Skip to contentSkip to conversation

nfs-utils: Something is wrong in is_vers4()

DISCUSSION NeilBrown 3 participants 7 messages Jul 31, 2012 – Aug 6, 2012

Subsystems (2)
lore
Share & tools
lore

NeilBrown

author
#0 lore raw

in nfs-utils, in utils/mount/mount_libmount.c there is a function

is_vers4()

which
/* returns: error = -1, success = 0 , unknown = 1 */

which seems odd to me... I would have chosen '1' for success (it is vers 4), 0 for failure (not vers 4), and maybe -1 for error (something went wrong).

This is used as follows:

			switch (is_vers4(cxt)) {
			case 0:
				/* We ignore the error from nfs_umount23.
				 * If the actual umount succeeds (in del_mtab),
				 * we don't want to signal an error, as that
				 * could cause /sbin/mount to retry!
				 */
				nfs_umount23(mnt_context_get_source(cxt), opts);
				break;
			case 1:			/* unknown */
				break;
			default:		/* error */
				goto err;
			}

so in the '0' (success, it is vers4) case we do nfs_umount23. Odd. In the '1' (unknown, so presumably vers2 or 3) we don't. Ever. Very odd.

So we don't currently send MOUNT_UMNT requests for v2 or v3. We don't for v4 either because nfs_umount_do_umnt contains:

	/* Skip UMNT call for vers=4 mounts */
	if (nfs_pmap.pm_vers == 4)
		return EX_SUCCESS;

so maybe is_vers4 isn't needed?

Looks like something needs to be fixed here but I'm not entirely sure what. Karel??

BTW Steve, Karel's "[PATCH] umount.nfs: ignore non-nfs filesystems" which appears in email:

From: Karel Zak <[email protected]>
To: NeilBrown <[email protected]>
Cc: Steve Dickson <[email protected]>, NFS <[email protected]>
Subject: Re: [PATCH] umount.nfs: restore correct error status when umount fails.
Date: Thu, 12 Jul 2012 18:44:20 +0200

hasn't been applied, but probably should be.

Thanks,
NeilBrown

#1 lore raw
Quoted 12 lines · NeilBrown

in nfs-utils, in utils/mount/mount_libmount.c there is a function

is_vers4()

which
/* returns: error = -1, success = 0 , unknown = 1 */

which seems odd to me... I would have chosen '1' for success (it is vers 4), 0 for failure (not vers 4), and maybe -1 for error (something went wrong).

This is used as follows:

Quoted 14 lines
			switch (is_vers4(cxt)) {
			case 0:
				/* We ignore the error from nfs_umount23.
				 * If the actual umount succeeds (in del_mtab),
				 * we don't want to signal an error, as that
				 * could cause /sbin/mount to retry!
				 */
				nfs_umount23(mnt_context_get_source(cxt), opts);
				break;
			case 1:			/* unknown */
				break;
			default:		/* error */
				goto err;
			}

The same code you can found in original non-libmount version in nfsumount.c *but* with correct nfs_umount_is_vers4() return codes.

Quoted 3 lines

so in the '0' (success, it is vers4) case we do nfs_umount23. Odd. In the '1' (unknown, so presumably vers2 or 3) we don't. Ever. Very odd.

You're right, should be (copy & past from nfsumount.c):

* Returns 1 if "mc" is an NFSv4 mount, zero if not, and
* -1 if some error occurred.

...probably my bug. Sorry.

Quoted 11 lines

So we don't currently send MOUNT_UMNT requests for v2 or v3. We don't for v4 either because nfs_umount_do_umnt contains:

	/* Skip UMNT call for vers=4 mounts */
	if (nfs_pmap.pm_vers == 4)
		return EX_SUCCESS;

so maybe is_vers4 isn't needed?

Looks like something needs to be fixed here but I'm not entirely sure what. Karel??

The is_vers4() should be fixed to be compatible with the original code.

Quoted 10 lines

BTW Steve, Karel's "[PATCH] umount.nfs: ignore non-nfs filesystems" which appears in email:

From: Karel Zak <[email protected]>
To: NeilBrown <[email protected]>
Cc: Steve Dickson <[email protected]>, NFS <[email protected]>
Subject: Re: [PATCH] umount.nfs: restore correct error status when umount fails.
Date: Thu, 12 Jul 2012 18:44:20 +0200

hasn't been applied, but probably should be.

Yes.

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#2 lore raw
In reply to NeilBrown
Quoted 45 lines · NeilBrown

in nfs-utils, in utils/mount/mount_libmount.c there is a function

is_vers4()

which
/* returns: error = -1, success = 0 , unknown = 1 */

which seems odd to me... I would have chosen '1' for success (it is vers 4), 0 for failure (not vers 4), and maybe -1 for error (something went wrong).

This is used as follows:

			switch (is_vers4(cxt)) {
			case 0:
				/* We ignore the error from nfs_umount23.
				 * If the actual umount succeeds (in del_mtab),
				 * we don't want to signal an error, as that
				 * could cause /sbin/mount to retry!
				 */
				nfs_umount23(mnt_context_get_source(cxt), opts);
				break;
			case 1:			/* unknown */
				break;
			default:		/* error */
				goto err;
			}

so in the '0' (success, it is vers4) case we do nfs_umount23. Odd. In the '1' (unknown, so presumably vers2 or 3) we don't. Ever. Very odd.

So we don't currently send MOUNT_UMNT requests for v2 or v3. We don't for v4 either because nfs_umount_do_umnt contains:

	/* Skip UMNT call for vers=4 mounts */
	if (nfs_pmap.pm_vers == 4)
		return EX_SUCCESS;

so maybe is_vers4 isn't needed?

Looks like something needs to be fixed here but I'm not entirely sure what. Karel??

BTW Steve, Karel's "[PATCH] umount.nfs: ignore non-nfs filesystems" which appears in email:

Quoted 7 lines

From: Karel Zak <[email protected]>
To: NeilBrown <[email protected]>
Cc: Steve Dickson <[email protected]>, NFS <[email protected]>
Subject: Re: [PATCH] umount.nfs: restore correct error status when umount fails.
Date: Thu, 12 Jul 2012 18:44:20 +0200

hasn't been applied, but probably should be.

I beg to differ....

commit 76908c3f14a12e865054ea5d6e4cad201c28839a

Author: NeilBrown <[email protected]>
Date:   Mon Jul 16 08:43:28 2012 -0400

mount.nfs: restore correct error status when umount fails

Or am I missing something?

steved.

#3 lore raw
Quoted 12 lines · Steve Dickson
Quoted 10 lines

BTW Steve, Karel's "[PATCH] umount.nfs: ignore non-nfs filesystems" which appears in email:

From: Karel Zak <[email protected]>
To: NeilBrown <[email protected]>
Cc: Steve Dickson <[email protected]>, NFS <[email protected]>
Subject: Re: [PATCH] umount.nfs: restore correct error status when umount fails.
Date: Thu, 12 Jul 2012 18:44:20 +0200

hasn't been applied, but probably should be.

I beg to differ....

Quoted 7 lines

commit 76908c3f14a12e865054ea5d6e4cad201c28839a

Author: NeilBrown <[email protected]>
Date:   Mon Jul 16 08:43:28 2012 -0400

mount.nfs: restore correct error status when umount fails

Or am I missing something?

Yes, Neil found two problems:

    http://www.spinics.net/lists/linux-nfs/msg31466.html

    Karel
Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#4 lore raw
Quoted 17 lines · Karel Zak
Quoted 16 lines · Steve Dickson
Quoted 10 lines

BTW Steve, Karel's "[PATCH] umount.nfs: ignore non-nfs filesystems" which appears in email:

From: Karel Zak <[email protected]>
To: NeilBrown <[email protected]>
Cc: Steve Dickson <[email protected]>, NFS <[email protected]>
Subject: Re: [PATCH] umount.nfs: restore correct error status when umount fails.
Date: Thu, 12 Jul 2012 18:44:20 +0200

hasn't been applied, but probably should be.

I beg to differ....

commit 76908c3f14a12e865054ea5d6e4cad201c28839a

Author: NeilBrown <[email protected]>
Date:   Mon Jul 16 08:43:28 2012 -0400
Quoted 7 lines
Quoted 3 lines

mount.nfs: restore correct error status when umount fails

Or am I missing something?

Yes, Neil found two problems:

    http://www.spinics.net/lists/linux-nfs/msg31466.html

Ah... Found it and committed it...

steved.

author
#5 lore raw
Quoted 22 lines · On Wed, 01 Aug 2012 11:22:04 -0400 Steve Dickson <[email protected]> wrote:
Quoted 19 lines · Karel Zak
Quoted 18 lines · Steve Dickson
Quoted 10 lines

BTW Steve, Karel's "[PATCH] umount.nfs: ignore non-nfs filesystems" which appears in email:

From: Karel Zak <[email protected]>
To: NeilBrown <[email protected]>
Cc: Steve Dickson <[email protected]>, NFS <[email protected]>
Subject: Re: [PATCH] umount.nfs: restore correct error status when umount fails.
Date: Thu, 12 Jul 2012 18:44:20 +0200

hasn't been applied, but probably should be.

I beg to differ....

commit 76908c3f14a12e865054ea5d6e4cad201c28839a

Author: NeilBrown <[email protected]>
Date:   Mon Jul 16 08:43:28 2012 -0400

mount.nfs: restore correct error status when umount fails

Quoted 8 lines
Quoted 5 lines
Quoted 1 line

Or am I missing something?

Yes, Neil found two problems:

    http://www.spinics.net/lists/linux-nfs/msg31466.html

Ah... Found it and committed it...

steved.

Thanks.
So here is one more :-) This fixes the the bug in $SUBJECT

NeilBrown

From: Neil Brown <[email protected]>
Date: Thu, 2 Aug 2012 10:20:13 +1000
Subject: [PATCH] umount: use correct return value for is_vers4.

is_vers4 in mount_libmount.c is based on nfs_umount_is_vers4 in nfsumount.c, except the return values are reversed. The result of this is:
- a MOUNT_UMNT call is not sent when an NFSv3 or NFSv2 filesystem is unmounted
- a MOUNT_UMNT call *is* sent when and 'nfs4' filesystem is unmounted (but not when an 'nfs -o vers=4 filesystem is unmounted, as that is checked elsewhere).

utils/mount/mount_libmount.c +3 -3
index ddf61b2..701d41e 100644
@@ -140,14 +140,14 @@ static int try_mount(struct libmnt_context *cxt, int bg)
 	return ret;
 }
 
-/* returns: error = -1, success = 0 , unknown = 1 */
+/* returns: error = -1, success = 1 , not vers4 == 0 */
 static int is_vers4(struct libmnt_context *cxt)
 {
 	struct libmnt_fs *fs = mnt_context_get_fs(cxt);
 	struct libmnt_table *tb = NULL;
 	const char *src = mnt_context_get_source(cxt),
 		   *tgt = mnt_context_get_target(cxt);
-	int rc = 1;
+	int rc = 0;
 
 	if (!src || !tgt)
 		return -1;
@@ -163,7 +163,7 @@ static int is_vers4(struct libmnt_context *cxt)
 	if (fs) {
 		const char *type = mnt_fs_get_fstype(fs);
 		if (type && strcmp(type, "nfs4") == 0)
-			rc = 0;
+			rc = 1;
 	}
 	mnt_free_table(tb);
 	return rc;
#6 lore raw
Quoted 10 lines · NeilBrown

is_vers4 in mount_libmount.c is based on nfs_umount_is_vers4 in nfsumount.c, except the return values are reversed. The result of this is:
- a MOUNT_UMNT call is not sent when an NFSv3 or NFSv2 filesystem is unmounted
- a MOUNT_UMNT call *is* sent when and 'nfs4' filesystem is unmounted (but not when an 'nfs -o vers=4 filesystem is unmounted, as that is checked elsewhere).

Committed...

steved.

Quoted 31 lines
utils/mount/mount_libmount.c +3 -3
index ddf61b2..701d41e 100644
@@ -140,14 +140,14 @@ static int try_mount(struct libmnt_context *cxt, int bg)
 	return ret;
 }
 
-/* returns: error = -1, success = 0 , unknown = 1 */
+/* returns: error = -1, success = 1 , not vers4 == 0 */
 static int is_vers4(struct libmnt_context *cxt)
 {
 	struct libmnt_fs *fs = mnt_context_get_fs(cxt);
 	struct libmnt_table *tb = NULL;
 	const char *src = mnt_context_get_source(cxt),
 		   *tgt = mnt_context_get_target(cxt);
-	int rc = 1;
+	int rc = 0;
 
 	if (!src || !tgt)
 		return -1;
@@ -163,7 +163,7 @@ static int is_vers4(struct libmnt_context *cxt)
 	if (fs) {
 		const char *type = mnt_fs_get_fstype(fs);
 		if (type && strcmp(type, "nfs4") == 0)
-			rc = 0;
+			rc = 1;
 	}
 	mnt_free_table(tb);
 	return rc;