Using MOBFMT_RANGE in the early days of guest backed objects was a major
performance win but that has changed a lot since. There's no more
a performance reason to use MOBFMT_RANGE. The device can/will still
profit from the pages being contiguous but marking them as MOBFMT_RANGE
no longer matters.
Benchmarks (e.g. heaven, valley) show that creating page tables
for mob memory is actually faster than using mobfmt ranges.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-12-zack@kde.org
The fb handle code assumes it deals with GEM objects. Because vmwgfx
buffer objects were not actually GEM objects we were not able to
implement that interface. Now that vmwgfx supports GEM buffer objects
we can trivially implement create_handle for buffer object backed
framebuffers.
Among others this gets IGT's kms_getfb test passing.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-6-zack@kde.org
This is initial change adding support for DRIVER_GEM to vmwgfx. vmwgfx
was written before GEM and has always used TTM. Over the years the
TTM buffers started inherting from GEM objects but vmwgfx never
implemented GEM making it quite awkward. We were directly setting
variables in GEM objects to not make DRM crash.
This change brings vmwgfx inline with other DRM drivers and allows us
to use a lot of DRM helpers which have depended on drivers with GEM
support.
Due to historical reasons vmwgfx splits the idea of a buffer and surface
which makes it a littly tricky since either one can be used in most
of our ioctl's which take user space handles. For now our BO's are
GEM objects and our surfaces are opaque objects which are backed by
GEM objects. In the future I'd like to combine those into a single
BO but we don't want to break any of our existing ioctl's so it will
take time to do it in a non-destructive way.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-5-zack@kde.org
Mob/GMR id resource manager was lacking the debug print callback
which meant that during memory errors we weren't getting the details
which are needed to fix those errors.
Kernel logs need to contain the information about used/max pages
by the Mob/GMR id resource manager as well as the maximum number
of id's they're allowed to allocate.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-3-zack@kde.org
vmwgfx shared very elaborate memory accounting with ttm. It was moved
from ttm to vmwgfx in change
f07069da6b ("drm/ttm: move memory accounting into vmwgfx v4")
but because of complexity it was hard to maintain. Some parts of the code
weren't freeing memory correctly and some were missing accounting all
together. While those would be fairly easy to fix the fundamental reason
for memory accounting in the driver was the ability to invoke shrinker
which is part of TTM code as well (with support for unified memory
hopefully coming soon).
That meant that vmwgfx had a lot of code that was either unused or
duplicating code from TTM. Removing this code also prevents excessive
calls to global swapout which were common during memory pressure
because both vmwgfx and TTM would invoke the shrinker when memory
usage reached half of RAM.
Fixes: f07069da6b ("drm/ttm: move memory accounting into vmwgfx v4")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206172620.3139754-2-zack@kde.org
Using the SET_RUNTIME_PM_OPS() macro causes a warning about the
referenced functions when they are marked static but not __maybe_unused:
drivers/gpu/drm/omapdrm/dss/dss.c:1572:12: error: unused function 'dss_runtime_suspend' [-Werror,-Wunused-function]
drivers/gpu/drm/omapdrm/dss/dss.c:1584:12: error: unused function 'dss_runtime_resume' [-Werror,-Wunused-function]
drivers/gpu/drm/omapdrm/dss/dispc.c:4845:12: error: unused function 'dispc_runtime_suspend' [-Werror,-Wunused-function]
drivers/gpu/drm/omapdrm/dss/dispc.c:4860:12: error: unused function 'dispc_runtime_resume' [-Werror,-Wunused-function]
Fixes: b92f7ea556 ("drm/omap: dss: Make use of the helper macro SET_RUNTIME_PM_OPS()")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211205131612.3192652-1-arnd@kernel.org
If the drm_plane has a source width that's greater than the max width
supported by a single hw overlay, then we assign a 'r_overlay' to it in
omap_plane_atomic_check().
Both overlays should have the capabilities required to handle the source
framebuffer. The only parameters that vary between the left and right
hwoverlays are the src_w, crtc_w, src_x and crtc_x as we just even chop
the fb into left and right halves.
We also take care of not creating odd width size when dealing with YUV
formats.
Since both halves need to be 'appear' side by side the zpos is
recalculated when dealing with dual overlay cases so that the other
planes zpos is consistent.
Depending on user space usage it is possible that on occasion the number
of requested planes exceeds the numbers of overlays required to display
them. In that case a failure would be returned for the plane that cannot
be handled at that time. It is up to user space to make sure the H/W
resource are not over-subscribed.
Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-10-narmstrong@baylibre.com
(re)assign the hw overlays to planes based on required caps, and to
handle situations where we could not modify an in-use plane.
This means all planes advertise the superset of formats and properties.
Userspace must (as always) use atomic TEST_ONLY step for atomic updates,
as not all planes may be available for use on every frame.
The mapping of hwoverlays to plane is stored in omap_global_state, so
that state updates are atomically committed in the same way that
plane/etc state updates are managed. This is needed because the
omap_plane_state keeps a pointer to the hwoverlay, and we don't want
global state to become out of sync with the plane state if an atomic
update fails, we hit deadlock/ backoff scenario, etc. The use of
global_state_lock keeps multiple parallel updates which both re-assign
hwoverlays properly serialized.
Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-8-narmstrong@baylibre.com
Global shared resources (like hw overlays) for omapdrm are implemented
as a part of atomic state using the drm_private_obj infrastructure
available in the atomic core.
omap_global_state is introduced as a drm atomic private object. The two
funcs omap_get_global_state() and omap_get_existing_global_state() are
the two variants that will be used to access omap_global_state.
drm_mode_config_init() needs to be called earlier because it
creates/initializes the private_obj link list maintained by the atomic
framework. The private_obj link list has to exist prior to calling
drm_atomic_private_obj_init(). Similarly the cleanup handler are
reordered appropriately.
Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-7-narmstrong@baylibre.com
When we added the support for the AUX channel in commit 13afcdd727
("drm/bridge: parade-ps8640: Add support for AUX channel") we forgot
to set "drm_dev" to avoid the warning splat at the beginning of
drm_dp_aux_register(). Since everything was working I guess I never
noticed the splat when testing against mainline. In any case, it's
easy to fix. This is basically just like commit 6cba3fe433 ("drm/dp:
Add backpointer to drm_device in drm_dp_aux") but just for the
parade-ps8640.
Fixes: 13afcdd727 ("drm/bridge: parade-ps8640: Add support for AUX channel")
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Philip Chen <philipchen@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206162907.1.I1f5d1eba741e4663050ec1b8e39a753f6e42e38b@changeid
The aspeed driver uses the gem_cma_helper code, but does
noto enforce enabling this through Kconfig:
x86_64-linux-ld: drivers/gpu/drm/aspeed/aspeed_gfx_drv.o:(.rodata+0x2c8): undefined reference to `drm_gem_cma_prime_import_sg_table'
x86_64-linux-ld: drivers/gpu/drm/aspeed/aspeed_gfx_drv.o:(.rodata+0x2d8): undefined reference to `drm_gem_cma_dumb_create'
x86_64-linux-ld: drivers/gpu/drm/aspeed/aspeed_gfx_crtc.o: in function `aspeed_gfx_pipe_update':
aspeed_gfx_crtc.c:(.text+0xe5): undefined reference to `drm_fb_cma_get_gem_obj'
Add the same 'select' that is used in other such drivers.
Fixes: 09717af7d1 ("drm: Remove CONFIG_DRM_KMS_CMA_HELPER option")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20211204174637.1160725-2-arnd@kernel.org
The fsl-dcu driver uses the gem_cma_helper code, but does
noto enforce enabling this through Kconfig:
arm-linux-gnueabi-ld: drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.o:(.rodata+0x32c): undefined reference to `drm_gem_cma_prime_import_sg_table'
arm-linux-gnueabi-ld: drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.o:(.rodata+0x334): undefined reference to `drm_gem_cma_dumb_create'
arm-linux-gnueabi-ld: drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.o: in function `fsl_dcu_drm_plane_atomic_update':
fsl_dcu_drm_plane.c:(.text+0x144): undefined reference to `drm_fb_cma_get_gem_obj'
Add the same 'select' that is used in other such drivers.
Fixes: 09717af7d1 ("drm: Remove CONFIG_DRM_KMS_CMA_HELPER option")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20211204174637.1160725-1-arnd@kernel.org
GEM helper libraries use struct drm_driver.gem_create_object to let
drivers override GEM object allocation. On failure, the call returns
NULL.
Change the semantics to make the calls return a pointer-encoded error.
This aligns the callback with its callers. Fixes the ingenic driver,
which already returns an error pointer.
Also update the callers to handle the involved types more strictly.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Steven Price <steven.price@arm.com>
Acked-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20211130095255.26710-1-tzimmermann@suse.de
TTM takes full control over TTM_PL_SYSTEM placed buffers. This makes
driver internal usage of TTM_PL_SYSTEM prone to errors because it
requires the drivers to manually handle all interactions between TTM
which can swap out those buffers whenever it thinks it's the right
thing to do and driver.
CPU buffers which need to be fenced and shared with accelerators should
be placed in driver specific placements that can explicitly handle
CPU/accelerator buffer fencing.
Currently, apart, from things silently failing nothing is enforcing
that requirement which means that it's easy for drivers and new
developers to get this wrong. To avoid the confusion we can document
this requirement and clarify the solution.
This came up during a discussion on dri-devel:
https://lore.kernel.org/dri-devel/232f45e9-8748-1243-09bf-56763e6668b3@amd.com
Signed-off-by: Zack Rusin <zackr@vmware.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211110145034.487512-1-zackr@vmware.com
For larger (bigger than a page) and noncontiguous mobs we have
to create page tables that allow the host to find the memory.
Those page tables just used regular system memory. Unfortunately
in TTM those BO's are not allowed to be busy thus can't be
fenced and we have to fence those bo's because we don't want
to destroy the page tables while the host is still executing
the command buffers which might be accessing them.
To solve it we introduce a new placement VMW_PL_SYSTEM which
is very similar to TTM_PL_SYSTEM except that it allows
fencing. This fixes kernel oops'es during unloading of the driver
(and pci hot remove/add) which were caused by busy BO's in
TTM_PL_SYSTEM being present in the delayed deletion list in
TTM (TTM_PL_SYSTEM manager is destroyed before the delayed
deletions are executed)
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-5-zackr@vmware.com
Some of our hosts have a bug where rescaning a pci bus results in stale
fifo memory being mapped on the host. This makes any fifo communication
impossible resulting in various kernel crashes.
Instead of unexpectedly crashing, predictably fail to load the driver
which will preserve the system.
Fixes: fb1d9738ca ("drm/vmwgfx: Add DRM driver for VMware Virtual GPU")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-4-zackr@vmware.com
Link drm_fb_cma_helper.o into drm_cma_helper.ko if CONFIG_DRM_KMS_HELPER
has been set. Remove CONFIG_DRM_KMS_CMA_HELPER config option. Selecting
KMS helpers and CMA will now automatically enable CMA KMS helpers.
Some drivers' Kconfig files did not correctly select KMS or CMA helpers.
Fix this as part of the change.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211106193509.17472-3-tzimmermann@suse.de
Change all GEM CMA object functions that receive a GEM object
of type struct drm_gem_object to expect an object of type
struct drm_gem_cma_object instead.
This change reduces the number of upcasts from struct drm_gem_object
by moving them into callers. The C compiler can now verify that the
GEM CMA functions are called with the correct type.
For consistency, the patch also renames drm_gem_cma_free_object to
drm_gem_cma_free. It further updates documentation for a number of
functions.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115120148.21766-4-tzimmermann@suse.de
