Maxime Ripard [Thu, 22 Feb 2024 18:14:21 +0000 (19:14 +0100)]
drm/sun4i: hdmi: Consolidate atomic_check and mode_valid
atomic_check and mode_valid do not check for the same things which can
lead to surprising result if the userspace commits a mode that didn't go
through mode_valid. Let's merge the two implementations into a function
called by both.
Tomi Valkeinen [Tue, 13 Feb 2024 08:16:37 +0000 (10:16 +0200)]
drm/tidss: Fix sync-lost issue with two displays
A sync lost issue can be observed with two displays, when moving a plane
from one disabled display to an another disabled display, and then
enabling the display to which the plane was moved to. The exact
requirements for this to trigger are not clear.
It looks like the issue is that the layers are left enabled in the first
display's OVR registers. Even if the corresponding VP is disabled, it
still causes an issue, as if the disabled VP and its OVR would still be
in use, leading to the same VID being used by two OVRs. However, this is
just speculation based on testing the DSS behavior.
Experimentation shows that as a workaround, we can disable all the
layers in the OVR when disabling a VP. There should be no downside to
this, as the OVR is anyway effectively disabled if its VP is disabled,
and it seems to solve the sync lost issue.
However, there may be a bigger issue in play here, related to J721e
erratum i2097 ("DSS: Disabling a Layer Connected to Overlay May Result
in Synclost During the Next Frame"). Experimentation also shows that the
OVR's CHANNELIN field has similar issue. So we may need to revisit this
when we find out more about the core issue.
Tomi Valkeinen [Tue, 13 Feb 2024 08:16:36 +0000 (10:16 +0200)]
drm/tidss: Fix initial plane zpos values
When the driver sets up the zpos property it sets the default zpos value
to the HW id of the plane. That is fine as such, but as on many DSS
versions the driver arranges the DRM planes in a different order than
the HW planes (to keep the non-scalable planes first), this leads to odd
initial zpos values. An example is J721e, where the initial zpos values
for DRM planes are 1, 3, 0, 2.
In theory the userspace should configure the zpos values properly when
using multiple planes, and in that sense the initial zpos values
shouldn't matter, but there's really no reason not to fix this and help
the userspace apps which don't handle zpos perfectly. In particular,
some versions of Weston seem to have issues dealing with the planes
with the current default zpos values.
So let's change the zpos values for the DRM planes to 0, 1, 2, 3.
Another option would be to configure the planes marked as primary planes
to zpos 0. On a two display system this would give us plane zpos values
of 0, 0, 1, 2. The end result and behavior would be very similar in this
option, and I'm not aware that this would actually help us in any way.
So, to keep the code simple, I opted for the 0, 1, 2, 3 values.
Rodrigo Vivi [Fri, 23 Feb 2024 19:15:48 +0000 (14:15 -0500)]
drm/i915: convert remaining intel_dp_vsc_sdp_pack
Commit 47f419e07111 ("drm/dp: move intel_dp_vsc_sdp_pack() to generic helper")
and commit b55b88d86fec ("drm/dp: drop the size parameter from drm_dp_vsc_sdp_pack()")
were based on top of a tree containing the
commit 31a5b6ed88c7 ("drm/i915/display: Unify VSC SPD preparation") but
landed in a tree where this commit didn't exist, leaving behind a spurious
case calling for a removed function: intel_dp_vsc_sdp_pack()
Let's convert the remaining case here so we can port this patch to
any tree that doesn't contain
commit 31a5b6ed88c7 ("drm/i915/display: Unify VSC SPD preparation")
In in kind of merge where this commit does exist, this line here will
be gone anyway and not needed any longer.
Maíra Canal [Wed, 14 Feb 2024 19:34:52 +0000 (16:34 -0300)]
drm/v3d: Enable V3D to use different PAGE_SIZE
Currently, the V3D driver uses PAGE_SHIFT over the assumption that
PAGE_SHIFT = 12, as the PAGE_SIZE = 4KB. But, the RPi 5 is using
PAGE_SIZE = 16KB, so the MMU PAGE_SHIFT is different than the system's
PAGE_SHIFT.
Enable V3D to be used in system's with any PAGE_SIZE by making sure that
everything MMU-related uses the MMU page shift.
Paloma Arellano [Thu, 15 Feb 2024 19:15:56 +0000 (11:15 -0800)]
drm/dp: add an API to indicate if sink supports VSC SDP
YUV420 format is supported only in the VSC SDP packet and not through
MSA. Hence add an API which indicates the sink support which can be used
by the rest of the DP programming.
changes in v5:
- rebased on top of drm-tip
changes in v4:
- bail out early if dpcd rev check fails
changes in v3:
- fix the commit title prefix to drm/dp
- get rid of redundant !!
- break out this change from series [1] to get acks from drm core
maintainers
Changes in v2:
- Move VSC SDP support check API from dp_panel.c to
drm_dp_helper.c
Fedor Pchelkin [Fri, 15 Dec 2023 09:33:55 +0000 (12:33 +0300)]
drm/tegra: put drm_gem_object ref on error in tegra_fb_create
Inside tegra_fb_create(), drm_gem_object_lookup() increments ref count of
the found object. But if the following size check fails then the last
found object's ref count should be put there as the unreferencing loop
can't detect this situation.
Found by Linux Verification Center (linuxtesting.org).
Biju Das [Thu, 22 Feb 2024 11:26:43 +0000 (11:26 +0000)]
drm: renesas: rz-du: Fix redefinition errors related to rzg2l_du_vsp_*()
CONFIG_DRM_RCAR_VSP is related to R-Car and not related to RZ/G2L
specific. This caused the below build issue. Fix this errors by replacing
CONFIG_DRM_RCAR_VSP->CONFIG_VIDEO_RENESAS_VSP1. While at it, use
IS_ENABLED() macro.
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.c:47:6: error: redefinition of 'rzg2l_du_vsp_enable'
47 | void rzg2l_du_vsp_enable(struct rzg2l_du_crtc *crtc)
| ^~~~~~~~~~~~~~~~~~~
In file included from drivers/gpu/drm/renesas/rz-du/rzg2l_du_drv.h:18,
from drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.c:30:
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.h:72:20: note: previous definition of 'rzg2l_du_vsp_enable' with type 'void(struct rzg2l_du_crtc *)'
72 | static inline void rzg2l_du_vsp_enable(struct rzg2l_du_crtc *crtc) { };
| ^~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.c:61:6: error: redefinition of 'rzg2l_du_vsp_disable'
61 | void rzg2l_du_vsp_disable(struct rzg2l_du_crtc *crtc)
| ^~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.h:73:20: note: previous definition of 'rzg2l_du_vsp_disable' with type 'void(struct rzg2l_du_crtc *)'
73 | static inline void rzg2l_du_vsp_disable(struct rzg2l_du_crtc *crtc) { };
| ^~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.c:66:6: error: redefinition of 'rzg2l_du_vsp_atomic_flush'
66 | void rzg2l_du_vsp_atomic_flush(struct rzg2l_du_crtc *crtc)
| ^~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.h:74:20: note: previous definition of 'rzg2l_du_vsp_atomic_flush' with type 'void(struct rzg2l_du_crtc *)'
74 | static inline void rzg2l_du_vsp_atomic_flush(struct rzg2l_du_crtc *crtc) { };
| ^~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.c:76:19: error: redefinition of 'rzg2l_du_vsp_get_drm_plane'
76 | struct drm_plane *rzg2l_du_vsp_get_drm_plane(struct rzg2l_du_crtc *crtc,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.h:75:33: note: previous definition of 'rzg2l_du_vsp_get_drm_plane' with type 'struct drm_plane *(struct rzg2l_du_crtc *, unsigned int)'
75 | static inline struct drm_plane *rzg2l_du_vsp_get_drm_plane(struct rzg2l_du_crtc *crtc,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.c:302:5: error: redefinition of 'rzg2l_du_vsp_init'
302 | int rzg2l_du_vsp_init(struct rzg2l_du_vsp *vsp, struct device_node *np,
| ^~~~~~~~~~~~~~~~~
drivers/gpu/drm/renesas/rz-du/rzg2l_du_vsp.h:66:19: note: previous definition of 'rzg2l_du_vsp_init' with type 'int(struct rzg2l_du_vsp *, struct device_node *, unsigned int)'
66 | static inline int rzg2l_du_vsp_init(struct rzg2l_du_vsp *vsp, struct device_node *np,
| ^~~~~~~~~~~~~~~~~
Erik Kurzinger [Fri, 19 Jan 2024 16:32:08 +0000 (08:32 -0800)]
drm/syncobj: call might_sleep before waiting for fence submission
If either the DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT or
DRM_SYNCOBJ_WAIT_FLAGS_WAIT_AVAILABLE flags are passed to
drm_syncobj_array_wait_timeout, the function might sleep if the fence at
one of the given timeline points has not yet been submitted. Therefore,
we should call might_sleep in that case to catch potential bugs.
Erik Kurzinger [Fri, 19 Jan 2024 16:32:07 +0000 (08:32 -0800)]
drm/syncobj: reject invalid flags in drm_syncobj_find_fence
The only flag that is meaningful to drm_syncobj_find_fence is
DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT. It should return -EINVAL for any
other flag bits.
Erik Kurzinger [Fri, 19 Jan 2024 16:32:06 +0000 (08:32 -0800)]
drm/syncobj: call drm_syncobj_fence_add_wait when WAIT_AVAILABLE flag is set
When waiting for a syncobj timeline point whose fence has not yet been
submitted with the WAIT_FOR_SUBMIT flag, a callback is registered using
drm_syncobj_fence_add_wait and the thread is put to sleep until the
timeout expires. If the fence is submitted before then,
drm_syncobj_add_point will wake up the sleeping thread immediately which
will proceed to wait for the fence to be signaled.
However, if the WAIT_AVAILABLE flag is used instead,
drm_syncobj_fence_add_wait won't get called, meaning the waiting thread
will always sleep for the full timeout duration, even if the fence gets
submitted earlier. If it turns out that the fence *has* been submitted
by the time it eventually wakes up, it will still indicate to userspace
that the wait completed successfully (it won't return -ETIME), but it
will have taken much longer than it should have.
To fix this, we must call drm_syncobj_fence_add_wait if *either* the
WAIT_FOR_SUBMIT flag or the WAIT_AVAILABLE flag is set. The only
difference being that with WAIT_FOR_SUBMIT we will also wait for the
fence to be signaled after it has been submitted while with
WAIT_AVAILABLE we will return immediately.
IGT test patch: https://lists.freedesktop.org/archives/igt-dev/2024-January/067537.html
Biju Das [Sun, 18 Feb 2024 16:48:38 +0000 (16:48 +0000)]
drm: renesas: Add RZ/G2L DU Support
The LCD controller is composed of Frame Compression Processor (FCPVD),
Video Signal Processor (VSPD), and Display Unit (DU).
It has DPI/DSI interfaces and supports a maximum resolution of 1080p
along with 2 RPFs to support the blending of two picture layers and
raster operations (ROPs).
The DU module is connected to VSPD. Add RZ/G2L DU support for RZ/G2L
alike SoCs.
Biju Das [Sun, 18 Feb 2024 16:48:36 +0000 (16:48 +0000)]
dt-bindings: display: Document Renesas RZ/G2L DU bindings
The RZ/G2L LCD controller is composed of Frame Compression Processor
(FCPVD), Video Signal Processor (VSPD), and Display Unit (DU).
The DU module supports the following hardware features
− Display Parallel Interface (DPI) and MIPI LINK Video Interface
− Display timing master
− Generates video timings
− Selecting the polarity of output DCLK, HSYNC, VSYNC, and DE
− Supports Progressive
− Input data format (from VSPD): RGB888, RGB666
− Output data format: same as Input data format
− Supporting Full HD (1920 pixels x 1080 lines) for MIPI-DSI Output
− Supporting WXGA (1280 pixels x 800 lines) for Parallel Output
This patch documents the DU module found on RZ/G2L LCDC.
Dmitry Baryshkov [Sun, 18 Feb 2024 04:12:04 +0000 (06:12 +0200)]
drm/ci: mark universal-plane-sanity as failing on SC7180
The test kms_universal_plane@universal-plane-sanity fails on both SC7180
platforms. The drm/msm returns -ERANGE as it can not handle passet
scaling range, however the test is not ready to handle that. Mark the
test as failing until it is fixed.
ERROR - Igt error: (kms_universal_plane:1554) CRITICAL: Test assertion failure function sanity_test_pipe, file ../tests/kms_universal_plane.c:438:
ERROR - Igt error: (kms_universal_plane:1554) CRITICAL: Failed assertion: drmModeSetPlane(data->drm_fd, primary->drm_plane->plane_id, output->config.crtc->crtc_id, test.oversized_fb.fb_id, 0, 0, 0, mode->hdisplay + 100, mode->vdisplay + 100, IGT_FIXED(0,0), IGT_FIXED(0,0), IGT_FIXED(mode->hdisplay,0), IGT_FIXED(mode->vdisplay,0)) == expect
ERROR - Igt error: (kms_universal_plane:1554) CRITICAL: Last errno: 34, Numerical result out of range
Dmitry Baryshkov [Sun, 18 Feb 2024 04:12:02 +0000 (06:12 +0200)]
drm/ci: update IGT test names
Since the addition of testlist.txt the IGT has changed some of test
names. Some test names were changed to use '-' instead of '_'. In other
cases tests were just renamed. Follow those changes.
Dmitry Baryshkov [Fri, 16 Feb 2024 20:56:50 +0000 (22:56 +0200)]
drm/ci: skip suspend tests for both msm-sc7180 machines
The commit ea489a3d983b ("drm/ci: add sc7180-trogdor-kingoftown")
dropped the msm-sc7180-skips.txt file, which disabled suspend-to-RAM
tests. However testing shows that STR tests still can fail. Restore the
skiplist, applying it to both limozeen and kingoftown machines.
Devarsh Thakkar [Fri, 16 Feb 2024 06:24:23 +0000 (11:54 +0530)]
dt-bindings: display: ti,am65x-dss: Add support for common1 region
TI keystone display subsystem present in AM65, AM62 and AM62A SoC support
two separate register spaces namely "common" and "common1" which can be
used by two separate hosts to program the display controller as described
in respective Technical Reference Manuals [1].
The common1 register space has similar set of configuration registers as
supported in common register space except the global configuration
registers which are exclusive to common region.
This adds binding for "common1" register region too as supported by the
hardware.
Wachowski, Karol [Wed, 14 Feb 2024 08:13:04 +0000 (09:13 +0100)]
accel/ivpu: Refactor BO creation functions
Rename BO allocate/create functions, so the code is more consistent.
There are now two matching buffer creation functions:
- ivpu_bo_create_ioctl() - create a BO from user space
- ivpu_bo_create() - create a BO from kernel space
ivpu_bo_alloc() is now only used to allocate struct ivpu_bo which better
matches its name.
Wachowski, Karol [Wed, 14 Feb 2024 08:13:02 +0000 (09:13 +0100)]
accel/ivpu: Use lazy allocation for doorbell IDs
Reserve/allocate and free doorbells for command queues when needed
using xarray. This allows to avoid reserving a doorbell for
a contexts that never issues a job.
accel/ivpu: Add support for FW boot param system_time_us
Add support for FW boot API param system_time_us.
According to the API description this field should
be set to system time in microseconds starting from 1970.
Douglas Anderson [Fri, 16 Feb 2024 20:31:12 +0000 (12:31 -0800)]
drm/panel: boe-tv101wum-nl6: make use of prepare_prev_first
The panel on sc7180-trogdor-wormdingler and
sc7180-trogdor-quackingstick hasn't been coming up since commit 9e15123eca79 ("drm/msm/dsi: Stop unconditionally powering up DSI hosts
at modeset"). Let's add "prepare_prev_first" as has been done for many
other DSI panels.
Rename meson_encoder_{cvbs,dsi,hdmi}_init() to
meson_encoder_{cvbs,dsi,hdmi}_probe() so it's clear that these functions
are used at probe time during driver initialization. Also switch all
error prints inside those functions to use dev_err_probe() for
consistency.
This makes the code more straight forward to read and makes the error
prints within those functions consistent (by logging all -EPROBE_DEFER
with dev_dbg(), while actual errors are logged with dev_err() and get
the error value printed).
Heiko Stuebner [Thu, 15 Feb 2024 09:05:14 +0000 (10:05 +0100)]
dt-bindings: display: ltk500hd1829: add variant compatible for ltk101b4029w
Add the compatible for the ltk101b4029w panel, that has the same
manufacturer, general bringup and supplies but a different dsi-init-
sequence compared to the ltk500hd1829 .
Colin Ian King [Thu, 15 Feb 2024 22:43:01 +0000 (22:43 +0000)]
gpu: host1x: remove redundant assignment to variable space
The variable space is being initialized with a value that is never read,
it is being re-assigned later on. The initialization is redundant and
can be removed. Also merge two declaration lines together.
Cleans up clang scan build warning:
drivers/gpu/host1x/cdma.c:628:15: warning: Value stored to 'space'
during its initialization is never read [deadcode.DeadStores]
The overridden mode fixes the panel glitching issue on mt8186 chromebook.
However, it causes the internal display not working on mt8173 chromebook.
Revert the overridden mode for now to let mt8173 have a functional display.
Dmitry Baryshkov [Wed, 14 Feb 2024 08:37:08 +0000 (10:37 +0200)]
drm: ci: use clk_ignore_unused for apq8016
If the ADV7511 bridge driver is compiled as a module, while DRM_MSM is
built-in, the clk_disable_unused congests with the runtime PM handling
of the DSI PHY for the clk_prepare_lock(). This causes apq8016 runner to
fail without completing any jobs ([1]). Drop the BM_CMDLINE which
duplicate the command line from the .baremetal-igt-arm64 clause and
enforce the clk_ignore_unused kernelarg instead to make apq8016 runner
work.
If the firmware framebuffer has been reloacted, the sysfb code
fixes the screen_info state before it creates the framebuffer's
platform device. Efifb will automatically receive a screen_info
with updated values. Hence remove the tracking from efifb.
firmware/sysfb: Update screen_info for relocated EFI framebuffers
On ARM PCI systems, the PCI hierarchy might be reconfigured during
boot and the firmware framebuffer might move as a result of that.
The values in screen_info will then be invalid.
Work around this problem by tracking the framebuffer's initial
location before it get relocated; then fix the screen_info state
between reloaction and creating the firmware framebuffer's device.
This functionality has been lifted from efifb. See the commit message
of commit 55d728a40d36 ("efi/fb: Avoid reconfiguration of BAR that
covers the framebuffer") for more information.
There will be no EFI framebuffer device for disabled parent devices
and thus we never probe efifb in that case. Hence remove the tracking
code from efifb.
firmware/sysfb: Create firmware device only for enabled PCI devices
Test if the firmware framebuffer's parent PCI device, if any, has
been enabled. If not, the firmware framebuffer is most likely not
working. Hence, do not create a device for the firmware framebuffer
on disabled PCI devices.
So far, efifb tracked the status of the PCI parent device internally
and did not bind if it was disabled. This patch implements the
functionality for all PCI-based firmware framebuffers.
v3:
* make commit message more precise (Sui)
v2:
* rework sysfb_pci_dev_is_enabled() (Javier)
The EFI device has the correct parent device set. This allows Linux
to handle the power management internally. Hence, remove the manual
PM management for the parent device from efifb.
firmware/sysfb: Set firmware-framebuffer parent device
Set the firmware framebuffer's parent device, which usually is the
graphics hardware's physical device. Integrates the framebuffer in
the Linux device hierarchy and lets Linux handle dependencies among
devices. For example, the graphics hardware won't be suspended while
the firmware device is still active.
v4:
* fix build for CONFIG_SYSFB_SIMPLEFB=n, again
v3:
* fix build for CONFIG_SYSFB_SIMPLEFB=n (Sui)
* test result of screen_info_pci_dev() for errors (Sui)
v2:
* detect parent device in sysfb_parent_dev()
The plain values as stored in struct screen_info need to be decoded
before being used. Add helpers that decode the type of video output
and the framebuffer I/O aperture.
Old or non-x86 systems may not set the type of video directly, but
only indicate the presence by storing 0x01 in orig_video_isVGA. The
decoding logic in screen_info_video_type() takes this into account.
It then follows similar code in vgacon's vgacon_startup() to detect
the video type from the given values.
A call to screen_info_resources() returns all known resources of the
given screen_info. The resources' values have been taken from existing
code in vgacon and vga16fb. These drivers can later be converted to
use the new interfaces.
Most Rockchip hdmi nodes are part of a power domain.
Add a power-domains property and include it to the example
with some reordering to align with the (new) documentation
about property ordering.
Johan Jonker [Wed, 31 Jan 2024 21:14:29 +0000 (22:14 +0100)]
dt-bindings: display: rockchip: rockchip,dw-hdmi: remove port property
The hdmi-connector nodes are now functional and the new way to model
hdmi ports nodes with both in and output port subnodes. Unfortunately
with the conversion to YAML the old method with only an input port node
was used. Later the new method was also added to the binding.
A binding must be unambiguously, so remove the old port property
entirely and make port@0 and port@1 a requirement as all
upstream dts files are updated as well and because checking
deprecated stuff is a bit pointless.
Update the example to avoid use of the removed property.
Chris Morgan [Mon, 12 Feb 2024 18:49:45 +0000 (12:49 -0600)]
drm/panel: st7703: Add Powkiddy RGB10MAX3 Panel Support
The Powkiddy RGB10MAX3 is a handheld device with a 5 inch 720x1280
display panel with a Sitronix ST7703 display controller. The panel
is installed rotated 270 degrees.
Since commit d492cc2573a0 ("driver core: device.h: make struct
bus_type a const *"), the driver core can properly handle constant
struct bus_type, move the host1x_bus_type variable to be a constant
structure as well, placing it into read-only memory which can not be
modified at runtime.
Erico Nunes [Wed, 24 Jan 2024 02:59:47 +0000 (03:59 +0100)]
drm/lima: standardize debug messages by ip name
Some debug messages carried the ip name, or included "lima", or
included both the ip name and then the numbered ip name again.
Make the messages more consistent by always looking up and showing
the ip name first.
Erico Nunes [Wed, 24 Jan 2024 02:59:46 +0000 (03:59 +0100)]
drm/lima: increase default job timeout to 10s
The previous 500ms default timeout was fairly optimistic and could be
hit by real world applications. Many distributions targeting devices
with a Mali-4xx already bumped this timeout to a higher limit.
We can be generous here with a high value as 10s since this should
mostly catch buggy jobs like infinite loop shaders, and these don't
seem to happen very often in real applications.
Marking the context as guilty currently only makes the application which
hits a single timeout problem to stop its rendering context entirely.
All jobs submitted later are dropped from the guilty context.
Lima runs on fairly underpowered hardware for modern standards and it is
not entirely unreasonable that a rendering job may time out occasionally
due to high system load or too demanding application stack. In this case
it would be generally preferred to report the error but try to keep the
application going.
Other similar embedded GPU drivers don't make use of the guilty context
flag. Now that there are reliability improvements to the lima timeout
recovery handling, drop the guilty contexts to let the application keep
running in this case.
Erico Nunes [Wed, 24 Jan 2024 02:59:44 +0000 (03:59 +0100)]
drm/lima: handle spurious timeouts due to high irq latency
There are several unexplained and unreproduced cases of rendering
timeouts with lima, for which one theory is high IRQ latency coming from
somewhere else in the system.
This kind of occurrence may cause applications to trigger unnecessary
resets of the GPU or even applications to hang if it hits an issue in
the recovery path.
Panfrost already does some special handling to account for such
"spurious timeouts", it makes sense to have this in lima too to reduce
the chance that it hit users.
Erico Nunes [Wed, 24 Jan 2024 02:59:43 +0000 (03:59 +0100)]
drm/lima: set gp bus_stop bit before hard reset
This is required for reliable hard resets. Otherwise, doing a hard reset
while a task is still running (such as a task which is being stopped by
the drm_sched timeout handler) may result in random mmu write timeouts
or lockups which cause the entire gpu to hang.
Erico Nunes [Wed, 24 Jan 2024 02:59:42 +0000 (03:59 +0100)]
drm/lima: set pp bus_stop bit before hard reset
This is required for reliable hard resets. Otherwise, doing a hard reset
while a task is still running (such as a task which is being stopped by
the drm_sched timeout handler) may result in random mmu write timeouts
or lockups which cause the entire gpu to hang.
Erico Nunes [Wed, 24 Jan 2024 02:59:41 +0000 (03:59 +0100)]
drm/lima: reset async_reset on gp hard reset
Lima gp jobs use an async reset to avoid having to wait for the soft
reset right after a job. The soft reset is done at the end of a job and
a reset_complete flag is expected to be set at the next job.
However, in case the user runs into a job timeout from any application,
a hard reset is issued to the hardware. This hard reset clears the
reset_complete flag, which causes an error message to show up before the
next job.
This is probably harmless for the execution but can be very confusing to
debug, as it blames a reset timeout on the next application to submit a
job.
Reset the async_reset flag when doing the hard reset so that we don't
get that message.
Erico Nunes [Wed, 24 Jan 2024 02:59:40 +0000 (03:59 +0100)]
drm/lima: reset async_reset on pp hard reset
Lima pp jobs use an async reset to avoid having to wait for the soft
reset right after a job. The soft reset is done at the end of a job and
a reset_complete flag is expected to be set at the next job.
However, in case the user runs into a job timeout from any application,
a hard reset is issued to the hardware. This hard reset clears the
reset_complete flag, which causes an error message to show up before the
next job.
This is probably harmless for the execution but can be very confusing to
debug, as it blames a reset timeout on the next application to submit a
job.
Reset the async_reset flag when doing the hard reset so that we don't
get that message.
Jani Nikula [Tue, 16 Jan 2024 13:07:35 +0000 (15:07 +0200)]
drm: remove drm_debug_printer in favor of drm_dbg_printer
Convert the remaining drm_debug_printer users over to drm_dbg_printer,
as it can handle the cases without struct drm_device pointer, and also
provides drm debug category and prefix support. Remove drm_debug_printer
altogether.
Jani Nikula [Tue, 16 Jan 2024 13:07:28 +0000 (15:07 +0200)]
drm/print: add drm_dbg_printer() for drm device specific printer
We've lacked a device specific debug printer. Add one. Take category
into account too.
__builtin_return_address(0) is inaccurate here, so don't use it. If
necessary, we can later pass __func__ to drm_dbg_printer() by wrapping
it inside a macro.
projects / linux.git / log