Re: [PATCH v3 0/9] Don't whitelist OA registers unconditionally

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Mon, Aug 10, 2026 at 04:45:07PM -0700, Dixit, Ashutosh wrote:
> On Mon, 22 Jun 2026 19:10:14 -0700, Dixit, Ashutosh wrote:
> >
> > On Mon, 22 Jun 2026 17:48:12 -0700, Dixit, Ashutosh wrote:
> > >
> > > On Mon, 15 Jun 2026 15:42:18 -0700, Ashutosh Dixit wrote:
> > > >
> > > > Whitelisting OA registers unconditionally is a security violation. In this
> > > > series we resolve this issue as follows:
> > > >
> > > > * Set the 'deny' bit (bit 30) for all OA registers, ensuring OA registers
> > > >   are not whitelisted by default after probe/reset/restart
> > > > * Reset the 'deny' bit when an OA stream is opened and certain conditions
> > > >   are met, whitelisting OA registers only for the duration when OA streams
> > > >   are open for a gt
> > > > * Set the 'deny' bit again, when OA streams are closed
> > > > * To manage this scheme, separate out OA whitelists from non-OA whitelists
> > > >   (into separate save-restore lists)
> > > >
> > > > v2: Address code review from Umesh. Patches changed in v2 have changelog
> > > >     appended to commit message
> > > > v3: Minor change to Patch 3
> > > >
> > > > Ashutosh Dixit (9):
> > > >   drm/xe/rtp: Add RING_FORCE_TO_NONPRIV_DENY to OA whitelists
> > > >   drm/xe/rtp: Maintain OA whitelists separately
> > > >   drm/xe/rtp: Keep track of non-OA nonpriv slots
> > > >   drm/xe/rtp: Generalize whitelist_apply_to_hwe
> > > >   drm/xe/rtp: Save OA nonpriv registers to register save/restore lists
> > > >   drm/xe/rtp: Toggle 'deny' bit to (de-)whitelist OA regs
> > > >   drm/xe/rtp: (De-)whitelist OA registers for all hwe's for a gt
> > > >   drm/xe/oa: (De-)whitelist OA registers on OA stream open/release
> > > >   drm/xe/rtp: Ensure locking/ref counting for OA whitelists
> > >
> > > I have added the following to all the patches here and merge this series:
> > >
> > >	Cc: stable@vger.kernel.org # v6.12+
> > >	Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
> >
> > Sorry I meant:
> >
> >	Fixes: 828a8eaf37c3 ("drm/xe/oa: Add MMIO trigger support")
> >	Cc: stable@vger.kernel.org # v6.12+
> >
> > >
> > > So the plan is to propagate this series to the previous LTS kernel versions
> > > too, in order to plug this security violation related to unconditional OA
> > > register whitelisting.
> > >
> > > I am preparing v6.12 and v6.18 stable kernel version series, based on this
> > > series. Since these are a bit old, this series needs porting to these old
> > > kernel versions. These will be sent after this series reaches Linus'
> > > master.
> >
> > This is needed for patches to be added to stable versions.
> 
> The series has now been applied to all existing stable branches.

Awesome! Thank you so much for all the great work and the follow-up.

> The
> version appied to the branches can be seen at the links below:
> 
> Remote:
> 	https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
> 
> Branches:
> 
> stable/linux-6.12.y:
> 	https://lore.kernel.org/stable/20260805024915.3407033-1-ashutosh.dixit@intel.com/
> 
> stable/linux-6.18.y:
> 	https://lore.kernel.org/stable/20260805024614.3406679-1-ashutosh.dixit@intel.com/
> 
> stable/linux-7.1.y:
> 	https://lore.kernel.org/stable/20260805024300.3406307-1-ashutosh.dixit@intel.com/
> 
> Thanks.
> --
> Ashutosh