netfilter-devel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Patrick McHardy <kaber@trash.net>
To: mirq-linux@rere.qmqm.pl
Cc: netfilter-devel@lists.netfilter.org
Subject: Re: [PATCH 2.6.20 07/10] nfnetlink_log: fix module reference	counting
Date: Wed, 14 Feb 2007 17:46:38 +0100	[thread overview]
Message-ID: <45D33CEE.6050404@trash.net> (raw)
In-Reply-To: <20070213162535.GA10544@rere.qmqm.pl>

[-- Attachment #1: Type: text/plain, Size: 1184 bytes --]

Micha³ Miros³aw wrote:
> On Tue, Feb 13, 2007 at 01:48:03PM +0100, Patrick McHardy wrote:
> 
>>I think we should just cancel the timer on destruction.
> 
> 
> It won't solve a race between destroying the instance and logging a
> packet. It could happen that:
> 
> CPU1				CPU2
> 				nfulnl_log_packet()
> 				 -> instance_lookup_get()
> 				 -> lock inst
> 				 ** got lock on inst
> _instance_destroy(inst)
>  -> remove inst from list
>  -> remove timer and unref inst
>  -> lock inst
>  				 -> start a timer
> 				 -> unlock inst
>  ** got lock on inst
>  -> clear the skb
>  -> unlock inst
>  -> unref inst
>  -> unref module... but we still have a timer pending!


This is easily fixable by adding a synchronize_rcu() call
after removing the instance from the global list.
nfulnl_log_packet() is called within a RCU read-side
critical section, so once synchronize_rcu() returns we're
guaranteed no CPU is within nfulnl_log_packet anymore with
this instance. And I think it is really preferable to having
the timer armed after destroying the instance.

This patch should take care of both the module reference
problem and the refcount leak.


[-- Attachment #2: x --]
[-- Type: text/plain, Size: 833 bytes --]

diff --git a/net/netfilter/nfnetlink_log.c b/net/netfilter/nfnetlink_log.c
index 8a460b1..1d45ff6 100644
--- a/net/netfilter/nfnetlink_log.c
+++ b/net/netfilter/nfnetlink_log.c
@@ -213,6 +213,9 @@ _instance_destroy2(struct nfulnl_instanc
 	if (lock)
 		write_unlock_bh(&instances_lock);
 
+	sychronize_rcu();
+	if (del_timer_sync(&inst->timer))
+		instance_put(inst);
 	/* then flush all pending packets from skb */
 
 	spin_lock_bh(&inst->lock);
@@ -363,9 +366,6 @@ __nfulnl_send(struct nfulnl_instance *in
 {
 	int status;
 
-	if (timer_pending(&inst->timer))
-		del_timer(&inst->timer);
-
 	if (!inst->skb)
 		return 0;
 
@@ -676,6 +676,9 @@ #endif
 		 * enough room in the skb left. flush to userspace. */
 		UDEBUG("flushing old skb\n");
 
+		if (del_timer(&inst->timer))
+			instance_put(inst);
+
 		__nfulnl_send(inst);
 	}
 

  reply	other threads:[~2007-02-14 16:46 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-02-12  0:39 [PATCH 2.6.20 07/10] nfnetlink_log: fix module reference counting Michał Mirosław
2007-02-13 12:48 ` Patrick McHardy
2007-02-13 16:25   ` Michał Mirosław
2007-02-14 16:46     ` Patrick McHardy [this message]
2007-02-14 17:07       ` Michał Mirosław
2007-02-14 17:22         ` Patrick McHardy

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=45D33CEE.6050404@trash.net \
    --to=kaber@trash.net \
    --cc=mirq-linux@rere.qmqm.pl \
    --cc=netfilter-devel@lists.netfilter.org \
    /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).