Re: [PATCH 5.15 00/57] 5.15.155-rc1 review

[Chuck Lever]

On Sat, Apr 13, 2024 at 01:41:52AM +0530, Harshit Mogalapalli wrote:
> Hi Greg, Chuck,
> 
> On 12/04/24 21:27, Chuck Lever III wrote:
> > 
> > 
> > > I have noticed a regression in lts test case with nfsv4 and this was overlooked in the previous cycle(5.15.154). So the regression is from 153-->154 update. And I think that is due to nfs backports we had in 5.15.154.
> > > 
> > > # ./runltp -d /tmpdir -s fcntl17
> > > 
> > > <<<test_start>>>
> > > tag=fcntl17 stime=1712915065
> ...
> > > fcntl17     1  TFAIL  :  fcntl17.c:429: Alarm expired, deadlock not detected
> > > fcntl17     0  TWARN  :  fcntl17.c:430: You may need to kill child processes by hand
> > > fcntl17     2  TPASS  :  Block 1 PASSED
> > > fcntl17     0  TINFO  :  Exit block 1
> > > fcntl17     0  TWARN  :  tst_tmpdir.c:342: tst_rmdir: rmobj(/tmpdir/ltp-jRFBtBQhhx/LTP_fcn9Xy4hM) failed: unlink(/tmpdir/ltp-jRFBtBQhhx/LTP_fcn9Xy4hM) failed; errno=2: ENOENT
> > > 
> > > 
> > > Steps used after installing latest ltp:
> > > 
> > > $ mkdir /tmpdir
> > > $ yum install nfs-utils  -y
> > > $ echo "/media *(rw,no_root_squash,sync)" >/etc/exports
> > > $ systemctl start nfs-server.service
> > > $ mount -o rw,nfsvers=3 127.0.0.1:/media /tmpdir
> > > $ cd /opt/ltp
> > > $ ./runltp -d /tmpdir -s fcntl17
> > > 
> > > 
> > > 
> > > This does not happen in 5.15.153 tag.
> > > 
> > > Adding nfs people to the CC list
> > 
> > The reproducer uses NFSv3, but the bug report says NFSv4
> > at the top.
> > 
> > I was able to reproduce this on my nfsd-5.15.y branch
> > with NFSv3.
> > 
> > A bisect would be most helpful.
> > 
> 
> I was able to bisect: here are the results:
> 
> 
> 
> 2267b2e84593bd3d61a1188e68fba06307fa9dab is the first bad commit
> commit 2267b2e84593bd3d61a1188e68fba06307fa9dab
> Author: Alexander Aring <aahringo@xxxxxxxxxx>
> Date:   Tue Sep 12 17:53:18 2023 -0400
> 
>     lockd: introduce safe async lock op
> 
>     [ Upstream commit 2dd10de8e6bcbacf85ad758b904543c294820c63 ]
> 
>     This patch reverts mostly commit 40595cdc93ed ("nfs: block notification
>     on fs with its own ->lock") and introduces an EXPORT_OP_ASYNC_LOCK
>     export flag to signal that the "own ->lock" implementation supports
>     async lock requests. The only main user is DLM that is used by GFS2 and
>     OCFS2 filesystem. Those implement their own lock() implementation and
>     return FILE_LOCK_DEFERRED as return value. Since commit 40595cdc93ed
>     ("nfs: block notification on fs with its own ->lock") the DLM
>     implementation were never updated. This patch should prepare for DLM
>     to set the EXPORT_OP_ASYNC_LOCK export flag and update the DLM
>     plock implementation regarding to it.
> 
>     Acked-by: Jeff Layton <jlayton@xxxxxxxxxx>
>     Signed-off-by: Alexander Aring <aahringo@xxxxxxxxxx>
>     Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx>
> 
>  Documentation/filesystems/nfs/exporting.rst |  7 +++++++
>  fs/lockd/svclock.c                          |  4 +---
>  fs/nfsd/nfs4state.c                         | 10 +++++++---
>  include/linux/exportfs.h                    | 14 ++++++++++++++
>  4 files changed, 29 insertions(+), 6 deletions(-)
> 
> Bisect log:
> ==========
> 
> git bisect start
> # status: waiting for both good and bad commits
> # bad: [cdfd0a7f01396303e9d4fb3513a1127636f12e5e] Linux 5.15.154
> git bisect bad cdfd0a7f01396303e9d4fb3513a1127636f12e5e
> # status: waiting for good commit(s), bad commit known
> # good: [9465fef4ae351749f7068da8c78af4ca27e61928] Linux 5.15.153
> git bisect good 9465fef4ae351749f7068da8c78af4ca27e61928
> # good: [4420d19ed4e4fe2adc9bed8a49bf195db1137458] NFSD: Report average age
> of filecache items
> git bisect good 4420d19ed4e4fe2adc9bed8a49bf195db1137458
> # good: [94e412c945e64579798204aee7bc669d0acfaf79] nfsd: fix courtesy client
> with deny mode handling in nfs4_upgrade_open
> git bisect good 94e412c945e64579798204aee7bc669d0acfaf79
> # bad: [254f1c2521716cafc63530750ce313059f5d5979] iwlwifi: mvm: rfi: use
> kmemdup() to replace kzalloc + memcpy
> git bisect bad 254f1c2521716cafc63530750ce313059f5d5979
> # bad: [e635f652696ef6f1230621cfd89c350cb5ec6169] serial: sc16is7xx: convert
> from _raw_ to _noinc_ regmap functions for FIFO
> git bisect bad e635f652696ef6f1230621cfd89c350cb5ec6169
> # good: [05b452e8748bcf92c00725691437e16d46af7c28] nfsd: Fix creation time
> serialization order
> git bisect good 05b452e8748bcf92c00725691437e16d46af7c28
> # bad: [ccd9fe71b9ee46ebcecec8aec5c4f1e1ddd35dfd] nfsd: Fix a regression in
> nfsd_setattr()
> git bisect bad ccd9fe71b9ee46ebcecec8aec5c4f1e1ddd35dfd
> # bad: [2267b2e84593bd3d61a1188e68fba06307fa9dab] lockd: introduce safe
> async lock op
> git bisect bad 2267b2e84593bd3d61a1188e68fba06307fa9dab
> # good: [56e5eeff6cfa4bd6ffa2b2ae5b8bfc1c28044faf] nfsd: separate
> nfsd_last_thread() from nfsd_put()
> git bisect good 56e5eeff6cfa4bd6ffa2b2ae5b8bfc1c28044faf
> # good: [6e5fed48d8b7b25f8517a1292b62a3a86a5aec91] NFSD: fix possible oops
> when nfsd/pool_stats is closed.
> git bisect good 6e5fed48d8b7b25f8517a1292b62a3a86a5aec91
> # first bad commit: [2267b2e84593bd3d61a1188e68fba06307fa9dab] lockd:
> introduce safe async lock op
> 
> 
> Hope the above might help.

Nice work. Thanks!


> I didnot test the revert of culprit commit on top of 5.15.154 yet.

Please try reverting that one -- it's very close to the top so one
or two others might need to be pulled off as well.

I expect this is due to a missing pre-requisite commit.


-- 
Chuck Lever