Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

Pull s390 updates from Martin Schwidefsky:
 "Among the traditional bug fixes and cleanups are some improvements:

   - A tool to generated the facility lists, generating the bit fields
     by hand has been a source of bugs in the past

   - The spinlock loop is reordered to avoid bursts of hypervisor calls

   - Add support for the open-for-business interface to the service
     element

   - The get_cpu call is added to the vdso

   - A set of tracepoints is defined for the common I/O layer

   - The deprecated sclp_cpi module is removed

   - Update default configuration"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (56 commits)
  s390/sclp: fix possible control register corruption
  s390: fix normalization bug in exception table sorting
  s390/configs: update default configurations
  s390/vdso: optimize getcpu system call
  s390: drop smp_mb in vdso_init
  s390: rename struct _lowcore to struct lowcore
  s390/mem_detect: use unsigned longs
  s390/ptrace: get rid of long longs in psw_bits
  s390/sysinfo: add missing SYSIB 1.2.2 multithreading fields
  s390: get rid of CONFIG_SCHED_MC and CONFIG_SCHED_BOOK
  s390/Kconfig: remove pointless 64 bit dependencies
  s390/dasd: fix failfast for disconnected devices
  s390/con3270: testing return kzalloc retval
  s390/hmcdrv: constify hmcdrv_ftp_ops structs
  s390/cio: add NULL test
  s390/cio: Change I/O instructions from inline to normal functions
  s390/cio: Introduce common I/O layer tracepoints
  s390/cio: Consolidate inline assemblies and related data definitions
  s390/cio: Fix incorrect xsch opcode specification
  s390/cio: Remove unused inline assemblies
  ...

24 files changed, 695 insertions, 618 deletions

diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile
index dc167a2..2f5586a 100644
--- a/arch/s390/kernel/Makefile
+++ b/arch/s390/kernel/Makefile
@@ -34,8 +34,10 @@ CFLAGS_sysinfo.o += -w
 #
 CFLAGS_REMOVE_sclp.o = $(CC_FLAGS_FTRACE)
 ifneq ($(CC_FLAGS_MARCH),-march=z900)
-CFLAGS_REMOVE_sclp.o += $(CC_FLAGS_MARCH)
-CFLAGS_sclp.o	+= -march=z900
+CFLAGS_REMOVE_sclp.o	+= $(CC_FLAGS_MARCH)
+CFLAGS_sclp.o		+= -march=z900
+AFLAGS_REMOVE_head.o	+= $(CC_FLAGS_MARCH)
+AFLAGS_head.o		+= -march=z900
 endif
 GCOV_PROFILE_sclp.o := n
 
@@ -50,7 +52,7 @@ extra-y				+= head.o head64.o vmlinux.lds
 
 obj-$(CONFIG_MODULES)		+= s390_ksyms.o module.o
 obj-$(CONFIG_SMP)		+= smp.o
-obj-$(CONFIG_SCHED_BOOK)	+= topology.o
+obj-$(CONFIG_SCHED_TOPOLOGY)	+= topology.o
 obj-$(CONFIG_HIBERNATION)	+= suspend.o swsusp.o
 obj-$(CONFIG_AUDIT)		+= audit.o
 compat-obj-$(CONFIG_AUDIT)	+= compat_audit.o
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index 9cd248f..53bbc9e 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -80,6 +80,8 @@ int main(void)
 	OFFSET(__VDSO_TK_SHIFT, vdso_data, tk_shift);
 	OFFSET(__VDSO_ECTG_BASE, vdso_per_cpu_data, ectg_timer_base);
 	OFFSET(__VDSO_ECTG_USER, vdso_per_cpu_data, ectg_user_time);
+	OFFSET(__VDSO_CPU_NR, vdso_per_cpu_data, cpu_nr);
+	OFFSET(__VDSO_NODE_ID, vdso_per_cpu_data, node_id);
 	BLANK();
 	/* constants used by the vdso */
 	DEFINE(__CLOCK_REALTIME, CLOCK_REALTIME);
@@ -97,95 +99,96 @@ int main(void)
 	OFFSET(__TIMER_IDLE_EXIT, s390_idle_data, timer_idle_exit);
 	BLANK();
 	/* hardware defined lowcore locations 0x000 - 0x1ff */
-	OFFSET(__LC_EXT_PARAMS, _lowcore, ext_params);
-	OFFSET(__LC_EXT_CPU_ADDR, _lowcore, ext_cpu_addr);
-	OFFSET(__LC_EXT_INT_CODE, _lowcore, ext_int_code);
-	OFFSET(__LC_SVC_ILC, _lowcore, svc_ilc);
-	OFFSET(__LC_SVC_INT_CODE, _lowcore, svc_code);
-	OFFSET(__LC_PGM_ILC, _lowcore, pgm_ilc);
-	OFFSET(__LC_PGM_INT_CODE, _lowcore, pgm_code);
-	OFFSET(__LC_DATA_EXC_CODE, _lowcore, data_exc_code);
-	OFFSET(__LC_MON_CLASS_NR, _lowcore, mon_class_num);
-	OFFSET(__LC_PER_CODE, _lowcore, per_code);
-	OFFSET(__LC_PER_ATMID, _lowcore, per_atmid);
-	OFFSET(__LC_PER_ADDRESS, _lowcore, per_address);
-	OFFSET(__LC_EXC_ACCESS_ID, _lowcore, exc_access_id);
-	OFFSET(__LC_PER_ACCESS_ID, _lowcore, per_access_id);
-	OFFSET(__LC_OP_ACCESS_ID, _lowcore, op_access_id);
-	OFFSET(__LC_AR_MODE_ID, _lowcore, ar_mode_id);
-	OFFSET(__LC_TRANS_EXC_CODE, _lowcore, trans_exc_code);
-	OFFSET(__LC_MON_CODE, _lowcore, monitor_code);
-	OFFSET(__LC_SUBCHANNEL_ID, _lowcore, subchannel_id);
-	OFFSET(__LC_SUBCHANNEL_NR, _lowcore, subchannel_nr);
-	OFFSET(__LC_IO_INT_PARM, _lowcore, io_int_parm);
-	OFFSET(__LC_IO_INT_WORD, _lowcore, io_int_word);
-	OFFSET(__LC_STFL_FAC_LIST, _lowcore, stfl_fac_list);
-	OFFSET(__LC_MCCK_CODE, _lowcore, mcck_interruption_code);
-	OFFSET(__LC_MCCK_FAIL_STOR_ADDR, _lowcore, failing_storage_address);
-	OFFSET(__LC_LAST_BREAK, _lowcore, breaking_event_addr);
-	OFFSET(__LC_RST_OLD_PSW, _lowcore, restart_old_psw);
-	OFFSET(__LC_EXT_OLD_PSW, _lowcore, external_old_psw);
-	OFFSET(__LC_SVC_OLD_PSW, _lowcore, svc_old_psw);
-	OFFSET(__LC_PGM_OLD_PSW, _lowcore, program_old_psw);
-	OFFSET(__LC_MCK_OLD_PSW, _lowcore, mcck_old_psw);
-	OFFSET(__LC_IO_OLD_PSW, _lowcore, io_old_psw);
-	OFFSET(__LC_RST_NEW_PSW, _lowcore, restart_psw);
-	OFFSET(__LC_EXT_NEW_PSW, _lowcore, external_new_psw);
-	OFFSET(__LC_SVC_NEW_PSW, _lowcore, svc_new_psw);
-	OFFSET(__LC_PGM_NEW_PSW, _lowcore, program_new_psw);
-	OFFSET(__LC_MCK_NEW_PSW, _lowcore, mcck_new_psw);
-	OFFSET(__LC_IO_NEW_PSW, _lowcore, io_new_psw);
+	OFFSET(__LC_EXT_PARAMS, lowcore, ext_params);
+	OFFSET(__LC_EXT_CPU_ADDR, lowcore, ext_cpu_addr);
+	OFFSET(__LC_EXT_INT_CODE, lowcore, ext_int_code);
+	OFFSET(__LC_SVC_ILC, lowcore, svc_ilc);
+	OFFSET(__LC_SVC_INT_CODE, lowcore, svc_code);
+	OFFSET(__LC_PGM_ILC, lowcore, pgm_ilc);
+	OFFSET(__LC_PGM_INT_CODE, lowcore, pgm_code);
+	OFFSET(__LC_DATA_EXC_CODE, lowcore, data_exc_code);
+	OFFSET(__LC_MON_CLASS_NR, lowcore, mon_class_num);
+	OFFSET(__LC_PER_CODE, lowcore, per_code);
+	OFFSET(__LC_PER_ATMID, lowcore, per_atmid);
+	OFFSET(__LC_PER_ADDRESS, lowcore, per_address);
+	OFFSET(__LC_EXC_ACCESS_ID, lowcore, exc_access_id);
+	OFFSET(__LC_PER_ACCESS_ID, lowcore, per_access_id);
+	OFFSET(__LC_OP_ACCESS_ID, lowcore, op_access_id);
+	OFFSET(__LC_AR_MODE_ID, lowcore, ar_mode_id);
+	OFFSET(__LC_TRANS_EXC_CODE, lowcore, trans_exc_code);
+	OFFSET(__LC_MON_CODE, lowcore, monitor_code);
+	OFFSET(__LC_SUBCHANNEL_ID, lowcore, subchannel_id);
+	OFFSET(__LC_SUBCHANNEL_NR, lowcore, subchannel_nr);
+	OFFSET(__LC_IO_INT_PARM, lowcore, io_int_parm);
+	OFFSET(__LC_IO_INT_WORD, lowcore, io_int_word);
+	OFFSET(__LC_STFL_FAC_LIST, lowcore, stfl_fac_list);
+	OFFSET(__LC_STFLE_FAC_LIST, lowcore, stfle_fac_list);
+	OFFSET(__LC_MCCK_CODE, lowcore, mcck_interruption_code);
+	OFFSET(__LC_MCCK_FAIL_STOR_ADDR, lowcore, failing_storage_address);
+	OFFSET(__LC_LAST_BREAK, lowcore, breaking_event_addr);
+	OFFSET(__LC_RST_OLD_PSW, lowcore, restart_old_psw);
+	OFFSET(__LC_EXT_OLD_PSW, lowcore, external_old_psw);
+	OFFSET(__LC_SVC_OLD_PSW, lowcore, svc_old_psw);
+	OFFSET(__LC_PGM_OLD_PSW, lowcore, program_old_psw);
+	OFFSET(__LC_MCK_OLD_PSW, lowcore, mcck_old_psw);
+	OFFSET(__LC_IO_OLD_PSW, lowcore, io_old_psw);
+	OFFSET(__LC_RST_NEW_PSW, lowcore, restart_psw);
+	OFFSET(__LC_EXT_NEW_PSW, lowcore, external_new_psw);
+	OFFSET(__LC_SVC_NEW_PSW, lowcore, svc_new_psw);
+	OFFSET(__LC_PGM_NEW_PSW, lowcore, program_new_psw);
+	OFFSET(__LC_MCK_NEW_PSW, lowcore, mcck_new_psw);
+	OFFSET(__LC_IO_NEW_PSW, lowcore, io_new_psw);
 	/* software defined lowcore locations 0x200 - 0xdff*/
-	OFFSET(__LC_SAVE_AREA_SYNC, _lowcore, save_area_sync);
-	OFFSET(__LC_SAVE_AREA_ASYNC, _lowcore, save_area_async);
-	OFFSET(__LC_SAVE_AREA_RESTART, _lowcore, save_area_restart);
-	OFFSET(__LC_CPU_FLAGS, _lowcore, cpu_flags);
-	OFFSET(__LC_RETURN_PSW, _lowcore, return_psw);
-	OFFSET(__LC_RETURN_MCCK_PSW, _lowcore, return_mcck_psw);
-	OFFSET(__LC_SYNC_ENTER_TIMER, _lowcore, sync_enter_timer);
-	OFFSET(__LC_ASYNC_ENTER_TIMER, _lowcore, async_enter_timer);
-	OFFSET(__LC_MCCK_ENTER_TIMER, _lowcore, mcck_enter_timer);
-	OFFSET(__LC_EXIT_TIMER, _lowcore, exit_timer);
-	OFFSET(__LC_USER_TIMER, _lowcore, user_timer);
-	OFFSET(__LC_SYSTEM_TIMER, _lowcore, system_timer);
-	OFFSET(__LC_STEAL_TIMER, _lowcore, steal_timer);
-	OFFSET(__LC_LAST_UPDATE_TIMER, _lowcore, last_update_timer);
-	OFFSET(__LC_LAST_UPDATE_CLOCK, _lowcore, last_update_clock);
-	OFFSET(__LC_INT_CLOCK, _lowcore, int_clock);
-	OFFSET(__LC_MCCK_CLOCK, _lowcore, mcck_clock);
-	OFFSET(__LC_CURRENT, _lowcore, current_task);
-	OFFSET(__LC_THREAD_INFO, _lowcore, thread_info);
-	OFFSET(__LC_KERNEL_STACK, _lowcore, kernel_stack);
-	OFFSET(__LC_ASYNC_STACK, _lowcore, async_stack);
-	OFFSET(__LC_PANIC_STACK, _lowcore, panic_stack);
-	OFFSET(__LC_RESTART_STACK, _lowcore, restart_stack);
-	OFFSET(__LC_RESTART_FN, _lowcore, restart_fn);
-	OFFSET(__LC_RESTART_DATA, _lowcore, restart_data);
-	OFFSET(__LC_RESTART_SOURCE, _lowcore, restart_source);
-	OFFSET(__LC_USER_ASCE, _lowcore, user_asce);
-	OFFSET(__LC_LPP, _lowcore, lpp);
-	OFFSET(__LC_CURRENT_PID, _lowcore, current_pid);
-	OFFSET(__LC_PERCPU_OFFSET, _lowcore, percpu_offset);
-	OFFSET(__LC_VDSO_PER_CPU, _lowcore, vdso_per_cpu_data);
-	OFFSET(__LC_MACHINE_FLAGS, _lowcore, machine_flags);
-	OFFSET(__LC_GMAP, _lowcore, gmap);
-	OFFSET(__LC_PASTE, _lowcore, paste);
+	OFFSET(__LC_SAVE_AREA_SYNC, lowcore, save_area_sync);
+	OFFSET(__LC_SAVE_AREA_ASYNC, lowcore, save_area_async);
+	OFFSET(__LC_SAVE_AREA_RESTART, lowcore, save_area_restart);
+	OFFSET(__LC_CPU_FLAGS, lowcore, cpu_flags);
+	OFFSET(__LC_RETURN_PSW, lowcore, return_psw);
+	OFFSET(__LC_RETURN_MCCK_PSW, lowcore, return_mcck_psw);
+	OFFSET(__LC_SYNC_ENTER_TIMER, lowcore, sync_enter_timer);
+	OFFSET(__LC_ASYNC_ENTER_TIMER, lowcore, async_enter_timer);
+	OFFSET(__LC_MCCK_ENTER_TIMER, lowcore, mcck_enter_timer);
+	OFFSET(__LC_EXIT_TIMER, lowcore, exit_timer);
+	OFFSET(__LC_USER_TIMER, lowcore, user_timer);
+	OFFSET(__LC_SYSTEM_TIMER, lowcore, system_timer);
+	OFFSET(__LC_STEAL_TIMER, lowcore, steal_timer);
+	OFFSET(__LC_LAST_UPDATE_TIMER, lowcore, last_update_timer);
+	OFFSET(__LC_LAST_UPDATE_CLOCK, lowcore, last_update_clock);
+	OFFSET(__LC_INT_CLOCK, lowcore, int_clock);
+	OFFSET(__LC_MCCK_CLOCK, lowcore, mcck_clock);
+	OFFSET(__LC_CURRENT, lowcore, current_task);
+	OFFSET(__LC_THREAD_INFO, lowcore, thread_info);
+	OFFSET(__LC_KERNEL_STACK, lowcore, kernel_stack);
+	OFFSET(__LC_ASYNC_STACK, lowcore, async_stack);
+	OFFSET(__LC_PANIC_STACK, lowcore, panic_stack);
+	OFFSET(__LC_RESTART_STACK, lowcore, restart_stack);
+	OFFSET(__LC_RESTART_FN, lowcore, restart_fn);
+	OFFSET(__LC_RESTART_DATA, lowcore, restart_data);
+	OFFSET(__LC_RESTART_SOURCE, lowcore, restart_source);
+	OFFSET(__LC_USER_ASCE, lowcore, user_asce);
+	OFFSET(__LC_LPP, lowcore, lpp);
+	OFFSET(__LC_CURRENT_PID, lowcore, current_pid);
+	OFFSET(__LC_PERCPU_OFFSET, lowcore, percpu_offset);
+	OFFSET(__LC_VDSO_PER_CPU, lowcore, vdso_per_cpu_data);
+	OFFSET(__LC_MACHINE_FLAGS, lowcore, machine_flags);
+	OFFSET(__LC_GMAP, lowcore, gmap);
+	OFFSET(__LC_PASTE, lowcore, paste);
 	/* software defined ABI-relevant lowcore locations 0xe00 - 0xe20 */
-	OFFSET(__LC_DUMP_REIPL, _lowcore, ipib);
+	OFFSET(__LC_DUMP_REIPL, lowcore, ipib);
 	/* hardware defined lowcore locations 0x1000 - 0x18ff */
-	OFFSET(__LC_VX_SAVE_AREA_ADDR, _lowcore, vector_save_area_addr);
-	OFFSET(__LC_EXT_PARAMS2, _lowcore, ext_params2);
-	OFFSET(SAVE_AREA_BASE, _lowcore, floating_pt_save_area);
-	OFFSET(__LC_FPREGS_SAVE_AREA, _lowcore, floating_pt_save_area);
-	OFFSET(__LC_GPREGS_SAVE_AREA, _lowcore, gpregs_save_area);
-	OFFSET(__LC_PSW_SAVE_AREA, _lowcore, psw_save_area);
-	OFFSET(__LC_PREFIX_SAVE_AREA, _lowcore, prefixreg_save_area);
-	OFFSET(__LC_FP_CREG_SAVE_AREA, _lowcore, fpt_creg_save_area);
-	OFFSET(__LC_CPU_TIMER_SAVE_AREA, _lowcore, cpu_timer_save_area);
-	OFFSET(__LC_CLOCK_COMP_SAVE_AREA, _lowcore, clock_comp_save_area);
-	OFFSET(__LC_AREGS_SAVE_AREA, _lowcore, access_regs_save_area);
-	OFFSET(__LC_CREGS_SAVE_AREA, _lowcore, cregs_save_area);
-	OFFSET(__LC_PGM_TDB, _lowcore, pgm_tdb);
+	OFFSET(__LC_VX_SAVE_AREA_ADDR, lowcore, vector_save_area_addr);
+	OFFSET(__LC_EXT_PARAMS2, lowcore, ext_params2);
+	OFFSET(__LC_FPREGS_SAVE_AREA, lowcore, floating_pt_save_area);
+	OFFSET(__LC_GPREGS_SAVE_AREA, lowcore, gpregs_save_area);
+	OFFSET(__LC_PSW_SAVE_AREA, lowcore, psw_save_area);
+	OFFSET(__LC_PREFIX_SAVE_AREA, lowcore, prefixreg_save_area);
+	OFFSET(__LC_FP_CREG_SAVE_AREA, lowcore, fpt_creg_save_area);
+	OFFSET(__LC_TOD_PROGREG_SAVE_AREA, lowcore, tod_progreg_save_area);
+	OFFSET(__LC_CPU_TIMER_SAVE_AREA, lowcore, cpu_timer_save_area);
+	OFFSET(__LC_CLOCK_COMP_SAVE_AREA, lowcore, clock_comp_save_area);
+	OFFSET(__LC_AREGS_SAVE_AREA, lowcore, access_regs_save_area);
+	OFFSET(__LC_CREGS_SAVE_AREA, lowcore, cregs_save_area);
+	OFFSET(__LC_PGM_TDB, lowcore, pgm_tdb);
 	BLANK();
 	/* gmap/sie offsets */
 	OFFSET(__GMAP_ASCE, gmap, asce);
diff --git a/arch/s390/kernel/crash_dump.c b/arch/s390/kernel/crash_dump.c
index 171e09b..a92b39f 100644
--- a/arch/s390/kernel/crash_dump.c
+++ b/arch/s390/kernel/crash_dump.c
@@ -13,6 +13,7 @@
 #include <linux/slab.h>
 #include <linux/bootmem.h>
 #include <linux/elf.h>
+#include <asm/asm-offsets.h>
 #include <linux/memblock.h>
 #include <asm/os_info.h>
 #include <asm/elf.h>
@@ -32,7 +33,84 @@ static struct memblock_type oldmem_type = {
 	.regions = &oldmem_region,
 };
 
-struct dump_save_areas dump_save_areas;
+struct save_area {
+	struct list_head list;
+	u64 psw[2];
+	u64 ctrs[16];
+	u64 gprs[16];
+	u32 acrs[16];
+	u64 fprs[16];
+	u32 fpc;
+	u32 prefix;
+	u64 todpreg;
+	u64 timer;
+	u64 todcmp;
+	u64 vxrs_low[16];
+	__vector128 vxrs_high[16];
+};
+
+static LIST_HEAD(dump_save_areas);
+
+/*
+ * Allocate a save area
+ */
+struct save_area * __init save_area_alloc(bool is_boot_cpu)
+{
+	struct save_area *sa;
+
+	sa = (void *) memblock_alloc(sizeof(*sa), 8);
+	if (!sa)
+		return NULL;
+	if (is_boot_cpu)
+		list_add(&sa->list, &dump_save_areas);
+	else
+		list_add_tail(&sa->list, &dump_save_areas);
+	return sa;
+}
+
+/*
+ * Return the address of the save area for the boot CPU
+ */
+struct save_area * __init save_area_boot_cpu(void)
+{
+	if (list_empty(&dump_save_areas))
+		return NULL;
+	return list_first_entry(&dump_save_areas, struct save_area, list);
+}
+
+/*
+ * Copy CPU registers into the save area
+ */
+void __init save_area_add_regs(struct save_area *sa, void *regs)
+{
+	struct lowcore *lc;
+
+	lc = (struct lowcore *)(regs - __LC_FPREGS_SAVE_AREA);
+	memcpy(&sa->psw, &lc->psw_save_area, sizeof(sa->psw));
+	memcpy(&sa->ctrs, &lc->cregs_save_area, sizeof(sa->ctrs));
+	memcpy(&sa->gprs, &lc->gpregs_save_area, sizeof(sa->gprs));
+	memcpy(&sa->acrs, &lc->access_regs_save_area, sizeof(sa->acrs));
+	memcpy(&sa->fprs, &lc->floating_pt_save_area, sizeof(sa->fprs));
+	memcpy(&sa->fpc, &lc->fpt_creg_save_area, sizeof(sa->fpc));
+	memcpy(&sa->prefix, &lc->prefixreg_save_area, sizeof(sa->prefix));
+	memcpy(&sa->todpreg, &lc->tod_progreg_save_area, sizeof(sa->todpreg));
+	memcpy(&sa->timer, &lc->cpu_timer_save_area, sizeof(sa->timer));
+	memcpy(&sa->todcmp, &lc->clock_comp_save_area, sizeof(sa->todcmp));
+}
+
+/*
+ * Copy vector registers into the save area
+ */
+void __init save_area_add_vxrs(struct save_area *sa, __vector128 *vxrs)
+{
+	int i;
+
+	/* Copy lower halves of vector registers 0-15 */
+	for (i = 0; i < 16; i++)
+		memcpy(&sa->vxrs_low[i], &vxrs[i].u[2], 8);
+	/* Copy vector registers 16-31 */
+	memcpy(sa->vxrs_high, vxrs + 16, 16 * sizeof(__vector128));
+}
 
 /*
  * Return physical address for virtual address
@@ -51,79 +129,85 @@ static inline void *load_real_addr(void *addr)
 }
 
 /*
- * Copy real to virtual or real memory
- */
-static int copy_from_realmem(void *dest, void *src, size_t count)
-{
-	unsigned long size;
-
-	if (!count)
-		return 0;
-	if (!is_vmalloc_or_module_addr(dest))
-		return memcpy_real(dest, src, count);
-	do {
-		size = min(count, PAGE_SIZE - (__pa(dest) & ~PAGE_MASK));
-		if (memcpy_real(load_real_addr(dest), src, size))
-			return -EFAULT;
-		count -= size;
-		dest += size;
-		src += size;
-	} while (count);
-	return 0;
-}
-
-/*
- * Pointer to ELF header in new kernel
- */
-static void *elfcorehdr_newmem;
-
-/*
- * Copy one page from zfcpdump "oldmem"
- *
- * For pages below HSA size memory from the HSA is copied. Otherwise
- * real memory copy is used.
+ * Copy memory of the old, dumped system to a kernel space virtual address
  */
-static ssize_t copy_oldmem_page_zfcpdump(char *buf, size_t csize,
-					 unsigned long src, int userbuf)
+int copy_oldmem_kernel(void *dst, void *src, size_t count)
 {
+	unsigned long from, len;
+	void *ra;
 	int rc;
 
-	if (src < sclp.hsa_size) {
-		rc = memcpy_hsa(buf, src, csize, userbuf);
-	} else {
-		if (userbuf)
-			rc = copy_to_user_real((void __force __user *) buf,
-					       (void *) src, csize);
-		else
-			rc = memcpy_real(buf, (void *) src, csize);
+	while (count) {
+		from = __pa(src);
+		if (!OLDMEM_BASE && from < sclp.hsa_size) {
+			/* Copy from zfcpdump HSA area */
+			len = min(count, sclp.hsa_size - from);
+			rc = memcpy_hsa_kernel(dst, from, len);
+			if (rc)
+				return rc;
+		} else {
+			/* Check for swapped kdump oldmem areas */
+			if (OLDMEM_BASE && from - OLDMEM_BASE < OLDMEM_SIZE) {
+				from -= OLDMEM_BASE;
+				len = min(count, OLDMEM_SIZE - from);
+			} else if (OLDMEM_BASE && from < OLDMEM_SIZE) {
+				len = min(count, OLDMEM_SIZE - from);
+				from += OLDMEM_BASE;
+			} else {
+				len = count;
+			}
+			if (is_vmalloc_or_module_addr(dst)) {
+				ra = load_real_addr(dst);
+				len = min(PAGE_SIZE - offset_in_page(ra), len);
+			} else {
+				ra = dst;
+			}
+			if (memcpy_real(ra, (void *) from, len))
+				return -EFAULT;
+		}
+		dst += len;
+		src += len;
+		count -= len;
 	}
-	return rc ? rc : csize;
+	return 0;
 }
 
 /*
- * Copy one page from kdump "oldmem"
- *
- * For the kdump reserved memory this functions performs a swap operation:
- *  - [OLDMEM_BASE - OLDMEM_BASE + OLDMEM_SIZE] is mapped to [0 - OLDMEM_SIZE].
- *  - [0 - OLDMEM_SIZE] is mapped to [OLDMEM_BASE - OLDMEM_BASE + OLDMEM_SIZE]
+ * Copy memory of the old, dumped system to a user space virtual address
  */
-static ssize_t copy_oldmem_page_kdump(char *buf, size_t csize,
-				      unsigned long src, int userbuf)
-
+int copy_oldmem_user(void __user *dst, void *src, size_t count)
 {
+	unsigned long from, len;
 	int rc;
 
-	if (src < OLDMEM_SIZE)
-		src += OLDMEM_BASE;
-	else if (src > OLDMEM_BASE &&
-		 src < OLDMEM_BASE + OLDMEM_SIZE)
-		src -= OLDMEM_BASE;
-	if (userbuf)
-		rc = copy_to_user_real((void __force __user *) buf,
-				       (void *) src, csize);
-	else
-		rc = copy_from_realmem(buf, (void *) src, csize);
-	return (rc == 0) ? rc : csize;
+	while (count) {
+		from = __pa(src);
+		if (!OLDMEM_BASE && from < sclp.hsa_size) {
+			/* Copy from zfcpdump HSA area */
+			len = min(count, sclp.hsa_size - from);
+			rc = memcpy_hsa_user(dst, from, len);
+			if (rc)
+				return rc;
+		} else {
+			/* Check for swapped kdump oldmem areas */
+			if (OLDMEM_BASE && from - OLDMEM_BASE < OLDMEM_SIZE) {
+				from -= OLDMEM_BASE;
+				len = min(count, OLDMEM_SIZE - from);
+			} else if (OLDMEM_BASE && from < OLDMEM_SIZE) {
+				len = min(count, OLDMEM_SIZE - from);
+				from += OLDMEM_BASE;
+			} else {
+				len = count;
+			}
+			rc = copy_to_user_real(dst, (void *) from, count);
+			if (rc)
+				return rc;
+		}
+		dst += len;
+		src += len;
+		count -= len;
+	}
+	return 0;
 }
 
 /*
@@ -132,15 +216,17 @@ static ssize_t copy_oldmem_page_kdump(char *buf, size_t csize,
 ssize_t copy_oldmem_page(unsigned long pfn, char *buf, size_t csize,
 			 unsigned long offset, int userbuf)
 {
-	unsigned long src;
+	void *src;
+	int rc;
 
 	if (!csize)
 		return 0;
-	src = (pfn << PAGE_SHIFT) + offset;
-	if (OLDMEM_BASE)
-		return copy_oldmem_page_kdump(buf, csize, src, userbuf);
+	src = (void *) (pfn << PAGE_SHIFT) + offset;
+	if (userbuf)
+		rc = copy_oldmem_user((void __force __user *) buf, src, csize);
 	else
-		return copy_oldmem_page_zfcpdump(buf, csize, src, userbuf);
+		rc = copy_oldmem_kernel((void *) buf, src, csize);
+	return rc;
 }
 
 /*
@@ -209,33 +295,6 @@ int remap_oldmem_pfn_range(struct vm_area_struct *vma, unsigned long from,
 }
 
 /*
- * Copy memory from old kernel
- */
-int copy_from_oldmem(void *dest, void *src, size_t count)
-{
-	unsigned long copied = 0;
-	int rc;
-
-	if (OLDMEM_BASE) {
-		if ((unsigned long) src < OLDMEM_SIZE) {
-			copied = min(count, OLDMEM_SIZE - (unsigned long) src);
-			rc = copy_from_realmem(dest, src + OLDMEM_BASE, copied);
-			if (rc)
-				return rc;
-		}
-	} else {
-		unsigned long hsa_end = sclp.hsa_size;
-		if ((unsigned long) src < hsa_end) {
-			copied = min(count, hsa_end - (unsigned long) src);
-			rc = memcpy_hsa(dest, (unsigned long) src, copied, 0);
-			if (rc)
-				return rc;
-		}
-	}
-	return copy_from_realmem(dest + copied, src + copied, count - copied);
-}
-
-/*
  * Alloc memory and panic in case of ENOMEM
  */
 static void *kzalloc_panic(int len)
@@ -251,8 +310,8 @@ static void *kzalloc_panic(int len)
 /*
  * Initialize ELF note
  */
-static void *nt_init(void *buf, Elf64_Word type, void *desc, int d_len,
-		     const char *name)
+static void *nt_init_name(void *buf, Elf64_Word type, void *desc, int d_len,
+			  const char *name)
 {
 	Elf64_Nhdr *note;
 	u64 len;
@@ -272,136 +331,42 @@ static void *nt_init(void *buf, Elf64_Word type, void *desc, int d_len,
 	return PTR_ADD(buf, len);
 }
 
-/*
- * Initialize prstatus note
- */
-static void *nt_prstatus(void *ptr, struct save_area *sa)
+static inline void *nt_init(void *buf, Elf64_Word type, void *desc, int d_len)
 {
-	struct elf_prstatus nt_prstatus;
-	static int cpu_nr = 1;
-
-	memset(&nt_prstatus, 0, sizeof(nt_prstatus));
-	memcpy(&nt_prstatus.pr_reg.gprs, sa->gp_regs, sizeof(sa->gp_regs));
-	memcpy(&nt_prstatus.pr_reg.psw, sa->psw, sizeof(sa->psw));
-	memcpy(&nt_prstatus.pr_reg.acrs, sa->acc_regs, sizeof(sa->acc_regs));
-	nt_prstatus.pr_pid = cpu_nr;
-	cpu_nr++;
-
-	return nt_init(ptr, NT_PRSTATUS, &nt_prstatus, sizeof(nt_prstatus),
-			 "CORE");
+	return nt_init_name(buf, type, desc, d_len, KEXEC_CORE_NOTE_NAME);
 }
 
 /*
- * Initialize fpregset (floating point) note
+ * Fill ELF notes for one CPU with save area registers
  */
-static void *nt_fpregset(void *ptr, struct save_area *sa)
+static void *fill_cpu_elf_notes(void *ptr, int cpu, struct save_area *sa)
 {
+	struct elf_prstatus nt_prstatus;
 	elf_fpregset_t nt_fpregset;
 
+	/* Prepare prstatus note */
+	memset(&nt_prstatus, 0, sizeof(nt_prstatus));
+	memcpy(&nt_prstatus.pr_reg.gprs, sa->gprs, sizeof(sa->gprs));
+	memcpy(&nt_prstatus.pr_reg.psw, sa->psw, sizeof(sa->psw));
+	memcpy(&nt_prstatus.pr_reg.acrs, sa->acrs, sizeof(sa->acrs));
+	nt_prstatus.pr_pid = cpu;
+	/* Prepare fpregset (floating point) note */
 	memset(&nt_fpregset, 0, sizeof(nt_fpregset));
-	memcpy(&nt_fpregset.fpc, &sa->fp_ctrl_reg, sizeof(sa->fp_ctrl_reg));
-	memcpy(&nt_fpregset.fprs, &sa->fp_regs, sizeof(sa->fp_regs));
-
-	return nt_init(ptr, NT_PRFPREG, &nt_fpregset, sizeof(nt_fpregset),
-		       "CORE");
-}
-
-/*
- * Initialize timer note
- */
-static void *nt_s390_timer(void *ptr, struct save_area *sa)
-{
-	return nt_init(ptr, NT_S390_TIMER, &sa->timer, sizeof(sa->timer),
-			 KEXEC_CORE_NOTE_NAME);
-}
-
-/*
- * Initialize TOD clock comparator note
- */
-static void *nt_s390_tod_cmp(void *ptr, struct save_area *sa)
-{
-	return nt_init(ptr, NT_S390_TODCMP, &sa->clk_cmp,
-		       sizeof(sa->clk_cmp), KEXEC_CORE_NOTE_NAME);
-}
-
-/*
- * Initialize TOD programmable register note
- */
-static void *nt_s390_tod_preg(void *ptr, struct save_area *sa)
-{
-	return nt_init(ptr, NT_S390_TODPREG, &sa->tod_reg,
-		       sizeof(sa->tod_reg), KEXEC_CORE_NOTE_NAME);
-}
-
-/*
- * Initialize control register note
- */
-static void *nt_s390_ctrs(void *ptr, struct save_area *sa)
-{
-	return nt_init(ptr, NT_S390_CTRS, &sa->ctrl_regs,
-		       sizeof(sa->ctrl_regs), KEXEC_CORE_NOTE_NAME);
-}
-
-/*
- * Initialize prefix register note
- */
-static void *nt_s390_prefix(void *ptr, struct save_area *sa)
-{
-	return nt_init(ptr, NT_S390_PREFIX, &sa->pref_reg,
-			 sizeof(sa->pref_reg), KEXEC_CORE_NOTE_NAME);
-}
-
-/*
- * Initialize vxrs high note (full 128 bit VX registers 16-31)
- */
-static void *nt_s390_vx_high(void *ptr, __vector128 *vx_regs)
-{
-	return nt_init(ptr, NT_S390_VXRS_HIGH, &vx_regs[16],
-		       16 * sizeof(__vector128), KEXEC_CORE_NOTE_NAME);
-}
-
-/*
- * Initialize vxrs low note (lower halves of VX registers 0-15)
- */
-static void *nt_s390_vx_low(void *ptr, __vector128 *vx_regs)
-{
-	Elf64_Nhdr *note;
-	u64 len;
-	int i;
-
-	note = (Elf64_Nhdr *)ptr;
-	note->n_namesz = strlen(KEXEC_CORE_NOTE_NAME) + 1;
-	note->n_descsz = 16 * 8;
-	note->n_type = NT_S390_VXRS_LOW;
-	len = sizeof(Elf64_Nhdr);
-
-	memcpy(ptr + len, KEXEC_CORE_NOTE_NAME, note->n_namesz);
-	len = roundup(len + note->n_namesz, 4);
-
-	ptr += len;
-	/* Copy lower halves of SIMD registers 0-15 */
-	for (i = 0; i < 16; i++) {
-		memcpy(ptr, &vx_regs[i].u[2], 8);
-		ptr += 8;
-	}
-	return ptr;
-}
-
-/*
- * Fill ELF notes for one CPU with save area registers
- */
-void *fill_cpu_elf_notes(void *ptr, struct save_area *sa, __vector128 *vx_regs)
-{
-	ptr = nt_prstatus(ptr, sa);
-	ptr = nt_fpregset(ptr, sa);
-	ptr = nt_s390_timer(ptr, sa);
-	ptr = nt_s390_tod_cmp(ptr, sa);
-	ptr = nt_s390_tod_preg(ptr, sa);
-	ptr = nt_s390_ctrs(ptr, sa);
-	ptr = nt_s390_prefix(ptr, sa);
-	if (MACHINE_HAS_VX && vx_regs) {
-		ptr = nt_s390_vx_low(ptr, vx_regs);
-		ptr = nt_s390_vx_high(ptr, vx_regs);
+	memcpy(&nt_fpregset.fpc, &sa->fpc, sizeof(sa->fpc));
+	memcpy(&nt_fpregset.fprs, &sa->fprs, sizeof(sa->fprs));
+	/* Create ELF notes for the CPU */
+	ptr = nt_init(ptr, NT_PRSTATUS, &nt_prstatus, sizeof(nt_prstatus));
+	ptr = nt_init(ptr, NT_PRFPREG, &nt_fpregset, sizeof(nt_fpregset));
+	ptr = nt_init(ptr, NT_S390_TIMER, &sa->timer, sizeof(sa->timer));
+	ptr = nt_init(ptr, NT_S390_TODCMP, &sa->todcmp, sizeof(sa->todcmp));
+	ptr = nt_init(ptr, NT_S390_TODPREG, &sa->todpreg, sizeof(sa->todpreg));
+	ptr = nt_init(ptr, NT_S390_CTRS, &sa->ctrs, sizeof(sa->ctrs));
+	ptr = nt_init(ptr, NT_S390_PREFIX, &sa->prefix, sizeof(sa->prefix));
+	if (MACHINE_HAS_VX) {
+		ptr = nt_init(ptr, NT_S390_VXRS_HIGH,
+			      &sa->vxrs_high, sizeof(sa->vxrs_high));
+		ptr = nt_init(ptr, NT_S390_VXRS_LOW,
+			      &sa->vxrs_low, sizeof(sa->vxrs_low));
 	}
 	return ptr;
 }
@@ -416,8 +381,7 @@ static void *nt_prpsinfo(void *ptr)
 	memset(&prpsinfo, 0, sizeof(prpsinfo));
 	prpsinfo.pr_sname = 'R';
 	strcpy(prpsinfo.pr_fname, "vmlinux");
-	return nt_init(ptr, NT_PRPSINFO, &prpsinfo, sizeof(prpsinfo),
-		       KEXEC_CORE_NOTE_NAME);
+	return nt_init(ptr, NT_PRPSINFO, &prpsinfo, sizeof(prpsinfo));
 }
 
 /*
@@ -429,17 +393,18 @@ static void *get_vmcoreinfo_old(unsigned long *size)
 	Elf64_Nhdr note;
 	void *addr;
 
-	if (copy_from_oldmem(&addr, &S390_lowcore.vmcore_info, sizeof(addr)))
+	if (copy_oldmem_kernel(&addr, &S390_lowcore.vmcore_info, sizeof(addr)))
 		return NULL;
 	memset(nt_name, 0, sizeof(nt_name));
-	if (copy_from_oldmem(&note, addr, sizeof(note)))
+	if (copy_oldmem_kernel(&note, addr, sizeof(note)))
 		return NULL;
-	if (copy_from_oldmem(nt_name, addr + sizeof(note), sizeof(nt_name) - 1))
+	if (copy_oldmem_kernel(nt_name, addr + sizeof(note),
+			       sizeof(nt_name) - 1))
 		return NULL;
 	if (strcmp(nt_name, "VMCOREINFO") != 0)
 		return NULL;
 	vmcoreinfo = kzalloc_panic(note.n_descsz);
-	if (copy_from_oldmem(vmcoreinfo, addr + 24, note.n_descsz))
+	if (copy_oldmem_kernel(vmcoreinfo, addr + 24, note.n_descsz))
 		return NULL;
 	*size = note.n_descsz;
 	return vmcoreinfo;
@@ -458,7 +423,7 @@ static void *nt_vmcoreinfo(void *ptr)
 		vmcoreinfo = get_vmcoreinfo_old(&size);
 	if (!vmcoreinfo)
 		return ptr;
-	return nt_init(ptr, 0, vmcoreinfo, size, "VMCOREINFO");
+	return nt_init_name(ptr, 0, vmcoreinfo, size, "VMCOREINFO");
 }
 
 /*
@@ -487,13 +452,12 @@ static void *ehdr_init(Elf64_Ehdr *ehdr, int mem_chunk_cnt)
  */
 static int get_cpu_cnt(void)
 {
-	int i, cpus = 0;
+	struct save_area *sa;
+	int cpus = 0;
 
-	for (i = 0; i < dump_save_areas.count; i++) {
-		if (dump_save_areas.areas[i]->sa.pref_reg == 0)
-			continue;
-		cpus++;
-	}
+	list_for_each_entry(sa, &dump_save_areas, list)
+		if (sa->prefix != 0)
+			cpus++;
 	return cpus;
 }
 
@@ -538,18 +502,16 @@ static void loads_init(Elf64_Phdr *phdr, u64 loads_offset)
  */
 static void *notes_init(Elf64_Phdr *phdr, void *ptr, u64 notes_offset)
 {
-	struct save_area_ext *sa_ext;
+	struct save_area *sa;
 	void *ptr_start = ptr;
-	int i;
+	int cpu;
 
 	ptr = nt_prpsinfo(ptr);
 
-	for (i = 0; i < dump_save_areas.count; i++) {
-		sa_ext = dump_save_areas.areas[i];
-		if (sa_ext->sa.pref_reg == 0)
-			continue;
-		ptr = fill_cpu_elf_notes(ptr, &sa_ext->sa, sa_ext->vx_regs);
-	}
+	cpu = 1;
+	list_for_each_entry(sa, &dump_save_areas, list)
+		if (sa->prefix != 0)
+			ptr = fill_cpu_elf_notes(ptr, cpu++, sa);
 	ptr = nt_vmcoreinfo(ptr);
 	memset(phdr, 0, sizeof(*phdr));
 	phdr->p_type = PT_NOTE;
@@ -573,9 +535,6 @@ int elfcorehdr_alloc(unsigned long long *addr, unsigned long long *size)
 	/* If we are not in kdump or zfcpdump mode return */
 	if (!OLDMEM_BASE && ipl_info.type != IPL_TYPE_FCP_DUMP)
 		return 0;
-	/* If elfcorehdr= has been passed via cmdline, we use that one */
-	if (elfcorehdr_addr != ELFCORE_ADDR_MAX)
-		return 0;
 	/* If we cannot get HSA size for zfcpdump return error */
 	if (ipl_info.type == IPL_TYPE_FCP_DUMP && !sclp.hsa_size)
 		return -ENODEV;
@@ -606,7 +565,6 @@ int elfcorehdr_alloc(unsigned long long *addr, unsigned long long *size)
 	hdr_off = PTR_DIFF(ptr, hdr);
 	loads_init(phdr_loads, hdr_off);
 	*addr = (unsigned long long) hdr;
-	elfcorehdr_newmem = hdr;
 	*size = (unsigned long long) hdr_off;
 	BUG_ON(elfcorehdr_size > alloc_size);
 	return 0;
@@ -617,8 +575,6 @@ int elfcorehdr_alloc(unsigned long long *addr, unsigned long long *size)
  */
 void elfcorehdr_free(unsigned long long addr)
 {
-	if (!elfcorehdr_newmem)
-		return;
 	kfree((void *)(unsigned long)addr);
 }
 
@@ -629,7 +585,6 @@ ssize_t elfcorehdr_read(char *buf, size_t count, u64 *ppos)
 {
 	void *src = (void *)(unsigned long)*ppos;
 
-	src = elfcorehdr_newmem ? src : src - OLDMEM_BASE;
 	memcpy(buf, src, count);
 	*ppos += count;
 	return count;
@@ -641,15 +596,8 @@ ssize_t elfcorehdr_read(char *buf, size_t count, u64 *ppos)
 ssize_t elfcorehdr_read_notes(char *buf, size_t count, u64 *ppos)
 {
 	void *src = (void *)(unsigned long)*ppos;
-	int rc;
 
-	if (elfcorehdr_newmem) {
-		memcpy(buf, src, count);
-	} else {
-		rc = copy_from_oldmem(buf, src, count);
-		if (rc)
-			return rc;
-	}
+	memcpy(buf, src, count);
 	*ppos += count;
 	return count;
 }
diff --git a/arch/s390/kernel/dis.c b/arch/s390/kernel/dis.c
index 6e72961..62973ef 100644
--- a/arch/s390/kernel/dis.c
+++ b/arch/s390/kernel/dis.c
@@ -2022,7 +2022,7 @@ void show_code(struct pt_regs *regs)
 			*ptr++ = '\t';
 		ptr += print_insn(ptr, code + start, addr);
 		start += opsize;
-		printk(buffer);
+		printk("%s", buffer);
 		ptr = buffer;
 		ptr += sprintf(ptr, "\n          ");
 		hops++;
@@ -2049,7 +2049,7 @@ void print_fn_code(unsigned char *code, unsigned long len)
 		ptr += print_insn(ptr, code, (unsigned long) code);
 		*ptr++ = '\n';
 		*ptr++ = 0;
-		printk(buffer);
+		printk("%s", buffer);
 		code += opsize;
 		len -= opsize;
 	}
diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c
index 3c31609..20a5caf 100644
--- a/arch/s390/kernel/early.c
+++ b/arch/s390/kernel/early.c
@@ -335,6 +335,14 @@ static __init void detect_machine_facilities(void)
 	}
 }
 
+static inline void save_vector_registers(void)
+{
+#ifdef CONFIG_CRASH_DUMP
+	if (test_facility(129))
+		save_vx_regs(boot_cpu_vector_save_area);
+#endif
+}
+
 static int __init disable_vector_extension(char *str)
 {
 	S390_lowcore.machine_flags &= ~MACHINE_FLAG_VX;
@@ -451,6 +459,7 @@ void __init startup_init(void)
 	detect_diag9c();
 	detect_diag44();
 	detect_machine_facilities();
+	save_vector_registers();
 	setup_topology();
 	sclp_early_detect();
 	lockdep_on();
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index 857b652..cd5a191 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -764,6 +764,7 @@ ENTRY(psw_idle)
 	.insn	rsy,0xeb0000000017,%r1,5,__SF_EMPTY+16(%r15)
 .Lpsw_idle_stcctm:
 #endif
+	oi	__LC_CPU_FLAGS+7,_CIF_ENABLED_WAIT
 	STCK	__CLOCK_IDLE_ENTER(%r2)
 	stpt	__TIMER_IDLE_ENTER(%r2)
 .Lpsw_idle_lpsw:
@@ -1146,6 +1147,7 @@ cleanup_critical:
 	.quad	.Lio_done - 4
 
 .Lcleanup_idle:
+	ni	__LC_CPU_FLAGS+7,255-_CIF_ENABLED_WAIT
 	# copy interrupt clock & cpu timer
 	mvc	__CLOCK_IDLE_EXIT(8,%r2),__LC_INT_CLOCK
 	mvc	__TIMER_IDLE_EXIT(8,%r2),__LC_ASYNC_ENTER_TIMER
diff --git a/arch/s390/kernel/head.S b/arch/s390/kernel/head.S
index 301ee9c..fcaefb0 100644
--- a/arch/s390/kernel/head.S
+++ b/arch/s390/kernel/head.S
@@ -25,6 +25,7 @@
 #include <linux/linkage.h>
 #include <asm/asm-offsets.h>
 #include <asm/thread_info.h>
+#include <asm/facility.h>
 #include <asm/page.h>
 #include <asm/ptrace.h>
 
@@ -300,27 +301,27 @@ ENTRY(startup_kdump)
 	xc	0x200(256),0x200	# partially clear lowcore
 	xc	0x300(256),0x300
 	xc	0xe00(256),0xe00
+	xc	0xf00(256),0xf00
 	lctlg	%c0,%c15,0x200(%r0)	# initialize control registers
 	stck	__LC_LAST_UPDATE_CLOCK
 	spt	6f-.LPG0(%r13)
 	mvc	__LC_LAST_UPDATE_TIMER(8),6f-.LPG0(%r13)
-	xc	__LC_STFL_FAC_LIST(8),__LC_STFL_FAC_LIST
-	# check capabilities against MARCH_{G5,Z900,Z990,Z9_109,Z10}
-	.insn	s,0xb2b10000,0		# store facilities @ __LC_STFL_FAC_LIST
-	tm	__LC_STFL_FAC_LIST,0x01	# stfle available ?
+	stfl	0(%r0)			# store facilities @ __LC_STFL_FAC_LIST
+	mvc	__LC_STFLE_FAC_LIST(4),__LC_STFL_FAC_LIST
+	tm	__LC_STFLE_FAC_LIST,0x01	# stfle available ?
 	jz	0f
-	la	%r0,1
-	.insn	s,0xb2b00000,__LC_STFL_FAC_LIST	# store facility list extended
+	lghi	%r0,FACILITIES_ALS_DWORDS-1
+	.insn	s,0xb2b00000,__LC_STFLE_FAC_LIST # store facility list extended
 	# verify if all required facilities are supported by the machine
-0:	la	%r1,__LC_STFL_FAC_LIST
+0:	la	%r1,__LC_STFLE_FAC_LIST
 	la	%r2,3f+8-.LPG0(%r13)
-	l	%r3,0(%r2)
-1:	l	%r0,0(%r1)
-	n	%r0,4(%r2)
-	cl	%r0,4(%r2)
+	lhi	%r3,FACILITIES_ALS_DWORDS
+1:	lg	%r0,0(%r1)
+	ng	%r0,0(%r2)
+	clg	%r0,0(%r2)
 	jne	2f
-	la	%r1,4(%r1)
-	la	%r2,4(%r2)
+	la	%r1,8(%r1)
+	la	%r2,8(%r2)
 	ahi	%r3,-1
 	jnz	1b
 	j	4f
@@ -340,24 +341,10 @@ ENTRY(startup_kdump)
 3:	.long	0x000a0000,0x8badcccc
 
 # List of facilities that are required. If not all facilities are present
-# the kernel will crash. Format is number of facility words with bits set,
-# followed by the facility words.
+# the kernel will crash.
+
+	.quad FACILITIES_ALS
 
-#if defined(CONFIG_MARCH_Z13)
-	.long 2, 0xc100eff2, 0xf46cc800
-#elif defined(CONFIG_MARCH_ZEC12)
-	.long 2, 0xc100eff2, 0xf46cc800
-#elif defined(CONFIG_MARCH_Z196)
-	.long 2, 0xc100eff2, 0xf46c0000
-#elif defined(CONFIG_MARCH_Z10)
-	.long 2, 0xc100eff2, 0xf0680000
-#elif defined(CONFIG_MARCH_Z9_109)
-	.long 1, 0xc100efc2
-#elif defined(CONFIG_MARCH_Z990)
-	.long 1, 0xc0002000
-#elif defined(CONFIG_MARCH_Z900)
-	.long 1, 0xc0000000
-#endif
 4:
 	/* Continue with startup code in head64.S */
 	jg	startup_continue
diff --git a/arch/s390/kernel/head64.S b/arch/s390/kernel/head64.S
index 58b719f..c5febe8 100644
--- a/arch/s390/kernel/head64.S
+++ b/arch/s390/kernel/head64.S
@@ -16,7 +16,7 @@
 
 __HEAD
 ENTRY(startup_continue)
-	tm	__LC_STFL_FAC_LIST+6,0x80	# LPP available ?
+	tm	__LC_STFLE_FAC_LIST+6,0x80	# LPP available ?
 	jz	0f
 	xc	__LC_LPP+1(7,0),__LC_LPP+1	# clear lpp and current_pid
 	mvi	__LC_LPP,0x80			#   and set LPP_MAGIC
diff --git a/arch/s390/kernel/ipl.c b/arch/s390/kernel/ipl.c
index b1f0a90..0a5a6b6 100644
--- a/arch/s390/kernel/ipl.c
+++ b/arch/s390/kernel/ipl.c
@@ -2039,21 +2039,15 @@ static void do_reset_calls(void)
 		reset->fn();
 }
 
-u32 dump_prefix_page;
-
-void s390_reset_system(void (*fn_pre)(void),
-		       void (*fn_post)(void *), void *data)
+void s390_reset_system(void)
 {
-	struct _lowcore *lc;
+	struct lowcore *lc;
 
-	lc = (struct _lowcore *)(unsigned long) store_prefix();
+	lc = (struct lowcore *)(unsigned long) store_prefix();
 
 	/* Stack for interrupt/machine check handler */
 	lc->panic_stack = S390_lowcore.panic_stack;
 
-	/* Save prefix page address for dump case */
-	dump_prefix_page = (u32)(unsigned long) lc;
-
 	/* Disable prefixing */
 	set_prefix(0);
 
@@ -2077,14 +2071,5 @@ void s390_reset_system(void (*fn_pre)(void),
 	S390_lowcore.subchannel_id = 0;
 	S390_lowcore.subchannel_nr = 0;
 
-	/* Store status at absolute zero */
-	store_status();
-
-	/* Call function before reset */
-	if (fn_pre)
-		fn_pre();
 	do_reset_calls();
-	/* Call function after reset */
-	if (fn_post)
-		fn_post(data);
 }
diff --git a/arch/s390/kernel/machine_kexec.c b/arch/s390/kernel/machine_kexec.c
index fb0901e..2f1b721 100644
--- a/arch/s390/kernel/machine_kexec.c
+++ b/arch/s390/kernel/machine_kexec.c
@@ -35,46 +35,6 @@ extern const unsigned long long relocate_kernel_len;
 #ifdef CONFIG_CRASH_DUMP
 
 /*
- * Create ELF notes for one CPU
- */
-static void add_elf_notes(int cpu)
-{
-	struct save_area *sa = (void *) 4608 + store_prefix();
-	void *ptr;
-
-	memcpy((void *) (4608UL + sa->pref_reg), sa, sizeof(*sa));
-	ptr = (u64 *) per_cpu_ptr(crash_notes, cpu);
-	ptr = fill_cpu_elf_notes(ptr, sa, NULL);
-	memset(ptr, 0, sizeof(struct elf_note));
-}
-
-/*
- * Initialize CPU ELF notes
- */
-static void setup_regs(void)
-{
-	unsigned long sa = S390_lowcore.prefixreg_save_area + SAVE_AREA_BASE;
-	struct _lowcore *lc;
-	int cpu, this_cpu;
-
-	/* Get lowcore pointer from store status of this CPU (absolute zero) */
-	lc = (struct _lowcore *)(unsigned long)S390_lowcore.prefixreg_save_area;
-	this_cpu = smp_find_processor_id(stap());
-	add_elf_notes(this_cpu);
-	for_each_online_cpu(cpu) {
-		if (cpu == this_cpu)
-			continue;
-		if (smp_store_status(cpu))
-			continue;
-		add_elf_notes(cpu);
-	}
-	if (MACHINE_HAS_VX)
-		save_vx_regs_safe((void *) lc->vector_save_area_addr);
-	/* Copy dump CPU store status info to absolute zero */
-	memcpy((void *) SAVE_AREA_BASE, (void *) sa, sizeof(struct save_area));
-}
-
-/*
  * PM notifier callback for kdump
  */
 static int machine_kdump_pm_cb(struct notifier_block *nb, unsigned long action,
@@ -105,14 +65,66 @@ static int __init machine_kdump_pm_init(void)
 arch_initcall(machine_kdump_pm_init);
 
 /*
- * Start kdump: We expect here that a store status has been done on our CPU
+ * Reset the system, copy boot CPU registers to absolute zero,
+ * and jump to the kdump image
  */
 static void __do_machine_kdump(void *image)
 {
-	int (*start_kdump)(int) = (void *)((struct kimage *) image)->start;
+	int (*start_kdump)(int);
+	unsigned long prefix;
+
+	/* store_status() saved the prefix register to lowcore */
+	prefix = (unsigned long) S390_lowcore.prefixreg_save_area;
+
+	/* Now do the reset  */
+	s390_reset_system();
+
+	/*
+	 * Copy dump CPU store status info to absolute zero.
+	 * This need to be done *after* s390_reset_system set the
+	 * prefix register of this CPU to zero
+	 */
+	memcpy((void *) __LC_FPREGS_SAVE_AREA,
+	       (void *)(prefix + __LC_FPREGS_SAVE_AREA), 512);
 
 	__load_psw_mask(PSW_MASK_BASE | PSW_DEFAULT_KEY | PSW_MASK_EA | PSW_MASK_BA);
+	start_kdump = (void *)((struct kimage *) image)->start;
 	start_kdump(1);
+
+	/* Die if start_kdump returns */
+	disabled_wait((unsigned long) __builtin_return_address(0));
+}
+
+/*
+ * Start kdump: create a LGR log entry, store status of all CPUs and
+ * branch to __do_machine_kdump.
+ */
+static noinline void __machine_kdump(void *image)
+{
+	int this_cpu, cpu;
+
+	lgr_info_log();
+	/* Get status of the other CPUs */
+	this_cpu = smp_find_processor_id(stap());
+	for_each_online_cpu(cpu) {
+		if (cpu == this_cpu)
+			continue;
+		if (smp_store_status(cpu))
+			continue;
+	}
+	/* Store status of the boot CPU */
+	if (MACHINE_HAS_VX)
+		save_vx_regs((void *) &S390_lowcore.vector_save_area);
+	/*
+	 * To create a good backchain for this CPU in the dump store_status
+	 * is passed the address of a function. The address is saved into
+	 * the PSW save area of the boot CPU and the function is invoked as
+	 * a tail call of store_status. The backchain in the dump will look
+	 * like this:
+	 *   restart_int_handler ->  __machine_kexec -> __do_machine_kdump
+	 * The call to store_status() will not return.
+	 */
+	store_status(__do_machine_kdump, image);
 }
 #endif
 
@@ -235,10 +247,14 @@ static void __do_machine_kexec(void *data)
 	relocate_kernel_t data_mover;
 	struct kimage *image = data;
 
+	s390_reset_system();
 	data_mover = (relocate_kernel_t) page_to_phys(image->control_code_page);
 
 	/* Call the moving routine */
 	(*data_mover)(&image->head, image->start);
+
+	/* Die if kexec returns */
+	disabled_wait((unsigned long) __builtin_return_address(0));
 }
 
 /*
@@ -251,14 +267,10 @@ static void __machine_kexec(void *data)
 	tracing_off();
 	debug_locks_off();
 #ifdef CONFIG_CRASH_DUMP
-	if (((struct kimage *) data)->type == KEXEC_TYPE_CRASH) {
-
-		lgr_info_log();
-		s390_reset_system(setup_regs, __do_machine_kdump, data);
-	} else
+	if (((struct kimage *) data)->type == KEXEC_TYPE_CRASH)
+		__machine_kdump(data);
 #endif
-		s390_reset_system(NULL, __do_machine_kexec, data);
-	disabled_wait((unsigned long) __builtin_return_address(0));
+	__do_machine_kexec(data);
 }
 
 /*
diff --git a/arch/s390/kernel/os_info.c b/arch/s390/kernel/os_info.c
index d112fc6..87f05e4 100644
--- a/arch/s390/kernel/os_info.c
+++ b/arch/s390/kernel/os_info.c
@@ -89,7 +89,7 @@ static void os_info_old_alloc(int nr, int align)
 		goto fail;
 	}
 	buf_align = PTR_ALIGN(buf, align);
-	if (copy_from_oldmem(buf_align, (void *) addr, size)) {
+	if (copy_oldmem_kernel(buf_align, (void *) addr, size)) {
 		msg = "copy failed";
 		goto fail_free;
 	}
@@ -122,14 +122,15 @@ static void os_info_old_init(void)
 		return;
 	if (!OLDMEM_BASE)
 		goto fail;
-	if (copy_from_oldmem(&addr, &S390_lowcore.os_info, sizeof(addr)))
+	if (copy_oldmem_kernel(&addr, &S390_lowcore.os_info, sizeof(addr)))
 		goto fail;
 	if (addr == 0 || addr % PAGE_SIZE)
 		goto fail;
 	os_info_old = kzalloc(sizeof(*os_info_old), GFP_KERNEL);
 	if (!os_info_old)
 		goto fail;
-	if (copy_from_oldmem(os_info_old, (void *) addr, sizeof(*os_info_old)))
+	if (copy_oldmem_kernel(os_info_old, (void *) addr,
+			       sizeof(*os_info_old)))
 		goto fail_free;
 	if (os_info_old->magic != OS_INFO_MAGIC)
 		goto fail_free;
diff --git a/arch/s390/kernel/reipl.S b/arch/s390/kernel/reipl.S
index 52aab0b..89ea8c2 100644
--- a/arch/s390/kernel/reipl.S
+++ b/arch/s390/kernel/reipl.S
@@ -9,60 +9,66 @@
 #include <asm/sigp.h>
 
 #
-# store_status
+# Issue "store status" for the current CPU to its prefix page
+# and call passed function afterwards
 #
-# Prerequisites to run this function:
-# - Prefix register is set to zero
-# - Original prefix register is stored in "dump_prefix_page"
-# - Lowcore protection is off
+# r2 = Function to be called after store status
+# r3 = Parameter for function
 #
 ENTRY(store_status)
 	/* Save register one and load save area base */
 	stg	%r1,__LC_SAVE_AREA_RESTART
-	lghi	%r1,SAVE_AREA_BASE
 	/* General purpose registers */
-	stmg	%r0,%r15,__LC_GPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	lg	%r2,__LC_SAVE_AREA_RESTART
-	stg	%r2,__LC_GPREGS_SAVE_AREA-SAVE_AREA_BASE+8(%r1)
+	lghi	%r1,__LC_GPREGS_SAVE_AREA
+	stmg	%r0,%r15,0(%r1)
+	mvc	8(8,%r1),__LC_SAVE_AREA_RESTART
 	/* Control registers */
-	stctg	%c0,%c15,__LC_CREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
+	lghi	%r1,__LC_CREGS_SAVE_AREA
+	stctg	%c0,%c15,0(%r1)
 	/* Access registers */
-	stam	%a0,%a15,__LC_AREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
+	lghi	%r1,__LC_AREGS_SAVE_AREA
+	stam	%a0,%a15,0(%r1)
 	/* Floating point registers */
-	std	%f0, 0x00 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f1, 0x08 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f2, 0x10 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f3, 0x18 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f4, 0x20 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f5, 0x28 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f6, 0x30 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f7, 0x38 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f8, 0x40 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f9, 0x48 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f10,0x50 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f11,0x58 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f12,0x60 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f13,0x68 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f14,0x70 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	std	%f15,0x78 + __LC_FPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
+	lghi	%r1,__LC_FPREGS_SAVE_AREA
+	std	%f0, 0x00(%r1)
+	std	%f1, 0x08(%r1)
+	std	%f2, 0x10(%r1)
+	std	%f3, 0x18(%r1)
+	std	%f4, 0x20(%r1)
+	std	%f5, 0x28(%r1)
+	std	%f6, 0x30(%r1)
+	std	%f7, 0x38(%r1)
+	std	%f8, 0x40(%r1)
+	std	%f9, 0x48(%r1)
+	std	%f10,0x50(%r1)
+	std	%f11,0x58(%r1)
+	std	%f12,0x60(%r1)
+	std	%f13,0x68(%r1)
+	std	%f14,0x70(%r1)
+	std	%f15,0x78(%r1)
 	/* Floating point control register */
-	stfpc	__LC_FP_CREG_SAVE_AREA-SAVE_AREA_BASE(%r1)
+	lghi	%r1,__LC_FP_CREG_SAVE_AREA
+	stfpc	0(%r1)
 	/* CPU timer */
-	stpt	__LC_CPU_TIMER_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	/* Saved prefix register */
-	larl	%r2,dump_prefix_page
-	mvc	__LC_PREFIX_SAVE_AREA-SAVE_AREA_BASE(4,%r1),0(%r2)
+	lghi	%r1,__LC_CPU_TIMER_SAVE_AREA
+	stpt	0(%r1)
+	/* Store prefix register */
+	lghi	%r1,__LC_PREFIX_SAVE_AREA
+	stpx	0(%r1)
 	/* Clock comparator - seven bytes */
-	larl	%r2,.Lclkcmp
-	stckc	0(%r2)
-	mvc	__LC_CLOCK_COMP_SAVE_AREA-SAVE_AREA_BASE + 1(7,%r1),1(%r2)
+	lghi	%r1,__LC_CLOCK_COMP_SAVE_AREA
+	larl	%r4,.Lclkcmp
+	stckc	0(%r4)
+	mvc	1(7,%r1),1(%r4)
 	/* Program status word */
-	epsw	%r2,%r3
-	st	%r2,__LC_PSW_SAVE_AREA-SAVE_AREA_BASE + 0(%r1)
-	st	%r3,__LC_PSW_SAVE_AREA-SAVE_AREA_BASE + 4(%r1)
-	larl	%r2,store_status
-	stg	%r2,__LC_PSW_SAVE_AREA-SAVE_AREA_BASE + 8(%r1)
-	br	%r14
+	lghi	%r1,__LC_PSW_SAVE_AREA
+	epsw	%r4,%r5
+	st	%r4,0(%r1)
+	st	%r5,4(%r1)
+	stg	%r2,8(%r1)
+	lgr	%r1,%r2
+	lgr	%r2,%r3
+	br	%r1
 
 	.section .bss
 	.align	8
@@ -77,9 +83,11 @@ ENTRY(store_status)
 ENTRY(do_reipl_asm)
 		basr	%r13,0
 .Lpg0:		lpswe	.Lnewpsw-.Lpg0(%r13)
-.Lpg1:		brasl	%r14,store_status
+.Lpg1:		lgr	%r3,%r2
+		larl	%r2,.Lstatus
+		brasl	%r14,store_status
 
-		lctlg	%c6,%c6,.Lall-.Lpg0(%r13)
+.Lstatus:	lctlg	%c6,%c6,.Lall-.Lpg0(%r13)
 		lgr	%r1,%r2
 		mvc	__LC_PGM_NEW_PSW(16),.Lpcnew-.Lpg0(%r13)
 		stsch	.Lschib-.Lpg0(%r13)
diff --git a/arch/s390/kernel/sclp.c b/arch/s390/kernel/sclp.c
index 9fe7781..d88db40 100644
--- a/arch/s390/kernel/sclp.c
+++ b/arch/s390/kernel/sclp.c
@@ -9,7 +9,11 @@
 #include <asm/processor.h>
 #include <asm/sclp.h>
 
+#define EVTYP_VT220MSG_MASK	0x00000040
+#define EVTYP_MSG_MASK		0x40000000
+
 static char _sclp_work_area[4096] __aligned(PAGE_SIZE);
+static bool have_vt220, have_linemode;
 
 static void _sclp_wait_int(void)
 {
@@ -68,7 +72,7 @@ static int _sclp_setup(int disable)
 		0x00, 0x1c,
 		0x00, 0x00, 0x00, 0x00,	0x00, 0x00, 0x00, 0x00,
 		0x00, 0x04,
-		0x80, 0x00, 0x00, 0x00,	0x40, 0x00, 0x00, 0x00,
+		0x80, 0x00, 0x00, 0x00,	0x40, 0x00, 0x00, 0x40,
 		0x00, 0x00, 0x00, 0x00,	0x00, 0x00, 0x00, 0x00
 	};
 	unsigned int *masks;
@@ -82,13 +86,13 @@ static int _sclp_setup(int disable)
 	rc = _sclp_servc(0x00780005, _sclp_work_area);
 	if (rc)
 		return rc;
-	if ((masks[0] & masks[3]) != masks[0] ||
-	    (masks[1] & masks[2]) != masks[1])
-		return -EIO;
+	have_vt220 = masks[2] & EVTYP_VT220MSG_MASK;
+	have_linemode = masks[2] & EVTYP_MSG_MASK;
 	return 0;
 }
 
-static int _sclp_print(const char *str)
+/* Output multi-line text using SCLP Message interface. */
+static void _sclp_print_lm(const char *str)
 {
 	static unsigned char write_head[] = {
 		/* sccb header */
@@ -143,18 +147,49 @@ static int _sclp_print(const char *str)
 	} while (ch != 0);
 
 	/* SCLP write data */
-	return _sclp_servc(0x00760005, _sclp_work_area);
+	_sclp_servc(0x00760005, _sclp_work_area);
 }
 
-int _sclp_print_early(const char *str)
+/* Output multi-line text (plus a newline) using SCLP VT220
+ * interface.
+ */
+static void _sclp_print_vt220(const char *str)
 {
-	int rc;
+	static unsigned char const write_head[] = {
+		/* sccb header */
+		0x00, 0x0e,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		/* evbuf header */
+		0x00, 0x06,
+		0x1a, 0x00, 0x00, 0x00,
+	};
+	size_t len = strlen(str);
 
-	rc = _sclp_setup(0);
-	if (rc)
-		return rc;
-	rc = _sclp_print(str);
-	if (rc)
-		return rc;
-	return _sclp_setup(1);
+	if (sizeof(write_head) + len >= sizeof(_sclp_work_area))
+		len = sizeof(_sclp_work_area) - sizeof(write_head) - 1;
+
+	memcpy(_sclp_work_area, write_head, sizeof(write_head));
+	memcpy(_sclp_work_area + sizeof(write_head), str, len);
+	_sclp_work_area[sizeof(write_head) + len] = '\n';
+
+	/* Update length fields in evbuf and sccb headers */
+	*(unsigned short *)(_sclp_work_area + 8) += len + 1;
+	*(unsigned short *)(_sclp_work_area + 0) += len + 1;
+
+	/* SCLP write data */
+	(void)_sclp_servc(0x00760005, _sclp_work_area);
+}
+
+/* Output one or more lines of text on the SCLP console (VT220 and /
+ * or line-mode). All lines get terminated; no need for a trailing LF.
+ */
+void _sclp_print_early(const char *str)
+{
+	if (_sclp_setup(0) != 0)
+		return;
+	if (have_linemode)
+		_sclp_print_lm(str);
+	if (have_vt220)
+		_sclp_print_vt220(str);
+	_sclp_setup(1);
 }
diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c
index dc83ae6..c6878fb 100644
--- a/arch/s390/kernel/setup.c
+++ b/arch/s390/kernel/setup.c
@@ -99,7 +99,7 @@ unsigned long MODULES_VADDR;
 unsigned long MODULES_END;
 
 /* An array with a pointer to the lowcore of every CPU. */
-struct _lowcore *lowcore_ptr[NR_CPUS];
+struct lowcore *lowcore_ptr[NR_CPUS];
 EXPORT_SYMBOL(lowcore_ptr);
 
 /*
@@ -293,12 +293,12 @@ void *restart_stack __attribute__((__section__(".data")));
 
 static void __init setup_lowcore(void)
 {
-	struct _lowcore *lc;
+	struct lowcore *lc;
 
 	/*
 	 * Setup lowcore for boot cpu
 	 */
-	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
+	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
 	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 	lc->restart_psw.mask = PSW_KERNEL_BITS;
 	lc->restart_psw.addr =
@@ -663,15 +663,6 @@ static void __init reserve_kernel(void)
 #endif
 }
 
-static void __init reserve_elfcorehdr(void)
-{
-#ifdef CONFIG_CRASH_DUMP
-	if (is_kdump_kernel())
-		memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
-				 PAGE_ALIGN(elfcorehdr_size));
-#endif
-}
-
 static void __init setup_memory(void)
 {
 	struct memblock_region *reg;
@@ -850,6 +841,11 @@ void __init setup_arch(char **cmdline_p)
 	init_mm.brk = (unsigned long) &_end;
 
 	parse_early_param();
+#ifdef CONFIG_CRASH_DUMP
+	/* Deactivate elfcorehdr= kernel parameter */
+	elfcorehdr_addr = ELFCORE_ADDR_MAX;
+#endif
+
 	os_info_init();
 	setup_ipl();
 
@@ -858,7 +854,6 @@ void __init setup_arch(char **cmdline_p)
 	reserve_oldmem();
 	reserve_kernel();
 	reserve_initrd();
-	reserve_elfcorehdr();
 	memblock_allow_resize();
 
 	/* Get information about *all* installed memory */
@@ -879,11 +874,13 @@ void __init setup_arch(char **cmdline_p)
 
 	check_initrd();
 	reserve_crashkernel();
+#ifdef CONFIG_CRASH_DUMP
 	/*
 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
 	 * Therefore CPU and device initialization should be done afterwards.
 	 */
 	smp_save_dump_cpus();
+#endif
 
 	setup_resources();
 	setup_vmcoreinfo();
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index 9062df5..a13468b 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -64,8 +64,9 @@ enum {
 static DEFINE_PER_CPU(struct cpu *, cpu_device);
 
 struct pcpu {
-	struct _lowcore *lowcore;	/* lowcore page(s) for the cpu */
+	struct lowcore *lowcore;	/* lowcore page(s) for the cpu */
 	unsigned long ec_mask;		/* bit mask for ec_xxx functions */
+	unsigned long ec_clk;		/* sigp timestamp for ec_xxx */
 	signed char state;		/* physical cpu state */
 	signed char polarization;	/* physical polarization */
 	u16 address;			/* physical cpu address */
@@ -80,6 +81,10 @@ EXPORT_SYMBOL(smp_cpu_mt_shift);
 unsigned int smp_cpu_mtid;
 EXPORT_SYMBOL(smp_cpu_mtid);
 
+#ifdef CONFIG_CRASH_DUMP
+__vector128 __initdata boot_cpu_vector_save_area[__NUM_VXRS];
+#endif
+
 static unsigned int smp_max_threads __initdata = -1U;
 
 static int __init early_nosmt(char *s)
@@ -105,8 +110,7 @@ DEFINE_MUTEX(smp_cpu_state_mutex);
 /*
  * Signal processor helper functions.
  */
-static inline int __pcpu_sigp_relax(u16 addr, u8 order, unsigned long parm,
-				    u32 *status)
+static inline int __pcpu_sigp_relax(u16 addr, u8 order, unsigned long parm)
 {
 	int cc;
 
@@ -171,6 +175,7 @@ static void pcpu_ec_call(struct pcpu *pcpu, int ec_bit)
 	if (test_and_set_bit(ec_bit, &pcpu->ec_mask))
 		return;
 	order = pcpu_running(pcpu) ? SIGP_EXTERNAL_CALL : SIGP_EMERGENCY_SIGNAL;
+	pcpu->ec_clk = get_tod_clock_fast();
 	pcpu_sigp_retry(pcpu, order, 0);
 }
 
@@ -180,10 +185,10 @@ static void pcpu_ec_call(struct pcpu *pcpu, int ec_bit)
 static int pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu)
 {
 	unsigned long async_stack, panic_stack;
-	struct _lowcore *lc;
+	struct lowcore *lc;
 
 	if (pcpu != &pcpu_devices[0]) {
-		pcpu->lowcore =	(struct _lowcore *)
+		pcpu->lowcore =	(struct lowcore *)
 			__get_free_pages(GFP_KERNEL | GFP_DMA, LC_ORDER);
 		async_stack = __get_free_pages(GFP_KERNEL, ASYNC_ORDER);
 		panic_stack = __get_free_page(GFP_KERNEL);
@@ -235,7 +240,7 @@ static void pcpu_free_lowcore(struct pcpu *pcpu)
 
 static void pcpu_prepare_secondary(struct pcpu *pcpu, int cpu)
 {
-	struct _lowcore *lc = pcpu->lowcore;
+	struct lowcore *lc = pcpu->lowcore;
 
 	if (MACHINE_HAS_TLB_LC)
 		cpumask_set_cpu(cpu, &init_mm.context.cpu_attach_mask);
@@ -255,7 +260,7 @@ static void pcpu_prepare_secondary(struct pcpu *pcpu, int cpu)
 
 static void pcpu_attach_task(struct pcpu *pcpu, struct task_struct *tsk)
 {
-	struct _lowcore *lc = pcpu->lowcore;
+	struct lowcore *lc = pcpu->lowcore;
 	struct thread_info *ti = task_thread_info(tsk);
 
 	lc->kernel_stack = (unsigned long) task_stack_page(tsk)
@@ -271,7 +276,7 @@ static void pcpu_attach_task(struct pcpu *pcpu, struct task_struct *tsk)
 
 static void pcpu_start_fn(struct pcpu *pcpu, void (*func)(void *), void *data)
 {
-	struct _lowcore *lc = pcpu->lowcore;
+	struct lowcore *lc = pcpu->lowcore;
 
 	lc->restart_stack = lc->kernel_stack;
 	lc->restart_fn = (unsigned long) func;
@@ -286,7 +291,7 @@ static void pcpu_start_fn(struct pcpu *pcpu, void (*func)(void *), void *data)
 static void pcpu_delegate(struct pcpu *pcpu, void (*func)(void *),
 			  void *data, unsigned long stack)
 {
-	struct _lowcore *lc = lowcore_ptr[pcpu - pcpu_devices];
+	struct lowcore *lc = lowcore_ptr[pcpu - pcpu_devices];
 	unsigned long source_cpu = stap();
 
 	__load_psw_mask(PSW_KERNEL_BITS);
@@ -538,53 +543,24 @@ EXPORT_SYMBOL(smp_ctl_clear_bit);
 
 #ifdef CONFIG_CRASH_DUMP
 
-static void __init __smp_store_cpu_state(struct save_area_ext *sa_ext,
-					 u16 address, int is_boot_cpu)
-{
-	void *lc = (void *)(unsigned long) store_prefix();
-	unsigned long vx_sa;
-
-	if (is_boot_cpu) {
-		/* Copy the registers of the boot CPU. */
-		copy_oldmem_page(1, (void *) &sa_ext->sa, sizeof(sa_ext->sa),
-				 SAVE_AREA_BASE - PAGE_SIZE, 0);
-		if (MACHINE_HAS_VX)
-			save_vx_regs_safe(sa_ext->vx_regs);
-		return;
-	}
-	/* Get the registers of a non-boot cpu. */
-	__pcpu_sigp_relax(address, SIGP_STOP_AND_STORE_STATUS, 0, NULL);
-	memcpy_real(&sa_ext->sa, lc + SAVE_AREA_BASE, sizeof(sa_ext->sa));
-	if (!MACHINE_HAS_VX)
-		return;
-	/* Get the VX registers */
-	vx_sa = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
-	if (!vx_sa)
-		panic("could not allocate memory for VX save area\n");
-	__pcpu_sigp_relax(address, SIGP_STORE_ADDITIONAL_STATUS, vx_sa, NULL);
-	memcpy(sa_ext->vx_regs, (void *) vx_sa, sizeof(sa_ext->vx_regs));
-	memblock_free(vx_sa, PAGE_SIZE);
-}
-
 int smp_store_status(int cpu)
 {
-	unsigned long vx_sa;
-	struct pcpu *pcpu;
+	struct pcpu *pcpu = pcpu_devices + cpu;
+	unsigned long pa;
 
-	pcpu = pcpu_devices + cpu;
-	if (__pcpu_sigp_relax(pcpu->address, SIGP_STOP_AND_STORE_STATUS,
-			      0, NULL) != SIGP_CC_ORDER_CODE_ACCEPTED)
+	pa = __pa(&pcpu->lowcore->floating_pt_save_area);
+	if (__pcpu_sigp_relax(pcpu->address, SIGP_STORE_STATUS_AT_ADDRESS,
+			      pa) != SIGP_CC_ORDER_CODE_ACCEPTED)
 		return -EIO;
 	if (!MACHINE_HAS_VX)
 		return 0;
-	vx_sa = __pa(pcpu->lowcore->vector_save_area_addr);
-	__pcpu_sigp_relax(pcpu->address, SIGP_STORE_ADDITIONAL_STATUS,
-			  vx_sa, NULL);
+	pa = __pa(pcpu->lowcore->vector_save_area_addr);
+	if (__pcpu_sigp_relax(pcpu->address, SIGP_STORE_ADDITIONAL_STATUS,
+			      pa) != SIGP_CC_ORDER_CODE_ACCEPTED)
+		return -EIO;
 	return 0;
 }
 
-#endif /* CONFIG_CRASH_DUMP */
-
 /*
  * Collect CPU state of the previous, crashed system.
  * There are four cases:
@@ -593,7 +569,7 @@ int smp_store_status(int cpu)
  *    The state for all CPUs except the boot CPU needs to be collected
  *    with sigp stop-and-store-status. The boot CPU state is located in
  *    the absolute lowcore of the memory stored in the HSA. The zcore code
- *    will allocate the save area and copy the boot CPU state from the HSA.
+ *    will copy the boot CPU state from the HSA.
  * 2) stand-alone kdump for SCSI (zfcp dump with swapped memory)
  *    condition: OLDMEM_BASE != NULL && ipl_info.type == IPL_TYPE_FCP_DUMP
  *    The state for all CPUs except the boot CPU needs to be collected
@@ -608,55 +584,78 @@ int smp_store_status(int cpu)
  *    stored the registers of the boot CPU in the memory of the old system.
  * 4) kdump and the old kernel stored the CPU state
  *    condition: OLDMEM_BASE != NULL && is_kdump_kernel()
- *    The state of all CPUs is stored in ELF sections in the memory of the
- *    old system. The ELF sections are picked up by the crash_dump code
- *    via elfcorehdr_addr.
+ *    This case does not exist for s390 anymore, setup_arch explicitly
+ *    deactivates the elfcorehdr= kernel parameter
  */
+static __init void smp_save_cpu_vxrs(struct save_area *sa, u16 addr,
+				     bool is_boot_cpu, unsigned long page)
+{
+	__vector128 *vxrs = (__vector128 *) page;
+
+	if (is_boot_cpu)
+		vxrs = boot_cpu_vector_save_area;
+	else
+		__pcpu_sigp_relax(addr, SIGP_STORE_ADDITIONAL_STATUS, page);
+	save_area_add_vxrs(sa, vxrs);
+}
+
+static __init void smp_save_cpu_regs(struct save_area *sa, u16 addr,
+				     bool is_boot_cpu, unsigned long page)
+{
+	void *regs = (void *) page;
+
+	if (is_boot_cpu)
+		copy_oldmem_kernel(regs, (void *) __LC_FPREGS_SAVE_AREA, 512);
+	else
+		__pcpu_sigp_relax(addr, SIGP_STORE_STATUS_AT_ADDRESS, page);
+	save_area_add_regs(sa, regs);
+}
+
 void __init smp_save_dump_cpus(void)
 {
-#ifdef CONFIG_CRASH_DUMP
-	int addr, cpu, boot_cpu_addr, max_cpu_addr;
-	struct save_area_ext *sa_ext;
+	int addr, boot_cpu_addr, max_cpu_addr;
+	struct save_area *sa;
+	unsigned long page;
 	bool is_boot_cpu;
 
-	if (is_kdump_kernel())
-		/* Previous system stored the CPU states. Nothing to do. */
-		return;
 	if (!(OLDMEM_BASE || ipl_info.type == IPL_TYPE_FCP_DUMP))
 		/* No previous system present, normal boot. */
 		return;
+	/* Allocate a page as dumping area for the store status sigps */
+	page = memblock_alloc_base(PAGE_SIZE, PAGE_SIZE, 1UL << 31);
+	if (!page)
+		panic("could not allocate memory for save area\n");
 	/* Set multi-threading state to the previous system. */
 	pcpu_set_smt(sclp.mtid_prev);
-	max_cpu_addr = SCLP_MAX_CORES << sclp.mtid_prev;
-	for (cpu = 0, addr = 0; addr <= max_cpu_addr; addr++) {
-		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0, NULL) ==
-		    SIGP_CC_NOT_OPERATIONAL)
-			continue;
-		cpu += 1;
-	}
-	dump_save_areas.areas = (void *)memblock_alloc(sizeof(void *) * cpu, 8);
-	dump_save_areas.count = cpu;
 	boot_cpu_addr = stap();
-	for (cpu = 0, addr = 0; addr <= max_cpu_addr; addr++) {
-		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0, NULL) ==
+	max_cpu_addr = SCLP_MAX_CORES << sclp.mtid_prev;
+	for (addr = 0; addr <= max_cpu_addr; addr++) {
+		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0) ==
 		    SIGP_CC_NOT_OPERATIONAL)
 			continue;
-		sa_ext = (void *) memblock_alloc(sizeof(*sa_ext), 8);
-		dump_save_areas.areas[cpu] = sa_ext;
-		if (!sa_ext)
-			panic("could not allocate memory for save area\n");
 		is_boot_cpu = (addr == boot_cpu_addr);
-		cpu += 1;
-		if (is_boot_cpu && !OLDMEM_BASE)
-			/* Skip boot CPU for standard zfcp dump. */
-			continue;
-		/* Get state for this CPU. */
-		__smp_store_cpu_state(sa_ext, addr, is_boot_cpu);
+		/* Allocate save area */
+		sa = save_area_alloc(is_boot_cpu);
+		if (!sa)
+			panic("could not allocate memory for save area\n");
+		if (MACHINE_HAS_VX)
+			/* Get the vector registers */
+			smp_save_cpu_vxrs(sa, addr, is_boot_cpu, page);
+		/*
+		 * For a zfcp dump OLDMEM_BASE == NULL and the registers
+		 * of the boot CPU are stored in the HSA. To retrieve
+		 * these registers an SCLP request is required which is
+		 * done by drivers/s390/char/zcore.c:init_cpu_info()
+		 */
+		if (!is_boot_cpu || OLDMEM_BASE)
+			/* Get the CPU registers */
+			smp_save_cpu_regs(sa, addr, is_boot_cpu, page);
 	}
+	memblock_free(page, PAGE_SIZE);
 	diag308_reset();
 	pcpu_set_smt(0);
-#endif /* CONFIG_CRASH_DUMP */
 }
+#endif /* CONFIG_CRASH_DUMP */
 
 void smp_cpu_set_polarization(int cpu, int val)
 {
@@ -680,7 +679,7 @@ static struct sclp_core_info *smp_get_core_info(void)
 		for (address = 0;
 		     address < (SCLP_MAX_CORES << smp_cpu_mt_shift);
 		     address += (1U << smp_cpu_mt_shift)) {
-			if (__pcpu_sigp_relax(address, SIGP_SENSE, 0, NULL) ==
+			if (__pcpu_sigp_relax(address, SIGP_SENSE, 0) ==
 			    SIGP_CC_NOT_OPERATIONAL)
 				continue;
 			info->core[info->configured].core_id =
@@ -924,7 +923,7 @@ void __init smp_prepare_boot_cpu(void)
 
 	pcpu->state = CPU_STATE_CONFIGURED;
 	pcpu->address = stap();
-	pcpu->lowcore = (struct _lowcore *)(unsigned long) store_prefix();
+	pcpu->lowcore = (struct lowcore *)(unsigned long) store_prefix();
 	S390_lowcore.percpu_offset = __per_cpu_offset[0];
 	smp_cpu_set_polarization(0, POLARIZATION_UNKNOWN);
 	set_cpu_present(0, true);
diff --git a/arch/s390/kernel/sysinfo.c b/arch/s390/kernel/sysinfo.c
index 99babea..f7dba388 100644
--- a/arch/s390/kernel/sysinfo.c
+++ b/arch/s390/kernel/sysinfo.c
@@ -111,8 +111,7 @@ static void stsi_1_1_1(struct seq_file *m, struct sysinfo_1_1_1 *info)
 
 static void stsi_15_1_x(struct seq_file *m, struct sysinfo_15_1_x *info)
 {
-	static int max_mnest;
-	int i, rc;
+	int i;
 
 	seq_putc(m, '\n');
 	if (!MACHINE_HAS_TOPOLOGY)
@@ -123,7 +122,7 @@ static void stsi_15_1_x(struct seq_file *m, struct sysinfo_15_1_x *info)
 	for (i = 0; i < TOPOLOGY_NR_MAG; i++)
 		seq_printf(m, " %d", info->mag[i]);
 	seq_putc(m, '\n');
-#ifdef CONFIG_SCHED_MC
+#ifdef CONFIG_SCHED_TOPOLOGY
 	store_topology(info);
 	seq_printf(m, "CPU Topology SW:     ");
 	for (i = 0; i < TOPOLOGY_NR_MAG; i++)
@@ -145,6 +144,10 @@ static void stsi_1_2_2(struct seq_file *m, struct sysinfo_1_2_2 *info)
 	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);
+	if (info->mt_installed) {
+		seq_printf(m, "CPUs G-MTID:          %d\n", info->mt_gtid);
+		seq_printf(m, "CPUs S-MTID:          %d\n", info->mt_stid);
+	}
 	/*
 	 * Sigh 2. According to the specification the alternate
 	 * capability field is a 32 bit floating point number
@@ -194,13 +197,10 @@ static void stsi_2_2_2(struct seq_file *m, struct sysinfo_2_2_2 *info)
 	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);
-	if (info->mt_installed & 0x80) {
-		seq_printf(m, "LPAR CPUs G-MTID:     %d\n",
-			   info->mt_general & 0x1f);
-		seq_printf(m, "LPAR CPUs S-MTID:     %d\n",
-			   info->mt_installed & 0x1f);
-		seq_printf(m, "LPAR CPUs PS-MTID:    %d\n",
-			   info->mt_psmtid & 0x1f);
+	if (info->mt_installed) {
+		seq_printf(m, "LPAR CPUs G-MTID:     %d\n", info->mt_gtid);
+		seq_printf(m, "LPAR CPUs S-MTID:     %d\n", info->mt_stid);
+		seq_printf(m, "LPAR CPUs PS-MTID:    %d\n", info->mt_psmtid);
 	}
 }
 
diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c
index 1b18118..d69d648 100644
--- a/arch/s390/kernel/traps.c
+++ b/arch/s390/kernel/traps.c
@@ -260,11 +260,8 @@ void vector_exception(struct pt_regs *regs)
 
 void data_exception(struct pt_regs *regs)
 {
-	__u16 __user *location;
 	int signal = 0;
 
-	location = get_trap_ip(regs);
-
 	save_fpu_regs();
 	if (current->thread.fpu.fpc & FPC_DXC_MASK)
 		signal = SIGFPE;
diff --git a/arch/s390/kernel/vdso.c b/arch/s390/kernel/vdso.c
index 59eddb0..94495ca 100644
--- a/arch/s390/kernel/vdso.c
+++ b/arch/s390/kernel/vdso.c
@@ -80,7 +80,7 @@ struct vdso_data *vdso_data = &vdso_data_store.data;
 /*
  * Setup vdso data page.
  */
-static void vdso_init_data(struct vdso_data *vd)
+static void __init vdso_init_data(struct vdso_data *vd)
 {
 	vd->ectg_available = test_facility(31);
 }
@@ -90,9 +90,10 @@ static void vdso_init_data(struct vdso_data *vd)
  */
 #define SEGMENT_ORDER	2
 
-int vdso_alloc_per_cpu(struct _lowcore *lowcore)
+int vdso_alloc_per_cpu(struct lowcore *lowcore)
 {
 	unsigned long segment_table, page_table, page_frame;
+	struct vdso_per_cpu_data *vd;
 	u32 *psal, *aste;
 	int i;
 
@@ -107,6 +108,12 @@ int vdso_alloc_per_cpu(struct _lowcore *lowcore)
 	if (!segment_table || !page_table || !page_frame)
 		goto out;
 
+	/* Initialize per-cpu vdso data page */
+	vd = (struct vdso_per_cpu_data *) page_frame;
+	vd->cpu_nr = lowcore->cpu_nr;
+	vd->node_id = cpu_to_node(vd->cpu_nr);
+
+	/* Set up access register mode page table */
 	clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
 		    PAGE_SIZE << SEGMENT_ORDER);
 	clear_table((unsigned long *) page_table, _PAGE_INVALID,
@@ -138,7 +145,7 @@ out:
 	return -ENOMEM;
 }
 
-void vdso_free_per_cpu(struct _lowcore *lowcore)
+void vdso_free_per_cpu(struct lowcore *lowcore)
 {
 	unsigned long segment_table, page_table, page_frame;
 	u32 *psal, *aste;
@@ -163,7 +170,7 @@ static void vdso_init_cr5(void)
 
 	if (!vdso_enabled)
 		return;
-	cr5 = offsetof(struct _lowcore, paste);
+	cr5 = offsetof(struct lowcore, paste);
 	__ctl_load(cr5, 5, 5);
 }
 
@@ -299,8 +306,6 @@ static int __init vdso_init(void)
 
 	get_page(virt_to_page(vdso_data));
 
-	smp_mb();
-
 	return 0;
 }
 early_initcall(vdso_init);
diff --git a/arch/s390/kernel/vdso32/Makefile b/arch/s390/kernel/vdso32/Makefile
index ee8a18e..f9c4595 100644
--- a/arch/s390/kernel/vdso32/Makefile
+++ b/arch/s390/kernel/vdso32/Makefile
@@ -1,6 +1,6 @@
 # List of files in the vdso, has to be asm only for now
 
-obj-vdso32 = gettimeofday.o clock_getres.o clock_gettime.o note.o
+obj-vdso32 = gettimeofday.o clock_getres.o clock_gettime.o note.o getcpu.o
 
 # Build rules
 
diff --git a/arch/s390/kernel/vdso32/getcpu.S b/arch/s390/kernel/vdso32/getcpu.S
new file mode 100644
index 0000000..c1ed0b7
--- /dev/null
+++ b/arch/s390/kernel/vdso32/getcpu.S
@@ -0,0 +1,43 @@
+/*
+ * Userland implementation of getcpu() for 32 bits processes in a
+ * s390 kernel for use in the vDSO
+ *
+ *  Copyright IBM Corp. 2016
+ *  Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com>
+ */
+#include <asm/vdso.h>
+#include <asm/asm-offsets.h>
+
+	.text
+	.align 4
+	.globl __kernel_getcpu
+	.type  __kernel_getcpu,@function
+__kernel_getcpu:
+	.cfi_startproc
+	ear	%r1,%a4
+	lhi	%r4,1
+	sll	%r4,24
+	sar	%a4,%r4
+	la	%r4,0
+	epsw	%r0,0
+	sacf	512
+	l	%r5,__VDSO_CPU_NR(%r4)
+	l	%r4,__VDSO_NODE_ID(%r4)
+	tml	%r0,0x4000
+	jo	1f
+	tml	%r0,0x8000
+	jno	0f
+	sacf	256
+	j	1f
+0:	sacf	0
+1:	sar	%a4,%r1
+	ltr	%r2,%r2
+	jz	2f
+	st	%r5,0(%r2)
+2:	ltr	%r3,%r3
+	jz	3f
+	st	%r4,0(%r3)
+3:	lhi	%r2,0
+	br	%r14
+	.cfi_endproc
+	.size	__kernel_getcpu,.-__kernel_getcpu
diff --git a/arch/s390/kernel/vdso32/vdso32.lds.S b/arch/s390/kernel/vdso32/vdso32.lds.S
index a8c379f..8f048c2 100644
--- a/arch/s390/kernel/vdso32/vdso32.lds.S
+++ b/arch/s390/kernel/vdso32/vdso32.lds.S
@@ -132,6 +132,7 @@ VERSION
 		__kernel_gettimeofday;
 		__kernel_clock_gettime;
 		__kernel_clock_getres;
+		__kernel_getcpu;
 
 	local: *;
 	};
diff --git a/arch/s390/kernel/vdso64/Makefile b/arch/s390/kernel/vdso64/Makefile
index c4b03f9..058659c 100644
--- a/arch/s390/kernel/vdso64/Makefile
+++ b/arch/s390/kernel/vdso64/Makefile
@@ -1,6 +1,6 @@
 # List of files in the vdso, has to be asm only for now
 
-obj-vdso64 = gettimeofday.o clock_getres.o clock_gettime.o note.o
+obj-vdso64 = gettimeofday.o clock_getres.o clock_gettime.o note.o getcpu.o
 
 # Build rules
 
diff --git a/arch/s390/kernel/vdso64/getcpu.S b/arch/s390/kernel/vdso64/getcpu.S
new file mode 100644
index 0000000..4cbe982
--- /dev/null
+++ b/arch/s390/kernel/vdso64/getcpu.S
@@ -0,0 +1,42 @@
+/*
+ * Userland implementation of getcpu() for 64 bits processes in a
+ * s390 kernel for use in the vDSO
+ *
+ *  Copyright IBM Corp. 2016
+ *  Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com>
+ */
+#include <asm/vdso.h>
+#include <asm/asm-offsets.h>
+
+	.text
+	.align 4
+	.globl __kernel_getcpu
+	.type  __kernel_getcpu,@function
+__kernel_getcpu:
+	.cfi_startproc
+	ear	%r1,%a4
+	llilh	%r4,0x0100
+	sar	%a4,%r4
+	la	%r4,0
+	epsw	%r0,0
+	sacf	512
+	l	%r5,__VDSO_CPU_NR(%r4)
+	l	%r4,__VDSO_NODE_ID(%r4)
+	tml	%r0,0x4000
+	jo	1f
+	tml	%r0,0x8000
+	jno	0f
+	sacf	256
+	j	1f
+0:	sacf	0
+1:	sar	%a4,%r1
+	ltgr	%r2,%r2
+	jz	2f
+	st	%r5,0(%r2)
+2:	ltgr	%r3,%r3
+	jz	3f
+	st	%r4,0(%r3)
+3:	lghi	%r2,0
+	br	%r14
+	.cfi_endproc
+	.size	__kernel_getcpu,.-__kernel_getcpu
diff --git a/arch/s390/kernel/vdso64/vdso64.lds.S b/arch/s390/kernel/vdso64/vdso64.lds.S
index 9f5979d..f35455d 100644
--- a/arch/s390/kernel/vdso64/vdso64.lds.S
+++ b/arch/s390/kernel/vdso64/vdso64.lds.S
@@ -132,6 +132,7 @@ VERSION
 		__kernel_gettimeofday;
 		__kernel_clock_gettime;
 		__kernel_clock_getres;
+		__kernel_getcpu;
 
 	local: *;
 	};