Request for backport of 358de8b4f201 to LTS kernels

Request for backport of 358de8b4f201 to LTS kernels

[Chuck Lever]

Hi Greg & Sasha !

I ran into some trouble in my nightly CI systems that test v6.6.y and
v6.1.y. Using "make binrpm-pkg" followed by "rpm -iv ..." results in the
test systems being unbootable because the vmlinuz file is never copied
to /boot. The test systems are imaged with Fedora 39.

I found a related Fedora bug:

  https://bugzilla.redhat.com/show_bug.cgi?id=2239008

It appears there is a missing fix in LTS kernels. I bisected the kernel
fix to:

  358de8b4f201 ("kbuild: rpm-pkg: simplify installkernel %post")

which includes a "Cc: stable" tag but does not appear in
origin/linux-6.6.y, origin/linux-6.1.y, or origin/5.15.y (I did not look
further back than that).

Would it be appropriate to apply 358de8b4f201 to LTS kernels?

-- 
Chuck Lever

Re: Request for backport of 358de8b4f201 to LTS kernels

[Greg KH]

On Mon, Jun 09, 2025 at 04:30:19PM -0400, Chuck Lever wrote:
> Hi Greg & Sasha !
> 
> I ran into some trouble in my nightly CI systems that test v6.6.y and
> v6.1.y. Using "make binrpm-pkg" followed by "rpm -iv ..." results in the
> test systems being unbootable because the vmlinuz file is never copied
> to /boot. The test systems are imaged with Fedora 39.
> 
> I found a related Fedora bug:
> 
>   https://bugzilla.redhat.com/show_bug.cgi?id=2239008
> 
> It appears there is a missing fix in LTS kernels. I bisected the kernel
> fix to:
> 
>   358de8b4f201 ("kbuild: rpm-pkg: simplify installkernel %post")
> 
> which includes a "Cc: stable" tag but does not appear in
> origin/linux-6.6.y, origin/linux-6.1.y, or origin/5.15.y (I did not look
> further back than that).
> 
> Would it be appropriate to apply 358de8b4f201 to LTS kernels?

Perhaps, if someone actually backported it to apply there, did you try
it?  :)

At the time, this was reported:
	https://lore.kernel.org/all/2024021932-lavish-expel-58e5@gregkh/
	https://lore.kernel.org/r/2024021934-spree-discard-c389@gregkh
	https://lore.kernel.org/r/2024021930-prozac-outfield-8653@gregkh
but no one did anything about it.

thanks,

greg k-h

Re: Request for backport of 358de8b4f201 to LTS kernels

[Chuck Lever]

On 6/17/25 9:51 AM, Greg KH wrote:
> On Mon, Jun 09, 2025 at 04:30:19PM -0400, Chuck Lever wrote:
>> Hi Greg & Sasha !
>>
>> I ran into some trouble in my nightly CI systems that test v6.6.y and
>> v6.1.y. Using "make binrpm-pkg" followed by "rpm -iv ..." results in the
>> test systems being unbootable because the vmlinuz file is never copied
>> to /boot. The test systems are imaged with Fedora 39.
>>
>> I found a related Fedora bug:
>>
>>   https://bugzilla.redhat.com/show_bug.cgi?id=2239008
>>
>> It appears there is a missing fix in LTS kernels. I bisected the kernel
>> fix to:
>>
>>   358de8b4f201 ("kbuild: rpm-pkg: simplify installkernel %post")
>>
>> which includes a "Cc: stable" tag but does not appear in
>> origin/linux-6.6.y, origin/linux-6.1.y, or origin/5.15.y (I did not look
>> further back than that).
>>
>> Would it be appropriate to apply 358de8b4f201 to LTS kernels?
> 
> Perhaps, if someone actually backported it to apply there, did you try
> it?  :)

I didn't try it (laziness). I'm happy to give it a shot.


> 
> At the time, this was reported:
> 	https://lore.kernel.org/all/2024021932-lavish-expel-58e5@gregkh/
> 	https://lore.kernel.org/r/2024021934-spree-discard-c389@gregkh
> 	https://lore.kernel.org/r/2024021930-prozac-outfield-8653@gregkh
> but no one did anything about it.
> 
> thanks,
> 
> greg k-h


-- 
Chuck Lever

Re: Request for backport of 358de8b4f201 to LTS kernels

[Chuck Lever]

On 6/17/25 9:51 AM, Greg KH wrote:
> On Mon, Jun 09, 2025 at 04:30:19PM -0400, Chuck Lever wrote:
>> Hi Greg & Sasha !
>>
>> I ran into some trouble in my nightly CI systems that test v6.6.y and
>> v6.1.y. Using "make binrpm-pkg" followed by "rpm -iv ..." results in the
>> test systems being unbootable because the vmlinuz file is never copied
>> to /boot. The test systems are imaged with Fedora 39.
>>
>> I found a related Fedora bug:
>>
>>   https://bugzilla.redhat.com/show_bug.cgi?id=2239008
>>
>> It appears there is a missing fix in LTS kernels. I bisected the kernel
>> fix to:
>>
>>   358de8b4f201 ("kbuild: rpm-pkg: simplify installkernel %post")
>>
>> which includes a "Cc: stable" tag but does not appear in
>> origin/linux-6.6.y, origin/linux-6.1.y, or origin/5.15.y (I did not look
>> further back than that).
>>
>> Would it be appropriate to apply 358de8b4f201 to LTS kernels?
> 
> Perhaps, if someone actually backported it to apply there, did you try
> it?  :)
> 
> At the time, this was reported:
> 	https://lore.kernel.org/all/2024021932-lavish-expel-58e5@gregkh/
> 	https://lore.kernel.org/r/2024021934-spree-discard-c389@gregkh
> 	https://lore.kernel.org/r/2024021930-prozac-outfield-8653@gregkh
> but no one did anything about it.

I've posted a clean and tested backport to v6.6 LTS.

However, the number of patches that need to be applied to v6.1 is much
larger than the provided possible dependency list. Hence my original
question "Would it be appropriate ..." I think therefore that the answer
is "It would not be appropriate to apply 358de8b4f201 to LTS kernels
prior to origin/linux-6.6.y".

And perhaps the solution for me is to ensure my v6.1.y test runners use
Fedora 37 instead of Fedora 39.


-- 
Chuck Lever