Hello,
recently I've debugged a deadlock issue in ext4 fast commit feature in our
openSUSE Leap kernel. Checking with upstream I've found out the issue was
accidentally fixed by commit 2729cfdcfa1c ("ext4: use
ext4_journal_start/stop for fast commit transactions"). Can you please pick
up this commit into the stable tree? The problem has been there since fast
commit began to exist (in 5.10-rc1) so 5.10 and 5.15 stable trees need the
fix. Thanks!
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
On Thu, Apr 27, 2023 at 06:24:59PM +0200, Jan Kara wrote:
> Hello,
>
> recently I've debugged a deadlock issue in ext4 fast commit feature in our
> openSUSE Leap kernel. Checking with upstream I've found out the issue was
> accidentally fixed by commit 2729cfdcfa1c ("ext4: use
> ext4_journal_start/stop for fast commit transactions"). Can you please pick
> up this commit into the stable tree? The problem has been there since fast
> commit began to exist (in 5.10-rc1) so 5.10 and 5.15 stable trees need the
> fix. Thanks!
This commit does not apply to those branches, so how was it tested?
Can you send us backported, and tested, versions of this commit so that
we can apply them?
thanks,
greg k-h
On Fri 28-04-23 09:32:34, Greg KH wrote:
> On Thu, Apr 27, 2023 at 06:24:59PM +0200, Jan Kara wrote:
> > Hello,
> >
> > recently I've debugged a deadlock issue in ext4 fast commit feature in our
> > openSUSE Leap kernel. Checking with upstream I've found out the issue was
> > accidentally fixed by commit 2729cfdcfa1c ("ext4: use
> > ext4_journal_start/stop for fast commit transactions"). Can you please pick
> > up this commit into the stable tree? The problem has been there since fast
> > commit began to exist (in 5.10-rc1) so 5.10 and 5.15 stable trees need the
> > fix. Thanks!
>
> This commit does not apply to those branches, so how was it tested?
Hum, I've picked up the patch to our 5.14-based distro kernel and it
applied (and passed testing) without issues so I was hoping 5.15 and 5.10
would work as well. Apparently I was wrong. Sorry for the trouble.
> Can you send us backported, and tested, versions of this commit so that
> we can apply them?
Yeah, I'll look into applying the patch directly to stable branches.
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
On Fri 28-04-23 14:02:46, Jan Kara wrote:
> On Fri 28-04-23 09:32:34, Greg KH wrote:
> > On Thu, Apr 27, 2023 at 06:24:59PM +0200, Jan Kara wrote:
> > > Hello,
> > >
> > > recently I've debugged a deadlock issue in ext4 fast commit feature in our
> > > openSUSE Leap kernel. Checking with upstream I've found out the issue was
> > > accidentally fixed by commit 2729cfdcfa1c ("ext4: use
> > > ext4_journal_start/stop for fast commit transactions"). Can you please pick
> > > up this commit into the stable tree? The problem has been there since fast
> > > commit began to exist (in 5.10-rc1) so 5.10 and 5.15 stable trees need the
> > > fix. Thanks!
> >
> > This commit does not apply to those branches, so how was it tested?
>
> Hum, I've picked up the patch to our 5.14-based distro kernel and it
> applied (and passed testing) without issues so I was hoping 5.15 and 5.10
> would work as well. Apparently I was wrong. Sorry for the trouble.
>
> > Can you send us backported, and tested, versions of this commit so that
> > we can apply them?
>
> Yeah, I'll look into applying the patch directly to stable branches.
Tested version of the patch for 5.15-stable is attached.
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
On Fri 28-04-23 14:02:46, Jan Kara wrote:
> > Can you send us backported, and tested, versions of this commit so that
> > we can apply them?
>
> Yeah, I'll look into applying the patch directly to stable branches.
Attached is backport of the fix to 5.10-stable kernel (passed xfstests run
for ext4 with fast_commit enabled).
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
On Fri, Apr 28, 2023 at 04:51:58PM +0200, Jan Kara wrote:
> On Fri 28-04-23 14:02:46, Jan Kara wrote:
> > > Can you send us backported, and tested, versions of this commit so that
> > > we can apply them?
> >
> > Yeah, I'll look into applying the patch directly to stable branches.
>
> Attached is backport of the fix to 5.10-stable kernel (passed xfstests run
> for ext4 with fast_commit enabled).
Thanks for both of these, now queued up.
greg k-h