Merge Qualcom Hexagon architecture

This is the fifth version of the patchset (with one tiny whitespace fix)
to the Linux kernel to support the Qualcomm Hexagon architecture.

Between now and the next pull requests, Richard Kuo should have his key
signed, etc., and should be back on kernel.org.  In the meantime, this
got merged as a emailed patch-series.

* Hexagon: (36 commits)
  Add extra arch overrides to asm-generic/checksum.h
  Hexagon: Add self to MAINTAINERS
  Hexagon: Add basic stacktrace functionality for Hexagon architecture.
  Hexagon: Add configuration and makefiles for the Hexagon architecture.
  Hexagon: Comet platform support
  Hexagon: kgdb support files
  Hexagon: Add page-fault support.
  Hexagon: Add page table header files & etc.
  Hexagon: Add ioremap support
  Hexagon: Provide DMA implementation
  Hexagon: Implement basic TLB management routines for Hexagon.
  Hexagon: Implement basic cache-flush support
  Hexagon: Provide basic implementation and/or stubs for I/O routines.
  Hexagon: Add user access functions
  Hexagon: Add locking types and functions
  Hexagon: Add SMP support
  Hexagon: Provide basic debugging and system trap support.
  Hexagon: Add ptrace support
  Hexagon: Add time and timer functions
  Hexagon: Add interrupts
  ...

1 files changed, 162 insertions, 0 deletions

diff --git a/arch/hexagon/kernel/head.S b/arch/hexagon/kernel/head.S
new file mode 100644
index 0000000..8e6b819
--- /dev/null
+++ b/arch/hexagon/kernel/head.S
@@ -0,0 +1,162 @@
+/*
+ * Early kernel startup code for Hexagon
+ *
+ * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
+ *
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * 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., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#include <linux/linkage.h>
+#include <linux/init.h>
+#include <asm/asm-offsets.h>
+#include <asm/mem-layout.h>
+#include <asm/vm_mmu.h>
+#include <asm/page.h>
+
+	__INIT
+ENTRY(stext)
+	/*
+	 * VMM will already have set up true vector page, MMU, etc.
+	 * To set up initial kernel identity map, we have to pass
+	 * the VMM a pointer to some canonical page tables. In
+	 * this implementation, we're assuming that we've got
+	 * them precompiled. Generate value in R24, as we'll need
+	 * it again shortly.
+	 */
+	r24.L = #LO(swapper_pg_dir)
+	r24.H = #HI(swapper_pg_dir)
+
+	/*
+	 * Symbol is kernel segment address, but we need
+	 * the logical/physical address.
+	 */
+	r24 = asl(r24, #2)
+	r24 = lsr(r24, #2)
+
+	r0 = r24
+
+	/*
+	 * Initialize a 16MB PTE to make the virtual and physical
+	 * addresses where the kernel was loaded be identical.
+	 */
+#define PTE_BITS ( __HVM_PTE_R | __HVM_PTE_W | __HVM_PTE_X	\
+		  | __HEXAGON_C_WB_L2 << 6			\
+		  | __HVM_PDE_S_4MB)
+
+	r1 = pc
+	r2.H = #0xffc0
+	r2.L = #0x0000
+	r1 = and(r1,r2)		/* round PC to 4MB boundary	*/
+	r2 = lsr(r1, #22)	/* 4MB page number		*/
+	r2 = asl(r2, #2)	/* times sizeof(PTE) (4bytes)	*/
+	r0 = add(r0,r2)		/* r0 = address of correct PTE	*/
+	r2 = #PTE_BITS
+	r1 = add(r1,r2)		/* r1 = 4MB PTE for the first entry	*/
+	r2.h = #0x0040
+	r2.l = #0x0000		/* 4MB	*/
+	memw(r0 ++ #4) = r1
+	r1 = add(r1, r2)
+	memw(r0 ++ #4) = r1
+
+	r0 = r24
+
+	/*
+	 * The subroutine wrapper around the virtual instruction touches
+	 * no memory, so we should be able to use it even here.
+	 */
+	call	__vmnewmap;
+
+	/*  Jump into virtual address range.  */
+
+	r31.h = #hi(__head_s_vaddr_target)
+	r31.l = #lo(__head_s_vaddr_target)
+	jumpr r31
+
+	/*  Insert trippy space effects.  */
+
+__head_s_vaddr_target:
+	/*
+	 * Tear down VA=PA translation now that we are running
+	 * in the desgnated kernel segments.
+	 */
+	r0 = #__HVM_PDE_S_INVALID
+	r1 = r24
+	loop0(1f,#0x100)
+1:
+	{
+		memw(R1 ++ #4) = R0
+	}:endloop0
+
+	r0 = r24
+	call __vmnewmap
+
+	/*  Go ahead and install the trap0 return so angel calls work  */
+	r0.h = #hi(_K_provisional_vec)
+	r0.l = #lo(_K_provisional_vec)
+	call __vmsetvec
+
+	/*
+	 * OK, at this point we should start to be much more careful,
+	 * we're going to enter C code and start touching memory
+	 * in all sorts of places.
+	 * This means:
+	 *      SGP needs to be OK
+	 *	Need to lock shared resources
+	 *	A bunch of other things that will cause
+	 * 	all kinds of painful bugs
+	 */
+
+	/*
+	 * Stack pointer should be pointed at the init task's
+	 * thread stack, which should have been declared in arch/init_task.c.
+	 * So uhhhhh...
+	 * It's accessible via the init_thread_union, which is a union
+	 * of a thread_info struct and a stack; of course, the top
+	 * of the stack is not for you.  The end of the stack
+	 * is simply init_thread_union + THREAD_SIZE.
+	 */
+
+	{r29.H = #HI(init_thread_union); r0.H = #HI(_THREAD_SIZE); }
+	{r29.L = #LO(init_thread_union); r0.L = #LO(_THREAD_SIZE); }
+
+	/*  initialize the register used to point to current_thread_info */
+	/*  Fixme:  THREADINFO_REG can't be R2 because of that memset thing. */
+	{r29 = add(r29,r0); THREADINFO_REG = r29; }
+
+	/*  Hack:  zero bss; */
+	{ r0.L = #LO(__bss_start);  r1 = #0; r2.l = #LO(__bss_stop); }
+	{ r0.H = #HI(__bss_start);           r2.h = #HI(__bss_stop); }
+
+	r2 = sub(r2,r0);
+	call memset;
+
+	/* Time to make the doughnuts.   */
+	call start_kernel
+
+	/*
+	 * Should not reach here.
+	 */
+1:
+	jump 1b
+
+.p2align PAGE_SHIFT
+ENTRY(external_cmdline_buffer)
+        .fill _PAGE_SIZE,1,0
+
+.data
+.p2align PAGE_SHIFT
+ENTRY(empty_zero_page)
+        .fill _PAGE_SIZE,1,0