Re: [PATCH v2 04/18] nfs: separate locked regions in nfs_clear_verifier_directory()

On Tue, 25 Aug 2026, Chuck Lever wrote:
> On Sat, Aug 15, 2026, at 2:21 PM, NeilBrown wrote:
> > nfs_clear_verifier_directory() locks the dir to call
> >   nfs_unset_verifier_delegated()
> > and also to walk the d_children list.
> > These are two separate needs and they don't need to both
> > be in the same locked region.
> >
> > Subsequent patches will use a helper for walking the d_children list and
> > that helper may drop the lock temporarily.  To simplify this transition,
> > split the locked range by dropping and retaking the lock between to make
> > two separate locked regions.
> 
> I hit this running xfstests against tmpfs with the series applied.  The
> trigger was incidental (an unrelated process chmod'ed a file on an
> NFSv4 mount while xfstests was running) so it is not deterministic, but
> the path looks reachable by any SETATTR that returns a delegation on a
> directory:
> 
>   BUG: sleeping function called from invalid context at fs/nfs/dir.c:1497
>   in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 2385380
>   preempt_count: 1, expected: 0
>   Call Trace:
>    __might_resched+0x165/0x180
>    nfs_clear_verifier_delegated+0x112/0x250 [nfs]
>    nfs_start_delegation_return+0xa1/0x130 [nfsv4]
>    nfs4_inode_return_delegation+0x1b/0xf0 [nfsv4]
>    nfs4_inode_make_writeable+0xa9/0x130 [nfsv4]
>    nfs4_proc_setattr+0xa2/0xf0 [nfsv4]
>    nfs_setattr+0x1f2/0x480 [nfs]
>    notify_change+0x321/0x380
>    chmod_common+0x10d/0x1c0
>    do_fchmodat+0x8a/0x110
>    __x64_sys_chmod+0x1d/0x30
> 
> Line 1497 is the d_for_each_positive_child() call.

Thanks for testing!

nfs_clear_verifier_deleted() holds the inode i_lock across
nfs_clear_verifier_directory() which walks the the d_children list.
I'll need to use d_find_alias() to get the dentry like I did in the
notify code.
Hopefully I'll send another RFC tomorrow.

Thanks,
NeilBrown