Re: important orangefs merge conflict request...

Re: important orangefs merge conflict request...

[Mike Marshall]

Hi Greg...

>> Sure, what is the upstream git commit id of the patch?

b36ddb9210 is the commit where Linus pulled my fix into 6.15.

Some of Matthew Wilcox's folio work also landed in orangefs in 6.15,
causing b36ddb9210 to have a merge conflict with 6.14.

I tested during 6.14-rc7 and Commit 665575cf was pulled into
6.14-rc7 after I tested, causing orangefs to be broken in 6.14.

So... there is not an upstream commit id for my patch that works
with 6.14.

Pulling 665575cf so late into rc7, even though it was known to
have caused deadlock problems with ext4 and (?) seems like
an "unlinuxy" thing to have done. I was just hoping there was
some way to get my un-committed 6.14 flavored patch
backported to 6.14... I figured if it was possible, you'd
know how...

Thanks...

-Mike

On Tue, Jun 3, 2025 at 12:38 AM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Mon, Jun 02, 2025 at 12:19:39PM -0400, Mike Marshall wrote:
> > Hi Greg..
> >
> > Commit 665575cf "filemap: move prefaulting out of hot write path", which was
> > pulled in 6.14-rc7, broke orangefs. I got a fix applied to to 6.15.
> > 6.15 saw more of Matthew Wilcox's folio work folded into orangefs,
> > and my 6.15 patch has conflicts with 6.14. Fedora (and others?) is
> > at 6.14 now.
> >
> > I have a tested 6.14 version of my patch, would it be possible to
> > get that backported to 6.14?
>
> Sure, what is the upstream git commit id of the patch?
>
> >
> > -Mike
>
> > From 572e2027f6111728877f20085e79d07c5b2238de Mon Sep 17 00:00:00 2001
> > From: Mike Marshall <hubcap@omnibond.com>
> > Date: Tue, 13 May 2025 17:00:37 -0400
> > Subject: [PATCH] orangefs: adjust counting code to recover from 665575cf
> >
> > A late commit to 6.14-rc7! broke orangefs. 665575cf seems like a
> > good change, but maybe should have been introduced during the merge
> > window. This patch adjusts the counting code associated with
> > writing out pages so that orangefs works in a 665575cf world.
> > ---
> >  fs/orangefs/inode.c | 9 +++++----
> >  1 file changed, 5 insertions(+), 4 deletions(-)
>
> <snip>
>
> This isn't in a format I can take it :(
>
> Also, it's not on a public mailing list for some reason...
>
> thanks,
>
> greg k-h

Re: important orangefs merge conflict request...

[Greg Kroah-Hartman]

On Tue, Jun 03, 2025 at 07:02:09AM -0400, Mike Marshall wrote:
> Hi Greg...
> 
> >> Sure, what is the upstream git commit id of the patch?
> 
> b36ddb9210 is the commit where Linus pulled my fix into 6.15.
> 
> Some of Matthew Wilcox's folio work also landed in orangefs in 6.15,
> causing b36ddb9210 to have a merge conflict with 6.14.
> 
> I tested during 6.14-rc7 and Commit 665575cf was pulled into
> 6.14-rc7 after I tested, causing orangefs to be broken in 6.14.
> 
> So... there is not an upstream commit id for my patch that works
> with 6.14.
> 
> Pulling 665575cf so late into rc7, even though it was known to
> have caused deadlock problems with ext4 and (?) seems like
> an "unlinuxy" thing to have done. I was just hoping there was
> some way to get my un-committed 6.14 flavored patch
> backported to 6.14... I figured if it was possible, you'd
> know how...

Just submit it to us at stable@vger.kernel.org as a backport of the
commit like any other normal stable backport?

thanks,

greg k-h