Allow to set priorities for buffer objects. Use priority 1 for surface
and cursor command releases. Use priority 0 for drawing command
releases. That way the short-living drawing commands are first in line
when it comes to eviction, making it *much* less likely that
ttm_bo_mem_force_space() picks something which can't be evicted and
throws an error after waiting a while without success.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: http://patchwork.freedesktop.org/patch/msgid/20210217123213.2199186-4-kraxel@redhat.com
It's possible to modeset a connector/mst port that has a 0 full_pbn
value: if the sink on the port deasserts its HPD and a branch device
reports this via a CSN with the port's ddps=0 and pdt!=NONE the driver
clears full_pbn, but the corresponding connector can be still
modesetted.
This happened on a DELL U2719D monitor as the branch device and an LG
27UL650-W daisy-chained to it, the LG monitor generating a long HPD
pulse (doing this for some reason always when waking up from some power
saving state).
Tune down the WARN about this scenario to a debug message.
v2: Use the correct atomic debug message level. (Lyude)
References: https://gitlab.freedesktop.org/drm/intel/-/issues/1917
Cc: Lyude Paul <lyude@redhat.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210216123448.410545-1-imre.deak@intel.com
The BOs of the hardware cursor are now mapped permanently while the
cursor plane is being used. This reduces the CPU overhead of the cursor
plane's atomic_update function.
The change also resolves a problem with the vmap call in the commit tail.
The vmap implementation could acquire the DMA reservation lock on the
BO, which is not allowed that late in the atomic update. Removing the
vmap call from atomic_update fixes the issue.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210209134632.12157-9-tzimmermann@suse.de
Our sysfs "modes" entries were broken because our preffered mode
never had its name set correctly. This resulted in the first
entry simply being called "preferred" followed by a list of
other resolutions. Lets fix it by actually setting the name of
mode (which is its resolution). This allows one to quickly
validate the modes set by the open-vm-tools.
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210209161700.335611-1-zackr@vmware.com
There are minor differences in the values for the threshold value and
the scan line size between families of ASPEED SoC. Additionally the SCU
registers for the output control and scratch registers differ between
families.
This adds device tree matching to parameterise these values, allowing us
to add support for the AST2400 now, and in the future the AST2600.
Reviewed-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Joel Stanley <joel@jms.id.au>
Link: https://patchwork.freedesktop.org/patch/msgid/20210209123734.130483-3-joel@jms.id.au
When run xrandr to change resolution on Beaglebone Black board, it will
print the error information:
root@beaglebone:~# xrandr -display :0 --output HDMI-1 --mode 720x400
[drm:drm_crtc_commit_wait] *ERROR* flip_done timed out
[drm:drm_atomic_helper_wait_for_dependencies] *ERROR* [CRTC:32:tilcdc crtc] commit wait timed out
[drm:drm_crtc_commit_wait] *ERROR* flip_done timed out
[drm:drm_atomic_helper_wait_for_dependencies] *ERROR* [CONNECTOR:34:HDMI-A-1] commit wait timed out
[drm:drm_crtc_commit_wait] *ERROR* flip_done timed out
[drm:drm_atomic_helper_wait_for_dependencies] *ERROR* [PLANE:31:plane-0] commit wait timed out
tilcdc 4830e000.lcdc: already pending page flip!
This is because there is operation sequence as below:
drm_atomic_connector_commit_dpms(mode is DRM_MODE_DPMS_OFF):
...
drm_atomic_helper_setup_commit <- init_completion(commit_A->flip_done)
drm_atomic_helper_commit_tail
tilcdc_crtc_atomic_disable
tilcdc_plane_atomic_update <- drm_crtc_send_vblank_event in tilcdc_crtc_irq
is skipped since tilcdc_crtc->enabled is 0
tilcdc_crtc_atomic_flush <- drm_crtc_send_vblank_event is skipped since
crtc->state->event is set to be NULL in
tilcdc_plane_atomic_update
drm_mode_setcrtc:
...
drm_atomic_helper_setup_commit <- init_completion(commit_B->flip_done)
drm_atomic_helper_wait_for_dependencies
drm_crtc_commit_wait <- wait for commit_A->flip_done completing
Just as shown above, the steps which could complete commit_A->flip_done
are all skipped and commit_A->flip_done will never be completed. This will
result a time-out ERROR when drm_crtc_commit_wait check the commit_A->flip_done.
So add drm_crtc_send_vblank_event in tilcdc_crtc_atomic_disable to
complete commit_A->flip_done.
Fixes: cb345decb4 ("drm/tilcdc: Use standard drm_atomic_helper_commit")
Signed-off-by: Quanyang Wang <quanyang.wang@windriver.com>
Reviewed-by: Jyri Sarha <jyri.sarha@iki.fi>
Tested-by: Jyri Sarha <jyri.sarha@iki.fi>
Signed-off-by: Jyri Sarha <jyri.sarha@iki.fi>
Link: https://patchwork.freedesktop.org/patch/msgid/20210209082415.382602-1-quanyang.wang@windriver.com
TTM implements a rather extensive accounting of allocated memory.
There are two reasons for this:
1. It tries to block userspace allocating a huge number of very small
BOs without accounting for the kmalloced memory.
2. Make sure we don't over allocate and run into an OOM situation
during swapout while trying to handle the memory shortage.
This is only partially a good idea. First of all it is perfectly
valid for an application to use all of system memory, limiting it to
50% is not really acceptable.
What we need to take care of is that the application is held
accountable for the memory it allocated. This is what control
mechanisms like memcg and the normal Linux page accounting already do.
Making sure that we don't run into an OOM situation while trying to
cope with a memory shortage is still a good idea, but this is also
not very well implemented since it means another opportunity of
recursion from the driver back into TTM.
So start to rework all of this by implementing a shrinker callback which
allows for TT object to be swapped out if necessary.
v2: Switch from limit to shrinker callback.
v3: fix gfp mask handling, use atomic for swapable_pages, add debugfs
v4: drop the extra gfp_mask checks
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20210208133226.36955-1-christian.koenig@amd.com
Functions in the atomic commit tail are not allowed to acquire the
dmabuf's reservation lock. So we cannot legally call the GEM object's
vmap functionality in atomic_update.
Instead use struct drm_shadow_plane_state and friends. It vmaps the
framebuffer BOs in prepare_fb and vunmaps them in cleanup_fb. The
cursor plane state stores the mapping's address. The pinning of the
BO is implicitly done by vmap.
As an extra benefit, there's no source of runtime errors left in
atomic_update.
v2:
* rebase patch onto struct drm_shadow_plane_state
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210209121042.24098-3-tzimmermann@suse.de
Vmap operations may acquire the dmabuf reservation lock, which is not
allowed within atomic commit-tail functions. Therefore move vmap and
vunmap from the damage handler into prepare_fb and cleanup_fb callbacks.
The mapping is provided as GEM shadow-buffered plane. The functions in
the commit tail use the pre-established mapping for damage handling.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210208115538.6430-8-tzimmermann@suse.de
Vmap operations may acquire the dmabuf reservation lock, which is not
allowed within atomic commit-tail functions. Therefore move vmap and
vunmap from the damage handler into prepare_fb and cleanup_fb callbacks.
The mapping is provided as GEM shadow-buffered plane. The functions in
the commit tail use the pre-established mapping for damage handling.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210208115538.6430-7-tzimmermann@suse.de
