From: NeilBrown <neilb@suse.de>
To: Al Viro <viro@zeniv.linux.org.uk>,
Linus Torvalds <torvalds@linux-foundation.org>,
Daire Byrne <daire@dneg.com>,
Trond Myklebust <trond.myklebust@hammerspace.com>,
Chuck Lever <chuck.lever@oracle.com>
Cc: Linux NFS Mailing List <linux-nfs@vger.kernel.org>,
linux-fsdevel@vger.kernel.org,
LKML <linux-kernel@vger.kernel.org>
Subject: [PATCH 02/10] VFS: move EEXIST and ENOENT tests into lookup_hash_update()
Date: Fri, 26 Aug 2022 12:10:43 +1000 [thread overview]
Message-ID: <166147984373.25420.5973159222199992210.stgit@noble.brown> (raw)
In-Reply-To: <166147828344.25420.13834885828450967910.stgit@noble.brown>
Moving common error handling into lookup_hash_update() simplifies
callers.
A future patch will export this functionality to nfsd, and the more code
we put in the interface, the less code will be needed in nfsd.
EEXIST is returned if LOOKUP_EXCL is set and dentry is positivie.
ENOENT is returned if LOOKUP_CREAT is NOT set and dentry is negative.
This involves seting LOOKUP_EXCL in cases where it wasn't before.
In particular, when creating a non-dir named "foo/", we want the
EEXIST error, but don't want to actually create anything.
Some filesystems assume LOOKUP_EXCL implies LOOKUP_CREATE, so ensure it
does when calling ->lookup().
Signed-off-by: NeilBrown <neilb@suse.de>
---
fs/namei.c | 58 ++++++++++++++++++++++++++++++++++------------------------
1 file changed, 34 insertions(+), 24 deletions(-)
diff --git a/fs/namei.c b/fs/namei.c
index c008dfd01e30..09c2d007814a 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -1562,7 +1562,13 @@ static struct dentry *lookup_dcache(const struct qstr *name,
{
struct dentry *dentry = d_lookup(dir, name);
if (dentry) {
- int error = d_revalidate(dentry, flags);
+ int error;
+ /* Some filesystems assume EXCL -> CREATE, so make
+ * sure it does.
+ */
+ if (!(flags & LOOKUP_CREATE))
+ flags &= ~LOOKUP_EXCL;
+ error = d_revalidate(dentry, flags);
if (unlikely(error <= 0)) {
if (!error)
d_invalidate(dentry);
@@ -1621,6 +1627,8 @@ static struct dentry *__lookup_hash(const struct qstr *name,
* or shared lock depending on the fs preference, then do a lookup,
* and then set the DCACHE_PAR_UPDATE bit on the child if a shared lock
* was taken on the parent.
+ * If LOOKUP_EXCL, name should not already exist, else -EEXIST
+ * If not LOOKUP_CREATE, name should already exist, else -ENOENT
*/
static struct dentry *lookup_hash_update(
const struct qstr *name,
@@ -1651,6 +1659,24 @@ static struct dentry *lookup_hash_update(
dput(dentry);
goto retry;
}
+ if (flags & LOOKUP_EXCL) {
+ if (d_is_positive(dentry)) {
+ d_lookup_done(dentry);
+ d_unlock_update(dentry);
+ dput(dentry);
+ err = -EEXIST;
+ goto out_err;
+ }
+ }
+ if (!(flags & LOOKUP_CREATE)) {
+ if (!dentry->d_inode) {
+ d_lookup_done(dentry);
+ d_unlock_update(dentry);
+ dput(dentry);
+ err = -ENOENT;
+ goto out_err;
+ }
+ }
return dentry;
out_err:
@@ -3868,7 +3894,7 @@ static struct dentry *filename_create_one(struct qstr *last, struct path *path,
struct dentry *dentry;
bool want_dir = lookup_flags & LOOKUP_DIRECTORY;
unsigned int reval_flag = lookup_flags & LOOKUP_REVAL;
- unsigned int create_flags = LOOKUP_CREATE | LOOKUP_EXCL;
+ unsigned int create_flag = LOOKUP_CREATE;
int err2;
int error;
@@ -3879,26 +3905,16 @@ static struct dentry *filename_create_one(struct qstr *last, struct path *path,
* '/', and a directory wasn't requested.
*/
if (last->name[last->len] && !want_dir)
- create_flags = 0;
+ /* Name was foo/bar/ but not creating a directory, so
+ * we won't try to create - result will be either -ENOENT
+ * or -EEXIST.
+ */
+ create_flag = 0;
dentry = lookup_hash_update(last, path->dentry,
- reval_flag | create_flags, wq);
+ reval_flag | create_flag | LOOKUP_EXCL, wq);
if (IS_ERR(dentry))
goto drop_write;
- error = -EEXIST;
- if (d_is_positive(dentry))
- goto fail;
-
- /*
- * Special case - lookup gave negative, but... we had foo/bar/
- * From the vfs_mknod() POV we just have a negative dentry -
- * all is fine. Let's be bastards - you had / on the end, you've
- * been asking for (non-existent) directory. -ENOENT for you.
- */
- if (unlikely(!create_flags)) {
- error = -ENOENT;
- goto fail;
- }
if (unlikely(err2)) {
error = err2;
goto fail;
@@ -4292,10 +4308,6 @@ int do_rmdir(int dfd, struct filename *name)
error = PTR_ERR(dentry);
if (IS_ERR(dentry))
goto exit3;
- if (!dentry->d_inode) {
- error = -ENOENT;
- goto exit4;
- }
error = security_path_rmdir(&path, dentry);
if (error)
goto exit4;
@@ -4435,8 +4447,6 @@ int do_unlinkat(int dfd, struct filename *name)
if (last.name[last.len])
goto slashes;
inode = dentry->d_inode;
- if (d_is_negative(dentry))
- goto slashes;
ihold(inode);
error = security_path_unlink(&path, dentry);
if (error)
next prev parent reply other threads:[~2022-08-26 2:17 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-26 2:10 [PATCH/RFC 00/10 v5] Improve scalability of directory operations NeilBrown
2022-08-26 2:10 ` [PATCH 09/10] VFS: add LOOKUP_SILLY_RENAME NeilBrown
2022-08-27 1:21 ` Al Viro
2022-08-29 3:15 ` NeilBrown
2022-08-26 2:10 ` [PATCH 01/10] VFS: support parallel updates in the one directory NeilBrown
2022-08-26 19:06 ` Linus Torvalds
2022-08-26 23:06 ` NeilBrown
2022-08-27 0:13 ` Linus Torvalds
2022-08-27 0:23 ` Al Viro
2022-08-27 21:14 ` Al Viro
2022-08-27 0:17 ` Al Viro
2022-09-01 0:31 ` NeilBrown
2022-09-01 3:44 ` Al Viro
2022-08-27 3:43 ` Al Viro
2022-08-29 1:59 ` NeilBrown
2022-09-03 0:06 ` Al Viro
2022-09-03 1:40 ` NeilBrown
2022-09-03 2:12 ` Al Viro
2022-09-03 17:52 ` Al Viro
2022-09-04 23:33 ` NeilBrown
2022-08-26 2:10 ` [PATCH 08/10] NFSD: allow parallel creates from nfsd NeilBrown
2022-08-27 4:37 ` Al Viro
2022-08-29 3:12 ` NeilBrown
2022-08-26 2:10 ` [PATCH 05/10] VFS: export done_path_update() NeilBrown
2022-08-26 2:10 ` NeilBrown [this message]
2022-08-26 2:10 ` [PATCH 06/10] VFS: support concurrent renames NeilBrown
2022-08-27 4:12 ` Al Viro
2022-08-29 3:08 ` NeilBrown
2022-08-26 2:10 ` [PATCH 10/10] NFS: support parallel updates in the one directory NeilBrown
2022-08-26 15:31 ` John Stoffel
2022-08-26 23:13 ` NeilBrown
2022-08-26 2:10 ` [PATCH 03/10] VFS: move want_write checks into lookup_hash_update() NeilBrown
2022-08-27 3:48 ` Al Viro
2022-08-26 2:10 ` [PATCH 04/10] VFS: move dput() and mnt_drop_write() into done_path_update() NeilBrown
2022-08-26 2:10 ` [PATCH 07/10] VFS: hold DCACHE_PAR_UPDATE lock across d_revalidate() NeilBrown
2022-08-26 14:42 ` [PATCH/RFC 00/10 v5] Improve scalability of directory operations John Stoffel
2022-08-26 23:30 ` NeilBrown
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=166147984373.25420.5973159222199992210.stgit@noble.brown \
--to=neilb@suse.de \
--cc=chuck.lever@oracle.com \
--cc=daire@dneg.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nfs@vger.kernel.org \
--cc=torvalds@linux-foundation.org \
--cc=trond.myklebust@hammerspace.com \
--cc=viro@zeniv.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).