From c1340c053be7a43d837a3acb352d5008be865a55 Mon Sep 17 00:00:00 2001 From: Matt Fleming Date: Sun, 28 Jun 2009 14:05:44 +0100 Subject: sh: Define HAVE_FUNCTION_TRACE_MCOUNT_TEST Enable HAVE_FUNCTION_TRACE_MCOUNT_TEST and test the value of function_trace_stop from our assembly code as opposed to using the generic C function. This should optimise our mcount/ftrace code path. Signed-off-by: Matt Fleming Signed-off-by: Paul Mundt --- arch/sh/lib/mcount.S | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S index 110fbfe1831f..cb87ef580b31 100644 --- a/arch/sh/lib/mcount.S +++ b/arch/sh/lib/mcount.S @@ -2,7 +2,7 @@ * arch/sh/lib/mcount.S * * Copyright (C) 2008 Paul Mundt - * Copyright (C) 2008 Matt Fleming + * Copyright (C) 2008, 2009 Matt Fleming * * 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 @@ -35,6 +35,12 @@ .type mcount,@function _mcount: mcount: +#ifndef CONFIG_DYNAMIC_FTRACE + mov.l .Lfunction_trace_stop, r0 + mov.l @r0, r0 + tst r0, r0 + bf ftrace_stub +#endif MCOUNT_ENTER() #ifdef CONFIG_DYNAMIC_FTRACE @@ -62,6 +68,11 @@ skip_trace: #ifdef CONFIG_DYNAMIC_FTRACE .globl ftrace_caller ftrace_caller: + mov.l .Lfunction_trace_stop, r0 + mov.l @r0, r0 + tst r0, r0 + bf ftrace_stub + MCOUNT_ENTER() .globl ftrace_call @@ -88,3 +99,7 @@ ftrace_call: ftrace_stub: rts nop + + .align 2 +.Lfunction_trace_stop: + .long function_trace_stop -- cgit v1.2.2 From c652d780c9cf7f860141de232b37160fe013feca Mon Sep 17 00:00:00 2001 From: Matt Fleming Date: Mon, 6 Jul 2009 20:16:33 +0900 Subject: sh: Add ftrace syscall tracing support Now that I've added TIF_SYSCALL_FTRACE the thread flags do not fit into a single byte any more. Code testing them now needs to be aware of the upper and lower bytes. Signed-off-by: Matt Fleming Signed-off-by: Paul Mundt --- arch/sh/lib/mcount.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S index cb87ef580b31..71e87f9b4fda 100644 --- a/arch/sh/lib/mcount.S +++ b/arch/sh/lib/mcount.S @@ -72,7 +72,7 @@ ftrace_caller: mov.l @r0, r0 tst r0, r0 bf ftrace_stub - + MCOUNT_ENTER() .globl ftrace_call -- cgit v1.2.2 From b99610fb9cdf390965c62c22322596d961591160 Mon Sep 17 00:00:00 2001 From: Matt Fleming Date: Sat, 11 Jul 2009 01:00:23 +0000 Subject: sh: Provide diagnostic kernel stack checks Enable kernel stack checking code in both the dynamic ftrace and mcount code paths. Check the stack to see if it's overflowing and make sure that the stack pointer contains an address that's either in init_stack or after the bss. Signed-off-by: Matt Fleming Signed-off-by: Paul Mundt --- arch/sh/lib/mcount.S | 85 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 85 insertions(+) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S index 71e87f9b4fda..8596483f7b41 100644 --- a/arch/sh/lib/mcount.S +++ b/arch/sh/lib/mcount.S @@ -9,6 +9,8 @@ * for more details. */ #include +#include +#include #define MCOUNT_ENTER() \ mov.l r4, @-r15; \ @@ -28,6 +30,55 @@ rts; \ mov.l @r15+, r4 +#ifdef CONFIG_STACK_DEBUG +/* + * Perform diagnostic checks on the state of the kernel stack. + * + * Check for stack overflow. If there is less than 1KB free + * then it has overflowed. + * + * Make sure the stack pointer contains a valid address. Valid + * addresses for kernel stacks are anywhere after the bss + * (after _ebss) and anywhere in init_thread_union (init_stack). + */ +#define STACK_CHECK() \ + mov #(THREAD_SIZE >> 10), r0; \ + shll8 r0; \ + shll2 r0; \ + \ + /* r1 = sp & (THREAD_SIZE - 1) */ \ + mov #-1, r1; \ + add r0, r1; \ + and r15, r1; \ + \ + mov #TI_SIZE, r3; \ + mov #(STACK_WARN >> 8), r2; \ + shll8 r2; \ + add r3, r2; \ + \ + /* Is the stack overflowing? */ \ + cmp/hi r2, r1; \ + bf stack_panic; \ + \ + /* If sp > _ebss then we're OK. */ \ + mov.l .L_ebss, r1; \ + cmp/hi r1, r15; \ + bt 1f; \ + \ + /* If sp < init_stack, we're not OK. */ \ + mov.l .L_init_thread_union, r1; \ + cmp/hs r1, r15; \ + bf stack_panic; \ + \ + /* If sp > init_stack && sp < _ebss, not OK. */ \ + add r0, r1; \ + cmp/hs r1, r15; \ + bt stack_panic; \ +1: +#else +#define STACK_CHECK() +#endif /* CONFIG_STACK_DEBUG */ + .align 2 .globl _mcount .type _mcount,@function @@ -41,6 +92,8 @@ mcount: tst r0, r0 bf ftrace_stub #endif + STACK_CHECK() + MCOUNT_ENTER() #ifdef CONFIG_DYNAMIC_FTRACE @@ -73,6 +126,8 @@ ftrace_caller: tst r0, r0 bf ftrace_stub + STACK_CHECK() + MCOUNT_ENTER() .globl ftrace_call @@ -100,6 +155,36 @@ ftrace_stub: rts nop +#ifdef CONFIG_STACK_DEBUG + .globl stack_panic +stack_panic: + mov.l .Ldump_stack, r0 + jsr @r0 + nop + + mov.l .Lpanic, r0 + jsr @r0 + mov.l .Lpanic_s, r4 + + rts + nop + .align 2 .Lfunction_trace_stop: .long function_trace_stop +.L_ebss: + .long _ebss +.L_init_thread_union: + .long init_thread_union +.Lpanic: + .long panic +.Lpanic_s: + .long .Lpanic_str +.Ldump_stack: + .long dump_stack + + .section .rodata + .align 2 +.Lpanic_str: + .string "Stack error" +#endif /* CONFIG_STACK_DEBUG */ -- cgit v1.2.2 From 327933f5d6cdf083284d3c06e0370d1de464aef4 Mon Sep 17 00:00:00 2001 From: Matt Fleming Date: Sat, 11 Jul 2009 00:29:03 +0000 Subject: sh: Function graph tracer support Add both dynamic and static function graph tracer support for sh. Signed-off-by: Matt Fleming Signed-off-by: Paul Mundt --- arch/sh/lib/Makefile | 1 + arch/sh/lib/mcount.S | 117 ++++++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 117 insertions(+), 1 deletion(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/Makefile b/arch/sh/lib/Makefile index aaea580b65bb..19328d90a2d1 100644 --- a/arch/sh/lib/Makefile +++ b/arch/sh/lib/Makefile @@ -25,6 +25,7 @@ memcpy-$(CONFIG_CPU_SH4) := memcpy-sh4.o lib-$(CONFIG_MMU) += copy_page.o clear_page.o lib-$(CONFIG_FUNCTION_TRACER) += mcount.o +lib-$(CONFIG_FUNCTION_GRAPH_TRACER) += mcount.o lib-y += $(memcpy-y) $(udivsi3-y) EXTRA_CFLAGS += -Werror diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S index 8596483f7b41..bd3ec648becc 100644 --- a/arch/sh/lib/mcount.S +++ b/arch/sh/lib/mcount.S @@ -111,14 +111,62 @@ mcount_call: jsr @r6 nop +#ifdef CONFIG_FUNCTION_GRAPH_TRACER + mov.l .Lftrace_graph_return, r6 + mov.l .Lftrace_stub, r7 + cmp/eq r6, r7 + bt 1f + + mov.l .Lftrace_graph_caller, r0 + jmp @r0 + nop + +1: + mov.l .Lftrace_graph_entry, r6 + mov.l .Lftrace_graph_entry_stub, r7 + cmp/eq r6, r7 + bt skip_trace + + mov.l .Lftrace_graph_caller, r0 + jmp @r0 + nop + + .align 2 +.Lftrace_graph_return: + .long ftrace_graph_return +.Lftrace_graph_entry: + .long ftrace_graph_entry +.Lftrace_graph_entry_stub: + .long ftrace_graph_entry_stub +.Lftrace_graph_caller: + .long ftrace_graph_caller +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ + + .globl skip_trace skip_trace: MCOUNT_LEAVE() .align 2 .Lftrace_trace_function: - .long ftrace_trace_function + .long ftrace_trace_function #ifdef CONFIG_DYNAMIC_FTRACE +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +/* + * NOTE: Do not move either ftrace_graph_call or ftrace_caller + * as this will affect the calculation of GRAPH_INSN_OFFSET. + */ + .globl ftrace_graph_call +ftrace_graph_call: + mov.l .Lskip_trace, r0 + jmp @r0 + nop + + .align 2 +.Lskip_trace: + .long skip_trace +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ + .globl ftrace_caller ftrace_caller: mov.l .Lfunction_trace_stop, r0 @@ -136,7 +184,12 @@ ftrace_call: jsr @r6 nop +#ifdef CONFIG_FUNCTION_GRAPH_TRACER + bra ftrace_graph_call + nop +#else MCOUNT_LEAVE() +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ #endif /* CONFIG_DYNAMIC_FTRACE */ /* @@ -188,3 +241,65 @@ stack_panic: .Lpanic_str: .string "Stack error" #endif /* CONFIG_STACK_DEBUG */ + +#ifdef CONFIG_FUNCTION_GRAPH_TRACER + .globl ftrace_graph_caller +ftrace_graph_caller: + mov.l 2f, r0 + mov.l @r0, r0 + tst r0, r0 + bt 1f + + mov.l 3f, r1 + jmp @r1 + nop +1: + /* + * MCOUNT_ENTER() pushed 5 registers onto the stack, so + * the stack address containing our return address is + * r15 + 20. + */ + mov #20, r0 + add r15, r0 + mov r0, r4 + + mov.l .Lprepare_ftrace_return, r0 + jsr @r0 + nop + + MCOUNT_LEAVE() + + .align 2 +2: .long function_trace_stop +3: .long skip_trace +.Lprepare_ftrace_return: + .long prepare_ftrace_return + + .globl return_to_handler +return_to_handler: + /* + * Save the return values. + */ + mov.l r0, @-r15 + mov.l r1, @-r15 + + mov #0, r4 + + mov.l .Lftrace_return_to_handler, r0 + jsr @r0 + nop + + /* + * The return value from ftrace_return_handler has the real + * address that we should return to. + */ + lds r0, pr + mov.l @r15+, r1 + rts + mov.l @r15+, r0 + + + .align 2 +.Lftrace_return_to_handler: + .long ftrace_return_to_handler +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ -- cgit v1.2.2 From 473d1cf4ee623b043790838bcf77e77958840bf2 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Sat, 11 Jul 2009 19:56:58 +0900 Subject: sh: Decouple mcount from ftrace. This adds a general CONFIG_MCOUNT in order to permit mcount generation without ftrace support. This is primarily for allowing platforms to enable aggressive stack overflow checking without having to enable ftrace support. Based on the sparc64 implementation. Signed-off-by: Paul Mundt --- arch/sh/lib/Makefile | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/Makefile b/arch/sh/lib/Makefile index 19328d90a2d1..c2b28d8b2dd1 100644 --- a/arch/sh/lib/Makefile +++ b/arch/sh/lib/Makefile @@ -24,8 +24,7 @@ memcpy-y := memcpy.o memcpy-$(CONFIG_CPU_SH4) := memcpy-sh4.o lib-$(CONFIG_MMU) += copy_page.o clear_page.o -lib-$(CONFIG_FUNCTION_TRACER) += mcount.o -lib-$(CONFIG_FUNCTION_GRAPH_TRACER) += mcount.o +lib-$(CONFIG_MCOUNT) += mcount.o lib-y += $(memcpy-y) $(udivsi3-y) EXTRA_CFLAGS += -Werror -- cgit v1.2.2 From a470b95e99ea77ef1e307ff181e59a4a16caa4f4 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Sat, 11 Jul 2009 20:33:34 +0900 Subject: sh: Fix up ftrace build error when STACK_DEBUG=n. Presently the closest reference to function_trace_stop is within a CONFIG_STACK_DEBUG block. When this is turned off, the build bails out with a pcrel too far error. Reorder things a bit to handle the various combinations. Signed-off-by: Paul Mundt --- arch/sh/lib/mcount.S | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S index bd3ec648becc..9e397aafc165 100644 --- a/arch/sh/lib/mcount.S +++ b/arch/sh/lib/mcount.S @@ -192,6 +192,10 @@ ftrace_call: #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ #endif /* CONFIG_DYNAMIC_FTRACE */ + .align 2 +.Lfunction_trace_stop: + .long function_trace_stop + /* * NOTE: From here on the locations of the .Lftrace_stub label and * ftrace_stub itself are fixed. Adding additional data here will skew @@ -199,7 +203,6 @@ ftrace_call: * Place new labels either after the ftrace_stub body, or before * ftrace_caller. You have been warned. */ - .align 2 .Lftrace_stub: .long ftrace_stub @@ -223,8 +226,6 @@ stack_panic: nop .align 2 -.Lfunction_trace_stop: - .long function_trace_stop .L_ebss: .long _ebss .L_init_thread_union: -- cgit v1.2.2 From e460ab27b6c3ea313762169713086529d5bfb8bc Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Sat, 11 Jul 2009 21:06:53 +0900 Subject: sh: Fix up stack overflow check with ftrace disabled. Presently the STACK_CHECK() code is called in to multiple times, although it's only necessary from the mcount entry. The code still attempts to treat the nop case as an ftrace path resulting in superfluous code flow for the case where ftrace is disabled. And finally, this also fixes up references to a few undefined symbols when FUNCTION_TRACER=n. Signed-off-by: Paul Mundt --- arch/sh/lib/mcount.S | 76 +++++++++++++++++++++++++++------------------------- 1 file changed, 40 insertions(+), 36 deletions(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S index 9e397aafc165..84a57761f17e 100644 --- a/arch/sh/lib/mcount.S +++ b/arch/sh/lib/mcount.S @@ -1,7 +1,7 @@ /* * arch/sh/lib/mcount.S * - * Copyright (C) 2008 Paul Mundt + * Copyright (C) 2008, 2009 Paul Mundt * Copyright (C) 2008, 2009 Matt Fleming * * This file is subject to the terms and conditions of the GNU General Public @@ -86,13 +86,18 @@ .type mcount,@function _mcount: mcount: + STACK_CHECK() + +#ifndef CONFIG_FUNCTION_TRACER + rts + nop +#else #ifndef CONFIG_DYNAMIC_FTRACE mov.l .Lfunction_trace_stop, r0 mov.l @r0, r0 tst r0, r0 bf ftrace_stub #endif - STACK_CHECK() MCOUNT_ENTER() @@ -174,8 +179,6 @@ ftrace_caller: tst r0, r0 bf ftrace_stub - STACK_CHECK() - MCOUNT_ENTER() .globl ftrace_call @@ -211,38 +214,6 @@ ftrace_stub: rts nop -#ifdef CONFIG_STACK_DEBUG - .globl stack_panic -stack_panic: - mov.l .Ldump_stack, r0 - jsr @r0 - nop - - mov.l .Lpanic, r0 - jsr @r0 - mov.l .Lpanic_s, r4 - - rts - nop - - .align 2 -.L_ebss: - .long _ebss -.L_init_thread_union: - .long init_thread_union -.Lpanic: - .long panic -.Lpanic_s: - .long .Lpanic_str -.Ldump_stack: - .long dump_stack - - .section .rodata - .align 2 -.Lpanic_str: - .string "Stack error" -#endif /* CONFIG_STACK_DEBUG */ - #ifdef CONFIG_FUNCTION_GRAPH_TRACER .globl ftrace_graph_caller ftrace_graph_caller: @@ -304,3 +275,36 @@ return_to_handler: .Lftrace_return_to_handler: .long ftrace_return_to_handler #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ +#endif /* CONFIG_FUNCTION_TRACER */ + +#ifdef CONFIG_STACK_DEBUG + .globl stack_panic +stack_panic: + mov.l .Ldump_stack, r0 + jsr @r0 + nop + + mov.l .Lpanic, r0 + jsr @r0 + mov.l .Lpanic_s, r4 + + rts + nop + + .align 2 +.L_ebss: + .long _ebss +.L_init_thread_union: + .long init_thread_union +.Lpanic: + .long panic +.Lpanic_s: + .long .Lpanic_str +.Ldump_stack: + .long dump_stack + + .section .rodata + .align 2 +.Lpanic_str: + .string "Stack error" +#endif /* CONFIG_STACK_DEBUG */ -- cgit v1.2.2 From dfff0fa65ab15db45acd64b3189787d37ab163cd Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Mon, 27 Jul 2009 20:53:22 +0900 Subject: sh: wire up clear_user_highpage() for sh4, convert sh7705. This wires up clear_user_highpage() on SH-4 and subsequently converts the SH7705 32kB cache mode over to using it. Now that the SH-4 implementation handles all of the dcache purging directly in the aliasing case, there is no need to do this in the default clear_page() implementation. Signed-off-by: Paul Mundt --- arch/sh/lib/clear_page.S | 46 ---------------------------------------------- 1 file changed, 46 deletions(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/clear_page.S b/arch/sh/lib/clear_page.S index 8342bfbde64c..bee9817e055d 100644 --- a/arch/sh/lib/clear_page.S +++ b/arch/sh/lib/clear_page.S @@ -8,52 +8,6 @@ #include #include -/* - * clear_page - * @to: P1 address - * - * void clear_page(void *to) - */ - -/* - * r0 --- scratch - * r4 --- to - * r5 --- to + PAGE_SIZE - */ -ENTRY(clear_page) - mov r4,r5 - mov.l .Llimit,r0 - add r0,r5 - mov #0,r0 - ! -1: -#if defined(CONFIG_CPU_SH4) - movca.l r0,@r4 - mov r4,r1 -#else - mov.l r0,@r4 -#endif - add #32,r4 - mov.l r0,@-r4 - mov.l r0,@-r4 - mov.l r0,@-r4 - mov.l r0,@-r4 - mov.l r0,@-r4 - mov.l r0,@-r4 - mov.l r0,@-r4 -#if defined(CONFIG_CPU_SH4) - ocbwb @r1 -#endif - cmp/eq r5,r4 - bf/s 1b - add #28,r4 - ! - rts - nop - - .balign 4 -.Llimit: .long (PAGE_SIZE-28) - ENTRY(__clear_user) ! mov #0, r0 -- cgit v1.2.2 From 221c007b028ebf663ebee4fc90483909547d92a7 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Mon, 27 Jul 2009 20:55:46 +0900 Subject: sh: Rename arch/sh/lib/clear_page.S -> __clear_user.S. Now that this only contains the __clear_user() function, rename it accordingly. Signed-off-by: Paul Mundt --- arch/sh/lib/Makefile | 2 +- arch/sh/lib/__clear_user.S | 108 +++++++++++++++++++++++++++++++++++++++++++++ arch/sh/lib/clear_page.S | 108 --------------------------------------------- 3 files changed, 109 insertions(+), 109 deletions(-) create mode 100644 arch/sh/lib/__clear_user.S delete mode 100644 arch/sh/lib/clear_page.S (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/Makefile b/arch/sh/lib/Makefile index c2b28d8b2dd1..a969b47c5463 100644 --- a/arch/sh/lib/Makefile +++ b/arch/sh/lib/Makefile @@ -23,7 +23,7 @@ obj-y += io.o memcpy-y := memcpy.o memcpy-$(CONFIG_CPU_SH4) := memcpy-sh4.o -lib-$(CONFIG_MMU) += copy_page.o clear_page.o +lib-$(CONFIG_MMU) += copy_page.o __clear_user.o lib-$(CONFIG_MCOUNT) += mcount.o lib-y += $(memcpy-y) $(udivsi3-y) diff --git a/arch/sh/lib/__clear_user.S b/arch/sh/lib/__clear_user.S new file mode 100644 index 000000000000..bee9817e055d --- /dev/null +++ b/arch/sh/lib/__clear_user.S @@ -0,0 +1,108 @@ +/* + * __clear_user_page, __clear_user, clear_page implementation of SuperH + * + * Copyright (C) 2001 Kaz Kojima + * Copyright (C) 2001, 2002 Niibe Yutaka + * Copyright (C) 2006 Paul Mundt + */ +#include +#include + +ENTRY(__clear_user) + ! + mov #0, r0 + mov #0xe0, r1 ! 0xffffffe0 + ! + ! r4..(r4+31)&~32 -------- not aligned [ Area 0 ] + ! (r4+31)&~32..(r4+r5)&~32 -------- aligned [ Area 1 ] + ! (r4+r5)&~32..r4+r5 -------- not aligned [ Area 2 ] + ! + ! Clear area 0 + mov r4, r2 + ! + tst r1, r5 ! length < 32 + bt .Larea2 ! skip to remainder + ! + add #31, r2 + and r1, r2 + cmp/eq r4, r2 + bt .Larea1 + mov r2, r3 + sub r4, r3 + mov r3, r7 + mov r4, r2 + ! +.L0: dt r3 +0: mov.b r0, @r2 + bf/s .L0 + add #1, r2 + ! + sub r7, r5 + mov r2, r4 +.Larea1: + mov r4, r3 + add r5, r3 + and r1, r3 + cmp/hi r2, r3 + bf .Larea2 + ! + ! Clear area 1 +#if defined(CONFIG_CPU_SH4) +1: movca.l r0, @r2 +#else +1: mov.l r0, @r2 +#endif + add #4, r2 +2: mov.l r0, @r2 + add #4, r2 +3: mov.l r0, @r2 + add #4, r2 +4: mov.l r0, @r2 + add #4, r2 +5: mov.l r0, @r2 + add #4, r2 +6: mov.l r0, @r2 + add #4, r2 +7: mov.l r0, @r2 + add #4, r2 +8: mov.l r0, @r2 + add #4, r2 + cmp/hi r2, r3 + bt/s 1b + nop + ! + ! Clear area 2 +.Larea2: + mov r4, r3 + add r5, r3 + cmp/hs r3, r2 + bt/s .Ldone + sub r2, r3 +.L2: dt r3 +9: mov.b r0, @r2 + bf/s .L2 + add #1, r2 + ! +.Ldone: rts + mov #0, r0 ! return 0 as normal return + + ! return the number of bytes remained +.Lbad_clear_user: + mov r4, r0 + add r5, r0 + rts + sub r2, r0 + +.section __ex_table,"a" + .align 2 + .long 0b, .Lbad_clear_user + .long 1b, .Lbad_clear_user + .long 2b, .Lbad_clear_user + .long 3b, .Lbad_clear_user + .long 4b, .Lbad_clear_user + .long 5b, .Lbad_clear_user + .long 6b, .Lbad_clear_user + .long 7b, .Lbad_clear_user + .long 8b, .Lbad_clear_user + .long 9b, .Lbad_clear_user +.previous diff --git a/arch/sh/lib/clear_page.S b/arch/sh/lib/clear_page.S deleted file mode 100644 index bee9817e055d..000000000000 --- a/arch/sh/lib/clear_page.S +++ /dev/null @@ -1,108 +0,0 @@ -/* - * __clear_user_page, __clear_user, clear_page implementation of SuperH - * - * Copyright (C) 2001 Kaz Kojima - * Copyright (C) 2001, 2002 Niibe Yutaka - * Copyright (C) 2006 Paul Mundt - */ -#include -#include - -ENTRY(__clear_user) - ! - mov #0, r0 - mov #0xe0, r1 ! 0xffffffe0 - ! - ! r4..(r4+31)&~32 -------- not aligned [ Area 0 ] - ! (r4+31)&~32..(r4+r5)&~32 -------- aligned [ Area 1 ] - ! (r4+r5)&~32..r4+r5 -------- not aligned [ Area 2 ] - ! - ! Clear area 0 - mov r4, r2 - ! - tst r1, r5 ! length < 32 - bt .Larea2 ! skip to remainder - ! - add #31, r2 - and r1, r2 - cmp/eq r4, r2 - bt .Larea1 - mov r2, r3 - sub r4, r3 - mov r3, r7 - mov r4, r2 - ! -.L0: dt r3 -0: mov.b r0, @r2 - bf/s .L0 - add #1, r2 - ! - sub r7, r5 - mov r2, r4 -.Larea1: - mov r4, r3 - add r5, r3 - and r1, r3 - cmp/hi r2, r3 - bf .Larea2 - ! - ! Clear area 1 -#if defined(CONFIG_CPU_SH4) -1: movca.l r0, @r2 -#else -1: mov.l r0, @r2 -#endif - add #4, r2 -2: mov.l r0, @r2 - add #4, r2 -3: mov.l r0, @r2 - add #4, r2 -4: mov.l r0, @r2 - add #4, r2 -5: mov.l r0, @r2 - add #4, r2 -6: mov.l r0, @r2 - add #4, r2 -7: mov.l r0, @r2 - add #4, r2 -8: mov.l r0, @r2 - add #4, r2 - cmp/hi r2, r3 - bt/s 1b - nop - ! - ! Clear area 2 -.Larea2: - mov r4, r3 - add r5, r3 - cmp/hs r3, r2 - bt/s .Ldone - sub r2, r3 -.L2: dt r3 -9: mov.b r0, @r2 - bf/s .L2 - add #1, r2 - ! -.Ldone: rts - mov #0, r0 ! return 0 as normal return - - ! return the number of bytes remained -.Lbad_clear_user: - mov r4, r0 - add r5, r0 - rts - sub r2, r0 - -.section __ex_table,"a" - .align 2 - .long 0b, .Lbad_clear_user - .long 1b, .Lbad_clear_user - .long 2b, .Lbad_clear_user - .long 3b, .Lbad_clear_user - .long 4b, .Lbad_clear_user - .long 5b, .Lbad_clear_user - .long 6b, .Lbad_clear_user - .long 7b, .Lbad_clear_user - .long 8b, .Lbad_clear_user - .long 9b, .Lbad_clear_user -.previous -- cgit v1.2.2 From fea966f7564205fcf5919af9bde031e753419c96 Mon Sep 17 00:00:00 2001 From: Stuart Menefy Date: Mon, 24 Aug 2009 17:09:53 +0900 Subject: sh: Remove implicit sign extension from assembler immediates The SH instruction set has several instructions which accept an 8 bit immediate operand. For logical instructions this operand is zero extended, for arithmetic instructions the operand is sign extended. After adding an option to the assembler to check this, it was found that several pieces of assembly code were assuming this behaviour, and in one case getting it wrong. So this patch explicitly sign extends any immediate operands, which makes it obvious what is happening, and fixes the one case which got it wrong. Signed-off-by: Stuart Menefy Signed-off-by: Paul Mundt --- arch/sh/lib/clear_page.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/clear_page.S b/arch/sh/lib/clear_page.S index 8342bfbde64c..c92244d4ff9d 100644 --- a/arch/sh/lib/clear_page.S +++ b/arch/sh/lib/clear_page.S @@ -57,7 +57,7 @@ ENTRY(clear_page) ENTRY(__clear_user) ! mov #0, r0 - mov #0xe0, r1 ! 0xffffffe0 + mov #0xffffffe0, r1 ! ! r4..(r4+31)&~32 -------- not aligned [ Area 0 ] ! (r4+31)&~32..(r4+r5)&~32 -------- aligned [ Area 1 ] -- cgit v1.2.2 From bd4fb4d4c1e4a5a2ffbf57a83817a749df1339dd Mon Sep 17 00:00:00 2001 From: Stuart Menefy Date: Mon, 24 Aug 2009 18:18:50 +0900 Subject: sh: Fix underflow in SH udelay() code. Signed-off-by: Stuart Menefy Signed-off-by: Paul Mundt --- arch/sh/lib/delay.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/delay.c b/arch/sh/lib/delay.c index f3ddd2133e6f..faa8f86c0db4 100644 --- a/arch/sh/lib/delay.c +++ b/arch/sh/lib/delay.c @@ -21,13 +21,14 @@ void __delay(unsigned long loops) inline void __const_udelay(unsigned long xloops) { + xloops *= 4; __asm__("dmulu.l %0, %2\n\t" "sts mach, %0" : "=r" (xloops) : "0" (xloops), - "r" (HZ * cpu_data[raw_smp_processor_id()].loops_per_jiffy) + "r" (cpu_data[raw_smp_processor_id()].loops_per_jiffy * (HZ/4)) : "macl", "mach"); - __delay(xloops); + __delay(++xloops); } void __udelay(unsigned long usecs) -- cgit v1.2.2 From a2494b9b5fb702becaf8d8e3138f7a1a0d3c537e Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Tue, 8 Sep 2009 16:23:08 +0900 Subject: sh: Kill off dcache writeback from copy_page(). Now that the cache purging is handled manually by all copy_page() callers, we can kill off copy_page()'s on writeback. This optimizes the non-aliasing case. Signed-off-by: Paul Mundt --- arch/sh/lib/copy_page.S | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) (limited to 'arch/sh/lib') diff --git a/arch/sh/lib/copy_page.S b/arch/sh/lib/copy_page.S index 43de7e8e4e17..9d7b8bc51866 100644 --- a/arch/sh/lib/copy_page.S +++ b/arch/sh/lib/copy_page.S @@ -30,7 +30,9 @@ ENTRY(copy_page) mov r4,r10 mov r5,r11 mov r5,r8 - mov.l .Lpsz,r0 + mov #(PAGE_SIZE >> 10), r0 + shll8 r0 + shll2 r0 add r0,r8 ! 1: mov.l @r11+,r0 @@ -43,7 +45,6 @@ ENTRY(copy_page) mov.l @r11+,r7 #if defined(CONFIG_CPU_SH4) movca.l r0,@r10 - mov r10,r0 #else mov.l r0,@r10 #endif @@ -55,9 +56,6 @@ ENTRY(copy_page) mov.l r3,@-r10 mov.l r2,@-r10 mov.l r1,@-r10 -#if defined(CONFIG_CPU_SH4) - ocbwb @r0 -#endif cmp/eq r11,r8 bf/s 1b add #28,r10 @@ -68,9 +66,6 @@ ENTRY(copy_page) rts nop - .balign 4 -.Lpsz: .long PAGE_SIZE - /* * __kernel_size_t __copy_user(void *to, const void *from, __kernel_size_t n); * Return the number of bytes NOT copied -- cgit v1.2.2