Re: [BUG] powerpc: userspace processes crash in ld64.so.2 when running recent linux-next kernels

On 03/08/26 6:40 pm, Venkat Rao Bagalkote wrote:
> Hi,
>
> We are seeing random userspace crashes while building the kernel on
> ppc64le when running recent upstream/mainline or linux-next kernels.
>
> The failures occur in different userspace programs, but the crash
> signature is consistently inside ld64.so.2.
>
> Environment
> -----------
>
> Architecture:
>   ppc64le
>
> Platform:
>   IBM Power systems (LPAR)
>
> Userspace:
>   RHEL userspace
>
> Workload:
>   Kernel build
>
>     make -j32 -s && make modules_install && make install
>
>
> gcc version 14.3.1 20251022 (Red Hat 14.3.1-4) (GCC)
>
> GNU ld version 2.41-63.el10
> ldd (GNU libc) 2.39
>
>
>
>
> Observation
> -----------
>
> The issue is observed when the system is booted with recent upstream
> kernels.
>
> The same LPAR, userspace, toolchain, and source tree are able to build
> successfully when booted with the base OS kernel.
>
> Typical failures include:
>
>   sh
>   fixdep
>   make
>
> The source file being compiled when the failure occurs varies from run
> to run.
>
> Examples
> --------
>
> Example 1:
>
> [ 1277.872746] fixdepsegfault (11) at 29ac0
> [ 1277.872746] fixdep[122982]: nip 3fffbebd8214 lr 3fffbebd7358
> [ 1277.872746] fixdep[122982]: code 1 in ld64.sofffbeba1000+50000]
>
> [ 1277.872793] fixdepcode:
> f90104a8 7ce903a6 f9410060 f9490008
> f9490010 f9490018 39290020 f9490000
>
> [ 1277.872805] fixdep[122982[37214,3fffbeba1000+50000 f9490018 
> <ea628000> 7e731050 f9410030 fa610020
>
> cripts/Makefile.build:289:
> drivers/gpu/drm/amd/amdgpu/../pm/powerplay/smumgr/vegam_smumgr.o]
> Error 139
>
> Example 2:
>
> [Sun Aug 2 12:36:38 2026] sh[115552]: segfaultbcd0
> nip 3fffbd73d004 lr 3fffbd73c148 code 1 in
> ld64.so.2[35004,3fffbd708000+4d000]
>
> Example 3:
>
> [Sun Aug 2 12:40:08 2026] fixdep[134469]: segfault (11) at 
> 2bcd0c3ca004 lr 3fff8c3c9148 code 1 in
> ld64.so.2[35004,3fff8c395000+4d000]
>
> Example 4:
>
> [Mon Aug 3 05:44:20 2026] make[600291]: segfault (11) ip 3fffaf3b5004 
> lr 3fffaf3b4148 code 1 in
> ld64.so.2[35004,3fffaf380000+4d000]
>
> Notes
> -----
>
> - The crashing executable varies (sh, fixdep, make, etc.).
> - The object file being compiled also varies.
> - The crash always appears to occur in ld64.so.2.
> - The issue has been observed on more than one system.
> - A fresh OS installation with ample free memory reproduces the issue.
>
> Let me know, if any more info is needed.
>
> If you happen to fix this issue, please add below tag.
>
> Reported-by: Venkat Rao Bagalkote <[email protected]>
>
>
>

As part of debugging and suggested by Maddy, I reverted the following 
commit:

   263e5159e00a ("powerpc: Fix exit_flags field placement in pt_regs for 
ptrace")

and rebuilt/booted linux-next with only that change reverted.

After booting the reverted kernel, I rebuilt the latest upstream kernel
using the same userspace and toolchain. The build completed
successfully and I did not observe any of the ld64.so.2 userspace
segfaults that were previously seen during kernel compilation.

Kernel used for testing:

   7.2.0-rc5-next-20260731-00001-g79d9c1ce90a9

Top of tree:

   commit 79d9c1ce90a9873422ca4d6d0124a1d1c10a098c
   Revert "powerpc: Fix exit_flags field placement in pt_regs for ptrace"

The upstream kernel build completed successfully:

   INSTALL /lib/modules/7.2.0-rc7/kernel/net/openvswitch/vport-gre.ko
   DEPMOD  /lib/modules/7.2.0-rc7
   INSTALL /boot

Based on this testing, reverting commit 263e5159e00a appears to
eliminate the random userspace crashes observed during kernel builds,
suggesting that the issue may be related to that change.

>
>
> Thanks,
> Venkat.