Re: [Qemu-devel] [PULL 1/4] curl: do not use aio_context_acquire/release

[Richard W.M. Jones]

On Wed, May 03, 2017 at 04:59:22PM +0200, Paolo Bonzini wrote:
> Maybe there is, can you grab an all-threads backtrace via gdb?

Program received signal SIGINT, Interrupt.
0x00007f1d57f861bd in __lll_lock_wait () from /lib64/libpthread.so.0
(gdb) t a a bt

Thread 9 (Thread 0x7f1d49967700 (LWP 29949)):
#0  0x00007f1d53a39bf9 in syscall () at /lib64/libc.so.6
#1  0x000055ea5c358856 in qemu_event_wait (val=<optimized out>, f=<optimized 
out>) at /home/rjones/d/qemu/include/qemu/futex.h:26
#2  0x000055ea5c358856 in qemu_event_wait (address@hidden 
<rcu_call_ready_event>) at util/qemu-thread-posix.c:399
#3  0x000055ea5c3680fe in call_rcu_thread (opaque=<optimized out>)
    at util/rcu.c:249
#4  0x00007f1d57f7fdc5 in start_thread () at /lib64/libpthread.so.0
#5  0x00007f1d53a3f73d in clone () at /lib64/libc.so.6

Thread 7 (Thread 0x7f1d46987700 (LWP 29955)):
#0  0x00007f1d57f861bd in __lll_lock_wait () at /lib64/libpthread.so.0
#1  0x00007f1d57f81d02 in _L_lock_791 () at /lib64/libpthread.so.0
#2  0x00007f1d57f81c08 in pthread_mutex_lock () at /lib64/libpthread.so.0
#3  0x000055ea5c3583e9 in qemu_mutex_lock (address@hidden <qemu_global_mutex>) 
at util/qemu-thread-posix.c:60
#4  0x000055ea5bff512c in qemu_mutex_lock_iothread ()
    at /home/rjones/d/qemu/cpus.c:1565
#5  0x000055ea5c0081c7 in kvm_cpu_exec (address@hidden)
    at /home/rjones/d/qemu/kvm-all.c:2096
#6  0x000055ea5bff5332 in qemu_kvm_cpu_thread_fn (arg=0x55ea5e7df020)
    at /home/rjones/d/qemu/cpus.c:1118
#7  0x00007f1d57f7fdc5 in start_thread () at /lib64/libpthread.so.0
#8  0x00007f1d53a3f73d in clone () at /lib64/libc.so.6

Thread 1 (Thread 0x7f1d5a4e2c00 (LWP 29947)):
#0  0x00007f1d57f861bd in __lll_lock_wait () at /lib64/libpthread.so.0
#1  0x00007f1d57f81d02 in _L_lock_791 () at /lib64/libpthread.so.0
#2  0x00007f1d57f81c08 in pthread_mutex_lock () at /lib64/libpthread.so.0
#3  0x000055ea5c3583e9 in qemu_mutex_lock (address@hidden)
    at util/qemu-thread-posix.c:60
#4  0x000055ea5c2fb2d1 in curl_readv_bh_cb (p=0x55ea6013c620)
    at block/curl.c:824
#5  0x000055ea5c352fe1 in aio_bh_poll (bh=0x55ea60eb99d0) at util/async.c:90
#6  0x000055ea5c352fe1 in aio_bh_poll (address@hidden)
    at util/async.c:118
#7  0x000055ea5c3563c4 in aio_poll (ctx=0x55ea5e555970, address@hidden) at 
util/aio-posix.c:682
#8  0x000055ea5c2fb09e in curl_init_state (bs=0x55ea5e5a56b0, address@hidden) 
at block/curl.c:470
#9  0x000055ea5c2fb39b in curl_readv_bh_cb (p=0x55ea5f650c30)
    at block/curl.c:839
#10 0x000055ea5c352fe1 in aio_bh_poll (bh=0x55ea60eb1910) at util/async.c:90
#11 0x000055ea5c352fe1 in aio_bh_poll (address@hidden)
    at util/async.c:118
#12 0x000055ea5c355f80 in aio_dispatch (ctx=0x55ea5e555970)
    at util/aio-posix.c:429
#13 0x000055ea5c352ebe in aio_ctx_dispatch (source=<optimized out>, 
callback=<optimized out>, user_data=<optimized out>) at util/async.c:261
#14 0x00007f1d55489d7a in g_main_context_dispatch () at /lib64/libglib-2.0.so.0
#15 0x000055ea5c35523c in main_loop_wait () at util/main-loop.c:213
#16 0x000055ea5c35523c in main_loop_wait (timeout=<optimized out>)
    at util/main-loop.c:261
---Type <return> to continue, or q <return> to quit---
#17 0x000055ea5c35523c in main_loop_wait (address@hidden)
    at util/main-loop.c:517
#18 0x000055ea5bfb44a7 in main () at vl.c:1899
#19 0x000055ea5bfb44a7 in main (argc=<optimized out>, argv=<optimized out>, 
envp=<optimized out>) at vl.c:4717


I'll get you one with more symbols in a minute, just installing
debuginfo ...

BTW this bug is reproducible on both Fedora 26 and RHEL 7.4
(in both cases with either qemu 2.9 or qemu from git).

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
Fedora Windows cross-compiler. Compile Windows programs, test, and
build Windows installers. Over 100 libraries supported.
http://fedoraproject.org/wiki/MinGW