** Changed in: spice-xpi
Status: Unknown => Fix Released
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1964777
Title:
qxl_send_events: spice-server bug: guest stopped, ignoring
To manage not
This bug was fixed in the package spice - 0.15.0-2ubuntu4
---
spice (0.15.0-2ubuntu4) jammy; urgency=medium
* d/p/Revert-reds-start-QXL-devices-if-VM-is-running-fix-r.patch: fix race
on spice init (LP: #1964777)
-- Christian Ehrhardt Mon, 21 Mar
2022 09:23:33 +0100
** Chang
MP approved, tests good, upstream ok with the change (some CI issues,
but nothing really stopping it) => uploaded the fix for Jammy.
** Changed in: spice (Ubuntu)
Assignee: Lena Voytek (lvoytek) => Christian Ehrhardt (paelzer)
--
You received this bug notification because you are a member
I did some more case analysis on this - this is complex and I might be
wrong - but for now it seems to me that a revert of the offending commit
is the best short term option. I've updated the upstream issue and
submitted a PR:
- https://gitlab.freedesktop.org/spice/spice/-/issues/64
- https://gitl
** Merge proposal linked:
https://code.launchpad.net/~paelzer/ubuntu/+source/spice/+git/spice/+merge/417153
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1964777
Title:
qxl_send_events: spice-se
** Tags removed: server-triage-discuss
** Changed in: spice (Ubuntu)
Assignee: (unassigned) => Lena Voytek (lvoytek)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1964777
Title:
qxl_send_event
In terms of annoyance this is "Critical", but since it correctly
initializes just a bit later and then works fine I can't consider it
worse than "Medium" at the moment.
** Tags added: server-todo server-triage-discuss
** Changed in: spice (Ubuntu)
Importance: Undecided => Medium
--
You recei
FYI I chimed in on
https://gitlab.freedesktop.org/spice/spice/-/issues/64 outlining what
I've found. I'll check back later if there is an easy conclusion and
follow up now that the people that know more about this code see what
happens.
--
You received this bug notification because you are a memb
In fact - as always - when you know what exactly you search for you find it.
Related are:
- https://gitlab.freedesktop.org/spice/spice/-/issues/64
- https://bugzilla.redhat.com/show_bug.cgi?id=2007363
- https://bugs.archlinux.org/task/71473
** Bug watch added: Red Hat Bugzilla #2007363
https://
Report exists, but no progress yet
https://gitlab.freedesktop.org/spice/spice/-/issues/64
** Also affects: spice-xpi via
https://gitlab.freedesktop.org/spice/spice/-/issues/64
Importance: Unknown
Status: Unknown
--
You received this bug notification because you are a member of Ubunt
Identified commit that causes this:
commit c23cbd6fa821fea8ac4ed97ca679afebe2333c8c
Author: Marc-André Lureau
Date: Thu Jan 28 13:59:24 2021 +0400
reds: start QXL devices if VM is running
Like char devices, QXL devices need to be explicily started.
For some historical reason, cha
With debuginfo into libspice this looks like:
(gdb) bt
#0 __pthread_create_2_1 (newthread=0x572d4f10, attr=0x0,
start_routine=0x77e33950 , arg=0x572d4f10) at
./nptl/pthread_create.c:621
#1 0x77eb11e6 in red_worker_run.isra.0
(worker=worker@entry=0x572d4f10) at ../serve
With the old lib, when the `qemu_spice_display_start` happens in the main
thread the spawned "SPICE worker" still is at:
(gdb) bt
#0 0x7770ad7f in __GI___poll (fds=0x7fffb80031b0, nfds=2,
timeout=2147483647) at ../sysdeps/unix/sysv/linux/poll.c:29
#1 0x77ab6686 in () at /lib/x8
It seems that when spawning the "SPICE worker" in the bad case
the worker passes events before the other side is ready.
In both cases the spawn goes through the same control path
#0 __pthread_create_2_1 (newthread=0x572d4f10, attr=0x0,
start_routine=0x77e33950, arg=0x572d4f10) at ./
Thread spawn for the spice worker:
Thread 1 "qemu-system-x86" hit Breakpoint 5, __pthread_create_2_1
(newthread=0x572d1690, attr=0x0, start_routine=0x77e3d140,
arg=0x572d1690) at pthread_create.c:597
597 in pthread_create.c
(gdb) info threads
Id Target Id
In the bad case the spice worker is at the fist qxl_send_events while at
the same time the main process initializing is still not done yet.
(gdb) info thread
Id Target IdFrame
1Thread 0x767dbc40 (LWP 144970) "qemu-system-x86"
qemu_spice_d
Good case (qemu_spice_display_start before qxl_send_events)
Thread 1 "qemu-system-x86" hit Breakpoint 4, qemu_spice_display_start () at
../ui/spice-core.c:995
995 if (spice_display_is_running) {
(gdb) bt
#0 qemu_spice_display_start () at ../ui/spice-core.c:995
#1 0x55a8b6b7 in q
Bad case (qemu_spice_display_start after qxl_send_events)
Thread 4 "SPICE Worker" hit Breakpoint 3, qxl_send_events (d=0x57194210,
events=16) at ../hw/display/qxl.c:1867
1867uint32_t le_events = cpu_to_le32(events);
(gdb) bt
#0 qxl_send_events (d=0x57194210, events=16) at ../hw/d
Status changed to 'Confirmed' because the bug affects multiple users.
** Changed in: qemu (Ubuntu)
Status: New => Confirmed
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1964777
Title:
qxl_s
Compare the same git build between Impish and Jammy:
~/qemu/build$ gdb ./qemu-system-x86_64
(gdb) watch spice_display_is_running
(gdb) b qxl_send_events
(gdb) b qemu_spice_display_is_running
(gdb) b qemu_spice_display_start
(gdb) handle SIGUSR1 nostop noprint pass
(gdb) run -cpu qemu64 -machine q3
We see that on Jammy the initialization ordering is different, here we get a
call to `qxl_send_events` before qemu_spice_display_start happened and
therefore triggering the error message.
Later on the init happens and that might explain why I see no fatal issues
later on on when using spice.
Never
An -O0 build later for debugging...
SimpleSpiceDisplay *ssd is ignored, the value qemu_spice_display_is_running
returns is the global spice_display_is_running of ui/spice-core.c
$6 = (int *) 0x5650d5b4
That is always 0 up to the point this checks it.
A watch on ~/work/qemu/qemu.git/ui/spice
I wanted to start bisecting to identify the offending commit.
But what I've found is that even old build e.g. v5.2.0 are broken
in this Ubuntu 22.04 test system.
In return I checked older Ubuntu systems if I might just have missed the
message before. But up until 21.10 with qemu 6.0 it was just fi
Looking at the code that warning is from here:
```
static void qxl_send_events(PCIQXLDevice *d, uint32_t events)
{
uint32_t old_pending;
uint32_t le_events = cpu_to_le32(events);
trace_qxl_send_events(d->id, events);
if (!qemu_spice_display_is_running(&d->ssd)) {
/*
24 matches
Mail list logo
