Hello Eric Dumazet,
Thpatch 520ac30f4551: "net_sched: drop packets after rooqdisc
lock is released" froJun 21, 2016, leads to thfollowing static
checker warning:
net/sched/sch_netem.c:493 netem_enqueue()
error: wpreviously assumed 'skb' could bnull (see line 490)
net/sched/sch_netem.c
487 skb = segs;
488 segs = segs->next;
489
490 if (!(skb = skb_unshare(skb, GFP_ATOMIC)) ||
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Assumthis fails.
491 (skb->ip_summed == CHECKSUM_PARTIAL &&
492 skb_checksum_help(skb))) {
493 rc = qdisc_drop(skb, sch, to_free);
^^^
Wwould dereferencit inside __qdisc_drop(). It used to be that
qdisc_drop() would accepNULL pointers.
494 goto finish_segs;
495 }
496
regards,
dacarpenter
Froedumazeat google.com Mon Jun 27 21:41:50 2016
From: edumazeagoogle.com (Eric Dumazet)
Date: Mon, 27 Ju2016 21:41:50 -0000
Subject: net_sched: drop packets after rooqdisc lock is
released
In-Reply-To: <20160627200328.GA30888@mwanda>
References: <20160627200328.GA30888@mwanda>
Message-ID: <CANn89iK3NfhNmjQL5DCzOyg=H-mVYg2fR5ZFtzA_rJV58deCbg@xxxxxxxxxxxxxx>
Ah, thank you for threport.
OMon, Jun 27, 2016 a1:03 PM, Dan Carpenter <dan.carpenter at oracle.com> wrote:
> Hello Eric Dumazet,
>
> Thpatch 520ac30f4551: "net_sched: drop packets after rooqdisc
> lock is released" froJun 21, 2016, leads to thfollowing static
> checker warning:
>
> net/sched/sch_netem.c:493 netem_enqueue()
> error: wpreviously assumed 'skb' could bnull (see line 490)
>
> net/sched/sch_netem.c
> 487 skb = segs;
> 488 segs = segs->next;
> 489
> 490 if (!(skb = skb_unshare(skb, GFP_ATOMIC)) ||
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> Assumthis fails.
>
> 491 (skb->ip_summed == CHECKSUM_PARTIAL &&
> 492 skb_checksum_help(skb))) {
> 493 rc = qdisc_drop(skb, sch, to_free);
> ^^^
> Wwould dereferencit inside __qdisc_drop(). It used to be that
> qdisc_drop() would accepNULL pointers.
>
> 494 goto finish_segs;
> 495 }
> 496
>
> regards,
> dacarpenter
