[GIT PULL] bcachefs fixes for 6.14.y

[GIT PULL] bcachefs fixes for 6.14.y

[Kent Overstreet]

The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:

  bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)

are available in the Git repository at:

  git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02

for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:

  bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)

----------------------------------------------------------------
bcachefs fixes for 6.15

remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
that have been hitting arch users

----------------------------------------------------------------
Alan Huang (1):
      bcachefs: Remove incorrect __counted_by annotation

 fs/bcachefs/xattr_format.h | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

Re: [GIT PULL] bcachefs fixes for 6.14.y

[Greg Kroah-Hartman]

On Fri, May 02, 2025 at 09:12:22PM -0400, Kent Overstreet wrote:
> 
> The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:
> 
>   bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)
> 
> are available in the Git repository at:
> 
>   git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02
> 
> for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:
> 
>   bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)
> 
> ----------------------------------------------------------------
> bcachefs fixes for 6.15
> 
> remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
> that have been hitting arch users
> 
> ----------------------------------------------------------------
> Alan Huang (1):
>       bcachefs: Remove incorrect __counted_by annotation
> 
>  fs/bcachefs/xattr_format.h | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)

You list 1 patch here, but if I pull this, I see 2 patches against the
latest linux-6.14.y branch.  When rebased, the "additional" one goes
away, as you already sent that to us in the past, so I'll just take the
one that's left here, but please, make this more obvious what is
happening.

Also, I see a lot of syzbot fixes going into bcachefs recently,
hopefully those are all for issues that only affected the tree after
6.14 was released.

thanks,

greg k-h

Re: [GIT PULL] bcachefs fixes for 6.14.y

[Greg Kroah-Hartman]

On Mon, May 05, 2025 at 10:32:49AM +0200, Greg Kroah-Hartman wrote:
> On Fri, May 02, 2025 at 09:12:22PM -0400, Kent Overstreet wrote:
> > 
> > The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:
> > 
> >   bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)
> > 
> > are available in the Git repository at:
> > 
> >   git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02
> > 
> > for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:
> > 
> >   bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)
> > 
> > ----------------------------------------------------------------
> > bcachefs fixes for 6.15
> > 
> > remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
> > that have been hitting arch users
> > 
> > ----------------------------------------------------------------
> > Alan Huang (1):
> >       bcachefs: Remove incorrect __counted_by annotation
> > 
> >  fs/bcachefs/xattr_format.h | 8 +++++++-
> >  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> You list 1 patch here, but if I pull this, I see 2 patches against the
> latest linux-6.14.y branch.  When rebased, the "additional" one goes
> away, as you already sent that to us in the past, so I'll just take the
> one that's left here, but please, make this more obvious what is
> happening.

Also, as this single commit is a revert of something in 6.12.y, should
it also be applied there to remove any false-positives happening with
those users?

thanks,

greg k-h

Re: [GIT PULL] bcachefs fixes for 6.14.y

[Kent Overstreet]

On Mon, May 05, 2025 at 10:34:35AM +0200, Greg Kroah-Hartman wrote:
> On Mon, May 05, 2025 at 10:32:49AM +0200, Greg Kroah-Hartman wrote:
> > On Fri, May 02, 2025 at 09:12:22PM -0400, Kent Overstreet wrote:
> > > 
> > > The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:
> > > 
> > >   bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)
> > > 
> > > are available in the Git repository at:
> > > 
> > >   git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02
> > > 
> > > for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:
> > > 
> > >   bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)
> > > 
> > > ----------------------------------------------------------------
> > > bcachefs fixes for 6.15
> > > 
> > > remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
> > > that have been hitting arch users
> > > 
> > > ----------------------------------------------------------------
> > > Alan Huang (1):
> > >       bcachefs: Remove incorrect __counted_by annotation
> > > 
> > >  fs/bcachefs/xattr_format.h | 8 +++++++-
> > >  1 file changed, 7 insertions(+), 1 deletion(-)
> > 
> > You list 1 patch here, but if I pull this, I see 2 patches against the
> > latest linux-6.14.y branch.  When rebased, the "additional" one goes
> > away, as you already sent that to us in the past, so I'll just take the
> > one that's left here, but please, make this more obvious what is
> > happening.
> 
> Also, as this single commit is a revert of something in 6.12.y, should
> it also be applied there to remove any false-positives happening with
> those users?

yup, pull request inc

(missed this because all the reports have been on 6.14, but no doubt
that's because gcc-15 has only recently been rolling out)

Re: [GIT PULL] bcachefs fixes for 6.14.y

[Kent Overstreet]

On Mon, May 05, 2025 at 10:32:49AM +0200, Greg Kroah-Hartman wrote:
> On Fri, May 02, 2025 at 09:12:22PM -0400, Kent Overstreet wrote:
> > 
> > The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:
> > 
> >   bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)
> > 
> > are available in the Git repository at:
> > 
> >   git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02
> > 
> > for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:
> > 
> >   bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)
> > 
> > ----------------------------------------------------------------
> > bcachefs fixes for 6.15
> > 
> > remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
> > that have been hitting arch users
> > 
> > ----------------------------------------------------------------
> > Alan Huang (1):
> >       bcachefs: Remove incorrect __counted_by annotation
> > 
> >  fs/bcachefs/xattr_format.h | 8 +++++++-
> >  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> You list 1 patch here, but if I pull this, I see 2 patches against the
> latest linux-6.14.y branch.  When rebased, the "additional" one goes
> away, as you already sent that to us in the past, so I'll just take the
> one that's left here, but please, make this more obvious what is
> happening.

That's because you're rebasing my patches.

> Also, I see a lot of syzbot fixes going into bcachefs recently,
> hopefully those are all for issues that only affected the tree after
> 6.14 was released.

Until the experimental label comes off I'm only doing critical
backports - it really doesn't make any sense to do anything else right
now.

The syzbot stuff has had zero overlap with user reported bugs, and since
it's fuzzing the on disk image (and we don't support unprivilidged
mounts - yet, at least) - they haven't been a security concern. There's
been one security bug since 6.7, and you have that fix.

Re: [GIT PULL] bcachefs fixes for 6.14.y

[Greg Kroah-Hartman]

On Mon, May 05, 2025 at 09:54:24AM -0400, Kent Overstreet wrote:
> On Mon, May 05, 2025 at 10:32:49AM +0200, Greg Kroah-Hartman wrote:
> > On Fri, May 02, 2025 at 09:12:22PM -0400, Kent Overstreet wrote:
> > > 
> > > The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:
> > > 
> > >   bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)
> > > 
> > > are available in the Git repository at:
> > > 
> > >   git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02
> > > 
> > > for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:
> > > 
> > >   bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)
> > > 
> > > ----------------------------------------------------------------
> > > bcachefs fixes for 6.15
> > > 
> > > remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
> > > that have been hitting arch users
> > > 
> > > ----------------------------------------------------------------
> > > Alan Huang (1):
> > >       bcachefs: Remove incorrect __counted_by annotation
> > > 
> > >  fs/bcachefs/xattr_format.h | 8 +++++++-
> > >  1 file changed, 7 insertions(+), 1 deletion(-)
> > 
> > You list 1 patch here, but if I pull this, I see 2 patches against the
> > latest linux-6.14.y branch.  When rebased, the "additional" one goes
> > away, as you already sent that to us in the past, so I'll just take the
> > one that's left here, but please, make this more obvious what is
> > happening.
> 
> That's because you're rebasing my patches.

Not really a "rebase", but rather a "cherry-pick", but we've been
through this before, so no need to go over it again :)

> > Also, I see a lot of syzbot fixes going into bcachefs recently,
> > hopefully those are all for issues that only affected the tree after
> > 6.14 was released.
> 
> Until the experimental label comes off I'm only doing critical
> backports - it really doesn't make any sense to do anything else right
> now.

Ok.

> The syzbot stuff has had zero overlap with user reported bugs, and since
> it's fuzzing the on disk image (and we don't support unprivilidged
> mounts - yet, at least) - they haven't been a security concern. There's
> been one security bug since 6.7, and you have that fix.

Great, thanks!

greg k-h

Re: [GIT PULL] bcachefs fixes for 6.14.y

[Kent Overstreet]

On Mon, May 05, 2025 at 04:06:55PM +0200, Greg Kroah-Hartman wrote:
> On Mon, May 05, 2025 at 09:54:24AM -0400, Kent Overstreet wrote:
> > On Mon, May 05, 2025 at 10:32:49AM +0200, Greg Kroah-Hartman wrote:
> > > On Fri, May 02, 2025 at 09:12:22PM -0400, Kent Overstreet wrote:
> > > > 
> > > > The following changes since commit 02a22be3c0003af08df510cba3d79d00c6495b74:
> > > > 
> > > >   bcachefs: bch2_ioctl_subvolume_destroy() fixes (2025-04-03 16:13:53 -0400)
> > > > 
> > > > are available in the Git repository at:
> > > > 
> > > >   git://evilpiepirate.org/bcachefs.git tags/bcachefs-for-6.14-2025-05-02
> > > > 
> > > > for you to fetch changes up to 52b17bca7b20663e5df6dbfc24cc2030259b64b6:
> > > > 
> > > >   bcachefs: Remove incorrect __counted_by annotation (2025-05-02 21:09:51 -0400)
> > > > 
> > > > ----------------------------------------------------------------
> > > > bcachefs fixes for 6.15
> > > > 
> > > > remove incorrect counted_by annotation, fixing FORTIFY_SOURCE crashes
> > > > that have been hitting arch users
> > > > 
> > > > ----------------------------------------------------------------
> > > > Alan Huang (1):
> > > >       bcachefs: Remove incorrect __counted_by annotation
> > > > 
> > > >  fs/bcachefs/xattr_format.h | 8 +++++++-
> > > >  1 file changed, 7 insertions(+), 1 deletion(-)
> > > 
> > > You list 1 patch here, but if I pull this, I see 2 patches against the
> > > latest linux-6.14.y branch.  When rebased, the "additional" one goes
> > > away, as you already sent that to us in the past, so I'll just take the
> > > one that's left here, but please, make this more obvious what is
> > > happening.
> > 
> > That's because you're rebasing my patches.
> 
> Not really a "rebase", but rather a "cherry-pick", but we've been
> through this before, so no need to go over it again :)
> 
> > > Also, I see a lot of syzbot fixes going into bcachefs recently,
> > > hopefully those are all for issues that only affected the tree after
> > > 6.14 was released.
> > 
> > Until the experimental label comes off I'm only doing critical
> > backports - it really doesn't make any sense to do anything else right
> > now.
> 
> Ok.
> 
> > The syzbot stuff has had zero overlap with user reported bugs, and since
> > it's fuzzing the on disk image (and we don't support unprivilidged
> > mounts - yet, at least) - they haven't been a security concern. There's
> > been one security bug since 6.7, and you have that fix.
> 
> Great, thanks!

BTW - in the interim, if we do want to backport a wider set of fixes the
sane thing to do will be to forklift all of fs/bcachefs to the stable
kernels. QA here has been good, so regressions in Linus's tree have been
a non issue - in 6.15, all (2?) regressions were found and fixed before
rc1 came out.

This hasn't been an option yet because the required on disk format
upgrades have been too disruptive for stable kernels (the 6.14 upgrade
took ~hours on the big 50-100TB filesystems), but as of 6.15 that should
be over, so I've been giving some thought to the idea, especially since
usage is steadily ticking up...