Re: [PATCH 0/2] bcache: two emergent fixes for Linux v5.2-rc5 (use-after-scope)

Re: [PATCH 0/2] bcache: two emergent fixes for Linux v5.2-rc5 (use-after-scope)

[Dmitry Vyukov]

Hi,

This is regarding the subj patch:
https://bugzilla.kernel.org/show_bug.cgi?id=203573
https://www.spinics.net/lists/linux-bcache/msg07474.html
(don't see a way to reply to the patch)

This looks like a serious bug that would have been caught by
use-after-scope mode in KASAN given any coverage of the involved code
(i.e. any tests that executes the function once) if I am reading this
correctly.
But use-after-scope detection was removed in:
7771bdbbfd3d kasan: remove use after scope bugs detection.
because it does not catch enough bugs.
Hard to say if this bug is enough rationale to bring use-after-scope
back, but it is a data point. FWIW this bug would have been detected
during patch testing with no debugging required.

Re: [PATCH 0/2] bcache: two emergent fixes for Linux v5.2-rc5 (use-after-scope)

[Coly Li]

On 2019/6/16 6:23 下午, Dmitry Vyukov wrote:
> Hi,
> 
> This is regarding the subj patch:
> https://bugzilla.kernel.org/show_bug.cgi?id=203573
> https://www.spinics.net/lists/linux-bcache/msg07474.html
> (don't see a way to reply to the patch)
> 
> This looks like a serious bug that would have been caught by
> use-after-scope mode in KASAN given any coverage of the involved code
> (i.e. any tests that executes the function once) if I am reading this
> correctly.
> But use-after-scope detection was removed in:
> 7771bdbbfd3d kasan: remove use after scope bugs detection.
> because it does not catch enough bugs.
> Hard to say if this bug is enough rationale to bring use-after-scope
> back, but it is a data point. FWIW this bug would have been detected
> during patch testing with no debugging required.
> 

Hi Dmitry,

I although thought it should be reported by compiler, but no idea why
compiler didn't complain.

Anyway, since now I start to enable KASAN in my testing.

Thanks.

-- 

Coly Li