Re: WARNING in bpf_bprintf_prepare

[Jiri Olsa <olsajiri@gmail.com>]

On Thu, Oct 27, 2022 at 07:45:16PM +0800, Hao Sun wrote:
> Jiri Olsa <olsajiri@gmail.com> 于2022年10月27日周四 19:24写道：
> >
> > On Thu, Oct 27, 2022 at 10:27:28AM +0800, Hao Sun wrote:
> > > Hi,
> > >
> > > The following warning can be triggered with the C reproducer in the link.
> > > Syzbot also reported this several days ago, Jiri posted a patch that
> > > uses bpf prog `active` field to fix this by 05b24ff9b2cfab (bpf:
> > > Prevent bpf program recursion...) according to syzbot dashboard
> > > (https://syzkaller.appspot.com/bug?id=179313fb375161d50a98311a28b8e2fc5f7350f9).
> > > But this warning can still be triggered on 247f34f7b803
> > > (Linux-v6.1-rc2) that already merged the patch, so it seems that this
> > > still is an issue.
> > >
> > > HEAD commit: 247f34f7b803 Linux 6.1-rc2
> > > git tree: upstream
> > > console output: https://pastebin.com/raw/kNw8JCu5
> > > kernel config: https://pastebin.com/raw/sE5QK5HL
> > > C reproducer: https://pastebin.com/raw/X96ASi27
> >
> > hi,
> > right, that fix addressed that issue for single bpf program,
> > and it won't prevent if there are multiple programs hook on
> > contention_begin tracepoint and calling bpf_trace_printk,
> >
> > I'm not sure we can do something there.. will check
> >
> > do you run just the reproducer, or you load the server somehow?
> > I cannot hit the issue so far
> >
> 
> Hi,
> 
> Last email has format issues, resend it here.
> 
> I built the kernel with the config in the link, which contains
> “CONFIG_CMDLINE="earlyprintk=serial net.ifnames=0
> sysctl.kernel.hung_task_all_cpu_backtrace=1 panic_on_warn=1 …”, and
> boot the kernel with normal qemu setup and then the warning can be
> triggered by executing the reproducer.
> 
> Also, I’m willing to test the proposed patch if any.

fyi I reproduced that.. will check if we can do anything about that

thanks,
jirka