On Tue, 6 Dec 2016 16:21:30 -0800 Miguel Angel Vico <mvicom...@nvidia.com> wrote:
> Hi Eric, > > The problem here is we are using crtc_id and connector_id as indexes, > while they aren't. > > Nothing keeps DRM drivers from advertising CRTC and connectors with ids > that aren't sequential and starting at 0 or 1. Hi, yes, we at Collabora have recently discovered this problem too. > The right fix is to use indexes instead of ids to update bitmasks. > > drm_output::pipe stores the CRTC index and can be used to update CRTC > bitmasks. Somehow I don't think that's a good plan. > For connectors, we need to add something similar. No, there are simpler solutions. We should keep track of the ids. The currently-in-use ids are already discoverable through the list of weston_outputs. We just needs arrays of ids for the other uses. The bit masks should just go. Thanks, pq > On Mon, 5 Dec 2016 21:49:40 +0000 > "Ruei, Eric" <e-ru...@ti.com> wrote: > > > To whom it may concern: > > > > We have encountered a Weston hotplug issue with DRM backend where the > > hotplug of HDMI connector caused unexpected Weston exit recently. The > > root cause of the problem is that the DRM connector id of the HDMI > > display exceeds 31, but the corresponding active bitmap variable > > connector_allocator is only 32-bit. The same issue is also applicable > > to the DRM instance variable crtc_allocator and all the local > > variables of CRTC or connector bitmap. A tentative quick fix is to > > increase the size of bitmap variable from uint32_t to uint64_t, but > > it will break again when the DRM id exceeds 63. > > > > Attached please find our tentative patch against Weston 1.9 as your > > reference. > > > > Best regards, > > > > Eric > > > > > >
pgpVIBFU2IfCp.pgp
Description: OpenPGP digital signature
_______________________________________________ wayland-devel mailing list wayland-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/wayland-devel