Do *not* use kernels with debugging options enabled when testing spice

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

 



Hi,

I've just spend the last 1.5 days tracking down some weird latency issues
which were showing up when testing spice usb-redir over a network connection
(rather then over loopback).

I managed to trace it to a simple read() call on the socket for the
(unencrypted) usb-redir channel taking up to 7 ms. Since this is done
from qemu's iothread, this blocks amongst other things the ehci
controller emulations timer from running for that amount, while it
should run every ms.

At first I thought the issue was because the socket was maybe not
set to non-blocking. But in the end it turns out to be a kernel issue,
one of the debugging options enabled in Fedora-18's / rawhides kernels
causes these latencies.

I've already verified that this is not an issue introduced in the
3.6 kernel versus the stable 3.5 kernel F-17 has, when ever a
new rc gets released one build is done in koji without the debugging
options, ie:
http://koji.fedoraproject.org/koji/buildinfo?buildID=354745

These builds work fine, only the builds with the debugging options show
these huge increases in latencies under certain circumstances.

Regards,

Hans

p.s.

The above is all just FYI, I don't expect anyone to do something about this,
I just wanted others to be aware of this issue, so that they don't end up
chasing ghosts like I did.
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/spice-devel


[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]     [Monitors]