This can be used to create a separate DRM file description, thus
creating a new GEM handle namespace.
My use-case is wlroots. The library splits responsibilities between
separate components: the GBM allocator creates buffers, the GLES2
renderer uses EGL to import them and render to them, the DRM
backend imports the buffers and displays them. wlroots has a
modular architecture, and any of these components can be swapped
and replaced with something else. For instance, the pipeline can
be set up so that the DRM dumb buffer allocator is used instead of
GBM and the Pixman renderer is used instead of GLES2. Library users
can also replace any of these components with their own custom one.
DMA-BUFs are used to pass buffer references across components. We
could use GEM handles instead, but this would result in pain if
multiple GPUs are in use: wlroots copies buffers across GPUs as
needed. Importing a GEM handle created on one GPU into a completely
different GPU will blow up (fail at best, mix unrelated buffers
otherwise).
Everything is fine if all components use Mesa. However, this isn't
always desirable. For instance when running with DRM dumb buffers
and the Pixman software renderer it's unfortunate to depend on GBM
in the DRM backend just to turn DMA-BUFs into FB IDs. GBM loads
Mesa drivers to perform an action which has nothing driver-specific.
Additionally, drivers will fail the import if the 3D engine can't
use the imported buffer, for instance amdgpu will refuse to import
DRM dumb buffers [1]. We might also want to be running with a Vulkan
renderer and a Vulkan allocator in the future, and GBM wouldn't be
welcome in this setup.
To address this, GBM can be side-stepped in the DRM backend, and
can be replaced with drmPrimeFDToHandle calls. However because of
GEM handle reference counting issues, care must be taken to avoid
double-closing the same GEM handle. In particular, it's not
possible to share a DRM FD with GBM or EGL and perform some
drmPrimeFDToHandle calls manually.
So wlroots needs to re-open the DRM FD to create a new GEM handle
namespace. However there's no guarantee that the file-system
permissions will be set up so that the primary FD can be opened
by the compsoitor. On modern systems seatd or logind is a privileged
process responsible for doing this, and other processes aren't
expected to do it. For historical reasons systemd still allows
physically logged in users to open primary DRM nodes, but this
doesn't work on non-systemd setups and it's desirable to lock
them down at some point.
Some might suggest to open the render node instead of re-opening
the primary node. However some systems don't have a render node
at all (e.g. no GPU, or a split render/display SoC).
Solutions to this issue have been discussed in [2]. One solution
would be to open the magic /proc/self/fd/<fd> file, but it's a
Linux-specific hack (wlroots supports BSDs too). Another solution
is to add support for re-opening a DRM primary node to seatd/logind,
but they don't support it now and really haven't been designed for
this (logind would need to grow a completely new API, because it
assumes unique dev_t IDs). Also this seems like pushing down a
kernel limitation to user-space a bit too hard.
Another solution is to allow creating empty DRM leases. The lessee
FD would have its own GEM handle namespace, so wouldn't conflict
wth GBM/EGL. It would have the master bit set, but would be able
to manage zero resources. wlroots doesn't intend to share this FD
with any other process.
All in all IMHO that seems like a pretty reasonable solution to the
issue at hand.
Note, I've discussed with Jonas Ådahl and Mutter plans to adopt a
similar design in the future.
Example usage in wlroots is available at [3]. IGT test available
at [4].
[1]: https://github.com/swaywm/wlroots/issues/2916
[2]: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/110
[3]: https://github.com/swaywm/wlroots/pull/3158
[4]: https://patchwork.freedesktop.org/series/94323/
Signed-off-by: Simon Ser <contact@emersion.fr>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Daniel Stone <daniels@collabora.com>
Cc: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Keith Packard <keithp@keithp.com>
Cc: Boris Brezillon <boris.brezillon@collabora.com>
Cc: Dave Airlie <airlied@redhat.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210903130000.1590-2-contact@emersion.fr
We may end up in i915_ttm_bo_destroy() in an error path before the
object is fully initialized. In that case it's not correct to call
__i915_gem_free_object(), because that function
a) Assumes the gem object refcount is 0, which it isn't.
b) frees the placements which are owned by the caller until the
init_object() region ops returns successfully. Fix this by providing
a lightweight cleanup function __i915_gem_object_fini() which is also
called by __i915_gem_free_object().
While doing this, also make sure we call dma_resv_fini() as part of
ordinary object destruction and not from the RCU callback that frees
the object. This will help track down bugs where the object is incorrectly
locked from an RCU lookup.
Finally, make sure the object isn't put on the region list until it's
either locked or fully initialized in order to block list processing of
partially initialized objects.
v2:
- The TTM object backend memory was freed before the gem pages were
put. Separate this functionality into __i915_gem_object_pages_fini()
and call it from the TTM delete_mem_notify() callback.
v3:
- Include i915_gem_object_free_mmaps() in __i915_gem_object_pages_fini()
to make sure we don't inadvertedly introduce a race.
Fixes: 48b0961269 ("drm/i915: Move __i915_gem_free_object to ttm_bo_destroy")
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com> #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20210930113236.583531-1-thomas.hellstrom@linux.intel.com
The Wa_14014971508 is required to fix scanout when a feature that i915
do not support is enabled and this feature is not planned to be enabled
for adlp.
Keeping this workaround enabled can badly hurt power-savings when
a full frame fetch is required(see psr2_sel_fetch_plane_state_supported()
and psr2_sel_fetch_pipe_state_supported()).
Here a example that could badly hurt power-savings, userspace does
a page flip to a rotated plane, so CONTINUOS_FULL_FRAME set.
But then for a whole 30 seconds nothing in the screen requires updates
but because CONTINUOS_FULL_FRAME is set, it will not go into DC5/DC6.
Reverting Wa_14014971508 fixes that, as only a single frame will be
sent and then display can go to DC5/DC6 for those 30 seconds of
idleness.
BSpec: 54369
Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930001409.254817-6-jose.souza@intel.com
Legacy cursor APIs are handled by intel_legacy_cursor_update(), that
calls drm_atomic_helper_update_plane() when going through the
slow/atomic path to update cursor, what was the case for PSR2
selective fetch.
drm_atomic_helper_update_plane() sets
drm_atomic_state->legacy_cursor_update to true when updating the
cursor plane, to allow several cursor updates to happen within the
same frame, as userspace does that.
If drivers waited for a vblank increment at the end of every cursor
movement that would cause a visible lag in the cursor.
But this optimization do not properly work with PSR2 selective fetch
dirt area calculation, for example if within a single frame the cursor
had 3 moves the final dirt area programmed to PSR2_MAN_TRK_CTL would
be based in the second movement as old state and third movement as new
state, not updating the area where cursor was in the first state.
So here switching back to the fast path approach in
intel_legacy_cursor_update() and handling cursor movements as
frontbuffer rendering(psr_force_hw_tracking_exit()), that is not the
most optimal for power-savings but is the solution that we have until
mailbox style updates is implemented.
Also removing the cursor workaround as not it is properly undestand
the issue and is know that it will never cover all the cases.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930001409.254817-5-jose.souza@intel.com
When PSR2 selective fetch is enabled writes to CURSURFLIVE alone do
not causes the panel to be updated when doing frontbuffer rendering.
From what I was able to figure from experiments the writes to
CURSURFLIVE takes PSR2 from deep sleep but panel is not updated
because PSR2_MAN_TRK_CTL has no start and end region set.
As we don't have the dirt area from current flush and invalidate API
and even if we did userspace could do several draws to frontbuffer and
we would need a way to append all the damaged areas of all the draws
that need to be part of next frame.
So here only programing PSR2_MAN_TRK_CTL to do a single full frame
fetch.
It is a safe approach as if scanout is in the visible area
the single full frame will only be visible for hardware in the next
frame because of the double buffering, and if scanout is in vblank
area it will be draw in the current frame.
No need to disable PSR and wait a few miliseconds to enable it again.
Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930001409.254817-4-jose.souza@intel.com
This unnecessary flushes are hurting power-savings are it causes
features like PSR, FBC and DRRS to disable it self to handle
frontbuffer rendering, below some explanation of why each removed
call is not necessary.
The flush in intel_prepare_plane_fb() is not required as framebuffer
will be flipped and power-saving features do the proper flip handling
in hardware.
intel_find_initial_plane_obj() flush is not required because it is
only executed during driver load and at this point the power-saving
features are not even enabled.
And the last one intelfb_create(), is also not required as at this
point the fbdev was just allocated, userspace will draw on
it what will trigger frontbuffer invalidates and flushes later on.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930001409.254817-3-jose.souza@intel.com
PSR2 selective is not supported over rotated and scaled planes.
We had the rotation check in intel_psr2_sel_fetch_config_valid()
but that code path is only execute when a modeset is needed and
those plane parameters can change without a modeset.
Pipe selective fetch restrictions are also needed, it could be added
in intel_psr_compute_config() but pippe scaling is computed after
it is executed, so leaving as is for now.
There is no much loss in this approach as it would cause selective
fetch to not enabled as for alderlake-P and newer will cause it to
switch to PSR1 that will have the same power-savings as do full pipe
fetch.
Also need to check those restricions in the second
for_each_oldnew_intel_plane_in_state() loop because the state could
only have a plane that is not affected by those restricitons but
the damaged area intersect with planes that has those restrictions,
so a full pipe fetch is required.
v2:
- also handling pipe restrictions
BSpec: 55229
Reviewed-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com> # v1
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930001409.254817-1-jose.souza@intel.com
Setting DP_PORT_EN in intel_dp->DP is already handled by
intel_dp_enable_port() so there is no point in setting it also
from the link training code.
For DDI platforms a bit with that name doesn't even exist. The
counterpart is DDI_BUF_CTL_ENABLE, which is already set up by
intel_ddi_prepare_link_retrain(). Fortunately it is the same bit
so there was no harm in doing this from the platform independent
code as well. But it's just confusing when platform independent
code sets platform specific bits in intel_dp->DP. Just get rid
of it.
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930134310.31669-3-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak.intel.com>
I want intel_dp->DP to be fully populated by the time the
initial vswing programming happens. To that end move the
intel_ddi_init_dp_buf_reg() call to an earlier spot.
Additionally we don't want intel_ddi_init_dp_buf_reg() to
set DDI_BUF_CTL_ENABLE since the port should only get enabled
at the start of link training (see intel_ddi_prepare_link_retrain()).
So any earlier write to the register should not set the enable bit.
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930134310.31669-2-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>
Currently we clear the leftover vswing/preemphasis values only
at the start of link training. That means the initial vswing
programming performed during modeset is going to use stale values
left over from the previous link training sequence, and then at
the start of link training we're going to reset the levels back
to 0. Seems much better to make sure we start with level 0 from
the get go.
Additionally if LTTPRs are present the leftover vswing/preemphasis
values are those of the last link in the chain, so not the values
that our PHY is even using after a successful link training sequence.
So let's make sure everything is cleared up before we start
programming anything.
Suggested-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210930134310.31669-1-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>
