diff mbox series

[14/16] nfsd: close cached files prior to a REMOVE or RENAME that would replace target

Message ID 20190630135240.7490-15-trond.myklebust@hammerspace.com (mailing list archive)
State New, archived
Headers show
Series Cache open file descriptors in knfsd | expand

Commit Message

Trond Myklebust June 30, 2019, 1:52 p.m. UTC 
From: Jeff Layton <jeff.layton@primarydata.com>

It's not uncommon for some workloads to do a bunch of I/O to a file and
delete it just afterward. If knfsd has a cached open file however, then
the file may still be open when the dentry is unlinked. If the
underlying filesystem is nfs, then that could trigger it to do a
sillyrename.

On a REMOVE or RENAME scan the nfsd_file cache for open files that
correspond to the inode, and proactively unhash and put their
references. This should prevent any delete-on-last-close activity from
occurring, solely due to knfsd's open file cache.

This must be done synchronously though so we use the variants that call
flush_delayed_fput. There are deadlock possibilities if you call
flush_delayed_fput while holding locks, however. In the case of
nfsd_rename, we don't even do the lookups of the dentries to be renamed
until we've locked for rename.

Once we've figured out what the target dentry is for a rename, check to
see whether there are cached open files associated with it. If there
are, then unwind all of the locking, close them all, and then reattempt
the rename.

Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfsd/vfs.c | 62 +++++++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 53 insertions(+), 9 deletions(-)
diff mbox series

Patch

diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 58b6d8df95d4..f5cf64a40112 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -1590,6 +1590,26 @@  nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
 	goto out_unlock;
 }
 
+static void
+nfsd_close_cached_files(struct dentry *dentry)
+{
+	struct inode *inode = d_inode(dentry);
+
+	if (inode && S_ISREG(inode->i_mode))
+		nfsd_file_close_inode_sync(inode);
+}
+
+static bool
+nfsd_has_cached_files(struct dentry *dentry)
+{
+	bool		ret = false;
+	struct inode *inode = d_inode(dentry);
+
+	if (inode && S_ISREG(inode->i_mode))
+		ret = nfsd_file_is_cached(inode);
+	return ret;
+}
+
 /*
  * Rename a file
  * N.B. After this call _both_ ffhp and tfhp need an fh_put
@@ -1602,6 +1622,7 @@  nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
 	struct inode	*fdir, *tdir;
 	__be32		err;
 	int		host_err;
+	bool		has_cached = false;
 
 	err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
 	if (err)
@@ -1620,6 +1641,7 @@  nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
 	if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
 		goto out;
 
+retry:
 	host_err = fh_want_write(ffhp);
 	if (host_err) {
 		err = nfserrno(host_err);
@@ -1659,11 +1681,16 @@  nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
 	if (ffhp->fh_export->ex_path.dentry != tfhp->fh_export->ex_path.dentry)
 		goto out_dput_new;
 
-	host_err = vfs_rename(fdir, odentry, tdir, ndentry, NULL, 0);
-	if (!host_err) {
-		host_err = commit_metadata(tfhp);
-		if (!host_err)
-			host_err = commit_metadata(ffhp);
+	if (nfsd_has_cached_files(ndentry)) {
+		has_cached = true;
+		goto out_dput_old;
+	} else {
+		host_err = vfs_rename(fdir, odentry, tdir, ndentry, NULL, 0);
+		if (!host_err) {
+			host_err = commit_metadata(tfhp);
+			if (!host_err)
+				host_err = commit_metadata(ffhp);
+		}
 	}
  out_dput_new:
 	dput(ndentry);
@@ -1676,12 +1703,26 @@  nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
 	 * as that would do the wrong thing if the two directories
 	 * were the same, so again we do it by hand.
 	 */
-	fill_post_wcc(ffhp);
-	fill_post_wcc(tfhp);
+	if (!has_cached) {
+		fill_post_wcc(ffhp);
+		fill_post_wcc(tfhp);
+	}
 	unlock_rename(tdentry, fdentry);
 	ffhp->fh_locked = tfhp->fh_locked = false;
 	fh_drop_write(ffhp);
 
+	/*
+	 * If the target dentry has cached open files, then we need to try to
+	 * close them prior to doing the rename. Flushing delayed fput
+	 * shouldn't be done with locks held however, so we delay it until this
+	 * point and then reattempt the whole shebang.
+	 */
+	if (has_cached) {
+		has_cached = false;
+		nfsd_close_cached_files(ndentry);
+		dput(ndentry);
+		goto retry;
+	}
 out:
 	return err;
 }
@@ -1728,10 +1769,13 @@  nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 	if (!type)
 		type = d_inode(rdentry)->i_mode & S_IFMT;
 
-	if (type != S_IFDIR)
+	if (type != S_IFDIR) {
+		nfsd_close_cached_files(rdentry);
 		host_err = vfs_unlink(dirp, rdentry, NULL);
-	else
+	} else {
 		host_err = vfs_rmdir(dirp, rdentry);
+	}
+
 	if (!host_err)
 		host_err = commit_metadata(fhp);
 	dput(rdentry);