netdev - Re: [PATCH net] net: netem: fix error path for corrupted GSO frames

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20191017114404.4fb502c6@cakuba.netronome.com>
Date:   Thu, 17 Oct 2019 11:44:04 -0700
From:   Jakub Kicinski <jakub.kicinski@...ronome.com>
To:     Cong Wang <xiyou.wangcong@...il.com>
Cc:     David Miller <davem@...emloft.net>,
        Linux Kernel Network Developers <netdev@...r.kernel.org>,
        oss-drivers@...ronome.com,
        Stephen Hemminger <stephen@...workplumber.org>,
        kbuild test robot <lkp@...el.com>,
        Dan Carpenter <dan.carpenter@...cle.com>,
        Ben Hutchings <ben@...adent.org.uk>,
        Simon Horman <simon.horman@...ronome.com>
Subject: Re: [PATCH net] net: netem: fix error path for corrupted GSO frames

On Thu, 17 Oct 2019 11:10:06 -0700, Cong Wang wrote:
> On Wed, Oct 16, 2019 at 4:22 PM Jakub Kicinski wrote:
> > On Wed, 16 Oct 2019 15:42:28 -0700, Cong Wang wrote:  
> > > > @@ -612,7 +613,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
> > > >                         }
> > > >                         segs = skb2;
> > > >                 }
> > > > -               qdisc_tree_reduce_backlog(sch, -nb, prev_len - len);
> > > > +               qdisc_tree_reduce_backlog(sch, !skb - nb, prev_len - len);  
> > >
> > > Am I the only one has trouble to understand the expression
> > > "!skb - nb"?  
> >
> > The backward logic of qdisc_tree_reduce_backlog() always gives me a
> > pause :S  
> 
> Yeah, reducing with a negative value is actually an add. Feel free
> to add a wrapper for this if you think it is better.

I was avoiding adding the wrapper due to stable, but perhaps it should
be okay.

How does this look?

--->8--------------

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 637548d54b3e..66c2dc6a4742 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -630,6 +630,13 @@ void qdisc_reset(struct Qdisc *qdisc);
 void qdisc_put(struct Qdisc *qdisc);
 void qdisc_put_unlocked(struct Qdisc *qdisc);
 void qdisc_tree_reduce_backlog(struct Qdisc *qdisc, int n, int len);
+
+static inline void
+qdisc_tree_increase_backlog(struct Qdisc *qdisc, int n, int len)
+{
+	qdisc_tree_reduce_backlog(qdisc, -n, -len);
+}
+
 #ifdef CONFIG_NET_SCHED
 int qdisc_offload_dump_helper(struct Qdisc *q, enum tc_setup_type type,
 			      void *type_data);
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index 0e44039e729c..e22c13b56bfc 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -509,6 +509,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		if (skb->ip_summed == CHECKSUM_PARTIAL &&
 		    skb_checksum_help(skb)) {
 			qdisc_drop(skb, sch, to_free);
+			skb = NULL;
 			goto finish_segs;
 		}
 
@@ -592,10 +593,13 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 
 finish_segs:
 	if (segs) {
-		unsigned int len, last_len;
+		unsigned int last_len, len = 0;
 		int nb = 0;
 
-		len = skb->len;
+		if (skb) {
+			len += skb->len;
+			nb++;
+		}
 
 		while (segs) {
 			skb2 = segs->next;
@@ -612,7 +616,8 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 			}
 			segs = skb2;
 		}
-		qdisc_tree_reduce_backlog(sch, -nb, prev_len - len);
+		/* Parent qdiscs accounted for 1 skb of size @prev_len */
+		qdisc_tree_increase_backlog(sch, nb - 1, len - prev_len);
 	}
 	return NET_XMIT_SUCCESS;
 }