On Mon, Jun 22, 2015 at 04:08:12PM +0200, Javier Celaya wrote: > El Lunes, 22 de junio de 2015 13:39:17 Christophe Fergeau escribió: > > Hey, > > > > On Fri, Jun 19, 2015 at 02:05:10PM +0200, Javier Celaya wrote: > > > When KMS is enabled, the mm_time value of a QXLDrawable is undefined. > > > This results in severe synchronization problems. This patch sets it to > > > zero to force the spice-server to use the local clock. > > > --- > > > > > > src/qxl_surface.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/src/qxl_surface.c b/src/qxl_surface.c > > > index c5e76e0..95d34db 100644 > > > --- a/src/qxl_surface.c > > > +++ b/src/qxl_surface.c > > > @@ -86,6 +86,7 @@ make_drawable (qxl_screen_t *qxl, qxl_surface_t *surf, > > > uint8_t type,> > > > if (!qxl->kms_enabled) > > > > > > drawable->mm_time = qxl->rom->mm_clock; > > > > > > + else drawable->mm_time = 0; > > > > With what spice-server version have you been having this issue? Since > > commit c541d7e29dc0053e6, mm_time should be unconditionnally set > > server-side. Is there anything causing qxl->rom->mm_clock not to work > > with older spice-server and kms? In other words, why not remove the if > > (!qxl->kms_enabled) check? > > > > Christophe > > Hi Christophe > > Indeed the version of spice-server might be a bit old, it is v0.12.4. I have not tested > with a newer one. However I tried to remove the (!qxl->kms_enabled) check and that > resulted in a segmentation fault. I suppose qxl->rom is NULL with kms, but I don't > know why.-- Ah right, this kind of low level interaction with the device is done in the kernel module, not in the xorg driver. I'm fairly sure this is obsolete with spice-server.git, but it does not hurt to make sure this field is initialized. Christophe
Attachment:
pgpW116mm3qOO.pgp
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel