public inbox for stable@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault
@ 2024-03-12  2:19 Samuel Holland
  2024-03-13  2:53 ` Charlie Jenkins
                   ` (2 more replies)
  0 siblings, 3 replies; 6+ messages in thread
From: Samuel Holland @ 2024-03-12  2:19 UTC (permalink / raw)
  To: Palmer Dabbelt, linux-riscv; +Cc: linux-kernel, Samuel Holland, stable

These macros did not initialize __kr_err, so they could fail even if
the access did not fault.

Cc: stable@vger.kernel.org
Fixes: d464118cdc41 ("riscv: implement __get_kernel_nofault and __put_user_nofault")
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
---
Found while testing the unaligned access speed series[1]. The observed
behavior was that with RISCV_EFFICIENT_UNALIGNED_ACCESS=y, the
copy_from_kernel_nofault() in prepend_copy() failed every time when
filling out /proc/self/mounts, so all of the mount points were "xxx".

I'm surprised this hasn't been seen before. For reference, I'm compiling
with clang 18.

[1]: https://lore.kernel.org/linux-riscv/20240308-disable_misaligned_probe_config-v9-0-a388770ba0ce@rivosinc.com/

 arch/riscv/include/asm/uaccess.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/include/asm/uaccess.h b/arch/riscv/include/asm/uaccess.h
index ec0cab9fbddd..72ec1d9bd3f3 100644
--- a/arch/riscv/include/asm/uaccess.h
+++ b/arch/riscv/include/asm/uaccess.h
@@ -319,7 +319,7 @@ unsigned long __must_check clear_user(void __user *to, unsigned long n)
 
 #define __get_kernel_nofault(dst, src, type, err_label)			\
 do {									\
-	long __kr_err;							\
+	long __kr_err = 0;						\
 									\
 	__get_user_nocheck(*((type *)(dst)), (type *)(src), __kr_err);	\
 	if (unlikely(__kr_err))						\
@@ -328,7 +328,7 @@ do {									\
 
 #define __put_kernel_nofault(dst, src, type, err_label)			\
 do {									\
-	long __kr_err;							\
+	long __kr_err = 0;						\
 									\
 	__put_user_nocheck(*((type *)(src)), (type *)(dst), __kr_err);	\
 	if (unlikely(__kr_err))						\
-- 
2.43.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
* Re: [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault
  2024-03-12  2:19 [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault Samuel Holland
@ 2024-03-13  2:53 ` Charlie Jenkins
  2024-03-13  3:05   ` Samuel Holland
  2024-03-25  7:42 ` Alexandre Ghiti
  2024-03-26 21:10 ` patchwork-bot+linux-riscv
  2 siblings, 1 reply; 6+ messages in thread
From: Charlie Jenkins @ 2024-03-13  2:53 UTC (permalink / raw)
  To: Samuel Holland; +Cc: Palmer Dabbelt, linux-riscv, linux-kernel, stable

On Mon, Mar 11, 2024 at 07:19:13PM -0700, Samuel Holland wrote:
> These macros did not initialize __kr_err, so they could fail even if
> the access did not fault.
> 
> Cc: stable@vger.kernel.org
> Fixes: d464118cdc41 ("riscv: implement __get_kernel_nofault and __put_user_nofault")
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> ---
> Found while testing the unaligned access speed series[1]. The observed
> behavior was that with RISCV_EFFICIENT_UNALIGNED_ACCESS=y, the
> copy_from_kernel_nofault() in prepend_copy() failed every time when
> filling out /proc/self/mounts, so all of the mount points were "xxx".
> 
> I'm surprised this hasn't been seen before. For reference, I'm compiling
> with clang 18.
> 
> [1]: https://lore.kernel.org/linux-riscv/20240308-disable_misaligned_probe_config-v9-0-a388770ba0ce@rivosinc.com/
> 
>  arch/riscv/include/asm/uaccess.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/riscv/include/asm/uaccess.h b/arch/riscv/include/asm/uaccess.h
> index ec0cab9fbddd..72ec1d9bd3f3 100644
> --- a/arch/riscv/include/asm/uaccess.h
> +++ b/arch/riscv/include/asm/uaccess.h
> @@ -319,7 +319,7 @@ unsigned long __must_check clear_user(void __user *to, unsigned long n)
>  
>  #define __get_kernel_nofault(dst, src, type, err_label)			\
>  do {									\
> -	long __kr_err;							\
> +	long __kr_err = 0;						\
>  									\
>  	__get_user_nocheck(*((type *)(dst)), (type *)(src), __kr_err);	\
>  	if (unlikely(__kr_err))						\
> @@ -328,7 +328,7 @@ do {									\
>  
>  #define __put_kernel_nofault(dst, src, type, err_label)			\
>  do {									\
> -	long __kr_err;							\
> +	long __kr_err = 0;						\
>  									\
>  	__put_user_nocheck(*((type *)(src)), (type *)(dst), __kr_err);	\
>  	if (unlikely(__kr_err))						\
> -- 
> 2.43.1
> 
> 
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv

I am not able to reproduce this using Clang 18 with
RISCV_EFFICIENT_UNALIGNED_ACCESS=y on 6.8. However I can see how this
could be an issue.

Going down the rabbit hold of macros here, I end up at
arch/riscv/include/asm/asm-extable.h where the register that hold 'err'
is written into the __ex_table section:

#define EX_DATA_REG(reg, gpr)						\
	"((.L__gpr_num_" #gpr ") << " __stringify(EX_DATA_REG_##reg##_SHIFT) ")"

#define _ASM_EXTABLE_UACCESS_ERR_ZERO(insn, fixup, err, zero)		\
	__DEFINE_ASM_GPR_NUMS						\
	__ASM_EXTABLE_RAW(#insn, #fixup, 				\
			  __stringify(EX_TYPE_UACCESS_ERR_ZERO),	\
			  "("						\
			    EX_DATA_REG(ERR, err) " | "			\
			    EX_DATA_REG(ZERO, zero)			\
			  ")")

I am wondering if setting this value to zero solves the problem by
hiding another issue. It seems like this shouldn't need to be
initialized to zero, however I am lost as to how this extable setup
works so perhaps this is the proper solution.

- Charlie


^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault
  2024-03-13  2:53 ` Charlie Jenkins
@ 2024-03-13  3:05   ` Samuel Holland
  2024-03-13  3:22     ` Charlie Jenkins
  0 siblings, 1 reply; 6+ messages in thread
From: Samuel Holland @ 2024-03-13  3:05 UTC (permalink / raw)
  To: Charlie Jenkins; +Cc: Palmer Dabbelt, linux-riscv, linux-kernel, stable

On 2024-03-12 9:53 PM, Charlie Jenkins wrote:
> On Mon, Mar 11, 2024 at 07:19:13PM -0700, Samuel Holland wrote:
>> These macros did not initialize __kr_err, so they could fail even if
>> the access did not fault.
>>
>> Cc: stable@vger.kernel.org
>> Fixes: d464118cdc41 ("riscv: implement __get_kernel_nofault and __put_user_nofault")
>> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
>> ---
>> Found while testing the unaligned access speed series[1]. The observed
>> behavior was that with RISCV_EFFICIENT_UNALIGNED_ACCESS=y, the
>> copy_from_kernel_nofault() in prepend_copy() failed every time when
>> filling out /proc/self/mounts, so all of the mount points were "xxx".
>>
>> I'm surprised this hasn't been seen before. For reference, I'm compiling
>> with clang 18.
>>
>> [1]: https://lore.kernel.org/linux-riscv/20240308-disable_misaligned_probe_config-v9-0-a388770ba0ce@rivosinc.com/
>>
>>  arch/riscv/include/asm/uaccess.h | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/riscv/include/asm/uaccess.h b/arch/riscv/include/asm/uaccess.h
>> index ec0cab9fbddd..72ec1d9bd3f3 100644
>> --- a/arch/riscv/include/asm/uaccess.h
>> +++ b/arch/riscv/include/asm/uaccess.h
>> @@ -319,7 +319,7 @@ unsigned long __must_check clear_user(void __user *to, unsigned long n)
>>  
>>  #define __get_kernel_nofault(dst, src, type, err_label)			\
>>  do {									\
>> -	long __kr_err;							\
>> +	long __kr_err = 0;						\
>>  									\
>>  	__get_user_nocheck(*((type *)(dst)), (type *)(src), __kr_err);	\
>>  	if (unlikely(__kr_err))						\
>> @@ -328,7 +328,7 @@ do {									\
>>  
>>  #define __put_kernel_nofault(dst, src, type, err_label)			\
>>  do {									\
>> -	long __kr_err;							\
>> +	long __kr_err = 0;						\
>>  									\
>>  	__put_user_nocheck(*((type *)(src)), (type *)(dst), __kr_err);	\
>>  	if (unlikely(__kr_err))						\
>> -- 
>> 2.43.1
>>
>>
>> _______________________________________________
>> linux-riscv mailing list
>> linux-riscv@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-riscv
> 
> I am not able to reproduce this using Clang 18 with
> RISCV_EFFICIENT_UNALIGNED_ACCESS=y on 6.8. However I can see how this
> could be an issue.
> 
> Going down the rabbit hold of macros here, I end up at
> arch/riscv/include/asm/asm-extable.h where the register that hold 'err'
> is written into the __ex_table section:
> 
> #define EX_DATA_REG(reg, gpr)						\
> 	"((.L__gpr_num_" #gpr ") << " __stringify(EX_DATA_REG_##reg##_SHIFT) ")"
> 
> #define _ASM_EXTABLE_UACCESS_ERR_ZERO(insn, fixup, err, zero)		\
> 	__DEFINE_ASM_GPR_NUMS						\
> 	__ASM_EXTABLE_RAW(#insn, #fixup, 				\
> 			  __stringify(EX_TYPE_UACCESS_ERR_ZERO),	\
> 			  "("						\
> 			    EX_DATA_REG(ERR, err) " | "			\
> 			    EX_DATA_REG(ZERO, zero)			\
> 			  ")")
> 
> I am wondering if setting this value to zero solves the problem by
> hiding another issue. It seems like this shouldn't need to be
> initialized to zero, however I am lost as to how this extable setup
> works so perhaps this is the proper solution.

extable works by running the handler (selected by EX_TYPE_*) if some exception
occurs while executing that instruction -- see the calls to fixup_exception() in
fault.c and traps.c. If there is no exception, then the handler does not run,
and the err register is not written by ex_handler_uaccess_err_zero().

If you look at __get_user_asm(), you can see that the err register is not
touched by the assembly code at all -- the only reference to %0 is in the
extable entry. So if the macro that declares the error variable doesn't
initialize it, nothing will.

Compare __get_user() and __put_user() which do initialize their error variable.

Regards,
Samuel


^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault
  2024-03-13  3:05   ` Samuel Holland
@ 2024-03-13  3:22     ` Charlie Jenkins
  0 siblings, 0 replies; 6+ messages in thread
From: Charlie Jenkins @ 2024-03-13  3:22 UTC (permalink / raw)
  To: Samuel Holland; +Cc: Palmer Dabbelt, linux-riscv, linux-kernel, stable

On Tue, Mar 12, 2024 at 10:05:37PM -0500, Samuel Holland wrote:
> On 2024-03-12 9:53 PM, Charlie Jenkins wrote:
> > On Mon, Mar 11, 2024 at 07:19:13PM -0700, Samuel Holland wrote:
> >> These macros did not initialize __kr_err, so they could fail even if
> >> the access did not fault.
> >>
> >> Cc: stable@vger.kernel.org
> >> Fixes: d464118cdc41 ("riscv: implement __get_kernel_nofault and __put_user_nofault")
> >> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> >> ---
> >> Found while testing the unaligned access speed series[1]. The observed
> >> behavior was that with RISCV_EFFICIENT_UNALIGNED_ACCESS=y, the
> >> copy_from_kernel_nofault() in prepend_copy() failed every time when
> >> filling out /proc/self/mounts, so all of the mount points were "xxx".
> >>
> >> I'm surprised this hasn't been seen before. For reference, I'm compiling
> >> with clang 18.
> >>
> >> [1]: https://lore.kernel.org/linux-riscv/20240308-disable_misaligned_probe_config-v9-0-a388770ba0ce@rivosinc.com/
> >>
> >>  arch/riscv/include/asm/uaccess.h | 4 ++--
> >>  1 file changed, 2 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/arch/riscv/include/asm/uaccess.h b/arch/riscv/include/asm/uaccess.h
> >> index ec0cab9fbddd..72ec1d9bd3f3 100644
> >> --- a/arch/riscv/include/asm/uaccess.h
> >> +++ b/arch/riscv/include/asm/uaccess.h
> >> @@ -319,7 +319,7 @@ unsigned long __must_check clear_user(void __user *to, unsigned long n)
> >>  
> >>  #define __get_kernel_nofault(dst, src, type, err_label)			\
> >>  do {									\
> >> -	long __kr_err;							\
> >> +	long __kr_err = 0;						\
> >>  									\
> >>  	__get_user_nocheck(*((type *)(dst)), (type *)(src), __kr_err);	\
> >>  	if (unlikely(__kr_err))						\
> >> @@ -328,7 +328,7 @@ do {									\
> >>  
> >>  #define __put_kernel_nofault(dst, src, type, err_label)			\
> >>  do {									\
> >> -	long __kr_err;							\
> >> +	long __kr_err = 0;						\
> >>  									\
> >>  	__put_user_nocheck(*((type *)(src)), (type *)(dst), __kr_err);	\
> >>  	if (unlikely(__kr_err))						\
> >> -- 
> >> 2.43.1
> >>
> >>
> >> _______________________________________________
> >> linux-riscv mailing list
> >> linux-riscv@lists.infradead.org
> >> http://lists.infradead.org/mailman/listinfo/linux-riscv
> > 
> > I am not able to reproduce this using Clang 18 with
> > RISCV_EFFICIENT_UNALIGNED_ACCESS=y on 6.8. However I can see how this
> > could be an issue.
> > 
> > Going down the rabbit hold of macros here, I end up at
> > arch/riscv/include/asm/asm-extable.h where the register that hold 'err'
> > is written into the __ex_table section:
> > 
> > #define EX_DATA_REG(reg, gpr)						\
> > 	"((.L__gpr_num_" #gpr ") << " __stringify(EX_DATA_REG_##reg##_SHIFT) ")"
> > 
> > #define _ASM_EXTABLE_UACCESS_ERR_ZERO(insn, fixup, err, zero)		\
> > 	__DEFINE_ASM_GPR_NUMS						\
> > 	__ASM_EXTABLE_RAW(#insn, #fixup, 				\
> > 			  __stringify(EX_TYPE_UACCESS_ERR_ZERO),	\
> > 			  "("						\
> > 			    EX_DATA_REG(ERR, err) " | "			\
> > 			    EX_DATA_REG(ZERO, zero)			\
> > 			  ")")
> > 
> > I am wondering if setting this value to zero solves the problem by
> > hiding another issue. It seems like this shouldn't need to be
> > initialized to zero, however I am lost as to how this extable setup
> > works so perhaps this is the proper solution.
> 
> extable works by running the handler (selected by EX_TYPE_*) if some exception
> occurs while executing that instruction -- see the calls to fixup_exception() in
> fault.c and traps.c. If there is no exception, then the handler does not run,
> and the err register is not written by ex_handler_uaccess_err_zero().

Hmm okay I understand thank you for explaining that. It's interesting to
me that in __get_user_asm 'err' is set as a read/write variable even
though __get_user_asm doesn't write to it. However, it seems like
changing it to a write-only variable the compiler incorrectly optimizes
err, and the kernel fails to boot.

> 
> If you look at __get_user_asm(), you can see that the err register is not
> touched by the assembly code at all -- the only reference to %0 is in the
> extable entry. So if the macro that declares the error variable doesn't
> initialize it, nothing will.


Reviewed-by: Charlie Jenkins <charlie@rivosinc.com>

> 
> Compare __get_user() and __put_user() which do initialize their error variable.
> 
> Regards,
> Samuel
> 

^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault
  2024-03-12  2:19 [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault Samuel Holland
  2024-03-13  2:53 ` Charlie Jenkins
@ 2024-03-25  7:42 ` Alexandre Ghiti
  2024-03-26 21:10 ` patchwork-bot+linux-riscv
  2 siblings, 0 replies; 6+ messages in thread
From: Alexandre Ghiti @ 2024-03-25  7:42 UTC (permalink / raw)
  To: Samuel Holland, Palmer Dabbelt, linux-riscv; +Cc: linux-kernel, stable

Hi Samuel,

On 12/03/2024 03:19, Samuel Holland wrote:
> These macros did not initialize __kr_err, so they could fail even if
> the access did not fault.
>
> Cc: stable@vger.kernel.org
> Fixes: d464118cdc41 ("riscv: implement __get_kernel_nofault and __put_user_nofault")
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> ---
> Found while testing the unaligned access speed series[1]. The observed
> behavior was that with RISCV_EFFICIENT_UNALIGNED_ACCESS=y, the
> copy_from_kernel_nofault() in prepend_copy() failed every time when
> filling out /proc/self/mounts, so all of the mount points were "xxx".
>
> I'm surprised this hasn't been seen before. For reference, I'm compiling
> with clang 18.
>
> [1]: https://lore.kernel.org/linux-riscv/20240308-disable_misaligned_probe_config-v9-0-a388770ba0ce@rivosinc.com/
>
>   arch/riscv/include/asm/uaccess.h | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/arch/riscv/include/asm/uaccess.h b/arch/riscv/include/asm/uaccess.h
> index ec0cab9fbddd..72ec1d9bd3f3 100644
> --- a/arch/riscv/include/asm/uaccess.h
> +++ b/arch/riscv/include/asm/uaccess.h
> @@ -319,7 +319,7 @@ unsigned long __must_check clear_user(void __user *to, unsigned long n)
>   
>   #define __get_kernel_nofault(dst, src, type, err_label)			\
>   do {									\
> -	long __kr_err;							\
> +	long __kr_err = 0;						\
>   									\
>   	__get_user_nocheck(*((type *)(dst)), (type *)(src), __kr_err);	\
>   	if (unlikely(__kr_err))						\
> @@ -328,7 +328,7 @@ do {									\
>   
>   #define __put_kernel_nofault(dst, src, type, err_label)			\
>   do {									\
> -	long __kr_err;							\
> +	long __kr_err = 0;						\
>   									\
>   	__put_user_nocheck(*((type *)(src)), (type *)(dst), __kr_err);	\
>   	if (unlikely(__kr_err))						\


That's a good a catch! Surprising indeed that we did not find this 
before, that's the second bug in the uaccess routines in the last 3 (or 
4) releases, maybe we should improve our coverage here, I'll see what we 
can do.

You can add:

Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>

Thanks,

Alex


^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault
  2024-03-12  2:19 [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault Samuel Holland
  2024-03-13  2:53 ` Charlie Jenkins
  2024-03-25  7:42 ` Alexandre Ghiti
@ 2024-03-26 21:10 ` patchwork-bot+linux-riscv
  2 siblings, 0 replies; 6+ messages in thread
From: patchwork-bot+linux-riscv @ 2024-03-26 21:10 UTC (permalink / raw)
  To: Samuel Holland; +Cc: linux-riscv, palmer, linux-kernel, stable

Hello:

This patch was applied to riscv/linux.git (fixes)
by Palmer Dabbelt <palmer@rivosinc.com>:

On Mon, 11 Mar 2024 19:19:13 -0700 you wrote:
> These macros did not initialize __kr_err, so they could fail even if
> the access did not fault.
> 
> Cc: stable@vger.kernel.org
> Fixes: d464118cdc41 ("riscv: implement __get_kernel_nofault and __put_user_nofault")
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> 
> [...]

Here is the summary with links:
  - riscv: Fix spurious errors from __get/put_kernel_nofault
    https://git.kernel.org/riscv/c/d080a08b06b6

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html



^ permalink raw reply	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2024-03-26 21:10 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-03-12  2:19 [PATCH] riscv: Fix spurious errors from __get/put_kernel_nofault Samuel Holland
2024-03-13  2:53 ` Charlie Jenkins
2024-03-13  3:05   ` Samuel Holland
2024-03-13  3:22     ` Charlie Jenkins
2024-03-25  7:42 ` Alexandre Ghiti
2024-03-26 21:10 ` patchwork-bot+linux-riscv

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox