[5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[Barry K. Nathan]

To be very clear, I want to emphasize up front that this is regarding
the *current 5.10.y queue*, and not any 5.10.y release. This does not
affect any currently released 5.10.y kernel.

When I apply the current 5.10.y queue on top of 5.10.252, the result
is a kernel that consistently freezes on shutdown, both when running
directly on my Lenovo ThinkPad T14 Gen 1 (Debian 12 bookworm running on
an Intel Core i5-10310U) and when running in a virt-manager VM (I didn't
put much thought into it and just happened to pick a Debian 13 trixie VM
for my testing).

(Just once in my testing, there was a visible kernel panic, but there
was also screen corruption, and it looks to me like it might've frozen
up partway through the panic being printed on the screen, so I'm not
sure how useful it would be. I did take a photo so I guess I could
post it up somewhere if it might be useful, or maybe I'll try to type
it in from the photo later today or tomorrow.)

I bisected manually, and it turned out to be this patch:
"driver core: platform: use bus_type functions"
(driver-core-platform-use-bus_type-functions.patch)

If I apply all of the patches in queue-5.10/series up to but not
including that patch, the resulting kernel shuts down fine. Once I add
that patch, the kernel freezes at shutdown.

Please let me know if there's anything else I should try, or any other
info I should try to provide. (The problem does not occur on any of
the recent stable releases, nor does it happen on 7.0-rc3.)

-- 
-Barry K. Nathan  <barryn@pobox.com>

Re: [5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[Greg Kroah-Hartman]

On Mon, Mar 09, 2026 at 07:03:33AM -0700, Barry K. Nathan wrote:
> To be very clear, I want to emphasize up front that this is regarding
> the *current 5.10.y queue*, and not any 5.10.y release. This does not
> affect any currently released 5.10.y kernel.
> 
> When I apply the current 5.10.y queue on top of 5.10.252, the result
> is a kernel that consistently freezes on shutdown, both when running
> directly on my Lenovo ThinkPad T14 Gen 1 (Debian 12 bookworm running on
> an Intel Core i5-10310U) and when running in a virt-manager VM (I didn't
> put much thought into it and just happened to pick a Debian 13 trixie VM
> for my testing).
> 
> (Just once in my testing, there was a visible kernel panic, but there
> was also screen corruption, and it looks to me like it might've frozen
> up partway through the panic being printed on the screen, so I'm not
> sure how useful it would be. I did take a photo so I guess I could
> post it up somewhere if it might be useful, or maybe I'll try to type
> it in from the photo later today or tomorrow.)
> 
> I bisected manually, and it turned out to be this patch:
> "driver core: platform: use bus_type functions"
> (driver-core-platform-use-bus_type-functions.patch)

That patch has been dropped from the queue earlier this morning, so all
should be fine.

thanks,

greg k-h

Re: [5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[Barry K. Nathan]

On 3/9/26 07:43, Greg Kroah-Hartman wrote:
> On Mon, Mar 09, 2026 at 07:03:33AM -0700, Barry K. Nathan wrote:
>> To be very clear, I want to emphasize up front that this is regarding
>> the *current 5.10.y queue*, and not any 5.10.y release. This does not
>> affect any currently released 5.10.y kernel.
>>
>> When I apply the current 5.10.y queue on top of 5.10.252, the result
>> is a kernel that consistently freezes on shutdown, both when running
>> directly on my Lenovo ThinkPad T14 Gen 1 (Debian 12 bookworm running on
>> an Intel Core i5-10310U) and when running in a virt-manager VM (I didn't
>> put much thought into it and just happened to pick a Debian 13 trixie VM
>> for my testing).
>>
>> (Just once in my testing, there was a visible kernel panic, but there
>> was also screen corruption, and it looks to me like it might've frozen
>> up partway through the panic being printed on the screen, so I'm not
>> sure how useful it would be. I did take a photo so I guess I could
>> post it up somewhere if it might be useful, or maybe I'll try to type
>> it in from the photo later today or tomorrow.)
>>
>> I bisected manually, and it turned out to be this patch:
>> "driver core: platform: use bus_type functions"
>> (driver-core-platform-use-bus_type-functions.patch)
> 
> That patch has been dropped from the queue earlier this morning, so all
> should be fine.
> 
> thanks,
> 
> greg k-h

$ git pull
Already up to date.
$ fgrep bus_type queue-5.10/series
driver-core-platform-use-bus_type-functions.patch

It doesn't look like the patch has been dropped yet from the
5.10.y queue in the stable-queue git repo on kernel.org.

(At least as of this writing, around 6:30 PM PDT, you can go to
https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-5.10
and see that the patch is still there.)


By the way, if dropping this patch turns out not to be the correct course
of action, I think backporting one additional patch will fix it.

I applied the following patch directly from mainline to test it; it
applied with an offset but no fuzz. I did some brief testing and it
seems to fix the problem (no more freezing on shutdown).

mainline commit 46e85af0cc53f35584e00bb5db7db6893d0e16e5
driver core: platform: don't oops in platform_shutdown() on unbound devices

-- 
-Barry K. Nathan  <barryn@pobox.com>

Re: [5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[Greg Kroah-Hartman]

On Mon, Mar 09, 2026 at 06:34:47PM -0700, Barry K. Nathan wrote:
> On 3/9/26 07:43, Greg Kroah-Hartman wrote:
> > On Mon, Mar 09, 2026 at 07:03:33AM -0700, Barry K. Nathan wrote:
> > > To be very clear, I want to emphasize up front that this is regarding
> > > the *current 5.10.y queue*, and not any 5.10.y release. This does not
> > > affect any currently released 5.10.y kernel.
> > > 
> > > When I apply the current 5.10.y queue on top of 5.10.252, the result
> > > is a kernel that consistently freezes on shutdown, both when running
> > > directly on my Lenovo ThinkPad T14 Gen 1 (Debian 12 bookworm running on
> > > an Intel Core i5-10310U) and when running in a virt-manager VM (I didn't
> > > put much thought into it and just happened to pick a Debian 13 trixie VM
> > > for my testing).
> > > 
> > > (Just once in my testing, there was a visible kernel panic, but there
> > > was also screen corruption, and it looks to me like it might've frozen
> > > up partway through the panic being printed on the screen, so I'm not
> > > sure how useful it would be. I did take a photo so I guess I could
> > > post it up somewhere if it might be useful, or maybe I'll try to type
> > > it in from the photo later today or tomorrow.)
> > > 
> > > I bisected manually, and it turned out to be this patch:
> > > "driver core: platform: use bus_type functions"
> > > (driver-core-platform-use-bus_type-functions.patch)
> > 
> > That patch has been dropped from the queue earlier this morning, so all
> > should be fine.
> > 
> > thanks,
> > 
> > greg k-h
> 
> $ git pull
> Already up to date.
> $ fgrep bus_type queue-5.10/series
> driver-core-platform-use-bus_type-functions.patch
> 
> It doesn't look like the patch has been dropped yet from the
> 5.10.y queue in the stable-queue git repo on kernel.org.
> 
> (At least as of this writing, around 6:30 PM PDT, you can go to
> https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-5.10
> and see that the patch is still there.)
> 
> 
> By the way, if dropping this patch turns out not to be the correct course
> of action, I think backporting one additional patch will fix it.
> 
> I applied the following patch directly from mainline to test it; it
> applied with an offset but no fuzz. I did some brief testing and it
> seems to fix the problem (no more freezing on shutdown).
> 
> mainline commit 46e85af0cc53f35584e00bb5db7db6893d0e16e5
> driver core: platform: don't oops in platform_shutdown() on unbound devices

Ugh, my fault, I was confusing this with a different patch.  I'll go
drop all of these now, thanks for confirming.

greg k-h

Re: [5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 3417 bytes --]

Hello Greg,

On Tue, Mar 10, 2026 at 08:07:33AM +0100, Greg Kroah-Hartman wrote:
> On Mon, Mar 09, 2026 at 06:34:47PM -0700, Barry K. Nathan wrote:
> > On 3/9/26 07:43, Greg Kroah-Hartman wrote:
> > > On Mon, Mar 09, 2026 at 07:03:33AM -0700, Barry K. Nathan wrote:
> > > > To be very clear, I want to emphasize up front that this is regarding
> > > > the *current 5.10.y queue*, and not any 5.10.y release. This does not
> > > > affect any currently released 5.10.y kernel.
> > > > 
> > > > When I apply the current 5.10.y queue on top of 5.10.252, the result
> > > > is a kernel that consistently freezes on shutdown, both when running
> > > > directly on my Lenovo ThinkPad T14 Gen 1 (Debian 12 bookworm running on
> > > > an Intel Core i5-10310U) and when running in a virt-manager VM (I didn't
> > > > put much thought into it and just happened to pick a Debian 13 trixie VM
> > > > for my testing).
> > > > 
> > > > (Just once in my testing, there was a visible kernel panic, but there
> > > > was also screen corruption, and it looks to me like it might've frozen
> > > > up partway through the panic being printed on the screen, so I'm not
> > > > sure how useful it would be. I did take a photo so I guess I could
> > > > post it up somewhere if it might be useful, or maybe I'll try to type
> > > > it in from the photo later today or tomorrow.)
> > > > 
> > > > I bisected manually, and it turned out to be this patch:
> > > > "driver core: platform: use bus_type functions"
> > > > (driver-core-platform-use-bus_type-functions.patch)
> > > 
> > > That patch has been dropped from the queue earlier this morning, so all
> > > should be fine.
> > > 
> > > thanks,
> > > 
> > > greg k-h
> > 
> > $ git pull
> > Already up to date.
> > $ fgrep bus_type queue-5.10/series
> > driver-core-platform-use-bus_type-functions.patch
> > 
> > It doesn't look like the patch has been dropped yet from the
> > 5.10.y queue in the stable-queue git repo on kernel.org.
> > 
> > (At least as of this writing, around 6:30 PM PDT, you can go to
> > https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-5.10
> > and see that the patch is still there.)
> > 
> > 
> > By the way, if dropping this patch turns out not to be the correct course
> > of action, I think backporting one additional patch will fix it.
> > 
> > I applied the following patch directly from mainline to test it; it
> > applied with an offset but no fuzz. I did some brief testing and it
> > seems to fix the problem (no more freezing on shutdown).
> > 
> > mainline commit 46e85af0cc53f35584e00bb5db7db6893d0e16e5
> > driver core: platform: don't oops in platform_shutdown() on unbound devices
> 
> Ugh, my fault, I was confusing this with a different patch.  I'll go
> drop all of these now, thanks for confirming.

I'm unsure why you wrote to "drop all of these" now that the reason for
the breakage is explained and the solution is to pick another patch from
mainline. If you don't really need
driver-core-platform-use-bus_type-functions.patch that's fine, but
otherwise just add 46e85af0cc53f35584e00bb5db7db6893d0e16e5?

I somehow expected that the patch selection procedure for stable notices
if you pick 9c30921fe799 ("driver core: platform: use bus_type
functions") that there are later changes declaring to fix this commit?!

Best regards
Uwe

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [5.10.y] "driver core: platform: use bus_type functions" causes freeze on shutdown

[Greg Kroah-Hartman]

On Tue, Mar 10, 2026 at 10:22:14AM +0100, Uwe Kleine-König wrote:
> Hello Greg,
> 
> On Tue, Mar 10, 2026 at 08:07:33AM +0100, Greg Kroah-Hartman wrote:
> > On Mon, Mar 09, 2026 at 06:34:47PM -0700, Barry K. Nathan wrote:
> > > On 3/9/26 07:43, Greg Kroah-Hartman wrote:
> > > > On Mon, Mar 09, 2026 at 07:03:33AM -0700, Barry K. Nathan wrote:
> > > > > To be very clear, I want to emphasize up front that this is regarding
> > > > > the *current 5.10.y queue*, and not any 5.10.y release. This does not
> > > > > affect any currently released 5.10.y kernel.
> > > > > 
> > > > > When I apply the current 5.10.y queue on top of 5.10.252, the result
> > > > > is a kernel that consistently freezes on shutdown, both when running
> > > > > directly on my Lenovo ThinkPad T14 Gen 1 (Debian 12 bookworm running on
> > > > > an Intel Core i5-10310U) and when running in a virt-manager VM (I didn't
> > > > > put much thought into it and just happened to pick a Debian 13 trixie VM
> > > > > for my testing).
> > > > > 
> > > > > (Just once in my testing, there was a visible kernel panic, but there
> > > > > was also screen corruption, and it looks to me like it might've frozen
> > > > > up partway through the panic being printed on the screen, so I'm not
> > > > > sure how useful it would be. I did take a photo so I guess I could
> > > > > post it up somewhere if it might be useful, or maybe I'll try to type
> > > > > it in from the photo later today or tomorrow.)
> > > > > 
> > > > > I bisected manually, and it turned out to be this patch:
> > > > > "driver core: platform: use bus_type functions"
> > > > > (driver-core-platform-use-bus_type-functions.patch)
> > > > 
> > > > That patch has been dropped from the queue earlier this morning, so all
> > > > should be fine.
> > > > 
> > > > thanks,
> > > > 
> > > > greg k-h
> > > 
> > > $ git pull
> > > Already up to date.
> > > $ fgrep bus_type queue-5.10/series
> > > driver-core-platform-use-bus_type-functions.patch
> > > 
> > > It doesn't look like the patch has been dropped yet from the
> > > 5.10.y queue in the stable-queue git repo on kernel.org.
> > > 
> > > (At least as of this writing, around 6:30 PM PDT, you can go to
> > > https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-5.10
> > > and see that the patch is still there.)
> > > 
> > > 
> > > By the way, if dropping this patch turns out not to be the correct course
> > > of action, I think backporting one additional patch will fix it.
> > > 
> > > I applied the following patch directly from mainline to test it; it
> > > applied with an offset but no fuzz. I did some brief testing and it
> > > seems to fix the problem (no more freezing on shutdown).
> > > 
> > > mainline commit 46e85af0cc53f35584e00bb5db7db6893d0e16e5
> > > driver core: platform: don't oops in platform_shutdown() on unbound devices
> > 
> > Ugh, my fault, I was confusing this with a different patch.  I'll go
> > drop all of these now, thanks for confirming.
> 
> I'm unsure why you wrote to "drop all of these" now that the reason for
> the breakage is explained and the solution is to pick another patch from
> mainline. If you don't really need
> driver-core-platform-use-bus_type-functions.patch that's fine, but
> otherwise just add 46e85af0cc53f35584e00bb5db7db6893d0e16e5?
> 
> I somehow expected that the patch selection procedure for stable notices
> if you pick 9c30921fe799 ("driver core: platform: use bus_type
> functions") that there are later changes declaring to fix this commit?!

Yes, we had not run the "find the fixes for the fixes" script, but in
the end, it was just not going to be worth it so I dropped these
changes.

thanks,

greg k-h