Linux NFS development
 help / color / mirror / Atom feed
From: Trond Myklebust <trond.myklebust@fys.uio.no>
To: Wendy Cheng <wcheng@redhat.com>
Cc: Linux NFS Mailing List <nfs@lists.sourceforge.net>
Subject: Re: [PATCH] fix recursive nlm_file_mutex deadlock
Date: Thu, 10 Aug 2006 12:05:01 -0400	[thread overview]
Message-ID: <1155225902.10547.53.camel@localhost> (raw)
In-Reply-To: <1155224407.10547.45.camel@localhost>

[-- Attachment #1: Type: text/plain, Size: 754 bytes --]

On Thu, 2006-08-10 at 11:40 -0400, Trond Myklebust wrote:
> On Thu, 2006-08-10 at 11:24 -0400, Wendy Cheng wrote:
> 
> > We simply can't call nlm_release_file() as long as it requires
> > nlm_file_mutex (no matter where you move the locking line) within
> > nlm_traverse_files() code path. I'm ok with atomic variable though.
> 
> As long as we can guarantee that file->f_count doesn't drop to zero (and
> we can make that guarantee in nlm_traverse_file()), then the code is
> correct.

Doh! There is no reason whatsoever to be holding nlm_file_mutex across
the call to nlm_inspect_file. As you said, the only thing it does is
protect the list, and we don't care about that.

The simplest fix is just to do something like the following.

Cheers,
  Trond

[-- Attachment #2: linux-2.6.18-009-fix_nlm_traverse_files_deadlock.dif --]
[-- Type: message/rfc822, Size: 1653 bytes --]

From: Trond Myklebust <Trond.Myklebust@netapp.com>
Subject: No Subject
Date: Thu, 10 Aug 2006 12:01:51 -0400
Message-ID: <1155225711.10547.49.camel@localhost>

nlm_traverse_files() is not allowed to hold the nlm_file_mutex while calling
nlm_inspect file, since it may end up calling nlm_release_file() when
releaseing the blocks.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---

 fs/lockd/svcsubs.c |   15 +++++++++------
 1 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/fs/lockd/svcsubs.c b/fs/lockd/svcsubs.c
index 2a4df9b..01b4db9 100644
--- a/fs/lockd/svcsubs.c
+++ b/fs/lockd/svcsubs.c
@@ -237,19 +237,22 @@ static int
 nlm_traverse_files(struct nlm_host *host, int action)
 {
 	struct nlm_file	*file, **fp;
-	int		i;
+	int i, ret = 0;
 
 	mutex_lock(&nlm_file_mutex);
 	for (i = 0; i < FILE_NRHASH; i++) {
 		fp = nlm_files + i;
 		while ((file = *fp) != NULL) {
+			file->f_count++;
+			mutex_unlock(&nlm_file_mutex);
+
 			/* Traverse locks, blocks and shares of this file
 			 * and update file->f_locks count */
-			if (nlm_inspect_file(host, file, action)) {
-				mutex_unlock(&nlm_file_mutex);
-				return 1;
-			}
+			if (nlm_inspect_file(host, file, action))
+				ret = 1;
 
+			mutex_lock(&nlm_file_mutex);
+			file->f_count--;
 			/* No more references to this file. Let go of it. */
 			if (!file->f_blocks && !file->f_locks
 			 && !file->f_shares && !file->f_count) {
@@ -262,7 +265,7 @@ nlm_traverse_files(struct nlm_host *host
 		}
 	}
 	mutex_unlock(&nlm_file_mutex);
-	return 0;
+	return ret;
 }
 
 /*

[-- Attachment #3: Type: text/plain, Size: 373 bytes --]

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642

[-- Attachment #4: Type: text/plain, Size: 140 bytes --]

_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

      reply	other threads:[~2006-08-10 16:05 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-08-09 18:13 [PATCH] fix recursive nlm_file_mutex deadlock Wendy Cheng
2006-08-09 18:32 ` Wendy Cheng
2006-08-09 21:45 ` Trond Myklebust
2006-08-09 22:07   ` Wendy Cheng
2006-08-09 22:41     ` Trond Myklebust
2006-08-09 23:57     ` Trond Myklebust
2006-08-10 15:24       ` Wendy Cheng
2006-08-10 15:40         ` Trond Myklebust
2006-08-10 16:05           ` Trond Myklebust [this message]

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=1155225902.10547.53.camel@localhost \
    --to=trond.myklebust@fys.uio.no \
    --cc=nfs@lists.sourceforge.net \
    --cc=wcheng@redhat.com \
    /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