AMD drm patch workflow is broken for stable trees

[Alex Deucher]

On Wed, Aug 14, 2024 at 4:55 PM Felix Kuehling <felix.kuehling at amd.com> wrote:
>
> On 2024-08-12 11:00, Greg KH wrote:
> > Hi all,
> >
> > As some of you have noticed, there's a TON of failure messages being
> > sent out for AMD gpu driver commits that are tagged for stable
> > backports.  In short, you all are doing something really wrong with how
> > you are tagging these.
> Hi Greg,
>
> I got notifications about one KFD patch failing to apply on six branches
> (6.10, 6.6, 6.1, 5.15, 5.10 and 5.4). The funny thing is, that you
> already applied this patch on two branches back in May. The emails had a
> suspicious looking date in the header (Sep 17, 2001). I wonder if there
> was some date glitch that caused a whole bunch of patches to be re-sent
> to stable somehow:

I think the crux of the problem is that sometimes patches go into
-next with stable tags and they end getting taken into -fixes as well
so after the merge window they end up getting picked up for stable
again.  Going forward, if they land in -next, I'll cherry-pick -x the
changes into -fixes so there is better traceability.

Alex

>
>     ------------------ original commit in Linus's tree
>     ------------------ From 24e82654e98e96cece5d8b919c522054456eeec6 Mon
>     Sep 17 00:00:00 2001 From: Alex Deucher
>     <alexander.deucher at amd.com>Date: Sun, 14 Apr 2024 13:06:39 -0400
>     Subject: [PATCH] drm/amdkfd: don't allow mapping the MMIO HDP page
>     with large pages ...
>
> On 6.1 and 6.6, the patch was already applied by you in May:
>
>     $ git log --pretty=fuller stable/linux-6.6.y --grep "drm/amdkfd: don't allow mapping the MMIO HDP page with large pages"
>     commit 4b4cff994a27ebf7bd3fb9a798a1cdfa8d01b724
>     Author:     Alex Deucher <alexander.deucher at amd.com>
>     AuthorDate: Sun Apr 14 13:06:39 2024 -0400
>     Commit:     Greg Kroah-Hartman <gregkh at linuxfoundation.org>
>     CommitDate: Fri May 17 12:02:34 2024 +0200
>
>          drm/amdkfd: don't allow mapping the MMIO HDP page with large pages
>     ...
>
> On 6.10 it was already upstream.
>
> On 5.4-5.15 it doesn't apply because of conflicts. I can resolve those
> and send the fixed patches out for you.
>
> Regards,
>    Felix
>
>
> >
> > Please fix it up to NOT have duplicates in multiple branches that end up
> > in Linus's tree at different times.  Or if you MUST do that, then give
> > us a chance to figure out that it IS a duplicate.  As-is, it's not
> > working at all, and I think I need to just drop all patches for this
> > driver that are tagged for stable going forward and rely on you all to
> > provide a proper set of backported fixes when you say they are needed.
> >
> > Again, what you are doing today is NOT ok and is broken.  Please fix.
> >
> > greg k-h