qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v3 1/2] io/channel: fix crash when qio_channel_readv_all retu


From: Daniel P . Berrangé
Subject: Re: [PATCH v3 1/2] io/channel: fix crash when qio_channel_readv_all return 0
Date: Mon, 11 May 2020 11:02:57 +0100
User-agent: Mutt/1.13.4 (2020-02-15)

On Fri, May 08, 2020 at 08:42:22PM +0800, Li Feng wrote:
> Marc-André Lureau <address@hidden> 于2020年5月8日周五 下午8:32写道:
> >
> > Hi
> >
> > On Fri, May 8, 2020 at 7:14 AM Li Feng <address@hidden> wrote:
> > >
> > > Root cause:
> > > From `man recvmsg`, the RETURN VALUE says:
> > > These  calls return the number of bytes received, or -1 if an error 
> > > occurred.
> > > In the event of an error, errno is set to indicate the error.
> > > The return value will be 0 when the peer has performed an orderly 
> > > shutdown.
> > >
> > > When an error happens, the socket will be closed, and recvmsg return 0,
> > > then error_setg will trigger a crash.
> > >
> > > This unit test could reproduce this issue:
> > > tests/test-char -p /char/socket/client/reconnect-error/unix
> >
> > Current master doesn't trigger the backtrace, it's only after your patch 2.
> Yes. However, the issue did exist in the master code base.
> The test case in patch 2 revealed this issue.
> 
> >
> > >
> > > The core file backtrace is :
> > >
> > > (gdb) bt
> > >     #0  0x00007ffff5ac3277 in raise () from /lib64/libc.so.6
> > >     #1  0x00007ffff5ac4968 in abort () from /lib64/libc.so.6
> > >     #2  0x00005555555aaa94 in error_handle_fatal (errp=<optimized out>, 
> > > err=0x7fffec0012d0) at util/error.c:40
> > >     #3  0x00005555555aab6d in error_setv (errp=0x555555802a08 
> > > <error_abort>, src=0x5555555c4280 "io/channel.c", line=148,
> > >         func=0x5555555c4580 <__func__.17489> "qio_channel_readv_all", 
> > > err_class=ERROR_CLASS_GENERIC_ERROR,
> > >         fmt=<optimized out>, ap=0x7ffff423bae0, suffix=0x0) at 
> > > util/error.c:73
> > >     #4  0x00005555555aacf0 in error_setg_internal 
> > > (errp=errp@entry=0x555555802a08 <error_abort>,
> > >         src=src@entry=0x5555555c4280 "io/channel.c", line=line@entry=148,
> > >         func=func@entry=0x5555555c4580 <__func__.17489> 
> > > "qio_channel_readv_all",
> > >         fmt=fmt@entry=0x5555555c43a0 "Unexpected end-of-file before all 
> > > bytes were read") at util/error.c:97
> > >     #5  0x000055555556c25c in qio_channel_readv_all (ioc=<optimized out>, 
> > > iov=<optimized out>, niov=<optimized out>,
> > >         errp=0x555555802a08 <error_abort>) at io/channel.c:147
> > >     #6  0x000055555556c29a in qio_channel_read_all (ioc=<optimized out>, 
> > > buf=<optimized out>, buflen=<optimized out>,
> > >         errp=<optimized out>) at io/channel.c:247
> > >     #7  0x000055555556ad22 in char_socket_ping_pong (ioc=0x7fffec0008c0) 
> > > at tests/test-char.c:732
> > >     #8  0x000055555556ae12 in char_socket_client_server_thread 
> > > (data=data@entry=0x55555582e350) at tests/test-char.c:891
> > >     #9  0x00005555555a95b6 in qemu_thread_start (args=<optimized out>) at 
> > > util/qemu-thread-posix.c:519
> > >     #10 0x00007ffff5e61e25 in start_thread () from /lib64/libpthread.so.0
> > >     #11 0x00007ffff5b8bbad in clone () from /lib64/libc.so.6
> > >
> > > Signed-off-by: Li Feng <address@hidden>
> > > ---
> > >  io/channel.c | 2 --
> > >  1 file changed, 2 deletions(-)
> > >
> > > diff --git a/io/channel.c b/io/channel.c
> > > index e4376eb0bc..1a4a505f01 100644
> > > --- a/io/channel.c
> > > +++ b/io/channel.c
> > > @@ -144,8 +144,6 @@ int qio_channel_readv_all(QIOChannel *ioc,
> > >
> > >      if (ret == 0) {
> > >          ret = -1;
> > > -        error_setg(errp,
> > > -                   "Unexpected end-of-file before all bytes were read");
> >
> > Nack, this code is fine.
> >
> > The problem is that the test case doesn't expect a disconnect in
> > char_socket_ping_pong().
> Yes, in the test case I try to inject an error in char_socket_ping_pong.
> Any concerns about these two patches?

I agree with Marc-André - this patch is wrong. qio_channel_readv_all
*MUST* always set 'errp' if the return value is -1. To not set 'errp'
is violating the calling convention.

The bug here is in your test case - it is passing '&error_abort' to the
'qio_channel_readv_all' call.  If you don't want the test to crash, then
don't pass &error_abort

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|




reply via email to

[Prev in Thread] Current Thread [Next in Thread]