Re: åå: INFO: rcu detected stall in tc_modify_qdisc

[Dmitry Vyukov]

On Thu, Jul 30, 2020 at 7:44 PM Vinicius Costa Gomes
<vinicius.gomes@xxxxxxxxx> wrote:
>
> Hi,
>
> Dmitry Vyukov <dvyukov@xxxxxxxxxx> writes:
>
> > On Wed, Jul 29, 2020 at 9:13 PM Vinicius Costa Gomes
> > <vinicius.gomes@xxxxxxxxx> wrote:
> >>
> >> Hi,
> >>
> >> "Zhang, Qiang" <Qiang.Zhang@xxxxxxxxxxxxx> writes:
> >>
> >> > ________________________________________
> >> > åää: linux-kernel-owner@xxxxxxxxxxxxxxx <linux-kernel-owner@xxxxxxxxxxxxxxx> äè syzbot <syzbot+9f78d5c664a8c33f4cce@xxxxxxxxxxxxxxxxxxxxxxxxx>
> >> > åéæé: 2020å7æ29æ 13:53
> >> > æää: davem@xxxxxxxxxxxxx; fweisbec@xxxxxxxxx; jhs@xxxxxxxxxxxx; jiri@xxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; mingo@xxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; syzkaller-bugs@xxxxxxxxxxxxxxxx; tglx@xxxxxxxxxxxxx; vinicius.gomes@xxxxxxxxx; xiyou.wangcong@xxxxxxxxx
> >> > äé: INFO: rcu detected stall in tc_modify_qdisc
> >> >
> >> > Hello,
> >> >
> >> > syzbot found the following issue on:
> >> >
> >> > HEAD commit:    181964e6 fix a braino in cmsghdr_from_user_compat_to_kern()
> >> > git tree:       net
> >> > console output: https://syzkaller.appspot.com/x/log.txt?x=12925e38900000
> >> > kernel config:  https://syzkaller.appspot.com/x/.config?x=f87a5e4232fdb267
> >> > dashboard link: https://syzkaller.appspot.com/bug?extid=9f78d5c664a8c33f4cce
> >> > compiler:       gcc (GCC) 10.1.0-syz 20200507
> >> > syz repro:
> >> > https://syzkaller.appspot.com/x/repro.syz?x=16587f8c900000
> >>
> >> It seems that syzkaller is generating an schedule with too small
> >> intervals (3ns in this case) which causes a hrtimer busy-loop which
> >> starves other kernel threads.
> >>
> >> We could put some limits on the interval when running in software mode,
> >> but I don't like this too much, because we are talking about users with
> >> CAP_NET_ADMIN and they have easier ways to do bad things to the system.
> >
> > Hi Vinicius,
> >
> > Could you explain why you don't like the argument if it's for CAP_NET_ADMIN?
> > Good code should check arguments regardless I think and it's useful to
> > protect root from, say, programming bugs rather than kill the machine
> > on any bug and misconfiguration. What am I missing?
>
> I admit that I am on the fence on that argument: do not let even root
> crash the system (the point that my code is crashing the system gives
> weight to this side) vs. root has great powers, they need to know what
> they are doing.
>
> The argument that I used to convince myself was: root can easily create
> a bunch of processes and give them the highest priority and do
> effectively the same thing as this issue, so I went with a the "they
> need to know what they are doing side".
>
> A bit more on the specifics here:
>
>   - Using a small interval size, is only a limitation of the taprio
>   software mode, when using hardware offloads (which I think most users
>   do), any interval size (supported by the hardware) can be used;
>
>   - Choosing a good lower limit for this seems kind of hard: something
>   below 1us would never work well, I think, but things 1us < x < 100us
>   will depend on the hardware/kernel config/system load, and this is the
>   range includes "useful" values for many systems.
>
> Perhaps a middle ground would be to impose a limit based on the link
> speed, the interval can never be smaller than the time it takes to send
> the minimum ethernet frame (for 1G links this would be ~480ns, should be
> enough to catch most programming mistakes). I am going to add this and
> see how it looks like.
>
> Sorry for the brain dump :-)
>
> >
> > Also are we talking about CAP_NET_ADMIN in a user ns as well
> > (effectively nobody)?
>
> Just checked, we are talking about CAP_NET_ADMIN in user namespace as
> well.

OK, so this is not root/admin, this is just any user.