[PATCH] MIPS: Fix a long-standing mistake in mips_atomic_set()

[PATCH] MIPS: Fix a long-standing mistake in mips_atomic_set()

[Huacai Chen]

This mistake comes from the commit f1e39a4a616cd99 ("MIPS: Rewrite
sysmips(MIPS_ATOMIC_SET, ...) in C with inline assembler"). In the
common case 'bnez' should be 'beqz' (as same as older kernels before
2.6.32), otherwise this syscall may cause an endless loop.

Cc: stable@vger.kernel.org
Signed-off-by: Huacai Chen <chenhc@lemote.com>
---
 arch/mips/kernel/syscall.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/mips/kernel/syscall.c b/arch/mips/kernel/syscall.c
index 1dfa7f5..95e1b30 100644
--- a/arch/mips/kernel/syscall.c
+++ b/arch/mips/kernel/syscall.c
@@ -134,7 +134,7 @@ static inline int mips_atomic_set(unsigned long addr, unsigned long new)
 		"1:	ll	%[old], (%[addr])			\n"
 		"	move	%[tmp], %[new]				\n"
 		"2:	sc	%[tmp], (%[addr])			\n"
-		"	bnez	%[tmp], 4f				\n"
+		"	beqz	%[tmp], 4f				\n"
 		"3:							\n"
 		"	.insn						\n"
 		"	.subsection 2					\n"
-- 
2.7.0

Re: [PATCH] MIPS: Fix a long-standing mistake in mips_atomic_set()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 1325 bytes --]

Hi Huacai,

On Thu, Jun 22, 2017 at 06:45:45PM +0800, Huacai Chen wrote:
> This mistake comes from the commit f1e39a4a616cd99 ("MIPS: Rewrite
> sysmips(MIPS_ATOMIC_SET, ...) in C with inline assembler"). In the
> common case 'bnez' should be 'beqz' (as same as older kernels before
> 2.6.32), otherwise this syscall may cause an endless loop.
> 
> Cc: stable@vger.kernel.org
> Signed-off-by: Huacai Chen <chenhc@lemote.com>

Thats a coincidence. 8 years its been broken and I submitted an
identical patch only a few weeks ago, along with some other related
fixes:

https://patchwork.linux-mips.org/project/linux-mips/list/?series=313&state=*

Cheers
James

> ---
>  arch/mips/kernel/syscall.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/mips/kernel/syscall.c b/arch/mips/kernel/syscall.c
> index 1dfa7f5..95e1b30 100644
> --- a/arch/mips/kernel/syscall.c
> +++ b/arch/mips/kernel/syscall.c
> @@ -134,7 +134,7 @@ static inline int mips_atomic_set(unsigned long addr, unsigned long new)
>  		"1:	ll	%[old], (%[addr])			\n"
>  		"	move	%[tmp], %[new]				\n"
>  		"2:	sc	%[tmp], (%[addr])			\n"
> -		"	bnez	%[tmp], 4f				\n"
> +		"	beqz	%[tmp], 4f				\n"
>  		"3:							\n"
>  		"	.insn						\n"
>  		"	.subsection 2					\n"
> -- 
> 2.7.0
> 
> 
> 
> 

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH] MIPS: Fix a long-standing mistake in mips_atomic_set()

[Ralf Baechle]

On Thu, Jun 22, 2017 at 01:31:59PM +0100, James Hogan wrote:

> Hi Huacai,
> 
> On Thu, Jun 22, 2017 at 06:45:45PM +0800, Huacai Chen wrote:
> > This mistake comes from the commit f1e39a4a616cd99 ("MIPS: Rewrite
> > sysmips(MIPS_ATOMIC_SET, ...) in C with inline assembler"). In the
> > common case 'bnez' should be 'beqz' (as same as older kernels before
> > 2.6.32), otherwise this syscall may cause an endless loop.
> > 
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Huacai Chen <chenhc@lemote.com>
> 
> Thats a coincidence. 8 years its been broken and I submitted an
> identical patch only a few weeks ago, along with some other related
> fixes:
> 
> https://patchwork.linux-mips.org/project/linux-mips/list/?series=313&state=*

I take it as a proof that everybody is using LL/SC (or the LL/SC emulation)
these days and sysmips(MIPS_ATOMIC_SET) finally has been obsoleted and is
only useful for stoneage binary compatibility.

Which is really good.  Unless people are using silly hacks such as $k0/$k1
being overwriten by exception handlers ...

  Ralf