Hi Aman,
On Fri, Mar 20, 2026 at 10:58:40PM +0530, Aman Dhoot wrote:
> As you told, I bisect the kernel, and this is the log
>
> ****************************************************************
>
> $ git bisect log
> git bisect start
> # status: waiting for both good and bad commits
> # good: [567bd8cbc2fe6b28b78864cbbbc41b0d405eb83c] Linux 6.12.63
> git bisect good 567bd8cbc2fe6b28b78864cbbbc41b0d405eb83c
> # status: waiting for bad commit, 1 good commit known
> # bad: [ff2177382799753070b71747f646963147eabc7c] Linux 6.12.69
> git bisect bad ff2177382799753070b71747f646963147eabc7c
> # good: [ebdbe19336f26ffe799db842d751745098dc11ff] ASoC: renesas: rz-ssi:
> Fix rz_ssi_priv::hw_params_cache::sample_width
> git bisect good ebdbe19336f26ffe799db842d751745098dc11ff
> # bad: [e79b03d386341e85a4f775e0a864e8aa7633a0a2] HID: intel-ish-hid: Use
> dedicated unbound workqueues to prevent resume blocking
> git bisect bad e79b03d386341e85a4f775e0a864e8aa7633a0a2
> # good: [feb28b6827ece47cce585599a00b02ee579532bc] powercap: fix sscanf()
> error return value handling
> git bisect good feb28b6827ece47cce585599a00b02ee579532bc
> # good: [68495f89a19b6835e388b89b2ffecc0c68f9666c] selftests/landlock: Fix
> TCP bind(AF_UNSPEC) test case
> git bisect good 68495f89a19b6835e388b89b2ffecc0c68f9666c
> # good: [4433ddc3700cea880c383a6ddfc0e2ab697f9bdf] EDAC/x38: Fix a resource
> leak in x38_probe1()
> git bisect good 4433ddc3700cea880c383a6ddfc0e2ab697f9bdf
> # bad: [94b010200a3c9a8420a9063344cedbcd71794c8f] LoongArch: dts:
> loongson-2k0500: Add default interrupt controller address cells
> git bisect bad 94b010200a3c9a8420a9063344cedbcd71794c8f
> # good: [654fa76032eee5df9ce8849bdff840595952c63d] mm/page_alloc: make
> percpu_pagelist_high_fraction reads lock-free
> git bisect good 654fa76032eee5df9ce8849bdff840595952c63d
> # bad: [8140ac7c55e75093a01c6110a2c4025fe7177c57] drm/amd: Clean up kfd
> node on surprise disconnect
> git bisect bad 8140ac7c55e75093a01c6110a2c4025fe7177c57
> # good: [df7a49b328928b6d6b174d954d63721d6f3848a2] LoongArch: Fix PMU
> counter allocation for mixed-type event groups
> git bisect good df7a49b328928b6d6b174d954d63721d6f3848a2
> # good: [ae5b1d291c814a2884c3d54a56e83bc99052b1eb] drm/amd/display: Bump
> the HDMI clock to 340MHz
> git bisect good ae5b1d291c814a2884c3d54a56e83bc99052b1eb
> # first bad commit: [8140ac7c55e75093a01c6110a2c4025fe7177c57] drm/amd:
> Clean up kfd node on surprise disconnect
>
> **********************************************************************************************
>
> When the bisect is an end, it provides this output:
>
>
> 8140ac7c55e75093a01c6110a2c4025fe7177c57 is the first bad commit
> commit 8140ac7c55e75093a01c6110a2c4025fe7177c57
> Author: Mario Limonciello (AMD) <[email protected]>
> Date: Wed Jan 7 15:37:28 2026 -0600
>
> drm/amd: Clean up kfd node on surprise disconnect
>
> commit 28695ca09d326461f8078332aa01db516983e8a2 upstream.
>
> When an eGPU is unplugged the KFD topology should also be destroyed
> for that GPU. This never happens because the fini_sw callbacks never
> get to run. Run them manually before calling
> amdgpu_device_ip_fini_early()
> when a device has already been disconnected.
>
> This location is intentionally chosen to make sure that the kfd locking
> refcount doesn't get incremented unintentionally.
>
> Cc: [email protected]
> Closes: https://community.frame.work/t/amd-egpu-on-linux/8691/33
> Signed-off-by: Mario Limonciello (AMD) <[email protected]>
> Reviewed-by: Kent Russell <[email protected]>
> Signed-off-by: Alex Deucher <[email protected]>
> (cherry picked from commit 6a23e7b4332c10f8b56c33a9c5431b52ecff9aab)
> Cc: [email protected]
> Signed-off-by: Greg Kroah-Hartman <[email protected]>
>
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
>
> According to me, this commit exists in the kernel version 6.12.66, and the
> problem also exists in v6.12.66
Thanks for doing that. It looks this is a regression fixed by
f7afda7fcd16 ("drm/amd: Fix hang on amdgpu unload by using
pci_dev_is_disconnected()"), which was backported to as well 6.22.77.
If possible it would be great if you can test that indeed this fixes
the problem. Cf.
https://kernel-team.pages.debian.net/kernel-handbook/ch-common-tasks.html#id-1.6.6.4
Regards,
Salvatore
