Smatch detected potential error pointer dereference.
drivers/gpu/drm/drm_syncobj.c:888 drm_syncobj_transfer_to_timeline()
error: 'fence' dereferencing possible ERR_PTR()
The error pointer comes from dma_fence_allocate_private_stub(). One
caller expected error pointers and one expected NULL pointers. Change
it to return NULL and update the caller which expected error pointers,
drm_syncobj_assign_null_handle(), to check for NULL instead.
Fixes: f781f661e8 ("dma-buf: keep the signaling time of merged fences v3")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Sumit Semwal <sumit.semwal@linaro.org>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/b09f1996-3838-4fa2-9193-832b68262e43@moroto.mountain
Try to deal with duplicate child devices for the same DDI port
by attempting to initialize them in VBT defined order The first
on to succeed for a specific DDI port will be the one we use.
We'll also get rid of i915->display.vbt.ports[] here as any conflicts
will now be handled at encoder registration time rather than during
VBT parsing. Note that intel_bios_encoder_data_lookup() still remaims
for pre-DDI DP/HDMI ports as those don't (at least yet) use VBT
driven initialization.
TODO: DSI dual link handling is sketchy at best
v2: Leave intel_bios_encoder_port() to the encoder callback (Jani)
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230630155846.29931-7-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Stop with the VBT AUX CH sanitation, and instead just check
that the appropriate AUX CH is still available when initializing
a DP/TC port.
The reason being that we want to start initializing ports in
VBT order to deal with VBTs that declare child devices with
seemingly conflicting ports. As the encoder initialization can
fail for other reasons (at least for eDP+AUX) we can't know
upfront which way the conflicts should be resolved.
Note that the old way of sanitizing gave priority to the last
port declared in the VBT, but now we sort of do the opposite by
favoring the first encoder to successfully initialize. The reason
for the old "last port wins" preference was eg. Asrock B250M-HDV
where port A (eDP) and port E (DP->VGA) have an AUX CH conflict
and we need to prefer port E. However with the new way port A (eDP)
will be probed first, but will fail to probe due to HPD and thus
port E will still win in the end.
v2: Pimp the commit message (Jani)
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230630155846.29931-5-ville.syrjala@linux.intel.com
Stop with the VBT DDC pin sanitation, and instead just check
that the appropriate DDC pin is still available when initializing
a HDMI connector.
The reason being that we want to start initializing ports in
VBT order to deal with VBTs that declare child devices with
seemingly conflicting ports. As the encoder initialization can
fail for other reasons (at least for eDP+AUX) we can't know
upfront which way the conflicts should be resolved.
Note that the old way of sanitizing gave priority to the last
port declared in the VBT, but now we sort of do the opposite by
favoring the first encoder to successfully initialize. So far
we're not aware of HDMI/DDC use cases where this would matter
but for AUX CH (will be subject to a similar change) there are
known cases where it matters.
Also note that the old code fell back to the platform default DDC
pin if the VBT pin was populated but invalid. That doesn't seem like
such a great idea because the VBT might have later declared another
port using that platform default pin, and so we might just be
creating more DDC pin conflicts here. So lets not second guess the
VBT and simply reject the entire HDMI encoder if the VBT DDC pin is
invalid.
v2: Pimp the commit message (Jani)
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230630155846.29931-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Currently we just clamp that value to the highest supported one, however that
means, we are not able to fit this into our available bandwidth range, so we
might see glitches or FIFO underruns.
While choosing less compressed bpp than min bpp required to handle the mode is
harmless and might even save some bandwidth, choosing higher compressed bpp than
min bpp required to handle the required mode config, can cause issues.
So in that case lets just conclude that even with DSC, we are not able to comply
with bandwidth requirements and fail.
v2: - s/clamp_t/min_t/ (Luca Coelho)
Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Reviewed-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230629122534.8815-1-stanislav.lisovskiy@intel.com
Loongson display controller IP has been integrated in both Loongson north
bridge chipset (ls7a1000/ls7a2000) and Loongson SoCs (ls2k1000/ls2k2000).
It has even been included in Loongson's BMC products. It has two display
pipes, and each display pipe supports a primary plane and a cursor plane.
For the DC in the LS7a1000, each display pipe has a DVO output interface,
which is able to support 1920x1080@60Hz. For the DC in the LS7A2000, each
display pipe is equipped with a built-in HDMI encoder, which is compliant
with the HDMI 1.4 specification. The first display pipe is also equipped
with a transparent VGA encoder, which is parallel with the HDMI encoder.
To get a decent performance for writing framebuffer data to the VRAM, the
write combine support should be enabled.
v1 -> v2:
1) Use hpd status reg when polling for ls7a2000.
2) Fix all warnings that emerged when compiling with W=1.
v2 -> v3:
1) Add COMPILE_TEST to Kconfig and make the driver off by default
2) Alphabetical sorting headers (Thomas)
3) Untangle register access functions as much as possible (Thomas)
4) Switch to TTM-based memory manager (Thomas)
5) Add the chip ID detection function which can be used to distinguish
chip models
6) Revise the built-in HDMI phy driver, nearly all main stream mode below
4K@30Hz is tested, and this driver supports clone(mirror) display mode
and extend(joint) display mode.
v3 -> v4:
1) Quickly fix a small mistake.
v4 -> v5:
1) Add per display pipe debugfs support to the builtin HDMI encoder.
v5 -> v6:
1) Remove stray code which didn't get used, say lsdc_of_get_reserved_ram
2) Fix all typos I could found, make sentences and code more readable
3) Untangle lsdc_hdmi*_connector_detect() function according to the pipe
4) Rename this driver as loongson.
v6 -> v7:
1) Add prime support for buffer self-sharing, sharing buffer with
drm/etnaviv is also tested and it works with limitations.
2) Implement buffer object tracking with list_head.
3) Add S3(sleep to RAM) support
4) Rewrite lsdc_bo_move since TTM core stop allocating resources
during BO creation. Patch V1 ~ V6 of this series no longer work.
Thus, we send V7.
v7 -> v8:
1) Zero a compile warning on a 32-bit platform, compile with W=1
2) Revise lsdc_bo_gpu_offset() and make minor cleanups.
3) Pageflip tested on the virtual terminal with the following commands:
modetest -M loongson -s 32:1920x1080 -v
modetest -M loongson -s 34:1920x1080 -v -F tiles
It works like a charm, when running the pageflip test with dual screens
configuration, another two additional BOs were created by the modetest,
VRAM usage up to 40+ MB, well we have at least 64MB, still enough.
# cat bos
bo[0000]: size: 8112kB VRAM
bo[0001]: size: 16kB VRAM
bo[0002]: size: 16kB VRAM
bo[0003]: size: 16208kB VRAM
bo[0004]: size: 8112kB VRAM
bo[0005]: size: 8112kB VRAM
v8 -> v9:
1) Select I2C and I2C_ALGOBIT in Kconfig, should depend on MMU.
2) Using pci_get_domain_bus_and_slot to get the GPU device.
v9 -> v10:
1) Revise lsdc_drm_freeze() to implement S3 correctly. We realized that
the pinned BO could not be moved, the VRAM lost power when sleeping
to RAM. Thus, the data in the buffer who is pinned in VRAM will get
lost when resumed. Yet it's not a big problem because this driver
relies on the CPU to update the front framebuffer. We can see the
garbage data when resume from S3, but the screen will show the right
image as I move the cursor. This is due to the CPU repaint. v10 of
this patch makes S3 perfect by unpin all of the BOs in VRAM, evict
them all to system RAM in lsdc_drm_freeze().
v10 -> v11:
1) On a double-screen case, The buffer object backing the single giant
framebuffer is referenced by two GEM objects; hence, it will be
pinned at least twice by prepare_fb() function. This causes its pin
count > 1. V10 of this patch only unpins VRAM BOs once when suspend,
which is not correct on double-screen case. V11 of this patch unpin
the BOs until its pin count reaches zero when suspend. Then, we make
the S3 support complete finally. With v11, I can't see any garbage
data when resume.
2) Fix vblank wait timeout when disable CRTC.
3) Test against IGT, at least fbdev test and kms_flip test passed.
4) Rewrite pixel PLL update function, magic numbers eliminated (Emil)
5) Drop a few common hardware features description in lsdc_desc (Emil)
6) Drop lsdc_mode_config_mode_valid(), instead add restrictions in dumb
create function. (Emil)
7) Untangle the ls7a1000 case and ls7a2000 case completely (Thomas)
v11 -> v12:
none
v12 -> v13:
1) Add benchmarks to figure out the bandwidth of the hardware platform.
Usage:
# cd /sys/kernel/debug/dri/0/
# cat benchmark
2) VRAM is filled with garbage data if uninitialized, add a buffer
clearing procedure (lsdc_bo_clear), clear the BO on creation time.
3) Update copyrights and adjust coding style (Huacai)
v13 -> v14:
1) Trying to add async update support for cursor plane.
v14 -> v15:
1) Add lsdc_vga_set_decode() funciton, which allow us remove multi-video
cards workaround, now it allow drm/loongson, drm/amdgpu, drm/etnaviv
co-exist in the system, more is also possible (Emil and Xuerui)
2) Fix typos and grammar mistakes as much as possible (Xuerui)
3) Unify copyrights as GPL-2.0+ (Xuerui)
4) Fix a bug introduce since V13, TTM may import BO from other drivers,
we shouldn't clear it on such a case.
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: David Airlie <airlied@gmail.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Geert Uytterhoeven <geert+renesas@glider.be>
Cc: loongson-kernel@lists.loongnix.cn
Tested-by: Liu Peibao <liupeibao@loongson.cn>
Tested-by: Li Yi <liyi@loongson.cn>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Sui Jingfeng <suijingfeng@loongson.cn>
Link: https://patchwork.freedesktop.org/patch/msgid/20230615143613.1236245-2-15330273260@189.cn
Most of the context workarounds tweak masked registers, but not all. For
masked registers, when writing the value it's sufficient to just write
the wa->set_bits since that will take care of both the clr and set bits
as well as not overwriting other bits.
However there are some workarounds, the registers are non-masked. Up
until now the driver was simply emitting a MI_LOAD_REGISTER_IMM with the
set_bits to program the register via the GPU in the WA bb. This has the
side effect of overwriting the content of the register outside of bits
that should be set and also doesn't handle the bits that should be
cleared.
Kenneth reported that on DG2, mesa was seeing a weird behavior due to
the kernel programming of L3SQCREG5 in dg2_ctx_gt_tuning_init(). With
the GPU idle, that register could be read via intel_reg as 0x00e001ff,
but during a 3D workload it would change to 0x0000007f. So the
programming of that tuning was affecting more than the bits in
L3_PWM_TIMER_INIT_VAL_MASK. Matt Roper noticed the lack of rmw for the
context workarounds due to the use of MI_LOAD_REGISTER_IMM.
So, for registers that are not masked, read its value via mmio, modify
and then set it in the buffer to be written by the GPU. This should take
care in a simple way of programming just the bits required by the
tuning/workaround. If in future there are registers that involved that
can't be read by the CPU, a more complex approach may be required like
a) issuing additional instructions to read and modify; or b) scan the
golden context and patch it in place before saving it; or something
else. But for now this should suffice.
Scanning the context workarounds for all platforms, these are the
impacted ones with the respective registers
mtl: DRAW_WATERMARK
mtl/dg2: XEHP_L3SQCREG5, XEHP_FF_MODE2
ICL has some non-masked registers in the context workarounds:
GEN8_L3CNTLREG, IVB_FBC_RT_BASE and VB_FBC_RT_BASE_UPPER, but there
shouldn't be an impact. The first is already being manually read and the
other 2 are intentionally overwriting the entire register. Same
reasoning applies to GEN12_FF_MODE2: the WA is intentionally
overwriting all the bits to avoid a read-modify-write.
v2: Reword commit message wrt GEN12_FF_MODE2 and the changed behavior
on preparatory patches.
v3: Also skip reading if clear|set bits covers everything
Cc: Kenneth Graunke <kenneth@whitecape.org>
Cc: Matt Roper <matthew.d.roper@intel.com>
Link: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23783#note_1968971
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230630203509.1635216-4-lucas.demarchi@intel.com
Right now context workarounds don't do a rmw and instead only write to
the register. Since 2 separate programmings to the same register are
coalesced into a single write, this is not problematic for
GEN12_FF_MODE2 since both TDS and GS timer are going to be written
together and the other remaining bits be zeroed.
However in order to fix other workarounds that may want to preserve the
unrelated bits in the same register, context workarounds need to
be changed to a rmw. To prepare for that, move the programming of
GEN12_FF_MODE2 to a single place so the value passed for "clear" can
be all the bits. Otherwise the second workaround would be dropped as
it'd be detected as overwriting a previously programmed workaround.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230630203509.1635216-3-lucas.demarchi@intel.com
Wait until the command transfer FIFO is empty before loading in the next
command. The previous behavior where the code waited until command transfer
FIFO was not full suffered from transfer corruption, where the last command
in the FIFO could be overwritten in case the FIFO indicates not full, but
also does not have enough space to store another transfer yet.
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>
Tested-by: Jagan Teki <jagan@amarulasolutions.com> # imx8mm-icore
Link: https://patchwork.freedesktop.org/patch/msgid/20230615201511.565923-1-marex@denx.de
fence Decrements the reference count before exiting.
Avoid Race Vulnerabilities for fence use-after-free.
v2 (chk): actually fix the use after free and not just move it.
Signed-off-by: shanzhulig <shanzhulig@gmail.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
An intentional delay is added on soft ctf triggered. Then there will
be a double check for the GPU temperature before taking further
action. This can avoid unintended shutdown due to temperature
momentary fluctuation.
Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Preparation for coming optimization which eliminates the influence of
GPU temperature momentary fluctuation.
Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This version brings along the following:
- Improve debugging mechanism for Gaming FAMS
- Add monitor specific edid quirks
- Fixes for Phantom pipe
- Fixes for Shapper LUT
- Clean up asserts
Acked-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Description]
- A full update is required if the number of planes for a given
stream changes
- The new fast update path only checked for stream and plane updates,
but there could be a plane addition or removal without one of the
stream and plane updates triggering a full update
- Add an explicit check for number of planes changing for a full update
Reviewed-by: Samson Tam <samson.tam@amd.com>
Acked-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Description]
- We want to limit the new fast update path to address and gamma updates
only.
- Add a check in dc_update_planes_and_stream to only take the new fast
update path if we only have the specific fast updates defined.
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Jun Lei <jun.lei@amd.com>
Acked-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The shaper LUT requires a 10-bit value of the delta between segments. We
were using dc_fixpt_clamp_u0d10() to do that but it doesn't do what we
want it to do. It will preserve 10-bit precision after the decimal
point, but that's not quite what we want. We want 14-bit precision and
discard the 4 most-significant bytes.
To do that we'll do dc_fixpt_clamp_u0d14() & 0x3ff instead.
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Krunoslav Kovac <krunoslav.kovac@amd.com>
Acked-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If shaper and 3D LUT data updates, lut_3d bit in update_flag is updated
and we need to call set_input_transfer_func to program DPP shaper and 3D
LUTs. Small cleanup of code style in the related if-condition.
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Krunoslav Kovac <krunoslav.kovac@amd.com>
Acked-by: Rodrigo Siqueira <rodrigo.siqueira@amd.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Need to unpause dpg first, or it will hit follow error during stop dpg:
"[drm] Register(1) [regUVD_POWER_STATUS] failed to reach value 0x00000001 != 0x00000000n"
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
So we can selectively enable it on certain devices. No
intended functional change.
Reviewed-and-tested-by: Jiadong Zhu <Jiadong.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
PSP functions are already set by psp_early_init() so initializing
them a second time is unnecessary.
No intended functional changes.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
