Re: [Qemu-devel] qemu segfauls with spiceport chardev and isa-serial

[Peter Crosthwaite]

On Tue, Feb 4, 2014 at 4:45 AM, Dr. David Alan Gilbert
<address@hidden> wrote:
> (cc'ing in Peter Crosthwaite and Michael Tokarev due to a serial fifo change
> - see below!)
>
> * Martin Kletzander (address@hidden) wrote:
>> Hello,
>
> Hi Martin,
>    I don't know about your spice warnings that triggered this but looking
> down the backtrace I can see something odd:
>
>> current HEAD (2f61120c10da9128357510debc8e66880cd2bfdc) segfaults when
>> I'm trying to do the following:
>>
>> I add this to qemu's command-line:
>>
>>  -chardev spiceport,id=charserial0,name=org.qemu.console.serial.0 \
>>  -device isa-serial,chardev=charserial0,id=serial0
>>
>> and then use spicy to connect to that machine.  That spits out the
>> following error:
>>
>>  GSpice-Message: main channel: opened
>>  port 0x7f74182366e0 org.qemu.console.serial.0: opened
>>
>>  (spicy:32386): GSpice-WARNING **: incomplete link header (-104/16)
>>
>>  (spicy:32386): GSpice-WARNING **: incomplete link header (-104/16)
>>  GSpice-Message: main channel: closed
>>
>> I can see that the console works when the window flashes, so there was
>> some communication done (Im running the kernel inside with
>> "console=tty0 console=ttyS0,115200n8" as suggested here:
>>
>> http://lists.freedesktop.org/archives/spice-devel/2014-January/015919.html
>>
>> The full command-line with backtrace of all the threads (with
>> abort()-ing thread being thread #1 follows.  Let me know if I can help
>> anyhow.
>>
>> Thanks,
>> Martin
>>
>> Command-line:
>>
>> qemu-system-x86_64 -name rhel7 -S -machine \
>> pc-i440fx-1.7,accel=kvm,usb=off,dump-guest-core=off -cpu SandyBridge \
>> -m 4101 -realtime mlock=off -smp 1,sockets=1,cores=1,threads=1 -uuid \
>> f49fa544-f21d-4267-8958-d82570644f39 -no-user-config -nodefaults \
>> -chardev \
>> socket,id=charmonitor,path=/var/lib/libvirt/qemu/rhel7.monitor,server,nowait 
>> \
>> -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc \
>> -no-shutdown -boot strict=on -device \
>> piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -device \
>> virtio-serial-pci,id=virtio-serial0,bus=pci.0,addr=0x6 -drive \
>> if=none,id=drive-ide0-0-0,readonly=on,format=raw -device \
>> ide-cd,bus=ide.0,unit=0,drive=drive-ide0-0-0,id=ide0-0-0 -drive \
>> file=/home/nert/.config/libvirt/images/rhel7.img,if=none,id=drive-virtio-disk0,format=qcow2
>>  \
>> -device \
>> virtio-blk-pci,scsi=off,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1
>>  \
>> -netdev tap,fd=20,id=hostnet0,vhost=on,vhostfd=21 -device \
>> virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:42:be:45,bus=pci.0,addr=0x3
>>  \
>> -chardev spiceport,id=charserial0,name=org.qemu.console.serial.0 \
>> -device isa-serial,chardev=charserial0,id=serial0 -device \
>> usb-tablet,id=input0 -vnc 127.0.0.1:0 -spice \
>> port=5901,tls-port=5902,addr=127.0.0.1,disable-ticketing,x509-dir=/etc/pki/libvirt-spice,seamless-migration=on
>>  \
>> -device \
>> qxl-vga,id=video0,ram_size=67108864,vram_size=67108864,bus=pci.0,addr=0x2 \
>> -device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5
>>
>> Backtrace:
>>
>
> <snipped boring threads in poll>
>
>> Thread 1 (Thread 0x7fee3da66980 (LWP 32022)):
>> #0  0x00007fee344f1f4e in __GI_raise (address@hidden) at 
>> ../nptl/sysdeps/unix/sysv/linux/raise.c:56
>> #1  0x00007fee344f369f in __GI_abort () at abort.c:89
>> #2  0x00007fee3de72baa in fifo8_pop (address@hidden) at util/fifo8.c:45
>
> fifo8_pop is aborting because the fifo is empty:
>     if (fifo->num == 0) {
>         abort();
>     }
>
> which seems fair enough
>
>> #3  0x00007fee3dc0c110 in serial_xmit (chan=<optimized out>, cond=<optimized 
>> out>, opaque=0x7fee3fc286a0)
>>     at hw/char/serial.c:228
>
>             s->tsr = fifo8_is_full(&s->xmit_fifo) ?
>                         0 : fifo8_pop(&s->xmit_fifo);
>
> Hmm, now I don't know anything about the tsr stuff; but that calls
> fifo8_pop whenever the fifo isn't *full* - i.e. it still gets called if empty.
>
> I think the change here comes from Peter's 8e8638fa87ff04 'char/serial: Use 
> generic Fifo8'
> changeset from June which did:
>
> -            s->tsr = fifo_get(s,XMIT_FIFO);
> -            if (!s->xmit_fifo.count) {
> +            s->tsr = fifo8_is_full(&s->xmit_fifo) ?
> +                        0 : fifo8_pop(&s->xmit_fifo);
> +            if (!s->xmit_fifo.num) {
>
> which makes me think (without having looked at the old data structure
> properly) if that should be   fifo8_is_empty ?
> (The old serial fifo_get routine returned 0 if empty rather than aborting).
>

Hi Dave,

Yes that does looks suss. My bad. Can you confirm your theory by
making the proposed change? does it fix the bug?

--- a/hw/char/serial.c
+++ b/hw/char/serial.c
@@ -225,7 +225,7 @@ static gboolean serial_xmit(GIOChannel *chan,
GIOCondition cond, void

     if (s->tsr_retry <= 0) {
         if (s->fcr & UART_FCR_FE) {
-            s->tsr = fifo8_is_full(&s->xmit_fifo) ?
+            s->tsr = fifo8_is_empty(&s->xmit_fifo) ?
                         0 : fifo8_pop(&s->xmit_fifo);
             if (!s->xmit_fifo.num) {
                 s->lsr |= UART_LSR_THRE;

Regards,
Peter

> Dave
>
>> #4  0x00007fee3d1a8957 in g_main_dispatch (context=0x7fee3fa49470)
>>     at 
>> /var/tmp/portage/dev-libs/glib-2.38.2/work/glib-2.38.2/glib/gmain.c:3066
>> #5  g_main_context_dispatch (address@hidden)
>>     at 
>> /var/tmp/portage/dev-libs/glib-2.38.2/work/glib-2.38.2/glib/gmain.c:3642
>> #6  0x00007fee3dccdde7 in glib_pollfds_poll () at main-loop.c:189
>> #7  os_host_main_loop_wait (timeout=<optimized out>) at main-loop.c:234
>> #8  main_loop_wait (nonblocking=<optimized out>) at main-loop.c:483
>> #9  0x00007fee3db61501 in main_loop () at vl.c:2018
>> #10 main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) 
>> at vl.c:4410
>
>
> --
> Dr. David Alan Gilbert / address@hidden / Manchester, UK
>