[Bug dynamic-link/23790] Data race in _dl_profile_fixup with reloc_result update from multiple threads.

[Bug dynamic-link/23790] Data race in _dl_profile_fixup with reloc_result update from multiple threads.

[carlos at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=23790

--- Comment #1 from Carlos O'Donell <carlos at redhat dot com> ---
This issue is only relevant when the PLT entry is a multi-word entry and
therefore needs multiple writes to complete.

I think that IA64 and HPPA (32-bit) are impacted by this issue because of the
function descriptor in the PLT.

The power code also likely suffers from this, but in binutils the PLT update
sequence is designed to be "thread safe" using architecture-related interlocks.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/23790] Data race in _dl_profile_fixup with reloc_result update from multiple threads.

[carlos at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=23790

--- Comment #2 from Carlos O'Donell <carlos at redhat dot com> ---
(In reply to Carlos O'Donell from comment #1)
> This issue is only relevant when the PLT entry is a multi-word entry and
> therefore needs multiple writes to complete.
> 
> I think that IA64 and HPPA (32-bit) are impacted by this issue because of
> the function descriptor in the PLT.
> 
> The power code also likely suffers from this, but in binutils the PLT update
> sequence is designed to be "thread safe" using architecture-related
> interlocks.

What I wrote here applies both to the traditional PLT fixup in the dynamic
loader *and* the fixup of the cached entries while profiling. This particular
bug is about only the fixup of the cached entries which are updated when
profiling. In this case glibc is entirely responsible for the update.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/23790] Data race in _dl_profile_fixup with reloc_result update from multiple threads.

[johnkaitlyn95 at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=23790

Kaitlyn John <johnkaitlyn95 at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |johnkaitlyn95 at gmail dot com

--- Comment #3 from Kaitlyn John <johnkaitlyn95 at gmail dot com> ---
Here is a race condition in dlprofile_fixup with reloc_result update from
multiple threads. The problem is that dlprofile_fixup can be executed from
different threads concurrently and the code which handles the result of the
relocation is not protected against concurrent writes, which leads to data
races on the reloc_result variable.
https://iheartcbdoil.com/cbd-benefits/how-cbd-can-be-helpful-for-heart-diseases/

-- 
You are receiving this mail because:
You are on the CC list for the bug.