linux-kernel - Re: [REGRESSION] 6.14.3 panic - kernel NULL pointer dereference in htb

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: <aAlAakEUu4XSEdXF@pop-os.localdomain>
Date: Wed, 23 Apr 2025 12:32:58 -0700
From: Cong Wang <xiyou.wangcong@...il.com>
To: "Alan J. Wylie" <alan@...ie.me.uk>
Cc: Holger Hoffstätte <holger@...lied-asynchrony.com>,
	Jamal Hadi Salim <jhs@...atatu.com>, regressions@...ts.linux.dev,
	Jiri Pirko <jiri@...nulli.us>, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org, Octavian Purdila <tavip@...gle.com>,
	Toke Høiland-Jørgensen <toke@...hat.com>,
	stable@...r.kernel.org, Greg KH <gregkh@...uxfoundation.org>
Subject: Re: [REGRESSION] 6.14.3 panic - kernel NULL pointer dereference in
 htb_dequeue

On Wed, Apr 23, 2025 at 10:51:49AM +0100, Alan J. Wylie wrote:
> On Tue, 22 Apr 2025 14:49:27 -0700
> Cong Wang <xiyou.wangcong@...il.com> wrote:
> 
> > Although I am still trying to understand the NULL pointer, which seems
> > likely from:
> > 
> >  478                         if (p->inner.clprio[prio].ptr == cl->node + prio) {
> >  479                                 /* we are removing child which is pointed to from
> >  480                                  * parent feed - forget the pointer but remember
> >  481                                  * classid
> >  482                                  */
> >  483                                 p->inner.clprio[prio].last_ptr_id = cl->common.classid;
> >  484                                 p->inner.clprio[prio].ptr = NULL;
> >  485                         }
> > 
> > Does the following patch work? I mean not just fixing the crash, but
> > also not causing any other problem.
> 
> > ---
> > 
> > diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
> > index 4b9a639b642e..0cdc778fddef 100644
> > --- a/net/sched/sch_htb.c
> > +++ b/net/sched/sch_htb.c
> > @@ -348,7 +348,8 @@ static void htb_add_to_wait_tree(struct htb_sched *q,
> >   */
> >  static inline void htb_next_rb_node(struct rb_node **n)
> >  {
> > -	*n = rb_next(*n);
> > +	if (*n)
> > +		*n = rb_next(*n);
> >  }
> >  
> >  /**
> 
> There's been three of these: 
> 
> Apr 23 08:08:32 bilbo kernel: WARNING: CPU: 0 PID: 0 at htb_deactivate+0xd/0x30 [sch_htb]
> Apr 23 08:08:32 bilbo kernel: WARNING: CPU: 0 PID: 0 at htb_deactivate+0xd/0x30 [sch_htb]
> Apr 23 10:41:36 bilbo kernel: WARNING: CPU: 1 PID: 0 at htb_deactivate+0xd/0x30 [sch_htb]
> 
> But no panic.
> 
> I've run scripts/decode.sh on the last one.
> 

Thanks a lot for testing! This helped a lot to verify how far we can go
beyond the panic and what I still missed. To me it looks a bit
complicated for -stable if we make everything idempotent along the path.

Do you mind testing the following one instead? Please revert the
above one for htb_next_rb_node(). I think maybe this is the safest fix
we could have for -stable.

Thanks!

--------->

diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index 4b9a639b642e..3786abbdc4c3 100644
--- a/net/sched/sch_htb.c
+++ b/net/sched/sch_htb.c
@@ -1487,7 +1487,8 @@ static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg)
 
 	if (!cl->prio_activity)
 		return;
-	htb_deactivate(qdisc_priv(sch), cl);
+	if (!cl->leaf.q->q.qlen)
+		htb_deactivate(qdisc_priv(sch), cl);
 }
 
 static inline int htb_parent_last_child(struct htb_class *cl)