Bug 370431

+++ This bug was initially created as a clone of Bug #370341 +++

+++ This bug was initially created as a clone of Bug #368931 +++

Description of problem:

When using 'xm create' to start a paravirt guest, the SDL window often
disappears part way through the boot.

Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1. Configure a pv guest to use SDL graphics
2. xm create <pv guest>
3. wait for the guest to boot all the way up
  
Actual results:
I've seen the SDL window disappear on both i386 and ia64 about 20% of the time. 
I see nothing suspicious in xend.log. There is no output to xend-debug.log.

Expected results:
The SDL window should remain present every time a paravirt guest is started.

Additional info:

Looks similar to bz# 217635

-- Additional comment from berrange on 2007-11-07 10:52 EST --
Co-incidentally yesterday i discovered a potential bug in the guest kernel PVFB
driver. It occassionally sends a framebuffer update event to the Dom0 with a x,y
of MAX_INT, MAX_INT, and a width+height of -MAX_INT, -MAX_INT.   When SDL then
tries to process this update event it gets an BadValue error from Xorg and
quits. In my tests against xen-unstable this happened on as many as 30% of all
bootup attempts with SDL, so I think its reasonably likely this is the same
problem you're describing. The VNC display doesn't seem to be affected, since
its update routine does region clipping. 



-- Additional comment from pm-rhel on 2007-11-07 10:54 EST --
This request was evaluated by Red Hat Product Management for inclusion in a Red
Hat Enterprise Linux maintenance release.  Product Management has requested
further review of this request by Red Hat Engineering, for potential
inclusion in a Red Hat Enterprise Linux Update release for currently deployed
products.  This request is not yet committed for inclusion in an Update
release.

-- Additional comment from armbru on 2007-11-07 16:22 EST --
Created an attachment (id=250871)
Fix tracking of screen dirtiness

The PVFB front end can get confused and send a bogus screen update to the back
end when the screen is clean.

------------------------------------------------------------------------------
Since the RHEL-4 fbfront stuff is based on the RHEL-5 stuff, I'm pretty sure it
has the same bug.  Cloning.