[PATCH] xtensa: Architecture support for Tensilica Xtensa Part 3

The attached patches provides part 3 of an architecture implementation for the
Tensilica Xtensa CPU series.
Signed-off-by: NChris Zankel <chris@zankel.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

arch/xtensa/kernel/Makefile

+#
+# Makefile for the Linux/Xtensa kernel.
+#
+
+extra-y := head.o vmlinux.lds
+
+
+obj-y := align.o entry.o irq.o coprocessor.o process.o ptrace.o semaphore.o  \
+	 setup.o signal.o syscalls.o time.o traps.o vectors.o platform.o  \
+	 pci-dma.o
+
+## windowspill.o
+
+obj-$(CONFIG_KGDB) += xtensa-stub.o
+obj-$(CONFIG_PCI) += pci.o
+obj-$(CONFIG_MODULES) += xtensa_ksyms.o module.o
+
+

arch/xtensa/kernel/align.S

+/*
+ * arch/xtensa/kernel/align.S
+ *
+ * Handle unalignment exceptions in kernel space.
+ *
+ * This file is subject to the terms and conditions of the GNU General
+ * Public License.  See the file "COPYING" in the main directory of
+ * this archive for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica, Inc.
+ *
+ * Rewritten by Chris Zankel <chris@zankel.net>
+ *
+ * Based on work from Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * and Marc Gauthier <marc@tensilica.com, marc@alimni.uwaterloo.ca>
+ */
+
+#include <linux/linkage.h>
+#include <asm/ptrace.h>
+#include <asm/ptrace.h>
+#include <asm/current.h>
+#include <asm/offsets.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/page.h>
+#include <asm/thread_info.h>
+
+#if XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION
+
+/*  First-level exception handler for unaligned exceptions.
+ *
+ *  Note: This handler works only for kernel exceptions.  Unaligned user
+ *        access should get a seg fault.
+ */
+
+/* Big and little endian 16-bit values are located in
+ * different halves of a register.  HWORD_START helps to
+ * abstract the notion of extracting a 16-bit value from a
+ * register.
+ * We also have to define new shifting instructions because
+ * lsb and msb are on 'opposite' ends in a register for
+ * different endian machines.
+ *
+ * Assume a memory region in ascending address:
+ *   	0 1 2 3|4 5 6 7
+ *
+ * When loading one word into a register, the content of that register is:
+ *  LE	3 2 1 0, 7 6 5 4
+ *  BE  0 1 2 3, 4 5 6 7
+ *
+ * Masking the bits of the higher/lower address means:
+ *  LE  X X 0 0, 0 0 X X
+ *  BE	0 0 X X, X X 0 0
+ *
+ * Shifting to higher/lower addresses, means:
+ *  LE  shift left / shift right
+ *  BE  shift right / shift left
+ *
+ * Extracting 16 bits from a 32 bit reg. value to higher/lower address means:
+ *  LE  mask 0 0 X X / shift left
+ *  BE  shift left / mask 0 0 X X
+ */
+
+#define UNALIGNED_USER_EXCEPTION
+
+#if XCHAL_HAVE_BE
+
+#define HWORD_START	16
+#define	INSN_OP0	28
+#define	INSN_T		24
+#define	INSN_OP1	16
+
+.macro __src_b	r, w0, w1;	src	\r, \w0, \w1;	.endm
+.macro __ssa8	r;		ssa8b	\r;		.endm
+.macro __ssa8r	r;		ssa8l	\r;		.endm
+.macro __sh	r, s;		srl	\r, \s;		.endm
+.macro __sl	r, s;		sll	\r, \s;		.endm
+.macro __exth	r, s;		extui	\r, \s, 0, 16;	.endm
+.macro __extl	r, s;		slli	\r, \s, 16;	.endm
+
+#else
+
+#define HWORD_START	0
+#define	INSN_OP0	0
+#define	INSN_T		4
+#define	INSN_OP1	12
+
+.macro __src_b	r, w0, w1;	src	\r, \w1, \w0;	.endm
+.macro __ssa8	r;		ssa8l	\r;		.endm
+.macro __ssa8r	r;		ssa8b	\r;		.endm
+.macro __sh	r, s;		sll	\r, \s;		.endm
+.macro __sl	r, s;		srl	\r, \s;		.endm
+.macro __exth	r, s;		slli	\r, \s, 16;	.endm
+.macro __extl	r, s;		extui	\r, \s, 0, 16;	.endm
+
+#endif
+
+/*
+ *	xxxx xxxx = imm8 field
+ *	     yyyy = imm4 field
+ *	     ssss = s field
+ *	     tttt = t field
+ *
+ *	       		 16		    0
+ *		          -------------------
+ *	L32I.N		  yyyy ssss tttt 1000
+ *	S32I.N	          yyyy ssss tttt 1001
+ *
+ *	       23			    0
+ *		-----------------------------
+ *	res	          0000           0010
+ *	L16UI	xxxx xxxx 0001 ssss tttt 0010
+ *	L32I	xxxx xxxx 0010 ssss tttt 0010
+ *	XXX	          0011 ssss tttt 0010
+ *	XXX	          0100 ssss tttt 0010
+ *	S16I	xxxx xxxx 0101 ssss tttt 0010
+ *	S32I	xxxx xxxx 0110 ssss tttt 0010
+ *	XXX	          0111 ssss tttt 0010
+ *	XXX	          1000 ssss tttt 0010
+ *	L16SI	xxxx xxxx 1001 ssss tttt 0010
+ *	XXX	          1010           0010
+ *      **L32AI	xxxx xxxx 1011 ssss tttt 0010 unsupported
+ *	XXX	          1100           0010
+ *	XXX	          1101           0010
+ *	XXX	          1110           0010
+ *	**S32RI	xxxx xxxx 1111 ssss tttt 0010 unsupported
+ *		-----------------------------
+ *                           ^         ^    ^
+ *    sub-opcode (NIBBLE_R) -+         |    |
+ *       t field (NIBBLE_T) -----------+    |
+ *  major opcode (NIBBLE_OP0) --------------+
+ */
+
+#define OP0_L32I_N	0x8		/* load immediate narrow */
+#define OP0_S32I_N	0x9		/* store immediate narrow */
+#define OP1_SI_MASK	0x4		/* OP1 bit set for stores */
+#define OP1_SI_BIT	2		/* OP1 bit number for stores */
+
+#define OP1_L32I	0x2
+#define OP1_L16UI	0x1
+#define OP1_L16SI	0x9
+#define OP1_L32AI	0xb
+
+#define OP1_S32I	0x6
+#define OP1_S16I	0x5
+#define OP1_S32RI	0xf
+
+/*
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ */
+
+
+ENTRY(fast_unaligned)
+
+	/* Note: We don't expect the address to be aligned on a word
+	 *       boundary. After all, the processor generated that exception
+	 *       and it would be a hardware fault.
+	 */
+
+	/* Save some working register */
+
+	s32i	a4, a2, PT_AREG4
+	s32i	a5, a2, PT_AREG5
+	s32i	a6, a2, PT_AREG6
+	s32i	a7, a2, PT_AREG7
+	s32i	a8, a2, PT_AREG8
+
+	rsr	a0, DEPC
+	xsr	a3, EXCSAVE_1
+	s32i	a0, a2, PT_AREG2
+	s32i	a3, a2, PT_AREG3
+
+	/* Keep value of SAR in a0 */
+
+	rsr	a0, SAR
+	rsr	a8, EXCVADDR		# load unaligned memory address
+
+	/* Now, identify one of the following load/store instructions.
+	 *
+	 * The only possible danger of a double exception on the
+	 * following l32i instructions is kernel code in vmalloc
+	 * memory. The processor was just executing at the EPC_1
+	 * address, and indeed, already fetched the instruction.  That
+	 * guarantees a TLB mapping, which hasn't been replaced by
+	 * this unaligned exception handler that uses only static TLB
+	 * mappings. However, high-level interrupt handlers might
+	 * modify TLB entries, so for the generic case, we register a
+	 * TABLE_FIXUP handler here, too.
+	 */
+
+	/* a3...a6 saved on stack, a2 = SP */
+
+	/* Extract the instruction that caused the unaligned access. */
+
+	rsr	a7, EPC_1	# load exception address
+	movi	a3, ~3
+	and	a3, a3, a7	# mask lower bits
+
+	l32i	a4, a3, 0	# load 2 words
+	l32i	a5, a3, 4
+
+	__ssa8	a7
+	__src_b	a4, a4, a5	# a4 has the instruction
+
+	/* Analyze the instruction (load or store?). */
+
+	extui	a5, a4, INSN_OP0, 4	# get insn.op0 nibble
+
+#if XCHAL_HAVE_NARROW
+	_beqi	a5, OP0_L32I_N, .Lload	# L32I.N, jump
+	addi	a6, a5, -OP0_S32I_N
+	_beqz	a6, .Lstore		# S32I.N, do a store
+#endif
+	/* 'store indicator bit' not set, jump */
+	_bbci.l	a4, OP1_SI_BIT + INSN_OP1, .Lload
+
+	/* Store: Jump to table entry to get the value in the source register.*/
+
+.Lstore:movi	a5, .Lstore_table	# table
+	extui	a6, a4, INSN_T, 4	# get source register
+	addx8	a5, a6, a5
+	jx	a5			# jump into table
+
+	/* Invalid instruction, CRITICAL! */
+.Linvalid_instruction_load:
+	j	.Linvalid_instruction
+
+	/* Load: Load memory address. */
+
+.Lload: movi	a3, ~3
+	and	a3, a3, a8		# align memory address
+
+	__ssa8	a8
+#ifdef UNALIGNED_USER_EXCEPTION
+	addi	a3, a3, 8
+	l32e	a5, a3, -8
+	l32e	a6, a3, -4
+#else
+	l32i	a5, a3, 0
+	l32i	a6, a3, 4
+#endif
+	__src_b	a3, a5, a6		# a3 has the data word
+
+#if XCHAL_HAVE_NARROW
+	addi	a7, a7, 2		# increment PC (assume 16-bit insn)
+
+	extui	a5, a4, INSN_OP0, 4
+	_beqi	a5, OP0_L32I_N, 1f	# l32i.n: jump
+
+	addi	a7, a7, 1
+#else
+	addi	a7, a7, 3
+#endif
+
+	extui	a5, a4, INSN_OP1, 4
+	_beqi	a5, OP1_L32I, 1f	# l32i: jump
+
+	extui	a3, a3, 0, 16		# extract lower 16 bits
+	_beqi	a5, OP1_L16UI, 1f
+	addi	a5, a5, -OP1_L16SI
+	_bnez	a5, .Linvalid_instruction_load
+
+	/* sign extend value */
+
+	slli	a3, a3, 16
+	srai	a3, a3, 16
+
+	/* Set target register. */
+
+1:
+
+#if XCHAL_HAVE_LOOP
+	rsr	a3, LEND		# check if we reached LEND
+	bne	a7, a3, 1f
+	rsr	a3, LCOUNT		# and LCOUNT != 0
+	beqz	a3, 1f
+	addi	a3, a3, -1		# decrement LCOUNT and set
+	rsr	a7, LBEG		# set PC to LBEGIN
+	wsr	a3, LCOUNT
+#endif
+
+1:	wsr	a7, EPC_1		# skip load instruction
+	extui	a4, a4, INSN_T, 4	# extract target register
+	movi	a5, .Lload_table
+	addx8	a4, a4, a5
+	jx	a4			# jump to entry for target register
+
+	.align	8
+.Lload_table:
+	s32i	a3, a2, PT_AREG0;	_j .Lexit;	.align 8
+	mov	a1, a3;			_j .Lexit;	.align 8 # fishy??
+	s32i	a3, a2, PT_AREG2;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG3;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG4;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG5;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG6;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG7;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG8;	_j .Lexit;	.align 8
+	mov	a9, a3		;	_j .Lexit;	.align 8
+	mov	a10, a3		;	_j .Lexit;	.align 8
+	mov	a11, a3		;	_j .Lexit;	.align 8
+	mov	a12, a3		;	_j .Lexit;	.align 8
+	mov	a13, a3		;	_j .Lexit;	.align 8
+	mov	a14, a3		;	_j .Lexit;	.align 8
+	mov	a15, a3		;	_j .Lexit;	.align 8
+
+.Lstore_table:
+	l32i	a3, a2, PT_AREG0;	_j 1f;	.align 8
+	mov	a3, a1;			_j 1f;	.align 8	# fishy??
+	l32i	a3, a2, PT_AREG2;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG3;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG4;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG5;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG6;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG7;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG8;	_j 1f;	.align 8
+	mov	a3, a9		;	_j 1f;	.align 8
+	mov	a3, a10		;	_j 1f;	.align 8
+	mov	a3, a11		;	_j 1f;	.align 8
+	mov	a3, a12		;	_j 1f;	.align 8
+	mov	a3, a13		;	_j 1f;	.align 8
+	mov	a3, a14		;	_j 1f;	.align 8
+	mov	a3, a15		;	_j 1f;	.align 8
+
+1: 	# a7: instruction pointer, a4: instruction, a3: value
+
+	movi	a6, 0			# mask: ffffffff:00000000
+
+#if XCHAL_HAVE_NARROW
+	addi	a7, a7, 2		# incr. PC,assume 16-bit instruction
+
+	extui	a5, a4, INSN_OP0, 4	# extract OP0
+	addi	a5, a5, -OP0_S32I_N
+	_beqz	a5, 1f			# s32i.n: jump
+
+	addi	a7, a7, 1		# increment PC, 32-bit instruction
+#else
+	addi	a7, a7, 3		# increment PC, 32-bit instruction
+#endif
+
+	extui	a5, a4, INSN_OP1, 4	# extract OP1
+	_beqi	a5, OP1_S32I, 1f	# jump if 32 bit store
+	_bnei	a5, OP1_S16I, .Linvalid_instruction_store
+
+	movi	a5, -1
+	__extl	a3, a3			# get 16-bit value
+	__exth	a6, a5			# get 16-bit mask ffffffff:ffff0000
+
+	/* Get memory address */
+
+1:
+#if XCHAL_HAVE_LOOP
+	rsr	a3, LEND		# check if we reached LEND
+	bne	a7, a3, 1f
+	rsr	a3, LCOUNT		# and LCOUNT != 0
+	beqz	a3, 1f
+	addi	a3, a3, -1		# decrement LCOUNT and set
+	rsr	a7, LBEG		# set PC to LBEGIN
+	wsr	a3, LCOUNT
+#endif
+
+1:	wsr	a7, EPC_1		# skip store instruction
+	movi	a4, ~3
+	and	a4, a4, a8		# align memory address
+
+	/* Insert value into memory */
+
+	movi	a5, -1			# mask: ffffffff:XXXX0000
+#ifdef UNALIGNED_USER_EXCEPTION
+	addi	a4, a4, 8
+#endif
+
+	__ssa8r a8
+	__src_b	a7, a5, a6		# lo-mask  F..F0..0 (BE) 0..0F..F (LE)
+	__src_b	a6, a6, a5		# hi-mask  0..0F..F (BE) F..F0..0 (LE)
+#ifdef UNALIGNED_USER_EXCEPTION
+	l32e	a5, a4, -8
+#else
+	l32i	a5, a4, 0		# load lower address word
+#endif
+	and	a5, a5, a7		# mask
+	__sh	a7, a3 			# shift value
+	or	a5, a5, a7		# or with original value
+#ifdef UNALIGNED_USER_EXCEPTION
+	s32e	a5, a4, -8
+	l32e	a7, a4, -4
+#else
+	s32i	a5, a4, 0		# store
+	l32i	a7, a4, 4		# same for upper address word
+#endif
+	__sl	a5, a3
+	and	a6, a7, a6
+	or	a6, a6, a5
+#ifdef UNALIGNED_USER_EXCEPTION
+	s32e	a6, a4, -4
+#else
+	s32i	a6, a4, 4
+#endif
+
+	/* Done. restore stack and return */
+
+.Lexit:
+	movi	a4, 0
+	rsr	a3, EXCSAVE_1
+	s32i	a4, a3, EXC_TABLE_FIXUP
+
+	/* Restore working register */
+
+	l32i	a7, a2, PT_AREG7
+	l32i	a6, a2, PT_AREG6
+	l32i	a5, a2, PT_AREG5
+	l32i	a4, a2, PT_AREG4
+	l32i	a3, a2, PT_AREG3
+
+	/* restore SAR and return */
+
+	wsr	a0, SAR
+	l32i	a0, a2, PT_AREG0
+	l32i	a2, a2, PT_AREG2
+	rfe
+
+	/* We cannot handle this exception. */
+
+	.extern _kernel_exception
+.Linvalid_instruction_store:
+.Linvalid_instruction:
+
+	/* Restore a4...a8 and SAR, set SP, and jump to default exception. */
+
+	l32i	a8, a2, PT_AREG8
+	l32i	a7, a2, PT_AREG7
+	l32i	a6, a2, PT_AREG6
+	l32i	a5, a2, PT_AREG5
+	l32i	a4, a2, PT_AREG4
+	wsr	a0, SAR
+	mov	a1, a2
+
+	rsr	a0, PS
+        bbsi.l  a2, PS_UM_SHIFT, 1f     # jump if user mode
+
+	movi	a0, _kernel_exception
+	jx	a0
+
+1:	movi	a0, _user_exception
+	jx	a0
+
+
+#endif /* XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION */
+

arch/xtensa/kernel/asm-offsets.c

+/*
+ * arch/xtensa/kernel/asm-offsets.c
+ *
+ * Generates definitions from c-type structures used by assembly sources.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ */
+
+#include <asm/processor.h>
+
+#include <linux/types.h>
+#include <linux/sched.h>
+#include <linux/stddef.h>
+#include <linux/thread_info.h>
+#include <linux/ptrace.h>
+#include <asm/ptrace.h>
+#include <asm/processor.h>
+#include <asm/uaccess.h>
+
+#define DEFINE(sym, val) asm volatile("\n->" #sym " %0 " #val : : "i" (val))
+#define BLANK() asm volatile("\n->" : : )
+
+int main(void)
+{
+	/* struct pt_regs */
+	DEFINE(PT_PC, offsetof (struct pt_regs, pc));
+	DEFINE(PT_PS, offsetof (struct pt_regs, ps));
+	DEFINE(PT_DEPC, offsetof (struct pt_regs, depc));
+	DEFINE(PT_EXCCAUSE, offsetof (struct pt_regs, exccause));
+	DEFINE(PT_EXCVADDR, offsetof (struct pt_regs, excvaddr));
+	DEFINE(PT_DEBUGCAUSE, offsetof (struct pt_regs, debugcause));
+	DEFINE(PT_WMASK, offsetof (struct pt_regs, wmask));
+	DEFINE(PT_LBEG, offsetof (struct pt_regs, lbeg));
+	DEFINE(PT_LEND, offsetof (struct pt_regs, lend));
+	DEFINE(PT_LCOUNT, offsetof (struct pt_regs, lcount));
+	DEFINE(PT_SAR, offsetof (struct pt_regs, sar));
+	DEFINE(PT_SYSCALL, offsetof (struct pt_regs, syscall));
+	DEFINE(PT_AREG, offsetof (struct pt_regs, areg[0]));
+	DEFINE(PT_AREG0, offsetof (struct pt_regs, areg[0]));
+	DEFINE(PT_AREG1, offsetof (struct pt_regs, areg[1]));
+	DEFINE(PT_AREG2, offsetof (struct pt_regs, areg[2]));
+	DEFINE(PT_AREG3, offsetof (struct pt_regs, areg[3]));
+	DEFINE(PT_AREG4, offsetof (struct pt_regs, areg[4]));
+	DEFINE(PT_AREG5, offsetof (struct pt_regs, areg[5]));
+	DEFINE(PT_AREG6, offsetof (struct pt_regs, areg[6]));
+	DEFINE(PT_AREG7, offsetof (struct pt_regs, areg[7]));
+	DEFINE(PT_AREG8, offsetof (struct pt_regs, areg[8]));
+	DEFINE(PT_AREG9, offsetof (struct pt_regs, areg[9]));
+	DEFINE(PT_AREG10, offsetof (struct pt_regs, areg[10]));
+	DEFINE(PT_AREG11, offsetof (struct pt_regs, areg[11]));
+	DEFINE(PT_AREG12, offsetof (struct pt_regs, areg[12]));
+	DEFINE(PT_AREG13, offsetof (struct pt_regs, areg[13]));
+	DEFINE(PT_AREG14, offsetof (struct pt_regs, areg[14]));
+	DEFINE(PT_AREG15, offsetof (struct pt_regs, areg[15]));
+	DEFINE(PT_WINDOWBASE, offsetof (struct pt_regs, windowbase));
+	DEFINE(PT_WINDOWSTART, offsetof(struct pt_regs, windowstart));
+	DEFINE(PT_SIZE, sizeof(struct pt_regs));
+	DEFINE(PT_AREG_END, offsetof (struct pt_regs, areg[XCHAL_NUM_AREGS]));
+	DEFINE(PT_USER_SIZE, offsetof(struct pt_regs, areg[XCHAL_NUM_AREGS]));
+	BLANK();
+
+	/* struct task_struct */
+	DEFINE(TASK_PTRACE, offsetof (struct task_struct, ptrace));
+	DEFINE(TASK_MM, offsetof (struct task_struct, mm));
+	DEFINE(TASK_ACTIVE_MM, offsetof (struct task_struct, active_mm));
+	DEFINE(TASK_PID, offsetof (struct task_struct, pid));
+	DEFINE(TASK_THREAD, offsetof (struct task_struct, thread));
+	DEFINE(TASK_THREAD_INFO, offsetof (struct task_struct, thread_info));
+	DEFINE(TASK_STRUCT_SIZE, sizeof (struct task_struct));
+	BLANK();
+
+	/* struct thread_info (offset from start_struct) */
+	DEFINE(THREAD_RA, offsetof (struct task_struct, thread.ra));
+	DEFINE(THREAD_SP, offsetof (struct task_struct, thread.sp));
+	DEFINE(THREAD_CP_SAVE, offsetof (struct task_struct, thread.cp_save));
+	DEFINE(THREAD_CURRENT_DS, offsetof (struct task_struct, thread.current_ds));
+	BLANK();
+
+	/* struct mm_struct */
+	DEFINE(MM_USERS, offsetof(struct mm_struct, mm_users));
+	DEFINE(MM_PGD, offsetof (struct mm_struct, pgd));
+	DEFINE(MM_CONTEXT, offsetof (struct mm_struct, context));
+	BLANK();
+	DEFINE(PT_SINGLESTEP_BIT, PT_SINGLESTEP_BIT);
+	return 0;
+}
+
+

arch/xtensa/kernel/coprocessor.S

+/*
+ * arch/xtensa/kernel/coprocessor.S
+ *
+ * Xtensa processor configuration-specific table of coprocessor and
+ * other custom register layout information.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2003 - 2005 Tensilica Inc.
+ *
+ * Marc Gauthier <marc@tensilica.com> <marc@alumni.uwaterloo.ca>
+ */
+
+/*
+ * This module contains a table that describes the layout of the various
+ * custom registers and states associated with each coprocessor, as well
+ * as those not associated with any coprocessor ("extra state").
+ * This table is included with core dumps and is available via the ptrace
+ * interface, allowing the layout of such register/state information to
+ * be modified in the kernel without affecting the debugger.  Each
+ * register or state is identified using a 32-bit "libdb target number"
+ * assigned when the Xtensa processor is generated.
+ */
+
+#include <linux/config.h>
+#include <linux/linkage.h>
+#include <asm/processor.h>
+
+#if XCHAL_HAVE_CP
+
+#define CP_LAST ((XCHAL_CP_MAX - 1) * COPROCESSOR_INFO_SIZE)
+
+ENTRY(release_coprocessors)
+
+	entry	a1, 16
+						# a2: task
+	movi	a3, 1 << XCHAL_CP_MAX 		# a3: coprocessor-bit
+	movi	a4, coprocessor_info+CP_LAST	# a4: owner-table
+						# a5: tmp
+	movi	a6, 0				# a6: 0
+	rsil	a7, LOCKLEVEL			# a7: PS
+
+1:	/* Check if task is coprocessor owner of coprocessor[i]. */
+
+	l32i	a5, a4, COPROCESSOR_INFO_OWNER
+	srli	a3, a3, 1
+	beqz	a3, 1f
+	addi	a4, a4, -8
+	beq	a2, a5, 1b
+
+	/* Found an entry: Clear entry CPENABLE bit to disable CP. */
+
+	rsr	a5, CPENABLE
+	s32i	a6, a4, COPROCESSOR_INFO_OWNER
+	xor	a5, a3, a5
+	wsr	a5, CPENABLE
+
+	bnez	a3, 1b
+
+1:	wsr	a7, PS
+	rsync
+	retw
+
+
+ENTRY(disable_coprocessor)
+	entry	sp, 16
+	rsil	a7, LOCKLEVEL
+	rsr	a3, CPENABLE
+	movi	a4, 1
+	ssl	a2
+	sll	a4, a4
+	and	a4, a3, a4
+	xor	a3, a3, a4
+	wsr	a3, CPENABLE
+	wsr	a7, PS
+	rsync
+	retw
+
+ENTRY(enable_coprocessor)
+	entry	sp, 16
+	rsil	a7, LOCKLEVEL
+	rsr	a3, CPENABLE
+	movi	a4, 1
+	ssl	a2
+	sll	a4, a4
+	or	a3, a3, a4
+	wsr	a3, CPENABLE
+	wsr	a7, PS
+	rsync
+	retw
+
+#endif
+
+ENTRY(save_coprocessor_extra)
+	entry	sp, 16
+	xchal_extra_store_funcbody
+	retw
+
+ENTRY(restore_coprocessor_extra)
+	entry	sp, 16
+	xchal_extra_load_funcbody
+	retw
+
+ENTRY(save_coprocessor_registers)
+	entry	sp, 16
+	xchal_cpi_store_funcbody
+	retw
+
+ENTRY(restore_coprocessor_registers)
+	entry	sp, 16
+	xchal_cpi_load_funcbody
+	retw
+
+
+/*
+ *  The Xtensa compile-time HAL (core.h) XCHAL_*_SA_CONTENTS_LIBDB macros
+ *  describe the contents of coprocessor & extra save areas in terms of
+ *  undefined CONTENTS_LIBDB_{SREG,UREG,REGF} macros.  We define these
+ *  latter macros here; they expand into a table of the format we want.
+ *  The general format is:
+ *
+ *	CONTENTS_LIBDB_SREG(libdbnum, offset, size, align, rsv1, name, sregnum,
+ *			    bitmask, rsv2, rsv3)
+ *	CONTENTS_LIBDB_UREG(libdbnum, offset, size, align, rsv1, name, uregnum,
+ *			    bitmask, rsv2, rsv3)
+ *	CONTENTS_LIBDB_REGF(libdbnum, offset, size, align, rsv1, name, index,
+ *			    numentries, contentsize, regname_base,
+ *			    regfile_name, rsv2, rsv3)
+ *
+ *  For this table, we only care about the <libdbnum>, <offset> and <size>
+ *  fields.
+ */
+
+/*  Map all XCHAL CONTENTS macros to the reg_entry asm macro defined below:  */
+
+#define CONTENTS_LIBDB_SREG(libdbnum,offset,size,align,rsv1,name,sregnum,     \
+			    bitmask, rsv2, rsv3)			      \
+		reg_entry libdbnum, offset, size ;
+#define CONTENTS_LIBDB_UREG(libdbnum,offset,size,align,rsv1,name,uregnum,     \
+			    bitmask, rsv2, rsv3)			      \
+		reg_entry libdbnum, offset, size ;
+#define CONTENTS_LIBDB_REGF(libdbnum, offset, size, align, rsv1, name, index, \
+			    numentries, contentsize, regname_base,	      \
+			    regfile_name, rsv2, rsv3)			      \
+		reg_entry libdbnum, offset, size ;
+
+/* A single table entry: */
+	.macro	reg_entry	libdbnum, offset, size
+	 .ifne	(__last_offset-(__last_group_offset+\offset))
+	  /* padding entry */
+	  .word	(0xFC000000+__last_offset-(__last_group_offset+\offset))
+	 .endif
+	 .word	\libdbnum				/* actual entry */
+	 .set	__last_offset, __last_group_offset+\offset+\size
+	.endm	/* reg_entry */
+
+
+/* Table entry that marks the beginning of a group (coprocessor or "extra"): */
+	.macro	reg_group	cpnum, num_entries, align
+	 .set	__last_group_offset, (__last_offset + \align- 1) & -\align
+	 .ifne	\num_entries
+	  .word	0xFD000000+(\cpnum<<16)+\num_entries
+	 .endif
+	.endm	/* reg_group */
+
+/*
+ * Register info tables.
+ */
+
+	.section .rodata, "a"
+	.globl	_xtensa_reginfo_tables
+	.globl	_xtensa_reginfo_table_size
+	.align	4
+_xtensa_reginfo_table_size:
+	.word	_xtensa_reginfo_table_end - _xtensa_reginfo_tables
+
+_xtensa_reginfo_tables:
+	.set	__last_offset, 0
+	reg_group 0xFF, XCHAL_EXTRA_SA_CONTENTS_LIBDB_NUM, XCHAL_EXTRA_SA_ALIGN
+	XCHAL_EXTRA_SA_CONTENTS_LIBDB
+	reg_group 0, XCHAL_CP0_SA_CONTENTS_LIBDB_NUM, XCHAL_CP0_SA_ALIGN
+	XCHAL_CP0_SA_CONTENTS_LIBDB
+	reg_group 1, XCHAL_CP1_SA_CONTENTS_LIBDB_NUM, XCHAL_CP1_SA_ALIGN
+	XCHAL_CP1_SA_CONTENTS_LIBDB
+	reg_group 2, XCHAL_CP2_SA_CONTENTS_LIBDB_NUM, XCHAL_CP2_SA_ALIGN
+	XCHAL_CP2_SA_CONTENTS_LIBDB
+	reg_group 3, XCHAL_CP3_SA_CONTENTS_LIBDB_NUM, XCHAL_CP3_SA_ALIGN
+	XCHAL_CP3_SA_CONTENTS_LIBDB
+	reg_group 4, XCHAL_CP4_SA_CONTENTS_LIBDB_NUM, XCHAL_CP4_SA_ALIGN
+	XCHAL_CP4_SA_CONTENTS_LIBDB
+	reg_group 5, XCHAL_CP5_SA_CONTENTS_LIBDB_NUM, XCHAL_CP5_SA_ALIGN
+	XCHAL_CP5_SA_CONTENTS_LIBDB
+	reg_group 6, XCHAL_CP6_SA_CONTENTS_LIBDB_NUM, XCHAL_CP6_SA_ALIGN
+	XCHAL_CP6_SA_CONTENTS_LIBDB
+	reg_group 7, XCHAL_CP7_SA_CONTENTS_LIBDB_NUM, XCHAL_CP7_SA_ALIGN
+	XCHAL_CP7_SA_CONTENTS_LIBDB
+	.word	0xFC000000	/* invalid register number,marks end of table*/
+_xtensa_reginfo_table_end:
+

arch/xtensa/kernel/head.S

+/*
+ * arch/xtensa/kernel/head.S
+ *
+ * Xtensa Processor startup code.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * Kevin Chea
+ */
+
+#include <xtensa/cacheasm.h>
+#include <linux/config.h>
+#include <asm/processor.h>
+#include <asm/page.h>
+
+/*
+ * This module contains the entry code for kernel images. It performs the
+ * minimal setup needed to call the generic C routines.
+ *
+ * Prerequisites:
+ *
+ * - The kernel image has been loaded to the actual address where it was
+ *   compiled to.
+ * - a2 contains either 0 or a pointer to a list of boot parameters.
+ *   (see setup.c for more details)
+ *
+ */
+
+	.macro	iterate	from, to , cmd
+		.ifeq	((\to - \from) & ~0xfff)
+			\cmd	\from
+			iterate "(\from+1)", \to, \cmd
+		.endif
+	.endm
+
+/*
+ *  _start
+ *
+ *  The bootloader passes a pointer to a list of boot parameters in a2.
+ */
+
+	/* The first bytes of the kernel image must be an instruction, so we
+	 * manually allocate and define the literal constant we need for a jx
+	 * instruction.
+	 */
+
+	.section .head.text, "ax"
+	.globl _start
+_start:	_j	2f
+	.align	4
+1:	.word	_startup
+2:	l32r	a0, 1b
+	jx	a0
+
+	.text
+	.align 4
+_startup:
+
+	/* Disable interrupts and exceptions. */
+
+	movi	a0, XCHAL_PS_EXCM_MASK
+	wsr	a0, PS
+
+	/* Preserve the pointer to the boot parameter list in EXCSAVE_1 */
+
+	wsr	a2, EXCSAVE_1
+
+	/* Start with a fresh windowbase and windowstart.  */
+
+	movi	a1, 1
+	movi	a0, 0
+	wsr	a1, WINDOWSTART
+	wsr	a0, WINDOWBASE
+	rsync
+
+	/* Set a0 to 0 for the remaining initialization. */
+
+	movi	a0, 0
+
+	/* Clear debugging registers. */
+
+#if XCHAL_HAVE_DEBUG
+	wsr	a0, IBREAKENABLE
+	wsr	a0, ICOUNT
+	movi	a1, 15
+	wsr	a0, ICOUNTLEVEL
+
+	.macro reset_dbreak num
+	wsr	a0, DBREAKC + \num
+	.endm
+
+        iterate 0, XCHAL_NUM_IBREAK-1, reset_dbreak
+#endif
+
+	/* Clear CCOUNT (not really necessary, but nice) */
+
+	wsr	a0, CCOUNT	# not really necessary, but nice
+
+	/* Disable zero-loops. */
+
+#if XCHAL_HAVE_LOOPS
+	wsr	a0, LCOUNT
+#endif
+
+	/* Disable all timers. */
+
+	.macro	reset_timer	num
+	wsr	a0, CCOMPARE_0 + \num
+	.endm
+	iterate	0, XCHAL_NUM_TIMERS-1, reset_timer
+
+	/* Interrupt initialization. */
+
+	movi	a2, XCHAL_INTTYPE_MASK_SOFTWARE | XCHAL_INTTYPE_MASK_EXTERN_EDGE
+	wsr	a0, INTENABLE
+	wsr	a2, INTCLEAR
+
+	/* Disable coprocessors. */
+
+#if XCHAL_CP_NUM > 0
+	wsr	a0, CPENABLE
+#endif
+
+	/* Set PS.INTLEVEL=1, PS.WOE=0, kernel stack, PS.EXCM=0
+	 *
+	 * Note: PS.EXCM must be cleared before using any loop
+	 *	 instructions; otherwise, they are silently disabled, and
+	 * 	 at most one iteration of the loop is executed.
+	 */
+
+	movi	a1, 1
+	wsr	a1, PS
+	rsync
+
+	/*  Initialize the caches.
+	 *  Does not include flushing writeback d-cache.
+	 *  a6, a7 are just working registers (clobbered).
+	 */
+
+	icache_reset  a2, a3
+	dcache_reset  a2, a3
+
+	/* Unpack data sections
+	 *
+	 * The linker script used to build the Linux kernel image
+	 * creates a table located at __boot_reloc_table_start
+	 * that contans the information what data needs to be unpacked.
+	 *
+	 * Uses a2-a7.
+	 */
+
+	movi	a2, __boot_reloc_table_start
+	movi	a3, __boot_reloc_table_end
+
+1:	beq	a2, a3, 3f	# no more entries?
+	l32i	a4, a2, 0	# start destination (in RAM)
+	l32i	a5, a2, 4	# end desination (in RAM)
+	l32i	a6, a2, 8	# start source (in ROM)
+	addi	a2, a2, 12	# next entry
+	beq	a4, a5, 1b	# skip, empty entry
+	beq	a4, a6, 1b	# skip, source and dest. are the same
+
+2:	l32i	a7, a6, 0	# load word
+	addi	a6, a6, 4
+	s32i	a7, a4, 0	# store word
+	addi	a4, a4, 4
+	bltu	a4, a5, 2b
+	j	1b
+
+3:
+	/* All code and initialized data segments have been copied.
+	 * Now clear the BSS segment.
+	 */
+
+	movi	a2, _bss_start	# start of BSS
+	movi	a3, _bss_end	# end of BSS
+
+1:	addi	a2, a2, 4
+	s32i	a0, a2, 0
+	blt	a2, a3, 1b
+
+#if XCHAL_DCACHE_IS_WRITEBACK
+
+	/* After unpacking, flush the writeback cache to memory so the
+	 * instructions/data are available.
+	 */
+
+	dcache_writeback_all	a2, a3
+#endif
+
+	/* Setup stack and enable window exceptions (keep irqs disabled) */
+
+	movi	a1, init_thread_union
+	addi	a1, a1, KERNEL_STACK_SIZE
+
+	movi	a2, 0x00040001		# WOE=1, INTLEVEL=1, UM=0
+	wsr	a2, PS			# (enable reg-windows; progmode stack)
+	rsync
+
+	/* Set up EXCSAVE[DEBUGLEVEL] to point to the Debug Exception Handler.*/
+
+	movi	a2, debug_exception
+	wsr	a2, EXCSAVE + XCHAL_DEBUGLEVEL
+
+	/* Set up EXCSAVE[1] to point to the exc_table. */
+
+	movi	a6, exc_table
+	xsr	a6, EXCSAVE_1
+
+	/* init_arch kick-starts the linux kernel */
+
+	movi	a4, init_arch
+	callx4	a4
+
+	movi	a4, start_kernel
+	callx4	a4
+
+should_never_return:
+	j	should_never_return
+
+	/* Define some common data structures here.  We define them
+	 * here in this assembly file due to their unusual alignment
+	 * requirements.
+	 */
+
+	.comm   swapper_pg_dir,PAGE_SIZE,PAGE_SIZE
+	.comm	empty_bad_page_table,PAGE_SIZE,PAGE_SIZE
+	.comm	empty_bad_page,PAGE_SIZE,PAGE_SIZE
+	.comm	empty_zero_page,PAGE_SIZE,PAGE_SIZE
+

arch/xtensa/kernel/irq.c

+/*
+ * linux/arch/xtensa/kernel/irq.c
+ *
+ * Xtensa built-in interrupt controller and some generic functions copied
+ * from i386.
+ *
+ * Copyright (C) 2002 - 2005 Tensilica, Inc.
+ * Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar
+ *
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Kevin Chea
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/seq_file.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel_stat.h>
+
+#include <asm/uaccess.h>
+#include <asm/platform.h>
+
+static void enable_xtensa_irq(unsigned int irq);
+static void disable_xtensa_irq(unsigned int irq);
+static void mask_and_ack_xtensa(unsigned int irq);
+static void end_xtensa_irq(unsigned int irq);
+
+static unsigned int cached_irq_mask;
+
+atomic_t irq_err_count;
+
+/*
+ * 'what should we do if we get a hw irq event on an illegal vector'.
+ * each architecture has to answer this themselves.
+ */
+void ack_bad_irq(unsigned int irq)
+{
+          printk("unexpected IRQ trap at vector %02x\n", irq);
+}
+
+/*
+ * do_IRQ handles all normal device IRQ's (the special
+ * SMP cross-CPU interrupts have their own specific
+ * handlers).
+ */
+
+unsigned int  do_IRQ(int irq, struct pt_regs *regs)
+{
+	irq_enter();
+
+#ifdef CONFIG_DEBUG_STACKOVERFLOW
+	/* Debugging check for stack overflow: is there less than 1KB free? */
+	{
+		unsigned long sp;
+
+		__asm__ __volatile__ ("mov %0, a1\n" : "=a" (sp));
+		sp &= THREAD_SIZE - 1;
+
+		if (unlikely(sp < (sizeof(thread_info) + 1024)))
+			printk("Stack overflow in do_IRQ: %ld\n",
+			       sp - sizeof(struct thread_info));
+	}
+#endif
+
+	__do_IRQ(irq, regs);
+
+	irq_exit();
+
+	return 1;
+}
+
+/*
+ * Generic, controller-independent functions:
+ */
+
+int show_interrupts(struct seq_file *p, void *v)
+{
+	int i = *(loff_t *) v, j;
+	struct irqaction * action;
+	unsigned long flags;
+
+	if (i == 0) {
+		seq_printf(p, "           ");
+		for (j=0; j<NR_CPUS; j++)
+			if (cpu_online(j))
+				seq_printf(p, "CPU%d       ",j);
+		seq_putc(p, '\n');
+	}
+
+	if (i < NR_IRQS) {
+		spin_lock_irqsave(&irq_desc[i].lock, flags);
+		action = irq_desc[i].action;
+		if (!action)
+			goto skip;
+		seq_printf(p, "%3d: ",i);
+#ifndef CONFIG_SMP
+		seq_printf(p, "%10u ", kstat_irqs(i));
+#else
+		for (j = 0; j < NR_CPUS; j++)
+			if (cpu_online(j))
+				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+#endif
+		seq_printf(p, " %14s", irq_desc[i].handler->typename);
+		seq_printf(p, "  %s", action->name);
+
+		for (action=action->next; action; action = action->next)
+			seq_printf(p, ", %s", action->name);
+
+		seq_putc(p, '\n');
+skip:
+		spin_unlock_irqrestore(&irq_desc[i].lock, flags);
+	} else if (i == NR_IRQS) {
+		seq_printf(p, "NMI: ");
+		for (j = 0; j < NR_CPUS; j++)
+			if (cpu_online(j))
+				seq_printf(p, "%10u ", nmi_count(j));
+		seq_putc(p, '\n');
+		seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
+	}
+	return 0;
+}
+/* shutdown is same as "disable" */
+#define shutdown_xtensa_irq disable_xtensa_irq
+
+static unsigned int startup_xtensa_irq(unsigned int irq)
+{
+	enable_xtensa_irq(irq);
+	return 0;               /* never anything pending */
+}
+
+static struct hw_interrupt_type xtensa_irq_type = {
+	"Xtensa-IRQ",
+	startup_xtensa_irq,
+	shutdown_xtensa_irq,
+	enable_xtensa_irq,
+	disable_xtensa_irq,
+	mask_and_ack_xtensa,
+	end_xtensa_irq
+};
+
+static inline void mask_irq(unsigned int irq)
+{
+	cached_irq_mask &= ~(1 << irq);
+	set_sr (cached_irq_mask, INTENABLE);
+}
+
+static inline void unmask_irq(unsigned int irq)
+{
+	cached_irq_mask |= 1 << irq;
+	set_sr (cached_irq_mask, INTENABLE);
+}
+
+static void disable_xtensa_irq(unsigned int irq)
+{
+	unsigned long flags;
+	local_save_flags(flags);
+	mask_irq(irq);
+	local_irq_restore(flags);
+}
+
+static void enable_xtensa_irq(unsigned int irq)
+{
+	unsigned long flags;
+	local_save_flags(flags);
+	unmask_irq(irq);
+	local_irq_restore(flags);
+}
+
+static void mask_and_ack_xtensa(unsigned int irq)
+{
+        disable_xtensa_irq(irq);
+}
+
+static void end_xtensa_irq(unsigned int irq)
+{
+        enable_xtensa_irq(irq);
+}
+
+
+void __init init_IRQ(void)
+{
+	int i;
+
+	for (i=0; i < XTENSA_NR_IRQS; i++)
+		irq_desc[i].handler = &xtensa_irq_type;
+
+	cached_irq_mask = 0;
+
+	platform_init_irq();
+}

arch/xtensa/kernel/module.c

+/*
+ * arch/xtensa/kernel/platform.c
+ *
+ * Module support.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/moduleloader.h>
+#include <linux/elf.h>
+#include <linux/vmalloc.h>
+#include <linux/fs.h>
+#include <linux/string.h>
+#include <linux/kernel.h>
+#include <linux/cache.h>
+
+LIST_HEAD(module_buf_list);
+
+void *module_alloc(unsigned long size)
+{
+  panic("module_alloc not implemented");
+}
+
+void module_free(struct module *mod, void *module_region)
+{
+  panic("module_free not implemented");
+}
+
+int module_frob_arch_sections(Elf32_Ehdr *hdr,
+    			      Elf32_Shdr *sechdrs,
+			      char *secstrings,
+			      struct module *me)
+{
+  panic("module_frob_arch_sections not implemented");
+}
+
+int apply_relocate(Elf32_Shdr *sechdrs,
+    		   const char *strtab,
+		   unsigned int symindex,
+		   unsigned int relsec,
+		   struct module *module)
+{
+  panic ("apply_relocate not implemented");
+}
+
+int apply_relocate_add(Elf32_Shdr *sechdrs,
+		       const char *strtab,
+		       unsigned int symindex,
+		       unsigned int relsec,
+		       struct module *module)
+{
+  panic("apply_relocate_add not implemented");
+}
+
+int module_finalize(const Elf_Ehdr *hdr,
+    		    const Elf_Shdr *sechdrs,
+		    struct module *me)
+{
+  panic ("module_finalize not implemented");
+}
+
+void module_arch_cleanup(struct module *mod)
+{
+  panic("module_arch_cleanup not implemented");
+}
+
+struct bug_entry *module_find_bug(unsigned long bugaddr)
+{
+  panic("module_find_bug not implemented");
+}

arch/xtensa/kernel/pci-dma.c

+/*
+ * arch/xtensa/pci-dma.c
+ *
+ * DMA coherent memory allocation.
+ *
+ * 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 of the  License, or (at your
+ * option) any later version.
+ *
+ * Copyright (C) 2002 - 2005 Tensilica Inc.
+ *
+ * Based on version for i386.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ */
+
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/string.h>
+#include <linux/pci.h>
+#include <asm/io.h>
+#include <asm/cacheflush.h>
+
+/*
+ * Note: We assume that the full memory space is always mapped to 'kseg'
+ *	 Otherwise we have to use page attributes (not implemented).
+ */
+
+void *
+dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *handle, int gfp)
+{
+	void *ret;
+
+	/* ignore region speicifiers */
+	gfp &= ~(__GFP_DMA | __GFP_HIGHMEM);
+
+	if (dev == NULL || (*dev->dma_mask < 0xffffffff))
+		gfp |= GFP_DMA;
+	ret = (void *)__get_free_pages(gfp, get_order(size));
+
+	if (ret != NULL) {
+		memset(ret, 0, size);
+		*handle = virt_to_bus(ret);
+	}
+	return (void*) BYPASS_ADDR((unsigned long)ret);
+}
+
+void dma_free_coherent(struct device *hwdev, size_t size,
+			 void *vaddr, dma_addr_t dma_handle)
+{
+	free_pages(CACHED_ADDR((unsigned long)vaddr), get_order(size));
+}
+
+
+void consistent_sync(void *vaddr, size_t size, int direction)
+{
+	switch (direction) {
+	case PCI_DMA_NONE:
+		BUG();
+	case PCI_DMA_FROMDEVICE:        /* invalidate only */
+		__invalidate_dcache_range((unsigned long)vaddr,
+				          (unsigned long)size);
+		break;
+
+	case PCI_DMA_TODEVICE:          /* writeback only */
+	case PCI_DMA_BIDIRECTIONAL:     /* writeback and invalidate */
+		__flush_invalidate_dcache_range((unsigned long)vaddr,
+				    		(unsigned long)size);
+		break;
+	}
+}

arch/xtensa/kernel/pci.c

+/*
+ * arch/xtensa/pcibios.c
+ *
+ * PCI bios-type initialisation for PCI machines
+ *
+ * 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 of the  License, or (at your
+ * option) any later version.
+ *
+ * Copyright (C) 2001-2005 Tensilica Inc.
+ *
+ * Based largely on work from Cort (ppc/kernel/pci.c)
+ * IO functions copied from sparc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ *
+ */
+
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/string.h>
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/errno.h>
+#include <linux/bootmem.h>
+
+#include <asm/pci-bridge.h>
+#include <asm/platform.h>
+
+#undef DEBUG
+
+#ifdef DEBUG
+#define DBG(x...) printk(x)
+#else
+#define DBG(x...)
+#endif
+
+/* PCI Controller */
+
+
+/*
+ * pcibios_alloc_controller
+ * pcibios_enable_device
+ * pcibios_fixups
+ * pcibios_align_resource
+ * pcibios_fixup_bus
+ * pcibios_setup
+ * pci_bus_add_device
+ * pci_mmap_page_range
+ */
+
+struct pci_controller* pci_ctrl_head;
+struct pci_controller** pci_ctrl_tail = &pci_ctrl_head;
+
+static int pci_bus_count;
+
+static void pcibios_fixup_resources(struct pci_dev* dev);
+
+#if 0 // FIXME
+struct pci_fixup pcibios_fixups[] = {
+	{ DECLARE_PCI_FIXUP_HEADER, PCI_ANY_ID, PCI_ANY_ID, pcibios_fixup_resources },
+	{ 0 }
+};
+#endif
+
+void
+pcibios_update_resource(struct pci_dev *dev, struct resource *root,
+			struct resource *res, int resource)
+{
+	u32 new, check, mask;
+	int reg;
+	struct pci_controller* pci_ctrl = dev->sysdata;
+
+	new = res->start;
+	if (pci_ctrl && res->flags & IORESOURCE_IO) {
+		new -= pci_ctrl->io_space.base;
+	}
+	new |= (res->flags & PCI_REGION_FLAG_MASK);
+	if (resource < 6) {
+		reg = PCI_BASE_ADDRESS_0 + 4*resource;
+	} else if (resource == PCI_ROM_RESOURCE) {
+		res->flags |= PCI_ROM_ADDRESS_ENABLE;
+		reg = dev->rom_base_reg;
+	} else {
+	/* Somebody might have asked allocation of a non-standard resource */
+		return;
+	}
+
+	pci_write_config_dword(dev, reg, new);
+	pci_read_config_dword(dev, reg, &check);
+	mask = (new & PCI_BASE_ADDRESS_SPACE_IO) ?
+		PCI_BASE_ADDRESS_IO_MASK : PCI_BASE_ADDRESS_MEM_MASK;
+
+	if ((new ^ check) & mask) {
+		printk(KERN_ERR "PCI: Error while updating region "
+		       "%s/%d (%08x != %08x)\n", dev->slot_name, resource,
+		       new, check);
+	}
+}
+
+/*
+ * We need to avoid collisions with `mirrored' VGA ports
+ * and other strange ISA hardware, so we always want the
+ * addresses to be allocated in the 0x000-0x0ff region
+ * modulo 0x400.
+ *
+ * Why? Because some silly external IO cards only decode
+ * the low 10 bits of the IO address. The 0x00-0xff region
+ * is reserved for motherboard devices that decode all 16
+ * bits, so it's ok to allocate at, say, 0x2800-0x28ff,
+ * but we want to try to avoid allocating at 0x2900-0x2bff
+ * which might have be mirrored at 0x0100-0x03ff..
+ */
+void
+pcibios_align_resource(void *data, struct resource *res, unsigned long size,
+    		       unsigned long align)
+{
+	struct pci_dev *dev = data;
+
+	if (res->flags & IORESOURCE_IO) {
+		unsigned long start = res->start;
+
+		if (size > 0x100) {
+			printk(KERN_ERR "PCI: I/O Region %s/%d too large"
+			       " (%ld bytes)\n", dev->slot_name,
+			       dev->resource - res, size);
+		}
+
+		if (start & 0x300) {
+			start = (start + 0x3ff) & ~0x3ff;
+			res->start = start;
+		}
+	}
+}
+
+int
+pcibios_enable_resources(struct pci_dev *dev, int mask)
+{
+	u16 cmd, old_cmd;
+	int idx;
+	struct resource *r;
+
+	pci_read_config_word(dev, PCI_COMMAND, &cmd);
+	old_cmd = cmd;
+	for(idx=0; idx<6; idx++) {
+		r = &dev->resource[idx];
+		if (!r->start && r->end) {
+			printk (KERN_ERR "PCI: Device %s not available because "
+				"of resource collisions\n", dev->slot_name);
+			return -EINVAL;
+		}
+		if (r->flags & IORESOURCE_IO)
+			cmd |= PCI_COMMAND_IO;
+		if (r->flags & IORESOURCE_MEM)
+			cmd |= PCI_COMMAND_MEMORY;
+	}
+	if (dev->resource[PCI_ROM_RESOURCE].start)
+		cmd |= PCI_COMMAND_MEMORY;
+	if (cmd != old_cmd) {
+		printk("PCI: Enabling device %s (%04x -> %04x)\n",
+			dev->slot_name, old_cmd, cmd);
+		pci_write_config_word(dev, PCI_COMMAND, cmd);
+	}
+	return 0;
+}
+
+struct pci_controller * __init pcibios_alloc_controller(void)
+{
+	struct pci_controller *pci_ctrl;
+
+	pci_ctrl = (struct pci_controller *)alloc_bootmem(sizeof(*pci_ctrl));
+	memset(pci_ctrl, 0, sizeof(struct pci_controller));
+
+	*pci_ctrl_tail = pci_ctrl;
+	pci_ctrl_tail = &pci_ctrl->next;
+
+	return pci_ctrl;
+}
+
+static int __init pcibios_init(void)
+{
+	struct pci_controller *pci_ctrl;
+	struct pci_bus *bus;
+	int next_busno = 0, i;
+
+	printk("PCI: Probing PCI hardware\n");
+
+	/* Scan all of the recorded PCI controllers.  */
+	for (pci_ctrl = pci_ctrl_head; pci_ctrl; pci_ctrl = pci_ctrl->next) {
+		pci_ctrl->last_busno = 0xff;
+		bus = pci_scan_bus(pci_ctrl->first_busno, pci_ctrl->ops,
+				   pci_ctrl);
+		if (pci_ctrl->io_resource.flags) {
+			unsigned long offs;
+
+			offs = (unsigned long)pci_ctrl->io_space.base;
+			pci_ctrl->io_resource.start += offs;
+			pci_ctrl->io_resource.end += offs;
+			bus->resource[0] = &pci_ctrl->io_resource;
+		}
+		for (i = 0; i < 3; ++i)
+			if (pci_ctrl->mem_resources[i].flags)
+				bus->resource[i+1] =&pci_ctrl->mem_resources[i];
+		pci_ctrl->bus = bus;
+		pci_ctrl->last_busno = bus->subordinate;
+		if (next_busno <= pci_ctrl->last_busno)
+			next_busno = pci_ctrl->last_busno+1;
+	}
+	pci_bus_count = next_busno;
+
+	return platform_pcibios_fixup();
+}
+
+subsys_initcall(pcibios_init);
+
+void __init pcibios_fixup_bus(struct pci_bus *bus)
+{
+	struct pci_controller *pci_ctrl = bus->sysdata;
+	struct resource *res;
+	unsigned long io_offset;
+	int i;
+
+	io_offset = (unsigned long)pci_ctrl->io_space.base;
+	if (bus->parent == NULL) {
+		/* this is a host bridge - fill in its resources */
+		pci_ctrl->bus = bus;
+
+		bus->resource[0] = res = &pci_ctrl->io_resource;
+		if (!res->flags) {
+			if (io_offset)
+				printk (KERN_ERR "I/O resource not set for host"
+					" bridge %d\n", pci_ctrl->index);
+			res->start = 0;
+			res->end = IO_SPACE_LIMIT;
+			res->flags = IORESOURCE_IO;
+		}
+		res->start += io_offset;
+		res->end += io_offset;
+
+		for (i = 0; i < 3; i++) {
+			res = &pci_ctrl->mem_resources[i];
+			if (!res->flags) {
+				if (i > 0)
+					continue;
+				printk(KERN_ERR "Memory resource not set for "
+				       "host bridge %d\n", pci_ctrl->index);
+				res->start = 0;
+				res->end = ~0U;
+				res->flags = IORESOURCE_MEM;
+			}
+			bus->resource[i+1] = res;
+		}
+	} else {
+		/* This is a subordinate bridge */
+		pci_read_bridge_bases(bus);
+
+		for (i = 0; i < 4; i++) {
+			if ((res = bus->resource[i]) == NULL || !res->flags)
+				continue;
+			if (io_offset && (res->flags & IORESOURCE_IO)) {
+				res->start += io_offset;
+				res->end += io_offset;
+			}
+		}
+	}
+}
+
+char __init *pcibios_setup(char *str)
+{
+	return str;
+}
+
+/* the next one is stolen from the alpha port... */
+
+void __init
+pcibios_update_irq(struct pci_dev *dev, int irq)
+{
+	pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq);
+}
+
+int pcibios_enable_device(struct pci_dev *dev, int mask)
+{
+	u16 cmd, old_cmd;
+	int idx;
+	struct resource *r;
+
+	pci_read_config_word(dev, PCI_COMMAND, &cmd);
+	old_cmd = cmd;
+	for (idx=0; idx<6; idx++) {
+		r = &dev->resource[idx];
+		if (!r->start && r->end) {
+			printk(KERN_ERR "PCI: Device %s not available because "
+			       "of resource collisions\n", dev->slot_name);
+			return -EINVAL;
+		}
+		if (r->flags & IORESOURCE_IO)
+			cmd |= PCI_COMMAND_IO;
+		if (r->flags & IORESOURCE_MEM)
+			cmd |= PCI_COMMAND_MEMORY;
+	}
+	if (cmd != old_cmd) {
+		printk("PCI: Enabling device %s (%04x -> %04x)\n",
+		       dev->slot_name, old_cmd, cmd);
+		pci_write_config_word(dev, PCI_COMMAND, cmd);
+	}
+
+	return 0;
+}
+
+#ifdef CONFIG_PROC_FS
+
+/*
+ * Return the index of the PCI controller for device pdev.
+ */
+
+int
+pci_controller_num(struct pci_dev *dev)
+{
+	struct pci_controller *pci_ctrl = (struct pci_controller*) dev->sysdata;
+	return pci_ctrl->index;
+}
+
+#endif /* CONFIG_PROC_FS */
+
+
+static void
+pcibios_fixup_resources(struct pci_dev *dev)
+{
+	struct pci_controller* pci_ctrl = (struct pci_controller *)dev->sysdata;
+	int i;
+	unsigned long offset;
+
+	if (!pci_ctrl) {
+		printk(KERN_ERR "No pci_ctrl for PCI dev %s!\n",dev->slot_name);
+		return;
+	}
+	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
+		struct resource *res = dev->resource + i;
+		if (!res->start || !res->flags)
+			continue;
+		if (res->end == 0xffffffff) {
+			DBG("PCI:%s Resource %d [%08lx-%08lx] is unassigned\n",
+			    dev->slot_name, i, res->start, res->end);
+			res->end -= res->start;
+			res->start = 0;
+			continue;
+		}
+		offset = 0;
+		if (res->flags & IORESOURCE_IO)
+			offset = (unsigned long) pci_ctrl->io_space.base;
+		else if (res->flags & IORESOURCE_MEM)
+			offset = (unsigned long) pci_ctrl->mem_space.base;
+
+		if (offset != 0) {
+			res->start += offset;
+			res->end += offset;
+#ifdef DEBUG
+			printk("Fixup res %d (%lx) of dev %s: %lx -> %lx\n",
+			       i, res->flags, dev->slot_name,
+			       res->start - offset, res->start);
+#endif
+		}
+	}
+}
+
+/*
+ * Platform support for /proc/bus/pci/X/Y mmap()s,
+ * modelled on the sparc64 implementation by Dave Miller.
+ *  -- paulus.
+ */
+
+/*
+ * Adjust vm_pgoff of VMA such that it is the physical page offset
+ * corresponding to the 32-bit pci bus offset for DEV requested by the user.
+ *
+ * Basically, the user finds the base address for his device which he wishes
+ * to mmap.  They read the 32-bit value from the config space base register,
+ * add whatever PAGE_SIZE multiple offset they wish, and feed this into the
+ * offset parameter of mmap on /proc/bus/pci/XXX for that device.
+ *
+ * Returns negative error code on failure, zero on success.
+ */
+static __inline__ int
+__pci_mmap_make_offset(struct pci_dev *dev, struct vm_area_struct *vma,
+		       enum pci_mmap_state mmap_state)
+{
+	struct pci_controller *pci_ctrl = (struct pci_controller*) dev->sysdata;
+	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
+	unsigned long io_offset = 0;
+	int i, res_bit;
+
+	if (pci_ctrl == 0)
+		return -EINVAL;		/* should never happen */
+
+	/* If memory, add on the PCI bridge address offset */
+	if (mmap_state == pci_mmap_mem) {
+		res_bit = IORESOURCE_MEM;
+	} else {
+		io_offset = (unsigned long)pci_ctrl->io_space.base;
+		offset += io_offset;
+		res_bit = IORESOURCE_IO;
+	}
+
+	/*
+	 * Check that the offset requested corresponds to one of the
+	 * resources of the device.
+	 */
+	for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
+		struct resource *rp = &dev->resource[i];
+		int flags = rp->flags;
+
+		/* treat ROM as memory (should be already) */
+		if (i == PCI_ROM_RESOURCE)
+			flags |= IORESOURCE_MEM;
+
+		/* Active and same type? */
+		if ((flags & res_bit) == 0)
+			continue;
+
+		/* In the range of this resource? */
+		if (offset < (rp->start & PAGE_MASK) || offset > rp->end)
+			continue;
+
+		/* found it! construct the final physical address */
+		if (mmap_state == pci_mmap_io)
+			offset += pci_ctrl->io_space.start - io_offset;
+		vma->vm_pgoff = offset >> PAGE_SHIFT;
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+/*
+ * Set vm_flags of VMA, as appropriate for this architecture, for a pci device
+ * mapping.
+ */
+static __inline__ void
+__pci_mmap_set_flags(struct pci_dev *dev, struct vm_area_struct *vma,
+		     enum pci_mmap_state mmap_state)
+{
+	vma->vm_flags |= VM_SHM | VM_LOCKED | VM_IO;
+}
+
+/*
+ * Set vm_page_prot of VMA, as appropriate for this architecture, for a pci
+ * device mapping.
+ */
+static __inline__ void
+__pci_mmap_set_pgprot(struct pci_dev *dev, struct vm_area_struct *vma,
+		      enum pci_mmap_state mmap_state, int write_combine)
+{
+	int prot = pgprot_val(vma->vm_page_prot);
+
+	/* Set to write-through */
+	prot &= ~_PAGE_NO_CACHE;
+#if 0
+	if (!write_combine)
+		prot |= _PAGE_WRITETHRU;
+#endif
+	vma->vm_page_prot = __pgprot(prot);
+}
+
+/*
+ * Perform the actual remap of the pages for a PCI device mapping, as
+ * appropriate for this architecture.  The region in the process to map
+ * is described by vm_start and vm_end members of VMA, the base physical
+ * address is found in vm_pgoff.
+ * The pci device structure is provided so that architectures may make mapping
+ * decisions on a per-device or per-bus basis.
+ *
+ * Returns a negative error code on failure, zero on success.
+ */
+int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
+			enum pci_mmap_state mmap_state,
+			int write_combine)
+{
+	int ret;
+
+	ret = __pci_mmap_make_offset(dev, vma, mmap_state);
+	if (ret < 0)
+		return ret;
+
+	__pci_mmap_set_flags(dev, vma, mmap_state);
+	__pci_mmap_set_pgprot(dev, vma, mmap_state, write_combine);
+
+	ret = io_remap_page_range(vma, vma->vm_start, vma->vm_pgoff<<PAGE_SHIFT,
+			       vma->vm_end - vma->vm_start, vma->vm_page_prot);
+
+	return ret;
+}
+
+/*
+ * This probably belongs here rather than ioport.c because
+ * we do not want this crud linked into SBus kernels.
+ * Also, think for a moment about likes of floppy.c that
+ * include architecture specific parts. They may want to redefine ins/outs.
+ *
+ * We do not use horroble macroses here because we want to
+ * advance pointer by sizeof(size).
+ */
+void outsb(unsigned long addr, const void *src, unsigned long count) {
+        while (count) {
+                count -= 1;
+                writeb(*(const char *)src, addr);
+                src += 1;
+                addr += 1;
+        }
+}
+
+void outsw(unsigned long addr, const void *src, unsigned long count) {
+        while (count) {
+                count -= 2;
+                writew(*(const short *)src, addr);
+                src += 2;
+                addr += 2;
+        }
+}
+
+void outsl(unsigned long addr, const void *src, unsigned long count) {
+        while (count) {
+                count -= 4;
+                writel(*(const long *)src, addr);
+                src += 4;
+                addr += 4;
+        }
+}
+
+void insb(unsigned long addr, void *dst, unsigned long count) {
+        while (count) {
+                count -= 1;
+                *(unsigned char *)dst = readb(addr);
+                dst += 1;
+                addr += 1;
+        }
+}
+
+void insw(unsigned long addr, void *dst, unsigned long count) {
+        while (count) {
+                count -= 2;
+                *(unsigned short *)dst = readw(addr);
+                dst += 2;
+                addr += 2;
+        }
+}
+
+void insl(unsigned long addr, void *dst, unsigned long count) {
+        while (count) {
+                count -= 4;
+                /*
+                 * XXX I am sure we are in for an unaligned trap here.
+                 */
+                *(unsigned long *)dst = readl(addr);
+                dst += 4;
+                addr += 4;
+        }
+}
+
+
+

arch/xtensa/kernel/platform.c

+/*
+ * arch/xtensa/kernel/platform.c
+ *
+ * Default platform functions.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ */
+
+#include <linux/config.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/time.h>
+#include <asm/platform.h>
+#include <asm/timex.h>
+
+#define _F(r,f,a,b)							\
+	r __platform_##f a b;                                   	\
+	r platform_##f a __attribute__((weak, alias("__platform_"#f)))
+
+/*
+ * Default functions that are used if no platform specific function is defined.
+ * (Please, refer to include/asm-xtensa/platform.h for more information)
+ */
+
+_F(void, setup, (char** cmd), { });
+_F(void, init_irq, (void), { });
+_F(void, restart, (void), { while(1); });
+_F(void, halt, (void), { while(1); });
+_F(void, power_off, (void), { while(1); });
+_F(void, idle, (void), { __asm__ __volatile__ ("waiti 0" ::: "memory"); });
+_F(void, heartbeat, (void), { });
+_F(int,  pcibios_fixup, (void), { return 0; });
+_F(int, get_rtc_time, (time_t* t), { return 0; });
+_F(int, set_rtc_time, (time_t t), { return 0; });
+
+#if CONFIG_XTENSA_CALIBRATE_CCOUNT
+_F(void, calibrate_ccount, (void),
+{
+  printk ("ERROR: Cannot calibrate cpu frequency! Assuming 100MHz.\n");
+  ccount_per_jiffy = 100 * (1000000UL/HZ);
+});
+#endif
+

arch/xtensa/kernel/process.c

+// TODO	verify coprocessor handling
+/*
+ * arch/xtensa/kernel/process.c
+ *
+ * Xtensa Processor version.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * Chris Zankel <chris@zankel.net>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Kevin Chea
+ */
+
+#include <linux/config.h>
+#include <linux/errno.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/stddef.h>
+#include <linux/unistd.h>
+#include <linux/ptrace.h>
+#include <linux/slab.h>
+#include <linux/elf.h>
+#include <linux/init.h>
+#include <linux/prctl.h>
+#include <linux/init_task.h>
+#include <linux/module.h>
+#include <linux/mqueue.h>
+
+#include <asm/pgtable.h>
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/io.h>
+#include <asm/processor.h>
+#include <asm/platform.h>
+#include <asm/mmu.h>
+#include <asm/irq.h>
+#include <asm/atomic.h>
+#include <asm/offsets.h>
+#include <asm/coprocessor.h>
+
+extern void ret_from_fork(void);
+
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS(init_signals);
+static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand);
+struct mm_struct init_mm = INIT_MM(init_mm);
+EXPORT_SYMBOL(init_mm);
+
+union thread_union init_thread_union
+	__attribute__((__section__(".data.init_task"))) =
+{ INIT_THREAD_INFO(init_task) };
+
+struct task_struct init_task = INIT_TASK(init_task);
+EXPORT_SYMBOL(init_task);
+
+struct task_struct *current_set[NR_CPUS] = {&init_task, };
+
+
+#if XCHAL_CP_NUM > 0
+
+/*
+ * Coprocessor ownership.
+ */
+
+coprocessor_info_t coprocessor_info[] = {
+	{ 0, XTENSA_CPE_CP0_OFFSET },
+	{ 0, XTENSA_CPE_CP1_OFFSET },
+	{ 0, XTENSA_CPE_CP2_OFFSET },
+	{ 0, XTENSA_CPE_CP3_OFFSET },
+	{ 0, XTENSA_CPE_CP4_OFFSET },
+	{ 0, XTENSA_CPE_CP5_OFFSET },
+	{ 0, XTENSA_CPE_CP6_OFFSET },
+	{ 0, XTENSA_CPE_CP7_OFFSET },
+};
+
+#endif
+
+/*
+ * Powermanagement idle function, if any is provided by the platform.
+ */
+
+void cpu_idle(void)
+{
+  	local_irq_enable();
+
+	/* endless idle loop with no priority at all */
+	while (1) {
+		while (!need_resched())
+			platform_idle();
+		preempt_enable();
+		schedule();
+	}
+}
+
+/*
+ * Free current thread data structures etc..
+ */
+
+void exit_thread(void)
+{
+	release_coprocessors(current);	/* Empty macro if no CPs are defined */
+}
+
+void flush_thread(void)
+{
+	release_coprocessors(current);	/* Empty macro if no CPs are defined */
+}
+
+/*
+ * Copy thread.
+ *
+ * The stack layout for the new thread looks like this:
+ *
+ *	+------------------------+ <- sp in childregs (= tos)
+ *	|       childregs        |
+ *	+------------------------+ <- thread.sp = sp in dummy-frame
+ *	|      dummy-frame       |    (saved in dummy-frame spill-area)
+ *	+------------------------+
+ *
+ * We create a dummy frame to return to ret_from_fork:
+ *   a0 points to ret_from_fork (simulating a call4)
+ *   sp points to itself (thread.sp)
+ *   a2, a3 are unused.
+ *
+ * Note: This is a pristine frame, so we don't need any spill region on top of
+ *       childregs.
+ */
+
+int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
+		unsigned long unused,
+                struct task_struct * p, struct pt_regs * regs)
+{
+	struct pt_regs *childregs;
+	unsigned long tos;
+	int user_mode = user_mode(regs);
+
+	/* Set up new TSS. */
+	tos = (unsigned long)p->thread_info + THREAD_SIZE;
+	if (user_mode)
+		childregs = (struct pt_regs*)(tos - PT_USER_SIZE);
+	else
+		childregs = (struct pt_regs*)tos - 1;
+
+	*childregs = *regs;
+
+	/* Create a call4 dummy-frame: a0 = 0, a1 = childregs. */
+	*((int*)childregs - 3) = (unsigned long)childregs;
+	*((int*)childregs - 4) = 0;
+
+	childregs->areg[1] = tos;
+	childregs->areg[2] = 0;
+	p->set_child_tid = p->clear_child_tid = NULL;
+	p->thread.ra = MAKE_RA_FOR_CALL((unsigned long)ret_from_fork, 0x1);
+	p->thread.sp = (unsigned long)childregs;
+	if (user_mode(regs)) {
+
+		int len = childregs->wmask & ~0xf;
+		childregs->areg[1] = usp;
+		memcpy(&childregs->areg[XCHAL_NUM_AREGS - len/4],
+		       &regs->areg[XCHAL_NUM_AREGS - len/4], len);
+
+		if (clone_flags & CLONE_SETTLS)
+			childregs->areg[2] = childregs->areg[6];
+
+	} else {
+		/* In kernel space, we start a new thread with a new stack. */
+		childregs->wmask = 1;
+	}
+	return 0;
+}
+
+
+/*
+ * Create a kernel thread
+ */
+
+int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
+{
+	long retval;
+	__asm__ __volatile__
+		("mov           a5, %4\n\t" /* preserve fn in a5 */
+		 "mov           a6, %3\n\t" /* preserve and setup arg in a6 */
+		 "movi		a2, %1\n\t" /* load __NR_clone for syscall*/
+		 "mov		a3, sp\n\t" /* sp check and sys_clone */
+		 "mov		a4, %5\n\t" /* load flags for syscall */
+		 "syscall\n\t"
+		 "beq		a3, sp, 1f\n\t" /* branch if parent */
+		 "callx4	a5\n\t"     /* call fn */
+		 "movi		a2, %2\n\t" /* load __NR_exit for syscall */
+		 "mov		a3, a6\n\t" /* load fn return value */
+		 "syscall\n"
+		 "1:\n\t"
+		 "mov		%0, a2\n\t" /* parent returns zero */
+		 :"=r" (retval)
+		 :"i" (__NR_clone), "i" (__NR_exit),
+		 "r" (arg), "r" (fn),
+		 "r" (flags | CLONE_VM)
+		 : "a2", "a3", "a4", "a5", "a6" );
+	return retval;
+}
+
+
+/*
+ * These bracket the sleeping functions..
+ */
+
+unsigned long get_wchan(struct task_struct *p)
+{
+	unsigned long sp, pc;
+	unsigned long stack_page = (unsigned long) p->thread_info;
+	int count = 0;
+
+	if (!p || p == current || p->state == TASK_RUNNING)
+		return 0;
+
+	sp = p->thread.sp;
+	pc = MAKE_PC_FROM_RA(p->thread.ra, p->thread.sp);
+
+	do {
+		if (sp < stack_page + sizeof(struct task_struct) ||
+		    sp >= (stack_page + THREAD_SIZE) ||
+		    pc == 0)
+			return 0;
+		if (!in_sched_functions(pc))
+			return pc;
+
+		/* Stack layout: sp-4: ra, sp-3: sp' */
+
+		pc = MAKE_PC_FROM_RA(*(unsigned long*)sp - 4, sp);
+		sp = *(unsigned long *)sp - 3;
+	} while (count++ < 16);
+	return 0;
+}
+
+/*
+ * do_copy_regs() gathers information from 'struct pt_regs' and
+ * 'current->thread.areg[]' to fill in the xtensa_gregset_t
+ * structure.
+ *
+ * xtensa_gregset_t and 'struct pt_regs' are vastly different formats
+ * of processor registers.  Besides different ordering,
+ * xtensa_gregset_t contains non-live register information that
+ * 'struct pt_regs' does not.  Exception handling (primarily) uses
+ * 'struct pt_regs'.  Core files and ptrace use xtensa_gregset_t.
+ *
+ */
+
+void do_copy_regs (xtensa_gregset_t *elfregs, struct pt_regs *regs,
+		   struct task_struct *tsk)
+{
+	int i, n, wb_offset;
+
+	elfregs->xchal_config_id0 = XCHAL_HW_CONFIGID0;
+	elfregs->xchal_config_id1 = XCHAL_HW_CONFIGID1;
+
+	__asm__ __volatile__ ("rsr  %0, 176\n" : "=a" (i));
+ 	elfregs->cpux = i;
+	__asm__ __volatile__ ("rsr  %0, 208\n" : "=a" (i));
+ 	elfregs->cpuy = i;
+
+	/* Note:  PS.EXCM is not set while user task is running; its
+	 * being set in regs->ps is for exception handling convenience.
+	 */
+
+	elfregs->pc		= regs->pc;
+	elfregs->ps		= (regs->ps & ~XCHAL_PS_EXCM_MASK);
+	elfregs->exccause	= regs->exccause;
+	elfregs->excvaddr	= regs->excvaddr;
+	elfregs->windowbase	= regs->windowbase;
+	elfregs->windowstart	= regs->windowstart;
+	elfregs->lbeg		= regs->lbeg;
+	elfregs->lend		= regs->lend;
+	elfregs->lcount		= regs->lcount;
+	elfregs->sar		= regs->sar;
+	elfregs->syscall	= regs->syscall;
+
+	/* Copy register file.
+	 * The layout looks like this:
+	 *
+	 * |  a0 ... a15  | Z ... Z |  arX ... arY  |
+	 *  current window  unused    saved frames
+	 */
+
+	memset (elfregs->ar, 0, sizeof(elfregs->ar));
+
+	wb_offset = regs->windowbase * 4;
+	n = (regs->wmask&1)? 4 : (regs->wmask&2)? 8 : (regs->wmask&4)? 12 : 16;
+
+	for (i = 0; i < n; i++)
+		elfregs->ar[(wb_offset + i) % XCHAL_NUM_AREGS] = regs->areg[i];
+
+	n = (regs->wmask >> 4) * 4;
+
+	for (i = XCHAL_NUM_AREGS - n; n > 0; i++, n--)
+		elfregs->ar[(wb_offset + i) % XCHAL_NUM_AREGS] = regs->areg[i];
+}
+
+void xtensa_elf_core_copy_regs (xtensa_gregset_t *elfregs, struct pt_regs *regs)
+{
+	do_copy_regs ((xtensa_gregset_t *)elfregs, regs, current);
+}
+
+
+/* The inverse of do_copy_regs().  No error or sanity checking. */
+
+void do_restore_regs (xtensa_gregset_t *elfregs, struct pt_regs *regs,
+		      struct task_struct *tsk)
+{
+	int i, n, wb_offset;
+
+	/* Note:  PS.EXCM is not set while user task is running; it
+	 * needs to be set in regs->ps is for exception handling convenience.
+	 */
+
+	regs->pc		= elfregs->pc;
+	regs->ps		= (elfregs->ps | XCHAL_PS_EXCM_MASK);
+	regs->exccause		= elfregs->exccause;
+	regs->excvaddr		= elfregs->excvaddr;
+	regs->windowbase	= elfregs->windowbase;
+	regs->windowstart	= elfregs->windowstart;
+	regs->lbeg		= elfregs->lbeg;
+	regs->lend		= elfregs->lend;
+	regs->lcount		= elfregs->lcount;
+	regs->sar		= elfregs->sar;
+	regs->syscall	= elfregs->syscall;
+
+	/* Clear everything. */
+
+	memset (regs->areg, 0, sizeof(regs->areg));
+
+	/* Copy regs from live window frame. */
+
+	wb_offset = regs->windowbase * 4;
+	n = (regs->wmask&1)? 4 : (regs->wmask&2)? 8 : (regs->wmask&4)? 12 : 16;
+
+	for (i = 0; i < n; i++)
+		regs->areg[(wb_offset+i) % XCHAL_NUM_AREGS] = elfregs->ar[i];
+
+	n = (regs->wmask >> 4) * 4;
+
+	for (i = XCHAL_NUM_AREGS - n; n > 0; i++, n--)
+		regs->areg[(wb_offset+i) % XCHAL_NUM_AREGS] = elfregs->ar[i];
+}
+
+/*
+ * do_save_fpregs() gathers information from 'struct pt_regs' and
+ * 'current->thread' to fill in the elf_fpregset_t structure.
+ *
+ * Core files and ptrace use elf_fpregset_t.
+ */
+
+void do_save_fpregs (elf_fpregset_t *fpregs, struct pt_regs *regs,
+		     struct task_struct *tsk)
+{
+#if XCHAL_HAVE_CP
+
+	extern unsigned char	_xtensa_reginfo_tables[];
+	extern unsigned		_xtensa_reginfo_table_size;
+	int i;
+	unsigned long flags;
+
+	/* Before dumping coprocessor state from memory,
+	 * ensure any live coprocessor contents for this
+	 * task are first saved to memory:
+	 */
+	local_irq_save(flags);
+
+	for (i = 0; i < XCHAL_CP_MAX; i++) {
+		if (tsk == coprocessor_info[i].owner) {
+			enable_coprocessor(i);
+			save_coprocessor_registers(
+			    tsk->thread.cp_save+coprocessor_info[i].offset,i);
+			disable_coprocessor(i);
+		}
+	}
+
+	local_irq_restore(flags);
+
+	/* Now dump coprocessor & extra state: */
+	memcpy((unsigned char*)fpregs,
+		_xtensa_reginfo_tables, _xtensa_reginfo_table_size);
+	memcpy((unsigned char*)fpregs + _xtensa_reginfo_table_size,
+		tsk->thread.cp_save, XTENSA_CP_EXTRA_SIZE);
+#endif
+}
+
+/*
+ * The inverse of do_save_fpregs().
+ * Copies coprocessor and extra state from fpregs into regs and tsk->thread.
+ * Returns 0 on success, non-zero if layout doesn't match.
+ */
+
+int  do_restore_fpregs (elf_fpregset_t *fpregs, struct pt_regs *regs,
+		        struct task_struct *tsk)
+{
+#if XCHAL_HAVE_CP
+
+	extern unsigned char	_xtensa_reginfo_tables[];
+	extern unsigned		_xtensa_reginfo_table_size;
+	int i;
+	unsigned long flags;
+
+	/* Make sure save area layouts match.
+	 * FIXME:  in the future we could allow restoring from
+	 * a different layout of the same registers, by comparing
+	 * fpregs' table with _xtensa_reginfo_tables and matching
+	 * entries and copying registers one at a time.
+	 * Not too sure yet whether that's very useful.
+	 */
+
+	if( memcmp((unsigned char*)fpregs,
+		_xtensa_reginfo_tables, _xtensa_reginfo_table_size) ) {
+	    return -1;
+	}
+
+	/* Before restoring coprocessor state from memory,
+	 * ensure any live coprocessor contents for this
+	 * task are first invalidated.
+	 */
+
+	local_irq_save(flags);
+
+	for (i = 0; i < XCHAL_CP_MAX; i++) {
+		if (tsk == coprocessor_info[i].owner) {
+			enable_coprocessor(i);
+			save_coprocessor_registers(
+			    tsk->thread.cp_save+coprocessor_info[i].offset,i);
+			coprocessor_info[i].owner = 0;
+			disable_coprocessor(i);
+		}
+	}
+
+	local_irq_restore(flags);
+
+	/*  Now restore coprocessor & extra state:  */
+
+	memcpy(tsk->thread.cp_save,
+		(unsigned char*)fpregs + _xtensa_reginfo_table_size,
+		XTENSA_CP_EXTRA_SIZE);
+#endif
+	return 0;
+}
+/*
+ * Fill in the CP structure for a core dump for a particular task.
+ */
+
+int
+dump_task_fpu(struct pt_regs *regs, struct task_struct *task, elf_fpregset_t *r)
+{
+/* see asm/coprocessor.h for this magic number 16 */
+#if TOTAL_CPEXTRA_SIZE > 16
+	do_save_fpregs (r, regs, task);
+
+	/*  For now, bit 16 means some extra state may be present:  */
+// FIXME!! need to track to return more accurate mask
+	return 0x10000 | XCHAL_CP_MASK;
+#else
+	return 0;	/* no coprocessors active on this processor */
+#endif
+}
+
+/*
+ * Fill in the CP structure for a core dump.
+ * This includes any FPU coprocessor.
+ * Here, we dump all coprocessors, and other ("extra") custom state.
+ *
+ * This function is called by elf_core_dump() in fs/binfmt_elf.c
+ * (in which case 'regs' comes from calls to do_coredump, see signals.c).
+ */
+int  dump_fpu(struct pt_regs *regs, elf_fpregset_t *r)
+{
+	return dump_task_fpu(regs, current, r);
+}

arch/xtensa/kernel/ptrace.c

+// TODO some minor issues
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005  Tensilica Inc.
+ *
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ * Chris Zankel <chris@zankel.net>
+ * Scott Foehner<sfoehner@yahoo.com>,
+ * Kevin Chea
+ * Marc Gauthier<marc@tensilica.com> <marc@alumni.uwaterloo.ca>
+ */
+
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/errno.h>
+#include <linux/ptrace.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/security.h>
+
+#include <asm/pgtable.h>
+#include <asm/page.h>
+#include <asm/system.h>
+#include <asm/uaccess.h>
+#include <asm/ptrace.h>
+#include <asm/elf.h>
+
+#define TEST_KERNEL	// verify kernel operations FIXME: remove
+
+
+/*
+ * Called by kernel/ptrace.c when detaching..
+ *
+ * Make sure single step bits etc are not set.
+ */
+
+void ptrace_disable(struct task_struct *child)
+{
+	/* Nothing to do.. */
+}
+
+int sys_ptrace(long request, long pid, long addr, long data)
+{
+	struct task_struct *child;
+	int ret = -EPERM;
+
+	lock_kernel();
+
+#if 0
+	if ((int)request != 1)
+	printk("ptrace(r=%d,pid=%d,addr=%08lx,data=%08lx)\n",
+	       (int) request, (int) pid, (unsigned long) addr,
+	       (unsigned long) data);
+#endif
+
+	if (request == PTRACE_TRACEME) {
+
+		/* Are we already being traced? */
+
+		if (current->ptrace & PT_PTRACED)
+			goto out;
+
+		if ((ret = security_ptrace(current->parent, current)))
+			goto out;
+
+		/* Set the ptrace bit in the process flags. */
+
+		current->ptrace |= PT_PTRACED;
+		ret = 0;
+		goto out;
+	}
+
+	ret = -ESRCH;
+	read_lock(&tasklist_lock);
+	child = find_task_by_pid(pid);
+	if (child)
+		get_task_struct(child);
+	read_unlock(&tasklist_lock);
+	if (!child)
+		goto out;
+
+	ret = -EPERM;
+	if (pid == 1)		/* you may not mess with init */
+		goto out;
+
+	if (request == PTRACE_ATTACH) {
+		ret = ptrace_attach(child);
+		goto out_tsk;
+	}
+
+	if ((ret = ptrace_check_attach(child, request == PTRACE_KILL)) < 0)
+		goto out_tsk;
+
+	switch (request) {
+	case PTRACE_PEEKTEXT: /* read word at location addr. */
+	case PTRACE_PEEKDATA:
+	{
+		unsigned long tmp;
+		int copied;
+
+		copied = access_process_vm(child, addr, &tmp, sizeof(tmp), 0);
+		ret = -EIO;
+		if (copied != sizeof(tmp))
+			break;
+		ret = put_user(tmp,(unsigned long *) data);
+
+		goto out;
+	}
+
+	/* Read the word at location addr in the USER area.  */
+
+	case PTRACE_PEEKUSR:
+		{
+		struct pt_regs *regs;
+		unsigned long tmp;
+
+		regs = xtensa_pt_regs(child);
+		tmp = 0;  /* Default return value. */
+
+		switch(addr) {
+
+		case REG_AR_BASE ... REG_AR_BASE + XCHAL_NUM_AREGS - 1:
+			{
+			int ar = addr - REG_AR_BASE - regs->windowbase * 4;
+			ar &= (XCHAL_NUM_AREGS - 1);
+			if (ar < 16 && ar + (regs->wmask >> 4) * 4 >= 0)
+				tmp = regs->areg[ar];
+			else
+				ret = -EIO;
+			break;
+			}
+		case REG_A_BASE ... REG_A_BASE + 15:
+			tmp = regs->areg[addr - REG_A_BASE];
+			break;
+		case REG_PC:
+			tmp = regs->pc;
+			break;
+		case REG_PS:
+			/* Note:  PS.EXCM is not set while user task is running;
+			 * its being set in regs is for exception handling
+			 * convenience.  */
+			tmp = (regs->ps & ~XCHAL_PS_EXCM_MASK);
+			break;
+		case REG_WB:
+			tmp = regs->windowbase;
+			break;
+		case REG_WS:
+			tmp = regs->windowstart;
+			break;
+		case REG_LBEG:
+			tmp = regs->lbeg;
+			break;
+		case REG_LEND:
+			tmp = regs->lend;
+			break;
+		case REG_LCOUNT:
+			tmp = regs->lcount;
+			break;
+		case REG_SAR:
+			tmp = regs->sar;
+			break;
+		case REG_DEPC:
+			tmp = regs->depc;
+			break;
+		case REG_EXCCAUSE:
+			tmp = regs->exccause;
+			break;
+		case REG_EXCVADDR:
+			tmp = regs->excvaddr;
+			break;
+		case SYSCALL_NR:
+			tmp = regs->syscall;
+			break;
+		default:
+			tmp = 0;
+			ret = -EIO;
+			goto out;
+		}
+		ret = put_user(tmp, (unsigned long *) data);
+		goto out;
+		}
+
+	case PTRACE_POKETEXT: /* write the word at location addr. */
+	case PTRACE_POKEDATA:
+		if (access_process_vm(child, addr, &data, sizeof(data), 1)
+		    == sizeof(data))
+			break;
+		ret = -EIO;
+		goto out;
+
+	case PTRACE_POKEUSR:
+		{
+		struct pt_regs *regs;
+		regs = xtensa_pt_regs(child);
+
+		switch (addr) {
+		case REG_AR_BASE ... REG_AR_BASE + XCHAL_NUM_AREGS - 1:
+			{
+			int ar = addr - REG_AR_BASE - regs->windowbase * 4;
+			if (ar < 16 && ar + (regs->wmask >> 4) * 4 >= 0)
+				regs->areg[ar & (XCHAL_NUM_AREGS - 1)] = data;
+			else
+				ret = -EIO;
+			break;
+			}
+		case REG_A_BASE ... REG_A_BASE + 15:
+			regs->areg[addr - REG_A_BASE] = data;
+			break;
+		case REG_PC:
+			regs->pc = data;
+			break;
+		case SYSCALL_NR:
+			regs->syscall = data;
+			break;
+#ifdef TEST_KERNEL
+		case REG_WB:
+			regs->windowbase = data;
+			break;
+		case REG_WS:
+			regs->windowstart = data;
+			break;
+#endif
+
+		default:
+			/* The rest are not allowed. */
+			ret = -EIO;
+			break;
+		}
+		break;
+		}
+
+	/* continue and stop at next (return from) syscall */
+	case PTRACE_SYSCALL:
+	case PTRACE_CONT: /* restart after signal. */
+	{
+		ret = -EIO;
+		if ((unsigned long) data > _NSIG)
+			break;
+		if (request == PTRACE_SYSCALL)
+			set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+		else
+			clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+		child->exit_code = data;
+		/* Make sure the single step bit is not set. */
+		child->ptrace &= ~PT_SINGLESTEP;
+		wake_up_process(child);
+		ret = 0;
+		break;
+	}
+
+	/*
+	 * make the child exit.  Best I can do is send it a sigkill.
+	 * perhaps it should be put in the status that it wants to
+	 * exit.
+	 */
+	case PTRACE_KILL:
+		ret = 0;
+		if (child->state == EXIT_ZOMBIE)	/* already dead */
+			break;
+		child->exit_code = SIGKILL;
+		child->ptrace &= ~PT_SINGLESTEP;
+		wake_up_process(child);
+		break;
+
+	case PTRACE_SINGLESTEP:
+		ret = -EIO;
+		if ((unsigned long) data > _NSIG)
+			break;
+		clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+		child->ptrace |= PT_SINGLESTEP;
+		child->exit_code = data;
+		wake_up_process(child);
+		ret = 0;
+		break;
+
+	case PTRACE_GETREGS:
+	{
+		/* 'data' points to user memory in which to write.
+		 * Mainly due to the non-live register values, we
+		 * reformat the register values into something more
+		 * standard.  For convenience, we use the handy
+		 * elf_gregset_t format. */
+
+		xtensa_gregset_t format;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		do_copy_regs (&format, regs, child);
+
+		/* Now, copy to user space nice and easy... */
+		ret = 0;
+		if (copy_to_user((void *)data, &format, sizeof(elf_gregset_t)))
+			ret = -EFAULT;
+		break;
+	}
+
+	case PTRACE_SETREGS:
+	{
+		/* 'data' points to user memory that contains the new
+		 * values in the elf_gregset_t format. */
+
+		xtensa_gregset_t format;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		if (copy_from_user(&format,(void *)data,sizeof(elf_gregset_t))){
+			ret = -EFAULT;
+			break;
+		}
+
+		/* FIXME: Perhaps we want some sanity checks on
+		 * these user-space values?  See ARM version.  Are
+		 * debuggers a security concern? */
+
+		do_restore_regs (&format, regs, child);
+
+		ret = 0;
+		break;
+	}
+
+	case PTRACE_GETFPREGS:
+	{
+		/* 'data' points to user memory in which to write.
+		 * For convenience, we use the handy
+		 * elf_fpregset_t format. */
+
+		elf_fpregset_t fpregs;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		do_save_fpregs (&fpregs, regs, child);
+
+		/* Now, copy to user space nice and easy... */
+		ret = 0;
+		if (copy_to_user((void *)data, &fpregs, sizeof(elf_fpregset_t)))
+			ret = -EFAULT;
+
+		break;
+	}
+
+	case PTRACE_SETFPREGS:
+	{
+		/* 'data' points to user memory that contains the new
+		 * values in the elf_fpregset_t format.
+		 */
+		elf_fpregset_t fpregs;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		ret = 0;
+		if (copy_from_user(&fpregs, (void *)data, sizeof(elf_fpregset_t))) {
+			ret = -EFAULT;
+			break;
+		}
+
+		if (do_restore_fpregs (&fpregs, regs, child))
+			ret = -EIO;
+		break;
+	}
+
+	case PTRACE_GETFPREGSIZE:
+		/* 'data' points to 'unsigned long' set to the size
+		 * of elf_fpregset_t
+		 */
+		ret = put_user(sizeof(elf_fpregset_t), (unsigned long *) data);
+		break;
+
+	case PTRACE_DETACH: /* detach a process that was attached. */
+		ret = ptrace_detach(child, data);
+		break;
+
+	default:
+		ret = ptrace_request(child, request, addr, data);
+		goto out;
+	}
+out_tsk:
+	put_task_struct(child);
+out:
+	unlock_kernel();
+	return ret;
+}
+
+void do_syscall_trace(void)
+{
+	if (!test_thread_flag(TIF_SYSCALL_TRACE))
+		return;
+
+	if (!(current->ptrace & PT_PTRACED))
+		return;
+
+	/*
+	 * The 0x80 provides a way for the tracing parent to distinguish
+	 * between a syscall stop and SIGTRAP delivery
+	 */
+	ptrace_notify(SIGTRAP|((current->ptrace & PT_TRACESYSGOOD) ? 0x80 : 0));
+
+	/*
+	 * this isn't the same as continuing with a signal, but it will do
+	 * for normal use.  strace only continues with a signal if the
+	 * stopping signal is not SIGTRAP.  -brl
+	 */
+	if (current->exit_code) {
+		send_sig(current->exit_code, current, 1);
+		current->exit_code = 0;
+	}
+}

arch/xtensa/kernel/xtensa_ksyms.c

+/*
+ * arch/xtensa/kernel/xtensa_ksyms.c
+ *
+ * Export Xtensa-specific functions for loadable modules.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005  Tensilica Inc.
+ *
+ * Joe Taylor <joe@tensilica.com>
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <asm/irq.h>
+#include <linux/in6.h>
+#include <linux/pci.h>
+#include <linux/ide.h>
+
+#include <asm/uaccess.h>
+#include <asm/checksum.h>
+#include <asm/dma.h>
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/semaphore.h>
+#ifdef CONFIG_BLK_DEV_FD
+#include <asm/floppy.h>
+#endif
+#ifdef CONFIG_NET
+#include <net/checksum.h>
+#endif /* CONFIG_NET */
+
+
+/*
+ * String functions
+ */
+EXPORT_SYMBOL(memcmp);
+EXPORT_SYMBOL(memset);
+EXPORT_SYMBOL(memcpy);
+EXPORT_SYMBOL(memmove);
+EXPORT_SYMBOL(memchr);
+EXPORT_SYMBOL(strcat);
+EXPORT_SYMBOL(strchr);
+EXPORT_SYMBOL(strlen);
+EXPORT_SYMBOL(strpbrk);
+EXPORT_SYMBOL(strncat);
+EXPORT_SYMBOL(strnlen);
+EXPORT_SYMBOL(strrchr);
+EXPORT_SYMBOL(strstr);
+
+EXPORT_SYMBOL(enable_irq);
+EXPORT_SYMBOL(disable_irq);
+EXPORT_SYMBOL(kernel_thread);
+
+/*
+ * gcc internal math functions
+ */
+extern long long __ashrdi3(long long, int);
+extern long long __ashldi3(long long, int);
+extern long long __lshrdi3(long long, int);
+extern int __divsi3(int, int);
+extern int __modsi3(int, int);
+extern long long __muldi3(long long, long long);
+extern int __mulsi3(int, int);
+extern unsigned int __udivsi3(unsigned int, unsigned int);
+extern unsigned int __umodsi3(unsigned int, unsigned int);
+extern unsigned long long __umoddi3(unsigned long long, unsigned long long);
+extern unsigned long long __udivdi3(unsigned long long, unsigned long long);
+
+EXPORT_SYMBOL(__ashldi3);
+EXPORT_SYMBOL(__ashrdi3);
+EXPORT_SYMBOL(__lshrdi3);
+EXPORT_SYMBOL(__divsi3);
+EXPORT_SYMBOL(__modsi3);
+EXPORT_SYMBOL(__muldi3);
+EXPORT_SYMBOL(__mulsi3);
+EXPORT_SYMBOL(__udivsi3);
+EXPORT_SYMBOL(__umodsi3);
+EXPORT_SYMBOL(__udivdi3);
+EXPORT_SYMBOL(__umoddi3);
+
+/*
+ * Semaphore operations
+ */
+EXPORT_SYMBOL(__down);
+EXPORT_SYMBOL(__down_interruptible);
+EXPORT_SYMBOL(__down_trylock);
+EXPORT_SYMBOL(__up);
+
+#ifdef CONFIG_NET
+/*
+ * Networking support
+ */
+EXPORT_SYMBOL(csum_partial_copy_generic);
+#endif /* CONFIG_NET */
+
+/*
+ * Architecture-specific symbols
+ */
+EXPORT_SYMBOL(__xtensa_copy_user);
+
+/*
+ * Kernel hacking ...
+ */
+
+#if defined(CONFIG_VGA_CONSOLE) || defined(CONFIG_DUMMY_CONSOLE)
+// FIXME EXPORT_SYMBOL(screen_info);
+#endif
+
+EXPORT_SYMBOL(get_wchan);
+
+EXPORT_SYMBOL(outsb);
+EXPORT_SYMBOL(outsw);
+EXPORT_SYMBOL(outsl);
+EXPORT_SYMBOL(insb);
+EXPORT_SYMBOL(insw);
+EXPORT_SYMBOL(insl);