On Wed, Jun 29, 2011 at 12:02:30PM +0200, Gerd Hoffmann wrote: > Hi, > > >Yeah, my bad. I did something wierd, and missed this. I'll send a v3. Do you > >have other comments? > > I stopped looking after figuring stuff doesn't apply to master. > > >Unrelated, but have you looked at the async patches that I sent, and yonit's > >comments? > > - (me) catch EINTR and EAGAIN in async thread read > > - (yonit) reset surfaces count and surfaces when doing destroy_all > > http://cgit.freedesktop.org/spice/qemu/log/?h=bz700134 is what I > currently have. Some of your patches are in there already.
I tried rebasing on that, testing showed there was a deadlock, you can't reuse wlock for track_command since it is called from interface_get_command, called from red_worker, which may be during update_area, which already took the wlock. Adding a separate track_lock will probably work. > > cheers, > Gerd >