On Fri, Oct 16, 2015 at 08:52:24AM +0100, Vladimir Murzin wrote:
> > On 15/10/15 14:02, Will Deacon wrote:
> >> diff --git a/arch/arm64/kernel/armv8_deprecated.c
> >> b/arch/arm64/kernel/armv8_deprecated.c
> >> index bcee7abac68e..6039d1eb5912 100644
> >> --- a/arch/arm64/kernel/armv8_deprecated.c
> >> +++ b/arch/arm64/kernel/armv8_deprecated.c
> >> @@ -284,12 +284,12 @@ static void register_insn_emulation_sysctl(struct
> >> ctl_table *table)
> >> __asm__ __volatile__( \
> >> ALTERNATIVE("nop", SET_PSTATE_PAN(0), ARM64_HAS_PAN, \
> >> CONFIG_ARM64_PAN) \
> >> - " mov %w2, %w1\n" \
> >> - "0: ldxr"B" %w1, [%3]\n" \
> >> - "1: stxr"B" %w0, %w2, [%3]\n" \
> >> + "0: ldxr"B" %w2, [%3]\n" \
> >> + "1: stxr"B" %w0, %w1, [%3]\n" \
> >> " cbz %w0, 2f\n" \
> >> " mov %w0, %w4\n" \
> >> "2:\n" \
> >> + " mov %w1, %w2\n" \
> >> " .pushsection .fixup,\"ax\"\n" \
> >> " .align 2\n" \
> >> "3: mov %w0, %w5\n" \
>
> On the second thought looks like we still update *data in case stxr
> fails (or I need more coffee).
I'm on the second cup and I see the same problem. Even if stxr fails, we
fall back through "mov %w1, %w2", so *data is always updated with the
loaded value. Maybe something like below on top of Will's patch:
diff --git a/arch/arm64/kernel/armv8_deprecated.c
b/arch/arm64/kernel/armv8_deprecated.c
index 6039d1eb5912..3fab37b3bc95 100644
--- a/arch/arm64/kernel/armv8_deprecated.c
+++ b/arch/arm64/kernel/armv8_deprecated.c
@@ -286,10 +286,10 @@ static void register_insn_emulation_sysctl(struct
ctl_table *table)
CONFIG_ARM64_PAN) \
"0: ldxr"B" %w2, [%3]\n" \
"1: stxr"B" %w0, %w1, [%3]\n" \
- " cbz %w0, 2f\n" \
- " mov %w0, %w4\n" \
+ " cmp %w0, #0\n" \
+ " csel %w0, %w4, wzr, ne\n" \
+ " csel %w1, %w2, %w1, eq\n" \
"2:\n" \
- " mov %w1, %w2\n" \
" .pushsection .fixup,\"ax\"\n" \
" .align 2\n" \
"3: mov %w0, %w5\n" \
@@ -303,7 +303,7 @@ static void register_insn_emulation_sysctl(struct ctl_table
*table)
ALTERNATIVE("nop", SET_PSTATE_PAN(1), ARM64_HAS_PAN, \
CONFIG_ARM64_PAN) \
: "=&r" (res), "+r" (data), "=&r" (temp) \
- : "r" (addr), "i" (-EAGAIN), "i" (-EFAULT) \
+ : "r" (addr), "r" (-EAGAIN), "i" (-EFAULT) \
: "memory")
#define __user_swp_asm(data, addr, res, temp) \
@@ -342,7 +342,7 @@ static void set_segfault(struct pt_regs *regs, unsigned
long addr)
static int emulate_swpX(unsigned int address, unsigned int *data,
unsigned int type)
{
- unsigned int res = 0;
+ unsigned int res;
if ((type != TYPE_SWPB) && (address & 0x3)) {
/* SWP to unaligned address not permitted */
Since Will is away for a week, I'll wait until he's back to push this
patch.
--
Catalin
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
