From: David Howells dhowells@redhat.com
[ Upstream commit 13fcc6356a94558a0a4857dc00cd26b3834a1b3e ]
When a lookup is done in an AFS directory, the filesystem will speculate and fetch up to 49 other statuses for files in the same directory and fetch those as well, turning them into inodes or updating inodes that already exist.
However, occasionally, a callback break might go missing due to NAT timing out, but the afs filesystem doesn't then realise that the directory is not up to date.
Alleviate this by using one of the status slots to check the directory in which the lookup is being done.
Reported-by: Dave Botsch botsch@cnf.cornell.edu Suggested-by: Jeffrey Altman jaltman@auristor.com Signed-off-by: David Howells dhowells@redhat.com Signed-off-by: Sasha Levin sashal@kernel.org --- fs/afs/dir.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/fs/afs/dir.c b/fs/afs/dir.c index d1e1caa23c8b3..3c486340b2208 100644 --- a/fs/afs/dir.c +++ b/fs/afs/dir.c @@ -658,7 +658,8 @@ static struct inode *afs_do_lookup(struct inode *dir, struct dentry *dentry,
cookie->ctx.actor = afs_lookup_filldir; cookie->name = dentry->d_name; - cookie->nr_fids = 1; /* slot 0 is saved for the fid we actually want */ + cookie->nr_fids = 2; /* slot 0 is saved for the fid we actually want + * and slot 1 for the directory */
read_seqlock_excl(&dvnode->cb_lock); dcbi = rcu_dereference_protected(dvnode->cb_interest, @@ -709,7 +710,11 @@ static struct inode *afs_do_lookup(struct inode *dir, struct dentry *dentry, if (!cookie->inodes) goto out_s;
- for (i = 1; i < cookie->nr_fids; i++) { + cookie->fids[1] = dvnode->fid; + cookie->statuses[1].cb_break = afs_calc_vnode_cb_break(dvnode); + cookie->inodes[1] = igrab(&dvnode->vfs_inode); + + for (i = 2; i < cookie->nr_fids; i++) { scb = &cookie->statuses[i];
/* Find any inodes that already exist and get their