On Mon, Feb 19, 2024 at 04:02:15AM +0100, Roland Mainz wrote: > On Sun, Feb 18, 2024 at 5:45 PM Chuck Lever <chuck.lever@xxxxxxxxxx> wrote: > > On Sun, Feb 18, 2024 at 01:00:33PM +0100, Roland Mainz wrote: > > > I'm getting the following errors from nfsd in the kernel log on a regular basis: > > > ---- snip ---- > > > [349278.877256] nfsd: inode locked twice during operation. > > > [349279.599457] nfsd: inode locked twice during operation. > > > [349280.302697] nfsd: inode locked twice during operation. > > > [349280.803115] nfsd: inode locked twice during operation. > > > ---- snip ---- > > > > > > nfsd runs on "Linux 5.10.0-22-rt-686-pae #1 SMP PREEMPT_RT Debian > > > 5.10.178-3 (2023-04-22) i686 GNU/Linux", exported filesystem is btrfs > > > on a SSD. > > > NFSv4.1 client is the Windows ms-nfs41-client (git > > > #e67a792c4249600164852cfc470ac0acdb9f043b) compiling gcc under Cygwin > > > 3.5.0. > > > > > > Is the nfsd issue known, and if "yes" is there a patch ? > > > > I believe that warning message vanishes as a side-effect of this > > series of commits: > > > > 7fe2a71dda34 NFSD: introduce struct nfsd_attrs > > 93adc1e391a7 NFSD: set attributes when creating symlinks > > d6a97d3f589a NFSD: add security label to struct nfsd_attrs > > c0cbe70742f4 NFSD: add posix ACLs to struct nfsd_attrs > > 927bfc5600cd NFSD: change nfsd_create()/nfsd_symlink() to unlock directory before returning. > > b677c0c63a13 NFSD: always drop directory lock in nfsd_unlink() > > e18bcb33bc5b NFSD: only call fh_unlock() once in nfsd_link() > > 19d008b46941 NFSD: reduce locking in nfsd_lookup() > > debf16f0c671 NFSD: use explicit lock/unlock for directory ops > > bb4d53d66e4b NFSD: use (un)lock_inode instead of fh_(un)lock for file operations > > dd8dd403d7b2 NFSD: discard fh_locked flag and fh_lock/fh_unlock > > > > plus this fix: > > > > 00801cd92d91 NFSD: fix regression with setting ACLs. > > Ouch... ;-( > ... so the patch at the bottom of > https://patchwork.kernel.org/project/linux-nfs/patch/533299E9.6010806@xxxxxxxxx/#8291331 > is not sufficient, right ? That one made it into upstream as: commit 2336745e87a646a3dc9570f082b85df519ee523e Author: Kinglong Mee <kinglongmee@xxxxxxxxx> AuthorDate: Sat Mar 29 10:23:47 2014 +0800 Commit: J. Bruce Fields <bfields@xxxxxxxxxxxx> CommitDate: Sun Mar 30 10:47:34 2014 -0400 NFSD: Clear wcc data between compound ops Testing NFS4.0 by pynfs, I got some messeages as, "nfsd: inode locked twice during operation." When one compound RPC contains two or more ops that locks the filehandle,the second op will cause the message. As two SETATTR ops, after the first SETATTR, nfsd will not call fh_put() to release current filehandle, it means filehandle have unlocked with fh_post_saved = 1. The second SETATTR find fh_post_saved = 1, and printk the message. v2: introduce helper fh_clear_wcc(). Signed-off-by: Kinglong Mee <kinglongmee@xxxxxxxxx> Signed-off-by: J. Bruce Fields <bfields@xxxxxxxxxx> So every Linux kernel after v3.15 has it already. > > Any upstream Linux kernel after v6.0 should operate without that > > warning. I don't see those commits in origin/linux-5.10.y. > > Are the kernels in the Linux 6.6-stable branch "safe" as NFSv4.x for > NFSv4.1 client development ? Generally speaking, later releases are better. If you notice a specific issue, I can respond to that. -- Chuck Lever
