Mailing List Archive

[PATCH] arm: fix unaligned memcpy() and memmove()
From: David Vrabel <david.vrabel@citrix.com>

If memcpy() and memmove() were used with source and destination of
different alignment then the result would be all jumbled.

When their implementations were imported from Linux some macros for
big-endian platforms were taken instead of the correct little-endian
ones (specifically, the push and pull macros in assembler.h).

Fix this by taking Linux's arch/include/asm/assembler.h as-is and
making only the minimum changes necessary for Xen.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
---
xen/arch/arm/lib/assembler.h | 324 ++++++++++++++++++++++++++++++++++++++---
1 files changed, 300 insertions(+), 24 deletions(-)

diff --git a/xen/arch/arm/lib/assembler.h b/xen/arch/arm/lib/assembler.h
index f8f0961..f8d4b3a 100644
--- a/xen/arch/arm/lib/assembler.h
+++ b/xen/arch/arm/lib/assembler.h
@@ -1,11 +1,72 @@
-#ifndef __ARCH_ARM_LIB_ASSEMBLER_H__
-#define __ARCH_ARM_LIB_ASSEMBLER_H__
-
/* From Linux arch/arm/include/asm/assembler.h */
/*
+ * arch/arm/include/asm/assembler.h
+ *
+ * Copyright (C) 1996-2000 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This file contains arm architecture specific defines
+ * for the different processors.
+ *
+ * Do not include any C declarations in this file - it is included by
+ * assembler source.
+ */
+#ifndef __ASM_ASSEMBLER_H__
+#define __ASM_ASSEMBLER_H__
+
+#ifndef __ASSEMBLY__
+#error "Only include this from assembly code"
+#endif
+
+// No Thumb, hence:
+#define W(instr) instr
+#define ARM(instr...) instr
+#define THUMB(instr...)
+
+#ifdef CONFIG_ARM_UNWIND
+#define UNWIND(code...) code
+#else
+#define UNWIND(code...)
+#endif
+
+/*
+ * Endian independent macros for shifting bytes within registers.
+ */
+#ifndef __ARMEB__
+#define pull lsr
+#define push lsl
+#define get_byte_0 lsl #0
+#define get_byte_1 lsr #8
+#define get_byte_2 lsr #16
+#define get_byte_3 lsr #24
+#define put_byte_0 lsl #0
+#define put_byte_1 lsl #8
+#define put_byte_2 lsl #16
+#define put_byte_3 lsl #24
+#else
+#define pull lsl
+#define push lsr
+#define get_byte_0 lsr #24
+#define get_byte_1 lsr #16
+#define get_byte_2 lsr #8
+#define get_byte_3 lsl #0
+#define put_byte_0 lsl #24
+#define put_byte_1 lsl #16
+#define put_byte_2 lsl #8
+#define put_byte_3 lsl #0
+#endif
+
+/*
* Data preload for architectures that support it
*/
-#define PLD(code...) code
+#if __LINUX_ARM_ARCH__ >= 5
+#define PLD(code...) code
+#else
+#define PLD(code...)
+#endif

/*
* This can be used to enable code to cacheline align the destination
@@ -16,34 +77,249 @@
*
* On Feroceon there is much to gain however, regardless of cache mode.
*/
-#ifdef CONFIG_CPU_FEROCEON /* Not in Xen... */
+#ifdef CONFIG_CPU_FEROCEON
#define CALGN(code...) code
#else
#define CALGN(code...)
#endif

-// No Thumb, hence:
-#define W(instr) instr
-#define ARM(instr...) instr
-#define THUMB(instr...)
+/*
+ * Enable and disable interrupts
+ */
+#if __LINUX_ARM_ARCH__ >= 6
+ .macro disable_irq_notrace
+ cpsid i
+ .endm

-#ifdef CONFIG_ARM_UNWIND
-#define UNWIND(code...) code
+ .macro enable_irq_notrace
+ cpsie i
+ .endm
#else
-#define UNWIND(code...)
+ .macro disable_irq_notrace
+ msr cpsr_c, #PSR_I_BIT | SVC_MODE
+ .endm
+
+ .macro enable_irq_notrace
+ msr cpsr_c, #SVC_MODE
+ .endm
#endif

-#define pull lsl
-#define push lsr
-#define get_byte_0 lsr #24
-#define get_byte_1 lsr #16
-#define get_byte_2 lsr #8
-#define get_byte_3 lsl #0
-#define put_byte_0 lsl #24
-#define put_byte_1 lsl #16
-#define put_byte_2 lsl #8
-#define put_byte_3 lsl #0
+ .macro asm_trace_hardirqs_off
+#if defined(CONFIG_TRACE_IRQFLAGS)
+ stmdb sp!, {r0-r3, ip, lr}
+ bl trace_hardirqs_off
+ ldmia sp!, {r0-r3, ip, lr}
+#endif
+ .endm
+
+ .macro asm_trace_hardirqs_on_cond, cond
+#if defined(CONFIG_TRACE_IRQFLAGS)
+ /*
+ * actually the registers should be pushed and pop'd conditionally, but
+ * after bl the flags are certainly clobbered
+ */
+ stmdb sp!, {r0-r3, ip, lr}
+ bl\cond trace_hardirqs_on
+ ldmia sp!, {r0-r3, ip, lr}
+#endif
+ .endm
+
+ .macro asm_trace_hardirqs_on
+ asm_trace_hardirqs_on_cond al
+ .endm
+
+ .macro disable_irq
+ disable_irq_notrace
+ asm_trace_hardirqs_off
+ .endm
+
+ .macro enable_irq
+ asm_trace_hardirqs_on
+ enable_irq_notrace
+ .endm
+/*
+ * Save the current IRQ state and disable IRQs. Note that this macro
+ * assumes FIQs are enabled, and that the processor is in SVC mode.
+ */
+ .macro save_and_disable_irqs, oldcpsr
+ mrs \oldcpsr, cpsr
+ disable_irq
+ .endm
+
+/*
+ * Restore interrupt state previously stored in a register. We don't
+ * guarantee that this will preserve the flags.
+ */
+ .macro restore_irqs_notrace, oldcpsr
+ msr cpsr_c, \oldcpsr
+ .endm
+
+ .macro restore_irqs, oldcpsr
+ tst \oldcpsr, #PSR_I_BIT
+ asm_trace_hardirqs_on_cond eq
+ restore_irqs_notrace \oldcpsr
+ .endm
+
+#define USER(x...) \
+9999: x; \
+ .pushsection __ex_table,"a"; \
+ .align 3; \
+ .long 9999b,9001f; \
+ .popsection
+
+#ifdef CONFIG_SMP
+#define ALT_SMP(instr...) \
+9998: instr
+/*
+ * Note: if you get assembler errors from ALT_UP() when building with
+ * CONFIG_THUMB2_KERNEL, you almost certainly need to use
+ * ALT_SMP( W(instr) ... )
+ */
+#define ALT_UP(instr...) \
+ .pushsection ".alt.smp.init", "a" ;\
+ .long 9998b ;\
+9997: instr ;\
+ .if . - 9997b != 4 ;\
+ .error "ALT_UP() content must assemble to exactly 4 bytes";\
+ .endif ;\
+ .popsection
+#define ALT_UP_B(label) \
+ .equ up_b_offset, label - 9998b ;\
+ .pushsection ".alt.smp.init", "a" ;\
+ .long 9998b ;\
+ W(b) . + up_b_offset ;\
+ .popsection
+#else
+#define ALT_SMP(instr...)
+#define ALT_UP(instr...) instr
+#define ALT_UP_B(label) b label
+#endif
+
+/*
+ * Instruction barrier
+ */
+ .macro instr_sync
+#if __LINUX_ARM_ARCH__ >= 7
+ isb
+#elif __LINUX_ARM_ARCH__ == 6
+ mcr p15, 0, r0, c7, c5, 4
+#endif
+ .endm
+
+/*
+ * SMP data memory barrier
+ */
+ .macro smp_dmb mode
+#ifdef CONFIG_SMP
+#if __LINUX_ARM_ARCH__ >= 7
+ .ifeqs "\mode","arm"
+ ALT_SMP(dmb)
+ .else
+ ALT_SMP(W(dmb))
+ .endif
+#elif __LINUX_ARM_ARCH__ == 6
+ ALT_SMP(mcr p15, 0, r0, c7, c10, 5) @ dmb
+#else
+#error Incompatible SMP platform
+#endif
+ .ifeqs "\mode","arm"
+ ALT_UP(nop)
+ .else
+ ALT_UP(W(nop))
+ .endif
+#endif
+ .endm
+
+#ifdef CONFIG_THUMB2_KERNEL
+ .macro setmode, mode, reg
+ mov \reg, #\mode
+ msr cpsr_c, \reg
+ .endm
+#else
+ .macro setmode, mode, reg
+ msr cpsr_c, #\mode
+ .endm
+#endif
+
+/*
+ * STRT/LDRT access macros with ARM and Thumb-2 variants
+ */
+#ifdef CONFIG_THUMB2_KERNEL
+
+ .macro usraccoff, instr, reg, ptr, inc, off, cond, abort, t=T()
+9999:
+ .if \inc == 1
+ \instr\cond\()b\()\t\().w \reg, [\ptr, #\off]
+ .elseif \inc == 4
+ \instr\cond\()\t\().w \reg, [\ptr, #\off]
+ .else
+ .error "Unsupported inc macro argument"
+ .endif
+
+ .pushsection __ex_table,"a"
+ .align 3
+ .long 9999b, \abort
+ .popsection
+ .endm
+
+ .macro usracc, instr, reg, ptr, inc, cond, rept, abort
+ @ explicit IT instruction needed because of the label
+ @ introduced by the USER macro
+ .ifnc \cond,al
+ .if \rept == 1
+ itt \cond
+ .elseif \rept == 2
+ ittt \cond
+ .else
+ .error "Unsupported rept macro argument"
+ .endif
+ .endif
+
+ @ Slightly optimised to avoid incrementing the pointer twice
+ usraccoff \instr, \reg, \ptr, \inc, 0, \cond, \abort
+ .if \rept == 2
+ usraccoff \instr, \reg, \ptr, \inc, \inc, \cond, \abort
+ .endif
+
+ add\cond \ptr, #\rept * \inc
+ .endm
+
+#else /* !CONFIG_THUMB2_KERNEL */
+
+ .macro usracc, instr, reg, ptr, inc, cond, rept, abort, t=T()
+ .rept \rept
+9999:
+ .if \inc == 1
+ \instr\cond\()b\()\t \reg, [\ptr], #\inc
+ .elseif \inc == 4
+ \instr\cond\()\t \reg, [\ptr], #\inc
+ .else
+ .error "Unsupported inc macro argument"
+ .endif
+
+ .pushsection __ex_table,"a"
+ .align 3
+ .long 9999b, \abort
+ .popsection
+ .endr
+ .endm
+
+#endif /* CONFIG_THUMB2_KERNEL */
+
+ .macro strusr, reg, ptr, inc, cond=al, rept=1, abort=9001f
+ usracc str, \reg, \ptr, \inc, \cond, \rept, \abort
+ .endm
+
+ .macro ldrusr, reg, ptr, inc, cond=al, rept=1, abort=9001f
+ usracc ldr, \reg, \ptr, \inc, \cond, \rept, \abort
+ .endm

-#define smp_dmb dmb
+/* Utility macro for declaring string literals */
+ .macro string name:req, string
+ .type \name , #object
+\name:
+ .asciz "\string"
+ .size \name , . - \name
+ .endm

-#endif /* __ARCH_ARM_LIB_ASSEMBLER_H__ */
+#endif /* __ASM_ASSEMBLER_H__ */
--
1.7.2.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
Re: [PATCH] arm: fix unaligned memcpy() and memmove() [ In reply to ]
On Fri, 2012-02-24 at 18:45 +0000, David Vrabel wrote:
> From: David Vrabel <david.vrabel@citrix.com>
>
> If memcpy() and memmove() were used with source and destination of
> different alignment then the result would be all jumbled.
>
> When their implementations were imported from Linux some macros for
> big-endian platforms were taken instead of the correct little-endian
> ones (specifically, the push and pull macros in assembler.h).
>
> Fix this by taking Linux's arch/include/asm/assembler.h as-is and
> making only the minimum changes necessary for Xen.
>
> Signed-off-by: David Vrabel <david.vrabel@citrix.com>
> Cc: Ian Campbell <ian.campbell@citrix.com>

Committed, thanks!



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel