diff options
| author | Trond Myklebust <trond.myklebust@hammerspace.com> | 2024-02-24 15:59:28 -0500 |
|---|---|---|
| committer | Trond Myklebust <trond.myklebust@hammerspace.com> | 2024-03-09 09:14:51 -0500 |
| commit | 0460253913e50a2aec911fe83090d60397f17664 (patch) | |
| tree | 516ae74bb5674b684efd4968a28635da1462bf65 /fs/nfs/nfs4proc.c | |
| parent | 2fdbc20036acda9e5694db74a032d3c605323005 (diff) | |
| download | linux-0460253913e50a2aec911fe83090d60397f17664.tar.gz linux-0460253913e50a2aec911fe83090d60397f17664.tar.bz2 linux-0460253913e50a2aec911fe83090d60397f17664.zip | |
NFSv4: nfs4_do_open() is incorrectly triggering state recovery
We're seeing spurious calls to nfs4_schedule_stateid_recovery() from
nfs4_do_open() in situations where there is no trigger coming from the
server.
In theory the code path being triggered is supposed to notice that state
recovery happened while we were processing the open call result from the
server, before the open stateid is published. However in the years since
that code was added, we've also added the 'session draining' mechanism,
which ensures that the state recovery will wait until all the session
slots have been returned. In nfs4_do_open() the session slot is only
returned on exit of the function, so we don't need the legacy mechanism.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Diffstat (limited to 'fs/nfs/nfs4proc.c')
| -rw-r--r-- | fs/nfs/nfs4proc.c | 7 |
1 files changed, 1 insertions, 6 deletions
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 206b4607b29a..e9c4f7b9d44e 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3069,10 +3069,8 @@ static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata, fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx); struct inode *dir = d_inode(opendata->dir); unsigned long dir_verifier; - unsigned int seq; int ret; - seq = raw_seqcount_begin(&sp->so_reclaim_seqcount); dir_verifier = nfs_save_change_attribute(dir); ret = _nfs4_proc_open(opendata, ctx); @@ -3125,11 +3123,8 @@ static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata, if (ret != 0) goto out; - if (d_inode(dentry) == state->inode) { + if (d_inode(dentry) == state->inode) nfs_inode_attach_open_context(ctx); - if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) - nfs4_schedule_stateid_recovery(server, state); - } out: if (!opendata->cancelled) { |