mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Chris Wright <chrisw@osdl.org>
To: Matthew Wilcox <matthew@wil.cx>
Cc: Chris Wright <chrisw@osdl.org>,
	Ivan Kalatchev <ivan.kalatchev@esg.ca>,
	linux-kernel@vger.kernel.org
Subject: [TEST PATCH] Re: kernel 2.6.8 bug in fs/locks.c
Date: Fri, 1 Oct 2004 17:52:30 -0700	[thread overview]
Message-ID: <20041001175230.F1973@build.pdx.osdl.net> (raw)
In-Reply-To: <20041001122332.E1973@build.pdx.osdl.net>; from chrisw@osdl.org on Fri, Oct 01, 2004 at 12:23:33PM -0700

* Chris Wright (chrisw@osdl.org) wrote:
> * Ivan Kalatchev (ivan.kalatchev@esg.ca) wrote:
> > I'm using pthreads for each user-application connections. To protect
> > configuration file from corruption I used file locking mechanism - fcntl
> > with F_WRLCK/F_RDLCK.
> 
> I must be confused.  pthreads and fcntl locking...that does't give
> proper exclusion?  The BUG, however, is no good.  Despite the fact
> that it appears to come at the result of an application bug, we should
> be able to handle this w/out a BUG.  AFAICT, one thread has closed the
> file descriptor, whilst another is mucking with the locks.  So the locker
> winds up holding the last ref to the filp.  This blows the logic of when
> locks_remove_posix gets called.  Thanks for the bug report.

This is a rather ugly fix, but it demonstrates the hole.  It's possible
for a posix lock to get applied after the final close of the file.  This
means locks_remove_posix misses the yet to be created posix lock, and
on final fput when leaving fcntl, locks_remove_flock will BUG().
I was able to reproduce Ivan's bugreport.  I doubt it's the right/best
fix, but it's functional (no more BUG()).  Against -bk-curr.

Signed-off-by: Chris Wright <chrisw@osdl.org>

===== fs/fcntl.c 1.40 vs edited =====
--- 1.40/fs/fcntl.c	2004-09-02 02:48:03 -07:00
+++ edited/fs/fcntl.c	2004-10-01 17:34:15 -07:00
@@ -364,7 +364,7 @@
 
 asmlinkage long sys_fcntl(int fd, unsigned int cmd, unsigned long arg)
 {	
-	struct file *filp;
+	struct file *filp, *filp2;
 	long err = -EBADF;
 
 	filp = fget(fd);
@@ -379,6 +379,11 @@
 
 	err = do_fcntl(fd, cmd, arg, filp);
 
+	filp2 = fget(fd);
+	if (filp2 != filp)
+		locks_remove_posix(filp, current->files);
+	if (filp2)
+		fput(filp2);
  	fput(filp);
 out:
 	return err;
@@ -387,7 +392,7 @@
 #if BITS_PER_LONG == 32
 asmlinkage long sys_fcntl64(unsigned int fd, unsigned int cmd, unsigned long arg)
 {	
-	struct file * filp;
+	struct file *filp, *filp2;
 	long err;
 
 	err = -EBADF;
@@ -414,6 +419,12 @@
 			err = do_fcntl(fd, cmd, arg, filp);
 			break;
 	}
+
+	filp2 = fget(fd);
+	if (filp2 != filp)
+		locks_remove_posix(filp, current->files);
+	if (filp2)
+		fput(filp2);
 	fput(filp);
 out:
 	return err;

      reply	other threads:[~2004-10-02  0:53 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-10-01 16:24 Ivan Kalatchev
2004-10-01 19:23 ` Chris Wright
2004-10-02  0:52   ` Chris Wright [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=20041001175230.F1973@build.pdx.osdl.net \
    --to=chrisw@osdl.org \
    --cc=ivan.kalatchev@esg.ca \
    --cc=linux-kernel@vger.kernel.org \
    --cc=matthew@wil.cx \
    /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

all inboxes | Powered by JetHome®