Re: INFO: task hung in filemap_fault
From: Tetsuo Handa
Date: Mon Jan 15 2018 - 05:44:54 EST
Dmitry Vyukov wrote:
> On Mon, Jan 8, 2018 at 11:48 AM, Tetsuo Handa
> <penguin-kernel@xxxxxxxxxxxxxxxxxxx> wrote:
> > Dmitry Vyukov wrote:
> >> >> Hi Tetsuo,
> >> >>
> >> >> syzbot always re-runs the same workload on a new machine. If it
> >> >> manages to reproduce the problem, it provides a reproducer. In this
> >> >> case it didn't.
> >> >
> >> > Even if it did not manage to reproduce the problem, showing raw.log in
> >> > C format is helpful for me. For example,
> >> >
> >> > ioctl$LOOP_CHANGE_FD(r3, 0x4c00, r1)
> >> >
> >> > is confusing. 0x4c00 is not LOOP_CHANGE_FD but LOOP_SET_FD.
> >> > If the message were
> >> >
> >> > ioctl(r3, 0x4c00, r1)
> >> >
> >> > more people will be able to read what the program tried to do.
> >> > There are many operations done on loop devices, but are too hard
> >> > for me to pick up only loop related actions.
> >>
> >>
> >> Hi Tetsuo,
> >>
> >> The main purpose of this format is different, this is a complete
> >> representation of programs that allows replaying them using syzkaller
> >> tools.
> >
> > What is ioctl$LOOP_CHANGE_FD(r3, 0x4c00, r1) ?
> > 0x4c00 is LOOP_SET_FD. Why LOOP_CHANGE_FD is there?
>
>
> In short, it specifies exact discrimination of the syscall which
> affects parsing of the rest of the arguments. For some syscalls
> (ioctl/setsockopt/sendmsg) kernel has hundreds of different
> discriminations with radically different arguments.
> Now if you are asking why the discrimination is LOOP_CHANGE_FD, but
> the actual command is LOOP_SET_FD, that's because this is a fuzzer,
> it's sole purpose is to mess things in unexpected ways.
??? I can't catch what you want to say.
I understand that a fuzzer intentionally tests various cases.
My question is simple. Why don't you use actual command name like
ioctl$LOOP_SET_FD(r3, 0x4c00, r1) ?
Writing like ioctl$LOOP_CHANGE_FD is confusing. I consider it as a bug.
>
>
> >> We can't simply drop info from there. Do you propose to add
> >> another attached file that contains the same info in a different
> >> format? What is the exact format you are proposing?
> >
> > Plain C program which can be compiled without installing additional
> > program/library packages (except those needed for building kernels).
> >
> >> Is it just
> >> dropping the syscall name part after $ sign? Note that it's still not
> >> C, more complex syscall generally look as follows:
> >>
> >> perf_event_open(&(0x7f0000b5a000)={0x4000000002, 0x78, 0x1e2, 0x0,
> >> 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff, 0x0,
> >> 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
> >> 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
> >> @perf_bp={&(0x7f0000000000)=0x0, 0x0}, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
> >> 0x0, 0x0}, 0x0, 0x0, 0xffffffffffffffff, 0x0)
> >> recvmmsg(0xffffffffffffffff, &(0x7f0000003000)=[{{0x0, 0x0,
> >> &(0x7f0000002000)=[{&(0x7f000000a000)=""/193, 0xc1},
> >> {&(0x7f0000007000-0x3d)=""/61, 0x3d}], 0x2,
> >> &(0x7f0000005000-0x67)=""/103, 0x67, 0x0}, 0x0}], 0x1, 0x0,
> >> &(0x7f0000003000-0x10)={0x77359400, 0x0})
> >> bpf$PROG_LOAD(0x5, &(0x7f0000000000)={0x1, 0x5,
> >> &(0x7f0000002000)=@framed={{0x18, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
> >> 0x0}, [@jmp={0x5, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}], {0x95, 0x0, 0x0,
> >> 0x0}}, &(0x7f0000004000-0xa)='syzkaller\x00', 0x3, 0xc3,
> >> &(0x7f0000386000)=""/195, 0x0, 0x0, [0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
> >> 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0], 0x0}, 0x48)
> >>
> >> Note: you can convert any syzkaller program to equivalent C code using
> >> syz-prog2c utility that comes with syzkaller.
> >
> > I won't install go language into my environment for analyzing/reproducing your
> > reports. If syz-prog2c is provided as a CGI service (e.g. receive URL containing
> > raw.log and print the converted C program), I might try it.
>
>
> raw.log is not a _program_, it's hundreds of separate programs that
> were executed before the crash. It's also very compressed
> representation as compared to equivalent C programs. For example for
> this program:
>
> mmap(&(0x7f0000000000/0xfff000)=nil, 0xfff000, 0x3, 0x32,
> 0xffffffffffffffff, 0x0)
> r0 = socket$nl_generic(0x10, 0x3, 0x10)
> sendmsg$nl_generic(r0,
> &(0x7f0000b3e000-0x38)={&(0x7f0000d4a000-0xc)={0x10, 0x0, 0x0, 0x0},
> 0xc, &(0x7f0000007000)={&(0x7f0000f7c000-0x15c)={0x24, 0x1c, 0x109,
> 0xffffffffffffffff, 0xffffffffffffffff, {0x4, 0x0, 0x0},
> [@nested={0x10, 0x9, [@typed={0xc, 0x0, @u32=0x0}]}]}, 0x24}, 0x1,
> 0x0, 0x0, 0x0}, 0x0)
>
> you can get up to this amount of C code:
> https://gist.githubusercontent.com/dvyukov/eeaeb4e4ac45c3a251f72098c9295bf9/raw/700cd583507eca90711ba11b42e406f317553371/gistfile1.txt
>
> that is, 700 lines of C source for 3 line program. So instead of a 1MB
> file that will be 100MB, and then it probably should be a gzip archive
> with hundreds of separate C files. There are people on this list
> complaining even about 200K of attachments. I don't see that this will
> be better and well accepted.
>
No problem. In the "tty: User triggerable soft lockup." case, I manually
trimmed the reproducer at https://marc.info/?l=linux-mm&m=151368630414963 .
That is,
(1) Can the problem be reproduced even if setup_tun(0, true); is commented out?
(2) Can the problem be reproduced even if NONFAILING(A = B); is replaced with
plain A = B; assignment?
(3) Can the problem be reproduced even if install_segv_handler(); is commented
out?
(4) Can the problem be reproduced even if some syscalls (e.g. __NR_memfd_create,
__NR_getsockopt, __NR_perf_event_open) are replaced with no-op?
and so on. Then, I finally reached a reproducer which I sent, and the bug was fixed.
What is important is that everyone can try simplifying the reproducer written
in plain C in order to narrow down the culprit. Providing a (e.g.) CGI service
which generates plain C reproducer like gistfile1.txt will be helpful to me.