Ben Skeggs [Fri, 26 Jul 2024 04:37:55 +0000 (14:37 +1000)]
drm/nouveau: create pci device once
HW isn't touched anymore (aside from detection) until the first
nvif_device has been allocated, so we no longer need a separate
probe-only step before kicking efifb (etc) off the HW.
Ben Skeggs [Fri, 26 Jul 2024 04:37:54 +0000 (14:37 +1000)]
drm/nouveau: replace drm_device* with nouveau_drm* as dev drvdata
We almost always want to cast the pointer from dev_get_drvdata() to
'struct nouveau_drm *', so just directly store that pointer instead,
simplifying callers, and fixing some clumsy naming of dev/drm_dev
variables at the same time.
Ben Skeggs [Fri, 26 Jul 2024 04:37:53 +0000 (14:37 +1000)]
drm/nouveau: handle pci/tegra drm_dev_{alloc, register} from common code
Unify some more of the PCI/Tegra DRM driver init, both as a general
cleanup, and because a subsequent commit changes the pointer stored
via dev_set_drvdata(), and this allows the change to be made in one
place.
Hamza Mahfooz [Thu, 25 Jul 2024 20:51:08 +0000 (16:51 -0400)]
drm/vblank: add dynamic per-crtc vblank configuration support
We would like to be able to enable vblank_disable_immediate
unconditionally, however there are a handful of cases where a small off
delay is necessary (e.g. with PSR enabled). So, we would like to be able
to adjust the vblank off delay and disable imminent values dynamically
for a given CRTC. Since, it will allow drivers to apply static screen
optimizations more quickly and consequently allow users to benefit more
so from the power savings afforded by the aforementioned optimizations,
while avoiding issues in cases where an off delay is still warranted.
In particular, the PSR case requires a small off delay of 2 frames,
otherwise display firmware isn't able to keep up with all of the
requests made to amdgpu. So, introduce drm_crtc_vblank_on_config() which
is like drm_crtc_vblank_on(), but it allows drivers to specify the
vblank CRTC configuration before enabling vblanking support for a given
CRTC.
accel/qaic: Remove the description of DRM_IOCTL_QAIC_PART_DEV
The partition device ioctl was removed during the development of the
initial version of qaic driver. Remove its description from the
documentation to avoid confusing readers.
Before building an image, the build script looks to see if there are fixes
to apply from an upstream repository. The link for the upstream repository
git://anongit.freedesktop.org/drm/drm became obsolete with the move to
Gitlab server in March 2024. Until recently, this obsolete link was
harmless because anongit would at least respond that there were no such
fixes available. In the last few days anongit has stopped responding to
requests causing the build script to hang indefinitely.
Update the link from anongit to the Gitlab server to prevent the build
script from hanging indefinitely.
Marco Pagani [Tue, 23 Jul 2024 19:10:35 +0000 (21:10 +0200)]
drm/test: use kunit action wrapper macro in the gem shmem test suite
Replace deferred action function wrappers with equivalent ones defined
using the macro introduced by commit 56778b49c9a2 ("kunit: Add a macro to
wrap a deferred action function")
Marek Vasut [Wed, 24 Jul 2024 00:55:53 +0000 (02:55 +0200)]
drm/panel/panel-ilitek-ili9806e: Add Densitron DMT028VGHMCMI-1D TFT to ILI9806E DSI TCON driver
Add Densitron DMT028VGHMCMI-1D 480x640 TFT matrix 2.83 inch panel
attached to Ilitek ILI9806E DSI TCON into the ILI9806E driver.
Note that the Densitron panels use different TCONs, this driver is for
the later panel, use panel-ilitek-st7701.c for the former panel:
DMT028VGHMCMI-1A - ST7701
DMT028VGHMCMI-1D - ILI9806E
drm/panel: boe-th101mb31ig002 : Fix the way to get porch parameters
The current driver can only obtain the porch parameters
of boe-th101mb31ig002. Modify it to obtain the porch
parameters of the panel currently being used.
Enable prepare_prev_first flag for BF060Y8M-AJ0 panel so that the
previous controller should be prepared first before the prepare for
the panel is called.
This makes sure that the previous controller, likely to be a DSI host
controller should be initialized to LP-11 before the panel is powered up.
Make the default DP port preemphasis configurable via new DT property
"toshiba,pre-emphasis". This is useful in case the DP link properties
are known and starting link training from preemphasis setting of 0 dB
is not useful. The preemphasis can be set separately for both DP lanes
in range 0=0dB, 1=3.5dB, 2=6dB .
Document default DP port preemphasis configurable via new DT property
"toshiba,pre-emphasis". This is useful in case the DP link properties
are known and starting link training from preemphasis setting of 0 dB
is not useful. The preemphasis can be set separately for both DP lanes
in range 0=0dB, 1=3.5dB, 2=6dB .
This is an endpoint property, not a port property, because the TC9595
datasheet does mention that the DP might operate in some sort of split
mode, where each DP lane is used to feed one display, so in that case
there might be two endpoints.
drm/ttm: Allow direct reclaim to allocate local memory v2
Limiting the allocation of higher order pages to the closest NUMA node
and enabling direct memory reclaim provides not only failsafe against
situations when memory becomes too much fragmented and the allocator is
not able to satisfy the request from the local node but falls back to
remote pages (HUGEPAGE) but also offers performance improvement.
Accessing remote pages suffers due to bandwidth limitations and could be
avoided if memory becomes defragmented and in most cases without using
manual compaction. (/proc/sys/vm/compact_memory)
Note: On certain distros such as RHEL, the proactive compaction is
disabled. (https://tinyurl.com/4f32f7rs)
v2 (chk): drop __GFP_RECLAIM since that is already set by GFP_USER
Cc: Dave Airlie <airlied@redhat.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240708160636.1147308-1-rajneesh.bhardwaj@amd.com Signed-off-by: Christian König <christian.koenig@amd.com>
Douglas Anderson [Fri, 21 Jun 2024 20:44:29 +0000 (13:44 -0700)]
drm/panel: Avoid warnings w/ panel-simple/panel-edp at shutdown
At shutdown if you've got a _properly_ coded DRM modeset driver then
you'll get these two warnings at shutdown time:
Skipping disable of already disabled panel
Skipping unprepare of already unprepared panel
These warnings are ugly and sound concerning, but they're actually a
sign of a properly working system. That's not great.
We're not ready to get rid of the calls to drm_panel_disable() and
drm_panel_unprepare() because we're not 100% convinced that all DRM
modeset drivers are properly calling drm_atomic_helper_shutdown() or
drm_helper_force_disable_all() at the right times. However, having the
warning show up for correctly working systems is bad.
As a bit of a workaround, add some "if" tests to try to avoid the
warning on correctly working systems. Also add some comments and
update the TODO items in the hopes that future developers won't be too
confused by what's going on here.
When the kmsg_dumper callback parameter changed, the reason variable
in mtdoops_do_dump() was not updated accordingly.
This breaks the build with mtdoops.
The added lvds driver and a change in the dsi driver resulted in failed
builds when COMMON_CLK is disabled:
x86_64-linux-ld: drivers/gpu/drm/stm/dw_mipi_dsi-stm.o: in function `dw_mipi_dsi_stm_remove':
dw_mipi_dsi-stm.c:(.text+0x51e): undefined reference to `clk_hw_unregister'
x86_64-linux-ld: drivers/gpu/drm/stm/lvds.o: in function `lvds_remove':
lvds.c:(.text+0xe3): undefined reference to `of_clk_del_provider'
x86_64-linux-ld: lvds.c:(.text+0xec): undefined reference to `clk_hw_unregister'
x86_64-linux-ld: drivers/gpu/drm/stm/lvds.o: in function `lvds_pll_config':
lvds.c:(.text+0xb5d): undefined reference to `clk_hw_get_rate'
x86_64-linux-ld: drivers/gpu/drm/stm/lvds.o: in function `lvds_probe':
lvds.c:(.text+0x1476): undefined reference to `clk_hw_register'
x86_64-linux-ld: lvds.c:(.text+0x148b): undefined reference to `of_clk_hw_simple_get'
x86_64-linux-ld: lvds.c:(.text+0x1493): undefined reference to `of_clk_add_hw_provider'
x86_64-linux-ld: lvds.c:(.text+0x1535): undefined reference to `clk_hw_unregister'
Add this as a dependency for the stm driver itself, since it will be
required in practice anyway.
drm/panic: Remove build time dependency with FRAMEBUFFER_CONSOLE
Now that fbcon has the skip_panic option, there is no more conflicts
between drm_panic and fbcon.
Remove the build time dependency, so they can be both enabled.
drm/fb-helper: Set skip_panic if the drm driver supports drm panic
fb_info->skip_panic flag prevents fbcon from writing to the
framebuffer during a kernel panic. So set it when the drm driver
implements drm panic, to avoid overwriting the panic screen.
This is required to avoid conflict between DRM_PANIC, and fbcon. If
a drm device already handle panic with drm_panic, it should set
the skip_panic field in fb_info, so that fbcon will stay quiet, and
not overwrite the panic_screen.
The reload interrupt is not used by the driver. To avoid
unnecessary calls of the interrupt routine, don't enable it.
Solve small typo and add mask to simplify the driver.
Implement struct drm_crtc_funcs.get_vblank_timestamp with the DRM
helper drm_crtc_vblank_helper_get_vblank_timestamp() with its helper
get_scanout_position. Read the scanout position from the MGAREG_VCOUNT
register.
Thomas Zimmermann [Thu, 18 Jul 2024 10:44:16 +0000 (12:44 +0200)]
drm/mgag200: Add vblank support
There's no VBLANK interrupt on Matrox chipsets. The workaround that is
being used here and in other free Matrox drivers is to program <linecomp>
to the value of <vblkstr> and enable the VLINE interrupt. This triggers
an interrupt at the time when VBLANK begins.
VLINE uses separate registers for enabling and clearing pending interrupts.
No extra synchronization between irq handler and the rest of the driver is
required.
v6:
- clear VLINE status bit before registering IRQ (Jocelyn)
v5:
- disable all interrupts before registering IRQ (Jocelyn)
- don't read from ICLEAR (Jocelyn)
v4:
- recreate patch on latest upstream
- use devm_request_irq() for managed cleanup
- fail if vblanking cannot be initialized
- rename register constants (Sam, Emil)
- clear interrupt before registering handler (Ville)
- move <linecomp> programming into separate commit
- set <linecomp> to <vblkstr>
- fix typo in commit message
v3:
- set <linecomp> to <vdisplay> + 1 to trigger at VBLANK
- expand comment on linecomp
v2:
- only signal vblank on CRTC 0
- use constants for registers and fields
- set VLINECLR before enabling interrupt
- test against STATUS and IEN in irq handler
- coding-style fixes
Thomas Zimmermann [Thu, 18 Jul 2024 10:44:14 +0000 (12:44 +0200)]
drm/mgag200: Add dedicated variables for blanking fields
Represent fields for horizontal and vertical blanking with <hblkstr>,
<hblkend>, <vblkstr> and <vblkend>. Aligns the code with the Matrox
programming manuals.
Thomas Zimmermann [Thu, 18 Jul 2024 10:44:13 +0000 (12:44 +0200)]
drm/mgag200: Use adjusted mode values for CRTCs
Use the values with the crtc_ prefix from struct drm_display_mode to
program hardware. The DRM core adjusted these values to the requirements
of CRTC hardware.
Thomas Zimmermann [Thu, 18 Jul 2024 10:44:11 +0000 (12:44 +0200)]
drm/mgag200: Use hexadecimal register indeces
In mgag200_set_mode_regs(), use hexadecimal indices for accessing
registers. Aligns the code with the register indices in the Matrox
programming manuals. Also convert to lower-case hexadecimal values.
commit 6a1688ae8794 ("drm/bridge: ptn3460: Convert to I2C driver model")
has dropped all the users of the struct bridge_init from the
exynos_dp_core, while retaining unused structure definition.
Later on the driver was reworked and the definition migrated
to the analogix_dp driver. Remove unused struct bridge_init definition.
Daniel Vetter [Wed, 10 Jul 2024 09:31:16 +0000 (11:31 +0200)]
drm: Add might_fault to drm_modeset_lock priming
We already teach lockdep that dma_resv nests within drm_modeset_lock,
but there's a lot more: All drm kms ioctl rely on being able to
put/get_user while holding modeset locks, so we really need a
might_fault in there too to complete the picture. Add it.
Motivated by a syzbot report that blew up on bcachefs doing an
unconditional console_lock way deep in the locking hierarchy, and
lockdep only noticing the depency loop in a drm ioctl instead of much
earlier. This annotation will make sure such issues have a much harder
time escaping.
References: https://lore.kernel.org/dri-devel/00000000000073db8b061cd43496@google.com/ Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Maxime Ripard <mripard@kernel.org> Cc: Thomas Zimmermann <tzimmermann@suse.de> Cc: Sumit Semwal <sumit.semwal@linaro.org> Cc: "Christian König" <christian.koenig@amd.com> Cc: linux-media@vger.kernel.org Cc: linaro-mm-sig@lists.linaro.org Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: https://patchwork.freedesktop.org/patch/msgid/20240710093120.732208-1-daniel.vetter@ffwll.ch
Alexander Stein [Wed, 3 Jul 2024 12:27:15 +0000 (14:27 +0200)]
drm/bridge: Silence error messages upon probe deferral
When -EPROBE_DEFER is returned do not raise an error, but silently return
this error instead. Fixes error like this:
[drm:drm_bridge_attach] *ERROR* failed to attach bridge
/soc@0/bus@30800000/mipi-dsi@30a00000 to encoder None-34: -517
[drm:drm_bridge_attach] *ERROR* failed to attach bridge
/soc@0/bus@30800000/mipi-dsi@30a00000 to encoder None-34: -517
printk: Add a short description string to kmsg_dump()
kmsg_dump doesn't forward the panic reason string to the kmsg_dumper
callback.
This patch adds a new struct kmsg_dump_detail, that will hold the
reason and description, and pass it to the dump() callback.
To avoid updating all kmsg_dump() call, it adds a kmsg_dump_desc()
function and a macro for backward compatibility.
I've written this for drm_panic, but it can be useful for other
kmsg_dumper.
It allows to see the panic reason, like "sysrq triggered crash"
or "VFS: Unable to mount root fs on xxxx" on the drm panic screen.
v2:
* Use a struct kmsg_dump_detail to hold the reason and description
pointer, for more flexibility if we want to add other parameters.
(Kees Cook)
* Fix powerpc/nvram_64 build, as I didn't update the forward
declaration of oops_to_nvram()
Marek Vasut [Fri, 31 May 2024 20:27:21 +0000 (22:27 +0200)]
drm: lcdif: Use adjusted_mode .clock instead of .crtc_clock
In case an upstream bridge modified the required clock frequency
in its .atomic_check callback by setting adjusted_mode.clock ,
make sure that clock frequency is generated by the LCDIFv3 block.
This is useful e.g. when LCDIFv3 feeds DSIM which feeds TC358767
with (e)DP output, where the TC358767 expects precise timing on
its input side, the precise timing must be generated by the LCDIF.
Alexander F. Lent [Tue, 9 Jul 2024 11:54:14 +0000 (07:54 -0400)]
accel/ivpu: Add missing MODULE_FIRMWARE metadata
Modules that load firmware from various paths at runtime must declare
those paths at compile time, via the MODULE_FIRMWARE macro, so that the
firmware paths are included in the module's metadata.
The accel/ivpu driver loads firmware but lacks this metadata,
preventing dracut from correctly locating firmware files. Fix it.
Fixes: 9ab43e95f922 ("accel/ivpu: Switch to generation based FW names") Fixes: 02d5b0aacd05 ("accel/ivpu: Implement firmware parsing and booting") Signed-off-by: Alexander F. Lent <lx@xanderlent.com> Reviewed-by: Jacek Lawrynowicz <jacek.lawrynowicz@linux.intel.com> Signed-off-by: Jacek Lawrynowicz <jacek.lawrynowicz@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240709-fix-ivpu-firmware-metadata-v3-1-55f70bba055b@xanderlent.com
Katya Orlova [Fri, 16 Feb 2024 12:50:40 +0000 (15:50 +0300)]
drm/stm: Avoid use-after-free issues with crtc and plane
ltdc_load() calls functions drm_crtc_init_with_planes(),
drm_universal_plane_init() and drm_encoder_init(). These functions
should not be called with parameters allocated with devm_kzalloc()
to avoid use-after-free issues [1].
Use allocations managed by the DRM framework.
Found by Linux Verification Center (linuxtesting.org).
Thomas Zimmermann [Thu, 11 Jul 2024 07:23:05 +0000 (09:23 +0200)]
drm/mgag200: Remove vidrst callbacks from struct mgag200_device_funcs
The callbacks disable_vidrst and enable_vidrst are obsolete. Remove
the fields from struct mgag200_device_funcs. Instead call their
implementations directly of the field 'has_vidrst' has been set in
struct mgag200_device_info.
Also change the logic slightly. The BMC used to start and stop scanout
during the CRTC's atomic_enable and atomic_disable. Plane updates were
done while the BMC scanned out the display. Now only stop once in
atomic_disable at the beginning of a modeset and then restart the
scanout at the end of a modeset in atomic_enable. While the modeset
takes place, the BMC does not scanout at all.
Thomas Zimmermann [Thu, 11 Jul 2024 07:23:04 +0000 (09:23 +0200)]
drm/mgag200: Only set VIDRST bits in CRTC modesetting
The VRSTEN and HRSTEN bits control whether a CRTC synchronizes its
display signal with an external source on the VIDRST pin. The G200WB
and G200EW3 models synchronize with a BMC chip, but different external
video encoders, such as the Matrox Maven, can also be attached to the
pin.
Only set VRSTEN and HRSTEN bits in the CRTC mode-setting code, so the
bits are independent from the BMC. Add the field set_vidrst to the CRTC
state for this purpose. Off by default, control the CRTC VIDRST setting
from the CRTC's atomic_check helper.
Use the common DRM function `drm_show_memory_stats()` to expose standard
fdinfo memory stats.
V3D exposes global GPU memory stats through debugfs. Those stats will be
preserved while the DRM subsystem doesn't have a standard solution to
expose global GPU stats.
drm/v3d: Move perfmon init completely into own unit
Now that the build time dependencies on various array sizes have been
removed, we can move the perfmon init completely into its own compilation
unit and remove the hardcoded defines.
This improves on the temporary fix quickly delivered in commit 9c3951ec27b9 ("drm/v3d: Fix perfmon build error/warning").
Instead of statically reserving pessimistic space for the kperfmon_ids
array, make the userspace extension code allocate the exactly required
amount of space.
Apart from saving some memory at runtime, this also removes the need for
the V3D_MAX_PERFMONS macro whose removal will benefit further driver
cleanup.
drm/v3d: Fix potential memory leak in the performance extension
If fetching of userspace memory fails during the main loop, all drm sync
objs looked up until that point will be leaked because of the missing
drm_syncobj_put.
Fix it by exporting and using a common cleanup helper.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com> Fixes: bae7cb5d6800 ("drm/v3d: Create a CPU job extension for the reset performance query job") Cc: Maíra Canal <mcanal@igalia.com> Cc: Iago Toral Quiroga <itoral@igalia.com> Cc: stable@vger.kernel.org # v6.8+ Signed-off-by: Maíra Canal <mcanal@igalia.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-4-tursulin@igalia.com
drm/v3d: Fix potential memory leak in the timestamp extension
If fetching of userspace memory fails during the main loop, all drm sync
objs looked up until that point will be leaked because of the missing
drm_syncobj_put.
Fix it by exporting and using a common cleanup helper.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com> Fixes: 9ba0ff3e083f ("drm/v3d: Create a CPU job extension for the timestamp query job") Cc: Maíra Canal <mcanal@igalia.com> Cc: Iago Toral Quiroga <itoral@igalia.com> Cc: stable@vger.kernel.org # v6.8+ Reviewed-by: Maíra Canal <mcanal@igalia.com> Signed-off-by: Maíra Canal <mcanal@igalia.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-3-tursulin@igalia.com
drm/v3d: Prevent out of bounds access in performance query extensions
Check that the number of perfmons userspace is passing in the copy and
reset extensions is not greater than the internal kernel storage where
the ids will be copied into.
Philipp Stanner [Fri, 12 Jul 2024 06:26:18 +0000 (08:26 +0200)]
drm/nouveau: Improve variable name in nouveau_sched_init()
nouveau_sched_init() uses the function drm_sched_init(). The latter
function has parameters called "hang_limit" and "timeout" in its API
documentation.
nouveau_sched_init(), however, defines a variable called
"job_hang_limit" which is passed to drm_sched_init()'s "timeout"
parameter. The actual "hang_limit" parameter is directly set to 0.
Mario Limonciello [Wed, 3 Jul 2024 05:17:22 +0000 (00:17 -0500)]
drm/amd: Add power_saving_policy drm property to eDP connectors
When the `power_saving_policy` property is set to bit mask
"Require color accuracy" ABM should be disabled immediately and
any requests by sysfs to update will return an -EBUSY error.
When the `power_saving_policy` property is set to bit mask
"Require low latency" PSR should be disabled.
When the property is restored to an empty bit mask ABM and PSR
can be enabled again.
Mario Limonciello [Wed, 3 Jul 2024 05:17:21 +0000 (00:17 -0500)]
drm: Introduce 'power saving policy' drm property
The `power saving policy` DRM property is an optional property that
can be added to a connector by a driver.
This property is for compositors to indicate intent of policy of
whether a driver can use power saving features that may compromise
the experience intended by the compositor.
Uprev IGT to the latest version, which includes a fix for the
writeback tests issue on MSM devices. Enable debugging for
igt-runner to log output such as 'Begin test' and 'End test'.
This will help identify which test causes system freeze or hangs.
Update xfails and add metadata header for each flake test.
drm/panel: boe-th101mb31ig002: Support for starry-er88577 MIPI-DSI panel
The init_code of the starry-er88577 panel is very similar to the
panel-boe-th101mb31ig002-28a.c driver, so we make it compatible with
the panel-boe-th101mb31ig002-28a.c driver
dt-bindings: display: panel: Add compatible for starry-er88577
The starry-er88577 is a 10.1" WXGA TFT-LCD panel, and the init_code
of the starry-er88577 panel is very similar to the boe-th101mb31ig002
panel, so We will add a new configuration based on
"boe,th101mb31ig002-28a.yaml".
Because the panel used reset gpio before but did not add the definition
of "reset gpio" in binding, reset gpio was added in binding, but since
the starry-er88577 panel did not use "reset gpio", a judgment was added
here.
Cong Yang [Thu, 4 Jul 2024 04:50:16 +0000 (12:50 +0800)]
drm/panel: jd9365da: Support for Melfas lmfbx101117480 MIPI-DSI panel
The Melfas lmfbx101117480 is a 10.1" WXGA TFT-LCD panel, use jd9365da
controller, which fits in nicely with the existing panel-jadard-jd9365da-h3
driver. Hence, we add a new compatible with panel specific config.
drm/panel: himax-hx8394: Add Support for Microchip AC40T08A MIPI Display Panel
Add support for the Microchip AC40T08A MIPI Display panel. This panel uses
a Himax HX8394 display controller and requires a vendor provided init
sequence. The display resolution is 720x1280@60Hz with width and height
of 76mm and 132mm respectively.
Add compatible string for the Microchip's AC40T08A MIPI Display
panel.This panel uses a Himax HX8394 display controller.
The reset line is not populated and leads to driver probe issues,
thus add conditional block to narrow reset-gpio property per variant.
Thomas Hellström [Fri, 5 Jul 2024 15:32:05 +0000 (17:32 +0200)]
drm/ttm: Use the LRU walker for eviction
Use the LRU walker for eviction. This helps
removing a lot of code with weird locking
semantics.
The functionality is slightly changed so that
when trylocked buffer objects are exhausted, we
continue to interleave walks with ticket-locks while
there is still progress made. The list walks are
not restarted in-between evictions.
Also provide a separate ttm_bo_evict_first()
function for its single user. The context of that
user allows sleeping dma_resv locks.
v6:
- Various cleanups suggested by Matthew Brost.
- Fix error return code of ttm_bo_evict_first(). (Matthew Brost)
- Fix an error check that was inverted. (Matthew Brost)
v7:
- Use s64 rather than long (Christian König)
- Early ttm_resource_cursor_fini() in ttm_bo_evict_first().
- Simplify check for bo_moved in ttm_bo_evict_first().
(Christian König)
- Don't evict pinned bos.
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> #v6 Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-8-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Thomas Hellström [Fri, 5 Jul 2024 15:32:04 +0000 (17:32 +0200)]
drm/ttm: Use the LRU walker helper for swapping
Rework the TTM swapping to use the LRU walker helper.
This helps fixing up the ttm_bo_swapout() interface
to be consistent about not requiring any locking.
For now mimic the current behaviour of using trylock
only. We could be using ticket-locks here but defer
that until it's deemed necessary. The TTM swapout
functionality is a bit weird anyway since it
alternates between memory types without exhausting
TTM_PL_SYSTEM first.
Intentionally keep pages as the unit of progress since
changing that to bytes is an unrelated change that can
be done later.
v6:
- Improve on error code translation in the swapout callback
(Matthew Brost).
v7:
- Use s64 rather than long.
- Remove ttm_resource_cursor_fini() since it's no longer used.
- Rename ttm_resource_cursor_fini_locked() to
ttm_resource_cursor_fini().
- Don't swap out pinned bos.
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> #v6 Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-7-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Thomas Hellström [Fri, 5 Jul 2024 15:32:03 +0000 (17:32 +0200)]
drm/ttm: Provide a generic LRU walker helper
Provide a generic LRU walker in TTM, in the spirit of drm_gem_lru_scan()
but building on the restartable TTM LRU functionality.
The LRU walker optionally supports locking objects as part of
a ww mutex locking transaction, to mimic to some extent the
current functionality in ttm. However any -EDEADLK return
is converted to -ENOSPC and then to -ENOMEM before reaching
the driver, so that the driver will need to backoff and possibly retry
without being able to keep the ticket.
v3:
- Move the helper to core ttm.
- Remove the drm_exec usage from it for now, it will be
reintroduced later in the series.
v4:
- Handle the -EALREADY case if ticketlocking.
v6:
- Some cleanup and added code comments (Matthew Brost)
- Clarified the ticketlock in the commit message (Matthew Brost)
v7:
- Use s64 rather than long for the target and progress
(Christian König)
- Update documentation to not encourage using pages as a
progress measure. (Christian König)
- Remove cond_resched(). (Christian König)
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> #v6 Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-6-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Thomas Hellström [Fri, 5 Jul 2024 15:32:02 +0000 (17:32 +0200)]
drm/ttm, drm/amdgpu, drm/xe: Consider hitch moves within bulk sublist moves
To address the problem with hitches moving when bulk move
sublists are lru-bumped, register the list cursors with the
ttm_lru_bulk_move structure when traversing its list, and
when lru-bumping the list, move the cursor hitch to the tail.
This also means it's mandatory for drivers to call
ttm_lru_bulk_move_init() and ttm_lru_bulk_move_fini() when
initializing and finalizing the bulk move structure, so add
those calls to the amdgpu- and xe driver.
Compared to v1 this is slightly more code but less fragile
and hopefully easier to understand.
Changes in previous series:
- Completely rework the functionality
- Avoid a NULL pointer dereference assigning manager->mem_type
- Remove some leftover code causing build problems
v2:
- For hitch bulk tail moves, store the mem_type in the cursor
instead of with the manager.
v3:
- Remove leftover mem_type member from change in v2.
v6:
- Add some lockdep asserts (Matthew Brost)
- Avoid NULL pointer dereference (Matthew Brost)
- No need to check bo->resource before dereferencing
bo->bulk_move (Matthew Brost)
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Acked-by: Christian König <christian.koenig@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-5-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Thomas Hellström [Fri, 5 Jul 2024 15:32:01 +0000 (17:32 +0200)]
drm/ttm: Use LRU hitches
Have iterators insert themselves into the list they are iterating
over using hitch list nodes. Since only the iterator owner
can remove these list nodes from the list, it's safe to unlock
the list and when continuing, use them as a starting point. Due to
the way LRU bumping works in TTM, newly added items will not be
missed, and bumped items will be iterated over a second time before
reaching the end of the list.
The exception is list with bulk move sublists. When bumping a
sublist, a hitch that is part of that sublist will also be moved
and we might miss items if restarting from it. This will be
addressed in a later patch.
Changes in previous series:
- Updated ttm_resource_cursor_fini() documentation.
v2:
- Don't reorder ttm_resource_manager_first() and _next().
(Christian König).
- Use list_add instead of list_move
(Christian König)
v3:
- Split into two patches, one cleanup, one new functionality
(Christian König)
- use ttm_resource_cursor_fini_locked() instead of open-coding
(Matthew Brost)
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Christian König <christian.koenig@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-4-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Thomas Hellström [Fri, 5 Jul 2024 15:32:00 +0000 (17:32 +0200)]
drm/ttm: Slightly clean up LRU list iteration
To make the transition to using lru hitches easier,
simplify the ttm_resource_manager_next() interface to only take
the cursor and reuse ttm_resource_manager_next() functionality
from ttm_resource_manager_first().
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Christian König <christian.koenig@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-3-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Thomas Hellström [Fri, 5 Jul 2024 15:31:59 +0000 (17:31 +0200)]
drm/ttm: Allow TTM LRU list nodes of different types
To be able to handle list unlocking while traversing the LRU
list, we want the iterators not only to point to the next
position of the list traversal, but to insert themselves as
list nodes at that point to work around the fact that the
next node might otherwise disappear from the list while
the iterator is pointing to it.
These list nodes need to be easily distinguishable from other
list nodes so that others traversing the list can skip
over them.
So declare a struct ttm_lru_item, with a struct list_head member
and a type enum. This will slightly increase the size of a
struct ttm_resource.
Changes in previous series:
- Update enum ttm_lru_item_type documentation.
v3:
- Introduce ttm_lru_first_res_or_null()
(Christian König, Thomas Hellström)
v5:
- Update also the TTM test code (Xe CI).
Cc: Christian König <christian.koenig@amd.com> Cc: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <dri-devel@lists.freedesktop.org> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Christian König <christian.koenig@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240705153206.68526-2-thomas.hellstrom@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
The initial idea of the Colorspace prop was that this maps 1:1 to
InfoFrames/SDP but KMS does not give user space enough information nor
control over the output format to figure out which variants can be used
for a given KMS commit. At the same time, properties like Broadcast RGB
expect full range quantization range being produced by user space from
the CRTC and drivers to convert to the range expected by the sink for
the chosen output format, mode, InfoFrames, etc.
This change documents the reality of the Colorspace property. The
Default variant unfortunately is very much driver specific and not
reflected by the EDID. The BT2020 variants are in active use by generic
compositors which have expectations from the driver about the
conversions it has to do when selecting certain output formats.
Everything else is also marked as undefined. Coming up with valid
behavior that makes it usable from user space and consistent with other
KMS properties for those variants is left as an exercise for whoever
wants to use them.
v2:
* Talk about "pixel operation properties" that user space configures
* Mention that user space is responsible for checking the EDID for sink
support
* Make it clear that drivers can choose between RGB and YCbCr on their
own
projects / users / jedix / linux-maple.git / log