int outputdatalength) +{ + /* Response type byte is the second byte in the response. */ + switch (((unsigned char *) reply->message)[1]) { + case TYPE82_RSP_CODE: + case TYPE88_RSP_CODE: + return convert_error(zdev, reply); + case TYPE80_RSP_CODE: + return convert_type80(zdev, reply, + outputdata, outputdatalength); + default: /* Unknown response type, this should NEVER EVER happen */ + zdev->online = 0; + return -EAGAIN; /* repeat the request on a different device. */ + } +} + +/** + * This function is called from the AP bus code after a crypto request + * "msg" has finished with the reply message "reply". + * It is called from tasklet context. + * @ap_dev: pointer to the AP device + * @msg: pointer to the AP message + * @reply: pointer to the AP reply message + */ +static void zcrypt_cex2a_receive(struct ap_device *ap_dev, + struct ap_message *msg, + struct ap_message *reply) +{ + static struct error_hdr error_reply = { + .type = TYPE82_RSP_CODE, + .reply_code = REP82_ERROR_MACHINE_FAILURE, + }; + struct type80_hdr *t80h; + int length; + + /* Copy the reply message to the request message buffer. */ + if (IS_ERR(reply)) { + memcpy(msg->message, &error_reply, sizeof(error_reply)); + goto out; + } + t80h = reply->message; + if (t80h->type == TYPE80_RSP_CODE) { + if (ap_dev->device_type == AP_DEVICE_TYPE_CEX2A) + length = min_t(int, + CEX2A_MAX_RESPONSE_SIZE, t80h->len); + else + length = min_t(int, + CEX3A_MAX_RESPONSE_SIZE, t80h->len); + memcpy(msg->message, reply->message, length); + } else + memcpy(msg->message, reply->message, sizeof(error_reply)); +out: + complete((struct completion *) msg->private); +} + +static atomic_t zcrypt_step = ATOMIC_INIT(0); + +/** + * The request distributor calls this function if it picked the CEX2A + * device to handle a modexpo request. + * @zdev: pointer to zcrypt_device structure that identifies the + * CEX2A device to the request distributor + * @mex: pointer to the modexpo request buffer + */ +static long zcrypt_cex2a_modexpo(struct zcrypt_device *zdev, + struct ica_rsa_modexpo *mex) +{ + struct ap_message ap_msg; + struct completion work; + int rc; + + ap_init_message(&ap_msg); + if (zdev->user_space_type == ZCRYPT_CEX2A) + ap_msg.message = kmalloc(MSGTYPE50_CRB2_MAX_MSG_SIZE, + GFP_KERNEL); + else + ap_msg.message = kmalloc(MSGTYPE50_CRB3_MAX_MSG_SIZE, + GFP_KERNEL); + if (!ap_msg.message) + return -ENOMEM; + ap_msg.receive = zcrypt_cex2a_receive; + ap_msg.psmid = (((unsigned long long) current->pid) << 32) + + atomic_inc_return(&zcrypt_step); + ap_msg.private = &work; + rc = ICAMEX_msg_to_type50MEX_msg(zdev, &ap_msg, mex); + if (rc) + goto out_free; + init_completion(&work); + ap_queue_message(zdev->ap_dev, &ap_msg); + rc = wait_for_completion_interruptible(&work); + if (rc == 0) + rc = convert_response(zdev, &ap_msg, mex->outputdata, + mex->outputdatalength); + else + /* Signal pending. */ + ap_cancel_message(zdev->ap_dev, &ap_msg); +out_free: + kfree(ap_msg.message); + return rc; +} + +/** + * The request distributor calls this function if it picked the CEX2A + * device to handle a modexpo_crt request. + * @zdev: pointer to zcrypt_device structure that identifies the + * CEX2A device to the request distributor + * @crt: pointer to the modexpoc_crt request buffer + */ +static long zcrypt_cex2a_modexpo_crt(struct zcrypt_device *zdev, + struct ica_rsa_modexpo_crt *crt) +{ + struct ap_message ap_msg; + struct completion work; + int rc; + + ap_init_message(&ap_msg); + if (zdev->user_space_type == ZCRYPT_CEX2A) + ap_msg.message = kmalloc(MSGTYPE50_CRB2_MAX_MSG_SIZE, + GFP_KERNEL); + else + ap_msg.message = kmalloc(MSGTYPE50_CRB3_MAX_MSG_SIZE, + GFP_KERNEL); + if (!ap_msg.message) + return -ENOMEM; + ap_msg.receive = zcrypt_cex2a_receive; + ap_msg.psmid = (((unsigned long long) current->pid) << 32) + + atomic_inc_return(&zcrypt_step); + ap_msg.private = &work; + rc = ICACRT_msg_to_type50CRT_msg(zdev, &ap_msg, crt); + if (rc) + goto out_free; + init_completion(&work); + ap_queue_message(zdev->ap_dev, &ap_msg); + rc = wait_for_completion_interruptible(&work); + if (rc == 0) + rc = convert_response(zdev, &ap_msg, crt->outputdata, + crt->outputdatalength); + else + /* Signal pending. */ + ap_cancel_message(zdev->ap_dev, &ap_msg); +out_free: + kfree(ap_msg.message); + return rc; +} + +/** + * The crypto operations for message type 50. + */ +static struct zcrypt_ops zcrypt_msgtype50_ops = { + .rsa_modexpo = zcrypt_cex2a_modexpo, + .rsa_modexpo_crt = zcrypt_cex2a_modexpo_crt, + .owner = THIS_MODULE, + .variant = MSGTYPE50_VARIANT_DEFAULT, +}; + +int __init zcrypt_msgtype50_init(void) +{ + zcrypt_msgtype_register(&zcrypt_msgtype50_ops); + return 0; +} + +void __exit zcrypt_msgtype50_exit(void) +{ + zcrypt_msgtype_unregister(&zcrypt_msgtype50_ops); +} + +module_init(zcrypt_msgtype50_init); +module_exit(zcrypt_msgtype50_exit); diff --git a/drivers/s390/crypto/zcrypt_msgtype50.h b/drivers/s390/crypto/zcrypt_msgtype50.h new file mode 100644 index 0000000..e56dc72 --- /dev/null +++ b/drivers/s390/crypto/zcrypt_msgtype50.h @@ -0,0 +1,39 @@ +/* + * zcrypt 2.1.0 + * + * Copyright IBM Corp. 2001, 2012 + * Author(s): Robert Burroughs + * Eric Rossman (edrossma@us.ibm.com) + * + * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) + * Major cleanup & driver split: Martin Schwidefsky + * MSGTYPE restruct: Holger Dengler + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2, or (at your option) + * any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#ifndef _ZCRYPT_MSGTYPE50_H_ +#define _ZCRYPT_MSGTYPE50_H_ + +#define MSGTYPE50_NAME "zcrypt_msgtype50" +#define MSGTYPE50_VARIANT_DEFAULT 0 + +#define MSGTYPE50_CRB2_MAX_MSG_SIZE 0x390 /*sizeof(struct type50_crb2_msg)*/ +#define MSGTYPE50_CRB3_MAX_MSG_SIZE 0x710 /*sizeof(struct type50_crb3_msg)*/ + +int zcrypt_msgtype50_init(void); +void zcrypt_msgtype50_exit(void); + +#endif /* _ZCRYPT_MSGTYPE50_H_ */ diff --git a/drivers/s390/crypto/zcrypt_msgtype6.c b/drivers/s390/crypto/zcrypt_msgtype6.c new file mode 100644 index 0000000..7d97fa5 --- /dev/null +++ b/drivers/s390/crypto/zcrypt_msgtype6.c @@ -0,0 +1,856 @@ +/* + * zcrypt 2.1.0 + * + * Copyright IBM Corp. 2001, 2012 + * Author(s): Robert Burroughs + * Eric Rossman (edrossma@us.ibm.com) + * + * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) + * Major cleanup & driver split: Martin Schwidefsky + * Ralph Wuerthner + * MSGTYPE restruct: Holger Dengler + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2, or (at your option) + * any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "ap_bus.h" +#include "zcrypt_api.h" +#include "zcrypt_error.h" +#include "zcrypt_msgtype6.h" +#include "zcrypt_cca_key.h" + +#define PCIXCC_MIN_MOD_SIZE_OLD 64 /* 512 bits */ +#define PCIXCC_MAX_ICA_RESPONSE_SIZE 0x77c /* max size type86 v2 reply */ + +#define CEIL4(x) ((((x)+3)/4)*4) + +struct response_type { + struct completion work; + int type; +}; +#define PCIXCC_RESPONSE_TYPE_ICA 0 +#define PCIXCC_RESPONSE_TYPE_XCRB 1 + +MODULE_AUTHOR("IBM Corporation"); +MODULE_DESCRIPTION("Cryptographic Coprocessor (message type 6), " \ + "Copyright IBM Corp. 2001, 2012"); +MODULE_LICENSE("GPL"); + +static void zcrypt_msgtype6_receive(struct ap_device *, struct ap_message *, + struct ap_message *); + +/** + * CPRB + * Note that all shorts, ints and longs are little-endian. + * All pointer fields are 32-bits long, and mean nothing + * + * A request CPRB is followed by a request_parameter_block. + * + * The request (or reply) parameter block is organized thus: + * function code + * VUD block + * key block + */ +struct CPRB { + unsigned short cprb_len; /* CPRB length */ + unsigned char cprb_ver_id; /* CPRB version id. */ + unsigned char pad_000; /* Alignment pad byte. */ + unsigned char srpi_rtcode[4]; /* SRPI return code LELONG */ + unsigned char srpi_verb; /* SRPI verb type */ + unsigned char flags; /* flags */ + unsigned char func_id[2]; /* function id */ + unsigned char checkpoint_flag; /* */ + unsigned char resv2; /* reserved */ + unsigned short req_parml; /* request parameter buffer */ + /* length 16-bit little endian */ + unsigned char req_parmp[4]; /* request parameter buffer * + * pointer (means nothing: the * + * parameter buffer follows * + * the CPRB). */ + unsigned char req_datal[4]; /* request data buffer */ + /* length ULELONG */ + unsigned char req_datap[4]; /* request data buffer */ + /* pointer */ + unsigned short rpl_parml; /* reply parameter buffer */ + /* length 16-bit little endian */ + unsigned char pad_001[2]; /* Alignment pad bytes. ULESHORT */ + unsigned char rpl_parmp[4]; /* reply parameter buffer * + * pointer (means nothing: the * + * parameter buffer follows * + * the CPRB). */ + unsigned char rpl_datal[4]; /* reply data buffer len ULELONG */ + unsigned char rpl_datap[4]; /* reply data buffer */ + /* pointer */ + unsigned short ccp_rscode; /* server reason code ULESHORT */ + unsigned short ccp_rtcode; /* server return code ULESHORT */ + unsigned char repd_parml[2]; /* replied parameter len ULESHORT*/ + unsigned char mac_data_len[2]; /* Mac Data Length ULESHORT */ + unsigned char repd_datal[4]; /* replied data length ULELONG */ + unsigned char req_pc[2]; /* PC identifier */ + unsigned char res_origin[8]; /* resource origin */ + unsigned char mac_value[8]; /* Mac Value */ + unsigned char logon_id[8]; /* Logon Identifier */ + unsigned char usage_domain[2]; /* cdx */ + unsigned char resv3[18]; /* reserved for requestor */ + unsigned short svr_namel; /* server name length ULESHORT */ + unsigned char svr_name[8]; /* server name */ +} __packed; + +struct function_and_rules_block { + unsigned char function_code[2]; + unsigned short ulen; + unsigned char only_rule[8]; +} __packed; + +/** + * The following is used to initialize the CPRBX passed to the PCIXCC/CEX2C + * card in a type6 message. The 3 fields that must be filled in at execution + * time are req_parml, rpl_parml and usage_domain. + * Everything about this interface is ascii/big-endian, since the + * device does *not* have 'Intel inside'. + * + * The CPRBX is followed immediately by the parm block. + * The parm block contains: + * - function code ('PD' 0x5044 or 'PK' 0x504B) + * - rule block (one of:) + * + 0x000A 'PKCS-1.2' (MCL2 'PD') + * + 0x000A 'ZERO-PAD' (MCL2 'PK') + * + 0x000A 'ZERO-PAD' (MCL3 'PD' or CEX2C 'PD') + * + 0x000A 'MRP ' (MCL3 'PK' or CEX2C 'PK') + * - VUD block + */ +static struct CPRBX static_cprbx = { + .cprb_len = 0x00DC, + .cprb_ver_id = 0x02, + .func_id = {0x54, 0x32}, +}; + +/** + * Convert a ICAMEX message to a type6 MEX message. + * + * @zdev: crypto device pointer + * @ap_msg: pointer to AP message + * @mex: pointer to user input data + * + * Returns 0 on success or -EFAULT. + */ +static int ICAMEX_msg_to_type6MEX_msgX(struct zcrypt_device *zdev, + struct ap_message *ap_msg, + struct ica_rsa_modexpo *mex) +{ + static struct type6_hdr static_type6_hdrX = { + .type = 0x06, + .offset1 = 0x00000058, + .agent_id = {'C', 'A',}, + .function_code = {'P', 'K'}, + }; + static struct function_and_rules_block static_pke_fnr = { + .function_code = {'P', 'K'}, + .ulen = 10, + .only_rule = {'M', 'R', 'P', ' ', ' ', ' ', ' ', ' '} + }; + static struct function_and_rules_block static_pke_fnr_MCL2 = { + .function_code = {'P', 'K'}, + .ulen = 10, + .only_rule = {'Z', 'E', 'R', 'O', '-', 'P', 'A', 'D'} + }; + struct { + struct type6_hdr hdr; + struct CPRBX cprbx; + struct function_and_rules_block fr; + unsigned short length; + char text[0]; + } __packed * msg = ap_msg->message; + int size; + + /* VUD.ciphertext */ + msg->length = mex->inputdatalength + 2; + if (copy_from_user(msg->text, mex->inputdata, mex->inputdatalength)) + return -EFAULT; + + /* Set up key which is located after the variable length text. */ + size = zcrypt_type6_mex_key_en(mex, msg->text+mex->inputdatalength, 1); + if (size < 0) + return size; + size += sizeof(*msg) + mex->inputdatalength; + + /* message header, cprbx and f&r */ + msg->hdr = static_type6_hdrX; + msg->hdr.ToCardLen1 = size - sizeof(msg->hdr); + msg->hdr.FromCardLen1 = PCIXCC_MAX_ICA_RESPONSE_SIZE - sizeof(msg->hdr); + + msg->cprbx = static_cprbx; + msg->cprbx.domain = AP_QID_QUEUE(zdev->ap_dev->qid); + msg->cprbx.rpl_msgbl = msg->hdr.FromCardLen1; + + msg->fr = (zdev->user_space_type == ZCRYPT_PCIXCC_MCL2) ? + static_pke_fnr_MCL2 : static_pke_fnr; + + msg->cprbx.req_parml = size - sizeof(msg->hdr) - sizeof(msg->cprbx); + + ap_msg->length = size; + return 0; +} + +/** + * Convert a ICACRT message to a type6 CRT message. + * + * @zdev: crypto device pointer + * @ap_msg: pointer to AP message + * @crt: pointer to user input data + * + * Returns 0 on success or -EFAULT. + */ +static int ICACRT_msg_to_type6CRT_msgX(struct zcrypt_device *zdev, + struct ap_message *ap_msg, + struct ica_rsa_modexpo_crt *crt) +{ + static struct type6_hdr static_type6_hdrX = { + .type = 0x06, + .offset1 = 0x00000058, + .agent_id = {'C', 'A',}, + .function_code = {'P', 'D'}, + }; + static struct function_and_rules_block static_pkd_fnr = { + .function_code = {'P', 'D'}, + .ulen = 10, + .only_rule = {'Z', 'E', 'R', 'O', '-', 'P', 'A', 'D'} + }; + + static struct function_and_rules_block static_pkd_fnr_MCL2 = { + .function_code = {'P', 'D'}, + .ulen = 10, + .only_rule = {'P', 'K', 'C', 'S', '-', '1', '.', '2'} + }; + struct { + struct type6_hdr hdr; + struct CPRBX cprbx; + struct function_and_rules_block fr; + unsigned short length; + char text[0]; + } __packed * msg = ap_msg->message; + int size; + + /* VUD.ciphertext */ + msg->length = crt->inputdatalength + 2; + if (copy_from_user(msg->text, crt->inputdata, crt->inputdatalength)) + return -EFAULT; + + /* Set up key which is located after the variable length text. */ + size = zcrypt_type6_crt_key(crt, msg->text + crt->inputdatalength, 1); + if (size < 0) + return size; + size += sizeof(*msg) + crt->inputdatalength; /* total size of msg */ + + /* message header, cprbx and f&r */ + msg->hdr = static_type6_hdrX; + msg->hdr.ToCardLen1 = size - sizeof(msg->hdr); + msg->hdr.FromCardLen1 = PCIXCC_MAX_ICA_RESPONSE_SIZE - sizeof(msg->hdr); + + msg->cprbx = static_cprbx; + msg->cprbx.domain = AP_QID_QUEUE(zdev->ap_dev->qid); + msg->cprbx.req_parml = msg->cprbx.rpl_msgbl = + size - sizeof(msg->hdr) - sizeof(msg->cprbx); + + msg->fr = (zdev->user_space_type == ZCRYPT_PCIXCC_MCL2) ? + static_pkd_fnr_MCL2 : static_pkd_fnr; + + ap_msg->length = size; + return 0; +} + +/** + * Convert a XCRB message to a type6 CPRB message. + * + * @zdev: crypto device pointer + * @ap_msg: pointer to AP message + * @xcRB: pointer to user input data + * + * Returns 0 on success or -EFAULT, -EINVAL. + */ +struct type86_fmt2_msg { + struct type86_hdr hdr; + struct type86_fmt2_ext fmt2; +} __packed; + +static int XCRB_msg_to_type6CPRB_msgX(struct zcrypt_device *zdev, + struct ap_message *ap_msg, + struct ica_xcRB *xcRB) +{ + static struct type6_hdr static_type6_hdrX = { + .type = 0x06, + .offset1 = 0x00000058, + }; + struct { + struct type6_hdr hdr; + struct CPRBX cprbx; + } __packed * msg = ap_msg->message; + + int rcblen = CEIL4(xcRB->request_control_blk_length); + int replylen; + char *req_data = ap_msg->message + sizeof(struct type6_hdr) + rcblen; + char *function_code; + + /* length checks */ + ap_msg->length = sizeof(struct type6_hdr) + + CEIL4(xcRB->request_control_blk_length) + + xcRB->request_data_length; + if (ap_msg->length > MSGTYPE06_MAX_MSG_SIZE) + return -EINVAL; + replylen = sizeof(struct type86_fmt2_msg) + + CEIL4(xcRB->reply_control_blk_length) + + xcRB->reply_data_length; + if (replylen > MSGTYPE06_MAX_MSG_SIZE) + return -EINVAL; + + /* prepare type6 header */ + msg->hdr = static_type6_hdrX; + memcpy(msg->hdr.agent_id , &(xcRB->agent_ID), sizeof(xcRB->agent_ID)); + msg->hdr.ToCardLen1 = xcRB->request_control_blk_length; + if (xcRB->request_data_length) { + msg->hdr.offset2 = msg->hdr.offset1 + rcblen; + msg->hdr.ToCardLen2 = xcRB->request_data_length; + } + msg->hdr.FromCardLen1 = xcRB->reply_control_blk_length; + msg->hdr.FromCardLen2 = xcRB->reply_data_length; + + /* prepare CPRB */ + if (copy_from_user(&(msg->cprbx), xcRB->request_control_blk_addr, + xcRB->request_control_blk_length)) + return -EFAULT; + if (msg->cprbx.cprb_len + sizeof(msg->hdr.function_code) > + xcRB->request_control_blk_length) + return -EINVAL; + function_code = ((unsigned char *)&msg->cprbx) + msg->cprbx.cprb_len; + memcpy(msg->hdr.function_code, function_code, + sizeof(msg->hdr.function_code)); + + if (memcmp(function_code, "US", 2) == 0) + ap_msg->special = 1; + else + ap_msg->special = 0; + + /* copy data block */ + if (xcRB->request_data_length && + copy_from_user(req_data, xcRB->request_data_address, + xcRB->request_data_length)) + return -EFAULT; + return 0; +} + +/** + * Copy results from a type 86 ICA reply message back to user space. + * + * @zdev: crypto device pointer + * @reply: reply AP message. + * @data: pointer to user output data + * @length: size of user output data + * + * Returns 0 on success or -EINVAL, -EFAULT, -EAGAIN in case of an error. + */ +struct type86x_reply { + struct type86_hdr hdr; + struct type86_fmt2_ext fmt2; + struct CPRBX cprbx; + unsigned char pad[4]; /* 4 byte function code/rules block ? */ + unsigned short length; + char text[0]; +} __packed; + +static int convert_type86_ica(struct zcrypt_device *zdev, + struct ap_message *reply, + char __user *outputdata, + unsigned int outputdatalength) +{ + static unsigned char static_pad[] = { + 0x00, 0x02, + 0x1B, 0x7B, 0x5D, 0xB5, 0x75, 0x01, 0x3D, 0xFD, + 0x8D, 0xD1, 0xC7, 0x03, 0x2D, 0x09, 0x23, 0x57, + 0x89, 0x49, 0xB9, 0x3F, 0xBB, 0x99, 0x41, 0x5B, + 0x75, 0x21, 0x7B, 0x9D, 0x3B, 0x6B, 0x51, 0x39, + 0xBB, 0x0D, 0x35, 0xB9, 0x89, 0x0F, 0x93, 0xA5, + 0x0B, 0x47, 0xF1, 0xD3, 0xBB, 0xCB, 0xF1, 0x9D, + 0x23, 0x73, 0x71, 0xFF, 0xF3, 0xF5, 0x45, 0xFB, + 0x61, 0x29, 0x23, 0xFD, 0xF1, 0x29, 0x3F, 0x7F, + 0x17, 0xB7, 0x1B, 0xA9, 0x19, 0xBD, 0x57, 0xA9, + 0xD7, 0x95, 0xA3, 0xCB, 0xED, 0x1D, 0xDB, 0x45, + 0x7D, 0x11, 0xD1, 0x51, 0x1B, 0xED, 0x71, 0xE9, + 0xB1, 0xD1, 0xAB, 0xAB, 0x21, 0x2B, 0x1B, 0x9F, + 0x3B, 0x9F, 0xF7, 0xF7, 0xBD, 0x63, 0xEB, 0xAD, + 0xDF, 0xB3, 0x6F, 0x5B, 0xDB, 0x8D, 0xA9, 0x5D, + 0xE3, 0x7D, 0x77, 0x49, 0x47, 0xF5, 0xA7, 0xFD, + 0xAB, 0x2F, 0x27, 0x35, 0x77, 0xD3, 0x49, 0xC9, + 0x09, 0xEB, 0xB1, 0xF9, 0xBF, 0x4B, 0xCB, 0x2B, + 0xEB, 0xEB, 0x05, 0xFF, 0x7D, 0xC7, 0x91, 0x8B, + 0x09, 0x83, 0xB9, 0xB9, 0x69, 0x33, 0x39, 0x6B, + 0x79, 0x75, 0x19, 0xBF, 0xBB, 0x07, 0x1D, 0xBD, + 0x29, 0xBF, 0x39, 0x95, 0x93, 0x1D, 0x35, 0xC7, + 0xC9, 0x4D, 0xE5, 0x97, 0x0B, 0x43, 0x9B, 0xF1, + 0x16, 0x93, 0x03, 0x1F, 0xA5, 0xFB, 0xDB, 0xF3, + 0x27, 0x4F, 0x27, 0x61, 0x05, 0x1F, 0xB9, 0x23, + 0x2F, 0xC3, 0x81, 0xA9, 0x23, 0x71, 0x55, 0x55, + 0xEB, 0xED, 0x41, 0xE5, 0xF3, 0x11, 0xF1, 0x43, + 0x69, 0x03, 0xBD, 0x0B, 0x37, 0x0F, 0x51, 0x8F, + 0x0B, 0xB5, 0x89, 0x5B, 0x67, 0xA9, 0xD9, 0x4F, + 0x01, 0xF9, 0x21, 0x77, 0x37, 0x73, 0x79, 0xC5, + 0x7F, 0x51, 0xC1, 0xCF, 0x97, 0xA1, 0x75, 0xAD, + 0x35, 0x9D, 0xD3, 0xD3, 0xA7, 0x9D, 0x5D, 0x41, + 0x6F, 0x65, 0x1B, 0xCF, 0xA9, 0x87, 0x91, 0x09 + }; + struct type86x_reply *msg = reply->message; + unsigned short service_rc, service_rs; + unsigned int reply_len, pad_len; + char *data; + + service_rc = msg->cprbx.ccp_rtcode; + if (unlikely(service_rc != 0)) { + service_rs = msg->cprbx.ccp_rscode; + if (service_rc == 8 && service_rs == 66) + return -EINVAL; + if (service_rc == 8 && service_rs == 65) + return -EINVAL; + if (service_rc == 8 && service_rs == 770) + return -EINVAL; + if (service_rc == 8 && service_rs == 783) { + zdev->min_mod_size = PCIXCC_MIN_MOD_SIZE_OLD; + return -EAGAIN; + } + if (service_rc == 12 && service_rs == 769) + return -EINVAL; + if (service_rc == 8 && service_rs == 72) + return -EINVAL; + zdev->online = 0; + return -EAGAIN; /* repeat the request on a different device. */ + } + data = msg->text; + reply_len = msg->length - 2; + if (reply_len > outputdatalength) + return -EINVAL; + /* + * For all encipher requests, the length of the ciphertext (reply_len) + * will always equal the modulus length. For MEX decipher requests + * the output needs to get padded. Minimum pad size is 10. + * + * Currently, the cases where padding will be added is for: + * - PCIXCC_MCL2 using a CRT form token (since PKD didn't support + * ZERO-PAD and CRT is only supported for PKD requests) + * - PCICC, always + */ + pad_len = outputdatalength - reply_len; + if (pad_len > 0) { + if (pad_len < 10) + return -EINVAL; + /* 'restore' padding left in the PCICC/PCIXCC card. */ + if (copy_to_user(outputdata, static_pad, pad_len - 1)) + return -EFAULT; + if (put_user(0, outputdata + pad_len - 1)) + return -EFAULT; + } + /* Copy the crypto response to user space. */ + if (copy_to_user(outputdata + pad_len, data, reply_len)) + return -EFAULT; + return 0; +} + +/** + * Copy results from a type 86 XCRB reply message back to user space. + * + * @zdev: crypto device pointer + * @reply: reply AP message. + * @xcRB: pointer to XCRB + * + * Returns 0 on success or -EINVAL, -EFAULT, -EAGAIN in case of an error. + */ +static int convert_type86_xcrb(struct zcrypt_device *zdev, + struct ap_message *reply, + struct ica_xcRB *xcRB) +{ + struct type86_fmt2_msg *msg = reply->message; + char *data = reply->message; + + /* Copy CPRB to user */ + if (copy_to_user(xcRB->reply_control_blk_addr, + data + msg->fmt2.offset1, msg->fmt2.count1)) + return -EFAULT; + xcRB->reply_control_blk_length = msg->fmt2.count1; + + /* Copy data buffer to user */ + if (msg->fmt2.count2) + if (copy_to_user(xcRB->reply_data_addr, + data + msg->fmt2.offset2, msg->fmt2.count2)) + return -EFAULT; + xcRB->reply_data_length = msg->fmt2.count2; + return 0; +} + +static int convert_type86_rng(struct zcrypt_device *zdev, + struct ap_message *reply, + char *buffer) +{ + struct { + struct type86_hdr hdr; + struct type86_fmt2_ext fmt2; + struct CPRBX cprbx; + } __packed * msg = reply->message; + char *data = reply->message; + + if (msg->cprbx.ccp_rtcode != 0 || msg->cprbx.ccp_rscode != 0) + return -EINVAL; + memcpy(buffer, data + msg->fmt2.offset2, msg->fmt2.count2); + return msg->fmt2.count2; +} + +static int convert_response_ica(struct zcrypt_device *zdev, + struct ap_message *reply, + char __user *outputdata, + unsigned int outputdatalength) +{ + struct type86x_reply *msg = reply->message; + + /* Response type byte is the second byte in the response. */ + switch (((unsigned char *) reply->message)[1]) { + case TYPE82_RSP_CODE: + case TYPE88_RSP_CODE: + return convert_error(zdev, reply); + case TYPE86_RSP_CODE: + if (msg->cprbx.ccp_rtcode && + (msg->cprbx.ccp_rscode == 0x14f) && + (outputdatalength > 256)) { + if (zdev->max_exp_bit_length <= 17) { + zdev->max_exp_bit_length = 17; + return -EAGAIN; + } else + return -EINVAL; + } + if (msg->hdr.reply_code) + return convert_error(zdev, reply); + if (msg->cprbx.cprb_ver_id == 0x02) + return convert_type86_ica(zdev, reply, + outputdata, outputdatalength); + /* Fall through, no break, incorrect cprb version is an unknown + * response */ + default: /* Unknown response type, this should NEVER EVER happen */ + zdev->online = 0; + return -EAGAIN; /* repeat the request on a different device. */ + } +} + +static int convert_response_xcrb(struct zcrypt_device *zdev, + struct ap_message *reply, + struct ica_xcRB *xcRB) +{ + struct type86x_reply *msg = reply->message; + + /* Response type byte is the second byte in the response. */ + switch (((unsigned char *) reply->message)[1]) { + case TYPE82_RSP_CODE: + case TYPE88_RSP_CODE: + xcRB->status = 0x0008044DL; /* HDD_InvalidParm */ + return convert_error(zdev, reply); + case TYPE86_RSP_CODE: + if (msg->hdr.reply_code) { + memcpy(&(xcRB->status), msg->fmt2.apfs, sizeof(u32)); + return convert_error(zdev, reply); + } + if (msg->cprbx.cprb_ver_id == 0x02) + return convert_type86_xcrb(zdev, reply, xcRB); + /* Fall through, no break, incorrect cprb version is an unknown + * response */ + default: /* Unknown response type, this should NEVER EVER happen */ + xcRB->status = 0x0008044DL; /* HDD_InvalidParm */ + zdev->online = 0; + return -EAGAIN; /* repeat the request on a different device. */ + } +} + +static int convert_response_rng(struct zcrypt_device *zdev, + struct ap_message *reply, + char *data) +{ + struct type86x_reply *msg = reply->message; + + switch (msg->hdr.type) { + case TYPE82_RSP_CODE: + case TYPE88_RSP_CODE: + return -EINVAL; + case TYPE86_RSP_CODE: + if (msg->hdr.reply_code) + return -EINVAL; + if (msg->cprbx.cprb_ver_id == 0x02) + return convert_type86_rng(zdev, reply, data); + /* Fall through, no break, incorrect cprb version is an unknown + * response */ + default: /* Unknown response type, this should NEVER EVER happen */ + zdev->online = 0; + return -EAGAIN; /* repeat the request on a different device. */ + } +} + +/** + * This function is called from the AP bus code after a crypto request + * "msg" has finished with the reply message "reply". + * It is called from tasklet context. + * @ap_dev: pointer to the AP device + * @msg: pointer to the AP message + * @reply: pointer to the AP reply message + */ +static void zcrypt_msgtype6_receive(struct ap_device *ap_dev, + struct ap_message *msg, + struct ap_message *reply) +{ + static struct error_hdr error_reply = { + .type = TYPE82_RSP_CODE, + .reply_code = REP82_ERROR_MACHINE_FAILURE, + }; + struct response_type *resp_type = + (struct response_type *) msg->private; + struct type86x_reply *t86r; + int length; + + /* Copy the reply message to the request message buffer. */ + if (IS_ERR(reply)) { + memcpy(msg->message, &error_reply, sizeof(error_reply)); + goto out; + } + t86r = reply->message; + if (t86r->hdr.type == TYPE86_RSP_CODE && + t86r->cprbx.cprb_ver_id == 0x02) { + switch (resp_type->type) { + case PCIXCC_RESPONSE_TYPE_ICA: + length = sizeof(struct type86x_reply) + + t86r->length - 2; + length = min(PCIXCC_MAX_ICA_RESPONSE_SIZE, length); + memcpy(msg->message, reply->message, length); + break; + case PCIXCC_RESPONSE_TYPE_XCRB: + length = t86r->fmt2.offset2 + t86r->fmt2.count2; + length = min(MSGTYPE06_MAX_MSG_SIZE, length); + memcpy(msg->message, reply->message, length); + break; + default: + memcpy(msg->message, &error_reply, + sizeof(error_reply)); + } + } else + memcpy(msg->message, reply->message, sizeof(error_reply)); +out: + complete(&(resp_type->work)); +} + +static atomic_t zcrypt_step = ATOMIC_INIT(0); + +/** + * The request distributor calls this function if it picked the PCIXCC/CEX2C + * device to handle a modexpo request. + * @zdev: pointer to zcrypt_device structure that identifies the + * PCIXCC/CEX2C device to the request distributor + * @mex: pointer to the modexpo request buffer + */ +static long zcrypt_msgtype6_modexpo(struct zcrypt_device *zdev, + struct ica_rsa_modexpo *mex) +{ + struct ap_message ap_msg; + struct response_type resp_type = { + .type = PCIXCC_RESPONSE_TYPE_ICA, + }; + int rc; + + ap_init_message(&ap_msg); + ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL); + if (!ap_msg.message) + return -ENOMEM; + ap_msg.receive = zcrypt_msgtype6_receive; + ap_msg.psmid = (((unsigned long long) current->pid) << 32) + + atomic_inc_return(&zcrypt_step); + ap_msg.private = &resp_type; + rc = ICAMEX_msg_to_type6MEX_msgX(zdev, &ap_msg, mex); + if (rc) + goto out_free; + init_completion(&resp_type.work); + ap_queue_message(zdev->ap_dev, &ap_msg); + rc = wait_for_completion_interruptible(&resp_type.work); + if (rc == 0) + rc = convert_response_ica(zdev, &ap_msg, mex->outputdata, + mex->outputdatalength); + else + /* Signal pending. */ + ap_cancel_message(zdev->ap_dev, &ap_msg); +out_free: + free_page((unsigned long) ap_msg.message); + return rc; +} + +/** + * The request distributor calls this function if it picked the PCIXCC/CEX2C + * device to handle a modexpo_crt request. + * @zdev: pointer to zcrypt_device structure that identifies the + * PCIXCC/CEX2C device to the request distributor + * @crt: pointer to the modexpoc_crt request buffer + */ +static long zcrypt_msgtype6_modexpo_crt(struct zcrypt_device *zdev, + struct ica_rsa_modexpo_crt *crt) +{ + struct ap_message ap_msg; + struct response_type resp_type = { + .type = PCIXCC_RESPONSE_TYPE_ICA, + }; + int rc; + + ap_init_message(&ap_msg); + ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL); + if (!ap_msg.message) + return -ENOMEM; + ap_msg.receive = zcrypt_msgtype6_receive; + ap_msg.psmid = (((unsigned long long) current->pid) << 32) + + atomic_inc_return(&zcrypt_step); + ap_msg.private = &resp_type; + rc = ICACRT_msg_to_type6CRT_msgX(zdev, &ap_msg, crt); + if (rc) + goto out_free; + init_completion(&resp_type.work); + ap_queue_message(zdev->ap_dev, &ap_msg); + rc = wait_for_completion_interruptible(&resp_type.work); + if (rc == 0) + rc = convert_response_ica(zdev, &ap_msg, crt->outputdata, + crt->outputdatalength); + else + /* Signal pending. */ + ap_cancel_message(zdev->ap_dev, &ap_msg); +out_free: + free_page((unsigned long) ap_msg.message); + return rc; +} + +/** + * The request distributor calls this function if it picked the PCIXCC/CEX2C + * device to handle a send_cprb request. + * @zdev: pointer to zcrypt_device structure that identifies the + * PCIXCC/CEX2C device to the request distributor + * @xcRB: pointer to the send_cprb request buffer + */ +static long zcrypt_msgtype6_send_cprb(struct zcrypt_device *zdev, + struct ica_xcRB *xcRB) +{ + struct ap_message ap_msg; + struct response_type resp_type = { + .type = PCIXCC_RESPONSE_TYPE_XCRB, + }; + int rc; + + ap_init_message(&ap_msg); + ap_msg.message = kmalloc(MSGTYPE06_MAX_MSG_SIZE, GFP_KERNEL); + if (!ap_msg.message) + return -ENOMEM; + ap_msg.receive = zcrypt_msgtype6_receive; + ap_msg.psmid = (((unsigned long long) current->pid) << 32) + + atomic_inc_return(&zcrypt_step); + ap_msg.private = &resp_type; + rc = XCRB_msg_to_type6CPRB_msgX(zdev, &ap_msg, xcRB); + if (rc) + goto out_free; + init_completion(&resp_type.work); + ap_queue_message(zdev->ap_dev, &ap_msg); + rc = wait_for_completion_interruptible(&resp_type.work); + if (rc == 0) + rc = convert_response_xcrb(zdev, &ap_msg, xcRB); + else + /* Signal pending. */ + ap_cancel_message(zdev->ap_dev, &ap_msg); +out_free: + kzfree(ap_msg.message); + return rc; +} + +/** + * The request distributor calls this function if it picked the PCIXCC/CEX2C + * device to generate random data. + * @zdev: pointer to zcrypt_device structure that identifies the + * PCIXCC/CEX2C device to the request distributor + * @buffer: pointer to a memory page to return random data + */ + +static long zcrypt_msgtype6_rng(struct zcrypt_device *zdev, + char *buffer) +{ + struct ap_message ap_msg; + struct response_type resp_type = { + .type = PCIXCC_RESPONSE_TYPE_XCRB, + }; + int rc; + + ap_init_message(&ap_msg); + ap_msg.message = kmalloc(MSGTYPE06_MAX_MSG_SIZE, GFP_KERNEL); + if (!ap_msg.message) + return -ENOMEM; + ap_msg.receive = zcrypt_msgtype6_receive; + ap_msg.psmid = (((unsigned long long) current->pid) << 32) + + atomic_inc_return(&zcrypt_step); + ap_msg.private = &resp_type; + rng_type6CPRB_msgX(zdev->ap_dev, &ap_msg, ZCRYPT_RNG_BUFFER_SIZE); + init_completion(&resp_type.work); + ap_queue_message(zdev->ap_dev, &ap_msg); + rc = wait_for_completion_interruptible(&resp_type.work); + if (rc == 0) + rc = convert_response_rng(zdev, &ap_msg, buffer); + else + /* Signal pending. */ + ap_cancel_message(zdev->ap_dev, &ap_msg); + kfree(ap_msg.message); + return rc; +} + +/** + * The crypto operations for a PCIXCC/CEX2C card. + */ +static struct zcrypt_ops zcrypt_msgtype6_norng_ops = { + .owner = THIS_MODULE, + .variant = MSGTYPE06_VARIANT_NORNG, + .rsa_modexpo = zcrypt_msgtype6_modexpo, + .rsa_modexpo_crt = zcrypt_msgtype6_modexpo_crt, + .send_cprb = zcrypt_msgtype6_send_cprb, +}; + +static struct zcrypt_ops zcrypt_msgtype6_ops = { + .owner = THIS_MODULE, + .variant = MSGTYPE06_VARIANT_DEFAULT, + .rsa_modexpo = zcrypt_msgtype6_modexpo, + .rsa_modexpo_crt = zcrypt_msgtype6_modexpo_crt, + .send_cprb = zcrypt_msgtype6_send_cprb, + .rng = zcrypt_msgtype6_rng, +}; + +int __init zcrypt_msgtype6_init(void) +{ + zcrypt_msgtype_register(&zcrypt_msgtype6_norng_ops); + zcrypt_msgtype_register(&zcrypt_msgtype6_ops); + return 0; +} + +void __exit zcrypt_msgtype6_exit(void) +{ + zcrypt_msgtype_unregister(&zcrypt_msgtype6_norng_ops); + zcrypt_msgtype_unregister(&zcrypt_msgtype6_ops); +} + +module_init(zcrypt_msgtype6_init); +module_exit(zcrypt_msgtype6_exit); diff --git a/drivers/s390/crypto/zcrypt_msgtype6.h b/drivers/s390/crypto/zcrypt_msgtype6.h new file mode 100644 index 0000000..1e500d3 --- /dev/null +++ b/drivers/s390/crypto/zcrypt_msgtype6.h @@ -0,0 +1,169 @@ +/* + * zcrypt 2.1.0 + * + * Copyright IBM Corp. 2001, 2012 + * Author(s): Robert Burroughs + * Eric Rossman (edrossma@us.ibm.com) + * + * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) + * Major cleanup & driver split: Martin Schwidefsky + * MSGTYPE restruct: Holger Dengler + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2, or (at your option) + * any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#ifndef _ZCRYPT_MSGTYPE6_H_ +#define _ZCRYPT_MSGTYPE6_H_ + +#include + +#define MSGTYPE06_NAME "zcrypt_msgtype6" +#define MSGTYPE06_VARIANT_DEFAULT 0 +#define MSGTYPE06_VARIANT_NORNG 1 + +#define MSGTYPE06_MAX_MSG_SIZE (12*1024) + +/** + * The type 6 message family is associated with PCICC or PCIXCC cards. + * + * It contains a message header followed by a CPRB, both of which + * are described below. + * + * Note that all reserved fields must be zeroes. + */ +struct type6_hdr { + unsigned char reserved1; /* 0x00 */ + unsigned char type; /* 0x06 */ + unsigned char reserved2[2]; /* 0x0000 */ + unsigned char right[4]; /* 0x00000000 */ + unsigned char reserved3[2]; /* 0x0000 */ + unsigned char reserved4[2]; /* 0x0000 */ + unsigned char apfs[4]; /* 0x00000000 */ + unsigned int offset1; /* 0x00000058 (offset to CPRB) */ + unsigned int offset2; /* 0x00000000 */ + unsigned int offset3; /* 0x00000000 */ + unsigned int offset4; /* 0x00000000 */ + unsigned char agent_id[16]; /* PCICC: */ + /* 0x0100 */ + /* 0x4343412d4150504c202020 */ + /* 0x010101 */ + /* PCIXCC: */ + /* 0x4341000000000000 */ + /* 0x0000000000000000 */ + unsigned char rqid[2]; /* rqid. internal to 603 */ + unsigned char reserved5[2]; /* 0x0000 */ + unsigned char function_code[2]; /* for PKD, 0x5044 (ascii 'PD') */ + unsigned char reserved6[2]; /* 0x0000 */ + unsigned int ToCardLen1; /* (request CPRB len + 3) & -4 */ + unsigned int ToCardLen2; /* db len 0x00000000 for PKD */ + unsigned int ToCardLen3; /* 0x00000000 */ + unsigned int ToCardLen4; /* 0x00000000 */ + unsigned int FromCardLen1; /* response buffer length */ + unsigned int FromCardLen2; /* db len 0x00000000 for PKD */ + unsigned int FromCardLen3; /* 0x00000000 */ + unsigned int FromCardLen4; /* 0x00000000 */ +} __packed; + +/** + * The type 86 message family is associated with PCICC and PCIXCC cards. + * + * It contains a message header followed by a CPRB. The CPRB is + * the same as the request CPRB, which is described above. + * + * If format is 1, an error condition exists and no data beyond + * the 8-byte message header is of interest. + * + * The non-error message is shown below. + * + * Note that all reserved fields must be zeroes. + */ +struct type86_hdr { + unsigned char reserved1; /* 0x00 */ + unsigned char type; /* 0x86 */ + unsigned char format; /* 0x01 (error) or 0x02 (ok) */ + unsigned char reserved2; /* 0x00 */ + unsigned char reply_code; /* reply code (see above) */ + unsigned char reserved3[3]; /* 0x000000 */ +} __packed; + +#define TYPE86_RSP_CODE 0x86 +#define TYPE86_FMT2 0x02 + +struct type86_fmt2_ext { + unsigned char reserved[4]; /* 0x00000000 */ + unsigned char apfs[4]; /* final status */ + unsigned int count1; /* length of CPRB + parameters */ + unsigned int offset1; /* offset to CPRB */ + unsigned int count2; /* 0x00000000 */ + unsigned int offset2; /* db offset 0x00000000 for PKD */ + unsigned int count3; /* 0x00000000 */ + unsigned int offset3; /* 0x00000000 */ + unsigned int count4; /* 0x00000000 */ + unsigned int offset4; /* 0x00000000 */ +} __packed; + +/** + * Prepare a type6 CPRB message for random number generation + * + * @ap_dev: AP device pointer + * @ap_msg: pointer to AP message + */ +static inline void rng_type6CPRB_msgX(struct ap_device *ap_dev, + struct ap_message *ap_msg, + unsigned random_number_length) +{ + struct { + struct type6_hdr hdr; + struct CPRBX cprbx; + char function_code[2]; + short int rule_length; + char rule[8]; + short int verb_length; + short int key_length; + } __packed * msg = ap_msg->message; + static struct type6_hdr static_type6_hdrX = { + .type = 0x06, + .offset1 = 0x00000058, + .agent_id = {'C', 'A'}, + .function_code = {'R', 'L'}, + .ToCardLen1 = sizeof(*msg) - sizeof(msg->hdr), + .FromCardLen1 = sizeof(*msg) - sizeof(msg->hdr), + }; + static struct CPRBX local_cprbx = { + .cprb_len = 0x00dc, + .cprb_ver_id = 0x02, + .func_id = {0x54, 0x32}, + .req_parml = sizeof(*msg) - sizeof(msg->hdr) - + sizeof(msg->cprbx), + .rpl_msgbl = sizeof(*msg) - sizeof(msg->hdr), + }; + + msg->hdr = static_type6_hdrX; + msg->hdr.FromCardLen2 = random_number_length, + msg->cprbx = local_cprbx; + msg->cprbx.rpl_datal = random_number_length, + msg->cprbx.domain = AP_QID_QUEUE(ap_dev->qid); + memcpy(msg->function_code, msg->hdr.function_code, 0x02); + msg->rule_length = 0x0a; + memcpy(msg->rule, "RANDOM ", 8); + msg->verb_length = 0x02; + msg->key_length = 0x02; + ap_msg->length = sizeof(*msg); +} + +int zcrypt_msgtype6_init(void); +void zcrypt_msgtype6_exit(void); + +#endif /* _ZCRYPT_MSGTYPE6_H_ */ diff --git a/drivers/s390/crypto/zcrypt_pcixcc.c b/drivers/s390/crypto/zcrypt_pcixcc.c index ccb4f8b6..c7275e3 100644 --- a/drivers/s390/crypto/zcrypt_pcixcc.c +++ b/drivers/s390/crypto/zcrypt_pcixcc.c @@ -1,13 +1,14 @@ /* * zcrypt 2.1.0 * - * Copyright IBM Corp. 2001, 2006 + * Copyright IBM Corp. 2001, 2012 * Author(s): Robert Burroughs * Eric Rossman (edrossma@us.ibm.com) * * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) * Major cleanup & driver split: Martin Schwidefsky * Ralph Wuerthner + * MSGTYPE restruct: Holger Dengler * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -35,9 +36,10 @@ #include "ap_bus.h" #include "zcrypt_api.h" #include "zcrypt_error.h" -#include "zcrypt_pcicc.h" +#include "zcrypt_msgtype6.h" #include "zcrypt_pcixcc.h" #include "zcrypt_cca_key.h" +#include "zcrypt_msgtype6.h" #define PCIXCC_MIN_MOD_SIZE 16 /* 128 bits */ #define PCIXCC_MIN_MOD_SIZE_OLD 64 /* 512 bits */ @@ -75,14 +77,12 @@ static struct ap_device_id zcrypt_pcixcc_ids[] = { MODULE_DEVICE_TABLE(ap, zcrypt_pcixcc_ids); MODULE_AUTHOR("IBM Corporation"); -MODULE_DESCRIPTION("PCIXCC Cryptographic Coprocessor device driver, " - "Copyright IBM Corp. 2001, 2006"); +MODULE_DESCRIPTION("PCIXCC Cryptographic Coprocessor device driver, " \ + "Copyright IBM Corp. 2001, 2012"); MODULE_LICENSE("GPL"); static int zcrypt_pcixcc_probe(struct ap_device *ap_dev); static void zcrypt_pcixcc_remove(struct ap_device *ap_dev); -static void zcrypt_pcixcc_receive(struct ap_device *, struct ap_message *, - struct ap_message *); static struct ap_driver zcrypt_pcixcc_driver = { .probe = zcrypt_pcixcc_probe, @@ -92,766 +92,6 @@ static struct ap_driver zcrypt_pcixcc_driver = { }; /** - * The following is used to initialize the CPRBX passed to the PCIXCC/CEX2C - * card in a type6 message. The 3 fields that must be filled in at execution - * time are req_parml, rpl_parml and usage_domain. - * Everything about this interface is ascii/big-endian, since the - * device does *not* have 'Intel inside'. - * - * The CPRBX is followed immediately by the parm block. - * The parm block contains: - * - function code ('PD' 0x5044 or 'PK' 0x504B) - * - rule block (one of:) - * + 0x000A 'PKCS-1.2' (MCL2 'PD') - * + 0x000A 'ZERO-PAD' (MCL2 'PK') - * + 0x000A 'ZERO-PAD' (MCL3 'PD' or CEX2C 'PD') - * + 0x000A 'MRP ' (MCL3 'PK' or CEX2C 'PK') - * - VUD block - */ -static struct CPRBX static_cprbx = { - .cprb_len = 0x00DC, - .cprb_ver_id = 0x02, - .func_id = {0x54,0x32}, -}; - -/** - * Convert a ICAMEX message to a type6 MEX message. - * - * @zdev: crypto device pointer - * @ap_msg: pointer to AP message - * @mex: pointer to user input data - * - * Returns 0 on success or -EFAULT. - */ -static int ICAMEX_msg_to_type6MEX_msgX(struct zcrypt_device *zdev, - struct ap_message *ap_msg, - struct ica_rsa_modexpo *mex) -{ - static struct type6_hdr static_type6_hdrX = { - .type = 0x06, - .offset1 = 0x00000058, - .agent_id = {'C','A',}, - .function_code = {'P','K'}, - }; - static struct function_and_rules_block static_pke_fnr = { - .function_code = {'P','K'}, - .ulen = 10, - .only_rule = {'M','R','P',' ',' ',' ',' ',' '} - }; - static struct function_and_rules_block static_pke_fnr_MCL2 = { - .function_code = {'P','K'}, - .ulen = 10, - .only_rule = {'Z','E','R','O','-','P','A','D'} - }; - struct { - struct type6_hdr hdr; - struct CPRBX cprbx; - struct function_and_rules_block fr; - unsigned short length; - char text[0]; - } __attribute__((packed)) *msg = ap_msg->message; - int size; - - /* VUD.ciphertext */ - msg->length = mex->inputdatalength + 2; - if (copy_from_user(msg->text, mex->inputdata, mex->inputdatalength)) - return -EFAULT; - - /* Set up key which is located after the variable length text. */ - size = zcrypt_type6_mex_key_en(mex, msg->text+mex->inputdatalength, 1); - if (size < 0) - return size; - size += sizeof(*msg) + mex->inputdatalength; - - /* message header, cprbx and f&r */ - msg->hdr = static_type6_hdrX; - msg->hdr.ToCardLen1 = size - sizeof(msg->hdr); - msg->hdr.FromCardLen1 = PCIXCC_MAX_ICA_RESPONSE_SIZE - sizeof(msg->hdr); - - msg->cprbx = static_cprbx; - msg->cprbx.domain = AP_QID_QUEUE(zdev->ap_dev->qid); - msg->cprbx.rpl_msgbl = msg->hdr.FromCardLen1; - - msg->fr = (zdev->user_space_type == ZCRYPT_PCIXCC_MCL2) ? - static_pke_fnr_MCL2 : static_pke_fnr; - - msg->cprbx.req_parml = size - sizeof(msg->hdr) - sizeof(msg->cprbx); - - ap_msg->length = size; - return 0; -} - -/** - * Convert a ICACRT message to a type6 CRT message. - * - * @zdev: crypto device pointer - * @ap_msg: pointer to AP message - * @crt: pointer to user input data - * - * Returns 0 on success or -EFAULT. - */ -static int ICACRT_msg_to_type6CRT_msgX(struct zcrypt_device *zdev, - struct ap_message *ap_msg, - struct ica_rsa_modexpo_crt *crt) -{ - static struct type6_hdr static_type6_hdrX = { - .type = 0x06, - .offset1 = 0x00000058, - .agent_id = {'C','A',}, - .function_code = {'P','D'}, - }; - static struct function_and_rules_block static_pkd_fnr = { - .function_code = {'P','D'}, - .ulen = 10, - .only_rule = {'Z','E','R','O','-','P','A','D'} - }; - - static struct function_and_rules_block static_pkd_fnr_MCL2 = { - .function_code = {'P','D'}, - .ulen = 10, - .only_rule = {'P','K','C','S','-','1','.','2'} - }; - struct { - struct type6_hdr hdr; - struct CPRBX cprbx; - struct function_and_rules_block fr; - unsigned short length; - char text[0]; - } __attribute__((packed)) *msg = ap_msg->message; - int size; - - /* VUD.ciphertext */ - msg->length = crt->inputdatalength + 2; - if (copy_from_user(msg->text, crt->inputdata, crt->inputdatalength)) - return -EFAULT; - - /* Set up key which is located after the variable length text. */ - size = zcrypt_type6_crt_key(crt, msg->text + crt->inputdatalength, 1); - if (size < 0) - return size; - size += sizeof(*msg) + crt->inputdatalength; /* total size of msg */ - - /* message header, cprbx and f&r */ - msg->hdr = static_type6_hdrX; - msg->hdr.ToCardLen1 = size - sizeof(msg->hdr); - msg->hdr.FromCardLen1 = PCIXCC_MAX_ICA_RESPONSE_SIZE - sizeof(msg->hdr); - - msg->cprbx = static_cprbx; - msg->cprbx.domain = AP_QID_QUEUE(zdev->ap_dev->qid); - msg->cprbx.req_parml = msg->cprbx.rpl_msgbl = - size - sizeof(msg->hdr) - sizeof(msg->cprbx); - - msg->fr = (zdev->user_space_type == ZCRYPT_PCIXCC_MCL2) ? - static_pkd_fnr_MCL2 : static_pkd_fnr; - - ap_msg->length = size; - return 0; -} - -/** - * Convert a XCRB message to a type6 CPRB message. - * - * @zdev: crypto device pointer - * @ap_msg: pointer to AP message - * @xcRB: pointer to user input data - * - * Returns 0 on success or -EFAULT, -EINVAL. - */ -struct type86_fmt2_msg { - struct type86_hdr hdr; - struct type86_fmt2_ext fmt2; -} __attribute__((packed)); - -static int XCRB_msg_to_type6CPRB_msgX(struct zcrypt_device *zdev, - struct ap_message *ap_msg, - struct ica_xcRB *xcRB) -{ - static struct type6_hdr static_type6_hdrX = { - .type = 0x06, - .offset1 = 0x00000058, - }; - struct { - struct type6_hdr hdr; - struct CPRBX cprbx; - } __attribute__((packed)) *msg = ap_msg->message; - - int rcblen = CEIL4(xcRB->request_control_blk_length); - int replylen; - char *req_data = ap_msg->message + sizeof(struct type6_hdr) + rcblen; - char *function_code; - - /* length checks */ - ap_msg->length = sizeof(struct type6_hdr) + - CEIL4(xcRB->request_control_blk_length) + - xcRB->request_data_length; - if (ap_msg->length > PCIXCC_MAX_XCRB_MESSAGE_SIZE) - return -EINVAL; - replylen = sizeof(struct type86_fmt2_msg) + - CEIL4(xcRB->reply_control_blk_length) + - xcRB->reply_data_length; - if (replylen > PCIXCC_MAX_XCRB_MESSAGE_SIZE) - return -EINVAL; - - /* prepare type6 header */ - msg->hdr = static_type6_hdrX; - memcpy(msg->hdr.agent_id , &(xcRB->agent_ID), sizeof(xcRB->agent_ID)); - msg->hdr.ToCardLen1 = xcRB->request_control_blk_length; - if (xcRB->request_data_length) { - msg->hdr.offset2 = msg->hdr.offset1 + rcblen; - msg->hdr.ToCardLen2 = xcRB->request_data_length; - } - msg->hdr.FromCardLen1 = xcRB->reply_control_blk_length; - msg->hdr.FromCardLen2 = xcRB->reply_data_length; - - /* prepare CPRB */ - if (copy_from_user(&(msg->cprbx), xcRB->request_control_blk_addr, - xcRB->request_control_blk_length)) - return -EFAULT; - if (msg->cprbx.cprb_len + sizeof(msg->hdr.function_code) > - xcRB->request_control_blk_length) - return -EINVAL; - function_code = ((unsigned char *)&msg->cprbx) + msg->cprbx.cprb_len; - memcpy(msg->hdr.function_code, function_code, sizeof(msg->hdr.function_code)); - - if (memcmp(function_code, "US", 2) == 0) - ap_msg->special = 1; - else - ap_msg->special = 0; - - /* copy data block */ - if (xcRB->request_data_length && - copy_from_user(req_data, xcRB->request_data_address, - xcRB->request_data_length)) - return -EFAULT; - return 0; -} - -/** - * Prepare a type6 CPRB message for random number generation - * - * @ap_dev: AP device pointer - * @ap_msg: pointer to AP message - */ -static void rng_type6CPRB_msgX(struct ap_device *ap_dev, - struct ap_message *ap_msg, - unsigned random_number_length) -{ - struct { - struct type6_hdr hdr; - struct CPRBX cprbx; - char function_code[2]; - short int rule_length; - char rule[8]; - short int verb_length; - short int key_length; - } __attribute__((packed)) *msg = ap_msg->message; - static struct type6_hdr static_type6_hdrX = { - .type = 0x06, - .offset1 = 0x00000058, - .agent_id = {'C', 'A'}, - .function_code = {'R', 'L'}, - .ToCardLen1 = sizeof *msg - sizeof(msg->hdr), - .FromCardLen1 = sizeof *msg - sizeof(msg->hdr), - }; - static struct CPRBX local_cprbx = { - .cprb_len = 0x00dc, - .cprb_ver_id = 0x02, - .func_id = {0x54, 0x32}, - .req_parml = sizeof *msg - sizeof(msg->hdr) - - sizeof(msg->cprbx), - .rpl_msgbl = sizeof *msg - sizeof(msg->hdr), - }; - - msg->hdr = static_type6_hdrX; - msg->hdr.FromCardLen2 = random_number_length, - msg->cprbx = local_cprbx; - msg->cprbx.rpl_datal = random_number_length, - msg->cprbx.domain = AP_QID_QUEUE(ap_dev->qid); - memcpy(msg->function_code, msg->hdr.function_code, 0x02); - msg->rule_length = 0x0a; - memcpy(msg->rule, "RANDOM ", 8); - msg->verb_length = 0x02; - msg->key_length = 0x02; - ap_msg->length = sizeof *msg; -} - -/** - * Copy results from a type 86 ICA reply message back to user space. - * - * @zdev: crypto device pointer - * @reply: reply AP message. - * @data: pointer to user output data - * @length: size of user output data - * - * Returns 0 on success or -EINVAL, -EFAULT, -EAGAIN in case of an error. - */ -struct type86x_reply { - struct type86_hdr hdr; - struct type86_fmt2_ext fmt2; - struct CPRBX cprbx; - unsigned char pad[4]; /* 4 byte function code/rules block ? */ - unsigned short length; - char text[0]; -} __attribute__((packed)); - -static int convert_type86_ica(struct zcrypt_device *zdev, - struct ap_message *reply, - char __user *outputdata, - unsigned int outputdatalength) -{ - static unsigned char static_pad[] = { - 0x00,0x02, - 0x1B,0x7B,0x5D,0xB5,0x75,0x01,0x3D,0xFD, - 0x8D,0xD1,0xC7,0x03,0x2D,0x09,0x23,0x57, - 0x89,0x49,0xB9,0x3F,0xBB,0x99,0x41,0x5B, - 0x75,0x21,0x7B,0x9D,0x3B,0x6B,0x51,0x39, - 0xBB,0x0D,0x35,0xB9,0x89,0x0F,0x93,0xA5, - 0x0B,0x47,0xF1,0xD3,0xBB,0xCB,0xF1,0x9D, - 0x23,0x73,0x71,0xFF,0xF3,0xF5,0x45,0xFB, - 0x61,0x29,0x23,0xFD,0xF1,0x29,0x3F,0x7F, - 0x17,0xB7,0x1B,0xA9,0x19,0xBD,0x57,0xA9, - 0xD7,0x95,0xA3,0xCB,0xED,0x1D,0xDB,0x45, - 0x7D,0x11,0xD1,0x51,0x1B,0xED,0x71,0xE9, - 0xB1,0xD1,0xAB,0xAB,0x21,0x2B,0x1B,0x9F, - 0x3B,0x9F,0xF7,0xF7,0xBD,0x63,0xEB,0xAD, - 0xDF,0xB3,0x6F,0x5B,0xDB,0x8D,0xA9,0x5D, - 0xE3,0x7D,0x77,0x49,0x47,0xF5,0xA7,0xFD, - 0xAB,0x2F,0x27,0x35,0x77,0xD3,0x49,0xC9, - 0x09,0xEB,0xB1,0xF9,0xBF,0x4B,0xCB,0x2B, - 0xEB,0xEB,0x05,0xFF,0x7D,0xC7,0x91,0x8B, - 0x09,0x83,0xB9,0xB9,0x69,0x33,0x39,0x6B, - 0x79,0x75,0x19,0xBF,0xBB,0x07,0x1D,0xBD, - 0x29,0xBF,0x39,0x95,0x93,0x1D,0x35,0xC7, - 0xC9,0x4D,0xE5,0x97,0x0B,0x43,0x9B,0xF1, - 0x16,0x93,0x03,0x1F,0xA5,0xFB,0xDB,0xF3, - 0x27,0x4F,0x27,0x61,0x05,0x1F,0xB9,0x23, - 0x2F,0xC3,0x81,0xA9,0x23,0x71,0x55,0x55, - 0xEB,0xED,0x41,0xE5,0xF3,0x11,0xF1,0x43, - 0x69,0x03,0xBD,0x0B,0x37,0x0F,0x51,0x8F, - 0x0B,0xB5,0x89,0x5B,0x67,0xA9,0xD9,0x4F, - 0x01,0xF9,0x21,0x77,0x37,0x73,0x79,0xC5, - 0x7F,0x51,0xC1,0xCF,0x97,0xA1,0x75,0xAD, - 0x35,0x9D,0xD3,0xD3,0xA7,0x9D,0x5D,0x41, - 0x6F,0x65,0x1B,0xCF,0xA9,0x87,0x91,0x09 - }; - struct type86x_reply *msg = reply->message; - unsigned short service_rc, service_rs; - unsigned int reply_len, pad_len; - char *data; - - service_rc = msg->cprbx.ccp_rtcode; - if (unlikely(service_rc != 0)) { - service_rs = msg->cprbx.ccp_rscode; - if (service_rc == 8 && service_rs == 66) - return -EINVAL; - if (service_rc == 8 && service_rs == 65) - return -EINVAL; - if (service_rc == 8 && service_rs == 770) - return -EINVAL; - if (service_rc == 8 && service_rs == 783) { - zdev->min_mod_size = PCIXCC_MIN_MOD_SIZE_OLD; - return -EAGAIN; - } - if (service_rc == 12 && service_rs == 769) - return -EINVAL; - if (service_rc == 8 && service_rs == 72) - return -EINVAL; - zdev->online = 0; - return -EAGAIN; /* repeat the request on a different device. */ - } - data = msg->text; - reply_len = msg->length - 2; - if (reply_len > outputdatalength) - return -EINVAL; - /* - * For all encipher requests, the length of the ciphertext (reply_len) - * will always equal the modulus length. For MEX decipher requests - * the output needs to get padded. Minimum pad size is 10. - * - * Currently, the cases where padding will be added is for: - * - PCIXCC_MCL2 using a CRT form token (since PKD didn't support - * ZERO-PAD and CRT is only supported for PKD requests) - * - PCICC, always - */ - pad_len = outputdatalength - reply_len; - if (pad_len > 0) { - if (pad_len < 10) - return -EINVAL; - /* 'restore' padding left in the PCICC/PCIXCC card. */ - if (copy_to_user(outputdata, static_pad, pad_len - 1)) - return -EFAULT; - if (put_user(0, outputdata + pad_len - 1)) - return -EFAULT; - } - /* Copy the crypto response to user space. */ - if (copy_to_user(outputdata + pad_len, data, reply_len)) - return -EFAULT; - return 0; -} - -/** - * Copy results from a type 86 XCRB reply message back to user space. - * - * @zdev: crypto device pointer - * @reply: reply AP message. - * @xcRB: pointer to XCRB - * - * Returns 0 on success or -EINVAL, -EFAULT, -EAGAIN in case of an error. - */ -static int convert_type86_xcrb(struct zcrypt_device *zdev, - struct ap_message *reply, - struct ica_xcRB *xcRB) -{ - struct type86_fmt2_msg *msg = reply->message; - char *data = reply->message; - - /* Copy CPRB to user */ - if (copy_to_user(xcRB->reply_control_blk_addr, - data + msg->fmt2.offset1, msg->fmt2.count1)) - return -EFAULT; - xcRB->reply_control_blk_length = msg->fmt2.count1; - - /* Copy data buffer to user */ - if (msg->fmt2.count2) - if (copy_to_user(xcRB->reply_data_addr, - data + msg->fmt2.offset2, msg->fmt2.count2)) - return -EFAULT; - xcRB->reply_data_length = msg->fmt2.count2; - return 0; -} - -static int convert_type86_rng(struct zcrypt_device *zdev, - struct ap_message *reply, - char *buffer) -{ - struct { - struct type86_hdr hdr; - struct type86_fmt2_ext fmt2; - struct CPRBX cprbx; - } __attribute__((packed)) *msg = reply->message; - char *data = reply->message; - - if (msg->cprbx.ccp_rtcode != 0 || msg->cprbx.ccp_rscode != 0) - return -EINVAL; - memcpy(buffer, data + msg->fmt2.offset2, msg->fmt2.count2); - return msg->fmt2.count2; -} - -static int convert_response_ica(struct zcrypt_device *zdev, - struct ap_message *reply, - char __user *outputdata, - unsigned int outputdatalength) -{ - struct type86x_reply *msg = reply->message; - - /* Response type byte is the second byte in the response. */ - switch (((unsigned char *) reply->message)[1]) { - case TYPE82_RSP_CODE: - case TYPE88_RSP_CODE: - return convert_error(zdev, reply); - case TYPE86_RSP_CODE: - if (msg->cprbx.ccp_rtcode && - (msg->cprbx.ccp_rscode == 0x14f) && - (outputdatalength > 256)) { - if (zdev->max_exp_bit_length <= 17) { - zdev->max_exp_bit_length = 17; - return -EAGAIN; - } else - return -EINVAL; - } - if (msg->hdr.reply_code) - return convert_error(zdev, reply); - if (msg->cprbx.cprb_ver_id == 0x02) - return convert_type86_ica(zdev, reply, - outputdata, outputdatalength); - /* Fall through, no break, incorrect cprb version is an unknown - * response */ - default: /* Unknown response type, this should NEVER EVER happen */ - zdev->online = 0; - return -EAGAIN; /* repeat the request on a different device. */ - } -} - -static int convert_response_xcrb(struct zcrypt_device *zdev, - struct ap_message *reply, - struct ica_xcRB *xcRB) -{ - struct type86x_reply *msg = reply->message; - - /* Response type byte is the second byte in the response. */ - switch (((unsigned char *) reply->message)[1]) { - case TYPE82_RSP_CODE: - case TYPE88_RSP_CODE: - xcRB->status = 0x0008044DL; /* HDD_InvalidParm */ - return convert_error(zdev, reply); - case TYPE86_RSP_CODE: - if (msg->hdr.reply_code) { - memcpy(&(xcRB->status), msg->fmt2.apfs, sizeof(u32)); - return convert_error(zdev, reply); - } - if (msg->cprbx.cprb_ver_id == 0x02) - return convert_type86_xcrb(zdev, reply, xcRB); - /* Fall through, no break, incorrect cprb version is an unknown - * response */ - default: /* Unknown response type, this should NEVER EVER happen */ - xcRB->status = 0x0008044DL; /* HDD_InvalidParm */ - zdev->online = 0; - return -EAGAIN; /* repeat the request on a different device. */ - } -} - -static int convert_response_rng(struct zcrypt_device *zdev, - struct ap_message *reply, - char *data) -{ - struct type86x_reply *msg = reply->message; - - switch (msg->hdr.type) { - case TYPE82_RSP_CODE: - case TYPE88_RSP_CODE: - return -EINVAL; - case TYPE86_RSP_CODE: - if (msg->hdr.reply_code) - return -EINVAL; - if (msg->cprbx.cprb_ver_id == 0x02) - return convert_type86_rng(zdev, reply, data); - /* Fall through, no break, incorrect cprb version is an unknown - * response */ - default: /* Unknown response type, this should NEVER EVER happen */ - zdev->online = 0; - return -EAGAIN; /* repeat the request on a different device. */ - } -} - -/** - * This function is called from the AP bus code after a crypto request - * "msg" has finished with the reply message "reply". - * It is called from tasklet context. - * @ap_dev: pointer to the AP device - * @msg: pointer to the AP message - * @reply: pointer to the AP reply message - */ -static void zcrypt_pcixcc_receive(struct ap_device *ap_dev, - struct ap_message *msg, - struct ap_message *reply) -{ - static struct error_hdr error_reply = { - .type = TYPE82_RSP_CODE, - .reply_code = REP82_ERROR_MACHINE_FAILURE, - }; - struct response_type *resp_type = - (struct response_type *) msg->private; - struct type86x_reply *t86r; - int length; - - /* Copy the reply message to the request message buffer. */ - if (IS_ERR(reply)) { - memcpy(msg->message, &error_reply, sizeof(error_reply)); - goto out; - } - t86r = reply->message; - if (t86r->hdr.type == TYPE86_RSP_CODE && - t86r->cprbx.cprb_ver_id == 0x02) { - switch (resp_type->type) { - case PCIXCC_RESPONSE_TYPE_ICA: - length = sizeof(struct type86x_reply) - + t86r->length - 2; - length = min(PCIXCC_MAX_ICA_RESPONSE_SIZE, length); - memcpy(msg->message, reply->message, length); - break; - case PCIXCC_RESPONSE_TYPE_XCRB: - length = t86r->fmt2.offset2 + t86r->fmt2.count2; - length = min(PCIXCC_MAX_XCRB_MESSAGE_SIZE, length); - memcpy(msg->message, reply->message, length); - break; - default: - memcpy(msg->message, &error_reply, sizeof error_reply); - } - } else - memcpy(msg->message, reply->message, sizeof error_reply); -out: - complete(&(resp_type->work)); -} - -static atomic_t zcrypt_step = ATOMIC_INIT(0); - -/** - * The request distributor calls this function if it picked the PCIXCC/CEX2C - * device to handle a modexpo request. - * @zdev: pointer to zcrypt_device structure that identifies the - * PCIXCC/CEX2C device to the request distributor - * @mex: pointer to the modexpo request buffer - */ -static long zcrypt_pcixcc_modexpo(struct zcrypt_device *zdev, - struct ica_rsa_modexpo *mex) -{ - struct ap_message ap_msg; - struct response_type resp_type = { - .type = PCIXCC_RESPONSE_TYPE_ICA, - }; - int rc; - - ap_init_message(&ap_msg); - ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL); - if (!ap_msg.message) - return -ENOMEM; - ap_msg.receive = zcrypt_pcixcc_receive; - ap_msg.psmid = (((unsigned long long) current->pid) << 32) + - atomic_inc_return(&zcrypt_step); - ap_msg.private = &resp_type; - rc = ICAMEX_msg_to_type6MEX_msgX(zdev, &ap_msg, mex); - if (rc) - goto out_free; - init_completion(&resp_type.work); - ap_queue_message(zdev->ap_dev, &ap_msg); - rc = wait_for_completion_interruptible(&resp_type.work); - if (rc == 0) - rc = convert_response_ica(zdev, &ap_msg, mex->outputdata, - mex->outputdatalength); - else - /* Signal pending. */ - ap_cancel_message(zdev->ap_dev, &ap_msg); -out_free: - free_page((unsigned long) ap_msg.message); - return rc; -} - -/** - * The request distributor calls this function if it picked the PCIXCC/CEX2C - * device to handle a modexpo_crt request. - * @zdev: pointer to zcrypt_device structure that identifies the - * PCIXCC/CEX2C device to the request distributor - * @crt: pointer to the modexpoc_crt request buffer - */ -static long zcrypt_pcixcc_modexpo_crt(struct zcrypt_device *zdev, - struct ica_rsa_modexpo_crt *crt) -{ - struct ap_message ap_msg; - struct response_type resp_type = { - .type = PCIXCC_RESPONSE_TYPE_ICA, - }; - int rc; - - ap_init_message(&ap_msg); - ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL); - if (!ap_msg.message) - return -ENOMEM; - ap_msg.receive = zcrypt_pcixcc_receive; - ap_msg.psmid = (((unsigned long long) current->pid) << 32) + - atomic_inc_return(&zcrypt_step); - ap_msg.private = &resp_type; - rc = ICACRT_msg_to_type6CRT_msgX(zdev, &ap_msg, crt); - if (rc) - goto out_free; - init_completion(&resp_type.work); - ap_queue_message(zdev->ap_dev, &ap_msg); - rc = wait_for_completion_interruptible(&resp_type.work); - if (rc == 0) - rc = convert_response_ica(zdev, &ap_msg, crt->outputdata, - crt->outputdatalength); - else - /* Signal pending. */ - ap_cancel_message(zdev->ap_dev, &ap_msg); -out_free: - free_page((unsigned long) ap_msg.message); - return rc; -} - -/** - * The request distributor calls this function if it picked the PCIXCC/CEX2C - * device to handle a send_cprb request. - * @zdev: pointer to zcrypt_device structure that identifies the - * PCIXCC/CEX2C device to the request distributor - * @xcRB: pointer to the send_cprb request buffer - */ -static long zcrypt_pcixcc_send_cprb(struct zcrypt_device *zdev, - struct ica_xcRB *xcRB) -{ - struct ap_message ap_msg; - struct response_type resp_type = { - .type = PCIXCC_RESPONSE_TYPE_XCRB, - }; - int rc; - - ap_init_message(&ap_msg); - ap_msg.message = kmalloc(PCIXCC_MAX_XCRB_MESSAGE_SIZE, GFP_KERNEL); - if (!ap_msg.message) - return -ENOMEM; - ap_msg.receive = zcrypt_pcixcc_receive; - ap_msg.psmid = (((unsigned long long) current->pid) << 32) + - atomic_inc_return(&zcrypt_step); - ap_msg.private = &resp_type; - rc = XCRB_msg_to_type6CPRB_msgX(zdev, &ap_msg, xcRB); - if (rc) - goto out_free; - init_completion(&resp_type.work); - ap_queue_message(zdev->ap_dev, &ap_msg); - rc = wait_for_completion_interruptible(&resp_type.work); - if (rc == 0) - rc = convert_response_xcrb(zdev, &ap_msg, xcRB); - else - /* Signal pending. */ - ap_cancel_message(zdev->ap_dev, &ap_msg); -out_free: - kzfree(ap_msg.message); - return rc; -} - -/** - * The request distributor calls this function if it picked the PCIXCC/CEX2C - * device to generate random data. - * @zdev: pointer to zcrypt_device structure that identifies the - * PCIXCC/CEX2C device to the request distributor - * @buffer: pointer to a memory page to return random data - */ - -static long zcrypt_pcixcc_rng(struct zcrypt_device *zdev, - char *buffer) -{ - struct ap_message ap_msg; - struct response_type resp_type = { - .type = PCIXCC_RESPONSE_TYPE_XCRB, - }; - int rc; - - ap_init_message(&ap_msg); - ap_msg.message = kmalloc(PCIXCC_MAX_XCRB_MESSAGE_SIZE, GFP_KERNEL); - if (!ap_msg.message) - return -ENOMEM; - ap_msg.receive = zcrypt_pcixcc_receive; - ap_msg.psmid = (((unsigned long long) current->pid) << 32) + - atomic_inc_return(&zcrypt_step); - ap_msg.private = &resp_type; - rng_type6CPRB_msgX(zdev->ap_dev, &ap_msg, ZCRYPT_RNG_BUFFER_SIZE); - init_completion(&resp_type.work); - ap_queue_message(zdev->ap_dev, &ap_msg); - rc = wait_for_completion_interruptible(&resp_type.work); - if (rc == 0) - rc = convert_response_rng(zdev, &ap_msg, buffer); - else - /* Signal pending. */ - ap_cancel_message(zdev->ap_dev, &ap_msg); - kfree(ap_msg.message); - return rc; -} - -/** - * The crypto operations for a PCIXCC/CEX2C card. - */ -static struct zcrypt_ops zcrypt_pcixcc_ops = { - .rsa_modexpo = zcrypt_pcixcc_modexpo, - .rsa_modexpo_crt = zcrypt_pcixcc_modexpo_crt, - .send_cprb = zcrypt_pcixcc_send_cprb, -}; - -static struct zcrypt_ops zcrypt_pcixcc_with_rng_ops = { - .rsa_modexpo = zcrypt_pcixcc_modexpo, - .rsa_modexpo_crt = zcrypt_pcixcc_modexpo_crt, - .send_cprb = zcrypt_pcixcc_send_cprb, - .rng = zcrypt_pcixcc_rng, -}; - -/** * Micro-code detection function. Its sends a message to a pcixcc card * to find out the microcode level. * @ap_dev: pointer to the AP device. @@ -1083,9 +323,11 @@ static int zcrypt_pcixcc_probe(struct ap_device *ap_dev) return rc; } if (rc) - zdev->ops = &zcrypt_pcixcc_with_rng_ops; + zdev->ops = zcrypt_msgtype_request(MSGTYPE06_NAME, + MSGTYPE06_VARIANT_DEFAULT); else - zdev->ops = &zcrypt_pcixcc_ops; + zdev->ops = zcrypt_msgtype_request(MSGTYPE06_NAME, + MSGTYPE06_VARIANT_NORNG); ap_dev->reply = &zdev->reply; ap_dev->private = zdev; rc = zcrypt_device_register(zdev); @@ -1095,6 +337,7 @@ static int zcrypt_pcixcc_probe(struct ap_device *ap_dev) out_free: ap_dev->private = NULL; + zcrypt_msgtype_release(zdev->ops); zcrypt_device_free(zdev); return rc; } @@ -1106,8 +349,10 @@ static int zcrypt_pcixcc_probe(struct ap_device *ap_dev) static void zcrypt_pcixcc_remove(struct ap_device *ap_dev) { struct zcrypt_device *zdev = ap_dev->private; + struct zcrypt_ops *zops = zdev->ops; zcrypt_device_unregister(zdev); + zcrypt_msgtype_release(zops); } int __init zcrypt_pcixcc_init(void) diff --git a/drivers/s390/crypto/zcrypt_pcixcc.h b/drivers/s390/crypto/zcrypt_pcixcc.h index c7cdf59..eacafc8 100644 --- a/drivers/s390/crypto/zcrypt_pcixcc.h +++ b/drivers/s390/crypto/zcrypt_pcixcc.h @@ -1,12 +1,13 @@ /* * zcrypt 2.1.0 * - * Copyright IBM Corp. 2001, 2006 + * Copyright IBM Corp. 2001, 2012 * Author(s): Robert Burroughs * Eric Rossman (edrossma@us.ibm.com) * * Hotplug & misc device support: Jochen Roehrig (roehrig@de.ibm.com) * Major cleanup & driver split: Martin Schwidefsky + * MSGTYPE restruct: Holger Dengler * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by -- cgit v0.10.2 From 1e2076f4527b5b0854d0ebe60102b6710a9ba64c Mon Sep 17 00:00:00 2001 From: Holger Dengler Date: Tue, 28 Aug 2012 16:48:29 +0200 Subject: s390/zcrypt: Add support for CEX4 crypto card New zcrypt module supports IBM CryptoExpress 4 cards. Signed-off-by: Holger Dengler Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/crypto/Makefile b/drivers/s390/crypto/Makefile index eca0bfa..771faf7 100644 --- a/drivers/s390/crypto/Makefile +++ b/drivers/s390/crypto/Makefile @@ -4,5 +4,5 @@ ap-objs := ap_bus.o obj-$(CONFIG_ZCRYPT) += ap.o zcrypt_api.o zcrypt_pcicc.o zcrypt_pcixcc.o -obj-$(CONFIG_ZCRYPT) += zcrypt_pcica.o zcrypt_cex2a.o +obj-$(CONFIG_ZCRYPT) += zcrypt_pcica.o zcrypt_cex2a.o zcrypt_cex4.o obj-$(CONFIG_ZCRYPT) += zcrypt_msgtype6.o zcrypt_msgtype50.o diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index f0b53ee..39e5eb2 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -1340,30 +1340,22 @@ static void ap_scan_bus(struct work_struct *unused) (unsigned long) ap_dev); switch (device_type) { case 0: + /* device type probing for old cards */ if (ap_probe_device_type(ap_dev)) { kfree(ap_dev); continue; } break; - case 10: - if (ap_query_functions(qid, &device_functions)) { - kfree(ap_dev); - continue; - } - ap_dev->functions = device_functions; - if (ap_test_bit(&device_functions, 3)) - ap_dev->device_type = AP_DEVICE_TYPE_CEX3C; - else if (ap_test_bit(&device_functions, 4)) - ap_dev->device_type = AP_DEVICE_TYPE_CEX3A; - else { - kfree(ap_dev); - continue; - } - break; default: ap_dev->device_type = device_type; } + rc = ap_query_functions(qid, &device_functions); + if (!rc) + ap_dev->functions = device_functions; + else + ap_dev->functions = 0u; + ap_dev->device.bus = &ap_bus_type; ap_dev->device.parent = ap_root_device; if (dev_set_name(&ap_dev->device, "card%02x", diff --git a/drivers/s390/crypto/ap_bus.h b/drivers/s390/crypto/ap_bus.h index e49bcf4..e9b963e 100644 --- a/drivers/s390/crypto/ap_bus.h +++ b/drivers/s390/crypto/ap_bus.h @@ -116,6 +116,7 @@ static inline int ap_test_bit(unsigned int *ptr, unsigned int nr) #define AP_DEVICE_TYPE_CEX2C 7 #define AP_DEVICE_TYPE_CEX3A 8 #define AP_DEVICE_TYPE_CEX3C 9 +#define AP_DEVICE_TYPE_CEX4 10 /* * Known function facilities diff --git a/drivers/s390/crypto/zcrypt_cex4.c b/drivers/s390/crypto/zcrypt_cex4.c new file mode 100644 index 0000000..ce12263 --- /dev/null +++ b/drivers/s390/crypto/zcrypt_cex4.c @@ -0,0 +1,149 @@ +/* + * Copyright IBM Corp. 2012 + * Author(s): Holger Dengler + */ + +#include +#include +#include +#include +#include +#include + +#include "ap_bus.h" +#include "zcrypt_api.h" +#include "zcrypt_msgtype6.h" +#include "zcrypt_msgtype50.h" +#include "zcrypt_error.h" +#include "zcrypt_cex4.h" + +#define CEX4A_MIN_MOD_SIZE 1 /* 8 bits */ +#define CEX4A_MAX_MOD_SIZE_2K 256 /* 2048 bits */ +#define CEX4A_MAX_MOD_SIZE_4K 512 /* 4096 bits */ + +#define CEX4C_MIN_MOD_SIZE 16 /* 256 bits */ +#define CEX4C_MAX_MOD_SIZE 512 /* 4096 bits */ + +#define CEX4A_SPEED_RATING 900 /* TODO new card, new speed rating */ +#define CEX4C_SPEED_RATING 6500 /* TODO new card, new speed rating */ + +#define CEX4A_MAX_MESSAGE_SIZE MSGTYPE50_CRB3_MAX_MSG_SIZE +#define CEX4C_MAX_MESSAGE_SIZE MSGTYPE06_MAX_MSG_SIZE + +#define CEX4_CLEANUP_TIME (15*HZ) + +static struct ap_device_id zcrypt_cex4_ids[] = { + { AP_DEVICE(AP_DEVICE_TYPE_CEX4) }, + { /* end of list */ }, +}; + +MODULE_DEVICE_TABLE(ap, zcrypt_cex4_ids); +MODULE_AUTHOR("IBM Corporation"); +MODULE_DESCRIPTION("CEX4 Cryptographic Card device driver, " \ + "Copyright IBM Corp. 2012"); +MODULE_LICENSE("GPL"); + +static int zcrypt_cex4_probe(struct ap_device *ap_dev); +static void zcrypt_cex4_remove(struct ap_device *ap_dev); + +static struct ap_driver zcrypt_cex4_driver = { + .probe = zcrypt_cex4_probe, + .remove = zcrypt_cex4_remove, + .ids = zcrypt_cex4_ids, + .request_timeout = CEX4_CLEANUP_TIME, +}; + +/** + * Probe function for CEX4 cards. It always accepts the AP device + * since the bus_match already checked the hardware type. + * @ap_dev: pointer to the AP device. + */ +static int zcrypt_cex4_probe(struct ap_device *ap_dev) +{ + struct zcrypt_device *zdev = NULL; + int rc = 0; + + switch (ap_dev->device_type) { + case AP_DEVICE_TYPE_CEX4: + if (ap_test_bit(&ap_dev->functions, AP_FUNC_ACCEL)) { + zdev = zcrypt_device_alloc(CEX4A_MAX_MESSAGE_SIZE); + if (!zdev) + return -ENOMEM; + zdev->type_string = "CEX4A"; + zdev->user_space_type = ZCRYPT_CEX3A; + zdev->min_mod_size = CEX4A_MIN_MOD_SIZE; + if (ap_test_bit(&ap_dev->functions, AP_FUNC_MEX4K) && + ap_test_bit(&ap_dev->functions, AP_FUNC_CRT4K)) { + zdev->max_mod_size = + CEX4A_MAX_MOD_SIZE_4K; + zdev->max_exp_bit_length = + CEX4A_MAX_MOD_SIZE_4K; + } else { + zdev->max_mod_size = + CEX4A_MAX_MOD_SIZE_2K; + zdev->max_exp_bit_length = + CEX4A_MAX_MOD_SIZE_2K; + } + zdev->short_crt = 1; + zdev->speed_rating = CEX4A_SPEED_RATING; + zdev->ops = zcrypt_msgtype_request(MSGTYPE50_NAME, + MSGTYPE50_VARIANT_DEFAULT); + } else if (ap_test_bit(&ap_dev->functions, AP_FUNC_COPRO)) { + zdev = zcrypt_device_alloc(CEX4C_MAX_MESSAGE_SIZE); + if (!zdev) + return -ENOMEM; + zdev->type_string = "CEX4C"; + zdev->user_space_type = ZCRYPT_CEX3C; + zdev->min_mod_size = CEX4C_MIN_MOD_SIZE; + zdev->max_mod_size = CEX4C_MAX_MOD_SIZE; + zdev->max_exp_bit_length = CEX4C_MAX_MOD_SIZE; + zdev->short_crt = 0; + zdev->speed_rating = CEX4C_SPEED_RATING; + zdev->ops = zcrypt_msgtype_request(MSGTYPE06_NAME, + MSGTYPE06_VARIANT_DEFAULT); + } + break; + } + if (!zdev) + return -ENODEV; + zdev->ap_dev = ap_dev; + zdev->online = 1; + ap_dev->reply = &zdev->reply; + ap_dev->private = zdev; + rc = zcrypt_device_register(zdev); + if (rc) { + zcrypt_msgtype_release(zdev->ops); + ap_dev->private = NULL; + zcrypt_device_free(zdev); + } + return rc; +} + +/** + * This is called to remove the extended CEX4 driver information + * if an AP device is removed. + */ +static void zcrypt_cex4_remove(struct ap_device *ap_dev) +{ + struct zcrypt_device *zdev = ap_dev->private; + struct zcrypt_ops *zops; + + if (zdev) { + zops = zdev->ops; + zcrypt_device_unregister(zdev); + zcrypt_msgtype_release(zops); + } +} + +int __init zcrypt_cex4_init(void) +{ + return ap_driver_register(&zcrypt_cex4_driver, THIS_MODULE, "cex4"); +} + +void __exit zcrypt_cex4_exit(void) +{ + ap_driver_unregister(&zcrypt_cex4_driver); +} + +module_init(zcrypt_cex4_init); +module_exit(zcrypt_cex4_exit); diff --git a/drivers/s390/crypto/zcrypt_cex4.h b/drivers/s390/crypto/zcrypt_cex4.h new file mode 100644 index 0000000..7195713 --- /dev/null +++ b/drivers/s390/crypto/zcrypt_cex4.h @@ -0,0 +1,12 @@ +/* + * Copyright IBM Corp. 2012 + * Author(s): Holger Dengler + */ + +#ifndef _ZCRYPT_CEX4_H_ +#define _ZCRYPT_CEX4_H_ + +int zcrypt_cex4_init(void); +void zcrypt_cex4_exit(void); + +#endif /* _ZCRYPT_CEX4_H_ */ -- cgit v0.10.2 From 01c5e6dc119bf2112e363477e3b056f6b54f2d79 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:42:37 +0200 Subject: s390/css: export css_sched_sch_todo Allow subchannel drivers build as a module to schedule subchannel work. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/cio/css.c b/drivers/s390/cio/css.c index 21908e6..b4d572f 100644 --- a/drivers/s390/cio/css.c +++ b/drivers/s390/cio/css.c @@ -445,6 +445,7 @@ void css_sched_sch_todo(struct subchannel *sch, enum sch_todo todo) put_device(&sch->dev); } } +EXPORT_SYMBOL_GPL(css_sched_sch_todo); static void css_sch_todo(struct work_struct *work) { -- cgit v0.10.2 From 382b73663529b756914cf17a9912c9cf00cf0025 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:43:36 +0200 Subject: s390: add eadm facility bits Add the eadm facility bits to the css characteristics and move them to a new header. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/chsc.h b/arch/s390/include/asm/chsc.h index bf115b4..aea451f 100644 --- a/arch/s390/include/asm/chsc.h +++ b/arch/s390/include/asm/chsc.h @@ -125,32 +125,4 @@ struct chsc_cpd_info { #define CHSC_INFO_CPD _IOWR(CHSC_IOCTL_MAGIC, 0x87, struct chsc_cpd_info) #define CHSC_INFO_DCAL _IOWR(CHSC_IOCTL_MAGIC, 0x88, struct chsc_dcal) -#ifdef __KERNEL__ - -struct css_general_char { - u64 : 12; - u32 dynio : 1; /* bit 12 */ - u32 : 28; - u32 aif : 1; /* bit 41 */ - u32 : 3; - u32 mcss : 1; /* bit 45 */ - u32 fcs : 1; /* bit 46 */ - u32 : 1; - u32 ext_mb : 1; /* bit 48 */ - u32 : 7; - u32 aif_tdd : 1; /* bit 56 */ - u32 : 1; - u32 qebsm : 1; /* bit 58 */ - u32 : 8; - u32 aif_osa : 1; /* bit 67 */ - u32 : 14; - u32 cib : 1; /* bit 82 */ - u32 : 5; - u32 fcx : 1; /* bit 88 */ - u32 : 7; -}__attribute__((packed)); - -extern struct css_general_char css_general_characteristics; - -#endif /* __KERNEL__ */ #endif diff --git a/arch/s390/include/asm/css_chars.h b/arch/s390/include/asm/css_chars.h new file mode 100644 index 0000000..a06ebc2 --- /dev/null +++ b/arch/s390/include/asm/css_chars.h @@ -0,0 +1,39 @@ +#ifndef _ASM_CSS_CHARS_H +#define _ASM_CSS_CHARS_H + +#include + +#ifdef __KERNEL__ + +struct css_general_char { + u64 : 12; + u32 dynio : 1; /* bit 12 */ + u32 : 4; + u32 eadm : 1; /* bit 17 */ + u32 : 23; + u32 aif : 1; /* bit 41 */ + u32 : 3; + u32 mcss : 1; /* bit 45 */ + u32 fcs : 1; /* bit 46 */ + u32 : 1; + u32 ext_mb : 1; /* bit 48 */ + u32 : 7; + u32 aif_tdd : 1; /* bit 56 */ + u32 : 1; + u32 qebsm : 1; /* bit 58 */ + u32 : 8; + u32 aif_osa : 1; /* bit 67 */ + u32 : 12; + u32 eadm_rf : 1; /* bit 80 */ + u32 : 1; + u32 cib : 1; /* bit 82 */ + u32 : 5; + u32 fcx : 1; /* bit 88 */ + u32 : 19; + u32 alt_ssi : 1; /* bit 108 */ +} __packed; + +extern struct css_general_char css_general_characteristics; + +#endif /* __KERNEL__ */ +#endif diff --git a/arch/s390/include/asm/scsw.h b/arch/s390/include/asm/scsw.h index 4071d00..8096bfe 100644 --- a/arch/s390/include/asm/scsw.h +++ b/arch/s390/include/asm/scsw.h @@ -9,7 +9,7 @@ #define _ASM_S390_SCSW_H_ #include -#include +#include #include /** diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index c48c72a..ed03c22 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -20,6 +20,7 @@ #include #include +#include #include #include #include @@ -31,8 +32,6 @@ #include "dasd_int.h" #include "dasd_eckd.h" -#include "../cio/chsc.h" - #ifdef PRINTK_HEADER #undef PRINTK_HEADER diff --git a/drivers/s390/cio/chsc.h b/drivers/s390/cio/chsc.h index 3f15b2a..b5261f3 100644 --- a/drivers/s390/cio/chsc.h +++ b/drivers/s390/cio/chsc.h @@ -3,6 +3,7 @@ #include #include +#include #include #include #include -- cgit v0.10.2 From d2fc439b99820cccd6978918c260730dd97bf373 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:44:51 +0200 Subject: s390: add eadm related structures Add structures to be used by the eadm subchannel driver. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/cio.h b/arch/s390/include/asm/cio.h index 77043aa..55bde60 100644 --- a/arch/s390/include/asm/cio.h +++ b/arch/s390/include/asm/cio.h @@ -80,6 +80,18 @@ struct erw { } __attribute__ ((packed)); /** + * struct erw_eadm - EADM Subchannel extended report word + * @b: aob error + * @r: arsb error + */ +struct erw_eadm { + __u32 : 16; + __u32 b : 1; + __u32 r : 1; + __u32 : 14; +} __packed; + +/** * struct sublog - subchannel logout area * @res0: reserved * @esf: extended status flags @@ -170,9 +182,22 @@ struct esw3 { } __attribute__ ((packed)); /** + * struct esw_eadm - EADM Subchannel Extended Status Word (ESW) + * @sublog: subchannel logout + * @erw: extended report word + */ +struct esw_eadm { + __u32 sublog; + struct erw_eadm erw; + __u32 : 32; + __u32 : 32; + __u32 : 32; +} __packed; + +/** * struct irb - interruption response block * @scsw: subchannel status word - * @esw: extened status word, 4 formats + * @esw: extened status word * @ecw: extended control word * * The irb that is handed to the device driver when an interrupt occurs. For @@ -191,6 +216,7 @@ struct irb { struct esw1 esw1; struct esw2 esw2; struct esw3 esw3; + struct esw_eadm eadm; } esw; __u8 ecw[32]; } __attribute__ ((packed,aligned(4))); diff --git a/arch/s390/include/asm/eadm.h b/arch/s390/include/asm/eadm.h new file mode 100644 index 0000000..4a65803 --- /dev/null +++ b/arch/s390/include/asm/eadm.h @@ -0,0 +1,74 @@ +#ifndef _ASM_S390_EADM_H +#define _ASM_S390_EADM_H + +#include + +struct arqb { + u64 data; + u16 fmt:4; + u16:12; + u16 cmd_code; + u16:16; + u16 msb_count; + u32 reserved[12]; +} __packed; + +#define ARQB_CMD_MOVE 1 + +struct arsb { + u16 fmt:4; + u32:28; + u8 ef; + u8:8; + u8 ecbi; + u8:8; + u8 fvf; + u16:16; + u8 eqc; + u32:32; + u64 fail_msb; + u64 fail_aidaw; + u64 fail_ms; + u64 fail_scm; + u32 reserved[4]; +} __packed; + +struct msb { + u8 fmt:4; + u8 oc:4; + u8 flags; + u16:12; + u16 bs:4; + u32 blk_count; + u64 data_addr; + u64 scm_addr; + u64:64; +} __packed; + +struct aidaw { + u8 flags; + u32 :24; + u32 :32; + u64 data_addr; +} __packed; + +#define MSB_OC_CLEAR 0 +#define MSB_OC_READ 1 +#define MSB_OC_WRITE 2 +#define MSB_OC_RELEASE 3 + +#define MSB_FLAG_BNM 0x80 +#define MSB_FLAG_IDA 0x40 + +#define MSB_BS_4K 0 +#define MSB_BS_1M 1 + +#define AOB_NR_MSB 124 + +struct aob { + struct arqb request; + struct arsb response; + struct msb msb[AOB_NR_MSB]; +} __packed __aligned(PAGE_SIZE); + +#endif /* _ASM_S390_EADM_H */ diff --git a/arch/s390/include/asm/scsw.h b/arch/s390/include/asm/scsw.h index 8096bfe..4af99cd 100644 --- a/arch/s390/include/asm/scsw.h +++ b/arch/s390/include/asm/scsw.h @@ -1,7 +1,7 @@ /* * Helper functions for scsw access. * - * Copyright IBM Corp. 2008, 2009 + * Copyright IBM Corp. 2008, 2012 * Author(s): Peter Oberparleiter */ @@ -100,14 +100,46 @@ struct tm_scsw { } __attribute__ ((packed)); /** + * struct eadm_scsw - subchannel status word for eadm subchannels + * @key: subchannel key + * @eswf: esw format + * @cc: deferred condition code + * @ectl: extended control + * @fctl: function control + * @actl: activity control + * @stctl: status control + * @aob: AOB address + * @dstat: device status + * @cstat: subchannel status + */ +struct eadm_scsw { + u32 key:4; + u32:1; + u32 eswf:1; + u32 cc:2; + u32:6; + u32 ectl:1; + u32:2; + u32 fctl:3; + u32 actl:7; + u32 stctl:5; + u32 aob; + u32 dstat:8; + u32 cstat:8; + u32:16; +} __packed; + +/** * union scsw - subchannel status word * @cmd: command-mode SCSW * @tm: transport-mode SCSW + * @eadm: eadm SCSW */ union scsw { struct cmd_scsw cmd; struct tm_scsw tm; -} __attribute__ ((packed)); + struct eadm_scsw eadm; +} __packed; #define SCSW_FCTL_CLEAR_FUNC 0x1 #define SCSW_FCTL_HALT_FUNC 0x2 diff --git a/drivers/s390/cio/orb.h b/drivers/s390/cio/orb.h index 45a9865..7a64053 100644 --- a/drivers/s390/cio/orb.h +++ b/drivers/s390/cio/orb.h @@ -59,9 +59,33 @@ struct tm_orb { u32:32; } __packed __aligned(4); +/* + * eadm operation request block + */ +struct eadm_orb { + u32 intparm; + u32 key:4; + u32:4; + u32 compat1:1; + u32 compat2:1; + u32:21; + u32 x:1; + u32 aob; + u32 css_prio:8; + u32:8; + u32 scm_prio:8; + u32:8; + u32:29; + u32 fmt:3; + u32:32; + u32:32; + u32:32; +} __packed __aligned(4); + union orb { struct cmd_orb cmd; struct tm_orb tm; + struct eadm_orb eadm; } __packed __aligned(4); #endif /* S390_ORB_H */ -- cgit v0.10.2 From 184b08afb5eab8a43d75f4aa0a0f912653f797d7 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:45:42 +0200 Subject: s390/chsc: implement store SCM information chsc_scm_info will be used to detect all available increments of storage class memory. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/cio/chsc.c b/drivers/s390/cio/chsc.c index cfe0c08..1218653 100644 --- a/drivers/s390/cio/chsc.c +++ b/drivers/s390/cio/chsc.c @@ -52,6 +52,11 @@ int chsc_error_from_response(int response) return -EINVAL; case 0x0004: return -EOPNOTSUPP; + case 0x000b: + return -EBUSY; + case 0x0100: + case 0x0102: + return -ENOMEM; default: return -EIO; } @@ -1047,3 +1052,33 @@ out: return rc; } EXPORT_SYMBOL_GPL(chsc_siosl); + +/** + * chsc_scm_info() - store SCM information (SSI) + * @scm_area: request and response block for SSI + * @token: continuation token + * + * Returns 0 on success. + */ +int chsc_scm_info(struct chsc_scm_info *scm_area, u64 token) +{ + int ccode, ret; + + memset(scm_area, 0, sizeof(*scm_area)); + scm_area->request.length = 0x0020; + scm_area->request.code = 0x004C; + scm_area->reqtok = token; + + ccode = chsc(scm_area); + if (ccode > 0) { + ret = (ccode == 3) ? -ENODEV : -EBUSY; + goto out; + } + ret = chsc_error_from_response(scm_area->response.code); + if (ret != 0) + CIO_MSG_EVENT(2, "chsc: scm info failed (rc=%04x)\n", + scm_area->response.code); +out: + return ret; +} +EXPORT_SYMBOL_GPL(chsc_scm_info); diff --git a/drivers/s390/cio/chsc.h b/drivers/s390/cio/chsc.h index b5261f3..3aa94fe 100644 --- a/drivers/s390/cio/chsc.h +++ b/drivers/s390/cio/chsc.h @@ -119,4 +119,39 @@ int chsc_error_from_response(int response); int chsc_siosl(struct subchannel_id schid); +/* Functions and definitions to query storage-class memory. */ +struct sale { + u64 sa; + u32 p:4; + u32 op_state:4; + u32 data_state:4; + u32 rank:4; + u32 r:1; + u32:7; + u32 rid:8; + u32:32; +} __packed; + +struct chsc_scm_info { + struct chsc_header request; + u32:32; + u64 reqtok; + u32 reserved1[4]; + struct chsc_header response; + u64:56; + u8 rq; + u32 mbc; + u64 msa; + u16 is; + u16 mmc; + u32 mci; + u64 nr_scm_ini; + u64 nr_scm_unini; + u32 reserved2[10]; + u64 restok; + struct sale scmal[248]; +} __packed; + +int chsc_scm_info(struct chsc_scm_info *scm_area, u64 token); + #endif -- cgit v0.10.2 From 1d1c8f78bed5f8e769757525bd9c2dec69f11a44 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:46:26 +0200 Subject: s390: add scm bus driver Bus driver to manage Storage Class Memory. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 443fac9..3781db2 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -433,6 +433,13 @@ config CHSC_SCH If unsure, say N. +config SCM_BUS + def_bool y + depends on 64BIT + prompt "SCM bus driver" + help + Bus driver for Storage Class Memory. + endmenu menu "Dump support" diff --git a/arch/s390/include/asm/eadm.h b/arch/s390/include/asm/eadm.h index 4a65803..3922f52 100644 --- a/arch/s390/include/asm/eadm.h +++ b/arch/s390/include/asm/eadm.h @@ -2,6 +2,8 @@ #define _ASM_S390_EADM_H #include +#include +#include struct arqb { u64 data; @@ -71,4 +73,53 @@ struct aob { struct msb msb[AOB_NR_MSB]; } __packed __aligned(PAGE_SIZE); +struct aob_rq_header { + struct scm_device *scmdev; + char data[0]; +}; + +struct scm_device { + u64 address; + u64 size; + unsigned int nr_max_block; + struct device dev; + spinlock_t lock; + struct { + unsigned int persistence:4; + unsigned int oper_state:4; + unsigned int data_state:4; + unsigned int rank:4; + unsigned int release:1; + unsigned int res_id:8; + } __packed attrs; +}; + +#define OP_STATE_GOOD 1 +#define OP_STATE_TEMP_ERR 2 +#define OP_STATE_PERM_ERR 3 + +struct scm_driver { + struct device_driver drv; + int (*probe) (struct scm_device *scmdev); + int (*remove) (struct scm_device *scmdev); + void (*handler) (struct scm_device *scmdev, void *data, int error); +}; + +int scm_driver_register(struct scm_driver *scmdrv); +void scm_driver_unregister(struct scm_driver *scmdrv); + +int scm_start_aob(struct aob *aob); +void scm_irq_handler(struct aob *aob, int error); + +struct eadm_ops { + int (*eadm_start) (struct aob *aob); + struct module *owner; +}; + +int scm_get_ref(void); +void scm_put_ref(void); + +void register_eadm_ops(struct eadm_ops *ops); +void unregister_eadm_ops(struct eadm_ops *ops); + #endif /* _ASM_S390_EADM_H */ diff --git a/drivers/s390/cio/Makefile b/drivers/s390/cio/Makefile index e1b700a..ac8f210 100644 --- a/drivers/s390/cio/Makefile +++ b/drivers/s390/cio/Makefile @@ -8,6 +8,7 @@ ccw_device-objs += device.o device_fsm.o device_ops.o ccw_device-objs += device_id.o device_pgid.o device_status.o obj-y += ccw_device.o cmf.o obj-$(CONFIG_CHSC_SCH) += chsc_sch.o +obj-$(CONFIG_SCM_BUS) += scm.o obj-$(CONFIG_CCWGROUP) += ccwgroup.o qdio-objs := qdio_main.o qdio_thinint.o qdio_debug.o qdio_setup.o diff --git a/drivers/s390/cio/scm.c b/drivers/s390/cio/scm.c new file mode 100644 index 0000000..874b64a --- /dev/null +++ b/drivers/s390/cio/scm.c @@ -0,0 +1,271 @@ +/* + * Recognize and maintain s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include "chsc.h" + +static struct device *scm_root; +static struct eadm_ops *eadm_ops; +static DEFINE_MUTEX(eadm_ops_mutex); + +#define to_scm_dev(n) container_of(n, struct scm_device, dev) +#define to_scm_drv(d) container_of(d, struct scm_driver, drv) + +static int scmdev_probe(struct device *dev) +{ + struct scm_device *scmdev = to_scm_dev(dev); + struct scm_driver *scmdrv = to_scm_drv(dev->driver); + + return scmdrv->probe ? scmdrv->probe(scmdev) : -ENODEV; +} + +static int scmdev_remove(struct device *dev) +{ + struct scm_device *scmdev = to_scm_dev(dev); + struct scm_driver *scmdrv = to_scm_drv(dev->driver); + + return scmdrv->remove ? scmdrv->remove(scmdev) : -ENODEV; +} + +static int scmdev_uevent(struct device *dev, struct kobj_uevent_env *env) +{ + return add_uevent_var(env, "MODALIAS=scm:scmdev"); +} + +static struct bus_type scm_bus_type = { + .name = "scm", + .probe = scmdev_probe, + .remove = scmdev_remove, + .uevent = scmdev_uevent, +}; + +/** + * scm_driver_register() - register a scm driver + * @scmdrv: driver to be registered + */ +int scm_driver_register(struct scm_driver *scmdrv) +{ + struct device_driver *drv = &scmdrv->drv; + + drv->bus = &scm_bus_type; + + return driver_register(drv); +} +EXPORT_SYMBOL_GPL(scm_driver_register); + +/** + * scm_driver_unregister() - deregister a scm driver + * @scmdrv: driver to be deregistered + */ +void scm_driver_unregister(struct scm_driver *scmdrv) +{ + driver_unregister(&scmdrv->drv); +} +EXPORT_SYMBOL_GPL(scm_driver_unregister); + +int scm_get_ref(void) +{ + int ret = 0; + + mutex_lock(&eadm_ops_mutex); + if (!eadm_ops || !try_module_get(eadm_ops->owner)) + ret = -ENOENT; + mutex_unlock(&eadm_ops_mutex); + + return ret; +} +EXPORT_SYMBOL_GPL(scm_get_ref); + +void scm_put_ref(void) +{ + mutex_lock(&eadm_ops_mutex); + module_put(eadm_ops->owner); + mutex_unlock(&eadm_ops_mutex); +} +EXPORT_SYMBOL_GPL(scm_put_ref); + +void register_eadm_ops(struct eadm_ops *ops) +{ + mutex_lock(&eadm_ops_mutex); + eadm_ops = ops; + mutex_unlock(&eadm_ops_mutex); +} +EXPORT_SYMBOL_GPL(register_eadm_ops); + +void unregister_eadm_ops(struct eadm_ops *ops) +{ + mutex_lock(&eadm_ops_mutex); + eadm_ops = NULL; + mutex_unlock(&eadm_ops_mutex); +} +EXPORT_SYMBOL_GPL(unregister_eadm_ops); + +int scm_start_aob(struct aob *aob) +{ + return eadm_ops->eadm_start(aob); +} +EXPORT_SYMBOL_GPL(scm_start_aob); + +void scm_irq_handler(struct aob *aob, int error) +{ + struct aob_rq_header *aobrq = (void *) aob->request.data; + struct scm_device *scmdev = aobrq->scmdev; + struct scm_driver *scmdrv = to_scm_drv(scmdev->dev.driver); + + scmdrv->handler(scmdev, aobrq->data, error); +} +EXPORT_SYMBOL_GPL(scm_irq_handler); + +#define scm_attr(name) \ +static ssize_t show_##name(struct device *dev, \ + struct device_attribute *attr, char *buf) \ +{ \ + struct scm_device *scmdev = to_scm_dev(dev); \ + int ret; \ + \ + spin_lock(&scmdev->lock); \ + ret = sprintf(buf, "%u\n", scmdev->attrs.name); \ + spin_unlock(&scmdev->lock); \ + \ + return ret; \ +} \ +static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL); + +scm_attr(persistence); +scm_attr(oper_state); +scm_attr(data_state); +scm_attr(rank); +scm_attr(release); +scm_attr(res_id); + +static struct attribute *scmdev_attrs[] = { + &dev_attr_persistence.attr, + &dev_attr_oper_state.attr, + &dev_attr_data_state.attr, + &dev_attr_rank.attr, + &dev_attr_release.attr, + &dev_attr_res_id.attr, + NULL, +}; + +static struct attribute_group scmdev_attr_group = { + .attrs = scmdev_attrs, +}; + +static const struct attribute_group *scmdev_attr_groups[] = { + &scmdev_attr_group, + NULL, +}; + +static void scmdev_release(struct device *dev) +{ + struct scm_device *scmdev = to_scm_dev(dev); + + kfree(scmdev); +} + +static void scmdev_setup(struct scm_device *scmdev, struct sale *sale, + unsigned int size, unsigned int max_blk_count) +{ + dev_set_name(&scmdev->dev, "%016llx", (unsigned long long) sale->sa); + scmdev->nr_max_block = max_blk_count; + scmdev->address = sale->sa; + scmdev->size = 1UL << size; + scmdev->attrs.rank = sale->rank; + scmdev->attrs.persistence = sale->p; + scmdev->attrs.oper_state = sale->op_state; + scmdev->attrs.data_state = sale->data_state; + scmdev->attrs.rank = sale->rank; + scmdev->attrs.release = sale->r; + scmdev->attrs.res_id = sale->rid; + scmdev->dev.parent = scm_root; + scmdev->dev.bus = &scm_bus_type; + scmdev->dev.release = scmdev_release; + scmdev->dev.groups = scmdev_attr_groups; + spin_lock_init(&scmdev->lock); +} + +static int scm_add(struct chsc_scm_info *scm_info, size_t num) +{ + struct sale *sale, *scmal = scm_info->scmal; + struct scm_device *scmdev; + int ret; + + for (sale = scmal; sale < scmal + num; sale++) { + scmdev = kzalloc(sizeof(*scmdev), GFP_KERNEL); + if (!scmdev) + return -ENODEV; + scmdev_setup(scmdev, sale, scm_info->is, scm_info->mbc); + ret = device_register(&scmdev->dev); + if (ret) { + /* Release reference from device_initialize(). */ + put_device(&scmdev->dev); + return ret; + } + } + + return 0; +} + +static int scm_update_information(void) +{ + struct chsc_scm_info *scm_info; + u64 token = 0; + size_t num; + int ret; + + scm_info = (void *)__get_free_page(GFP_KERNEL | GFP_DMA); + if (!scm_info) + return -ENOMEM; + + do { + ret = chsc_scm_info(scm_info, token); + if (ret) + break; + + num = (scm_info->response.length - + (offsetof(struct chsc_scm_info, scmal) - + offsetof(struct chsc_scm_info, response)) + ) / sizeof(struct sale); + + ret = scm_add(scm_info, num); + if (ret) + break; + + token = scm_info->restok; + } while (token); + + free_page((unsigned long)scm_info); + + return ret; +} + +static int __init scm_init(void) +{ + int ret; + + ret = bus_register(&scm_bus_type); + if (ret) + return ret; + + scm_root = root_device_register("scm"); + if (IS_ERR(scm_root)) { + bus_unregister(&scm_bus_type); + return PTR_ERR(scm_root); + } + + scm_update_information(); + return 0; +} +subsys_initcall_sync(scm_init); -- cgit v0.10.2 From 40ff4cc06697e8ba3f8ce93b0592ddbcf70cd444 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:47:02 +0200 Subject: s390: add scm notification Detect an scm change notification in store event information. Update affected scm devices and notify their drivers. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/eadm.h b/arch/s390/include/asm/eadm.h index 3922f52..4d6e103 100644 --- a/arch/s390/include/asm/eadm.h +++ b/arch/s390/include/asm/eadm.h @@ -102,6 +102,7 @@ struct scm_driver { struct device_driver drv; int (*probe) (struct scm_device *scmdev); int (*remove) (struct scm_device *scmdev); + void (*notify) (struct scm_device *scmdev); void (*handler) (struct scm_device *scmdev, void *data, int error); }; diff --git a/drivers/s390/cio/chsc.c b/drivers/s390/cio/chsc.c index 1218653..4d51a7c 100644 --- a/drivers/s390/cio/chsc.c +++ b/drivers/s390/cio/chsc.c @@ -398,6 +398,20 @@ static void chsc_process_sei_chp_config(struct chsc_sei_area *sei_area) } } +static void chsc_process_sei_scm_change(struct chsc_sei_area *sei_area) +{ + int ret; + + CIO_CRW_EVENT(4, "chsc: scm change notification\n"); + if (sei_area->rs != 7) + return; + + ret = scm_update_information(); + if (ret) + CIO_CRW_EVENT(0, "chsc: updating change notification" + " failed (rc=%d).\n", ret); +} + static void chsc_process_sei(struct chsc_sei_area *sei_area) { /* Check if we might have lost some information. */ @@ -419,6 +433,9 @@ static void chsc_process_sei(struct chsc_sei_area *sei_area) case 8: /* channel-path-configuration notification */ chsc_process_sei_chp_config(sei_area); break; + case 12: /* scm change notification */ + chsc_process_sei_scm_change(sei_area); + break; default: /* other stuff */ CIO_CRW_EVENT(4, "chsc: unhandled sei content code %d\n", sei_area->cc); diff --git a/drivers/s390/cio/chsc.h b/drivers/s390/cio/chsc.h index 3aa94fe..662dab4 100644 --- a/drivers/s390/cio/chsc.h +++ b/drivers/s390/cio/chsc.h @@ -154,4 +154,11 @@ struct chsc_scm_info { int chsc_scm_info(struct chsc_scm_info *scm_area, u64 token); +#ifdef CONFIG_SCM_BUS +int scm_update_information(void); +#else /* CONFIG_SCM_BUS */ +#define scm_update_information() 0 +#endif /* CONFIG_SCM_BUS */ + + #endif diff --git a/drivers/s390/cio/scm.c b/drivers/s390/cio/scm.c index 874b64a..aa4476e 100644 --- a/drivers/s390/cio/scm.c +++ b/drivers/s390/cio/scm.c @@ -196,6 +196,47 @@ static void scmdev_setup(struct scm_device *scmdev, struct sale *sale, spin_lock_init(&scmdev->lock); } +/* + * Check for state-changes, notify the driver and userspace. + */ +static void scmdev_update(struct scm_device *scmdev, struct sale *sale) +{ + struct scm_driver *scmdrv; + bool changed; + + device_lock(&scmdev->dev); + changed = scmdev->attrs.rank != sale->rank || + scmdev->attrs.oper_state != sale->op_state; + scmdev->attrs.rank = sale->rank; + scmdev->attrs.oper_state = sale->op_state; + if (!scmdev->dev.driver) + goto out; + scmdrv = to_scm_drv(scmdev->dev.driver); + if (changed && scmdrv->notify) + scmdrv->notify(scmdev); +out: + device_unlock(&scmdev->dev); + if (changed) + kobject_uevent(&scmdev->dev.kobj, KOBJ_CHANGE); +} + +static int check_address(struct device *dev, void *data) +{ + struct scm_device *scmdev = to_scm_dev(dev); + struct sale *sale = data; + + return scmdev->address == sale->sa; +} + +static struct scm_device *scmdev_find(struct sale *sale) +{ + struct device *dev; + + dev = bus_find_device(&scm_bus_type, NULL, sale, check_address); + + return dev ? to_scm_dev(dev) : NULL; +} + static int scm_add(struct chsc_scm_info *scm_info, size_t num) { struct sale *sale, *scmal = scm_info->scmal; @@ -203,6 +244,13 @@ static int scm_add(struct chsc_scm_info *scm_info, size_t num) int ret; for (sale = scmal; sale < scmal + num; sale++) { + scmdev = scmdev_find(sale); + if (scmdev) { + scmdev_update(scmdev, sale); + /* Release reference from scm_find(). */ + put_device(&scmdev->dev); + continue; + } scmdev = kzalloc(sizeof(*scmdev), GFP_KERNEL); if (!scmdev) return -ENODEV; @@ -218,7 +266,7 @@ static int scm_add(struct chsc_scm_info *scm_info, size_t num) return 0; } -static int scm_update_information(void) +int scm_update_information(void) { struct chsc_scm_info *scm_info; u64 token = 0; -- cgit v0.10.2 From eadb86ab80545d04a0ee576e92ba4447621cdb02 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:48:16 +0200 Subject: s390/cio: add eadm subchannel driver This driver allows usage of EADM subchannels. EADM subchannels act as a communication vehicle for SCM increments. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 3781db2..3f42161 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -440,6 +440,17 @@ config SCM_BUS help Bus driver for Storage Class Memory. +config EADM_SCH + def_tristate m + prompt "Support for EADM subchannels" + depends on SCM_BUS + help + This driver allows usage of EADM subchannels. EADM subchannels act + as a communication vehicle for SCM increments. + + To compile this driver as a module, choose M here: the + module will be called eadm_sch. + endmenu menu "Dump support" diff --git a/arch/s390/include/asm/isc.h b/arch/s390/include/asm/isc.h index 1420a11..5ae6064 100644 --- a/arch/s390/include/asm/isc.h +++ b/arch/s390/include/asm/isc.h @@ -14,6 +14,7 @@ /* Regular I/O interrupts. */ #define IO_SCH_ISC 3 /* regular I/O subchannels */ #define CONSOLE_ISC 1 /* console I/O subchannel */ +#define EADM_SCH_ISC 4 /* EADM subchannels */ #define CHSC_SCH_ISC 7 /* CHSC subchannels */ /* Adapter interrupts. */ #define QDIO_AIRQ_ISC IO_SCH_ISC /* I/O subchannel in qdio mode */ diff --git a/drivers/s390/cio/Makefile b/drivers/s390/cio/Makefile index ac8f210..8c4a386 100644 --- a/drivers/s390/cio/Makefile +++ b/drivers/s390/cio/Makefile @@ -8,6 +8,7 @@ ccw_device-objs += device.o device_fsm.o device_ops.o ccw_device-objs += device_id.o device_pgid.o device_status.o obj-y += ccw_device.o cmf.o obj-$(CONFIG_CHSC_SCH) += chsc_sch.o +obj-$(CONFIG_EADM_SCH) += eadm_sch.o obj-$(CONFIG_SCM_BUS) += scm.o obj-$(CONFIG_CCWGROUP) += ccwgroup.o diff --git a/drivers/s390/cio/eadm_sch.c b/drivers/s390/cio/eadm_sch.c new file mode 100644 index 0000000..3fc882f --- /dev/null +++ b/drivers/s390/cio/eadm_sch.c @@ -0,0 +1,398 @@ +/* + * Driver for s390 eadm subchannels + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott + */ + +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include + +#include "eadm_sch.h" +#include "ioasm.h" +#include "cio.h" +#include "css.h" +#include "orb.h" + +MODULE_DESCRIPTION("driver for s390 eadm subchannels"); +MODULE_LICENSE("GPL"); + +#define EADM_TIMEOUT (5 * HZ) +static DEFINE_SPINLOCK(list_lock); +static LIST_HEAD(eadm_list); + +static debug_info_t *eadm_debug; + +#define EADM_LOG(imp, txt) do { \ + debug_text_event(eadm_debug, imp, txt); \ + } while (0) + +static void EADM_LOG_HEX(int level, void *data, int length) +{ + if (level > eadm_debug->level) + return; + while (length > 0) { + debug_event(eadm_debug, level, data, length); + length -= eadm_debug->buf_size; + data += eadm_debug->buf_size; + } +} + +static void orb_init(union orb *orb) +{ + memset(orb, 0, sizeof(union orb)); + orb->eadm.compat1 = 1; + orb->eadm.compat2 = 1; + orb->eadm.fmt = 1; + orb->eadm.x = 1; +} + +static int eadm_subchannel_start(struct subchannel *sch, struct aob *aob) +{ + union orb *orb = &get_eadm_private(sch)->orb; + int cc; + + orb_init(orb); + orb->eadm.aob = (u32)__pa(aob); + orb->eadm.intparm = (u32)(addr_t)sch; + orb->eadm.key = PAGE_DEFAULT_KEY >> 4; + + EADM_LOG(6, "start"); + EADM_LOG_HEX(6, &sch->schid, sizeof(sch->schid)); + + cc = ssch(sch->schid, orb); + switch (cc) { + case 0: + sch->schib.scsw.eadm.actl |= SCSW_ACTL_START_PEND; + break; + case 1: /* status pending */ + case 2: /* busy */ + return -EBUSY; + case 3: /* not operational */ + return -ENODEV; + } + return 0; +} + +static int eadm_subchannel_clear(struct subchannel *sch) +{ + int cc; + + cc = csch(sch->schid); + if (cc) + return -ENODEV; + + sch->schib.scsw.eadm.actl |= SCSW_ACTL_CLEAR_PEND; + return 0; +} + +static void eadm_subchannel_timeout(unsigned long data) +{ + struct subchannel *sch = (struct subchannel *) data; + + spin_lock_irq(sch->lock); + EADM_LOG(1, "timeout"); + EADM_LOG_HEX(1, &sch->schid, sizeof(sch->schid)); + if (eadm_subchannel_clear(sch)) + EADM_LOG(0, "clear failed"); + spin_unlock_irq(sch->lock); +} + +static void eadm_subchannel_set_timeout(struct subchannel *sch, int expires) +{ + struct eadm_private *private = get_eadm_private(sch); + + if (expires == 0) { + del_timer(&private->timer); + return; + } + if (timer_pending(&private->timer)) { + if (mod_timer(&private->timer, jiffies + expires)) + return; + } + private->timer.function = eadm_subchannel_timeout; + private->timer.data = (unsigned long) sch; + private->timer.expires = jiffies + expires; + add_timer(&private->timer); +} + +static void eadm_subchannel_irq(struct subchannel *sch) +{ + struct eadm_private *private = get_eadm_private(sch); + struct eadm_scsw *scsw = &sch->schib.scsw.eadm; + struct irb *irb = (struct irb *)&S390_lowcore.irb; + int error = 0; + + EADM_LOG(6, "irq"); + EADM_LOG_HEX(6, irb, sizeof(*irb)); + + if ((scsw->stctl & (SCSW_STCTL_ALERT_STATUS | SCSW_STCTL_STATUS_PEND)) + && scsw->eswf == 1 && irb->esw.eadm.erw.r) + error = -EIO; + + if (scsw->fctl & SCSW_FCTL_CLEAR_FUNC) + error = -ETIMEDOUT; + + eadm_subchannel_set_timeout(sch, 0); + + if (private->state != EADM_BUSY) { + EADM_LOG(1, "irq unsol"); + EADM_LOG_HEX(1, irb, sizeof(*irb)); + private->state = EADM_NOT_OPER; + css_sched_sch_todo(sch, SCH_TODO_EVAL); + return; + } + scm_irq_handler((struct aob *)(unsigned long)scsw->aob, error); + private->state = EADM_IDLE; +} + +static struct subchannel *eadm_get_idle_sch(void) +{ + struct eadm_private *private; + struct subchannel *sch; + unsigned long flags; + + spin_lock_irqsave(&list_lock, flags); + list_for_each_entry(private, &eadm_list, head) { + sch = private->sch; + spin_lock(sch->lock); + if (private->state == EADM_IDLE) { + private->state = EADM_BUSY; + list_move_tail(&private->head, &eadm_list); + spin_unlock(sch->lock); + spin_unlock_irqrestore(&list_lock, flags); + + return sch; + } + spin_unlock(sch->lock); + } + spin_unlock_irqrestore(&list_lock, flags); + + return NULL; +} + +static int eadm_start_aob(struct aob *aob) +{ + struct eadm_private *private; + struct subchannel *sch; + unsigned long flags; + int ret; + + sch = eadm_get_idle_sch(); + if (!sch) + return -EBUSY; + + spin_lock_irqsave(sch->lock, flags); + eadm_subchannel_set_timeout(sch, EADM_TIMEOUT); + ret = eadm_subchannel_start(sch, aob); + if (!ret) + goto out_unlock; + + /* Handle start subchannel failure. */ + eadm_subchannel_set_timeout(sch, 0); + private = get_eadm_private(sch); + private->state = EADM_NOT_OPER; + css_sched_sch_todo(sch, SCH_TODO_EVAL); + +out_unlock: + spin_unlock_irqrestore(sch->lock, flags); + + return ret; +} + +static int eadm_subchannel_probe(struct subchannel *sch) +{ + struct eadm_private *private; + int ret; + + private = kzalloc(sizeof(*private), GFP_KERNEL | GFP_DMA); + if (!private) + return -ENOMEM; + + INIT_LIST_HEAD(&private->head); + init_timer(&private->timer); + + spin_lock_irq(sch->lock); + set_eadm_private(sch, private); + private->state = EADM_IDLE; + private->sch = sch; + sch->isc = EADM_SCH_ISC; + ret = cio_enable_subchannel(sch, (u32)(unsigned long)sch); + if (ret) { + set_eadm_private(sch, NULL); + spin_unlock_irq(sch->lock); + kfree(private); + goto out; + } + spin_unlock_irq(sch->lock); + + spin_lock_irq(&list_lock); + list_add(&private->head, &eadm_list); + spin_unlock_irq(&list_lock); + + if (dev_get_uevent_suppress(&sch->dev)) { + dev_set_uevent_suppress(&sch->dev, 0); + kobject_uevent(&sch->dev.kobj, KOBJ_ADD); + } +out: + return ret; +} + +static void eadm_quiesce(struct subchannel *sch) +{ + int ret; + + do { + spin_lock_irq(sch->lock); + ret = cio_disable_subchannel(sch); + spin_unlock_irq(sch->lock); + } while (ret == -EBUSY); +} + +static int eadm_subchannel_remove(struct subchannel *sch) +{ + struct eadm_private *private = get_eadm_private(sch); + + spin_lock_irq(&list_lock); + list_del(&private->head); + spin_unlock_irq(&list_lock); + + eadm_quiesce(sch); + + spin_lock_irq(sch->lock); + set_eadm_private(sch, NULL); + spin_unlock_irq(sch->lock); + + kfree(private); + + return 0; +} + +static void eadm_subchannel_shutdown(struct subchannel *sch) +{ + eadm_quiesce(sch); +} + +static int eadm_subchannel_freeze(struct subchannel *sch) +{ + return cio_disable_subchannel(sch); +} + +static int eadm_subchannel_restore(struct subchannel *sch) +{ + return cio_enable_subchannel(sch, (u32)(unsigned long)sch); +} + +/** + * eadm_subchannel_sch_event - process subchannel event + * @sch: subchannel + * @process: non-zero if function is called in process context + * + * An unspecified event occurred for this subchannel. Adjust data according + * to the current operational state of the subchannel. Return zero when the + * event has been handled sufficiently or -EAGAIN when this function should + * be called again in process context. + */ +static int eadm_subchannel_sch_event(struct subchannel *sch, int process) +{ + struct eadm_private *private; + unsigned long flags; + int ret = 0; + + spin_lock_irqsave(sch->lock, flags); + if (!device_is_registered(&sch->dev)) + goto out_unlock; + + if (work_pending(&sch->todo_work)) + goto out_unlock; + + if (cio_update_schib(sch)) { + css_sched_sch_todo(sch, SCH_TODO_UNREG); + goto out_unlock; + } + private = get_eadm_private(sch); + if (private->state == EADM_NOT_OPER) + private->state = EADM_IDLE; + +out_unlock: + spin_unlock_irqrestore(sch->lock, flags); + + return ret; +} + +static struct css_device_id eadm_subchannel_ids[] = { + { .match_flags = 0x1, .type = SUBCHANNEL_TYPE_ADM, }, + { /* end of list */ }, +}; +MODULE_DEVICE_TABLE(css, eadm_subchannel_ids); + +static struct css_driver eadm_subchannel_driver = { + .drv = { + .name = "eadm_subchannel", + .owner = THIS_MODULE, + }, + .subchannel_type = eadm_subchannel_ids, + .irq = eadm_subchannel_irq, + .probe = eadm_subchannel_probe, + .remove = eadm_subchannel_remove, + .shutdown = eadm_subchannel_shutdown, + .sch_event = eadm_subchannel_sch_event, + .freeze = eadm_subchannel_freeze, + .thaw = eadm_subchannel_restore, + .restore = eadm_subchannel_restore, +}; + +static struct eadm_ops eadm_ops = { + .eadm_start = eadm_start_aob, + .owner = THIS_MODULE, +}; + +static int __init eadm_sch_init(void) +{ + int ret; + + if (!css_general_characteristics.eadm) + return -ENXIO; + + eadm_debug = debug_register("eadm_log", 16, 1, 16); + if (!eadm_debug) + return -ENOMEM; + + debug_register_view(eadm_debug, &debug_hex_ascii_view); + debug_set_level(eadm_debug, 2); + + isc_register(EADM_SCH_ISC); + ret = css_driver_register(&eadm_subchannel_driver); + if (ret) + goto cleanup; + + register_eadm_ops(&eadm_ops); + return ret; + +cleanup: + isc_unregister(EADM_SCH_ISC); + debug_unregister(eadm_debug); + return ret; +} + +static void __exit eadm_sch_exit(void) +{ + unregister_eadm_ops(&eadm_ops); + css_driver_unregister(&eadm_subchannel_driver); + isc_unregister(EADM_SCH_ISC); + debug_unregister(eadm_debug); +} +module_init(eadm_sch_init); +module_exit(eadm_sch_exit); diff --git a/drivers/s390/cio/eadm_sch.h b/drivers/s390/cio/eadm_sch.h new file mode 100644 index 0000000..2779be0 --- /dev/null +++ b/drivers/s390/cio/eadm_sch.h @@ -0,0 +1,20 @@ +#ifndef EADM_SCH_H +#define EADM_SCH_H + +#include +#include +#include +#include "orb.h" + +struct eadm_private { + union orb orb; + enum {EADM_IDLE, EADM_BUSY, EADM_NOT_OPER} state; + struct timer_list timer; + struct list_head head; + struct subchannel *sch; +} __aligned(8); + +#define get_eadm_private(n) ((struct eadm_private *)dev_get_drvdata(&n->dev)) +#define set_eadm_private(n, p) (dev_set_drvdata(&n->dev, p)) + +#endif -- cgit v0.10.2 From 2e73c2cf78f797f3ff299ca39b210bceb40ab804 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:48:47 +0200 Subject: s390/eadm_sch: add support for irq statistics Add support for EADM interrupt statistics in /proc/interrupts. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/irq.h b/arch/s390/include/asm/irq.h index 2b9d418..33cc590 100644 --- a/arch/s390/include/asm/irq.h +++ b/arch/s390/include/asm/irq.h @@ -30,6 +30,7 @@ enum interruption_class { IOINT_CLW, IOINT_CTC, IOINT_APB, + IOINT_ADM, IOINT_CSC, NMI_NMI, NR_IRQS, diff --git a/arch/s390/kernel/irq.c b/arch/s390/kernel/irq.c index dd7630d..a22fdca 100644 --- a/arch/s390/kernel/irq.c +++ b/arch/s390/kernel/irq.c @@ -55,6 +55,7 @@ static const struct irq_class intrclass_names[] = { {.name = "CLW", .desc = "[I/O] CLAW" }, {.name = "CTC", .desc = "[I/O] CTC" }, {.name = "APB", .desc = "[I/O] AP Bus" }, + {.name = "ADM", .desc = "[I/O] EADM Subchannel" }, {.name = "CSC", .desc = "[I/O] CHSC Subchannel" }, {.name = "NMI", .desc = "[NMI] Machine Check" }, }; diff --git a/drivers/s390/cio/eadm_sch.c b/drivers/s390/cio/eadm_sch.c index 3fc882f..6c96734 100644 --- a/drivers/s390/cio/eadm_sch.c +++ b/drivers/s390/cio/eadm_sch.c @@ -5,6 +5,7 @@ * Author(s): Sebastian Ott */ +#include #include #include #include @@ -138,6 +139,8 @@ static void eadm_subchannel_irq(struct subchannel *sch) EADM_LOG(6, "irq"); EADM_LOG_HEX(6, irb, sizeof(*irb)); + kstat_cpu(smp_processor_id()).irqs[IOINT_ADM]++; + if ((scsw->stctl & (SCSW_STCTL_ALERT_STATUS | SCSW_STCTL_STATUS_PEND)) && scsw->eswf == 1 && irb->esw.eadm.erw.r) error = -EIO; -- cgit v0.10.2 From f30664e2c85c7804f07c636bbe99f35e0b2d4c76 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:50:38 +0200 Subject: s390: add scm block driver Block device driver for Storage Class Memory (SCM). This driver provides a block device interface for each available SCM increment. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/block/Kconfig b/drivers/s390/block/Kconfig index 8e477bb..18178b6 100644 --- a/drivers/s390/block/Kconfig +++ b/drivers/s390/block/Kconfig @@ -70,3 +70,14 @@ config DASD_EER This driver provides a character device interface to the DASD extended error reporting. This is only needed if you want to use applications written for the EER facility. + +config SCM_BLOCK + def_tristate m + prompt "Support for Storage Class Memory" + depends on S390 && BLOCK && EADM_SCH && SCM_BUS + help + Block device driver for Storage Class Memory (SCM). This driver + provides a block device interface for each available SCM increment. + + To compile this driver as a module, choose M here: the + module will be called scm_block. diff --git a/drivers/s390/block/Makefile b/drivers/s390/block/Makefile index 0a89e08..b64e2b3 100644 --- a/drivers/s390/block/Makefile +++ b/drivers/s390/block/Makefile @@ -17,3 +17,6 @@ obj-$(CONFIG_DASD_ECKD) += dasd_eckd_mod.o obj-$(CONFIG_DASD_FBA) += dasd_fba_mod.o obj-$(CONFIG_BLK_DEV_XPRAM) += xpram.o obj-$(CONFIG_DCSSBLK) += dcssblk.o + +scm_block-objs := scm_drv.o scm_blk.o +obj-$(CONFIG_SCM_BLOCK) += scm_block.o diff --git a/drivers/s390/block/scm_blk.c b/drivers/s390/block/scm_blk.c new file mode 100644 index 0000000..634ad58 --- /dev/null +++ b/drivers/s390/block/scm_blk.c @@ -0,0 +1,414 @@ +/* + * Block driver for s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott + */ + +#define KMSG_COMPONENT "scm_block" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include +#include +#include +#include +#include +#include +#include +#include +#include "scm_blk.h" + +debug_info_t *scm_debug; +static int scm_major; +static DEFINE_SPINLOCK(list_lock); +static LIST_HEAD(inactive_requests); +static unsigned int nr_requests = 64; +static atomic_t nr_devices = ATOMIC_INIT(0); +module_param(nr_requests, uint, S_IRUGO); +MODULE_PARM_DESC(nr_requests, "Number of parallel requests."); + +MODULE_DESCRIPTION("Block driver for s390 storage class memory."); +MODULE_LICENSE("GPL"); +MODULE_ALIAS("scm:scmdev*"); + +static void __scm_free_rq(struct scm_request *scmrq) +{ + struct aob_rq_header *aobrq = to_aobrq(scmrq); + + free_page((unsigned long) scmrq->aob); + free_page((unsigned long) scmrq->aidaw); + kfree(aobrq); +} + +static void scm_free_rqs(void) +{ + struct list_head *iter, *safe; + struct scm_request *scmrq; + + spin_lock_irq(&list_lock); + list_for_each_safe(iter, safe, &inactive_requests) { + scmrq = list_entry(iter, struct scm_request, list); + list_del(&scmrq->list); + __scm_free_rq(scmrq); + } + spin_unlock_irq(&list_lock); +} + +static int __scm_alloc_rq(void) +{ + struct aob_rq_header *aobrq; + struct scm_request *scmrq; + + aobrq = kzalloc(sizeof(*aobrq) + sizeof(*scmrq), GFP_KERNEL); + if (!aobrq) + return -ENOMEM; + + scmrq = (void *) aobrq->data; + scmrq->aidaw = (void *) get_zeroed_page(GFP_DMA); + scmrq->aob = (void *) get_zeroed_page(GFP_DMA); + if (!scmrq->aob || !scmrq->aidaw) { + __scm_free_rq(scmrq); + return -ENOMEM; + } + INIT_LIST_HEAD(&scmrq->list); + spin_lock_irq(&list_lock); + list_add(&scmrq->list, &inactive_requests); + spin_unlock_irq(&list_lock); + + return 0; +} + +static int scm_alloc_rqs(unsigned int nrqs) +{ + int ret = 0; + + while (nrqs-- && !ret) + ret = __scm_alloc_rq(); + + return ret; +} + +static struct scm_request *scm_request_fetch(void) +{ + struct scm_request *scmrq = NULL; + + spin_lock(&list_lock); + if (list_empty(&inactive_requests)) + goto out; + scmrq = list_first_entry(&inactive_requests, struct scm_request, list); + list_del(&scmrq->list); +out: + spin_unlock(&list_lock); + return scmrq; +} + +static void scm_request_done(struct scm_request *scmrq) +{ + unsigned long flags; + + spin_lock_irqsave(&list_lock, flags); + list_add(&scmrq->list, &inactive_requests); + spin_unlock_irqrestore(&list_lock, flags); +} + +static int scm_open(struct block_device *blkdev, fmode_t mode) +{ + return scm_get_ref(); +} + +static int scm_release(struct gendisk *gendisk, fmode_t mode) +{ + scm_put_ref(); + return 0; +} + +static const struct block_device_operations scm_blk_devops = { + .owner = THIS_MODULE, + .open = scm_open, + .release = scm_release, +}; + +static void scm_request_prepare(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + struct scm_device *scmdev = bdev->gendisk->private_data; + struct aidaw *aidaw = scmrq->aidaw; + struct msb *msb = &scmrq->aob->msb[0]; + struct req_iterator iter; + struct bio_vec *bv; + + msb->bs = MSB_BS_4K; + scmrq->aob->request.msb_count = 1; + msb->scm_addr = scmdev->address + + ((u64) blk_rq_pos(scmrq->request) << 9); + msb->oc = (rq_data_dir(scmrq->request) == READ) ? + MSB_OC_READ : MSB_OC_WRITE; + msb->flags |= MSB_FLAG_IDA; + msb->data_addr = (u64) aidaw; + + rq_for_each_segment(bv, scmrq->request, iter) { + WARN_ON(bv->bv_offset); + msb->blk_count += bv->bv_len >> 12; + aidaw->data_addr = (u64) page_address(bv->bv_page); + aidaw++; + } +} + +static inline void scm_request_init(struct scm_blk_dev *bdev, + struct scm_request *scmrq, + struct request *req) +{ + struct aob_rq_header *aobrq = to_aobrq(scmrq); + struct aob *aob = scmrq->aob; + + memset(aob, 0, sizeof(*aob)); + memset(scmrq->aidaw, 0, PAGE_SIZE); + aobrq->scmdev = bdev->scmdev; + aob->request.cmd_code = ARQB_CMD_MOVE; + aob->request.data = (u64) aobrq; + scmrq->request = req; + scmrq->bdev = bdev; + scmrq->retries = 4; + scmrq->error = 0; +} + +static void scm_ensure_queue_restart(struct scm_blk_dev *bdev) +{ + if (atomic_read(&bdev->queued_reqs)) { + /* Queue restart is triggered by the next interrupt. */ + return; + } + blk_delay_queue(bdev->rq, SCM_QUEUE_DELAY); +} + +static void scm_request_requeue(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + + blk_requeue_request(bdev->rq, scmrq->request); + scm_request_done(scmrq); + scm_ensure_queue_restart(bdev); +} + +static void scm_request_finish(struct scm_request *scmrq) +{ + blk_end_request_all(scmrq->request, scmrq->error); + scm_request_done(scmrq); +} + +static void scm_blk_request(struct request_queue *rq) +{ + struct scm_device *scmdev = rq->queuedata; + struct scm_blk_dev *bdev = dev_get_drvdata(&scmdev->dev); + struct scm_request *scmrq; + struct request *req; + int ret; + + while ((req = blk_peek_request(rq))) { + if (req->cmd_type != REQ_TYPE_FS) + continue; + + scmrq = scm_request_fetch(); + if (!scmrq) { + SCM_LOG(5, "no request"); + scm_ensure_queue_restart(bdev); + return; + } + scm_request_init(bdev, scmrq, req); + scm_request_prepare(scmrq); + blk_start_request(req); + + ret = scm_start_aob(scmrq->aob); + if (ret) { + SCM_LOG(5, "no subchannel"); + scm_request_requeue(scmrq); + return; + } + atomic_inc(&bdev->queued_reqs); + } +} + +static void __scmrq_log_error(struct scm_request *scmrq) +{ + struct aob *aob = scmrq->aob; + + if (scmrq->error == -ETIMEDOUT) + SCM_LOG(1, "Request timeout"); + else { + SCM_LOG(1, "Request error"); + SCM_LOG_HEX(1, &aob->response, sizeof(aob->response)); + } + if (scmrq->retries) + SCM_LOG(1, "Retry request"); + else + pr_err("An I/O operation to SCM failed with rc=%d\n", + scmrq->error); +} + +void scm_blk_irq(struct scm_device *scmdev, void *data, int error) +{ + struct scm_request *scmrq = data; + struct scm_blk_dev *bdev = scmrq->bdev; + + scmrq->error = error; + if (error) + __scmrq_log_error(scmrq); + + spin_lock(&bdev->lock); + list_add_tail(&scmrq->list, &bdev->finished_requests); + spin_unlock(&bdev->lock); + tasklet_hi_schedule(&bdev->tasklet); +} + +static void scm_blk_tasklet(struct scm_blk_dev *bdev) +{ + struct scm_request *scmrq; + unsigned long flags; + + spin_lock_irqsave(&bdev->lock, flags); + while (!list_empty(&bdev->finished_requests)) { + scmrq = list_first_entry(&bdev->finished_requests, + struct scm_request, list); + list_del(&scmrq->list); + spin_unlock_irqrestore(&bdev->lock, flags); + + if (scmrq->error && scmrq->retries-- > 0) { + if (scm_start_aob(scmrq->aob)) { + spin_lock_irqsave(&bdev->rq_lock, flags); + scm_request_requeue(scmrq); + spin_unlock_irqrestore(&bdev->rq_lock, flags); + } + /* Request restarted or requeued, handle next. */ + spin_lock_irqsave(&bdev->lock, flags); + continue; + } + scm_request_finish(scmrq); + atomic_dec(&bdev->queued_reqs); + spin_lock_irqsave(&bdev->lock, flags); + } + spin_unlock_irqrestore(&bdev->lock, flags); + /* Look out for more requests. */ + blk_run_queue(bdev->rq); +} + +int scm_blk_dev_setup(struct scm_blk_dev *bdev, struct scm_device *scmdev) +{ + struct request_queue *rq; + int len, ret = -ENOMEM; + unsigned int devindex, nr_max_blk; + + devindex = atomic_inc_return(&nr_devices) - 1; + /* scma..scmz + scmaa..scmzz */ + if (devindex > 701) { + ret = -ENODEV; + goto out; + } + + bdev->scmdev = scmdev; + spin_lock_init(&bdev->rq_lock); + spin_lock_init(&bdev->lock); + INIT_LIST_HEAD(&bdev->finished_requests); + atomic_set(&bdev->queued_reqs, 0); + tasklet_init(&bdev->tasklet, + (void (*)(unsigned long)) scm_blk_tasklet, + (unsigned long) bdev); + + rq = blk_init_queue(scm_blk_request, &bdev->rq_lock); + if (!rq) + goto out; + + bdev->rq = rq; + nr_max_blk = min(scmdev->nr_max_block, + (unsigned int) (PAGE_SIZE / sizeof(struct aidaw))); + + blk_queue_logical_block_size(rq, 1 << 12); + blk_queue_max_hw_sectors(rq, nr_max_blk << 3); /* 8 * 512 = blk_size */ + blk_queue_max_segments(rq, nr_max_blk); + queue_flag_set_unlocked(QUEUE_FLAG_NONROT, rq); + + bdev->gendisk = alloc_disk(SCM_NR_PARTS); + if (!bdev->gendisk) + goto out_queue; + + rq->queuedata = scmdev; + bdev->gendisk->driverfs_dev = &scmdev->dev; + bdev->gendisk->private_data = scmdev; + bdev->gendisk->fops = &scm_blk_devops; + bdev->gendisk->queue = rq; + bdev->gendisk->major = scm_major; + bdev->gendisk->first_minor = devindex * SCM_NR_PARTS; + + len = snprintf(bdev->gendisk->disk_name, DISK_NAME_LEN, "scm"); + if (devindex > 25) { + len += snprintf(bdev->gendisk->disk_name + len, + DISK_NAME_LEN - len, "%c", + 'a' + (devindex / 26) - 1); + devindex = devindex % 26; + } + snprintf(bdev->gendisk->disk_name + len, DISK_NAME_LEN - len, "%c", + 'a' + devindex); + + /* 512 byte sectors */ + set_capacity(bdev->gendisk, scmdev->size >> 9); + add_disk(bdev->gendisk); + return 0; + +out_queue: + blk_cleanup_queue(rq); +out: + atomic_dec(&nr_devices); + return ret; +} + +void scm_blk_dev_cleanup(struct scm_blk_dev *bdev) +{ + tasklet_kill(&bdev->tasklet); + del_gendisk(bdev->gendisk); + blk_cleanup_queue(bdev->gendisk->queue); + put_disk(bdev->gendisk); +} + +static int __init scm_blk_init(void) +{ + int ret; + + ret = register_blkdev(0, "scm"); + if (ret < 0) + goto out; + + scm_major = ret; + if (scm_alloc_rqs(nr_requests)) + goto out_unreg; + + scm_debug = debug_register("scm_log", 16, 1, 16); + if (!scm_debug) + goto out_free; + + debug_register_view(scm_debug, &debug_hex_ascii_view); + debug_set_level(scm_debug, 2); + + ret = scm_drv_init(); + if (ret) + goto out_dbf; + + return ret; + +out_dbf: + debug_unregister(scm_debug); +out_free: + scm_free_rqs(); +out_unreg: + unregister_blkdev(scm_major, "scm"); +out: + return ret; +} +module_init(scm_blk_init); + +static void __exit scm_blk_cleanup(void) +{ + scm_drv_cleanup(); + debug_unregister(scm_debug); + scm_free_rqs(); + unregister_blkdev(scm_major, "scm"); +} +module_exit(scm_blk_cleanup); diff --git a/drivers/s390/block/scm_blk.h b/drivers/s390/block/scm_blk.h new file mode 100644 index 0000000..5aba561 --- /dev/null +++ b/drivers/s390/block/scm_blk.h @@ -0,0 +1,79 @@ +#ifndef SCM_BLK_H +#define SCM_BLK_H + +#include +#include +#include +#include +#include + +#include +#include + +#define SCM_NR_PARTS 8 +#define SCM_QUEUE_DELAY 5 + +struct scm_blk_dev { + struct tasklet_struct tasklet; + struct request_queue *rq; + struct gendisk *gendisk; + struct scm_device *scmdev; + spinlock_t rq_lock; /* guard the request queue */ + spinlock_t lock; /* guard the rest of the blockdev */ + atomic_t queued_reqs; + struct list_head finished_requests; +}; + +struct scm_request { + struct scm_blk_dev *bdev; + struct request *request; + struct aidaw *aidaw; + struct aob *aob; + struct list_head list; + u8 retries; + int error; +}; + +#define to_aobrq(rq) container_of((void *) rq, struct aob_rq_header, data) + +int scm_blk_dev_setup(struct scm_blk_dev *, struct scm_device *); +void scm_blk_dev_cleanup(struct scm_blk_dev *); +void scm_blk_irq(struct scm_device *, void *, int); + +int scm_drv_init(void); +void scm_drv_cleanup(void); + + +extern debug_info_t *scm_debug; + +#define SCM_LOG(imp, txt) do { \ + debug_text_event(scm_debug, imp, txt); \ + } while (0) + +static inline void SCM_LOG_HEX(int level, void *data, int length) +{ + if (level > scm_debug->level) + return; + while (length > 0) { + debug_event(scm_debug, level, data, length); + length -= scm_debug->buf_size; + data += scm_debug->buf_size; + } +} + +static inline void SCM_LOG_STATE(int level, struct scm_device *scmdev) +{ + struct { + u64 address; + u8 oper_state; + u8 rank; + } __packed data = { + .address = scmdev->address, + .oper_state = scmdev->attrs.oper_state, + .rank = scmdev->attrs.rank, + }; + + SCM_LOG_HEX(level, &data, sizeof(data)); +} + +#endif /* SCM_BLK_H */ diff --git a/drivers/s390/block/scm_drv.c b/drivers/s390/block/scm_drv.c new file mode 100644 index 0000000..fce711a --- /dev/null +++ b/drivers/s390/block/scm_drv.c @@ -0,0 +1,90 @@ +/* + * Device driver for s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott + */ + +#define KMSG_COMPONENT "scm_block" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include +#include +#include +#include +#include "scm_blk.h" + +static void notify(struct scm_device *scmdev) +{ + pr_info("%lu: The capabilities of the SCM increment changed\n", + (unsigned long) scmdev->address); + SCM_LOG(2, "State changed"); + SCM_LOG_STATE(2, scmdev); +} + +static int scm_probe(struct scm_device *scmdev) +{ + struct scm_blk_dev *bdev; + int ret; + + SCM_LOG(2, "probe"); + SCM_LOG_STATE(2, scmdev); + + if (scmdev->attrs.oper_state != OP_STATE_GOOD) + return -EINVAL; + + bdev = kzalloc(sizeof(*bdev), GFP_KERNEL); + if (!bdev) + return -ENOMEM; + + spin_lock_irq(&scmdev->lock); + dev_set_drvdata(&scmdev->dev, bdev); + spin_unlock_irq(&scmdev->lock); + + ret = scm_blk_dev_setup(bdev, scmdev); + if (ret) { + spin_lock_irq(&scmdev->lock); + dev_set_drvdata(&scmdev->dev, NULL); + spin_unlock_irq(&scmdev->lock); + kfree(bdev); + goto out; + } + +out: + return ret; +} + +static int scm_remove(struct scm_device *scmdev) +{ + struct scm_blk_dev *bdev; + + spin_lock_irq(&scmdev->lock); + bdev = dev_get_drvdata(&scmdev->dev); + dev_set_drvdata(&scmdev->dev, NULL); + spin_unlock_irq(&scmdev->lock); + scm_blk_dev_cleanup(bdev); + kfree(bdev); + + return 0; +} + +static struct scm_driver scm_drv = { + .drv = { + .name = "scm_block", + .owner = THIS_MODULE, + }, + .notify = notify, + .probe = scm_probe, + .remove = scm_remove, + .handler = scm_blk_irq, +}; + +int __init scm_drv_init(void) +{ + return scm_driver_register(&scm_drv); +} + +void scm_drv_cleanup(void) +{ + scm_driver_unregister(&scm_drv); +} -- cgit v0.10.2 From 0d804b20735d974534abb422f723d404b779433a Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 28 Aug 2012 16:51:19 +0200 Subject: s390/scm_block: force cluster writes Force writes to Storage Class Memory (SCM) to be in done in clusters. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/block/Kconfig b/drivers/s390/block/Kconfig index 18178b6..4a3b623 100644 --- a/drivers/s390/block/Kconfig +++ b/drivers/s390/block/Kconfig @@ -81,3 +81,10 @@ config SCM_BLOCK To compile this driver as a module, choose M here: the module will be called scm_block. + +config SCM_BLOCK_CLUSTER_WRITE + def_bool y + prompt "SCM force cluster writes" + depends on SCM_BLOCK + help + Force writes to Storage Class Memory (SCM) to be in done in clusters. diff --git a/drivers/s390/block/Makefile b/drivers/s390/block/Makefile index b64e2b3..c2f4e67 100644 --- a/drivers/s390/block/Makefile +++ b/drivers/s390/block/Makefile @@ -19,4 +19,7 @@ obj-$(CONFIG_BLK_DEV_XPRAM) += xpram.o obj-$(CONFIG_DCSSBLK) += dcssblk.o scm_block-objs := scm_drv.o scm_blk.o +ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE +scm_block-objs += scm_blk_cluster.o +endif obj-$(CONFIG_SCM_BLOCK) += scm_block.o diff --git a/drivers/s390/block/scm_blk.c b/drivers/s390/block/scm_blk.c index 634ad58..9978ad4 100644 --- a/drivers/s390/block/scm_blk.c +++ b/drivers/s390/block/scm_blk.c @@ -37,6 +37,7 @@ static void __scm_free_rq(struct scm_request *scmrq) free_page((unsigned long) scmrq->aob); free_page((unsigned long) scmrq->aidaw); + __scm_free_rq_cluster(scmrq); kfree(aobrq); } @@ -70,6 +71,12 @@ static int __scm_alloc_rq(void) __scm_free_rq(scmrq); return -ENOMEM; } + + if (__scm_alloc_rq_cluster(scmrq)) { + __scm_free_rq(scmrq); + return -ENOMEM; + } + INIT_LIST_HEAD(&scmrq->list); spin_lock_irq(&list_lock); list_add(&scmrq->list, &inactive_requests); @@ -170,6 +177,7 @@ static inline void scm_request_init(struct scm_blk_dev *bdev, scmrq->bdev = bdev; scmrq->retries = 4; scmrq->error = 0; + scm_request_cluster_init(scmrq); } static void scm_ensure_queue_restart(struct scm_blk_dev *bdev) @@ -181,17 +189,19 @@ static void scm_ensure_queue_restart(struct scm_blk_dev *bdev) blk_delay_queue(bdev->rq, SCM_QUEUE_DELAY); } -static void scm_request_requeue(struct scm_request *scmrq) +void scm_request_requeue(struct scm_request *scmrq) { struct scm_blk_dev *bdev = scmrq->bdev; + scm_release_cluster(scmrq); blk_requeue_request(bdev->rq, scmrq->request); scm_request_done(scmrq); scm_ensure_queue_restart(bdev); } -static void scm_request_finish(struct scm_request *scmrq) +void scm_request_finish(struct scm_request *scmrq) { + scm_release_cluster(scmrq); blk_end_request_all(scmrq->request, scmrq->error); scm_request_done(scmrq); } @@ -215,6 +225,16 @@ static void scm_blk_request(struct request_queue *rq) return; } scm_request_init(bdev, scmrq, req); + if (!scm_reserve_cluster(scmrq)) { + SCM_LOG(5, "cluster busy"); + scm_request_done(scmrq); + return; + } + if (scm_need_cluster_request(scmrq)) { + blk_start_request(req); + scm_initiate_cluster_request(scmrq); + return; + } scm_request_prepare(scmrq); blk_start_request(req); @@ -282,6 +302,13 @@ static void scm_blk_tasklet(struct scm_blk_dev *bdev) spin_lock_irqsave(&bdev->lock, flags); continue; } + + if (scm_test_cluster_request(scmrq)) { + scm_cluster_request_irq(scmrq); + spin_lock_irqsave(&bdev->lock, flags); + continue; + } + scm_request_finish(scmrq); atomic_dec(&bdev->queued_reqs); spin_lock_irqsave(&bdev->lock, flags); @@ -325,6 +352,7 @@ int scm_blk_dev_setup(struct scm_blk_dev *bdev, struct scm_device *scmdev) blk_queue_max_hw_sectors(rq, nr_max_blk << 3); /* 8 * 512 = blk_size */ blk_queue_max_segments(rq, nr_max_blk); queue_flag_set_unlocked(QUEUE_FLAG_NONROT, rq); + scm_blk_dev_cluster_setup(bdev); bdev->gendisk = alloc_disk(SCM_NR_PARTS); if (!bdev->gendisk) @@ -370,7 +398,10 @@ void scm_blk_dev_cleanup(struct scm_blk_dev *bdev) static int __init scm_blk_init(void) { - int ret; + int ret = -EINVAL; + + if (!scm_cluster_size_valid()) + goto out; ret = register_blkdev(0, "scm"); if (ret < 0) diff --git a/drivers/s390/block/scm_blk.h b/drivers/s390/block/scm_blk.h index 5aba561..7ac6bad 100644 --- a/drivers/s390/block/scm_blk.h +++ b/drivers/s390/block/scm_blk.h @@ -22,6 +22,9 @@ struct scm_blk_dev { spinlock_t lock; /* guard the rest of the blockdev */ atomic_t queued_reqs; struct list_head finished_requests; +#ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE + struct list_head cluster_list; +#endif }; struct scm_request { @@ -32,6 +35,13 @@ struct scm_request { struct list_head list; u8 retries; int error; +#ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE + struct { + enum {CLUSTER_NONE, CLUSTER_READ, CLUSTER_WRITE} state; + struct list_head list; + void **buf; + } cluster; +#endif }; #define to_aobrq(rq) container_of((void *) rq, struct aob_rq_header, data) @@ -40,9 +50,37 @@ int scm_blk_dev_setup(struct scm_blk_dev *, struct scm_device *); void scm_blk_dev_cleanup(struct scm_blk_dev *); void scm_blk_irq(struct scm_device *, void *, int); +void scm_request_finish(struct scm_request *); +void scm_request_requeue(struct scm_request *); + int scm_drv_init(void); void scm_drv_cleanup(void); +#ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE +void __scm_free_rq_cluster(struct scm_request *); +int __scm_alloc_rq_cluster(struct scm_request *); +void scm_request_cluster_init(struct scm_request *); +bool scm_reserve_cluster(struct scm_request *); +void scm_release_cluster(struct scm_request *); +void scm_blk_dev_cluster_setup(struct scm_blk_dev *); +bool scm_need_cluster_request(struct scm_request *); +void scm_initiate_cluster_request(struct scm_request *); +void scm_cluster_request_irq(struct scm_request *); +bool scm_test_cluster_request(struct scm_request *); +bool scm_cluster_size_valid(void); +#else +#define __scm_free_rq_cluster(scmrq) {} +#define __scm_alloc_rq_cluster(scmrq) 0 +#define scm_request_cluster_init(scmrq) {} +#define scm_reserve_cluster(scmrq) true +#define scm_release_cluster(scmrq) {} +#define scm_blk_dev_cluster_setup(bdev) {} +#define scm_need_cluster_request(scmrq) false +#define scm_initiate_cluster_request(scmrq) {} +#define scm_cluster_request_irq(scmrq) {} +#define scm_test_cluster_request(scmrq) false +#define scm_cluster_size_valid() true +#endif extern debug_info_t *scm_debug; diff --git a/drivers/s390/block/scm_blk_cluster.c b/drivers/s390/block/scm_blk_cluster.c new file mode 100644 index 0000000..f4bb61b --- /dev/null +++ b/drivers/s390/block/scm_blk_cluster.c @@ -0,0 +1,228 @@ +/* + * Block driver for s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott + */ + +#include +#include +#include +#include +#include +#include +#include +#include "scm_blk.h" + +static unsigned int write_cluster_size = 64; +module_param(write_cluster_size, uint, S_IRUGO); +MODULE_PARM_DESC(write_cluster_size, + "Number of pages used for contiguous writes."); + +#define CLUSTER_SIZE (write_cluster_size * PAGE_SIZE) + +void __scm_free_rq_cluster(struct scm_request *scmrq) +{ + int i; + + if (!scmrq->cluster.buf) + return; + + for (i = 0; i < 2 * write_cluster_size; i++) + free_page((unsigned long) scmrq->cluster.buf[i]); + + kfree(scmrq->cluster.buf); +} + +int __scm_alloc_rq_cluster(struct scm_request *scmrq) +{ + int i; + + scmrq->cluster.buf = kzalloc(sizeof(void *) * 2 * write_cluster_size, + GFP_KERNEL); + if (!scmrq->cluster.buf) + return -ENOMEM; + + for (i = 0; i < 2 * write_cluster_size; i++) { + scmrq->cluster.buf[i] = (void *) get_zeroed_page(GFP_DMA); + if (!scmrq->cluster.buf[i]) + return -ENOMEM; + } + INIT_LIST_HEAD(&scmrq->cluster.list); + return 0; +} + +void scm_request_cluster_init(struct scm_request *scmrq) +{ + scmrq->cluster.state = CLUSTER_NONE; +} + +static bool clusters_intersect(struct scm_request *A, struct scm_request *B) +{ + unsigned long firstA, lastA, firstB, lastB; + + firstA = ((u64) blk_rq_pos(A->request) << 9) / CLUSTER_SIZE; + lastA = (((u64) blk_rq_pos(A->request) << 9) + + blk_rq_bytes(A->request) - 1) / CLUSTER_SIZE; + + firstB = ((u64) blk_rq_pos(B->request) << 9) / CLUSTER_SIZE; + lastB = (((u64) blk_rq_pos(B->request) << 9) + + blk_rq_bytes(B->request) - 1) / CLUSTER_SIZE; + + return (firstB <= lastA && firstA <= lastB); +} + +bool scm_reserve_cluster(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + struct scm_request *iter; + + if (write_cluster_size == 0) + return true; + + spin_lock(&bdev->lock); + list_for_each_entry(iter, &bdev->cluster_list, cluster.list) { + if (clusters_intersect(scmrq, iter) && + (rq_data_dir(scmrq->request) == WRITE || + rq_data_dir(iter->request) == WRITE)) { + spin_unlock(&bdev->lock); + return false; + } + } + list_add(&scmrq->cluster.list, &bdev->cluster_list); + spin_unlock(&bdev->lock); + + return true; +} + +void scm_release_cluster(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + unsigned long flags; + + if (write_cluster_size == 0) + return; + + spin_lock_irqsave(&bdev->lock, flags); + list_del(&scmrq->cluster.list); + spin_unlock_irqrestore(&bdev->lock, flags); +} + +void scm_blk_dev_cluster_setup(struct scm_blk_dev *bdev) +{ + INIT_LIST_HEAD(&bdev->cluster_list); + blk_queue_io_opt(bdev->rq, CLUSTER_SIZE); +} + +static void scm_prepare_cluster_request(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + struct scm_device *scmdev = bdev->gendisk->private_data; + struct request *req = scmrq->request; + struct aidaw *aidaw = scmrq->aidaw; + struct msb *msb = &scmrq->aob->msb[0]; + struct req_iterator iter; + struct bio_vec *bv; + int i = 0; + u64 addr; + + switch (scmrq->cluster.state) { + case CLUSTER_NONE: + scmrq->cluster.state = CLUSTER_READ; + /* fall through */ + case CLUSTER_READ: + scmrq->aob->request.msb_count = 1; + msb->bs = MSB_BS_4K; + msb->oc = MSB_OC_READ; + msb->flags = MSB_FLAG_IDA; + msb->data_addr = (u64) aidaw; + msb->blk_count = write_cluster_size; + + addr = scmdev->address + ((u64) blk_rq_pos(req) << 9); + msb->scm_addr = round_down(addr, CLUSTER_SIZE); + + if (msb->scm_addr != + round_down(addr + (u64) blk_rq_bytes(req) - 1, + CLUSTER_SIZE)) + msb->blk_count = 2 * write_cluster_size; + + for (i = 0; i < msb->blk_count; i++) { + aidaw->data_addr = (u64) scmrq->cluster.buf[i]; + aidaw++; + } + + break; + case CLUSTER_WRITE: + msb->oc = MSB_OC_WRITE; + + for (addr = msb->scm_addr; + addr < scmdev->address + ((u64) blk_rq_pos(req) << 9); + addr += PAGE_SIZE) { + aidaw->data_addr = (u64) scmrq->cluster.buf[i]; + aidaw++; + i++; + } + rq_for_each_segment(bv, req, iter) { + aidaw->data_addr = (u64) page_address(bv->bv_page); + aidaw++; + i++; + } + for (; i < msb->blk_count; i++) { + aidaw->data_addr = (u64) scmrq->cluster.buf[i]; + aidaw++; + } + break; + } +} + +bool scm_need_cluster_request(struct scm_request *scmrq) +{ + if (rq_data_dir(scmrq->request) == READ) + return false; + + return blk_rq_bytes(scmrq->request) < CLUSTER_SIZE; +} + +/* Called with queue lock held. */ +void scm_initiate_cluster_request(struct scm_request *scmrq) +{ + scm_prepare_cluster_request(scmrq); + if (scm_start_aob(scmrq->aob)) + scm_request_requeue(scmrq); +} + +bool scm_test_cluster_request(struct scm_request *scmrq) +{ + return scmrq->cluster.state != CLUSTER_NONE; +} + +void scm_cluster_request_irq(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + unsigned long flags; + + switch (scmrq->cluster.state) { + case CLUSTER_NONE: + BUG(); + break; + case CLUSTER_READ: + if (scmrq->error) { + scm_request_finish(scmrq); + break; + } + scmrq->cluster.state = CLUSTER_WRITE; + spin_lock_irqsave(&bdev->rq_lock, flags); + scm_initiate_cluster_request(scmrq); + spin_unlock_irqrestore(&bdev->rq_lock, flags); + break; + case CLUSTER_WRITE: + scm_request_finish(scmrq); + break; + } +} + +bool scm_cluster_size_valid(void) +{ + return write_cluster_size == 0 || write_cluster_size == 32 || + write_cluster_size == 64 || write_cluster_size == 128; +} -- cgit v0.10.2 From e4b8b3f33fcaa0ed6e6b5482a606091d8cd20beb Mon Sep 17 00:00:00 2001 From: Jan Glauber Date: Tue, 31 Jul 2012 10:52:05 +0200 Subject: s390: add support for runtime instrumentation Allow user-space threads to use runtime instrumentation (RI). To enable RI for a thread there is a new s390 specific system call, sys_s390_runtime_instr, that takes as parameter a realtime signal number. If the RI facility is available the system call sets up a control block for the calling thread with the appropriate permissions for the thread to modify the control block. The user-space thread can then use the store and modify RI instructions to alter the control block and start/stop the instrumentation via RION/RIOFF. If the user specified program buffer runs full RI triggers an external interrupt. The external interrupt is translated to a real-time signal that is delivered to the thread that enabled RI on that CPU. The number of the real-time signal is the number specified in the RI system call. So, user-space can select any available real-time signal number in case the application itself uses real-time signals for other purposes. The kernel saves the RI control blocks on task switch only if the running thread was enabled for RI. Therefore, the performance impact on task switch should be negligible if RI is not used. RI is only enabled for user-space mode and is disabled for the supervisor state. Reviewed-by: Heiko Carstens Signed-off-by: Jan Glauber Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/cpu_mf.h b/arch/s390/include/asm/cpu_mf.h index a3afecd..35f0020 100644 --- a/arch/s390/include/asm/cpu_mf.h +++ b/arch/s390/include/asm/cpu_mf.h @@ -21,11 +21,15 @@ #define CPU_MF_INT_SF_LSDA (1 << 22) /* loss of sample data alert */ #define CPU_MF_INT_CF_CACA (1 << 7) /* counter auth. change alert */ #define CPU_MF_INT_CF_LCDA (1 << 6) /* loss of counter data alert */ +#define CPU_MF_INT_RI_HALTED (1 << 5) /* run-time instr. halted */ +#define CPU_MF_INT_RI_BUF_FULL (1 << 4) /* run-time instr. program + buffer full */ #define CPU_MF_INT_CF_MASK (CPU_MF_INT_CF_CACA|CPU_MF_INT_CF_LCDA) #define CPU_MF_INT_SF_MASK (CPU_MF_INT_SF_IAE|CPU_MF_INT_SF_ISE| \ CPU_MF_INT_SF_PRA|CPU_MF_INT_SF_SACA| \ CPU_MF_INT_SF_LSDA) +#define CPU_MF_INT_RI_MASK (CPU_MF_INT_RI_HALTED|CPU_MF_INT_RI_BUF_FULL) /* CPU measurement facility support */ static inline int cpum_cf_avail(void) diff --git a/arch/s390/include/asm/irq.h b/arch/s390/include/asm/irq.h index 33cc590..6703dd9 100644 --- a/arch/s390/include/asm/irq.h +++ b/arch/s390/include/asm/irq.h @@ -19,6 +19,7 @@ enum interruption_class { EXTINT_IUC, EXTINT_CMS, EXTINT_CMC, + EXTINT_CMR, IOINT_CIO, IOINT_QAI, IOINT_DAS, diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h index d4477ba..0fff583 100644 --- a/arch/s390/include/asm/processor.h +++ b/arch/s390/include/asm/processor.h @@ -17,6 +17,7 @@ #include #include #include +#include /* * Default implementation of macro that returns current @@ -78,6 +79,9 @@ struct thread_struct { /* pfault_wait is used to block the process on a pfault event */ unsigned long pfault_wait; struct list_head list; + /* cpu runtime instrumentation */ + struct runtime_instr_cb *ri_cb; + int ri_signum; }; typedef struct thread_struct thread_struct; diff --git a/arch/s390/include/asm/ptrace.h b/arch/s390/include/asm/ptrace.h index d5f08ea..5c32bae 100644 --- a/arch/s390/include/asm/ptrace.h +++ b/arch/s390/include/asm/ptrace.h @@ -235,6 +235,7 @@ typedef struct #define PSW_MASK_ASC 0x0000C000UL #define PSW_MASK_CC 0x00003000UL #define PSW_MASK_PM 0x00000F00UL +#define PSW_MASK_RI 0x00000000UL #define PSW_MASK_EA 0x00000000UL #define PSW_MASK_BA 0x00000000UL @@ -264,10 +265,11 @@ typedef struct #define PSW_MASK_ASC 0x0000C00000000000UL #define PSW_MASK_CC 0x0000300000000000UL #define PSW_MASK_PM 0x00000F0000000000UL +#define PSW_MASK_RI 0x0000008000000000UL #define PSW_MASK_EA 0x0000000100000000UL #define PSW_MASK_BA 0x0000000080000000UL -#define PSW_MASK_USER 0x00003F0180000000UL +#define PSW_MASK_USER 0x00003F8180000000UL #define PSW_ADDR_AMODE 0x0000000000000000UL #define PSW_ADDR_INSN 0xFFFFFFFFFFFFFFFFUL diff --git a/arch/s390/include/asm/runtime_instr.h b/arch/s390/include/asm/runtime_instr.h new file mode 100644 index 0000000..830da73 --- /dev/null +++ b/arch/s390/include/asm/runtime_instr.h @@ -0,0 +1,98 @@ +#ifndef _RUNTIME_INSTR_H +#define _RUNTIME_INSTR_H + +#define S390_RUNTIME_INSTR_START 0x1 +#define S390_RUNTIME_INSTR_STOP 0x2 + +struct runtime_instr_cb { + __u64 buf_current; + __u64 buf_origin; + __u64 buf_limit; + + __u32 valid : 1; + __u32 pstate : 1; + __u32 pstate_set_buf : 1; + __u32 home_space : 1; + __u32 altered : 1; + __u32 : 3; + __u32 pstate_sample : 1; + __u32 sstate_sample : 1; + __u32 pstate_collect : 1; + __u32 sstate_collect : 1; + __u32 : 1; + __u32 halted_int : 1; + __u32 int_requested : 1; + __u32 buffer_full_int : 1; + __u32 key : 4; + __u32 : 9; + __u32 rgs : 3; + + __u32 mode : 4; + __u32 next : 1; + __u32 mae : 1; + __u32 : 2; + __u32 call_type_br : 1; + __u32 return_type_br : 1; + __u32 other_type_br : 1; + __u32 bc_other_type : 1; + __u32 emit : 1; + __u32 tx_abort : 1; + __u32 : 2; + __u32 bp_xn : 1; + __u32 bp_xt : 1; + __u32 bp_ti : 1; + __u32 bp_ni : 1; + __u32 suppr_y : 1; + __u32 suppr_z : 1; + + __u32 dc_miss_extra : 1; + __u32 lat_lev_ignore : 1; + __u32 ic_lat_lev : 4; + __u32 dc_lat_lev : 4; + + __u64 reserved1; + __u64 scaling_factor; + __u64 rsic; + __u64 reserved2; +} __packed __aligned(8); + +extern struct runtime_instr_cb runtime_instr_empty_cb; + +static inline void load_runtime_instr_cb(struct runtime_instr_cb *cb) +{ + asm volatile(".insn rsy,0xeb0000000060,0,0,%0" /* LRIC */ + : : "Q" (*cb)); +} + +static inline void store_runtime_instr_cb(struct runtime_instr_cb *cb) +{ + asm volatile(".insn rsy,0xeb0000000061,0,0,%0" /* STRIC */ + : "=Q" (*cb) : : "cc"); +} + +static inline void save_ri_cb(struct runtime_instr_cb *cb_prev) +{ +#ifdef CONFIG_64BIT + if (cb_prev) + store_runtime_instr_cb(cb_prev); +#endif +} + +static inline void restore_ri_cb(struct runtime_instr_cb *cb_next, + struct runtime_instr_cb *cb_prev) +{ +#ifdef CONFIG_64BIT + if (cb_next) + load_runtime_instr_cb(cb_next); + else if (cb_prev) + load_runtime_instr_cb(&runtime_instr_empty_cb); +#endif +} + +#ifdef CONFIG_64BIT +extern void exit_thread_runtime_instr(void); +#else +static inline void exit_thread_runtime_instr(void) { } +#endif + +#endif /* _RUNTIME_INSTR_H */ diff --git a/arch/s390/include/asm/switch_to.h b/arch/s390/include/asm/switch_to.h index f223068..dc4967b 100644 --- a/arch/s390/include/asm/switch_to.h +++ b/arch/s390/include/asm/switch_to.h @@ -80,10 +80,12 @@ static inline void restore_access_regs(unsigned int *acrs) if (prev->mm) { \ save_fp_regs(&prev->thread.fp_regs); \ save_access_regs(&prev->thread.acrs[0]); \ + save_ri_cb(prev->thread.ri_cb); \ } \ if (next->mm) { \ restore_fp_regs(&next->thread.fp_regs); \ restore_access_regs(&next->thread.acrs[0]); \ + restore_ri_cb(next->thread.ri_cb, prev->thread.ri_cb); \ update_per_regs(next); \ } \ prev = __switch_to(prev,next); \ diff --git a/arch/s390/include/asm/unistd.h b/arch/s390/include/asm/unistd.h index 6756e78..da2c2f4 100644 --- a/arch/s390/include/asm/unistd.h +++ b/arch/s390/include/asm/unistd.h @@ -277,7 +277,8 @@ #define __NR_setns 339 #define __NR_process_vm_readv 340 #define __NR_process_vm_writev 341 -#define NR_syscalls 342 +#define __NR_s390_runtime_instr 342 +#define NR_syscalls 343 /* * There are some system calls that are not present on 64 bit, some diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile index 9a111b6..bba01bf 100644 --- a/arch/s390/kernel/Makefile +++ b/arch/s390/kernel/Makefile @@ -49,7 +49,7 @@ obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o obj-$(CONFIG_FTRACE_SYSCALLS) += ftrace.o obj-$(CONFIG_CRASH_DUMP) += crash_dump.o obj-$(CONFIG_PERF_EVENTS) += perf_event.o perf_cpum_cf.o -obj-$(CONFIG_64BIT) += cache.o +obj-$(CONFIG_64BIT) += runtime_instr.o cache.o # Kexec part S390_KEXEC_OBJS := machine_kexec.o diff --git a/arch/s390/kernel/compat_wrapper.S b/arch/s390/kernel/compat_wrapper.S index 2d82cfc..5f7928f 100644 --- a/arch/s390/kernel/compat_wrapper.S +++ b/arch/s390/kernel/compat_wrapper.S @@ -1646,3 +1646,8 @@ ENTRY(compat_sys_process_vm_writev_wrapper) llgf %r0,164(%r15) # unsigned long stg %r0,160(%r15) jg compat_sys_process_vm_writev + +ENTRY(sys_s390_runtime_instr_wrapper) + lgfr %r2,%r2 # int + lgfr %r3,%r3 # int + jg sys_s390_runtime_instr diff --git a/arch/s390/kernel/dis.c b/arch/s390/kernel/dis.c index 84fd7e9..4bc67db 100644 --- a/arch/s390/kernel/dis.c +++ b/arch/s390/kernel/dis.c @@ -315,6 +315,8 @@ enum { LONG_INSN_POPCNT, LONG_INSN_RISBHG, LONG_INSN_RISBLG, + LONG_INSN_RINEXT, + LONG_INSN_RIEMIT, }; static char *long_insn_name[] = { @@ -330,6 +332,8 @@ static char *long_insn_name[] = { [LONG_INSN_POPCNT] = "popcnt", [LONG_INSN_RISBHG] = "risbhg", [LONG_INSN_RISBLG] = "risblk", + [LONG_INSN_RINEXT] = "rinext", + [LONG_INSN_RIEMIT] = "riemit", }; static struct insn opcode[] = { @@ -582,6 +586,17 @@ static struct insn opcode_a7[] = { { "", 0, INSTR_INVALID } }; +static struct insn opcode_aa[] = { +#ifdef CONFIG_64BIT + { { 0, LONG_INSN_RINEXT }, 0x00, INSTR_RI_RI }, + { "rion", 0x01, INSTR_RI_RI }, + { "tric", 0x02, INSTR_RI_RI }, + { "rioff", 0x03, INSTR_RI_RI }, + { { 0, LONG_INSN_RIEMIT }, 0x04, INSTR_RI_RI }, +#endif + { "", 0, INSTR_INVALID } +}; + static struct insn opcode_b2[] = { #ifdef CONFIG_64BIT { "sske", 0x2b, INSTR_RRF_M0RR }, @@ -1210,6 +1225,9 @@ static struct insn opcode_eb[] = { { "cliy", 0x55, INSTR_SIY_URD }, { "oiy", 0x56, INSTR_SIY_URD }, { "xiy", 0x57, INSTR_SIY_URD }, + { "lric", 0x60, INSTR_RSY_RDRM }, + { "stric", 0x61, INSTR_RSY_RDRM }, + { "mric", 0x62, INSTR_RSY_RDRM }, { "icmh", 0x80, INSTR_RSE_RURD }, { "icmh", 0x80, INSTR_RSY_RURD }, { "icmy", 0x81, INSTR_RSY_RURD }, @@ -1408,6 +1426,9 @@ static struct insn *find_insn(unsigned char *code) case 0xa7: table = opcode_a7; break; + case 0xaa: + table = opcode_aa; + break; case 0xb2: table = opcode_b2; break; diff --git a/arch/s390/kernel/irq.c b/arch/s390/kernel/irq.c index a22fdca..c1b4493 100644 --- a/arch/s390/kernel/irq.c +++ b/arch/s390/kernel/irq.c @@ -44,6 +44,7 @@ static const struct irq_class intrclass_names[] = { {.name = "IUC", .desc = "[EXT] IUCV" }, {.name = "CMS", .desc = "[EXT] CPU-Measurement: Sampling" }, {.name = "CMC", .desc = "[EXT] CPU-Measurement: Counter" }, + {.name = "CMR", .desc = "[EXT] CPU-Measurement: RI" }, {.name = "CIO", .desc = "[I/O] Common I/O Layer Interrupt" }, {.name = "QAI", .desc = "[I/O] QDIO Adapter Interrupt" }, {.name = "DAS", .desc = "[I/O] DASD" }, diff --git a/arch/s390/kernel/process.c b/arch/s390/kernel/process.c index 7331753..cb4328e 100644 --- a/arch/s390/kernel/process.c +++ b/arch/s390/kernel/process.c @@ -30,6 +30,7 @@ #include #include #include +#include #include "entry.h" asmlinkage void ret_from_fork(void) asm ("ret_from_fork"); @@ -132,6 +133,7 @@ EXPORT_SYMBOL(kernel_thread); */ void exit_thread(void) { + exit_thread_runtime_instr(); } void flush_thread(void) @@ -170,6 +172,11 @@ int copy_thread(unsigned long clone_flags, unsigned long new_stackp, /* Save access registers to new thread structure. */ save_access_regs(&p->thread.acrs[0]); + /* Don't copy runtime instrumentation info */ + p->thread.ri_cb = NULL; + p->thread.ri_signum = 0; + frame->childregs.psw.mask &= ~PSW_MASK_RI; + #ifndef CONFIG_64BIT /* * save fprs to current->thread.fp_regs to merge them with diff --git a/arch/s390/kernel/runtime_instr.c b/arch/s390/kernel/runtime_instr.c new file mode 100644 index 0000000..e27e23d --- /dev/null +++ b/arch/s390/kernel/runtime_instr.c @@ -0,0 +1,150 @@ +/* + * Copyright IBM Corp. 2012 + * Author(s): Jan Glauber + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* empty control block to disable RI by loading it */ +struct runtime_instr_cb runtime_instr_empty_cb; + +static int runtime_instr_avail(void) +{ + return test_facility(64); +} + +static void disable_runtime_instr(void) +{ + struct pt_regs *regs = task_pt_regs(current); + + load_runtime_instr_cb(&runtime_instr_empty_cb); + + /* + * Make sure the RI bit is deleted from the PSW. If the user did not + * switch off RI before the system call the process will get a + * specification exception otherwise. + */ + regs->psw.mask &= ~PSW_MASK_RI; +} + +static void init_runtime_instr_cb(struct runtime_instr_cb *cb) +{ + cb->buf_limit = 0xfff; + if (addressing_mode == HOME_SPACE_MODE) + cb->home_space = 1; + cb->int_requested = 1; + cb->pstate = 1; + cb->pstate_set_buf = 1; + cb->pstate_sample = 1; + cb->pstate_collect = 1; + cb->key = PAGE_DEFAULT_KEY; + cb->valid = 1; +} + +void exit_thread_runtime_instr(void) +{ + struct task_struct *task = current; + + if (!task->thread.ri_cb) + return; + disable_runtime_instr(); + kfree(task->thread.ri_cb); + task->thread.ri_signum = 0; + task->thread.ri_cb = NULL; +} + +static void runtime_instr_int_handler(struct ext_code ext_code, + unsigned int param32, unsigned long param64) +{ + struct siginfo info; + + if (!(param32 & CPU_MF_INT_RI_MASK)) + return; + + kstat_cpu(smp_processor_id()).irqs[EXTINT_CMR]++; + + if (!current->thread.ri_cb) + return; + if (current->thread.ri_signum < SIGRTMIN || + current->thread.ri_signum > SIGRTMAX) { + WARN_ON_ONCE(1); + return; + } + + memset(&info, 0, sizeof(info)); + info.si_signo = current->thread.ri_signum; + info.si_code = SI_QUEUE; + if (param32 & CPU_MF_INT_RI_BUF_FULL) + info.si_int = ENOBUFS; + else if (param32 & CPU_MF_INT_RI_HALTED) + info.si_int = ECANCELED; + else + return; /* unknown reason */ + + send_sig_info(current->thread.ri_signum, &info, current); +} + +SYSCALL_DEFINE2(s390_runtime_instr, int, command, int, signum) +{ + struct runtime_instr_cb *cb; + + if (!runtime_instr_avail()) + return -EOPNOTSUPP; + + if (command == S390_RUNTIME_INSTR_STOP) { + preempt_disable(); + exit_thread_runtime_instr(); + preempt_enable(); + return 0; + } + + if (command != S390_RUNTIME_INSTR_START || + (signum < SIGRTMIN || signum > SIGRTMAX)) + return -EINVAL; + + if (!current->thread.ri_cb) { + cb = kzalloc(sizeof(*cb), GFP_KERNEL); + if (!cb) + return -ENOMEM; + } else { + cb = current->thread.ri_cb; + memset(cb, 0, sizeof(*cb)); + } + + init_runtime_instr_cb(cb); + current->thread.ri_signum = signum; + + /* now load the control block to make it available */ + preempt_disable(); + current->thread.ri_cb = cb; + load_runtime_instr_cb(cb); + preempt_enable(); + return 0; +} + +static int __init runtime_instr_init(void) +{ + int rc; + + if (!runtime_instr_avail()) + return 0; + + measurement_alert_subclass_register(); + rc = register_external_interrupt(0x1407, runtime_instr_int_handler); + if (rc) + measurement_alert_subclass_unregister(); + else + pr_info("Runtime instrumentation facility initialized\n"); + return rc; +} +device_initcall(runtime_instr_init); diff --git a/arch/s390/kernel/syscalls.S b/arch/s390/kernel/syscalls.S index bcab2f0..539f0a7 100644 --- a/arch/s390/kernel/syscalls.S +++ b/arch/s390/kernel/syscalls.S @@ -350,3 +350,4 @@ SYSCALL(sys_syncfs,sys_syncfs,sys_syncfs_wrapper) SYSCALL(sys_setns,sys_setns,sys_setns_wrapper) SYSCALL(sys_process_vm_readv,sys_process_vm_readv,compat_sys_process_vm_readv_wrapper) /* 340 */ SYSCALL(sys_process_vm_writev,sys_process_vm_writev,compat_sys_process_vm_writev_wrapper) +SYSCALL(sys_ni_syscall,sys_s390_runtime_instr,sys_s390_runtime_instr_wrapper) -- cgit v0.10.2 From d35339a42dd1f53b0bb86cf75418a9b7cf5f0f30 Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Tue, 31 Jul 2012 11:03:04 +0200 Subject: s390: add support for transactional memory Allow user-space processes to use transactional execution (TX). If the TX facility is available user space programs can use transactions for fine-grained serialization based on the data objects that are referenced during a transaction. This is useful for lockless data structures and speculative compiler optimizations. Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/elf.h b/arch/s390/include/asm/elf.h index 9b94a16..db57594 100644 --- a/arch/s390/include/asm/elf.h +++ b/arch/s390/include/asm/elf.h @@ -101,6 +101,7 @@ #define HWCAP_S390_HPAGE 128 #define HWCAP_S390_ETF3EH 256 #define HWCAP_S390_HIGH_GPRS 512 +#define HWCAP_S390_TE 1024 /* * These are used to set parameters in the core dumps. diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h index aab5555..bbf8141 100644 --- a/arch/s390/include/asm/lowcore.h +++ b/arch/s390/include/asm/lowcore.h @@ -329,9 +329,13 @@ struct _lowcore { __u8 pad_0x1338[0x1340-0x1338]; /* 0x1338 */ __u32 access_regs_save_area[16]; /* 0x1340 */ __u64 cregs_save_area[16]; /* 0x1380 */ + __u8 pad_0x1400[0x1800-0x1400]; /* 0x1400 */ + + /* Transaction abort diagnostic block */ + __u8 pgm_tdb[256]; /* 0x1800 */ /* align to the top of the prefix area */ - __u8 pad_0x1400[0x2000-0x1400]; /* 0x1400 */ + __u8 pad_0x1900[0x2000-0x1900]; /* 0x1900 */ } __packed; #endif /* CONFIG_32BIT */ diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h index 0fff583..46fe1fb 100644 --- a/arch/s390/include/asm/processor.h +++ b/arch/s390/include/asm/processor.h @@ -76,14 +76,20 @@ struct thread_struct { unsigned long gmap_addr; /* address of last gmap fault. */ struct per_regs per_user; /* User specified PER registers */ struct per_event per_event; /* Cause of the last PER trap */ + unsigned long per_flags; /* Flags to control debug behavior */ /* pfault_wait is used to block the process on a pfault event */ unsigned long pfault_wait; struct list_head list; /* cpu runtime instrumentation */ struct runtime_instr_cb *ri_cb; int ri_signum; +#ifdef CONFIG_64BIT + unsigned char trap_tdb[256]; /* Transaction abort diagnose block */ +#endif }; +#define PER_FLAG_NO_TE 1UL /* Flag to disable transactions. */ + typedef struct thread_struct thread_struct; /* diff --git a/arch/s390/include/asm/ptrace.h b/arch/s390/include/asm/ptrace.h index 5c32bae..ce20a53 100644 --- a/arch/s390/include/asm/ptrace.h +++ b/arch/s390/include/asm/ptrace.h @@ -361,17 +361,19 @@ struct per_struct_kernel { unsigned char access_id; /* PER trap access identification */ }; -#define PER_EVENT_MASK 0xE9000000UL +#define PER_EVENT_MASK 0xEB000000UL #define PER_EVENT_BRANCH 0x80000000UL #define PER_EVENT_IFETCH 0x40000000UL #define PER_EVENT_STORE 0x20000000UL #define PER_EVENT_STORE_REAL 0x08000000UL +#define PER_EVENT_TRANSACTION_END 0x02000000UL #define PER_EVENT_NULLIFICATION 0x01000000UL -#define PER_CONTROL_MASK 0x00a00000UL +#define PER_CONTROL_MASK 0x00e00000UL #define PER_CONTROL_BRANCH_ADDRESS 0x00800000UL +#define PER_CONTROL_SUSPENSION 0x00400000UL #define PER_CONTROL_ALTERATION 0x00200000UL #endif @@ -485,6 +487,8 @@ typedef struct #define PTRACE_GET_LAST_BREAK 0x5006 #define PTRACE_PEEK_SYSTEM_CALL 0x5007 #define PTRACE_POKE_SYSTEM_CALL 0x5008 +#define PTRACE_ENABLE_TE 0x5009 +#define PTRACE_DISABLE_TE 0x5010 /* * PT_PROT definition is loosely based on hppa bsd definition in diff --git a/arch/s390/include/asm/setup.h b/arch/s390/include/asm/setup.h index e6859d1..908f688 100644 --- a/arch/s390/include/asm/setup.h +++ b/arch/s390/include/asm/setup.h @@ -80,6 +80,7 @@ extern unsigned int addressing_mode; #define MACHINE_FLAG_LPAR (1UL << 12) #define MACHINE_FLAG_SPP (1UL << 13) #define MACHINE_FLAG_TOPOLOGY (1UL << 14) +#define MACHINE_FLAG_TE (1UL << 15) #define MACHINE_IS_VM (S390_lowcore.machine_flags & MACHINE_FLAG_VM) #define MACHINE_IS_KVM (S390_lowcore.machine_flags & MACHINE_FLAG_KVM) @@ -98,6 +99,7 @@ extern unsigned int addressing_mode; #define MACHINE_HAS_PFMF (0) #define MACHINE_HAS_SPP (0) #define MACHINE_HAS_TOPOLOGY (0) +#define MACHINE_HAS_TE (0) #else /* CONFIG_64BIT */ #define MACHINE_HAS_IEEE (1) #define MACHINE_HAS_CSP (1) @@ -109,6 +111,7 @@ extern unsigned int addressing_mode; #define MACHINE_HAS_PFMF (S390_lowcore.machine_flags & MACHINE_FLAG_PFMF) #define MACHINE_HAS_SPP (S390_lowcore.machine_flags & MACHINE_FLAG_SPP) #define MACHINE_HAS_TOPOLOGY (S390_lowcore.machine_flags & MACHINE_FLAG_TOPOLOGY) +#define MACHINE_HAS_TE (S390_lowcore.machine_flags & MACHINE_FLAG_TE) #endif /* CONFIG_64BIT */ #define ZFCPDUMP_HSA_SIZE (32UL<<20) diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c index 45ef1a7..fface87 100644 --- a/arch/s390/kernel/asm-offsets.c +++ b/arch/s390/kernel/asm-offsets.c @@ -157,6 +157,8 @@ int main(void) DEFINE(__LC_LAST_BREAK, offsetof(struct _lowcore, breaking_event_addr)); DEFINE(__LC_VDSO_PER_CPU, offsetof(struct _lowcore, vdso_per_cpu_data)); DEFINE(__LC_GMAP, offsetof(struct _lowcore, gmap)); + DEFINE(__LC_PGM_TDB, offsetof(struct _lowcore, pgm_tdb)); + DEFINE(__THREAD_trap_tdb, offsetof(struct task_struct, thread.trap_tdb)); DEFINE(__GMAP_ASCE, offsetof(struct gmap, asce)); #endif /* CONFIG_32BIT */ return 0; diff --git a/arch/s390/kernel/dis.c b/arch/s390/kernel/dis.c index 4bc67db..7b6ad27 100644 --- a/arch/s390/kernel/dis.c +++ b/arch/s390/kernel/dis.c @@ -317,6 +317,9 @@ enum { LONG_INSN_RISBLG, LONG_INSN_RINEXT, LONG_INSN_RIEMIT, + LONG_INSN_TABORT, + LONG_INSN_TBEGIN, + LONG_INSN_TBEGINC, }; static char *long_insn_name[] = { @@ -334,6 +337,9 @@ static char *long_insn_name[] = { [LONG_INSN_RISBLG] = "risblk", [LONG_INSN_RINEXT] = "rinext", [LONG_INSN_RIEMIT] = "riemit", + [LONG_INSN_TABORT] = "tabort", + [LONG_INSN_TBEGIN] = "tbegin", + [LONG_INSN_TBEGINC] = "tbeginc", }; static struct insn opcode[] = { @@ -609,6 +615,9 @@ static struct insn opcode_b2[] = { { "lpswe", 0xb2, INSTR_S_RD }, { "srnmt", 0xb9, INSTR_S_RD }, { "lfas", 0xbd, INSTR_S_RD }, + { "etndg", 0xec, INSTR_RRE_R0 }, + { { 0, LONG_INSN_TABORT }, 0xfc, INSTR_S_RD }, + { "tend", 0xf8, INSTR_S_RD }, #endif { "stidp", 0x02, INSTR_S_RD }, { "sck", 0x04, INSTR_S_RD }, @@ -1165,6 +1174,7 @@ static struct insn opcode_e3[] = { { "stfh", 0xcb, INSTR_RXY_RRRD }, { "chf", 0xcd, INSTR_RXY_RRRD }, { "clhf", 0xcf, INSTR_RXY_RRRD }, + { "ntstg", 0x25, INSTR_RXY_RRRD }, #endif { "lrv", 0x1e, INSTR_RXY_RRRD }, { "lrvh", 0x1f, INSTR_RXY_RRRD }, @@ -1188,6 +1198,8 @@ static struct insn opcode_e5[] = { { "mvhhi", 0x44, INSTR_SIL_RDI }, { "mvhi", 0x4c, INSTR_SIL_RDI }, { "mvghi", 0x48, INSTR_SIL_RDI }, + { { 0, LONG_INSN_TBEGIN }, 0x60, INSTR_SIL_RDU }, + { { 0, LONG_INSN_TBEGINC }, 0x61, INSTR_SIL_RDU }, #endif { "lasp", 0x00, INSTR_SSE_RDRD }, { "tprot", 0x01, INSTR_SSE_RDRD }, diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c index e484079..1345ba4 100644 --- a/arch/s390/kernel/early.c +++ b/arch/s390/kernel/early.c @@ -370,6 +370,8 @@ static __init void detect_machine_facilities(void) S390_lowcore.machine_flags |= MACHINE_FLAG_MVCOS; if (test_facility(40)) S390_lowcore.machine_flags |= MACHINE_FLAG_SPP; + if (test_facility(50) && test_facility(73)) + S390_lowcore.machine_flags |= MACHINE_FLAG_TE; #endif } @@ -439,7 +441,6 @@ static void __init setup_boot_command_line(void) append_to_cmdline(append_ipl_scpdata); } - /* * Save ipl parameters, clear bss memory, initialize storage keys * and create a kernel NSS at startup if the SAVESYS= parm is defined diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S index 349b7ee..95e9d93 100644 --- a/arch/s390/kernel/entry64.S +++ b/arch/s390/kernel/entry64.S @@ -412,6 +412,11 @@ ENTRY(pgm_check_handler) 1: UPDATE_VTIME %r14,__LC_SYNC_ENTER_TIMER LAST_BREAK %r14 lg %r15,__LC_KERNEL_STACK + lg %r14,__TI_task(%r12) + lghi %r13,__LC_PGM_TDB + tm __LC_PGM_ILC+2,0x02 # check for transaction abort + jz 2f + mvc __THREAD_trap_tdb(256,%r14),0(%r13) 2: aghi %r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE) la %r11,STACK_FRAME_OVERHEAD(%r15) stmg %r0,%r7,__PT_R0(%r11) @@ -422,13 +427,12 @@ ENTRY(pgm_check_handler) stg %r10,__PT_ARGS(%r11) tm __LC_PGM_ILC+3,0x80 # check for per exception jz 0f - lg %r1,__TI_task(%r12) tmhh %r8,0x0001 # kernel per event ? jz pgm_kprobe oi __TI_flags+7(%r12),_TIF_PER_TRAP - mvc __THREAD_per_address(8,%r1),__LC_PER_ADDRESS - mvc __THREAD_per_cause(2,%r1),__LC_PER_CAUSE - mvc __THREAD_per_paid(1,%r1),__LC_PER_PAID + mvc __THREAD_per_address(8,%r14),__LC_PER_ADDRESS + mvc __THREAD_per_cause(2,%r14),__LC_PER_CAUSE + mvc __THREAD_per_paid(1,%r14),__LC_PER_PAID 0: REENABLE_IRQS xc __SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15) larl %r1,pgm_check_table diff --git a/arch/s390/kernel/processor.c b/arch/s390/kernel/processor.c index 572d4c9..22bf8f0 100644 --- a/arch/s390/kernel/processor.c +++ b/arch/s390/kernel/processor.c @@ -39,9 +39,9 @@ void __cpuinit cpu_init(void) */ static int show_cpuinfo(struct seq_file *m, void *v) { - static const char *hwcap_str[10] = { + static const char *hwcap_str[11] = { "esan3", "zarch", "stfle", "msa", "ldisp", "eimm", "dfp", - "edat", "etf3eh", "highgprs" + "edat", "etf3eh", "highgprs", "te" }; unsigned long n = (unsigned long) v - 1; int i; @@ -54,7 +54,7 @@ static int show_cpuinfo(struct seq_file *m, void *v) num_online_cpus(), loops_per_jiffy/(500000/HZ), (loops_per_jiffy/(5000/HZ))%100); seq_puts(m, "features\t: "); - for (i = 0; i < 10; i++) + for (i = 0; i < 11; i++) if (hwcap_str[i] && (elf_hwcap & (1UL << i))) seq_printf(m, "%s ", hwcap_str[i]); seq_puts(m, "\n"); diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c index e4be113..b817cc5 100644 --- a/arch/s390/kernel/ptrace.c +++ b/arch/s390/kernel/ptrace.c @@ -42,6 +42,7 @@ enum s390_regset { REGSET_GENERAL, REGSET_FP, REGSET_LAST_BREAK, + REGSET_TDB, REGSET_SYSTEM_CALL, REGSET_GENERAL_EXTENDED, }; @@ -52,6 +53,21 @@ void update_per_regs(struct task_struct *task) struct thread_struct *thread = &task->thread; struct per_regs old, new; + /* Take care of the enable/disable of transactional execution. */ + if (MACHINE_HAS_TE) { + unsigned long cr0, cr0_new; + + __ctl_store(cr0, 0, 0); + /* set or clear transaction execution bits 8 and 9. */ + if (task->thread.per_flags & PER_FLAG_NO_TE) + cr0_new = cr0 & ~(3UL << 54); + else + cr0_new = cr0 | (3UL << 54); + /* Only load control register 0 if necessary. */ + if (cr0 != cr0_new) + __ctl_load(cr0_new, 0, 0); + } + /* Copy user specified PER registers */ new.control = thread->per_user.control; new.start = thread->per_user.start; @@ -60,6 +76,10 @@ void update_per_regs(struct task_struct *task) /* merge TIF_SINGLE_STEP into user specified PER registers. */ if (test_tsk_thread_flag(task, TIF_SINGLE_STEP)) { new.control |= PER_EVENT_IFETCH; +#ifdef CONFIG_64BIT + new.control |= PER_CONTROL_SUSPENSION; + new.control |= PER_EVENT_TRANSACTION_END; +#endif new.start = 0; new.end = PSW_ADDR_INSN; } @@ -100,6 +120,7 @@ void ptrace_disable(struct task_struct *task) memset(&task->thread.per_event, 0, sizeof(task->thread.per_event)); clear_tsk_thread_flag(task, TIF_SINGLE_STEP); clear_tsk_thread_flag(task, TIF_PER_TRAP); + task->thread.per_flags = 0; } #ifndef CONFIG_64BIT @@ -416,6 +437,16 @@ long arch_ptrace(struct task_struct *child, long request, put_user(task_thread_info(child)->last_break, (unsigned long __user *) data); return 0; + case PTRACE_ENABLE_TE: + if (!MACHINE_HAS_TE) + return -EIO; + child->thread.per_flags &= ~PER_FLAG_NO_TE; + return 0; + case PTRACE_DISABLE_TE: + if (!MACHINE_HAS_TE) + return -EIO; + child->thread.per_flags |= PER_FLAG_NO_TE; + return 0; default: /* Removing high order bit from addr (only for 31 bit). */ addr &= PSW_ADDR_INSN; @@ -903,6 +934,28 @@ static int s390_last_break_set(struct task_struct *target, return 0; } +static int s390_tdb_get(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + void *kbuf, void __user *ubuf) +{ + struct pt_regs *regs = task_pt_regs(target); + unsigned char *data; + + if (!(regs->int_code & 0x200)) + return -ENODATA; + data = target->thread.trap_tdb; + return user_regset_copyout(&pos, &count, &kbuf, &ubuf, data, 0, 256); +} + +static int s390_tdb_set(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + const void *kbuf, const void __user *ubuf) +{ + return 0; +} + #endif static int s390_system_call_get(struct task_struct *target, @@ -951,6 +1004,14 @@ static const struct user_regset s390_regsets[] = { .get = s390_last_break_get, .set = s390_last_break_set, }, + [REGSET_TDB] = { + .core_note_type = NT_S390_TDB, + .n = 1, + .size = 256, + .align = 1, + .get = s390_tdb_get, + .set = s390_tdb_set, + }, #endif [REGSET_SYSTEM_CALL] = { .core_note_type = NT_S390_SYSTEM_CALL, @@ -1148,6 +1209,14 @@ static const struct user_regset s390_compat_regsets[] = { .get = s390_compat_last_break_get, .set = s390_compat_last_break_set, }, + [REGSET_TDB] = { + .core_note_type = NT_S390_TDB, + .n = 1, + .size = 256, + .align = 1, + .get = s390_tdb_get, + .set = s390_tdb_set, + }, [REGSET_SYSTEM_CALL] = { .core_note_type = NT_S390_SYSTEM_CALL, .n = 1, diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index 40b5769..7b59fff 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -980,6 +980,12 @@ static void __init setup_hwcaps(void) * HWCAP_S390_HIGH_GPRS is bit 9. */ elf_hwcap |= HWCAP_S390_HIGH_GPRS; + + /* + * Transactional execution support HWCAP_S390_TE is bit 10. + */ + if (test_facility(50) && test_facility(73)) + elf_hwcap |= HWCAP_S390_TE; #endif get_cpu_id(&cpu_id); diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c index 01775c0..29af526 100644 --- a/arch/s390/kernel/traps.c +++ b/arch/s390/kernel/traps.c @@ -57,6 +57,23 @@ static int kstack_depth_to_print = 12; static int kstack_depth_to_print = 20; #endif /* CONFIG_64BIT */ +static inline void __user *get_trap_ip(struct pt_regs *regs) +{ +#ifdef CONFIG_64BIT + unsigned long address; + + if (regs->int_code & 0x200) + address = *(unsigned long *)(current->thread.trap_tdb + 24); + else + address = regs->psw.addr; + return (void __user *) + ((address - (regs->int_code >> 16)) & PSW_ADDR_INSN); +#else + return (void __user *) + ((regs->psw.addr - (regs->int_code >> 16)) & PSW_ADDR_INSN); +#endif +} + /* * For show_trace we have tree different stack to consider: * - the panic stack which is used if the kernel stack has overflown @@ -285,12 +302,6 @@ int is_valid_bugaddr(unsigned long addr) return 1; } -static inline void __user *get_psw_address(struct pt_regs *regs) -{ - return (void __user *) - ((regs->psw.addr - (regs->int_code >> 16)) & PSW_ADDR_INSN); -} - static void __kprobes do_trap(struct pt_regs *regs, int si_signo, int si_code, char *str) { @@ -304,7 +315,7 @@ static void __kprobes do_trap(struct pt_regs *regs, info.si_signo = si_signo; info.si_errno = 0; info.si_code = si_code; - info.si_addr = get_psw_address(regs); + info.si_addr = get_trap_ip(regs); force_sig_info(si_signo, &info, current); report_user_fault(regs, si_signo); } else { @@ -381,6 +392,11 @@ DO_ERROR_INFO(special_op_exception, SIGILL, ILL_ILLOPN, DO_ERROR_INFO(translation_exception, SIGILL, ILL_ILLOPN, "translation exception") +#ifdef CONFIG_64BIT +DO_ERROR_INFO(transaction_exception, SIGILL, ILL_ILLOPN, + "transaction constraint exception") +#endif + static inline void do_fp_trap(struct pt_regs *regs, int fpc) { int si_code = 0; @@ -408,7 +424,7 @@ static void __kprobes illegal_op(struct pt_regs *regs) __u16 __user *location; int signal = 0; - location = get_psw_address(regs); + location = get_trap_ip(regs); if (user_mode(regs)) { if (get_user(*((__u16 *) opcode), (__u16 __user *) location)) @@ -476,7 +492,7 @@ void specification_exception(struct pt_regs *regs) __u16 __user *location = NULL; int signal = 0; - location = (__u16 __user *) get_psw_address(regs); + location = (__u16 __user *) get_trap_ip(regs); if (user_mode(regs)) { get_user(*((__u16 *) opcode), location); @@ -525,7 +541,7 @@ static void data_exception(struct pt_regs *regs) __u16 __user *location; int signal = 0; - location = get_psw_address(regs); + location = get_trap_ip(regs); if (MACHINE_HAS_IEEE) asm volatile("stfpc %0" : "=m" (current->thread.fp_regs.fpc)); @@ -641,6 +657,7 @@ void __init trap_init(void) pgm_check_table[0x12] = &translation_exception; pgm_check_table[0x13] = &special_op_exception; #ifdef CONFIG_64BIT + pgm_check_table[0x18] = &transaction_exception; pgm_check_table[0x38] = &do_asce_exception; pgm_check_table[0x39] = &do_dat_exception; pgm_check_table[0x3A] = &do_dat_exception; diff --git a/include/linux/elf.h b/include/linux/elf.h index 999b4f5..f930b1a 100644 --- a/include/linux/elf.h +++ b/include/linux/elf.h @@ -387,6 +387,7 @@ typedef struct elf64_shdr { #define NT_S390_PREFIX 0x305 /* s390 prefix register */ #define NT_S390_LAST_BREAK 0x306 /* s390 breaking event address */ #define NT_S390_SYSTEM_CALL 0x307 /* s390 system call restart data */ +#define NT_S390_TDB 0x308 /* s390 transaction diagnostic block */ #define NT_ARM_VFP 0x400 /* ARM VFP/NEON registers */ -- cgit v0.10.2 From 6668022c7bde3fdc96d3d257294a7216c7a46829 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Wed, 29 Aug 2012 14:12:20 +0200 Subject: s390/cache: add cpu cache information to /proc/cpuinfo Add a line for each cpu cache to /proc/cpuinfo. Since we only have information of private cpu caches in sysfs we add a line for each cpu cache in /proc/cpuinfo which will also contain information about shared caches. For a z196 machine /proc/cpuinfo now looks like: vendor_id : IBM/S390 bogomips per cpu: 14367.00 features : esan3 zarch stfle msa ldisp eimm dfp etf3eh highgprs cache0 : level=1 type=Data scope=Private size=64K line_size=256 associativity=4 cache1 : level=1 type=Instruction scope=Private size=128K line_size=256 associativity=8 cache2 : level=2 type=Unified scope=Private size=1536K line_size=256 associativity=12 cache3 : level=3 type=Unified scope=Shared size=24576K line_size=256 associativity=12 cache4 : level=4 type=Unified scope=Shared size=196608K line_size=256 associativity=24 processor 0: version = FF, identification = 000123, machine = 2817 processor 1: version = FF, identification = 100123, machine = 2817 processor 2: version = FF, identification = 200123, machine = 2817 processor 3: version = FF, identification = 200123, machine = 2817 Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h index 46fe1fb..7e81ff1 100644 --- a/arch/s390/include/asm/processor.h +++ b/arch/s390/include/asm/processor.h @@ -140,6 +140,12 @@ struct task_struct; struct mm_struct; struct seq_file; +#ifdef CONFIG_64BIT +extern void show_cacheinfo(struct seq_file *m); +#else +static inline void show_cacheinfo(struct seq_file *m) { } +#endif + /* Free all resources held by a thread. */ extern void release_thread(struct task_struct *); extern int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags); diff --git a/arch/s390/kernel/cache.c b/arch/s390/kernel/cache.c index 5e20bab..8df8d8a1 100644 --- a/arch/s390/kernel/cache.c +++ b/arch/s390/kernel/cache.c @@ -6,6 +6,7 @@ */ #include +#include #include #include #include @@ -17,8 +18,9 @@ struct cache { unsigned int line_size; unsigned int associativity; unsigned int nr_sets; - int level; - int type; + unsigned int level : 3; + unsigned int type : 2; + unsigned int private : 1; struct list_head list; }; @@ -83,6 +85,24 @@ static const char * const cache_type_string[] = { static struct cache_dir *cache_dir_cpu[NR_CPUS]; static LIST_HEAD(cache_list); +void show_cacheinfo(struct seq_file *m) +{ + struct cache *cache; + int index = 0; + + list_for_each_entry(cache, &cache_list, list) { + seq_printf(m, "cache%-11d: ", index); + seq_printf(m, "level=%d ", cache->level); + seq_printf(m, "type=%s ", cache_type_string[cache->type]); + seq_printf(m, "scope=%s ", cache->private ? "Private" : "Shared"); + seq_printf(m, "size=%luK ", cache->size >> 10); + seq_printf(m, "line_size=%u ", cache->line_size); + seq_printf(m, "associativity=%d", cache->associativity); + seq_puts(m, "\n"); + index++; + } +} + static inline unsigned long ecag(int ai, int li, int ti) { unsigned long cmd, val; @@ -93,7 +113,7 @@ static inline unsigned long ecag(int ai, int li, int ti) return val; } -static int __init cache_add(int level, int type) +static int __init cache_add(int level, int private, int type) { struct cache *cache; int ti; @@ -107,8 +127,9 @@ static int __init cache_add(int level, int type) cache->associativity = ecag(EXTRACT_ASSOCIATIVITY, level, ti); cache->nr_sets = cache->size / cache->associativity; cache->nr_sets /= cache->line_size; + cache->private = private; cache->level = level + 1; - cache->type = type; + cache->type = type - 1; list_add_tail(&cache->list, &cache_list); return 0; } @@ -117,23 +138,26 @@ static void __init cache_build_info(void) { struct cache *cache, *next; union cache_topology ct; - int level, rc; + int level, private, rc; ct.raw = ecag(EXTRACT_TOPOLOGY, 0, 0); for (level = 0; level < CACHE_MAX_LEVEL; level++) { switch (ct.ci[level].scope) { case CACHE_SCOPE_NOTEXISTS: case CACHE_SCOPE_RESERVED: - case CACHE_SCOPE_SHARED: return; + case CACHE_SCOPE_SHARED: + private = 0; + break; case CACHE_SCOPE_PRIVATE: + private = 1; break; } if (ct.ci[level].type == CACHE_TYPE_SEPARATE) { - rc = cache_add(level, CACHE_TYPE_DATA); - rc |= cache_add(level, CACHE_TYPE_INSTRUCTION); + rc = cache_add(level, private, CACHE_TYPE_DATA); + rc |= cache_add(level, private, CACHE_TYPE_INSTRUCTION); } else { - rc = cache_add(level, ct.ci[level].type); + rc = cache_add(level, private, ct.ci[level].type); } if (rc) goto error; @@ -208,7 +232,7 @@ DEFINE_CACHE_ATTR(size, "%luK\n", index->cache->size >> 10); DEFINE_CACHE_ATTR(coherency_line_size, "%u\n", index->cache->line_size); DEFINE_CACHE_ATTR(number_of_sets, "%u\n", index->cache->nr_sets); DEFINE_CACHE_ATTR(ways_of_associativity, "%u\n", index->cache->associativity); -DEFINE_CACHE_ATTR(type, "%s\n", cache_type_string[index->cache->type - 1]); +DEFINE_CACHE_ATTR(type, "%s\n", cache_type_string[index->cache->type]); DEFINE_CACHE_ATTR(level, "%d\n", index->cache->level); static ssize_t shared_cpu_map_func(struct kobject *kobj, int type, char *buf) @@ -298,6 +322,8 @@ static int __cpuinit cache_add_cpu(int cpu) if (!cache_dir) return -ENOMEM; list_for_each_entry(cache, &cache_list, list) { + if (!cache->private) + break; rc = cache_create_index_dir(cache_dir, cache, index, cpu); if (rc) return rc; diff --git a/arch/s390/kernel/processor.c b/arch/s390/kernel/processor.c index 22bf8f0..63f3b76 100644 --- a/arch/s390/kernel/processor.c +++ b/arch/s390/kernel/processor.c @@ -58,6 +58,7 @@ static int show_cpuinfo(struct seq_file *m, void *v) if (hwcap_str[i] && (elf_hwcap & (1UL << i))) seq_printf(m, "%s ", hwcap_str[i]); seq_puts(m, "\n"); + show_cacheinfo(m); } get_online_cpus(); if (cpu_online(n)) { -- cgit v0.10.2 From fbf3c54239f77a82218002e0c511819e274e7cd4 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Wed, 29 Aug 2012 14:54:38 +0200 Subject: s390/processor: use ARRAY_SIZE instead of hard coded value Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/processor.c b/arch/s390/kernel/processor.c index 63f3b76..753c41d 100644 --- a/arch/s390/kernel/processor.c +++ b/arch/s390/kernel/processor.c @@ -39,7 +39,7 @@ void __cpuinit cpu_init(void) */ static int show_cpuinfo(struct seq_file *m, void *v) { - static const char *hwcap_str[11] = { + static const char *hwcap_str[] = { "esan3", "zarch", "stfle", "msa", "ldisp", "eimm", "dfp", "edat", "etf3eh", "highgprs", "te" }; @@ -54,7 +54,7 @@ static int show_cpuinfo(struct seq_file *m, void *v) num_online_cpus(), loops_per_jiffy/(500000/HZ), (loops_per_jiffy/(5000/HZ))%100); seq_puts(m, "features\t: "); - for (i = 0; i < 11; i++) + for (i = 0; i < ARRAY_SIZE(hwcap_str); i++) if (hwcap_str[i] && (elf_hwcap & (1UL << i))) seq_printf(m, "%s ", hwcap_str[i]); seq_puts(m, "\n"); -- cgit v0.10.2 From c5e3acd666543544d402c83d24d44e487fb51f11 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 30 Aug 2012 14:24:42 +0200 Subject: s390/smp: avoid concurrent write to sigp status field When a sigp instruction is issued it may store a status. This status is currently stored in a per cpu field of the target cpu. If multiple cpus issue a sigp instruction with the same target cpu and a status is stored the result is not necessarily as expected. Currently this is not an issue: - on cpu hotplug no sigps, except "restart" and "sense" are sent to the target cpu. - on external call we don't look at the status if it is stored - on sense running the condition code "status stored" is sufficient to tell if a cpu is running or not Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index 87be102..c88fe85 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -66,7 +66,6 @@ struct pcpu { unsigned long panic_stack; /* panic stack for the cpu */ unsigned long ec_mask; /* bit mask for ec_xxx functions */ int state; /* physical cpu state */ - u32 status; /* last status received via sigp */ u16 address; /* physical cpu address */ }; @@ -99,7 +98,7 @@ static inline int __pcpu_sigp_relax(u16 addr, u8 order, u32 parm, u32 *status) int cc; while (1) { - cc = __pcpu_sigp(addr, order, parm, status); + cc = __pcpu_sigp(addr, order, parm, NULL); if (cc != SIGP_CC_BUSY) return cc; cpu_relax(); @@ -111,7 +110,7 @@ static int pcpu_sigp_retry(struct pcpu *pcpu, u8 order, u32 parm) int cc, retry; for (retry = 0; ; retry++) { - cc = __pcpu_sigp(pcpu->address, order, parm, &pcpu->status); + cc = __pcpu_sigp(pcpu->address, order, parm, NULL); if (cc != SIGP_CC_BUSY) break; if (retry >= 3) @@ -122,16 +121,18 @@ static int pcpu_sigp_retry(struct pcpu *pcpu, u8 order, u32 parm) static inline int pcpu_stopped(struct pcpu *pcpu) { + u32 status; + if (__pcpu_sigp(pcpu->address, SIGP_SENSE, - 0, &pcpu->status) != SIGP_CC_STATUS_STORED) + 0, &status) != SIGP_CC_STATUS_STORED) return 0; - return !!(pcpu->status & (SIGP_STATUS_CHECK_STOP|SIGP_STATUS_STOPPED)); + return !!(status & (SIGP_STATUS_CHECK_STOP|SIGP_STATUS_STOPPED)); } static inline int pcpu_running(struct pcpu *pcpu) { if (__pcpu_sigp(pcpu->address, SIGP_SENSE_RUNNING, - 0, &pcpu->status) != SIGP_CC_STATUS_STORED) + 0, NULL) != SIGP_CC_STATUS_STORED) return 1; /* Status stored condition code is equivalent to cpu not running. */ return 0; -- cgit v0.10.2 From 3d5787c9976889bcf26cbd8f3124b91678a33275 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 31 Aug 2012 13:35:25 +0200 Subject: s390/mm: change default addressing mode Change the default addressing mode so that user space runs in primary space and the kernel runs in home space. In addition remove the "switch_amode" kernel parameter so all users who already specified they want the new default behaviour will stay in the "switched" mode instead of in the opposite they intended. If there is a need to switch addressing modes, this can be done with the "user_mode" kernel parameter: user_mode=home Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index 7b59fff..39d609b 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -302,7 +302,7 @@ static int __init parse_vmalloc(char *arg) } early_param("vmalloc", parse_vmalloc); -unsigned int addressing_mode = HOME_SPACE_MODE; +unsigned int addressing_mode = PRIMARY_SPACE_MODE; EXPORT_SYMBOL_GPL(addressing_mode); static int set_amode_primary(void) @@ -323,16 +323,6 @@ static int set_amode_primary(void) } } -/* - * Switch kernel/user addressing modes? - */ -static int __init early_parse_switch_amode(char *p) -{ - addressing_mode = PRIMARY_SPACE_MODE; - return 0; -} -early_param("switch_amode", early_parse_switch_amode); - static int __init early_parse_user_mode(char *p) { if (p && strcmp(p, "primary") == 0) -- cgit v0.10.2 From 46e8894786327caf52cf686e27ba0795bddfcd63 Mon Sep 17 00:00:00 2001 From: Stefan Weinhuber Date: Fri, 31 Aug 2012 10:52:13 +0200 Subject: s390/partitions: make partition detection independent from DASD ioctls In some usage scenarios it is desireable to work with disk images or virtualized DASD devices. One problem that prevents such applications is the partition detection in ibm.c. Currently it works only for devices that support the BIODASDINFO2 ioctl, in other words, it only works for devices that belong to the DASD device driver. The information gained from the BIODASDINFO2 ioctl is only for a small set of legacy cases abolutely necessary. All current VOL1, LNX1 and CMS1 type of disk labels can be interpreted correctly without this information, as long as the generic HDIO_GETGEO ioctl works and provides a correct disk geometry. This patch makes the ibm.c partition detection as independent as possible from the BIODASDINFO2 ioctl. Only the following two cases are still restricted to real DASDs: - An FBA DASD, or LDL formatted ECKD DASD without any disk label. - An old style LNX1 label (without large volume support) on a disk with inconsistent device geometry. Signed-off-by: Stefan Weinhuber Signed-off-by: Martin Schwidefsky diff --git a/block/partitions/ibm.c b/block/partitions/ibm.c index 1104aca..47a6147 100644 --- a/block/partitions/ibm.c +++ b/block/partitions/ibm.c @@ -1,9 +1,8 @@ /* - * File...........: linux/fs/partitions/ibm.c * Author(s)......: Holger Smolinski * Volker Sameske * Bugreports.to..: - * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000 + * Copyright IBM Corp. 1999, 2012 */ #include @@ -17,17 +16,23 @@ #include "check.h" #include "ibm.h" + +union label_t { + struct vtoc_volume_label_cdl vol; + struct vtoc_volume_label_ldl lnx; + struct vtoc_cms_label cms; +}; + /* * compute the block number from a * cyl-cyl-head-head structure */ -static sector_t -cchh2blk (struct vtoc_cchh *ptr, struct hd_geometry *geo) { - +static sector_t cchh2blk(struct vtoc_cchh *ptr, struct hd_geometry *geo) +{ sector_t cyl; __u16 head; - /*decode cylinder and heads for large volumes */ + /* decode cylinder and heads for large volumes */ cyl = ptr->hh & 0xFFF0; cyl <<= 12; cyl |= ptr->cc; @@ -40,13 +45,12 @@ cchh2blk (struct vtoc_cchh *ptr, struct hd_geometry *geo) { * compute the block number from a * cyl-cyl-head-head-block structure */ -static sector_t -cchhb2blk (struct vtoc_cchhb *ptr, struct hd_geometry *geo) { - +static sector_t cchhb2blk(struct vtoc_cchhb *ptr, struct hd_geometry *geo) +{ sector_t cyl; __u16 head; - /*decode cylinder and heads for large volumes */ + /* decode cylinder and heads for large volumes */ cyl = ptr->hh & 0xFFF0; cyl <<= 12; cyl |= ptr->cc; @@ -56,26 +60,243 @@ cchhb2blk (struct vtoc_cchhb *ptr, struct hd_geometry *geo) { ptr->b; } +static int find_label(struct parsed_partitions *state, + dasd_information2_t *info, + struct hd_geometry *geo, + int blocksize, + sector_t *labelsect, + char name[], + char type[], + union label_t *label) +{ + Sector sect; + unsigned char *data; + sector_t testsect[3]; + unsigned char temp[5]; + int found = 0; + int i, testcount; + + /* There a three places where we may find a valid label: + * - on an ECKD disk it's block 2 + * - on an FBA disk it's block 1 + * - on an CMS formatted FBA disk it is sector 1, even if the block size + * is larger than 512 bytes (possible if the DIAG discipline is used) + * If we have a valid info structure, then we know exactly which case we + * have, otherwise we just search through all possebilities. + */ + if (info) { + if ((info->cu_type == 0x6310 && info->dev_type == 0x9336) || + (info->cu_type == 0x3880 && info->dev_type == 0x3370)) + testsect[0] = info->label_block; + else + testsect[0] = info->label_block * (blocksize >> 9); + testcount = 1; + } else { + testsect[0] = 1; + testsect[1] = (blocksize >> 9); + testsect[2] = 2 * (blocksize >> 9); + testcount = 3; + } + for (i = 0; i < testcount; ++i) { + data = read_part_sector(state, testsect[i], §); + if (data == NULL) + continue; + memcpy(label, data, sizeof(*label)); + memcpy(temp, data, 4); + temp[4] = 0; + EBCASC(temp, 4); + put_dev_sector(sect); + if (!strcmp(temp, "VOL1") || + !strcmp(temp, "LNX1") || + !strcmp(temp, "CMS1")) { + if (!strcmp(temp, "VOL1")) { + strncpy(type, label->vol.vollbl, 4); + strncpy(name, label->vol.volid, 6); + } else { + strncpy(type, label->lnx.vollbl, 4); + strncpy(name, label->lnx.volid, 6); + } + EBCASC(type, 4); + EBCASC(name, 6); + *labelsect = testsect[i]; + found = 1; + break; + } + } + if (!found) + memset(label, 0, sizeof(*label)); + + return found; +} + +static int find_vol1_partitions(struct parsed_partitions *state, + struct hd_geometry *geo, + int blocksize, + char name[], + union label_t *label) +{ + sector_t blk; + int counter; + char tmp[64]; + Sector sect; + unsigned char *data; + loff_t offset, size; + struct vtoc_format1_label f1; + int secperblk; + + snprintf(tmp, sizeof(tmp), "VOL1/%8s:", name); + strlcat(state->pp_buf, tmp, PAGE_SIZE); + /* + * get start of VTOC from the disk label and then search for format1 + * and format8 labels + */ + secperblk = blocksize >> 9; + blk = cchhb2blk(&label->vol.vtoc, geo) + 1; + counter = 0; + data = read_part_sector(state, blk * secperblk, §); + while (data != NULL) { + memcpy(&f1, data, sizeof(struct vtoc_format1_label)); + put_dev_sector(sect); + /* skip FMT4 / FMT5 / FMT7 labels */ + if (f1.DS1FMTID == _ascebc['4'] + || f1.DS1FMTID == _ascebc['5'] + || f1.DS1FMTID == _ascebc['7'] + || f1.DS1FMTID == _ascebc['9']) { + blk++; + data = read_part_sector(state, blk * secperblk, §); + continue; + } + /* only FMT1 and 8 labels valid at this point */ + if (f1.DS1FMTID != _ascebc['1'] && + f1.DS1FMTID != _ascebc['8']) + break; + /* OK, we got valid partition data */ + offset = cchh2blk(&f1.DS1EXT1.llimit, geo); + size = cchh2blk(&f1.DS1EXT1.ulimit, geo) - + offset + geo->sectors; + offset *= secperblk; + size *= secperblk; + if (counter >= state->limit) + break; + put_partition(state, counter + 1, offset, size); + counter++; + blk++; + data = read_part_sector(state, blk * secperblk, §); + } + strlcat(state->pp_buf, "\n", PAGE_SIZE); + + if (!data) + return -1; + + return 1; +} + +static int find_lnx1_partitions(struct parsed_partitions *state, + struct hd_geometry *geo, + int blocksize, + char name[], + union label_t *label, + sector_t labelsect, + loff_t i_size, + dasd_information2_t *info) +{ + loff_t offset, geo_size, size; + char tmp[64]; + int secperblk; + + snprintf(tmp, sizeof(tmp), "LNX1/%8s:", name); + strlcat(state->pp_buf, tmp, PAGE_SIZE); + secperblk = blocksize >> 9; + if (label->lnx.ldl_version == 0xf2) { + size = label->lnx.formatted_blocks * secperblk; + } else { + /* + * Formated w/o large volume support. If the sanity check + * 'size based on geo == size based on i_size' is true, then + * we can safely assume that we know the formatted size of + * the disk, otherwise we need additional information + * that we can only get from a real DASD device. + */ + geo_size = geo->cylinders * geo->heads + * geo->sectors * secperblk; + size = i_size >> 9; + if (size != geo_size) { + if (!info) { + strlcat(state->pp_buf, "\n", PAGE_SIZE); + return 1; + } + if (!strcmp(info->type, "ECKD")) + if (geo_size < size) + size = geo_size; + /* else keep size based on i_size */ + } + } + /* first and only partition starts in the first block after the label */ + offset = labelsect + secperblk; + put_partition(state, 1, offset, size - offset); + strlcat(state->pp_buf, "\n", PAGE_SIZE); + return 1; +} + +static int find_cms1_partitions(struct parsed_partitions *state, + struct hd_geometry *geo, + int blocksize, + char name[], + union label_t *label, + sector_t labelsect) +{ + loff_t offset, size; + char tmp[64]; + int secperblk; + + /* + * VM style CMS1 labeled disk + */ + blocksize = label->cms.block_size; + secperblk = blocksize >> 9; + if (label->cms.disk_offset != 0) { + snprintf(tmp, sizeof(tmp), "CMS1/%8s(MDSK):", name); + strlcat(state->pp_buf, tmp, PAGE_SIZE); + /* disk is reserved minidisk */ + offset = label->cms.disk_offset * secperblk; + size = (label->cms.block_count - 1) * secperblk; + } else { + snprintf(tmp, sizeof(tmp), "CMS1/%8s:", name); + strlcat(state->pp_buf, tmp, PAGE_SIZE); + /* + * Special case for FBA devices: + * If an FBA device is CMS formatted with blocksize > 512 byte + * and the DIAG discipline is used, then the CMS label is found + * in sector 1 instead of block 1. However, the partition is + * still supposed to start in block 2. + */ + if (labelsect == 1) + offset = 2 * secperblk; + else + offset = labelsect + secperblk; + size = label->cms.block_count * secperblk; + } + + put_partition(state, 1, offset, size-offset); + strlcat(state->pp_buf, "\n", PAGE_SIZE); + return 1; +} + + /* + * This is the main function, called by check.c */ int ibm_partition(struct parsed_partitions *state) { struct block_device *bdev = state->bdev; int blocksize, res; - loff_t i_size, offset, size, fmt_size; + loff_t i_size, offset, size; dasd_information2_t *info; struct hd_geometry *geo; char type[5] = {0,}; char name[7] = {0,}; - union label_t { - struct vtoc_volume_label_cdl vol; - struct vtoc_volume_label_ldl lnx; - struct vtoc_cms_label cms; - } *label; - unsigned char *data; - Sector sect; sector_t labelsect; - char tmp[64]; + union label_t *label; res = 0; blocksize = bdev_logical_block_size(bdev); @@ -84,7 +305,6 @@ int ibm_partition(struct parsed_partitions *state) i_size = i_size_read(bdev->bd_inode); if (i_size == 0) goto out_exit; - info = kmalloc(sizeof(dasd_information2_t), GFP_KERNEL); if (info == NULL) goto out_exit; @@ -94,176 +314,45 @@ int ibm_partition(struct parsed_partitions *state) label = kmalloc(sizeof(union label_t), GFP_KERNEL); if (label == NULL) goto out_nolab; - - if (ioctl_by_bdev(bdev, BIODASDINFO2, (unsigned long)info) != 0 || - ioctl_by_bdev(bdev, HDIO_GETGEO, (unsigned long)geo) != 0) + if (ioctl_by_bdev(bdev, HDIO_GETGEO, (unsigned long)geo) != 0) goto out_freeall; - - /* - * Special case for FBA disks: label sector does not depend on - * blocksize. - */ - if ((info->cu_type == 0x6310 && info->dev_type == 0x9336) || - (info->cu_type == 0x3880 && info->dev_type == 0x3370)) - labelsect = info->label_block; - else - labelsect = info->label_block * (blocksize >> 9); - - /* - * Get volume label, extract name and type. - */ - data = read_part_sector(state, labelsect, §); - if (data == NULL) - goto out_readerr; - - memcpy(label, data, sizeof(union label_t)); - put_dev_sector(sect); - - if ((!info->FBA_layout) && (!strcmp(info->type, "ECKD"))) { - strncpy(type, label->vol.vollbl, 4); - strncpy(name, label->vol.volid, 6); - } else { - strncpy(type, label->lnx.vollbl, 4); - strncpy(name, label->lnx.volid, 6); + if (ioctl_by_bdev(bdev, BIODASDINFO2, (unsigned long)info) != 0) { + kfree(info); + info = NULL; } - EBCASC(type, 4); - EBCASC(name, 6); - - res = 1; - /* - * Three different formats: LDL, CDL and unformated disk - * - * identified by info->format - * - * unformated disks we do not have to care about - */ - if (info->format == DASD_FORMAT_LDL) { - if (strncmp(type, "CMS1", 4) == 0) { - /* - * VM style CMS1 labeled disk - */ - blocksize = label->cms.block_size; - if (label->cms.disk_offset != 0) { - snprintf(tmp, sizeof(tmp), "CMS1/%8s(MDSK):", name); - strlcat(state->pp_buf, tmp, PAGE_SIZE); - /* disk is reserved minidisk */ - offset = label->cms.disk_offset; - size = (label->cms.block_count - 1) - * (blocksize >> 9); - } else { - snprintf(tmp, sizeof(tmp), "CMS1/%8s:", name); - strlcat(state->pp_buf, tmp, PAGE_SIZE); - offset = (info->label_block + 1); - size = label->cms.block_count - * (blocksize >> 9); - } - put_partition(state, 1, offset*(blocksize >> 9), - size-offset*(blocksize >> 9)); - } else { - if (strncmp(type, "LNX1", 4) == 0) { - snprintf(tmp, sizeof(tmp), "LNX1/%8s:", name); - strlcat(state->pp_buf, tmp, PAGE_SIZE); - if (label->lnx.ldl_version == 0xf2) { - fmt_size = label->lnx.formatted_blocks - * (blocksize >> 9); - } else if (!strcmp(info->type, "ECKD")) { - /* formated w/o large volume support */ - fmt_size = geo->cylinders * geo->heads - * geo->sectors * (blocksize >> 9); - } else { - /* old label and no usable disk geometry - * (e.g. DIAG) */ - fmt_size = i_size >> 9; - } - size = i_size >> 9; - if (fmt_size < size) - size = fmt_size; - offset = (info->label_block + 1); - } else { - /* unlabeled disk */ - strlcat(state->pp_buf, "(nonl)", PAGE_SIZE); - size = i_size >> 9; - offset = (info->label_block + 1); - } - put_partition(state, 1, offset*(blocksize >> 9), - size-offset*(blocksize >> 9)); + if (find_label(state, info, geo, blocksize, &labelsect, name, type, + label)) { + if (!strncmp(type, "VOL1", 4)) { + res = find_vol1_partitions(state, geo, blocksize, name, + label); + } else if (!strncmp(type, "LNX1", 4)) { + res = find_lnx1_partitions(state, geo, blocksize, name, + label, labelsect, i_size, + info); + } else if (!strncmp(type, "CMS1", 4)) { + res = find_cms1_partitions(state, geo, blocksize, name, + label, labelsect); } - } else if (info->format == DASD_FORMAT_CDL) { - /* - * New style CDL formatted disk - */ - sector_t blk; - int counter; - + } else if (info) { /* - * check if VOL1 label is available - * if not, something is wrong, skipping partition detection + * ugly but needed for backward compatibility: + * If the block device is a DASD (i.e. BIODASDINFO2 works), + * then we claim it in any case, even though it has no valid + * label. If it has the LDL format, then we simply define a + * partition as if it had an LNX1 label. */ - if (strncmp(type, "VOL1", 4) == 0) { - snprintf(tmp, sizeof(tmp), "VOL1/%8s:", name); - strlcat(state->pp_buf, tmp, PAGE_SIZE); - /* - * get block number and read then go through format1 - * labels - */ - blk = cchhb2blk(&label->vol.vtoc, geo) + 1; - counter = 0; - data = read_part_sector(state, blk * (blocksize/512), - §); - while (data != NULL) { - struct vtoc_format1_label f1; - - memcpy(&f1, data, - sizeof(struct vtoc_format1_label)); - put_dev_sector(sect); - - /* skip FMT4 / FMT5 / FMT7 labels */ - if (f1.DS1FMTID == _ascebc['4'] - || f1.DS1FMTID == _ascebc['5'] - || f1.DS1FMTID == _ascebc['7'] - || f1.DS1FMTID == _ascebc['9']) { - blk++; - data = read_part_sector(state, - blk * (blocksize/512), §); - continue; - } - - /* only FMT1 and 8 labels valid at this point */ - if (f1.DS1FMTID != _ascebc['1'] && - f1.DS1FMTID != _ascebc['8']) - break; - - /* OK, we got valid partition data */ - offset = cchh2blk(&f1.DS1EXT1.llimit, geo); - size = cchh2blk(&f1.DS1EXT1.ulimit, geo) - - offset + geo->sectors; - if (counter >= state->limit) - break; - put_partition(state, counter + 1, - offset * (blocksize >> 9), - size * (blocksize >> 9)); - counter++; - blk++; - data = read_part_sector(state, - blk * (blocksize/512), §); - } - - if (!data) - /* Are we not supposed to report this ? */ - goto out_readerr; - } else - printk(KERN_INFO "Expected Label VOL1 not " - "found, treating as CDL formated Disk"); - - } - - strlcat(state->pp_buf, "\n", PAGE_SIZE); - goto out_freeall; - + res = 1; + if (info->format == DASD_FORMAT_LDL) { + strlcat(state->pp_buf, "(nonl)", PAGE_SIZE); + size = i_size >> 9; + offset = (info->label_block + 1) * (blocksize >> 9); + put_partition(state, 1, offset, size-offset); + strlcat(state->pp_buf, "\n", PAGE_SIZE); + } + } else + res = 0; -out_readerr: - res = -1; out_freeall: kfree(label); out_nolab: -- cgit v0.10.2 From d1b0d842c4450e410053083db837ef16532a4139 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Sun, 2 Sep 2012 11:02:23 +0200 Subject: s390/mm: rename addressing_mode to s390_user_mode Renaming the globally visible variable "user_mode" to "addressing_mode" in order to fix a name clash was not a good idea. (Commit 37fe1d73 "s390/mm: rename user_mode variable to addressing_mode") Looking at the code after a couple of weeks one thinks: addressing mode of what? So rename the variable again. This time to s390_user_mode. Which hopefully makes more sense. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/mmu_context.h b/arch/s390/include/asm/mmu_context.h index b749c57..084e775 100644 --- a/arch/s390/include/asm/mmu_context.h +++ b/arch/s390/include/asm/mmu_context.h @@ -57,7 +57,7 @@ static inline void update_mm(struct mm_struct *mm, struct task_struct *tsk) pgd_t *pgd = mm->pgd; S390_lowcore.user_asce = mm->context.asce_bits | __pa(pgd); - if (addressing_mode != HOME_SPACE_MODE) { + if (s390_user_mode != HOME_SPACE_MODE) { /* Load primary space page table origin. */ asm volatile(LCTL_OPCODE" 1,1,%0\n" : : "m" (S390_lowcore.user_asce) ); diff --git a/arch/s390/include/asm/setup.h b/arch/s390/include/asm/setup.h index 908f688..87b47ca 100644 --- a/arch/s390/include/asm/setup.h +++ b/arch/s390/include/asm/setup.h @@ -60,7 +60,7 @@ void create_mem_hole(struct mem_chunk memory_chunk[], unsigned long addr, #define SECONDARY_SPACE_MODE 2 #define HOME_SPACE_MODE 3 -extern unsigned int addressing_mode; +extern unsigned int s390_user_mode; /* * Machine features detected in head.S diff --git a/arch/s390/kernel/runtime_instr.c b/arch/s390/kernel/runtime_instr.c index e27e23d..61066f6 100644 --- a/arch/s390/kernel/runtime_instr.c +++ b/arch/s390/kernel/runtime_instr.c @@ -40,7 +40,7 @@ static void disable_runtime_instr(void) static void init_runtime_instr_cb(struct runtime_instr_cb *cb) { cb->buf_limit = 0xfff; - if (addressing_mode == HOME_SPACE_MODE) + if (s390_user_mode == HOME_SPACE_MODE) cb->home_space = 1; cb->int_requested = 1; cb->pstate = 1; diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index 39d609b..555f985 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -302,8 +302,8 @@ static int __init parse_vmalloc(char *arg) } early_param("vmalloc", parse_vmalloc); -unsigned int addressing_mode = PRIMARY_SPACE_MODE; -EXPORT_SYMBOL_GPL(addressing_mode); +unsigned int s390_user_mode = PRIMARY_SPACE_MODE; +EXPORT_SYMBOL_GPL(s390_user_mode); static int set_amode_primary(void) { @@ -326,9 +326,9 @@ static int set_amode_primary(void) static int __init early_parse_user_mode(char *p) { if (p && strcmp(p, "primary") == 0) - addressing_mode = PRIMARY_SPACE_MODE; + s390_user_mode = PRIMARY_SPACE_MODE; else if (!p || strcmp(p, "home") == 0) - addressing_mode = HOME_SPACE_MODE; + s390_user_mode = HOME_SPACE_MODE; else return 1; return 0; @@ -337,7 +337,7 @@ early_param("user_mode", early_parse_user_mode); static void setup_addressing_mode(void) { - if (addressing_mode == PRIMARY_SPACE_MODE) { + if (s390_user_mode == PRIMARY_SPACE_MODE) { if (set_amode_primary()) pr_info("Address spaces switched, " "mvcos available\n"); diff --git a/arch/s390/kernel/vdso.c b/arch/s390/kernel/vdso.c index 9a19ca3..d777628 100644 --- a/arch/s390/kernel/vdso.c +++ b/arch/s390/kernel/vdso.c @@ -85,7 +85,7 @@ struct vdso_data *vdso_data = &vdso_data_store.data; static void vdso_init_data(struct vdso_data *vd) { vd->ectg_available = - addressing_mode != HOME_SPACE_MODE && test_facility(31); + s390_user_mode != HOME_SPACE_MODE && test_facility(31); } #ifdef CONFIG_64BIT @@ -102,7 +102,7 @@ int vdso_alloc_per_cpu(struct _lowcore *lowcore) lowcore->vdso_per_cpu_data = __LC_PASTE; - if (addressing_mode == HOME_SPACE_MODE || !vdso_enabled) + if (s390_user_mode == HOME_SPACE_MODE || !vdso_enabled) return 0; segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER); @@ -147,7 +147,7 @@ void vdso_free_per_cpu(struct _lowcore *lowcore) unsigned long segment_table, page_table, page_frame; u32 *psal, *aste; - if (addressing_mode == HOME_SPACE_MODE || !vdso_enabled) + if (s390_user_mode == HOME_SPACE_MODE || !vdso_enabled) return; psal = (u32 *)(addr_t) lowcore->paste[4]; @@ -165,7 +165,7 @@ static void vdso_init_cr5(void) { unsigned long cr5; - if (addressing_mode == HOME_SPACE_MODE || !vdso_enabled) + if (s390_user_mode == HOME_SPACE_MODE || !vdso_enabled) return; cr5 = offsetof(struct _lowcore, paste); __ctl_load(cr5, 5, 5); diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c index 12c887b..8b2cac1 100644 --- a/arch/s390/mm/fault.c +++ b/arch/s390/mm/fault.c @@ -111,7 +111,7 @@ static inline int user_space_fault(unsigned long trans_exc_code) if (trans_exc_code == 2) /* Access via secondary space, set_fs setting decides */ return current->thread.mm_segment.ar4; - if (addressing_mode == HOME_SPACE_MODE) + if (s390_user_mode == HOME_SPACE_MODE) /* User space if the access has been done via home space. */ return trans_exc_code == 3; /* diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c index 18df31d..d83ecbf 100644 --- a/arch/s390/mm/pgtable.c +++ b/arch/s390/mm/pgtable.c @@ -796,7 +796,7 @@ int s390_enable_sie(void) struct mm_struct *mm, *old_mm; /* Do we have switched amode? If no, we cannot do sie */ - if (addressing_mode == HOME_SPACE_MODE) + if (s390_user_mode == HOME_SPACE_MODE) return -EINVAL; /* Do we have pgstes? if yes, we are done */ -- cgit v0.10.2 From bad4c9c82d7ecb3b6ce4dbe25b7810018b52b8c1 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 3 Sep 2012 07:08:06 +0200 Subject: s390/mm: shorten addressing mode initialization Shorten the code for addressing mode initialization. Also add missing __init annotations, since this code is only used during kernel initialization. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index 555f985..7d27b0a 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -305,7 +305,7 @@ early_param("vmalloc", parse_vmalloc); unsigned int s390_user_mode = PRIMARY_SPACE_MODE; EXPORT_SYMBOL_GPL(s390_user_mode); -static int set_amode_primary(void) +static void __init set_user_mode_primary(void) { psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME; psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY; @@ -313,14 +313,7 @@ static int set_amode_primary(void) psw32_user_bits = (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY; #endif - - if (MACHINE_HAS_MVCOS) { - memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess)); - return 1; - } else { - memcpy(&uaccess, &uaccess_pt, sizeof(uaccess)); - return 0; - } + uaccess = MACHINE_HAS_MVCOS ? uaccess_mvcos_switch : uaccess_pt; } static int __init early_parse_user_mode(char *p) @@ -335,16 +328,15 @@ static int __init early_parse_user_mode(char *p) } early_param("user_mode", early_parse_user_mode); -static void setup_addressing_mode(void) +static void __init setup_addressing_mode(void) { - if (s390_user_mode == PRIMARY_SPACE_MODE) { - if (set_amode_primary()) - pr_info("Address spaces switched, " - "mvcos available\n"); - else - pr_info("Address spaces switched, " - "mvcos not available\n"); - } + if (s390_user_mode != PRIMARY_SPACE_MODE) + return; + set_user_mode_primary(); + if (MACHINE_HAS_MVCOS) + pr_info("Address spaces switched, mvcos available\n"); + else + pr_info("Address spaces switched, mvcos not available\n"); } void *restart_stack __attribute__((__section__(".data"))); -- cgit v0.10.2 From 78609132795b4e3d6d51c6b67d461bf1626afb2f Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 3 Sep 2012 14:11:32 +0200 Subject: s390/topology: remove sysinfo header include, add forward declaration instead Any change to sysinfo.h causes a whole kernel recompile since sysinfo.h is included by topology.h, which again is used nearly everywhere. So remove that include and add a forward declaration instead. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/topology.h b/arch/s390/include/asm/topology.h index 0837de8..1e6c446 100644 --- a/arch/s390/include/asm/topology.h +++ b/arch/s390/include/asm/topology.h @@ -2,8 +2,8 @@ #define _ASM_S390_TOPOLOGY_H #include -#include +struct sysinfo_15_1_x; struct cpu; #ifdef CONFIG_SCHED_BOOK diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c index 05151e0..fc957f2 100644 --- a/arch/s390/kernel/topology.c +++ b/arch/s390/kernel/topology.c @@ -17,6 +17,7 @@ #include #include #include +#include #define PTF_HORIZONTAL (0UL) #define PTF_VERTICAL (1UL) -- cgit v0.10.2 From 0facaa170a6a0255092d3b32325a15672465ac4c Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 3 Sep 2012 09:38:30 +0200 Subject: s390/sysinfo: convert /proc/sysinfo to seqfile The current proc implementation of the /proc/sysinfo file writes all informations contained in all system information blocks to a single page. This is done by calling sprintf all the time in the expectation that everything will fit into a single page. This however is not necessarily true if the configuration of a machine is very large. So convert /proc/sysinfo to avoid writing into random memory regions. For readability reasons a couple of lines are longer than 80 characters. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/sysinfo.c b/arch/s390/kernel/sysinfo.c index fa0eb23..d3b543c 100644 --- a/arch/s390/kernel/sysinfo.c +++ b/arch/s390/kernel/sysinfo.c @@ -25,14 +25,14 @@ static inline int stsi_0(void) { int rc = stsi(NULL, 0, 0, 0); + return rc == -ENOSYS ? rc : (((unsigned int) rc) >> 28); } -static int stsi_1_1_1(struct sysinfo_1_1_1 *info, char *page, int len) +static void stsi_1_1_1(struct seq_file *m, struct sysinfo_1_1_1 *info) { if (stsi(info, 1, 1, 1) == -ENOSYS) - return len; - + return; EBCASC(info->manufacturer, sizeof(info->manufacturer)); EBCASC(info->type, sizeof(info->type)); EBCASC(info->model, sizeof(info->model)); @@ -41,58 +41,45 @@ static int stsi_1_1_1(struct sysinfo_1_1_1 *info, char *page, int len) EBCASC(info->model_capacity, sizeof(info->model_capacity)); EBCASC(info->model_perm_cap, sizeof(info->model_perm_cap)); EBCASC(info->model_temp_cap, sizeof(info->model_temp_cap)); - len += sprintf(page + len, "Manufacturer: %-16.16s\n", - info->manufacturer); - len += sprintf(page + len, "Type: %-4.4s\n", - info->type); + seq_printf(m, "Manufacturer: %-16.16s\n", info->manufacturer); + seq_printf(m, "Type: %-4.4s\n", info->type); + /* + * Sigh: the model field has been renamed with System z9 + * to model_capacity and a new model field has been added + * after the plant field. To avoid confusing older programs + * the "Model:" prints "model_capacity model" or just + * "model_capacity" if the model string is empty . + */ + seq_printf(m, "Model: %-16.16s", info->model_capacity); if (info->model[0] != '\0') - /* - * Sigh: the model field has been renamed with System z9 - * to model_capacity and a new model field has been added - * after the plant field. To avoid confusing older programs - * the "Model:" prints "model_capacity model" or just - * "model_capacity" if the model string is empty . - */ - len += sprintf(page + len, - "Model: %-16.16s %-16.16s\n", - info->model_capacity, info->model); - else - len += sprintf(page + len, "Model: %-16.16s\n", - info->model_capacity); - len += sprintf(page + len, "Sequence Code: %-16.16s\n", - info->sequence); - len += sprintf(page + len, "Plant: %-4.4s\n", - info->plant); - len += sprintf(page + len, "Model Capacity: %-16.16s %08u\n", - info->model_capacity, *(u32 *) info->model_cap_rating); + seq_printf(m, " %-16.16s", info->model); + seq_putc(m, '\n'); + seq_printf(m, "Sequence Code: %-16.16s\n", info->sequence); + seq_printf(m, "Plant: %-4.4s\n", info->plant); + seq_printf(m, "Model Capacity: %-16.16s %08u\n", + info->model_capacity, *(u32 *) info->model_cap_rating); if (info->model_perm_cap[0] != '\0') - len += sprintf(page + len, - "Model Perm. Capacity: %-16.16s %08u\n", - info->model_perm_cap, - *(u32 *) info->model_perm_cap_rating); + seq_printf(m, "Model Perm. Capacity: %-16.16s %08u\n", + info->model_perm_cap, + *(u32 *) info->model_perm_cap_rating); if (info->model_temp_cap[0] != '\0') - len += sprintf(page + len, - "Model Temp. Capacity: %-16.16s %08u\n", - info->model_temp_cap, - *(u32 *) info->model_temp_cap_rating); + seq_printf(m, "Model Temp. Capacity: %-16.16s %08u\n", + info->model_temp_cap, + *(u32 *) info->model_temp_cap_rating); if (info->cai) { - len += sprintf(page + len, - "Capacity Adj. Ind.: %d\n", - info->cai); - len += sprintf(page + len, "Capacity Ch. Reason: %d\n", - info->ccr); + seq_printf(m, "Capacity Adj. Ind.: %d\n", info->cai); + seq_printf(m, "Capacity Ch. Reason: %d\n", info->ccr); } - return len; } -static int stsi_15_1_x(struct sysinfo_15_1_x *info, char *page, int len) +static void stsi_15_1_x(struct seq_file *m, struct sysinfo_15_1_x *info) { static int max_mnest; int i, rc; - len += sprintf(page + len, "\n"); + seq_putc(m, '\n'); if (!MACHINE_HAS_TOPOLOGY) - return len; + return; if (max_mnest) { stsi(info, 15, 1, max_mnest); } else { @@ -102,181 +89,142 @@ static int stsi_15_1_x(struct sysinfo_15_1_x *info, char *page, int len) break; } } - len += sprintf(page + len, "CPU Topology HW: "); + seq_printf(m, "CPU Topology HW: "); for (i = 0; i < TOPOLOGY_NR_MAG; i++) - len += sprintf(page + len, " %d", info->mag[i]); - len += sprintf(page + len, "\n"); + seq_printf(m, " %d", info->mag[i]); + seq_putc(m, '\n'); #ifdef CONFIG_SCHED_MC store_topology(info); - len += sprintf(page + len, "CPU Topology SW: "); + seq_printf(m, "CPU Topology SW: "); for (i = 0; i < TOPOLOGY_NR_MAG; i++) - len += sprintf(page + len, " %d", info->mag[i]); - len += sprintf(page + len, "\n"); + seq_printf(m, " %d", info->mag[i]); + seq_putc(m, '\n'); #endif - return len; } -static int stsi_1_2_2(struct sysinfo_1_2_2 *info, char *page, int len) +static void stsi_1_2_2(struct seq_file *m, struct sysinfo_1_2_2 *info) { struct sysinfo_1_2_2_extension *ext; int i; if (stsi(info, 1, 2, 2) == -ENOSYS) - return len; + return; ext = (struct sysinfo_1_2_2_extension *) ((unsigned long) info + info->acc_offset); - - len += sprintf(page + len, "CPUs Total: %d\n", - info->cpus_total); - len += sprintf(page + len, "CPUs Configured: %d\n", - info->cpus_configured); - len += sprintf(page + len, "CPUs Standby: %d\n", - info->cpus_standby); - len += sprintf(page + len, "CPUs Reserved: %d\n", - info->cpus_reserved); - - if (info->format == 1) { - /* - * Sigh 2. According to the specification the alternate - * capability field is a 32 bit floating point number - * if the higher order 8 bits are not zero. Printing - * a floating point number in the kernel is a no-no, - * always print the number as 32 bit unsigned integer. - * The user-space needs to know about the strange - * encoding of the alternate cpu capability. - */ - len += sprintf(page + len, "Capability: %u %u\n", - info->capability, ext->alt_capability); - for (i = 2; i <= info->cpus_total; i++) - len += sprintf(page + len, - "Adjustment %02d-way: %u %u\n", - i, info->adjustment[i-2], - ext->alt_adjustment[i-2]); - - } else { - len += sprintf(page + len, "Capability: %u\n", - info->capability); - for (i = 2; i <= info->cpus_total; i++) - len += sprintf(page + len, - "Adjustment %02d-way: %u\n", - i, info->adjustment[i-2]); + seq_printf(m, "CPUs Total: %d\n", info->cpus_total); + seq_printf(m, "CPUs Configured: %d\n", info->cpus_configured); + seq_printf(m, "CPUs Standby: %d\n", info->cpus_standby); + seq_printf(m, "CPUs Reserved: %d\n", info->cpus_reserved); + /* + * Sigh 2. According to the specification the alternate + * capability field is a 32 bit floating point number + * if the higher order 8 bits are not zero. Printing + * a floating point number in the kernel is a no-no, + * always print the number as 32 bit unsigned integer. + * The user-space needs to know about the strange + * encoding of the alternate cpu capability. + */ + seq_printf(m, "Capability: %u", info->capability); + if (info->format == 1) + seq_printf(m, " %u", ext->alt_capability); + seq_putc(m, '\n'); + for (i = 2; i <= info->cpus_total; i++) { + seq_printf(m, "Adjustment %02d-way: %u", + i, info->adjustment[i-2]); + if (info->format == 1) + seq_printf(m, " %u", ext->alt_adjustment[i-2]); + seq_putc(m, '\n'); } - - if (info->secondary_capability != 0) - len += sprintf(page + len, "Secondary Capability: %d\n", - info->secondary_capability); - return len; + if (info->secondary_capability) + seq_printf(m, "Secondary Capability: %d\n", + info->secondary_capability); } -static int stsi_2_2_2(struct sysinfo_2_2_2 *info, char *page, int len) +static void stsi_2_2_2(struct seq_file *m, struct sysinfo_2_2_2 *info) { if (stsi(info, 2, 2, 2) == -ENOSYS) - return len; - + return; EBCASC(info->name, sizeof(info->name)); - - len += sprintf(page + len, "\n"); - len += sprintf(page + len, "LPAR Number: %d\n", - info->lpar_number); - - len += sprintf(page + len, "LPAR Characteristics: "); + seq_putc(m, '\n'); + seq_printf(m, "LPAR Number: %d\n", info->lpar_number); + seq_printf(m, "LPAR Characteristics: "); if (info->characteristics & LPAR_CHAR_DEDICATED) - len += sprintf(page + len, "Dedicated "); + seq_printf(m, "Dedicated "); if (info->characteristics & LPAR_CHAR_SHARED) - len += sprintf(page + len, "Shared "); + seq_printf(m, "Shared "); if (info->characteristics & LPAR_CHAR_LIMITED) - len += sprintf(page + len, "Limited "); - len += sprintf(page + len, "\n"); - - len += sprintf(page + len, "LPAR Name: %-8.8s\n", - info->name); - - len += sprintf(page + len, "LPAR Adjustment: %d\n", - info->caf); - - len += sprintf(page + len, "LPAR CPUs Total: %d\n", - info->cpus_total); - len += sprintf(page + len, "LPAR CPUs Configured: %d\n", - info->cpus_configured); - len += sprintf(page + len, "LPAR CPUs Standby: %d\n", - info->cpus_standby); - len += sprintf(page + len, "LPAR CPUs Reserved: %d\n", - info->cpus_reserved); - len += sprintf(page + len, "LPAR CPUs Dedicated: %d\n", - info->cpus_dedicated); - len += sprintf(page + len, "LPAR CPUs Shared: %d\n", - info->cpus_shared); - return len; + seq_printf(m, "Limited "); + seq_putc(m, '\n'); + seq_printf(m, "LPAR Name: %-8.8s\n", info->name); + seq_printf(m, "LPAR Adjustment: %d\n", info->caf); + seq_printf(m, "LPAR CPUs Total: %d\n", info->cpus_total); + seq_printf(m, "LPAR CPUs Configured: %d\n", info->cpus_configured); + seq_printf(m, "LPAR CPUs Standby: %d\n", info->cpus_standby); + seq_printf(m, "LPAR CPUs Reserved: %d\n", info->cpus_reserved); + seq_printf(m, "LPAR CPUs Dedicated: %d\n", info->cpus_dedicated); + seq_printf(m, "LPAR CPUs Shared: %d\n", info->cpus_shared); } -static int stsi_3_2_2(struct sysinfo_3_2_2 *info, char *page, int len) +static void stsi_3_2_2(struct seq_file *m, struct sysinfo_3_2_2 *info) { int i; if (stsi(info, 3, 2, 2) == -ENOSYS) - return len; + return; for (i = 0; i < info->count; i++) { EBCASC(info->vm[i].name, sizeof(info->vm[i].name)); EBCASC(info->vm[i].cpi, sizeof(info->vm[i].cpi)); - len += sprintf(page + len, "\n"); - len += sprintf(page + len, "VM%02d Name: %-8.8s\n", - i, info->vm[i].name); - len += sprintf(page + len, "VM%02d Control Program: %-16.16s\n", - i, info->vm[i].cpi); - - len += sprintf(page + len, "VM%02d Adjustment: %d\n", - i, info->vm[i].caf); - - len += sprintf(page + len, "VM%02d CPUs Total: %d\n", - i, info->vm[i].cpus_total); - len += sprintf(page + len, "VM%02d CPUs Configured: %d\n", - i, info->vm[i].cpus_configured); - len += sprintf(page + len, "VM%02d CPUs Standby: %d\n", - i, info->vm[i].cpus_standby); - len += sprintf(page + len, "VM%02d CPUs Reserved: %d\n", - i, info->vm[i].cpus_reserved); + seq_putc(m, '\n'); + seq_printf(m, "VM%02d Name: %-8.8s\n", i, info->vm[i].name); + seq_printf(m, "VM%02d Control Program: %-16.16s\n", i, info->vm[i].cpi); + seq_printf(m, "VM%02d Adjustment: %d\n", i, info->vm[i].caf); + seq_printf(m, "VM%02d CPUs Total: %d\n", i, info->vm[i].cpus_total); + seq_printf(m, "VM%02d CPUs Configured: %d\n", i, info->vm[i].cpus_configured); + seq_printf(m, "VM%02d CPUs Standby: %d\n", i, info->vm[i].cpus_standby); + seq_printf(m, "VM%02d CPUs Reserved: %d\n", i, info->vm[i].cpus_reserved); } - return len; } -static int proc_read_sysinfo(char *page, char **start, - off_t off, int count, - int *eof, void *data) +static int sysinfo_show(struct seq_file *m, void *v) { - unsigned long info = get_zeroed_page(GFP_KERNEL); - int level, len; + void *info = (void *)get_zeroed_page(GFP_KERNEL); + int level; if (!info) return 0; - - len = 0; level = stsi_0(); if (level >= 1) - len = stsi_1_1_1((struct sysinfo_1_1_1 *) info, page, len); - + stsi_1_1_1(m, info); if (level >= 1) - len = stsi_15_1_x((struct sysinfo_15_1_x *) info, page, len); - + stsi_15_1_x(m, info); if (level >= 1) - len = stsi_1_2_2((struct sysinfo_1_2_2 *) info, page, len); - + stsi_1_2_2(m, info); if (level >= 2) - len = stsi_2_2_2((struct sysinfo_2_2_2 *) info, page, len); - + stsi_2_2_2(m, info); if (level >= 3) - len = stsi_3_2_2((struct sysinfo_3_2_2 *) info, page, len); + stsi_3_2_2(m, info); + free_page((unsigned long)info); + return 0; +} - free_page(info); - return len; +static int sysinfo_open(struct inode *inode, struct file *file) +{ + return single_open(file, sysinfo_show, NULL); } -static __init int create_proc_sysinfo(void) +static const struct file_operations sysinfo_fops = { + .open = sysinfo_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static int __init sysinfo_create_proc(void) { - create_proc_read_entry("sysinfo", 0444, NULL, - proc_read_sysinfo, NULL); + proc_create("sysinfo", 0444, NULL, &sysinfo_fops); return 0; } -device_initcall(create_proc_sysinfo); +device_initcall(sysinfo_create_proc); /* * Service levels interface. -- cgit v0.10.2 From 25502f0015a8d6dd4bb34336ddb3eac6b1a55317 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 3 Sep 2012 14:05:05 +0200 Subject: s390/sysinfo: add additional z196 fields to output Add a couple of missing fields that were introduced with z196. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/sysinfo.h b/arch/s390/include/asm/sysinfo.h index 282ee36..12e5256 100644 --- a/arch/s390/include/asm/sysinfo.h +++ b/arch/s390/include/asm/sysinfo.h @@ -17,7 +17,10 @@ #include struct sysinfo_1_1_1 { - unsigned short :16; + unsigned char p:1; + unsigned char :6; + unsigned char t:1; + unsigned char :8; unsigned char ccr; unsigned char cai; char reserved_0[28]; @@ -30,9 +33,14 @@ struct sysinfo_1_1_1 { char model[16]; char model_perm_cap[16]; char model_temp_cap[16]; - char model_cap_rating[4]; - char model_perm_cap_rating[4]; - char model_temp_cap_rating[4]; + unsigned int model_cap_rating; + unsigned int model_perm_cap_rating; + unsigned int model_temp_cap_rating; + unsigned char typepct[5]; + unsigned char reserved_2[3]; + unsigned int ncr; + unsigned int npr; + unsigned int ntr; }; struct sysinfo_1_2_1 { @@ -47,8 +55,9 @@ struct sysinfo_1_2_2 { char format; char reserved_0[1]; unsigned short acc_offset; - char reserved_1[24]; - unsigned int secondary_capability; + char reserved_1[20]; + unsigned int nominal_cap; + unsigned int secondary_cap; unsigned int capability; unsigned short cpus_total; unsigned short cpus_configured; diff --git a/arch/s390/kernel/sysinfo.c b/arch/s390/kernel/sysinfo.c index d3b543c..2249b0c 100644 --- a/arch/s390/kernel/sysinfo.c +++ b/arch/s390/kernel/sysinfo.c @@ -31,6 +31,8 @@ static inline int stsi_0(void) static void stsi_1_1_1(struct seq_file *m, struct sysinfo_1_1_1 *info) { + int i; + if (stsi(info, 1, 1, 1) == -ENOSYS) return; EBCASC(info->manufacturer, sizeof(info->manufacturer)); @@ -57,18 +59,31 @@ static void stsi_1_1_1(struct seq_file *m, struct sysinfo_1_1_1 *info) seq_printf(m, "Sequence Code: %-16.16s\n", info->sequence); seq_printf(m, "Plant: %-4.4s\n", info->plant); seq_printf(m, "Model Capacity: %-16.16s %08u\n", - info->model_capacity, *(u32 *) info->model_cap_rating); - if (info->model_perm_cap[0] != '\0') + info->model_capacity, info->model_cap_rating); + if (info->model_perm_cap_rating) seq_printf(m, "Model Perm. Capacity: %-16.16s %08u\n", info->model_perm_cap, - *(u32 *) info->model_perm_cap_rating); - if (info->model_temp_cap[0] != '\0') + info->model_perm_cap_rating); + if (info->model_temp_cap_rating) seq_printf(m, "Model Temp. Capacity: %-16.16s %08u\n", info->model_temp_cap, - *(u32 *) info->model_temp_cap_rating); + info->model_temp_cap_rating); + if (info->ncr) + seq_printf(m, "Nominal Cap. Rating: %08u\n", info->ncr); + if (info->npr) + seq_printf(m, "Nominal Perm. Rating: %08u\n", info->npr); + if (info->ntr) + seq_printf(m, "Nominal Temp. Rating: %08u\n", info->ntr); if (info->cai) { seq_printf(m, "Capacity Adj. Ind.: %d\n", info->cai); seq_printf(m, "Capacity Ch. Reason: %d\n", info->ccr); + seq_printf(m, "Capacity Transient: %d\n", info->t); + } + if (info->p) { + for (i = 1; i <= ARRAY_SIZE(info->typepct); i++) { + seq_printf(m, "Type %d Percentage: %d\n", + i, info->typepct[i - 1]); + } } } @@ -128,6 +143,10 @@ static void stsi_1_2_2(struct seq_file *m, struct sysinfo_1_2_2 *info) if (info->format == 1) seq_printf(m, " %u", ext->alt_capability); seq_putc(m, '\n'); + if (info->nominal_cap) + seq_printf(m, "Nominal Capability: %d\n", info->nominal_cap); + if (info->secondary_cap) + seq_printf(m, "Secondary Capability: %d\n", info->secondary_cap); for (i = 2; i <= info->cpus_total; i++) { seq_printf(m, "Adjustment %02d-way: %u", i, info->adjustment[i-2]); @@ -135,9 +154,6 @@ static void stsi_1_2_2(struct seq_file *m, struct sysinfo_1_2_2 *info) seq_printf(m, " %u", ext->alt_adjustment[i-2]); seq_putc(m, '\n'); } - if (info->secondary_capability) - seq_printf(m, "Secondary Capability: %d\n", - info->secondary_capability); } static void stsi_2_2_2(struct seq_file *m, struct sysinfo_2_2_2 *info) -- cgit v0.10.2 From 34cda99260247873df53ae00885fb0f426b149a5 Mon Sep 17 00:00:00 2001 From: Gerald Schaefer Date: Tue, 4 Sep 2012 15:37:55 +0200 Subject: s390/mm: implement __get_user_pages_fast() This patch introduces the __get_user_pages_fast() function on s390, which will be needed with CONFIG_TRANSPARENT_HUGEPAGE and futex. Signed-off-by: Gerald Schaefer Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/mm/gup.c b/arch/s390/mm/gup.c index 65cb06e..eeaf802 100644 --- a/arch/s390/mm/gup.c +++ b/arch/s390/mm/gup.c @@ -154,6 +154,43 @@ static inline int gup_pud_range(pgd_t *pgdp, pgd_t pgd, unsigned long addr, return 1; } +/* + * Like get_user_pages_fast() except its IRQ-safe in that it won't fall + * back to the regular GUP. + */ +int __get_user_pages_fast(unsigned long start, int nr_pages, int write, + struct page **pages) +{ + struct mm_struct *mm = current->mm; + unsigned long addr, len, end; + unsigned long next, flags; + pgd_t *pgdp, pgd; + int nr = 0; + + start &= PAGE_MASK; + addr = start; + len = (unsigned long) nr_pages << PAGE_SHIFT; + end = start + len; + if (unlikely(!access_ok(write ? VERIFY_WRITE : VERIFY_READ, + (void __user *)start, len))) + return 0; + + local_irq_save(flags); + pgdp = pgd_offset(mm, addr); + do { + pgd = *pgdp; + barrier(); + next = pgd_addr_end(addr, end); + if (pgd_none(pgd)) + break; + if (!gup_pud_range(pgdp, pgd, addr, next, write, pages, &nr)) + break; + } while (pgdp++, addr = next, addr != end); + local_irq_restore(flags); + + return nr; +} + /** * get_user_pages_fast() - pin user pages in memory * @start: starting user address -- cgit v0.10.2 From fade4dc49101e3b68fb375fd2b00d0ef1f31a36f Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 4 Sep 2012 14:26:03 +0200 Subject: s390/sysinfo,topology: fix cpu topology maximum nesting detection The maximum nesting of the cpu topology is evaluated when /proc/sysinfo is the first time read. This happens without a lock and a concurrent reader on a different cpu can see and use an invalid intermediate value. Besides the fact that this race is quite unlikely the worst thing that could happen is that /proc/sysinfo would contain bogus information about the machine's cpu topology. Nevertheless this should be fixed. So move the detection code to the early machine detection code and since now the value is early available use it in the topology code as well. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/sysinfo.h b/arch/s390/include/asm/sysinfo.h index 12e5256..04e6e97 100644 --- a/arch/s390/include/asm/sysinfo.h +++ b/arch/s390/include/asm/sysinfo.h @@ -118,6 +118,8 @@ struct sysinfo_3_2_2 { char reserved_544[3552]; }; +extern int topology_max_mnest; + #define TOPOLOGY_CPU_BITS 64 #define TOPOLOGY_NR_MAG 6 diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c index 1345ba4..f4bcdc0 100644 --- a/arch/s390/kernel/early.c +++ b/arch/s390/kernel/early.c @@ -215,26 +215,44 @@ static noinline __init void init_kernel_storage_key(void) PAGE_DEFAULT_KEY, 0); } -static __initdata struct sysinfo_3_2_2 vmms __aligned(PAGE_SIZE); +static __initdata char sysinfo_page[PAGE_SIZE] __aligned(PAGE_SIZE); static noinline __init void detect_machine_type(void) { + struct sysinfo_3_2_2 *vmms = (struct sysinfo_3_2_2 *)&sysinfo_page; + /* Check current-configuration-level */ if ((stsi(NULL, 0, 0, 0) >> 28) <= 2) { S390_lowcore.machine_flags |= MACHINE_FLAG_LPAR; return; } /* Get virtual-machine cpu information. */ - if (stsi(&vmms, 3, 2, 2) == -ENOSYS || !vmms.count) + if (stsi(vmms, 3, 2, 2) == -ENOSYS || !vmms->count) return; /* Running under KVM? If not we assume z/VM */ - if (!memcmp(vmms.vm[0].cpi, "\xd2\xe5\xd4", 3)) + if (!memcmp(vmms->vm[0].cpi, "\xd2\xe5\xd4", 3)) S390_lowcore.machine_flags |= MACHINE_FLAG_KVM; else S390_lowcore.machine_flags |= MACHINE_FLAG_VM; } +static __init void setup_topology(void) +{ +#ifdef CONFIG_64BIT + int max_mnest; + + if (!test_facility(11)) + return; + S390_lowcore.machine_flags |= MACHINE_FLAG_TOPOLOGY; + for (max_mnest = 6; max_mnest > 1; max_mnest--) { + if (stsi(&sysinfo_page, 15, 1, max_mnest) != -ENOSYS) + break; + } + topology_max_mnest = max_mnest; +#endif +} + static void early_pgm_check_handler(void) { unsigned long addr; @@ -364,8 +382,6 @@ static __init void detect_machine_facilities(void) S390_lowcore.machine_flags |= MACHINE_FLAG_IDTE; if (test_facility(8)) S390_lowcore.machine_flags |= MACHINE_FLAG_PFMF; - if (test_facility(11)) - S390_lowcore.machine_flags |= MACHINE_FLAG_TOPOLOGY; if (test_facility(27)) S390_lowcore.machine_flags |= MACHINE_FLAG_MVCOS; if (test_facility(40)) @@ -467,6 +483,7 @@ void __init startup_init(void) detect_diag44(); detect_machine_facilities(); setup_hpage(); + setup_topology(); sclp_facilities_detect(); detect_memory_layout(memory_chunk); #ifdef CONFIG_DYNAMIC_FTRACE diff --git a/arch/s390/kernel/sysinfo.c b/arch/s390/kernel/sysinfo.c index 2249b0c..2af4ee6 100644 --- a/arch/s390/kernel/sysinfo.c +++ b/arch/s390/kernel/sysinfo.c @@ -22,6 +22,8 @@ #include #include +int topology_max_mnest; + static inline int stsi_0(void) { int rc = stsi(NULL, 0, 0, 0); @@ -95,15 +97,7 @@ static void stsi_15_1_x(struct seq_file *m, struct sysinfo_15_1_x *info) seq_putc(m, '\n'); if (!MACHINE_HAS_TOPOLOGY) return; - if (max_mnest) { - stsi(info, 15, 1, max_mnest); - } else { - for (max_mnest = 6; max_mnest > 1; max_mnest--) { - rc = stsi(info, 15, 1, max_mnest); - if (rc != -ENOSYS) - break; - } - } + stsi(info, 15, 1, topology_max_mnest); seq_printf(m, "CPU Topology HW: "); for (i = 0; i < TOPOLOGY_NR_MAG; i++) seq_printf(m, " %d", info->mag[i]); diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c index fc957f2..80ca4ba 100644 --- a/arch/s390/kernel/topology.c +++ b/arch/s390/kernel/topology.c @@ -251,12 +251,10 @@ static void update_cpu_core_map(void) void store_topology(struct sysinfo_15_1_x *info) { - int rc; - - rc = stsi(info, 15, 1, 3); - if (rc != -ENOSYS) - return; - stsi(info, 15, 1, 2); + if (topology_max_mnest >= 3) + stsi(info, 15, 1, 3); + else + stsi(info, 15, 1, 2); } int arch_update_cpu_topology(void) -- cgit v0.10.2 From 50ab9a9a60fc83b8e8db36b54f365226e2b139ac Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 4 Sep 2012 17:36:16 +0200 Subject: s390/smp,topology: add polarization member to pcpu struct The cpu polarization member is the only per cpu state that is not part of the pcpu structure. So add it there and have everything in one place. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/smp.h b/arch/s390/include/asm/smp.h index da7dcd7..b64f15c 100644 --- a/arch/s390/include/asm/smp.h +++ b/arch/s390/include/asm/smp.h @@ -30,6 +30,8 @@ extern int smp_vcpu_scheduled(int cpu); extern void smp_yield_cpu(int cpu); extern void smp_yield(void); extern void smp_stop_cpu(void); +extern void smp_cpu_set_polarization(int cpu, int val); +extern int smp_cpu_get_polarization(int cpu); #else /* CONFIG_SMP */ diff --git a/arch/s390/include/asm/topology.h b/arch/s390/include/asm/topology.h index 1e6c446..9ca3053 100644 --- a/arch/s390/include/asm/topology.h +++ b/arch/s390/include/asm/topology.h @@ -51,24 +51,6 @@ static inline void topology_expect_change(void) { } #define POLARIZATION_VM (2) #define POLARIZATION_VH (3) -extern int cpu_polarization[]; - -static inline void cpu_set_polarization(int cpu, int val) -{ -#ifdef CONFIG_SCHED_BOOK - cpu_polarization[cpu] = val; -#endif -} - -static inline int cpu_read_polarization(int cpu) -{ -#ifdef CONFIG_SCHED_BOOK - return cpu_polarization[cpu]; -#else - return POLARIZATION_HRZ; -#endif -} - #ifdef CONFIG_SCHED_BOOK void s390_init_cpu_topology(void); #else diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index c88fe85..8676aab 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -66,6 +66,7 @@ struct pcpu { unsigned long panic_stack; /* panic stack for the cpu */ unsigned long ec_mask; /* bit mask for ec_xxx functions */ int state; /* physical cpu state */ + int polarization; /* physical polarization */ u16 address; /* physical cpu address */ }; @@ -73,6 +74,10 @@ static u8 boot_cpu_type; static u16 boot_cpu_address; static struct pcpu pcpu_devices[NR_CPUS]; +/* + * The smp_cpu_state_mutex must be held when changing the state or polarization + * member of a pcpu data structure within the pcpu_devices arreay. + */ DEFINE_MUTEX(smp_cpu_state_mutex); /* @@ -587,6 +592,16 @@ static inline void smp_get_save_area(int cpu, u16 address) { } #endif /* CONFIG_ZFCPDUMP || CONFIG_CRASH_DUMP */ +void smp_cpu_set_polarization(int cpu, int val) +{ + pcpu_devices[cpu].polarization = val; +} + +int smp_cpu_get_polarization(int cpu) +{ + return pcpu_devices[cpu].polarization; +} + static struct sclp_cpu_info *smp_get_cpu_info(void) { static int use_sigp_detection; @@ -629,7 +644,7 @@ static int __devinit __smp_rescan_cpus(struct sclp_cpu_info *info, pcpu->address = info->cpu[i].address; pcpu->state = (cpu >= info->configured) ? CPU_STATE_STANDBY : CPU_STATE_CONFIGURED; - cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); set_cpu_present(cpu, true); if (sysfs_add && smp_add_present_cpu(cpu) != 0) set_cpu_present(cpu, false); @@ -797,7 +812,7 @@ void __init smp_prepare_boot_cpu(void) pcpu->async_stack = S390_lowcore.async_stack - ASYNC_SIZE; pcpu->panic_stack = S390_lowcore.panic_stack - PAGE_SIZE; S390_lowcore.percpu_offset = __per_cpu_offset[0]; - cpu_set_polarization(0, POLARIZATION_UNKNOWN); + smp_cpu_set_polarization(0, POLARIZATION_UNKNOWN); set_cpu_present(0, true); set_cpu_online(0, true); } @@ -863,7 +878,7 @@ static ssize_t cpu_configure_store(struct device *dev, if (rc) break; pcpu->state = CPU_STATE_STANDBY; - cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); topology_expect_change(); break; case 1: @@ -873,7 +888,7 @@ static ssize_t cpu_configure_store(struct device *dev, if (rc) break; pcpu->state = CPU_STATE_CONFIGURED; - cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); topology_expect_change(); break; default: diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c index 80ca4ba..30c8533 100644 --- a/arch/s390/kernel/topology.c +++ b/arch/s390/kernel/topology.c @@ -45,9 +45,6 @@ static struct mask_info book_info; cpumask_t cpu_book_map[NR_CPUS]; unsigned char cpu_book_id[NR_CPUS]; -/* smp_cpu_state_mutex must be held when accessing this array */ -int cpu_polarization[NR_CPUS]; - static cpumask_t cpu_group_map(struct mask_info *info, unsigned int cpu) { cpumask_t mask; @@ -95,7 +92,7 @@ static struct mask_info *add_cpus_to_mask(struct topology_cpu *tl_cpu, } else { cpu_core_id[lcpu] = core->id; } - cpu_set_polarization(lcpu, tl_cpu->pp); + smp_cpu_set_polarization(lcpu, tl_cpu->pp); } } return core; @@ -202,7 +199,7 @@ static void topology_update_polarization_simple(void) mutex_lock(&smp_cpu_state_mutex); for_each_possible_cpu(cpu) - cpu_set_polarization(cpu, POLARIZATION_HRZ); + smp_cpu_set_polarization(cpu, POLARIZATION_HRZ); mutex_unlock(&smp_cpu_state_mutex); } @@ -232,7 +229,7 @@ int topology_set_cpu_management(int fc) if (rc) return -EBUSY; for_each_possible_cpu(cpu) - cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); return rc; } @@ -414,7 +411,7 @@ static ssize_t cpu_polarization_show(struct device *dev, ssize_t count; mutex_lock(&smp_cpu_state_mutex); - switch (cpu_read_polarization(cpu)) { + switch (smp_cpu_get_polarization(cpu)) { case POLARIZATION_HRZ: count = sprintf(buf, "horizontal\n"); break; -- cgit v0.10.2 From c3e6d407c0c09fb46a391bdd0a28827472825de7 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 4 Sep 2012 19:36:41 +0200 Subject: s390/scm: remove superfluous lock Remove the spinlock from struct scm_device. drvdata and attributes are guarded via device_lock. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/eadm.h b/arch/s390/include/asm/eadm.h index 4d6e103..8d48471 100644 --- a/arch/s390/include/asm/eadm.h +++ b/arch/s390/include/asm/eadm.h @@ -3,7 +3,6 @@ #include #include -#include struct arqb { u64 data; @@ -83,7 +82,6 @@ struct scm_device { u64 size; unsigned int nr_max_block; struct device dev; - spinlock_t lock; struct { unsigned int persistence:4; unsigned int oper_state:4; diff --git a/drivers/s390/block/scm_drv.c b/drivers/s390/block/scm_drv.c index fce711a..349966f 100644 --- a/drivers/s390/block/scm_drv.c +++ b/drivers/s390/block/scm_drv.c @@ -9,7 +9,6 @@ #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt #include -#include #include #include #include "scm_blk.h" @@ -37,15 +36,10 @@ static int scm_probe(struct scm_device *scmdev) if (!bdev) return -ENOMEM; - spin_lock_irq(&scmdev->lock); dev_set_drvdata(&scmdev->dev, bdev); - spin_unlock_irq(&scmdev->lock); - ret = scm_blk_dev_setup(bdev, scmdev); if (ret) { - spin_lock_irq(&scmdev->lock); dev_set_drvdata(&scmdev->dev, NULL); - spin_unlock_irq(&scmdev->lock); kfree(bdev); goto out; } @@ -56,12 +50,9 @@ out: static int scm_remove(struct scm_device *scmdev) { - struct scm_blk_dev *bdev; + struct scm_blk_dev *bdev = dev_get_drvdata(&scmdev->dev); - spin_lock_irq(&scmdev->lock); - bdev = dev_get_drvdata(&scmdev->dev); dev_set_drvdata(&scmdev->dev, NULL); - spin_unlock_irq(&scmdev->lock); scm_blk_dev_cleanup(bdev); kfree(bdev); diff --git a/drivers/s390/cio/scm.c b/drivers/s390/cio/scm.c index aa4476e..bcf20f3 100644 --- a/drivers/s390/cio/scm.c +++ b/drivers/s390/cio/scm.c @@ -5,7 +5,6 @@ * Author(s): Sebastian Ott */ -#include #include #include #include @@ -134,9 +133,9 @@ static ssize_t show_##name(struct device *dev, \ struct scm_device *scmdev = to_scm_dev(dev); \ int ret; \ \ - spin_lock(&scmdev->lock); \ + device_lock(dev); \ ret = sprintf(buf, "%u\n", scmdev->attrs.name); \ - spin_unlock(&scmdev->lock); \ + device_unlock(dev); \ \ return ret; \ } \ @@ -193,7 +192,6 @@ static void scmdev_setup(struct scm_device *scmdev, struct sale *sale, scmdev->dev.bus = &scm_bus_type; scmdev->dev.release = scmdev_release; scmdev->dev.groups = scmdev_attr_groups; - spin_lock_init(&scmdev->lock); } /* -- cgit v0.10.2 From 24996edce547fd981c089db9a12717fd76a51160 Mon Sep 17 00:00:00 2001 From: Sebastian Ott Date: Tue, 4 Sep 2012 19:37:51 +0200 Subject: s390/scm: reorder scm_remove Do not reset drvdata before the block device is cleaned up. With a non-empty block queue drvdata could be accessed. Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/block/scm_drv.c b/drivers/s390/block/scm_drv.c index 349966f..9fa0a90 100644 --- a/drivers/s390/block/scm_drv.c +++ b/drivers/s390/block/scm_drv.c @@ -52,8 +52,8 @@ static int scm_remove(struct scm_device *scmdev) { struct scm_blk_dev *bdev = dev_get_drvdata(&scmdev->dev); - dev_set_drvdata(&scmdev->dev, NULL); scm_blk_dev_cleanup(bdev); + dev_set_drvdata(&scmdev->dev, NULL); kfree(bdev); return 0; -- cgit v0.10.2 From eb608fb366de123a97227437e5306f731f4a63c5 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Wed, 5 Sep 2012 13:26:11 +0200 Subject: s390/exceptions: switch to relative exception table entries This is the s390 port of 70627654 "x86, extable: Switch to relative exception table entries". Reduces the size of our exception tables by 50% on 64 bit builds. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h index 7e81ff1..f3e0aab 100644 --- a/arch/s390/include/asm/processor.h +++ b/arch/s390/include/asm/processor.h @@ -11,6 +11,8 @@ #ifndef __ASM_S390_PROCESSOR_H #define __ASM_S390_PROCESSOR_H +#ifndef __ASSEMBLY__ + #include #include #include @@ -348,23 +350,6 @@ extern void (*s390_base_ext_handler_fn)(void); #define ARCH_LOW_ADDRESS_LIMIT 0x7fffffffUL -/* - * Helper macro for exception table entries - */ -#ifndef CONFIG_64BIT -#define EX_TABLE(_fault,_target) \ - ".section __ex_table,\"a\"\n" \ - " .align 4\n" \ - " .long " #_fault "," #_target "\n" \ - ".previous\n" -#else -#define EX_TABLE(_fault,_target) \ - ".section __ex_table,\"a\"\n" \ - " .align 8\n" \ - " .quad " #_fault "," #_target "\n" \ - ".previous\n" -#endif - extern int memcpy_real(void *, void *, size_t); extern void memcpy_absolute(void *, void *, size_t); @@ -375,4 +360,25 @@ extern void memcpy_absolute(void *, void *, size_t); memcpy_absolute(&(dest), &__tmp, sizeof(__tmp)); \ } -#endif /* __ASM_S390_PROCESSOR_H */ +/* + * Helper macro for exception table entries + */ +#define EX_TABLE(_fault, _target) \ + ".section __ex_table,\"a\"\n" \ + ".align 4\n" \ + ".long (" #_fault ") - .\n" \ + ".long (" #_target ") - .\n" \ + ".previous\n" + +#else /* __ASSEMBLY__ */ + +#define EX_TABLE(_fault, _target) \ + .section __ex_table,"a" ; \ + .align 4 ; \ + .long (_fault) - . ; \ + .long (_target) - . ; \ + .previous + +#endif /* __ASSEMBLY__ */ + +#endif /* __ASM_S390_PROCESSOR_H */ diff --git a/arch/s390/include/asm/uaccess.h b/arch/s390/include/asm/uaccess.h index a8ab18b..34268df 100644 --- a/arch/s390/include/asm/uaccess.h +++ b/arch/s390/include/asm/uaccess.h @@ -76,9 +76,22 @@ static inline int __range_ok(unsigned long addr, unsigned long size) struct exception_table_entry { - unsigned long insn, fixup; + int insn, fixup; }; +static inline unsigned long extable_insn(const struct exception_table_entry *x) +{ + return (unsigned long)&x->insn + x->insn; +} + +static inline unsigned long extable_fixup(const struct exception_table_entry *x) +{ + return (unsigned long)&x->fixup + x->fixup; +} + +#define ARCH_HAS_SORT_EXTABLE +#define ARCH_HAS_SEARCH_EXTABLE + struct uaccess_ops { size_t (*copy_from_user)(size_t, const void __user *, void *); size_t (*copy_from_user_small)(size_t, const void __user *, void *); diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c index f4bcdc0..e8000d5 100644 --- a/arch/s390/kernel/early.c +++ b/arch/s390/kernel/early.c @@ -255,14 +255,14 @@ static __init void setup_topology(void) static void early_pgm_check_handler(void) { - unsigned long addr; const struct exception_table_entry *fixup; + unsigned long addr; addr = S390_lowcore.program_old_psw.addr; fixup = search_exception_tables(addr & PSW_ADDR_INSN); if (!fixup) disabled_wait(0); - S390_lowcore.program_old_psw.addr = fixup->fixup | PSW_ADDR_AMODE; + S390_lowcore.program_old_psw.addr = extable_fixup(fixup)|PSW_ADDR_AMODE; } static noinline __init void setup_lowcore_early(void) diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S index 95e9d93..7549985 100644 --- a/arch/s390/kernel/entry64.S +++ b/arch/s390/kernel/entry64.S @@ -10,6 +10,7 @@ #include #include +#include #include #include #include @@ -1008,9 +1009,7 @@ sie_fault: .Lhost_id: .quad 0 - .section __ex_table,"a" - .quad sie_loop,sie_fault - .previous + EX_TABLE(sie_loop,sie_fault) #endif .section .rodata, "a" diff --git a/arch/s390/kernel/kprobes.c b/arch/s390/kernel/kprobes.c index 8aa634f..d1c7214 100644 --- a/arch/s390/kernel/kprobes.c +++ b/arch/s390/kernel/kprobes.c @@ -547,7 +547,7 @@ static int __kprobes kprobe_trap_handler(struct pt_regs *regs, int trapnr) */ entry = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN); if (entry) { - regs->psw.addr = entry->fixup | PSW_ADDR_AMODE; + regs->psw.addr = extable_fixup(entry) | PSW_ADDR_AMODE; return 1; } diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c index 29af526..befc761 100644 --- a/arch/s390/kernel/traps.c +++ b/arch/s390/kernel/traps.c @@ -322,7 +322,7 @@ static void __kprobes do_trap(struct pt_regs *regs, const struct exception_table_entry *fixup; fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN); if (fixup) - regs->psw.addr = fixup->fixup | PSW_ADDR_AMODE; + regs->psw.addr = extable_fixup(fixup) | PSW_ADDR_AMODE; else { enum bug_trap_type btt; diff --git a/arch/s390/mm/Makefile b/arch/s390/mm/Makefile index d98fe90..0f5536b 100644 --- a/arch/s390/mm/Makefile +++ b/arch/s390/mm/Makefile @@ -3,7 +3,7 @@ # obj-y := init.o fault.o extmem.o mmap.o vmem.o pgtable.o maccess.o \ - page-states.o gup.o + page-states.o gup.o extable.o obj-$(CONFIG_CMM) += cmm.o obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o obj-$(CONFIG_DEBUG_SET_MODULE_RONX) += pageattr.o diff --git a/arch/s390/mm/extable.c b/arch/s390/mm/extable.c new file mode 100644 index 0000000..4d1ee88 --- /dev/null +++ b/arch/s390/mm/extable.c @@ -0,0 +1,81 @@ +#include +#include +#include + +/* + * Search one exception table for an entry corresponding to the + * given instruction address, and return the address of the entry, + * or NULL if none is found. + * We use a binary search, and thus we assume that the table is + * already sorted. + */ +const struct exception_table_entry * +search_extable(const struct exception_table_entry *first, + const struct exception_table_entry *last, + unsigned long value) +{ + const struct exception_table_entry *mid; + unsigned long addr; + + while (first <= last) { + mid = ((last - first) >> 1) + first; + addr = extable_insn(mid); + if (addr < value) + first = mid + 1; + else if (addr > value) + last = mid - 1; + else + return mid; + } + return NULL; +} + +/* + * The exception table needs to be sorted so that the binary + * search that we use to find entries in it works properly. + * This is used both for the kernel exception table and for + * the exception tables of modules that get loaded. + * + */ +static int cmp_ex(const void *a, const void *b) +{ + const struct exception_table_entry *x = a, *y = b; + + /* This compare is only valid after normalization. */ + return x->insn - y->insn; +} + +void sort_extable(struct exception_table_entry *start, + struct exception_table_entry *finish) +{ + struct exception_table_entry *p; + int i; + + /* Normalize entries to being relative to the start of the section */ + for (p = start, i = 0; p < finish; p++, i += 8) + p->insn += i; + sort(start, finish - start, sizeof(*start), cmp_ex, NULL); + /* Denormalize all entries */ + for (p = start, i = 0; p < finish; p++, i += 8) + p->insn -= i; +} + +#ifdef CONFIG_MODULES +/* + * If the exception table is sorted, any referring to the module init + * will be at the beginning or the end. + */ +void trim_init_extable(struct module *m) +{ + /* Trim the beginning */ + while (m->num_exentries && + within_module_init(extable_insn(&m->extable[0]), m)) { + m->extable++; + m->num_exentries--; + } + /* Trim the end */ + while (m->num_exentries && + within_module_init(extable_insn(&m->extable[m->num_exentries-1]), m)) + m->num_exentries--; +} +#endif /* CONFIG_MODULES */ diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c index 8b2cac1..ac9122c 100644 --- a/arch/s390/mm/fault.c +++ b/arch/s390/mm/fault.c @@ -163,7 +163,7 @@ static noinline void do_no_context(struct pt_regs *regs) /* Are we prepared to handle this kernel fault? */ fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN); if (fixup) { - regs->psw.addr = fixup->fixup | PSW_ADDR_AMODE; + regs->psw.addr = extable_fixup(fixup) | PSW_ADDR_AMODE; return; } diff --git a/scripts/sortextable.c b/scripts/sortextable.c index 6acf834..f19ddc4 100644 --- a/scripts/sortextable.c +++ b/scripts/sortextable.c @@ -161,7 +161,7 @@ typedef void (*table_sort_t)(char *, int); #define SORTEXTABLE_64 #include "sortextable.h" -static int compare_x86_table(const void *a, const void *b) +static int compare_relative_table(const void *a, const void *b) { int32_t av = (int32_t)r(a); int32_t bv = (int32_t)r(b); @@ -173,7 +173,7 @@ static int compare_x86_table(const void *a, const void *b) return 0; } -static void sort_x86_table(char *extab_image, int image_size) +static void sort_relative_table(char *extab_image, int image_size) { int i; @@ -188,7 +188,7 @@ static void sort_x86_table(char *extab_image, int image_size) i += 4; } - qsort(extab_image, image_size / 8, 8, compare_x86_table); + qsort(extab_image, image_size / 8, 8, compare_relative_table); /* Now denormalize. */ i = 0; @@ -245,9 +245,9 @@ do_file(char const *const fname) break; case EM_386: case EM_X86_64: - custom_sort = sort_x86_table; - break; case EM_S390: + custom_sort = sort_relative_table; + break; case EM_MIPS: break; } /* end switch */ -- cgit v0.10.2 From 94f2b9e25dc4d71b12cd16b6786cdef52b34a931 Mon Sep 17 00:00:00 2001 From: Jean Delvare Date: Thu, 6 Sep 2012 09:55:56 +0200 Subject: s390/3270: drop unused spinlock from struct con3270 I was checking why this spinlock was never initialized, but it turns out it's not used anywhere, so we can drop it. Signed-off-by: Jean Delvare Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/char/con3270.c b/drivers/s390/char/con3270.c index bb07577..699fd3e 100644 --- a/drivers/s390/char/con3270.c +++ b/drivers/s390/char/con3270.c @@ -35,7 +35,6 @@ static struct raw3270_fn con3270_fn; */ struct con3270 { struct raw3270_view view; - spinlock_t lock; struct list_head freemem; /* list of free memory for strings. */ /* Output stuff. */ -- cgit v0.10.2 From caf757c609445b01ad845df160369d8ccfd97d5b Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 14:42:13 +0200 Subject: s390/sysinfo,stsi: change return code handling Change return code handling of the stsi() function: In case function code 0 was specified the return value is the current configuration level (already shifted). That way all the code that actually copied the stsi_0() function can go away. Otherwise the return value is 0 (success) or negative to indicate an error (currently only -EOPNOTSUPP). Also stsi() is no longer an inline function. The function is not performance critical, but every caller would generate an exception table entry for this function. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/sysinfo.h b/arch/s390/include/asm/sysinfo.h index 04e6e97..f92428e 100644 --- a/arch/s390/include/asm/sysinfo.h +++ b/arch/s390/include/asm/sysinfo.h @@ -153,21 +153,7 @@ struct sysinfo_15_1_x { union topology_entry tle[0]; }; -static inline int stsi(void *sysinfo, int fc, int sel1, int sel2) -{ - register int r0 asm("0") = (fc << 28) | sel1; - register int r1 asm("1") = sel2; - - asm volatile( - " stsi 0(%2)\n" - "0: jz 2f\n" - "1: lhi %0,%3\n" - "2:\n" - EX_TABLE(0b, 1b) - : "+d" (r0) : "d" (r1), "a" (sysinfo), "K" (-ENOSYS) - : "cc", "memory"); - return r0; -} +int stsi(void *sysinfo, int fc, int sel1, int sel2); /* * Service level reporting interface. diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c index e8000d5..7f47176 100644 --- a/arch/s390/kernel/early.c +++ b/arch/s390/kernel/early.c @@ -222,12 +222,12 @@ static noinline __init void detect_machine_type(void) struct sysinfo_3_2_2 *vmms = (struct sysinfo_3_2_2 *)&sysinfo_page; /* Check current-configuration-level */ - if ((stsi(NULL, 0, 0, 0) >> 28) <= 2) { + if (stsi(NULL, 0, 0, 0) <= 2) { S390_lowcore.machine_flags |= MACHINE_FLAG_LPAR; return; } /* Get virtual-machine cpu information. */ - if (stsi(vmms, 3, 2, 2) == -ENOSYS || !vmms->count) + if (stsi(vmms, 3, 2, 2) || !vmms->count) return; /* Running under KVM? If not we assume z/VM */ @@ -246,7 +246,7 @@ static __init void setup_topology(void) return; S390_lowcore.machine_flags |= MACHINE_FLAG_TOPOLOGY; for (max_mnest = 6; max_mnest > 1; max_mnest--) { - if (stsi(&sysinfo_page, 15, 1, max_mnest) != -ENOSYS) + if (stsi(&sysinfo_page, 15, 1, max_mnest) == 0) break; } topology_max_mnest = max_mnest; diff --git a/arch/s390/kernel/lgr.c b/arch/s390/kernel/lgr.c index eca94e7..6ea6d69 100644 --- a/arch/s390/kernel/lgr.c +++ b/arch/s390/kernel/lgr.c @@ -51,16 +51,6 @@ static struct lgr_info lgr_info_cur; static struct debug_info *lgr_dbf; /* - * Return number of valid stsi levels - */ -static inline int stsi_0(void) -{ - int rc = stsi(NULL, 0, 0, 0); - - return rc == -ENOSYS ? rc : (((unsigned int) rc) >> 28); -} - -/* * Copy buffer and then convert it to ASCII */ static void cpascii(char *dst, char *src, int size) @@ -76,7 +66,7 @@ static void lgr_stsi_1_1_1(struct lgr_info *lgr_info) { struct sysinfo_1_1_1 *si = (void *) lgr_page; - if (stsi(si, 1, 1, 1) == -ENOSYS) + if (stsi(si, 1, 1, 1)) return; cpascii(lgr_info->manufacturer, si->manufacturer, sizeof(si->manufacturer)); @@ -93,7 +83,7 @@ static void lgr_stsi_2_2_2(struct lgr_info *lgr_info) { struct sysinfo_2_2_2 *si = (void *) lgr_page; - if (stsi(si, 2, 2, 2) == -ENOSYS) + if (stsi(si, 2, 2, 2)) return; cpascii(lgr_info->name, si->name, sizeof(si->name)); memcpy(&lgr_info->lpar_number, &si->lpar_number, @@ -108,7 +98,7 @@ static void lgr_stsi_3_2_2(struct lgr_info *lgr_info) struct sysinfo_3_2_2 *si = (void *) lgr_page; int i; - if (stsi(si, 3, 2, 2) == -ENOSYS) + if (stsi(si, 3, 2, 2)) return; for (i = 0; i < min_t(u8, si->count, VM_LEVEL_MAX); i++) { cpascii(lgr_info->vm[i].name, si->vm[i].name, @@ -124,16 +114,17 @@ static void lgr_stsi_3_2_2(struct lgr_info *lgr_info) */ static void lgr_info_get(struct lgr_info *lgr_info) { + int level; + memset(lgr_info, 0, sizeof(*lgr_info)); stfle(lgr_info->stfle_fac_list, ARRAY_SIZE(lgr_info->stfle_fac_list)); - lgr_info->level = stsi_0(); - if (lgr_info->level == -ENOSYS) - return; - if (lgr_info->level >= 1) + level = stsi(NULL, 0, 0, 0); + lgr_info->level = level; + if (level >= 1) lgr_stsi_1_1_1(lgr_info); - if (lgr_info->level >= 2) + if (level >= 2) lgr_stsi_2_2_2(lgr_info); - if (lgr_info->level >= 3) + if (level >= 3) lgr_stsi_3_2_2(lgr_info); } diff --git a/arch/s390/kernel/sysinfo.c b/arch/s390/kernel/sysinfo.c index 2af4ee6..62f89d9 100644 --- a/arch/s390/kernel/sysinfo.c +++ b/arch/s390/kernel/sysinfo.c @@ -24,18 +24,38 @@ int topology_max_mnest; -static inline int stsi_0(void) +/* + * stsi - store system information + * + * Returns the current configuration level if function code 0 was specified. + * Otherwise returns 0 on success or a negative value on error. + */ +int stsi(void *sysinfo, int fc, int sel1, int sel2) { - int rc = stsi(NULL, 0, 0, 0); - - return rc == -ENOSYS ? rc : (((unsigned int) rc) >> 28); + register int r0 asm("0") = (fc << 28) | sel1; + register int r1 asm("1") = sel2; + int rc = 0; + + asm volatile( + " stsi 0(%3)\n" + "0: jz 2f\n" + "1: lhi %1,%4\n" + "2:\n" + EX_TABLE(0b, 1b) + : "+d" (r0), "+d" (rc) + : "d" (r1), "a" (sysinfo), "K" (-EOPNOTSUPP) + : "cc", "memory"); + if (rc) + return rc; + return fc ? 0 : ((unsigned int) r0) >> 28; } +EXPORT_SYMBOL(stsi); static void stsi_1_1_1(struct seq_file *m, struct sysinfo_1_1_1 *info) { int i; - if (stsi(info, 1, 1, 1) == -ENOSYS) + if (stsi(info, 1, 1, 1)) return; EBCASC(info->manufacturer, sizeof(info->manufacturer)); EBCASC(info->type, sizeof(info->type)); @@ -97,7 +117,8 @@ static void stsi_15_1_x(struct seq_file *m, struct sysinfo_15_1_x *info) seq_putc(m, '\n'); if (!MACHINE_HAS_TOPOLOGY) return; - stsi(info, 15, 1, topology_max_mnest); + if (stsi(info, 15, 1, topology_max_mnest)) + return; seq_printf(m, "CPU Topology HW: "); for (i = 0; i < TOPOLOGY_NR_MAG; i++) seq_printf(m, " %d", info->mag[i]); @@ -116,7 +137,7 @@ static void stsi_1_2_2(struct seq_file *m, struct sysinfo_1_2_2 *info) struct sysinfo_1_2_2_extension *ext; int i; - if (stsi(info, 1, 2, 2) == -ENOSYS) + if (stsi(info, 1, 2, 2)) return; ext = (struct sysinfo_1_2_2_extension *) ((unsigned long) info + info->acc_offset); @@ -152,7 +173,7 @@ static void stsi_1_2_2(struct seq_file *m, struct sysinfo_1_2_2 *info) static void stsi_2_2_2(struct seq_file *m, struct sysinfo_2_2_2 *info) { - if (stsi(info, 2, 2, 2) == -ENOSYS) + if (stsi(info, 2, 2, 2)) return; EBCASC(info->name, sizeof(info->name)); seq_putc(m, '\n'); @@ -179,7 +200,7 @@ static void stsi_3_2_2(struct seq_file *m, struct sysinfo_3_2_2 *info) { int i; - if (stsi(info, 3, 2, 2) == -ENOSYS) + if (stsi(info, 3, 2, 2)) return; for (i = 0; i < info->count; i++) { EBCASC(info->vm[i].name, sizeof(info->vm[i].name)); @@ -202,7 +223,7 @@ static int sysinfo_show(struct seq_file *m, void *v) if (!info) return 0; - level = stsi_0(); + level = stsi(NULL, 0, 0, 0); if (level >= 1) stsi_1_1_1(m, info); if (level >= 1) @@ -365,7 +386,7 @@ void s390_adjust_jiffies(void) if (!info) return; - if (stsi(info, 1, 2, 2) != -ENOSYS) { + if (stsi(info, 1, 2, 2) == 0) { /* * Major sigh. The cpu capability encoding is "special". * If the first 9 bits of info->capability are 0 then it diff --git a/arch/s390/kvm/priv.c b/arch/s390/kvm/priv.c index 60da903..310be61 100644 --- a/arch/s390/kvm/priv.c +++ b/arch/s390/kvm/priv.c @@ -211,7 +211,7 @@ static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem) spin_unlock(&fi->lock); /* deal with other level 3 hypervisors */ - if (stsi(mem, 3, 2, 2) == -ENOSYS) + if (stsi(mem, 3, 2, 2)) mem->count = 0; if (mem->count < 8) mem->count++; @@ -259,7 +259,7 @@ static int handle_stsi(struct kvm_vcpu *vcpu) mem = get_zeroed_page(GFP_KERNEL); if (!mem) goto out_fail; - if (stsi((void *) mem, fc, sel1, sel2) == -ENOSYS) + if (stsi((void *) mem, fc, sel1, sel2)) goto out_mem; break; case 3: diff --git a/drivers/s390/net/qeth_core_main.c b/drivers/s390/net/qeth_core_main.c index 7a8b096..cf6da7f 100644 --- a/drivers/s390/net/qeth_core_main.c +++ b/drivers/s390/net/qeth_core_main.c @@ -2993,7 +2993,7 @@ static void qeth_get_trap_id(struct qeth_card *card, struct qeth_trap_id *tid) struct sysinfo_2_2_2 *info222 = (struct sysinfo_2_2_2 *)info; struct sysinfo_3_2_2 *info322 = (struct sysinfo_3_2_2 *)info; struct ccw_dev_id ccwid; - int level, rc; + int level; tid->chpid = card->info.chpid; ccw_device_get_id(CARD_RDEV(card), &ccwid); @@ -3001,17 +3001,10 @@ static void qeth_get_trap_id(struct qeth_card *card, struct qeth_trap_id *tid) tid->devno = ccwid.devno; if (!info) return; - - rc = stsi(NULL, 0, 0, 0); - if (rc == -ENOSYS) - level = rc; - else - level = (((unsigned int) rc) >> 28); - - if ((level >= 2) && (stsi(info222, 2, 2, 2) != -ENOSYS)) + level = stsi(NULL, 0, 0, 0); + if ((level >= 2) && (stsi(info222, 2, 2, 2) == 0)) tid->lparnr = info222->lpar_number; - - if ((level >= 3) && (stsi(info322, 3, 2, 2) != -ENOSYS)) { + if ((level >= 3) && (stsi(info322, 3, 2, 2) == 0)) { EBCASC(info322->vm[0].name, sizeof(info322->vm[0].name)); memcpy(tid->vmname, info322->vm[0].name, sizeof(tid->vmname)); } -- cgit v0.10.2 From 2267a0c85c4314b64440b9a1e3442d2eb90a719f Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 14:47:30 +0200 Subject: s390/dasd: fix return value for non-existent ioctls For non-existent ioctls -ENOTTY should be returned. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/block/dasd_ioctl.c b/drivers/s390/block/dasd_ioctl.c index 654c692..8252f37 100644 --- a/drivers/s390/block/dasd_ioctl.c +++ b/drivers/s390/block/dasd_ioctl.c @@ -292,12 +292,12 @@ out: #else static int dasd_ioctl_reset_profile(struct dasd_block *block) { - return -ENOSYS; + return -ENOTTY; } static int dasd_ioctl_read_profile(struct dasd_block *block, void __user *argp) { - return -ENOSYS; + return -ENOTTY; } #endif -- cgit v0.10.2 From d06cbda69ffc89bdfcf4cb9d78b0e64c5c28e549 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 15:00:07 +0200 Subject: s390/sclp: use -EOPNOTSUPP instead of -ENOSYS The return value was only internally used, so it's ok to change. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/char/sclp.c b/drivers/s390/char/sclp.c index 3fcc000..4fa21f7 100644 --- a/drivers/s390/char/sclp.c +++ b/drivers/s390/char/sclp.c @@ -334,7 +334,7 @@ sclp_dispatch_evbufs(struct sccb_header *sccb) reg->receiver_fn(evbuf); spin_lock_irqsave(&sclp_lock, flags); } else if (reg == NULL) - rc = -ENOSYS; + rc = -EOPNOTSUPP; } spin_unlock_irqrestore(&sclp_lock, flags); return rc; diff --git a/drivers/s390/char/sclp_rw.c b/drivers/s390/char/sclp_rw.c index 4be63be..3b13d58 100644 --- a/drivers/s390/char/sclp_rw.c +++ b/drivers/s390/char/sclp_rw.c @@ -463,7 +463,7 @@ sclp_emit_buffer(struct sclp_buffer *buffer, /* Use write priority message */ sccb->msg_buf.header.type = EVTYP_PMSGCMD; else - return -ENOSYS; + return -EOPNOTSUPP; buffer->request.command = SCLP_CMDW_WRITE_EVENT_DATA; buffer->request.status = SCLP_REQ_FILLED; buffer->request.callback = sclp_writedata_callback; -- cgit v0.10.2 From a8f6db4d2990d6866ba5d9c699cfc0835b9d8859 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 15:13:34 +0200 Subject: s390/etr,stp: use -EOPNOTSUPP instead of -ENOSYS Change -ENOSYS to -EOPNOTSUPP. Return value is used only internally. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/etr.h b/arch/s390/include/asm/etr.h index a24b03b..629b79a 100644 --- a/arch/s390/include/asm/etr.h +++ b/arch/s390/include/asm/etr.h @@ -140,7 +140,7 @@ struct etr_ptff_qto { /* Inline assembly helper functions */ static inline int etr_setr(struct etr_eacr *ctrl) { - int rc = -ENOSYS; + int rc = -EOPNOTSUPP; asm volatile( " .insn s,0xb2160000,%1\n" @@ -154,7 +154,7 @@ static inline int etr_setr(struct etr_eacr *ctrl) /* Stores a format 1 aib with 64 bytes */ static inline int etr_stetr(struct etr_aib *aib) { - int rc = -ENOSYS; + int rc = -EOPNOTSUPP; asm volatile( " .insn s,0xb2170000,%1\n" @@ -169,7 +169,7 @@ static inline int etr_stetr(struct etr_aib *aib) static inline int etr_steai(struct etr_aib *aib, unsigned int func) { register unsigned int reg0 asm("0") = func; - int rc = -ENOSYS; + int rc = -EOPNOTSUPP; asm volatile( " .insn s,0xb2b30000,%1\n" @@ -190,7 +190,7 @@ static inline int etr_ptff(void *ptff_block, unsigned int func) { register unsigned int reg0 asm("0") = func; register unsigned long reg1 asm("1") = (unsigned long) ptff_block; - int rc = -ENOSYS; + int rc = -EOPNOTSUPP; asm volatile( " .word 0x0104\n" diff --git a/arch/s390/kernel/time.c b/arch/s390/kernel/time.c index dcec960..2db1011 100644 --- a/arch/s390/kernel/time.c +++ b/arch/s390/kernel/time.c @@ -329,7 +329,7 @@ static unsigned long clock_sync_flags; * The synchronous get_clock function. It will write the current clock * value to the clock pointer and return 0 if the clock is in sync with * the external time source. If the clock mode is local it will return - * -ENOSYS and -EAGAIN if the clock is not in sync with the external + * -EOPNOTSUPP and -EAGAIN if the clock is not in sync with the external * reference. */ int get_sync_clock(unsigned long long *clock) @@ -347,7 +347,7 @@ int get_sync_clock(unsigned long long *clock) return 0; if (!test_bit(CLOCK_SYNC_HAS_ETR, &clock_sync_flags) && !test_bit(CLOCK_SYNC_HAS_STP, &clock_sync_flags)) - return -ENOSYS; + return -EOPNOTSUPP; if (!test_bit(CLOCK_SYNC_ETR, &clock_sync_flags) && !test_bit(CLOCK_SYNC_STP, &clock_sync_flags)) return -EACCES; diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index ed03c22..e1bb930 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -211,7 +211,7 @@ check_XRC (struct ccw1 *de_ccw, rc = get_sync_clock(&data->ep_sys_time); /* Ignore return code if sync clock is switched off. */ - if (rc == -ENOSYS || rc == -EACCES) + if (rc == -EOPNOTSUPP || rc == -EACCES) rc = 0; de_ccw->count = sizeof(struct DE_eckd_data); @@ -322,7 +322,7 @@ static int check_XRC_on_prefix(struct PFX_eckd_data *pfxdata, rc = get_sync_clock(&pfxdata->define_extent.ep_sys_time); /* Ignore return code if sync clock is switched off. */ - if (rc == -ENOSYS || rc == -EACCES) + if (rc == -EOPNOTSUPP || rc == -EACCES) rc = 0; return rc; } -- cgit v0.10.2 From d978386e79412e45d1aea81eee345dac23d20865 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 15:19:08 +0200 Subject: s390/kexec: change return value of machine_kexec_prepare Returning -ENOSYS on kexec_load() is a bad idea since user space cannot tell if the system call is not implmented or if it failed. Use -EOPNOTSUPP in case somebody tries a kexec_load on a NSS image based kernel instead. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/machine_kexec.c b/arch/s390/kernel/machine_kexec.c index 0d5063e0..cc91386 100644 --- a/arch/s390/kernel/machine_kexec.c +++ b/arch/s390/kernel/machine_kexec.c @@ -159,7 +159,7 @@ int machine_kexec_prepare(struct kimage *image) /* Can't replace kernel image since it is read-only. */ if (ipl_flags & IPL_NSS_VALID) - return -ENOSYS; + return -EOPNOTSUPP; if (image->type == KEXEC_TYPE_CRASH) return machine_kexec_prepare_kdump(); -- cgit v0.10.2 From a11b2ef7bb69136dfac2d6367574151223ee7712 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 16:53:44 +0200 Subject: s390/appldata: change return value of appldata_asm Change return value of appldata_asm() to -EOPNOTSUPP in case of an error. The return value was only used internally and not passed to user space. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/appldata.h b/arch/s390/include/asm/appldata.h index f328294..32a7059 100644 --- a/arch/s390/include/asm/appldata.h +++ b/arch/s390/include/asm/appldata.h @@ -70,7 +70,7 @@ static inline int appldata_asm(struct appldata_product_id *id, int ry; if (!MACHINE_IS_VM) - return -ENOSYS; + return -EOPNOTSUPP; parm_list.diag = 0xdc; parm_list.function = fn; parm_list.parlist_length = sizeof(parm_list); -- cgit v0.10.2 From 745e967a49b26725cc9f30105ff67c8fee8926d6 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 6 Sep 2012 17:35:00 +0200 Subject: s390/vmlogrdr: change return value from -ENOSYS to -EOPNOTSUPP Changing the return value of vmlogrdr_open() to -EOPNOTSUPP if O_NONBLOCK is specified shouldn't have any negative side effects. Any existing user wouldn't specify that flag since it wouldn't work anyway. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/char/vmlogrdr.c b/drivers/s390/char/vmlogrdr.c index c131bc4..9b3a24e 100644 --- a/drivers/s390/char/vmlogrdr.c +++ b/drivers/s390/char/vmlogrdr.c @@ -321,7 +321,7 @@ static int vmlogrdr_open (struct inode *inode, struct file *filp) * only allow for blocking reads to be open */ if (filp->f_flags & O_NONBLOCK) - return -ENOSYS; + return -EOPNOTSUPP; /* Besure this device hasn't already been opened */ spin_lock_bh(&logptr->priv_lock); -- cgit v0.10.2 From dabecb2933f7ae901c88cb10c71ab38ca7dfc38f Mon Sep 17 00:00:00 2001 From: Holger Dengler Date: Mon, 10 Sep 2012 21:34:26 +0200 Subject: s390/zcryt: Handle AP configuration changes Detect external AP bus configuration changes and request an AP device rescan. Signed-off-by: Holger Dengler Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index 39e5eb2..cf5d9dc 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -950,6 +950,20 @@ void ap_driver_unregister(struct ap_driver *ap_drv) } EXPORT_SYMBOL(ap_driver_unregister); +void ap_bus_force_rescan(void) +{ + /* Delete the AP bus rescan timer. */ + del_timer(&ap_config_timer); + + /* processing a synchonuous bus rescan */ + ap_scan_bus(NULL); + + /* Setup the AP bus rescan timer again. */ + ap_config_timer.expires = jiffies + ap_config_time * HZ; + add_timer(&ap_config_timer); +} +EXPORT_SYMBOL(ap_bus_force_rescan); + /* * AP bus attributes. */ diff --git a/drivers/s390/crypto/ap_bus.h b/drivers/s390/crypto/ap_bus.h index e9b963e..685f6cc 100644 --- a/drivers/s390/crypto/ap_bus.h +++ b/drivers/s390/crypto/ap_bus.h @@ -231,6 +231,7 @@ int ap_recv(ap_qid_t, unsigned long long *, void *, size_t); void ap_queue_message(struct ap_device *ap_dev, struct ap_message *ap_msg); void ap_cancel_message(struct ap_device *ap_dev, struct ap_message *ap_msg); void ap_flush_queue(struct ap_device *ap_dev); +void ap_bus_force_rescan(void); int ap_module_init(void); void ap_module_exit(void); diff --git a/drivers/s390/crypto/zcrypt_api.c b/drivers/s390/crypto/zcrypt_api.c index f1f026e..31cfaa5 100644 --- a/drivers/s390/crypto/zcrypt_api.c +++ b/drivers/s390/crypto/zcrypt_api.c @@ -38,7 +38,10 @@ #include #include #include +#include +#include +#include "zcrypt_debug.h" #include "zcrypt_api.h" /* @@ -53,6 +56,10 @@ static DEFINE_SPINLOCK(zcrypt_device_lock); static LIST_HEAD(zcrypt_device_list); static int zcrypt_device_count = 0; static atomic_t zcrypt_open_count = ATOMIC_INIT(0); +static atomic_t zcrypt_rescan_count = ATOMIC_INIT(0); + +atomic_t zcrypt_rescan_req = ATOMIC_INIT(0); +EXPORT_SYMBOL(zcrypt_rescan_req); static int zcrypt_rng_device_add(void); static void zcrypt_rng_device_remove(void); @@ -60,6 +67,10 @@ static void zcrypt_rng_device_remove(void); static DEFINE_SPINLOCK(zcrypt_ops_list_lock); static LIST_HEAD(zcrypt_ops_list); +static debug_info_t *zcrypt_dbf_common; +static debug_info_t *zcrypt_dbf_devices; +static struct dentry *debugfs_root; + /* * Device attributes common for all crypto devices. */ @@ -89,6 +100,8 @@ static ssize_t zcrypt_online_store(struct device *dev, if (sscanf(buf, "%d\n", &online) != 1 || online < 0 || online > 1) return -EINVAL; zdev->online = online; + ZCRYPT_DBF_DEV(DBF_INFO, zdev, "dev%04xo%dman", zdev->ap_dev->qid, + zdev->online); if (!online) ap_flush_queue(zdev->ap_dev); return count; @@ -107,6 +120,24 @@ static struct attribute_group zcrypt_device_attr_group = { }; /** + * Process a rescan of the transport layer. + * + * Returns 1, if the rescan has been processed, otherwise 0. + */ +static inline int zcrypt_process_rescan(void) +{ + if (atomic_read(&zcrypt_rescan_req)) { + atomic_set(&zcrypt_rescan_req, 0); + atomic_inc(&zcrypt_rescan_count); + ap_bus_force_rescan(); + ZCRYPT_DBF_COMMON(DBF_INFO, "rescan%07d", + atomic_inc_return(&zcrypt_rescan_count)); + return 1; + } + return 0; +} + +/** * __zcrypt_increase_preference(): Increase preference of a crypto device. * @zdev: Pointer the crypto device * @@ -194,6 +225,7 @@ struct zcrypt_device *zcrypt_device_alloc(size_t max_response_size) zdev->reply.length = max_response_size; spin_lock_init(&zdev->lock); INIT_LIST_HEAD(&zdev->list); + zdev->dbf_area = zcrypt_dbf_devices; return zdev; out_free: @@ -229,6 +261,8 @@ int zcrypt_device_register(struct zcrypt_device *zdev) kref_init(&zdev->refcount); spin_lock_bh(&zcrypt_device_lock); zdev->online = 1; /* New devices are online by default. */ + ZCRYPT_DBF_DEV(DBF_INFO, zdev, "dev%04xo%dreg", zdev->ap_dev->qid, + zdev->online); list_add_tail(&zdev->list, &zcrypt_device_list); __zcrypt_increase_preference(zdev); zcrypt_device_count++; @@ -707,6 +741,11 @@ static long zcrypt_unlocked_ioctl(struct file *filp, unsigned int cmd, do { rc = zcrypt_rsa_modexpo(&mex); } while (rc == -EAGAIN); + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + do { + rc = zcrypt_rsa_modexpo(&mex); + } while (rc == -EAGAIN); if (rc) return rc; return put_user(mex.outputdatalength, &umex->outputdatalength); @@ -719,6 +758,11 @@ static long zcrypt_unlocked_ioctl(struct file *filp, unsigned int cmd, do { rc = zcrypt_rsa_crt(&crt); } while (rc == -EAGAIN); + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + do { + rc = zcrypt_rsa_crt(&crt); + } while (rc == -EAGAIN); if (rc) return rc; return put_user(crt.outputdatalength, &ucrt->outputdatalength); @@ -731,6 +775,11 @@ static long zcrypt_unlocked_ioctl(struct file *filp, unsigned int cmd, do { rc = zcrypt_send_cprb(&xcRB); } while (rc == -EAGAIN); + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + do { + rc = zcrypt_send_cprb(&xcRB); + } while (rc == -EAGAIN); if (copy_to_user(uxcRB, &xcRB, sizeof(xcRB))) return -EFAULT; return rc; @@ -837,10 +886,15 @@ static long trans_modexpo32(struct file *filp, unsigned int cmd, do { rc = zcrypt_rsa_modexpo(&mex64); } while (rc == -EAGAIN); - if (!rc) - rc = put_user(mex64.outputdatalength, - &umex32->outputdatalength); - return rc; + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + do { + rc = zcrypt_rsa_modexpo(&mex64); + } while (rc == -EAGAIN); + if (rc) + return rc; + return put_user(mex64.outputdatalength, + &umex32->outputdatalength); } struct compat_ica_rsa_modexpo_crt { @@ -877,10 +931,15 @@ static long trans_modexpo_crt32(struct file *filp, unsigned int cmd, do { rc = zcrypt_rsa_crt(&crt64); } while (rc == -EAGAIN); - if (!rc) - rc = put_user(crt64.outputdatalength, - &ucrt32->outputdatalength); - return rc; + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + do { + rc = zcrypt_rsa_crt(&crt64); + } while (rc == -EAGAIN); + if (rc) + return rc; + return put_user(crt64.outputdatalength, + &ucrt32->outputdatalength); } struct compat_ica_xcRB { @@ -936,6 +995,11 @@ static long trans_xcRB32(struct file *filp, unsigned int cmd, do { rc = zcrypt_send_cprb(&xcRB64); } while (rc == -EAGAIN); + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + do { + rc = zcrypt_send_cprb(&xcRB64); + } while (rc == -EAGAIN); xcRB32.reply_control_blk_length = xcRB64.reply_control_blk_length; xcRB32.reply_data_length = xcRB64.reply_data_length; xcRB32.status = xcRB64.status; @@ -1193,6 +1257,9 @@ static int zcrypt_rng_data_read(struct hwrng *rng, u32 *data) */ if (zcrypt_rng_buffer_index == 0) { rc = zcrypt_rng((char *) zcrypt_rng_buffer); + /* on failure: retry once again after a requested rescan */ + if ((rc == -ENODEV) && (zcrypt_process_rescan())) + rc = zcrypt_rng((char *) zcrypt_rng_buffer); if (rc < 0) return -EIO; zcrypt_rng_buffer_index = rc / sizeof *data; @@ -1245,6 +1312,30 @@ static void zcrypt_rng_device_remove(void) mutex_unlock(&zcrypt_rng_mutex); } +int __init zcrypt_debug_init(void) +{ + debugfs_root = debugfs_create_dir("zcrypt", NULL); + + zcrypt_dbf_common = debug_register("zcrypt_common", 1, 1, 16); + debug_register_view(zcrypt_dbf_common, &debug_hex_ascii_view); + debug_set_level(zcrypt_dbf_common, DBF_ERR); + + zcrypt_dbf_devices = debug_register("zcrypt_devices", 1, 1, 16); + debug_register_view(zcrypt_dbf_devices, &debug_hex_ascii_view); + debug_set_level(zcrypt_dbf_devices, DBF_ERR); + + return 0; +} + +void zcrypt_debug_exit(void) +{ + debugfs_remove(debugfs_root); + if (zcrypt_dbf_common) + debug_unregister(zcrypt_dbf_common); + if (zcrypt_dbf_devices) + debug_unregister(zcrypt_dbf_devices); +} + /** * zcrypt_api_init(): Module initialization. * @@ -1254,6 +1345,12 @@ int __init zcrypt_api_init(void) { int rc; + rc = zcrypt_debug_init(); + if (rc) + goto out; + + atomic_set(&zcrypt_rescan_req, 0); + /* Register the request sprayer. */ rc = misc_register(&zcrypt_misc_device); if (rc < 0) @@ -1283,6 +1380,7 @@ void zcrypt_api_exit(void) { remove_proc_entry("driver/z90crypt", NULL); misc_deregister(&zcrypt_misc_device); + zcrypt_debug_exit(); } module_init(zcrypt_api_init); diff --git a/drivers/s390/crypto/zcrypt_api.h b/drivers/s390/crypto/zcrypt_api.h index 02b2d35..8963291 100644 --- a/drivers/s390/crypto/zcrypt_api.h +++ b/drivers/s390/crypto/zcrypt_api.h @@ -29,8 +29,10 @@ #ifndef _ZCRYPT_API_H_ #define _ZCRYPT_API_H_ -#include "ap_bus.h" +#include +#include #include +#include "ap_bus.h" /* deprecated status calls */ #define ICAZ90STATUS _IOR(ZCRYPT_IOCTL_MAGIC, 0x10, struct ica_z90_status) @@ -112,8 +114,13 @@ struct zcrypt_device { struct ap_message reply; /* Per-device reply structure. */ int max_exp_bit_length; + + debug_info_t *dbf_area; /* debugging */ }; +/* transport layer rescanning */ +extern atomic_t zcrypt_rescan_req; + struct zcrypt_device *zcrypt_device_alloc(size_t); void zcrypt_device_free(struct zcrypt_device *); void zcrypt_device_get(struct zcrypt_device *); diff --git a/drivers/s390/crypto/zcrypt_debug.h b/drivers/s390/crypto/zcrypt_debug.h new file mode 100644 index 0000000..841ea72 --- /dev/null +++ b/drivers/s390/crypto/zcrypt_debug.h @@ -0,0 +1,59 @@ +/* + * Copyright IBM Corp. 2012 + * Author(s): Holger Dengler (hd@linux.vnet.ibm.com) + */ +#ifndef ZCRYPT_DEBUG_H +#define ZCRYPT_DEBUG_H + +#include +#include "zcrypt_api.h" + +/* that gives us 15 characters in the text event views */ +#define ZCRYPT_DBF_LEN 16 + +/* sort out low debug levels early to avoid wasted sprints */ +static inline int zcrypt_dbf_passes(debug_info_t *dbf_grp, int level) +{ + return (level <= dbf_grp->level); +} + +#define DBF_ERR 3 /* error conditions */ +#define DBF_WARN 4 /* warning conditions */ +#define DBF_INFO 6 /* informational */ + +#define RC2WARN(rc) ((rc) ? DBF_WARN : DBF_INFO) + +#define ZCRYPT_DBF_COMMON(level, text...) \ + do { \ + if (zcrypt_dbf_passes(zcrypt_dbf_common, level)) { \ + char debug_buffer[ZCRYPT_DBF_LEN]; \ + snprintf(debug_buffer, ZCRYPT_DBF_LEN, text); \ + debug_text_event(zcrypt_dbf_common, level, \ + debug_buffer); \ + } \ + } while (0) + +#define ZCRYPT_DBF_DEVICES(level, text...) \ + do { \ + if (zcrypt_dbf_passes(zcrypt_dbf_devices, level)) { \ + char debug_buffer[ZCRYPT_DBF_LEN]; \ + snprintf(debug_buffer, ZCRYPT_DBF_LEN, text); \ + debug_text_event(zcrypt_dbf_devices, level, \ + debug_buffer); \ + } \ + } while (0) + +#define ZCRYPT_DBF_DEV(level, device, text...) \ + do { \ + if (zcrypt_dbf_passes(device->dbf_area, level)) { \ + char debug_buffer[ZCRYPT_DBF_LEN]; \ + snprintf(debug_buffer, ZCRYPT_DBF_LEN, text); \ + debug_text_event(device->dbf_area, level, \ + debug_buffer); \ + } \ + } while (0) + +int zcrypt_debug_init(void); +void zcrypt_debug_exit(void); + +#endif /* ZCRYPT_DEBUG_H */ diff --git a/drivers/s390/crypto/zcrypt_error.h b/drivers/s390/crypto/zcrypt_error.h index 0965e26..0079b66 100644 --- a/drivers/s390/crypto/zcrypt_error.h +++ b/drivers/s390/crypto/zcrypt_error.h @@ -26,6 +26,8 @@ #ifndef _ZCRYPT_ERROR_H_ #define _ZCRYPT_ERROR_H_ +#include +#include "zcrypt_debug.h" #include "zcrypt_api.h" /** @@ -108,16 +110,27 @@ static inline int convert_error(struct zcrypt_device *zdev, * and then repeat the request. */ WARN_ON(1); + atomic_set(&zcrypt_rescan_req, 1); zdev->online = 0; + ZCRYPT_DBF_DEV(DBF_ERR, zdev, "dev%04xo%drc%d", + zdev->ap_dev->qid, + zdev->online, ehdr->reply_code); return -EAGAIN; case REP82_ERROR_TRANSPORT_FAIL: case REP82_ERROR_MACHINE_FAILURE: // REP88_ERROR_MODULE_FAILURE // '10' CEX2A /* If a card fails disable it and repeat the request. */ + atomic_set(&zcrypt_rescan_req, 1); zdev->online = 0; + ZCRYPT_DBF_DEV(DBF_ERR, zdev, "dev%04xo%drc%d", + zdev->ap_dev->qid, + zdev->online, ehdr->reply_code); return -EAGAIN; default: zdev->online = 0; + ZCRYPT_DBF_DEV(DBF_ERR, zdev, "dev%04xo%drc%d", + zdev->ap_dev->qid, + zdev->online, ehdr->reply_code); return -EAGAIN; /* repeat the request on a different device. */ } } -- cgit v0.10.2 From c0162b07b34ce8b33548f4b2658ec3133f960e71 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 13 Sep 2012 15:55:38 +0200 Subject: s390/syscalls: wire up kcmp system call Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/unistd.h b/arch/s390/include/asm/unistd.h index da2c2f4..4e64b5c 100644 --- a/arch/s390/include/asm/unistd.h +++ b/arch/s390/include/asm/unistd.h @@ -278,7 +278,8 @@ #define __NR_process_vm_readv 340 #define __NR_process_vm_writev 341 #define __NR_s390_runtime_instr 342 -#define NR_syscalls 343 +#define __NR_kcmp 343 +#define NR_syscalls 344 /* * There are some system calls that are not present on 64 bit, some diff --git a/arch/s390/kernel/compat_wrapper.S b/arch/s390/kernel/compat_wrapper.S index 5f7928f..3afba80 100644 --- a/arch/s390/kernel/compat_wrapper.S +++ b/arch/s390/kernel/compat_wrapper.S @@ -1651,3 +1651,11 @@ ENTRY(sys_s390_runtime_instr_wrapper) lgfr %r2,%r2 # int lgfr %r3,%r3 # int jg sys_s390_runtime_instr + +ENTRY(sys_kcmp_wrapper) + lgfr %r2,%r2 # pid_t + lgfr %r3,%r3 # pid_t + lgfr %r4,%r4 # int + llgfr %r5,%r5 # unsigned long + llgfr %r6,%r6 # unsigned long + jg sys_kcmp diff --git a/arch/s390/kernel/syscalls.S b/arch/s390/kernel/syscalls.S index 539f0a7..4817485 100644 --- a/arch/s390/kernel/syscalls.S +++ b/arch/s390/kernel/syscalls.S @@ -351,3 +351,4 @@ SYSCALL(sys_setns,sys_setns,sys_setns_wrapper) SYSCALL(sys_process_vm_readv,sys_process_vm_readv,compat_sys_process_vm_readv_wrapper) /* 340 */ SYSCALL(sys_process_vm_writev,sys_process_vm_writev,compat_sys_process_vm_writev_wrapper) SYSCALL(sys_ni_syscall,sys_s390_runtime_instr,sys_s390_runtime_instr_wrapper) +SYSCALL(sys_kcmp,sys_kcmp,sys_kcmp_wrapper) -- cgit v0.10.2 From d3feb481083636296f25daf36dbf44493fa7e1f2 Mon Sep 17 00:00:00 2001 From: Jan Glauber Date: Tue, 11 Sep 2012 13:42:06 +0200 Subject: s390/qdio: fix truncated debug output of hex values Calling debug_event the s390 debug feature only logs up to buf_size bytes of the debug view. If debug_event is called with more bytes than buf_size the additional data is ignored and not logged in the debug view. Use multiple calls to debug_event if the length exceeds buf_size. Signed-off-by: Jan Glauber Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/cio/qdio_debug.h b/drivers/s390/cio/qdio_debug.h index e1f6468..7f8b973 100644 --- a/drivers/s390/cio/qdio_debug.h +++ b/drivers/s390/cio/qdio_debug.h @@ -37,10 +37,14 @@ static inline int qdio_dbf_passes(debug_info_t *dbf_grp, int level) debug_text_event(qdio_dbf_setup, DBF_ERR, debug_buffer); \ } while (0) -#define DBF_HEX(addr, len) \ - do { \ - debug_event(qdio_dbf_setup, DBF_ERR, (void*)(addr), len); \ - } while (0) +static inline void DBF_HEX(void *addr, int len) +{ + while (len > 0) { + debug_event(qdio_dbf_setup, DBF_ERR, addr, len); + len -= qdio_dbf_setup->buf_size; + addr += qdio_dbf_setup->buf_size; + } +} #define DBF_ERROR(text...) \ do { \ @@ -49,11 +53,14 @@ static inline int qdio_dbf_passes(debug_info_t *dbf_grp, int level) debug_text_event(qdio_dbf_error, DBF_ERR, debug_buffer); \ } while (0) -#define DBF_ERROR_HEX(addr, len) \ - do { \ - debug_event(qdio_dbf_error, DBF_ERR, (void*)(addr), len); \ - } while (0) - +static inline void DBF_ERROR_HEX(void *addr, int len) +{ + while (len > 0) { + debug_event(qdio_dbf_error, DBF_ERR, addr, len); + len -= qdio_dbf_error->buf_size; + addr += qdio_dbf_error->buf_size; + } +} #define DBF_DEV_EVENT(level, device, text...) \ do { \ @@ -64,10 +71,15 @@ static inline int qdio_dbf_passes(debug_info_t *dbf_grp, int level) } \ } while (0) -#define DBF_DEV_HEX(level, device, addr, len) \ - do { \ - debug_event(device->debug_area, level, (void*)(addr), len); \ - } while (0) +static inline void DBF_DEV_HEX(struct qdio_irq *dev, void *addr, + int len, int level) +{ + while (len > 0) { + debug_event(dev->debug_area, level, addr, len); + len -= dev->debug_area->buf_size; + addr += dev->debug_area->buf_size; + } +} void qdio_allocate_dbf(struct qdio_initialize *init_data, struct qdio_irq *irq_ptr); -- cgit v0.10.2 From 0327dab0e88b2f4be6ad7dcd5af86d085d2c0603 Mon Sep 17 00:00:00 2001 From: Wei Yongjun Date: Fri, 14 Sep 2012 10:31:21 +0800 Subject: s390/topology: use for_each_set_bit to simplify the code Use for_each_set_bit() to simplify the code. spatch with a semantic match is used to found this. (http://coccinelle.lip6.fr/) Signed-off-by: Wei Yongjun Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c index 30c8533..54d93f4 100644 --- a/arch/s390/kernel/topology.c +++ b/arch/s390/kernel/topology.c @@ -73,10 +73,7 @@ static struct mask_info *add_cpus_to_mask(struct topology_cpu *tl_cpu, { unsigned int cpu; - for (cpu = find_first_bit(&tl_cpu->mask[0], TOPOLOGY_CPU_BITS); - cpu < TOPOLOGY_CPU_BITS; - cpu = find_next_bit(&tl_cpu->mask[0], TOPOLOGY_CPU_BITS, cpu + 1)) - { + for_each_set_bit(cpu, &tl_cpu->mask[0], TOPOLOGY_CPU_BITS) { unsigned int rcpu; int lcpu; -- cgit v0.10.2 From c80773e5a24df5f0604ed796b56edd8a14dbcce7 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 11:21:17 +0200 Subject: s390/ap_bus: add missing ifdef MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Get rid of this compile warning for CONFIG_32BIT: drivers/s390/crypto/ap_bus.c:168:12: warning: ‘ap_configuration_available’ defined but not used [-Wunused-function] Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index cf5d9dc..7b865a7 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -165,10 +165,12 @@ static int ap_interrupts_available(void) * * Returns 1 if AP configuration information is available. */ +#ifdef CONFIG_64BIT static int ap_configuration_available(void) { return test_facility(2) && test_facility(12); } +#endif /** * ap_test_queue(): Test adjunct processor queue. -- cgit v0.10.2 From 0e37b0cbb8edf64dc87ae95831c177db6c6d3b5c Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 12:30:10 +0200 Subject: s390/tape: remove even more tape block leftovers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix this compile error: In file included from drivers/s390/char/tape_core.c:29:0: drivers/s390/char/tape_std.h:103:66: warning: ‘struct request’ declared inside parameter list [enabled by default] Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/char/tape.h b/drivers/s390/char/tape.h index c06be6c..ea664dd 100644 --- a/drivers/s390/char/tape.h +++ b/drivers/s390/char/tape.h @@ -15,7 +15,6 @@ #include #include #include -#include #include #include #include diff --git a/drivers/s390/char/tape_std.h b/drivers/s390/char/tape_std.h index c5816ad..8c760c0 100644 --- a/drivers/s390/char/tape_std.h +++ b/drivers/s390/char/tape_std.h @@ -100,11 +100,7 @@ struct tape_request *tape_std_read_block(struct tape_device *, size_t); void tape_std_read_backward(struct tape_device *device, struct tape_request *request); struct tape_request *tape_std_write_block(struct tape_device *, size_t); -struct tape_request *tape_std_bread(struct tape_device *, struct request *); -void tape_std_free_bread(struct tape_request *); void tape_std_check_locate(struct tape_device *, struct tape_request *); -struct tape_request *tape_std_bwrite(struct request *, - struct tape_device *, int); /* Some non-mtop commands. */ int tape_std_assign(struct tape_device *); -- cgit v0.10.2 From 305e4f108b683a96254d36f916ecab79dcc3ea4f Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 12:57:39 +0200 Subject: s390/perf_events: compile only for CONFIG_64BIT MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The whole hardware support is only available in zArch mode. Fixes also this compile warning: arch/s390/kernel/perf_cpum_cf.c: In function ‘cpumf_pmu_init’: arch/s390/kernel/perf_cpum_cf.c:670:2: warning: left shift count >= width of type [enabled by default] Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile index bba01bf..249aaeb 100644 --- a/arch/s390/kernel/Makefile +++ b/arch/s390/kernel/Makefile @@ -48,8 +48,11 @@ obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o obj-$(CONFIG_FTRACE_SYSCALLS) += ftrace.o obj-$(CONFIG_CRASH_DUMP) += crash_dump.o + +ifdef CONFIG_64BIT obj-$(CONFIG_PERF_EVENTS) += perf_event.o perf_cpum_cf.o -obj-$(CONFIG_64BIT) += runtime_instr.o cache.o +obj-y += runtime_instr.o cache.o +endif # Kexec part S390_KEXEC_OBJS := machine_kexec.o -- cgit v0.10.2 From 1a0f24878eb5ca56dff5f20b9f0e716947dc0394 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 13:22:48 +0200 Subject: s390/ipl,decrompressor: disable branch profiling The early kernel decrompressing code can't call into the kernel in order to profile branches. Fixes this link error: arch/s390/boot/compressed/misc.o: In function `decompress_kernel': misc.c:(.text+0x9a6): undefined reference to `ftrace_likely_update' Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/boot/compressed/Makefile b/arch/s390/boot/compressed/Makefile index 10e22c4..3ad8f61 100644 --- a/arch/s390/boot/compressed/Makefile +++ b/arch/s390/boot/compressed/Makefile @@ -11,6 +11,7 @@ targets := vmlinux.lds vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 \ sizes.h head$(BITS).o KBUILD_CFLAGS := -m$(BITS) -D__KERNEL__ $(LINUX_INCLUDE) -O2 +KBUILD_CFLAGS += -DDISABLE_BRANCH_PROFILING KBUILD_CFLAGS += $(cflags-y) KBUILD_CFLAGS += $(call cc-option,-mpacked-stack) KBUILD_CFLAGS += $(call cc-option,-ffreestanding) -- cgit v0.10.2 From 66389e8583af71f8e01fab7c8e704a6b393663b9 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 12:59:47 +0200 Subject: s390/ptrace: add missing ifdef MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit if (MACHINE_HAS_TE) translates to if (0) on !CONFIG_64BIT however the compiler still warns about invalid shifts within non-reachable code. So add an explicit ifdef to get rid of this warning: arch/s390/kernel/ptrace.c: In function ‘update_per_regs’: arch/s390/kernel/ptrace.c:63:4: warning: left shift count >= width of type [enabled by default] arch/s390/kernel/ptrace.c:65:4: warning: left shift count >= width of type [enabled by default] Reported-by: Fengguang Wu Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c index b817cc5..a314c57 100644 --- a/arch/s390/kernel/ptrace.c +++ b/arch/s390/kernel/ptrace.c @@ -53,6 +53,7 @@ void update_per_regs(struct task_struct *task) struct thread_struct *thread = &task->thread; struct per_regs old, new; +#ifdef CONFIG_64BIT /* Take care of the enable/disable of transactional execution. */ if (MACHINE_HAS_TE) { unsigned long cr0, cr0_new; @@ -67,7 +68,7 @@ void update_per_regs(struct task_struct *task) if (cr0 != cr0_new) __ctl_load(cr0_new, 0, 0); } - +#endif /* Copy user specified PER registers */ new.control = thread->per_user.control; new.start = thread->per_user.start; -- cgit v0.10.2 From f6e38691a2f410e3333941e535c7b20352f98ddb Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 13:45:24 +0200 Subject: s390/process: add missing header include Add appropriate header file: arch/s390/kernel/process.c:327:15: warning: symbol 'arch_align_stack' was not declared. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/process.c b/arch/s390/kernel/process.c index cb4328e..5024be2 100644 --- a/arch/s390/kernel/process.c +++ b/arch/s390/kernel/process.c @@ -26,6 +26,7 @@ #include #include #include +#include #include #include #include -- cgit v0.10.2 From 6b563d8c26735a76d84664d1b48271c432d63983 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 14:11:32 +0200 Subject: s390/crashdump: move fill_cpu_elf_notes() prototype to header file Move fill_cpu_elf_notes() prototype to header file. This way we get compile errors if e.g. the number of function parameters get changed. Otherwise it's possible to change just the definition and everything else still compiles fine, but the result is broken code. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/elf.h b/arch/s390/include/asm/elf.h index db57594..178ff96 100644 --- a/arch/s390/include/asm/elf.h +++ b/arch/s390/include/asm/elf.h @@ -213,4 +213,6 @@ int arch_setup_additional_pages(struct linux_binprm *, int); extern unsigned long arch_randomize_brk(struct mm_struct *mm); #define arch_randomize_brk arch_randomize_brk +void *fill_cpu_elf_notes(void *ptr, struct save_area *sa); + #endif diff --git a/arch/s390/kernel/crash_dump.c b/arch/s390/kernel/crash_dump.c index cc1172b..fb8d878 100644 --- a/arch/s390/kernel/crash_dump.c +++ b/arch/s390/kernel/crash_dump.c @@ -13,8 +13,9 @@ #include #include #include -#include #include +#include +#include #define PTR_ADD(x, y) (((char *) (x)) + ((unsigned long) (y))) #define PTR_SUB(x, y) (((char *) (x)) - ((unsigned long) (y))) diff --git a/arch/s390/kernel/machine_kexec.c b/arch/s390/kernel/machine_kexec.c index cc91386..b3de277 100644 --- a/arch/s390/kernel/machine_kexec.c +++ b/arch/s390/kernel/machine_kexec.c @@ -21,6 +21,7 @@ #include #include #include +#include #include #include @@ -31,8 +32,6 @@ extern const unsigned long long relocate_kernel_len; #ifdef CONFIG_CRASH_DUMP -void *fill_cpu_elf_notes(void *ptr, struct save_area *sa); - /* * Create ELF notes for one CPU */ -- cgit v0.10.2 From 41459d36cf0d57813017dae6080a879cc038e5fe Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 14 Sep 2012 11:09:52 +0200 Subject: s390: add uninitialized_var() to suppress false positive compiler warnings MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Get rid of these: arch/s390/kernel/smp.c:134:19: warning: ‘status’ may be used uninitialized in this function [-Wuninitialized] arch/s390/mm/pgtable.c:641:10: warning: ‘table’ may be used uninitialized in this function [-Wuninitialized] arch/s390/mm/pgtable.c:644:12: warning: ‘page’ may be used uninitialized in this function [-Wuninitialized] drivers/s390/cio/cio.c:1037:14: warning: ‘schid’ may be used uninitialized in this function [-Wuninitialized] Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index 8676aab..ea431e5 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -126,7 +126,7 @@ static int pcpu_sigp_retry(struct pcpu *pcpu, u8 order, u32 parm) static inline int pcpu_stopped(struct pcpu *pcpu) { - u32 status; + u32 uninitialized_var(status); if (__pcpu_sigp(pcpu->address, SIGP_SENSE, 0, &status) != SIGP_CC_STATUS_STORED) diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c index d83ecbf..b402991 100644 --- a/arch/s390/mm/pgtable.c +++ b/arch/s390/mm/pgtable.c @@ -609,8 +609,8 @@ static inline unsigned int atomic_xor_bits(atomic_t *v, unsigned int bits) */ unsigned long *page_table_alloc(struct mm_struct *mm, unsigned long vmaddr) { - struct page *page; - unsigned long *table; + unsigned long *uninitialized_var(table); + struct page *uninitialized_var(page); unsigned int mask, bit; if (mm_has_pgste(mm)) diff --git a/drivers/s390/cio/cio.c b/drivers/s390/cio/cio.c index 33d1ef7..8e927b9 100644 --- a/drivers/s390/cio/cio.c +++ b/drivers/s390/cio/cio.c @@ -1029,7 +1029,7 @@ extern void do_reipl_asm(__u32 schid); /* Make sure all subchannels are quiet before we re-ipl an lpar. */ void reipl_ccw_dev(struct ccw_dev_id *devid) { - struct subchannel_id schid; + struct subchannel_id uninitialized_var(schid); s390_reset_system(NULL, NULL); if (reipl_find_schid(devid, &schid) != 0) -- cgit v0.10.2 From 708c39db0ec8461c3ef1ac333ac7bb9f28833d5a Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 17 Sep 2012 06:05:16 +0200 Subject: s390/irq: use designated initializers for irq class array Use designated initializers for the irq class array in irq.c so it's always guaranteed that the order of elements is equal to their corresponding parts in irq.h. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/irq.c b/arch/s390/kernel/irq.c index c1b4493..6cdc55b 100644 --- a/arch/s390/kernel/irq.c +++ b/arch/s390/kernel/irq.c @@ -30,35 +30,35 @@ struct irq_class { }; static const struct irq_class intrclass_names[] = { - {.name = "EXT" }, - {.name = "I/O" }, - {.name = "CLK", .desc = "[EXT] Clock Comparator" }, - {.name = "EXC", .desc = "[EXT] External Call" }, - {.name = "EMS", .desc = "[EXT] Emergency Signal" }, - {.name = "TMR", .desc = "[EXT] CPU Timer" }, - {.name = "TAL", .desc = "[EXT] Timing Alert" }, - {.name = "PFL", .desc = "[EXT] Pseudo Page Fault" }, - {.name = "DSD", .desc = "[EXT] DASD Diag" }, - {.name = "VRT", .desc = "[EXT] Virtio" }, - {.name = "SCP", .desc = "[EXT] Service Call" }, - {.name = "IUC", .desc = "[EXT] IUCV" }, - {.name = "CMS", .desc = "[EXT] CPU-Measurement: Sampling" }, - {.name = "CMC", .desc = "[EXT] CPU-Measurement: Counter" }, - {.name = "CMR", .desc = "[EXT] CPU-Measurement: RI" }, - {.name = "CIO", .desc = "[I/O] Common I/O Layer Interrupt" }, - {.name = "QAI", .desc = "[I/O] QDIO Adapter Interrupt" }, - {.name = "DAS", .desc = "[I/O] DASD" }, - {.name = "C15", .desc = "[I/O] 3215" }, - {.name = "C70", .desc = "[I/O] 3270" }, - {.name = "TAP", .desc = "[I/O] Tape" }, - {.name = "VMR", .desc = "[I/O] Unit Record Devices" }, - {.name = "LCS", .desc = "[I/O] LCS" }, - {.name = "CLW", .desc = "[I/O] CLAW" }, - {.name = "CTC", .desc = "[I/O] CTC" }, - {.name = "APB", .desc = "[I/O] AP Bus" }, - {.name = "ADM", .desc = "[I/O] EADM Subchannel" }, - {.name = "CSC", .desc = "[I/O] CHSC Subchannel" }, - {.name = "NMI", .desc = "[NMI] Machine Check" }, + [EXTERNAL_INTERRUPT] = {.name = "EXT"}, + [IO_INTERRUPT] = {.name = "I/O"}, + [EXTINT_CLK] = {.name = "CLK", .desc = "[EXT] Clock Comparator"}, + [EXTINT_EXC] = {.name = "EXC", .desc = "[EXT] External Call"}, + [EXTINT_EMS] = {.name = "EMS", .desc = "[EXT] Emergency Signal"}, + [EXTINT_TMR] = {.name = "TMR", .desc = "[EXT] CPU Timer"}, + [EXTINT_TLA] = {.name = "TAL", .desc = "[EXT] Timing Alert"}, + [EXTINT_PFL] = {.name = "PFL", .desc = "[EXT] Pseudo Page Fault"}, + [EXTINT_DSD] = {.name = "DSD", .desc = "[EXT] DASD Diag"}, + [EXTINT_VRT] = {.name = "VRT", .desc = "[EXT] Virtio"}, + [EXTINT_SCP] = {.name = "SCP", .desc = "[EXT] Service Call"}, + [EXTINT_IUC] = {.name = "IUC", .desc = "[EXT] IUCV"}, + [EXTINT_CMS] = {.name = "CMS", .desc = "[EXT] CPU-Measurement: Sampling"}, + [EXTINT_CMC] = {.name = "CMC", .desc = "[EXT] CPU-Measurement: Counter"}, + [EXTINT_CMR] = {.name = "CMR", .desc = "[EXT] CPU-Measurement: RI"}, + [IOINT_CIO] = {.name = "CIO", .desc = "[I/O] Common I/O Layer Interrupt"}, + [IOINT_QAI] = {.name = "QAI", .desc = "[I/O] QDIO Adapter Interrupt"}, + [IOINT_DAS] = {.name = "DAS", .desc = "[I/O] DASD"}, + [IOINT_C15] = {.name = "C15", .desc = "[I/O] 3215"}, + [IOINT_C70] = {.name = "C70", .desc = "[I/O] 3270"}, + [IOINT_TAP] = {.name = "TAP", .desc = "[I/O] Tape"}, + [IOINT_VMR] = {.name = "VMR", .desc = "[I/O] Unit Record Devices"}, + [IOINT_LCS] = {.name = "LCS", .desc = "[I/O] LCS"}, + [IOINT_CLW] = {.name = "CLW", .desc = "[I/O] CLAW"}, + [IOINT_CTC] = {.name = "CTC", .desc = "[I/O] CTC"}, + [IOINT_APB] = {.name = "APB", .desc = "[I/O] AP Bus"}, + [IOINT_ADM] = {.name = "ADM", .desc = "[I/O] EADM Subchannel"}, + [IOINT_CSC] = {.name = "CSC", .desc = "[I/O] CHSC Subchannel"}, + [NMI_NMI] = {.name = "NMI", .desc = "[NMI] Machine Check"}, }; /* -- cgit v0.10.2 From 0ed23b3e495807809873bb607d12c857ae9ab7c2 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 17 Sep 2012 06:24:58 +0200 Subject: s390/kexec: remove CONFIG_KEXEC Since "Kconfig: split the s390 base menu" CONFIG_KEXEC gets always selected. Therefore there is no point in keeping CONFIG_KEXEC anywhere. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile index 249aaeb..4da52fe 100644 --- a/arch/s390/kernel/Makefile +++ b/arch/s390/kernel/Makefile @@ -23,10 +23,11 @@ CFLAGS_sysinfo.o += -Iinclude/math-emu -Iarch/s390/math-emu -w obj-y := bitmap.o traps.o time.o process.o base.o early.o setup.o vtime.o \ processor.o sys_s390.o ptrace.o signal.o cpcmd.o ebcdic.o nmi.o \ debug.o irq.o ipl.o dis.o diag.o mem_detect.o sclp.o vdso.o \ - sysinfo.o jump_label.o lgr.o os_info.o + sysinfo.o jump_label.o lgr.o os_info.o machine_kexec.o obj-y += $(if $(CONFIG_64BIT),entry64.o,entry.o) obj-y += $(if $(CONFIG_64BIT),reipl64.o,reipl.o) +obj-y += $(if $(CONFIG_64BIT),relocate_kernel64.o,relocate_kernel.o) extra-y += head.o vmlinux.lds extra-y += $(if $(CONFIG_64BIT),head64.o,head31.o) @@ -54,11 +55,6 @@ obj-$(CONFIG_PERF_EVENTS) += perf_event.o perf_cpum_cf.o obj-y += runtime_instr.o cache.o endif -# Kexec part -S390_KEXEC_OBJS := machine_kexec.o -S390_KEXEC_OBJS += $(if $(CONFIG_64BIT),relocate_kernel64.o,relocate_kernel.o) -obj-$(CONFIG_KEXEC) += $(S390_KEXEC_OBJS) - # vdso obj-$(CONFIG_64BIT) += vdso64/ obj-$(CONFIG_32BIT) += vdso32/ diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index 7d27b0a..afa9fdb 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -584,9 +584,7 @@ static void __init setup_memory_end(void) static void __init setup_vmcoreinfo(void) { -#ifdef CONFIG_KEXEC mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note()); -#endif } #ifdef CONFIG_CRASH_DUMP -- cgit v0.10.2 From 28634a07d3f72f2c186ad465e58f72e728c5cd4e Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 17 Sep 2012 06:38:22 +0200 Subject: s390/percpu: implement this_cpu_xchg() The generic variant has a local_irq_save/restore pair which is quite expensive. It is sufficient to disable preemption, which is a no-op with !CONFIG_PREEMPT and then use the regular xchg macro. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/percpu.h b/arch/s390/include/asm/percpu.h index 6537e72..7adebfc 100644 --- a/arch/s390/include/asm/percpu.h +++ b/arch/s390/include/asm/percpu.h @@ -84,6 +84,24 @@ do { \ #define this_cpu_cmpxchg_4(pcp, oval, nval) arch_this_cpu_cmpxchg(pcp, oval, nval) #define this_cpu_cmpxchg_8(pcp, oval, nval) arch_this_cpu_cmpxchg(pcp, oval, nval) +#define arch_this_cpu_xchg(pcp, nval) \ +({ \ + typeof(pcp) *ptr__; \ + typeof(pcp) ret__; \ + preempt_disable(); \ + ptr__ = __this_cpu_ptr(&(pcp)); \ + ret__ = xchg(ptr__, nval); \ + preempt_enable(); \ + ret__; \ +}) + +#define this_cpu_xchg_1(pcp, nval) arch_this_cpu_xchg(pcp, nval) +#define this_cpu_xchg_2(pcp, nval) arch_this_cpu_xchg(pcp, nval) +#define this_cpu_xchg_4(pcp, nval) arch_this_cpu_xchg(pcp, nval) +#ifdef CONFIG_64BIT +#define this_cpu_xchg_8(pcp, nval) arch_this_cpu_xchg(pcp, nval) +#endif + #include #endif /* __ARCH_S390_PERCPU__ */ -- cgit v0.10.2 From ba6f5c2a8da8f7d6e14702d8c9b99b5767f6db49 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 17 Sep 2012 06:46:55 +0200 Subject: s390/percpu: implement this_cpu_add_return() Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/include/asm/percpu.h b/arch/s390/include/asm/percpu.h index 7adebfc..964e7ee 100644 --- a/arch/s390/include/asm/percpu.h +++ b/arch/s390/include/asm/percpu.h @@ -20,7 +20,7 @@ #endif #define arch_this_cpu_to_op(pcp, val, op) \ -do { \ +({ \ typedef typeof(pcp) pcp_op_T__; \ pcp_op_T__ old__, new__, prev__; \ pcp_op_T__ *ptr__; \ @@ -39,13 +39,19 @@ do { \ } \ } while (prev__ != old__); \ preempt_enable(); \ -} while (0) + new__; \ +}) #define this_cpu_add_1(pcp, val) arch_this_cpu_to_op(pcp, val, +) #define this_cpu_add_2(pcp, val) arch_this_cpu_to_op(pcp, val, +) #define this_cpu_add_4(pcp, val) arch_this_cpu_to_op(pcp, val, +) #define this_cpu_add_8(pcp, val) arch_this_cpu_to_op(pcp, val, +) +#define this_cpu_add_return_1(pcp, val) arch_this_cpu_to_op(pcp, val, +) +#define this_cpu_add_return_2(pcp, val) arch_this_cpu_to_op(pcp, val, +) +#define this_cpu_add_return_4(pcp, val) arch_this_cpu_to_op(pcp, val, +) +#define this_cpu_add_return_8(pcp, val) arch_this_cpu_to_op(pcp, val, +) + #define this_cpu_and_1(pcp, val) arch_this_cpu_to_op(pcp, val, &) #define this_cpu_and_2(pcp, val) arch_this_cpu_to_op(pcp, val, &) #define this_cpu_and_4(pcp, val) arch_this_cpu_to_op(pcp, val, &) -- cgit v0.10.2 From b1d6b40cbd0d6ff475b6a0a7a807a1e3bee7c033 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 17 Sep 2012 07:37:13 +0200 Subject: s390/cmpxchg,percpu: implement cmpxchg_double() Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 3f42161..e601340 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -96,6 +96,7 @@ config S390 select HAVE_MEMBLOCK select HAVE_MEMBLOCK_NODE_MAP select HAVE_CMPXCHG_LOCAL + select HAVE_CMPXCHG_DOUBLE select ARCH_DISCARD_MEMBLOCK select BUILDTIME_EXTABLE_SORT select ARCH_INLINE_SPIN_TRYLOCK diff --git a/arch/s390/include/asm/cmpxchg.h b/arch/s390/include/asm/cmpxchg.h index 8d798e9..0f636cb 100644 --- a/arch/s390/include/asm/cmpxchg.h +++ b/arch/s390/include/asm/cmpxchg.h @@ -7,7 +7,9 @@ #ifndef __ASM_CMPXCHG_H #define __ASM_CMPXCHG_H +#include #include +#include extern void __xchg_called_with_bad_pointer(void); @@ -203,6 +205,65 @@ static inline unsigned long long __cmpxchg64(void *ptr, }) #endif /* CONFIG_64BIT */ +#define __cmpxchg_double_op(p1, p2, o1, o2, n1, n2, insn) \ +({ \ + register __typeof__(*(p1)) __old1 asm("2") = (o1); \ + register __typeof__(*(p2)) __old2 asm("3") = (o2); \ + register __typeof__(*(p1)) __new1 asm("4") = (n1); \ + register __typeof__(*(p2)) __new2 asm("5") = (n2); \ + int cc; \ + asm volatile( \ + insn " %[old],%[new],%[ptr]\n" \ + " ipm %[cc]\n" \ + " srl %[cc],28" \ + : [cc] "=d" (cc), [old] "+d" (__old1), "+d" (__old2) \ + : [new] "d" (__new1), "d" (__new2), \ + [ptr] "Q" (*(p1)), "Q" (*(p2)) \ + : "memory", "cc"); \ + !cc; \ +}) + +#define __cmpxchg_double_4(p1, p2, o1, o2, n1, n2) \ + __cmpxchg_double_op(p1, p2, o1, o2, n1, n2, "cds") + +#define __cmpxchg_double_8(p1, p2, o1, o2, n1, n2) \ + __cmpxchg_double_op(p1, p2, o1, o2, n1, n2, "cdsg") + +extern void __cmpxchg_double_called_with_bad_pointer(void); + +#define __cmpxchg_double(p1, p2, o1, o2, n1, n2) \ +({ \ + int __ret; \ + switch (sizeof(*(p1))) { \ + case 4: \ + __ret = __cmpxchg_double_4(p1, p2, o1, o2, n1, n2); \ + break; \ + case 8: \ + __ret = __cmpxchg_double_8(p1, p2, o1, o2, n1, n2); \ + break; \ + default: \ + __cmpxchg_double_called_with_bad_pointer(); \ + } \ + __ret; \ +}) + +#define cmpxchg_double(p1, p2, o1, o2, n1, n2) \ +({ \ + __typeof__(p1) __p1 = (p1); \ + __typeof__(p2) __p2 = (p2); \ + int __ret; \ + BUILD_BUG_ON(sizeof(*(p1)) != sizeof(long)); \ + BUILD_BUG_ON(sizeof(*(p2)) != sizeof(long)); \ + VM_BUG_ON((unsigned long)((__p1) + 1) != (unsigned long)(__p2));\ + if (sizeof(long) == 4) \ + __ret = __cmpxchg_double_4(__p1, __p2, o1, o2, n1, n2); \ + else \ + __ret = __cmpxchg_double_8(__p1, __p2, o1, o2, n1, n2); \ + __ret; \ +}) + +#define system_has_cmpxchg_double() 1 + #include static inline unsigned long __cmpxchg_local(void *ptr, diff --git a/arch/s390/include/asm/percpu.h b/arch/s390/include/asm/percpu.h index 964e7ee..86fe0ee 100644 --- a/arch/s390/include/asm/percpu.h +++ b/arch/s390/include/asm/percpu.h @@ -67,7 +67,7 @@ #define this_cpu_xor_4(pcp, val) arch_this_cpu_to_op(pcp, val, ^) #define this_cpu_xor_8(pcp, val) arch_this_cpu_to_op(pcp, val, ^) -#define arch_this_cpu_cmpxchg(pcp, oval, nval) \ +#define arch_this_cpu_cmpxchg(pcp, oval, nval) \ ({ \ typedef typeof(pcp) pcp_op_T__; \ pcp_op_T__ ret__; \ @@ -108,6 +108,26 @@ #define this_cpu_xchg_8(pcp, nval) arch_this_cpu_xchg(pcp, nval) #endif +#define arch_this_cpu_cmpxchg_double(pcp1, pcp2, o1, o2, n1, n2) \ +({ \ + typeof(pcp1) o1__ = (o1), n1__ = (n1); \ + typeof(pcp2) o2__ = (o2), n2__ = (n2); \ + typeof(pcp1) *p1__; \ + typeof(pcp2) *p2__; \ + int ret__; \ + preempt_disable(); \ + p1__ = __this_cpu_ptr(&(pcp1)); \ + p2__ = __this_cpu_ptr(&(pcp2)); \ + ret__ = __cmpxchg_double(p1__, p2__, o1__, o2__, n1__, n2__); \ + preempt_enable(); \ + ret__; \ +}) + +#define this_cpu_cmpxchg_double_4 arch_this_cpu_cmpxchg_double +#ifdef CONFIG_64BIT +#define this_cpu_cmpxchg_double_8 arch_this_cpu_cmpxchg_double +#endif + #include #endif /* __ARCH_S390_PERCPU__ */ -- cgit v0.10.2 From 72abaadf25f2bc6098740622e4ed526775b197f9 Mon Sep 17 00:00:00 2001 From: Peter Senna Tschudin Date: Mon, 17 Sep 2012 10:12:00 +0200 Subject: drivers/s390/char/monreader.c: fix error return code Convert a nonnegative error return code to a negative one, as returned elsewhere in the function. A simplified version of the semantic match that finds this problem is as follows: (http://coccinelle.lip6.fr/) // ( if@p1 (\(ret < 0\|ret != 0\)) { ... return ret; } | ret@p1 = 0 ) ... when != ret = e1 when != &ret *if(...) { ... when != ret = e2 when forall return ret; } // Signed-off-by: Peter Senna Tschudin Signed-off-by: Gerald Schaefer Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/char/monreader.c b/drivers/s390/char/monreader.c index 5b8b859..f4ff515 100644 --- a/drivers/s390/char/monreader.c +++ b/drivers/s390/char/monreader.c @@ -571,8 +571,11 @@ static int __init mon_init(void) if (rc) goto out_iucv; monreader_device = kzalloc(sizeof(struct device), GFP_KERNEL); - if (!monreader_device) + if (!monreader_device) { + rc = -ENOMEM; goto out_driver; + } + dev_set_name(monreader_device, "monreader-dev"); monreader_device->bus = &iucv_bus; monreader_device->parent = iucv_root; -- cgit v0.10.2 From 3bc9fef9cc1e4047c3a3c51d84cc1c5d2ef03cea Mon Sep 17 00:00:00 2001 From: Stefan Haberland Date: Thu, 20 Sep 2012 18:37:36 +0200 Subject: s390/dasd: check count address during online setting A common way to prepare a z/VM mini disk is to format the real device with a z/VM tool like CPFMTXA and then define a mini disk that excludes the first cylinder, i.e. the cylinder 0 of the virtual disk is located at cylinder 1 of the real device. The DASD device driver will recognize such a mini disk as formatted, as the uniform record layout on the disk matches that of an LDL formatted device. However, the cylinder value in the 'count' field of the ECKD records matches the geometry of the real device, and not that of the mini disk, so I/O requests will fail with 'record not found' errors. To make the mini disk usable, it needs to be formatted with a tool like dasdfmt. To enable tools like distribution installation tools to recognize this situation, the DASD device driver should report such a mini disk as 'not formatted'. To this end we need to extend the device recognition code to check not just for proper record sizes, but also for proper cylinder/head/record values. Signed-off-by: Stefan Haberland Reviewed-by: Stefan Weinhuber Signed-off-by: Martin Schwidefsky diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index e1bb930..108332b 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -139,6 +139,10 @@ dasd_eckd_set_online(struct ccw_device *cdev) static const int sizes_trk0[] = { 28, 148, 84 }; #define LABEL_SIZE 140 +/* head and record addresses of count_area read in analysis ccw */ +static const int count_area_head[] = { 0, 0, 0, 0, 2 }; +static const int count_area_rec[] = { 1, 2, 3, 4, 1 }; + static inline unsigned int round_up_multiple(unsigned int no, unsigned int mult) { @@ -1939,7 +1943,10 @@ static int dasd_eckd_end_analysis(struct dasd_block *block) count_area = NULL; for (i = 0; i < 3; i++) { if (private->count_area[i].kl != 4 || - private->count_area[i].dl != dasd_eckd_cdl_reclen(i) - 4) { + private->count_area[i].dl != dasd_eckd_cdl_reclen(i) - 4 || + private->count_area[i].cyl != 0 || + private->count_area[i].head != count_area_head[i] || + private->count_area[i].record != count_area_rec[i]) { private->uses_cdl = 0; break; } @@ -1951,7 +1958,10 @@ static int dasd_eckd_end_analysis(struct dasd_block *block) for (i = 0; i < 5; i++) { if ((private->count_area[i].kl != 0) || (private->count_area[i].dl != - private->count_area[0].dl)) + private->count_area[0].dl) || + private->count_area[i].cyl != 0 || + private->count_area[i].head != count_area_head[i] || + private->count_area[i].record != count_area_rec[i]) break; } if (i == 5) -- cgit v0.10.2 From 5e249d6e11bd92161d2dcb92691b0a6ec361627a Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 24 Sep 2012 08:17:58 +0200 Subject: s390/mm: mark free_initrd_mem() as __init Same as 0d26d1d8 "x86/mm: Mark free_initrd_mem() as __init". In addition also add the __init annotation to setup_zero_pages(). Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c index 6adbc08..81e596c 100644 --- a/arch/s390/mm/init.c +++ b/arch/s390/mm/init.c @@ -42,7 +42,7 @@ pgd_t swapper_pg_dir[PTRS_PER_PGD] __attribute__((__aligned__(PAGE_SIZE))); unsigned long empty_zero_page, zero_page_mask; EXPORT_SYMBOL(empty_zero_page); -static unsigned long setup_zero_pages(void) +static unsigned long __init setup_zero_pages(void) { struct cpuid cpu_id; unsigned int order; @@ -212,7 +212,7 @@ void free_initmem(void) } #ifdef CONFIG_BLK_DEV_INITRD -void free_initrd_mem(unsigned long start, unsigned long end) +void __init free_initrd_mem(unsigned long start, unsigned long end) { free_init_pages("initrd memory", start, end); } -- cgit v0.10.2 From 022912575ce3135517ec1cbfb8d4b240e0821feb Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 24 Sep 2012 08:22:00 +0200 Subject: s390/traps: move call to print_modules() out of show_regs() Same as 0fa0e2f0 "x86: Move call to print_modules() out of show_regs()". Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c index befc761..3d2b0fa 100644 --- a/arch/s390/kernel/traps.c +++ b/arch/s390/kernel/traps.c @@ -231,7 +231,6 @@ void show_registers(struct pt_regs *regs) void show_regs(struct pt_regs *regs) { - print_modules(); printk("CPU: %d %s %s %.*s\n", task_thread_info(current)->cpu, print_tainted(), init_utsname()->release, @@ -271,6 +270,7 @@ void die(struct pt_regs *regs, const char *str) #endif printk("\n"); notify_die(DIE_OOPS, str, regs, 0, regs->int_code & 0xffff, SIGSEGV); + print_modules(); show_regs(regs); bust_spinlocks(0); add_taint(TAINT_DIE); -- cgit v0.10.2 From c59eed111bff70bbb8c85abc73374d4da0848c79 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 24 Sep 2012 08:31:35 +0200 Subject: s390/bpf,jit: add support for BPF_S_ANC_ALU_XOR_X instruction Add support for new BPF_S_ANC_ALU_XOR_X instruction which got added with ffe06c17 "filter: add XOR operation". s390 version of 4bfaddf1 "x86 bpf_jit: support BPF_S_ANC_ALU_XOR_X instruction". Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/net/bpf_jit_comp.c b/arch/s390/net/bpf_jit_comp.c index f57f0c3..9b355b4 100644 --- a/arch/s390/net/bpf_jit_comp.c +++ b/arch/s390/net/bpf_jit_comp.c @@ -367,6 +367,11 @@ static int bpf_jit_insn(struct bpf_jit *jit, struct sock_filter *filter, /* o %r5,(%r13) */ EMIT4_DISP(0x5650d000, EMIT_CONST(K)); break; + case BPF_S_ANC_ALU_XOR_X: /* A ^= X; */ + jit->seen |= SEEN_XREG; + /* xr %r5,%r12 */ + EMIT2(0x175c); + break; case BPF_S_ALU_LSH_X: /* A <<= X; */ jit->seen |= SEEN_XREG; /* sll %r5,0(%r12) */ -- cgit v0.10.2 From f7ed68a47fa8da76c0e92ae72e0a1a04dbb4566d Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 24 Sep 2012 15:29:44 +0200 Subject: s390/disassembler: fix decoding of risblg instruction Fix typo: risblk -> risblg. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/kernel/dis.c b/arch/s390/kernel/dis.c index 7b6ad27..cc84a24 100644 --- a/arch/s390/kernel/dis.c +++ b/arch/s390/kernel/dis.c @@ -334,7 +334,7 @@ static char *long_insn_name[] = { [LONG_INSN_LLGHRL] = "llghrl", [LONG_INSN_POPCNT] = "popcnt", [LONG_INSN_RISBHG] = "risbhg", - [LONG_INSN_RISBLG] = "risblk", + [LONG_INSN_RISBLG] = "risblg", [LONG_INSN_RINEXT] = "rinext", [LONG_INSN_RIEMIT] = "riemit", [LONG_INSN_TABORT] = "tabort", -- cgit v0.10.2 From 0eccc783d85f5edd28497c1f1a62faba392b04eb Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 25 Sep 2012 14:53:15 +0200 Subject: s390/jump label,nss: let shared kernel support depend on !JUMP_LABEL We can have either shared kernel or jump label support, but not both. If a kernel gets IPL'ed from an NSS it's not possible to patch the text segment, since it's read-only. Therefore any static branches cannot be updated. So we need to make sure that shared kernel support is disabled if jump label support is enabled. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index e601340..95303f6 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -545,8 +545,8 @@ config PFAULT this option. config SHARED_KERNEL - def_bool y - prompt "VM shared kernel support" + bool "VM shared kernel support" + depends on !JUMP_LABEL help Select this option, if you want to share the text segment of the Linux kernel between different VM guests. This reduces memory -- cgit v0.10.2 From c397031f58f9a0a5d808dc998105781b1945b6fe Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Thu, 16 Aug 2012 12:07:02 +0200 Subject: s390: update defconfig Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky diff --git a/arch/s390/defconfig b/arch/s390/defconfig index f39cd71..b74400e 100644 --- a/arch/s390/defconfig +++ b/arch/s390/defconfig @@ -16,8 +16,8 @@ CONFIG_CGROUPS=y CONFIG_CPUSETS=y CONFIG_CGROUP_CPUACCT=y CONFIG_RESOURCE_COUNTERS=y -CONFIG_CGROUP_MEMCG=y -CONFIG_CGROUP_MEM_RES_CTLR_SWAP=y +CONFIG_MEMCG=y +CONFIG_MEMCG_SWAP=y CONFIG_CGROUP_SCHED=y CONFIG_RT_GROUP_SCHED=y CONFIG_BLK_CGROUP=y @@ -32,20 +32,19 @@ CONFIG_EXPERT=y CONFIG_PROFILING=y CONFIG_OPROFILE=y CONFIG_KPROBES=y +CONFIG_JUMP_LABEL=y CONFIG_MODULES=y CONFIG_MODULE_UNLOAD=y CONFIG_MODVERSIONS=y CONFIG_PARTITION_ADVANCED=y CONFIG_IBM_PARTITION=y CONFIG_DEFAULT_DEADLINE=y -CONFIG_PREEMPT=y +CONFIG_HZ_100=y CONFIG_MEMORY_HOTPLUG=y CONFIG_MEMORY_HOTREMOVE=y CONFIG_KSM=y -CONFIG_BINFMT_MISC=m -CONFIG_CMM=m -CONFIG_HZ_100=y CONFIG_CRASH_DUMP=y +CONFIG_BINFMT_MISC=m CONFIG_HIBERNATION=y CONFIG_PACKET=y CONFIG_UNIX=y @@ -75,6 +74,7 @@ CONFIG_NET_CLS_RSVP=m CONFIG_NET_CLS_RSVP6=m CONFIG_NET_CLS_ACT=y CONFIG_NET_ACT_POLICE=y +CONFIG_BPF_JIT=y CONFIG_UEVENT_HELPER_PATH="/sbin/hotplug" CONFIG_DEVTMPFS=y CONFIG_BLK_DEV_LOOP=m @@ -121,7 +121,6 @@ CONFIG_DEBUG_NOTIFIERS=y CONFIG_RCU_TRACE=y CONFIG_KPROBES_SANITY_TEST=y CONFIG_DEBUG_FORCE_WEAK_PER_CPU=y -CONFIG_CPU_NOTIFIER_ERROR_INJECT=m CONFIG_LATENCYTOP=y CONFIG_DEBUG_PAGEALLOC=y CONFIG_BLK_DEV_IO_TRACE=y @@ -173,3 +172,4 @@ CONFIG_CRYPTO_SHA512_S390=m CONFIG_CRYPTO_DES_S390=m CONFIG_CRYPTO_AES_S390=m CONFIG_CRC7=m +CONFIG_CMM=m -- cgit v0.10.2