[PATCH 1/3] drm: change DROP_MASTER permissions to allow DRM_MASTER

Emil Velikov emil.l.velikov at gmail.com
Thu Dec 20 19:09:27 UTC 2018


On Thu, 20 Dec 2018 at 14:45, Daniel Vetter <daniel at ffwll.ch> wrote:
>
> On Thu, Dec 20, 2018 at 01:50:26PM +0000, Emil Velikov wrote:
> > On Wed, 19 Dec 2018 at 20:36, Daniel Vetter <daniel at ffwll.ch> wrote:
> > >
> > > On Wed, Dec 19, 2018 at 07:22:45PM +0000, Emil Velikov wrote:
> > > > From: Emil Velikov <emil.velikov at collabora.com>
> > > >
> > > > Currently only DRM_ROOT_ONLY is allowed to call the ioctl.
> > > >
> > > > Change that to DRM_MASTER, which means that only a process that is the
> > > > current DRM master can drop it. Which makes sense, the process should
> > > > be able to opt-out without any specific requirements.
> > > >
> > > > Signed-off-by: Emil Velikov <emil.velikov at collabora.com>
> > >
> > > I guess this makes sense, but then you already need someone else to do the
> > > setmaster for you if you want to run as non-root and be able to switch
> > > between compositors. So no idea where this will be useful.
> > >
> > X, Weston and the Gnome/KDE wayland compositors use logind for managing that.
> > Some have codepaths to manage drm{Set,Drop}Master manually, although
> > they don't seems to bother adjusting privileges, I'd imagine due to VT
> > switching.
> >
> > If ones has CONFIG_VT=n system, then it should be a matter of once-off
> > drmSetMaster + lower priv.
> >
> > > Either way: New uapi -> needs the userspace patches to exist.
> >
> > Slightly confused - apps already use the uapi, what do you mean with
> > "new uapi" here?
> > I'm OK with adding an IGT, although beyond that I'm not sure what
> > other userspace patches I could provide.
>
> You change the uapi to allow more stuff (dropmaster without having
> CAP_SYS_ADMIN), that needs userspace. Since current userspace has no use
> for calling drop_master without being root.
>
> Same way your patch to automatically auth clients if the driver supports
> rendernodes is a uapi extension, and it's good to know what code exactly
> it's meant for.
>
Ack. Upon further testing I've spotted some issues, so this will go on
the back-burner.

Thanks
Emil


More information about the dri-devel mailing list