Linux Device Mapper development
 help / color / mirror / Atom feed
From: malahal@us.ibm.com
To: dm-devel@redhat.com
Subject: Re: [PATCH] a deadlock bug in the kernel-side device mapper code
Date: Mon, 9 Nov 2009 14:24:25 -0800	[thread overview]
Message-ID: <20091109222425.GA3577@us.ibm.com> (raw)
In-Reply-To: <20091109175730.GA12556@us.ibm.com>

malahal@us.ibm.com [malahal@us.ibm.com] wrote:
> dm_get_from_kobject() seems to be a culprit though. It checks
> DMF_FREEING without a lock and then calls dm_get. Here is an untested
> patch. Made on top of _name_read_lock patch.

Found another place where the lock is not held correctly. This is an
update to my previous patch [dmf_freeing_lock.patch]

Description:

There are places where the code checks for DMF_FREEING bit flag and
conditionally calls dm_get(). This should be done under _minor_lock,
otherwise there is no point in checking for the bit flag.

Signed-off-by: malahal@us.ibm.com

diff -r e4c5c66b9a17 drivers/md/dm-ioctl.c
--- a/drivers/md/dm-ioctl.c	Mon Nov 09 13:38:38 2009 -0800
+++ b/drivers/md/dm-ioctl.c	Mon Nov 09 14:21:13 2009 -0800
@@ -1595,7 +1595,6 @@
 	if (!md)
 		return -ENXIO;
 
-	dm_get(md);
 	mutex_lock(&_name_read_lock);
 	hc = dm_get_mdptr(md);
 	if (!hc || hc->md != md) {
@@ -1610,7 +1609,6 @@
 
 out:
 	mutex_unlock(&_name_read_lock);
-	dm_put(md);
 
 	return r;
 }
diff -r e4c5c66b9a17 drivers/md/dm.c
--- a/drivers/md/dm.c	Mon Nov 09 13:38:38 2009 -0800
+++ b/drivers/md/dm.c	Mon Nov 09 14:21:13 2009 -0800
@@ -1998,28 +1998,24 @@
 	if (MAJOR(dev) != _major || minor >= (1 << MINORBITS))
 		return NULL;
 
-	spin_lock(&_minor_lock);
-
 	md = idr_find(&_minor_idr, minor);
 	if (md && (md == MINOR_ALLOCED ||
 		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
-		   test_bit(DMF_FREEING, &md->flags))) {
-		md = NULL;
-		goto out;
-	}
-
-out:
-	spin_unlock(&_minor_lock);
+		   test_bit(DMF_FREEING, &md->flags)))
+		return NULL;
 
 	return md;
 }
 
 struct mapped_device *dm_get_md(dev_t dev)
 {
-	struct mapped_device *md = dm_find_md(dev);
+	struct mapped_device *md;
 
+	spin_lock(&_minor_lock);
+	md = dm_find_md(dev);
 	if (md)
 		dm_get(md);
+	spin_unlock(&_minor_lock);
 
 	return md;
 }
@@ -2584,11 +2580,17 @@
 	if (&md->kobj != kobj)
 		return NULL;
 
+	spin_lock(&_minor_lock);
+
 	if (test_bit(DMF_FREEING, &md->flags) ||
 	    test_bit(DMF_DELETING, &md->flags))
-		return NULL;
+		md = NULL;
 
-	dm_get(md);
+	if (md)
+		dm_get(md);
+
+	spin_unlock(&_minor_lock);
+
 	return md;
 }
 

  reply	other threads:[~2009-11-09 22:24 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-11-05 13:21 a deadlock bug in the kernel-side device mapper code guy keren
2009-11-05 14:24 ` Alasdair G Kergon
2009-11-06  2:58   ` [PATCH] " Mikulas Patocka
     [not found]     ` <20091106151504.GS13375@agk-dp.fab.redhat.com>
2009-11-06 16:30       ` Mikulas Patocka
2009-11-06 17:27     ` malahal
2009-11-09  8:51     ` Mike Anderson
2009-11-09 17:57       ` malahal
2009-11-09 22:24         ` malahal [this message]
2009-11-10  0:24         ` Alasdair G Kergon
2009-11-10  1:50           ` malahal
2009-11-10  5:24             ` Mike Anderson
2009-11-10  6:13       ` Mikulas Patocka
2009-11-06  0:24 ` Kiyoshi Ueda

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=20091109222425.GA3577@us.ibm.com \
    --to=malahal@us.ibm.com \
    --cc=dm-devel@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