[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAM0EoM==XCDWdj7n955uRkeJeyLrOgW_0k+d1qpPVjgz4ax4KQ@mail.gmail.com>
Date: Fri, 21 Jul 2023 15:55:14 -0400
From: Jamal Hadi Salim <jhs@...atatu.com>
To: M A Ramdhan <ramdhan@...rlabs.sg>
Cc: valis <sec@...is.email>, netdev@...r.kernel.org, xiyou.wangcong@...il.com,
jiri@...nulli.us, davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org,
pabeni@...hat.com, pctammela@...atatu.com, victor@...atatu.com,
billy@...rlabs.sg
Subject: Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf_result on
update to avoid use-after-free
On Fri, Jul 21, 2023 at 2:05 PM M A Ramdhan <ramdhan@...rlabs.sg> wrote:
>
> On Sat, Jul 22, 2023 at 12:51 AM valis <sec@...is.email> wrote:
> >
> > When u32_change() is called on an existing filter, the whole
> > tcf_result struct is always copied into the new instance of the filter.
> >
> > This causes a problem when updating a filter bound to a class,
> > as tcf_unbind_filter() is always called on the old instance in the
> > success path, decreasing filter_cnt of the still referenced class
> > and allowing it to be deleted, leading to a use-after-free.
> >
> > Fix this by no longer copying the tcf_result struct from the old filter.
> >
> > Fixes: de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
> > Reported-by: valis <sec@...is.email>
> > Reported-by: M A Ramdhan <ramdhan@...rlabs.sg>
> > Signed-off-by: valis <sec@...is.email>
> > Cc: stable@...r.kernel.org
> > ---
> > net/sched/cls_u32.c | 1 -
> > 1 file changed, 1 deletion(-)
> >
> > diff --git a/net/sched/cls_u32.c b/net/sched/cls_u32.c
> > index 5abf31e432ca..19aa60d1eea7 100644
> > --- a/net/sched/cls_u32.c
> > +++ b/net/sched/cls_u32.c
> > @@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
> >
> > new->ifindex = n->ifindex;
> > new->fshift = n->fshift;
> > - new->res = n->res;
> > new->flags = n->flags;
> > RCU_INIT_POINTER(new->ht_down, ht);
> >
> > --
> > 2.30.2
> >
> Hi,
>
> We also thought it's also the correct fixes,
> but we're not sure because it will always remove the already bound
> qdisc class when we change the filter, even tho we never specify
> the new TCA_U32_CLASSID in the new filter.
I am assuming you are referring to the u32 classifier here and from
what you are describing you first create a filter with reference to an
existing class and then you replace it to not reference the class - am
i correct? If yes, then please provide an example policy setup? We
tested this scenario on u32 extensively and it should be fine.
cheers,
jamal
> I also look at the implementation of cls_tcindex and cls_rsvp which still copy
> the old tcf_result, but don't call the tcf_unbind_filter when changing
> the filter.
>
> If it's the intended behaviour, then I'm good with this patch.
>
> Thanks & Regards,
> M A Ramdhan
Powered by blists - more mailing lists