Re: [PATCH 1/1] net:sched fix a bug about memery leak

[Christoph Paasch <christoph.paasch@uclouvain.be>]

On 24/10/13 - 02:33:38, Eric Dumazet wrote:
> On Thu, 2013-10-24 at 17:12 +0800, Jing Wang wrote:
> > From: Jing Wang <windsdaemon@gmail.com>
> > 
> > the code isn't properly release memory
> > 
> > Signed-off-by: Jing Wang <windsdaemon@gmail.com>
> > ---
> >  net/sched/cls_route.c |    9 ++++++---
> >  1 files changed, 6 insertions(+), 3 deletions(-)
> > 
> > diff --git a/net/sched/cls_route.c b/net/sched/cls_route.c
> > index 37da567..118f8d5 100644
> > --- a/net/sched/cls_route.c
> > +++ b/net/sched/cls_route.c
> > @@ -466,11 +466,11 @@ static int route4_change(struct net *net, struct sk_buff *in_skb,
> >  		goto reinsert;
> >  	}
> >  
> > -	err = -ENOBUFS;
> > +	err = -ENOMEM;
> >  	if (head == NULL) {
> >  		head = kzalloc(sizeof(struct route4_head), GFP_KERNEL);
> >  		if (head == NULL)
> > -			goto errout;
> > +			goto errhead;
> >  
> >  		tcf_tree_lock(tp);
> >  		tp->root = head;
> > @@ -479,7 +479,7 @@ static int route4_change(struct net *net, struct sk_buff *in_skb,
> >  
> >  	f = kzalloc(sizeof(struct route4_filter), GFP_KERNEL);
> >  	if (f == NULL)
> > -		goto errout;
> > +		goto errflt;
> >  
> >  	err = route4_set_parms(net, tp, base, f, handle, head, tb,
> >  		tca[TCA_RATE], 1);
> > @@ -517,6 +517,9 @@ reinsert:
> >  
> >  errout:
> >  	kfree(f);
> > +errflt:
> > +    kfree(head);
> > +errhead:
> >  	return err;
> >  }
> >  
> 
> I don't think this patch is needed or correct.
> 
> tp->root is the head, you cannot free it like that.
> 
> It will be freed properly in route4_destroy()
> 
> Please elaborate, thanks.

I think there is something else wrong in route4_change:

----
>From 1409402bf964bef79667755a5d0d5e0c2bd663f3 Mon Sep 17 00:00:00 2001
From: Christoph Paasch <christoph.paasch@uclouvain.be>
Date: Thu, 24 Oct 2013 15:33:28 +0200
Subject: [PATCH net] net: sched: Don't free f before it is allocated in
 route4_change

f is set to *arg in route4_change at the beginning, which points to a                                                                                                                                                                                                          
route4_filter in the hash-table (gotten through route4_get, called by
tc_ctl_filter). If the alloc of head fails, we should not goto errout,
because this will free f and thus freed memory will be referenced by
the hash-table.
Only later the pointer f will change to an allocated route4_filter.

This patch returns err if the allocation of head fails as f has not yet
been allocated inside route4_change.

Seems the code has been like this since Linus's original git-commit.

Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
---
 net/sched/cls_route.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/sched/cls_route.c b/net/sched/cls_route.c
index 37da567..f17c67f 100644
--- a/net/sched/cls_route.c
+++ b/net/sched/cls_route.c
@@ -470,7 +470,7 @@ static int route4_change(struct net *net, struct sk_buff *in_skb,
        if (head == NULL) {
                head = kzalloc(sizeof(struct route4_head), GFP_KERNEL);
                if (head == NULL)
-                       goto errout;
+                       return err;

                tcf_tree_lock(tp);
                tp->root = head;
-- 
1.8.3.2