[GIT PULL] Ceph updates for 6.4-rc1

[GIT PULL] Ceph updates for 6.4-rc1

[Ilya Dryomov]

Hi Linus,

The following changes since commit 457391b0380335d5e9a5babdec90ac53928b23b4:

  Linux 6.3 (2023-04-23 12:02:52 -0700)

are available in the Git repository at:

  https://github.com/ceph/ceph-client.git tags/ceph-for-6.4-rc1

for you to fetch changes up to db2993a423e3fd0e4878f4d3ac66fe717f5f072e:

  ceph: reorder fields in 'struct ceph_snapid_map' (2023-04-30 12:37:28 +0200)

----------------------------------------------------------------
A few filesystem improvements, with a rather nasty use-after-free fix
from Xiubo intended for stable.

----------------------------------------------------------------
Christophe JAILLET (1):
      ceph: reorder fields in 'struct ceph_snapid_map'

Xiubo Li (4):
      ceph: do not print the whole xattr value if it's too long
      ceph: implement writeback livelock avoidance using page tagging
      ceph: fix potential use-after-free bug when trimming caps
      ceph: pass ino# instead of old_dentry if it's disconnected

 fs/ceph/addr.c       | 11 +++++++-
 fs/ceph/caps.c       |  2 +-
 fs/ceph/debugfs.c    | 18 ++++++++----
 fs/ceph/dir.c        | 13 +++++++--
 fs/ceph/mds_client.c | 78 ++++++++++++++++++++++++++++++++++------------------
 fs/ceph/mds_client.h |  5 ++--
 fs/ceph/super.h      |  2 ++
 fs/ceph/xattr.c      | 20 +++++++++-----
 8 files changed, 102 insertions(+), 47 deletions(-)

Re: [GIT PULL] Ceph updates for 6.4-rc1

[pr-tracker-bot]

The pull request you sent on Thu,  4 May 2023 20:28:10 +0200:

> https://github.com/ceph/ceph-client.git tags/ceph-for-6.4-rc1

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/3c4aa44343777844e425c28f1427127f3e55826f

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html

Re: [GIT PULL] Ceph updates for 6.4-rc1

[Luís Henriques]

[re-arranged CC list]

Ilya Dryomov <idryomov@gmail.com> writes:

> Hi Linus,
>
> The following changes since commit 457391b0380335d5e9a5babdec90ac53928b23b4:
>
>   Linux 6.3 (2023-04-23 12:02:52 -0700)
>
> are available in the Git repository at:
>
>   https://github.com/ceph/ceph-client.git tags/ceph-for-6.4-rc1
>
> for you to fetch changes up to db2993a423e3fd0e4878f4d3ac66fe717f5f072e:
>
>   ceph: reorder fields in 'struct ceph_snapid_map' (2023-04-30 12:37:28 +0200)
>
> ----------------------------------------------------------------
> A few filesystem improvements, with a rather nasty use-after-free fix
> from Xiubo intended for stable.

Thank you, Ilya.  It's unfortunate that fscrypt support misses yet another
merge window, but I guess there are still a few loose ends.

Is there a public list of issues (kernel or ceph proper) still to be
sorted out before this feature gets merged?  Or is this just a lack of
confidence on the implementation stability?

Cheers
-- 
Luís

Re: [GIT PULL] Ceph updates for 6.4-rc1

[Ilya Dryomov]

On Fri, May 5, 2023 at 1:42 PM Luís Henriques <lhenriques@suse.de> wrote:
>
>
> [re-arranged CC list]
>
> Ilya Dryomov <idryomov@gmail.com> writes:
>
> > Hi Linus,
> >
> > The following changes since commit 457391b0380335d5e9a5babdec90ac53928b23b4:
> >
> >   Linux 6.3 (2023-04-23 12:02:52 -0700)
> >
> > are available in the Git repository at:
> >
> >   https://github.com/ceph/ceph-client.git tags/ceph-for-6.4-rc1
> >
> > for you to fetch changes up to db2993a423e3fd0e4878f4d3ac66fe717f5f072e:
> >
> >   ceph: reorder fields in 'struct ceph_snapid_map' (2023-04-30 12:37:28 +0200)
> >
> > ----------------------------------------------------------------
> > A few filesystem improvements, with a rather nasty use-after-free fix
> > from Xiubo intended for stable.
>
> Thank you, Ilya.  It's unfortunate that fscrypt support misses yet another
> merge window, but I guess there are still a few loose ends.
>
> Is there a public list of issues (kernel or ceph proper) still to be
> sorted out before this feature gets merged?  Or is this just a lack of
> confidence on the implementation stability?

Hi Luís,

When fscrypt work got supposedly finalized it was already pretty late
in the cycle and it just didn't help that upon pulling it I encountered
a subtly broken patch which was NACKed before ("libceph: defer removing
the req from osdc just after req->r_callback") and also that "optionally
bypass content encryption" leftover.  It got addressed but too late for
such a large change to be staged for 6.4 merge window.

I would encourage everyone to make another pass over the entire series
to make sure that there is nothing eyebrows-raising left there and that
it really feels solid.

Thanks,

                Ilya

Re: [GIT PULL] Ceph updates for 6.4-rc1

[Luís Henriques]

Ilya Dryomov <idryomov@gmail.com> writes:

> On Fri, May 5, 2023 at 1:42 PM Luís Henriques <lhenriques@suse.de> wrote:
>>
>>
>> [re-arranged CC list]
>>
>> Ilya Dryomov <idryomov@gmail.com> writes:
>>
>> > Hi Linus,
>> >
>> > The following changes since commit 457391b0380335d5e9a5babdec90ac53928b23b4:
>> >
>> >   Linux 6.3 (2023-04-23 12:02:52 -0700)
>> >
>> > are available in the Git repository at:
>> >
>> >   https://github.com/ceph/ceph-client.git tags/ceph-for-6.4-rc1
>> >
>> > for you to fetch changes up to db2993a423e3fd0e4878f4d3ac66fe717f5f072e:
>> >
>> >   ceph: reorder fields in 'struct ceph_snapid_map' (2023-04-30 12:37:28 +0200)
>> >
>> > ----------------------------------------------------------------
>> > A few filesystem improvements, with a rather nasty use-after-free fix
>> > from Xiubo intended for stable.
>>
>> Thank you, Ilya.  It's unfortunate that fscrypt support misses yet another
>> merge window, but I guess there are still a few loose ends.
>>
>> Is there a public list of issues (kernel or ceph proper) still to be
>> sorted out before this feature gets merged?  Or is this just a lack of
>> confidence on the implementation stability?
>
> Hi Luís,
>
> When fscrypt work got supposedly finalized it was already pretty late
> in the cycle and it just didn't help that upon pulling it I encountered
> a subtly broken patch which was NACKed before ("libceph: defer removing
> the req from osdc just after req->r_callback") and also that "optionally
> bypass content encryption" leftover.  It got addressed but too late for
> such a large change to be staged for 6.4 merge window.
> 
> I would encourage everyone to make another pass over the entire series
> to make sure that there is nothing eyebrows-raising left there and that
> it really feels solid.

Thanks for the clarification, Ilya.  I'll definitely restart testing and
reviewing the whole series, that's something I've been doing every time
the patchset is rebased.  But yeah my eyes are already so used to look
into that code that any issues I may be able to find are probably related
with the rebase itself :-)

Cheers,
-- 
Luís

Re: [GIT PULL] Ceph updates for 6.4-rc1

[Xiubo Li]

On 5/6/23 02:59, Ilya Dryomov wrote:
> On Fri, May 5, 2023 at 1:42 PM Luís Henriques <lhenriques@suse.de> wrote:
>>
>> [re-arranged CC list]
>>
>> Ilya Dryomov <idryomov@gmail.com> writes:
>>
>>> Hi Linus,
>>>
>>> The following changes since commit 457391b0380335d5e9a5babdec90ac53928b23b4:
>>>
>>>    Linux 6.3 (2023-04-23 12:02:52 -0700)
>>>
>>> are available in the Git repository at:
>>>
>>>    https://github.com/ceph/ceph-client.git tags/ceph-for-6.4-rc1
>>>
>>> for you to fetch changes up to db2993a423e3fd0e4878f4d3ac66fe717f5f072e:
>>>
>>>    ceph: reorder fields in 'struct ceph_snapid_map' (2023-04-30 12:37:28 +0200)
>>>
>>> ----------------------------------------------------------------
>>> A few filesystem improvements, with a rather nasty use-after-free fix
>>> from Xiubo intended for stable.
>> Thank you, Ilya.  It's unfortunate that fscrypt support misses yet another
>> merge window, but I guess there are still a few loose ends.
>>
>> Is there a public list of issues (kernel or ceph proper) still to be
>> sorted out before this feature gets merged?  Or is this just a lack of
>> confidence on the implementation stability?
> Hi Luís,
>
> When fscrypt work got supposedly finalized it was already pretty late
> in the cycle and it just didn't help that upon pulling it I encountered
> a subtly broken patch which was NACKed before ("libceph: defer removing
> the req from osdc just after req->r_callback")

We still need this patch or a new patch to fix this.

Another patch will only fix the mds request case, but not the osd 
request case.

thanks

>   and also that "optionally
> bypass content encryption" leftover.  It got addressed but too late for
> such a large change to be staged for 6.4 merge window.
>
> I would encourage everyone to make another pass over the entire series
> to make sure that there is nothing eyebrows-raising left there and that
> it really feels solid.
>
> Thanks,
>
>                  Ilya
>