We need to save the reservation object pointer associated with the
imported dmabuf in the newly created GEM object to allow
drm_gem_plane_helper_prepare_fb() to extract the exclusive fence
from it and attach it to the plane state during prepare phase.
This is needed to ensure that drm_atomic_helper_wait_for_fences()
correctly waits for the relevant fences (move, etc) associated with
the reservation object, thereby implementing proper synchronization.
Otherwise, artifacts or slight flickering can be seen when apps
are dragged across the screen when running Gnome (Wayland). This
problem is mostly seen with dGPUs in the case where the FBs are
allocated in VRAM but need to be migrated to System RAM as they
are shared with virtio-gpu.
Fixes: ca77f27a26 ("drm/virtio: Import prime buffers from other devices as guest blobs")
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Cc: Gurchetan Singh <gurchetansingh@chromium.org>
Cc: Chia-I Wu <olvaffe@gmail.com>
Signed-off-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
Reviewed-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
[dmitry.osipenko@collabora.com: Moved assignment before object_init()]
Link: https://patchwork.freedesktop.org/patch/msgid/20250325201021.1315080-1-vivek.kasireddy@intel.com
For paired jobs, have the fragment job take a reference on the
geometry job, so that the geometry job cannot be freed until
the fragment job has finished with it.
The geometry job structure is accessed when the fragment job is being
prepared by the GPU scheduler. Taking the reference prevents the
geometry job being freed until the fragment job no longer requires it.
Fixes a use after free bug detected by KASAN:
[ 124.256386] BUG: KASAN: slab-use-after-free in pvr_queue_prepare_job+0x108/0x868 [powervr]
[ 124.264893] Read of size 1 at addr ffff0000084cb960 by task kworker/u16:4/63
Cc: stable@vger.kernel.org
Fixes: eaf01ee5ba ("drm/imagination: Implement job submission and scheduling")
Signed-off-by: Brendan King <brendan.king@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://lore.kernel.org/r/20250318-ddkopsrc-1337-use-after-free-in-pvr_queue_prepare_job-v1-1-80fb30d044a6@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
When slicing a BO, we need to iterate through the BO's sgt to find the
right pieces to construct the slice. Some of the data types chosen for
this process are incorrectly too small, and can overflow. This can
result in the incorrect slice construction, which can lead to data
corruption in workload execution.
The device can only handle 32-bit sized transfers, and the scatterlist
struct only supports 32-bit buffer sizes, so our upper limit for an
individual transfer is an unsigned int. Using an int is incorrect due to
the reservation of the sign bit. Upgrade the length of a scatterlist
entry and the offsets into a scatterlist entry to unsigned int for a
correct representation.
While each transfer may be limited to 32-bits, the overall BO may exceed
that size. For counting the total length of the BO, we need a type that
can represent the largest allocation possible on the system. That is the
definition of size_t, so use it.
Fixes: ff13be8303 ("accel/qaic: Add datapath")
Signed-off-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Signed-off-by: Jeff Hugo <jeff.hugo@oss.qualcomm.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Reviewed-by: Troy Hanson <quic_thanson@quicinc.com>
Reviewed-by: Youssef Samir <quic_yabdulra@quicinc.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250306171959.853466-1-jeff.hugo@oss.qualcomm.com
Similar to commit e4b5ccd392 ("drm/v3d: Ensure job pointer is set to
NULL after job completion"), ensure the job pointer is set to `NULL` when
a job's fence has an error. Failing to do so can trigger kernel warnings
in specific scenarios, such as:
1. v3d_csd_job_run() assigns `v3d->csd_job = job`
2. CSD job exceeds hang limit, causing a timeout → v3d_gpu_reset_for_timeout()
3. GPU reset
4. drm_sched_resubmit_jobs() sets the job's fence to `-ECANCELED`.
5. v3d_csd_job_run() detects the fence error and returns NULL, not
submitting the job to the GPU
6. User-space runs `modprobe -r v3d`
7. v3d_gem_destroy()
v3d_gem_destroy() triggers a warning indicating that the CSD job never
ended, as we didn't set `v3d->csd_job` to NULL after the timeout. The same
can also happen to BIN, RENDER, and TFU jobs.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250313-v3d-gpu-reset-fixes-v4-2-c1e780d8e096@igalia.com
The V3D driver still relies on `drm_sched_increase_karma()` and
`drm_sched_resubmit_jobs()` for resubmissions when a timeout occurs.
The function `drm_sched_increase_karma()` marks the job as guilty, while
`drm_sched_resubmit_jobs()` sets an error (-ECANCELED) in the DMA fence of
that guilty job.
Because of this, we must check whether the job’s DMA fence has been
flagged with an error before executing the job. Otherwise, the same guilty
job may be resubmitted indefinitely, causing repeated GPU resets.
This patch adds a check for an error on the job's fence to prevent running
a guilty job that was previously flagged when the GPU timed out.
Note that the CPU and CACHE_CLEAN queues do not require this check, as
their jobs are executed synchronously once the DRM scheduler starts them.
Cc: stable@vger.kernel.org
Fixes: d223f98f02 ("drm/v3d: Add support for compute shader dispatch.")
Fixes: 1584f16ca9 ("drm/v3d: Add support for submitting jobs to the TFU.")
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250313-v3d-gpu-reset-fixes-v4-1-c1e780d8e096@igalia.com
The handling of the MST Connection Status Notify message is skipped if
the probing of the topology is still pending. Acquiring the
drm_dp_mst_topology_mgr::probe_lock for this in
drm_dp_mst_handle_up_req() is problematic: the task/work this function
is called from is also responsible for handling MST down-request replies
(in drm_dp_mst_handle_down_rep()). Thus drm_dp_mst_link_probe_work() -
holding already probe_lock - could be blocked waiting for an MST
down-request reply while drm_dp_mst_handle_up_req() is waiting for
probe_lock while processing a CSN message. This leads to the probe
work's down-request message timing out.
A scenario similar to the above leading to a down-request timeout is
handling a CSN message in drm_dp_mst_handle_conn_stat(), holding the
probe_lock and sending down-request messages while a second CSN message
sent by the sink subsequently is handled by drm_dp_mst_handle_up_req().
Fix the above by moving the logic to skip the CSN handling to
drm_dp_mst_process_up_req(). This function is called from a work
(separate from the task/work handling new up/down messages), already
holding probe_lock. This solves the above timeout issue, since handling
of down-request replies won't be blocked by probe_lock.
Fixes: ddf983488c ("drm/dp_mst: Skip CSN if topology probing is not done yet")
Cc: Wayne Lin <Wayne.Lin@amd.com>
Cc: Lyude Paul <lyude@redhat.com>
Cc: stable@vger.kernel.org # v6.6+
Reviewed-by: Wayne Lin <Wayne.Lin@amd.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250307183152.3822170-1-imre.deak@intel.com
Video players (eg. mpv) do periodic XResetScreenSaver() calls to
keep the screen on while the video playing. The modesetting ddx
plumbs these straight through into the kernel as DPMS setproperty
ioctls, without any filtering whatsoever. When implemented via
atomic these end up as empty commits on the crtc (which will
nonetheless take one full frame), which leads to a dropped
frame every time XResetScreenSaver() is called.
Let's just filter out redundant DPMS property changes in the
kernel to avoid this issue.
v2: Explain the resulting commits a bit better (Sima)
Document the behaviour in uapi docs (Sima)
Cc: stable@vger.kernel.org
Testcase: igt/kms_flip/flip-vs-dpms-on-nop
Reviewed-by: Simona Vetter <simona.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250219160239.17502-1-ville.syrjala@linux.intel.com
Since pci_get_domain_bus_and_slot() can return NULL, add NULL check for
pci_gfx_root in the mid_get_vbt_data().
This change is similar to the checks implemented in mid_get_fuse_settings()
and mid_get_pci_revID(), which were introduced by commit 0cecdd818c
("gma500: Final enables for Oaktrail") as "additional minor
bulletproofing".
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: f910b41105 ("gma500: Add the glue to the various BIOS and firmware interfaces")
Signed-off-by: Ivan Abramov <i.abramov@mt-integration.ru>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250306112046.17144-1-i.abramov@mt-integration.ru
When firmware traces are enabled, the firmware dumps 48-bit timestamps
for each trace as two 32-bit values, highest 32 bits (of which only 16
useful) first.
The driver was reassembling them the other way round i.e. interpreting
the first value in memory as the lowest 32 bits, and the second value
as the highest 32 bits (then truncated to 16 bits).
Due to this, firmware trace dumps showed very large timestamps even for
traces recorded shortly after GPU boot. The timestamps in these dumps
would also sometimes jump backwards because of the truncation.
Example trace dumped after loading the powervr module and enabling
firmware traces, where each line is commented with the timestamp value
in hexadecimal to better show both issues:
[93540092739584] : Host Sync Partition marker: 1 // 0x551300000000
[28419798597632] : GPU units deinit // 0x19d900000000
[28548647616512] : GPU deinit // 0x19f700000000
Update logic to reassemble the timestamps halves in the correct order.
Fixes: cb56cd6108 ("drm/imagination: Add firmware trace to debugfs")
Signed-off-by: Alessio Belle <alessio.belle@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250221-fix-fw-trace-timestamps-v1-1-dba4aeb030ca@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Do scheduler queue fence release processing on a workqueue, rather
than in the release function itself.
Fixes deadlock issues such as the following:
[ 607.400437] ============================================
[ 607.405755] WARNING: possible recursive locking detected
[ 607.415500] --------------------------------------------
[ 607.420817] weston:zfq0/24149 is trying to acquire lock:
[ 607.426131] ffff000017d041a0 (reservation_ww_class_mutex){+.+.}-{3:3}, at: pvr_gem_object_vunmap+0x40/0xc0 [powervr]
[ 607.436728]
but task is already holding lock:
[ 607.442554] ffff000017d105a0 (reservation_ww_class_mutex){+.+.}-{3:3}, at: dma_buf_ioctl+0x250/0x554
[ 607.451727]
other info that might help us debug this:
[ 607.458245] Possible unsafe locking scenario:
[ 607.464155] CPU0
[ 607.466601] ----
[ 607.469044] lock(reservation_ww_class_mutex);
[ 607.473584] lock(reservation_ww_class_mutex);
[ 607.478114]
*** DEADLOCK ***
Cc: stable@vger.kernel.org
Fixes: eaf01ee5ba ("drm/imagination: Implement job submission and scheduling")
Signed-off-by: Brendan King <brendan.king@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250226-fence-release-deadlock-v2-1-6fed2fc1fe88@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
DMA areas are not necessarily backed by struct page, so we cannot
rely on it for deferred I/O. Allocate a shadow buffer for drivers
that require deferred I/O and use it as framebuffer memory.
Fixes driver errors about being "Unable to handle kernel NULL pointer
dereference at virtual address" or "Unable to handle kernel paging
request at virtual address".
The patch splits drm_fbdev_dma_driver_fbdev_probe() in an initial
allocation, which creates the DMA-backed buffer object, and a tail
that sets up the fbdev data structures. There is a tail function for
direct memory mappings and a tail function for deferred I/O with
the shadow buffer.
It is no longer possible to use deferred I/O without shadow buffer.
It can be re-added if there exists a reliably test for usable struct
page in the allocated DMA-backed buffer object.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reported-by: Nuno Gonçalves <nunojpg@gmail.com>
CLoses: https://lore.kernel.org/dri-devel/CAEXMXLR55DziAMbv_+2hmLeH-jP96pmit6nhs6siB22cpQFr9w@mail.gmail.com/
Tested-by: Nuno Gonçalves <nunojpg@gmail.com>
Fixes: 5ab91447aa ("drm/tiny/ili9225: Use fbdev-dma")
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: <stable@vger.kernel.org> # v6.11+
Reviewed-by: Simona Vetter <simona.vetter@ffwll.ch>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211090643.74250-1-tzimmermann@suse.de
When compiling without CONFIG_IA32_EMULATION, there can be some errors:
drivers/accel/amdxdna/amdxdna_mailbox.c: In function ‘mailbox_release_msg’:
drivers/accel/amdxdna/amdxdna_mailbox.c:197:2: error: implicit declaration
of function ‘kfree’.
197 | kfree(mb_msg);
| ^~~~~
drivers/accel/amdxdna/amdxdna_mailbox.c: In function ‘xdna_mailbox_send_msg’:
drivers/accel/amdxdna/amdxdna_mailbox.c:418:11: error:implicit declaration
of function ‘kzalloc’.
418 | mb_msg = kzalloc(sizeof(*mb_msg) + pkg_size, GFP_KERNEL);
| ^~~~~~~
Add the missing include.
Fixes: b87f920b93 ("accel/amdxdna: Support hardware mailbox")
Signed-off-by: Su Hui <suhui@nfschina.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250211015354.3388171-1-suhui@nfschina.com
I was pondering with myself for a while if I should just make it official
that I'm not really involved in the kernel community anymore, neither as a
reviewer, nor as a maintainer.
Most of the time I simply excused myself with "if something urgent comes
up, I can chime in and help out". Lyude and Danilo are doing a wonderful
job and I've put all my trust into them.
However, there is one thing I can't stand and it's hurting me the most.
I'm convinced, no, my core believe is, that inclusivity and respect,
working with others as equals, no power plays involved, is how we should
work together within the Free and Open Source community.
I can understand maintainers needing to learn, being concerned on
technical points. Everybody deserves the time to understand and learn. It
is my true belief that most people are capable of change eventually. I
truly believe this community can change from within, however this doesn't
mean it's going to be a smooth process.
The moment I made up my mind about this was reading the following words
written by a maintainer within the kernel community:
"we are the thin blue line"
This isn't okay. This isn't creating an inclusive environment. This isn't
okay with the current political situation especially in the US. A
maintainer speaking those words can't be kept. No matter how important
or critical or relevant they are. They need to be removed until they
learn. Learn what those words mean for a lot of marginalized people. Learn
about what horrors it evokes in their minds.
I can't in good faith remain to be part of a project and its community
where those words are tolerated. Those words are not technical, they are
a political statement. Even if unintentionally, such words carry power,
they carry meanings one needs to be aware of. They do cause an immense
amount of harm.
I wish the best of luck for everybody to continue to try to work from
within. You got my full support and I won't hold it against anybody trying
to improve the community, it's a thankless job, it's a lot of work. People
will continue to burn out.
I got burned out enough by myself caring about the bits I maintained, but
eventually I had to realize my limits. The obligation I felt was eating me
from inside. It stopped being fun at some point and I reached a point
where I simply couldn't continue the work I was so motivated doing as I've
did in the early days.
Please respect my wishes and put this statement as is into the tree.
Leaving anything out destroys its entire meaning.
Respectfully
Karol
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20250215073753.1217002-2-kherbst@redhat.com
The current implementation has a bug: If the current css doesn't
contain any pool that is a descendant of the "pool" (i.e. when
found_descendant == false), then "pool" will point to some unrelated
pool. If the current css has a child, we'll overwrite parent_pool with
this unrelated pool on the next iteration.
Since we can just check whether a pool refers to the same region to
determine whether or not it's related, all the additional pool tracking
is unnecessary, so just switch to using css_for_each_descendant_pre for
traversal.
Fixes: b168ed458d ("kernel/cgroup: Add "dmem" memory accounting cgroup")
Signed-off-by: Friedrich Vock <friedrich.vock@gmx.de>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250127152754.21325-1-friedrich.vock@gmx.de
Signed-off-by: Maarten Lankhorst <dev@lankhorst.se>
In case we have to retry the loop, we are missing to unlock+put the
folio. In that case, we will keep failing make_device_exclusive_range()
because we cannot grab the folio lock, and even return from the function
with the folio locked and referenced, effectively never succeeding the
make_device_exclusive_range().
While at it, convert the other unlock+put to use a folio as well.
This was found by code inspection.
Fixes: 8f187163eb ("nouveau/svm: implement atomic SVM access")
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Tested-by: Alistair Popple <apopple@nvidia.com>
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20250124181524.3584236-2-david@redhat.com
In jadard_prepare() a reset pulse is generated with the following
statements (delays ommited for clarity):
gpiod_set_value(jadard->reset, 1); --> Deassert reset
gpiod_set_value(jadard->reset, 0); --> Assert reset for 10ms
gpiod_set_value(jadard->reset, 1); --> Deassert reset
However, specifying second argument of "0" to gpiod_set_value() means to
deassert the GPIO, and "1" means to assert it. If the reset signal is
defined as GPIO_ACTIVE_LOW in the DTS, the above statements will
incorrectly generate the reset pulse (inverted) and leave it asserted
(LOW) at the end of jadard_prepare().
Fix reset behavior by inverting gpiod_set_value() second argument
in jadard_prepare(). Also modify second argument to devm_gpiod_get()
in jadard_dsi_probe() to assert the reset when probing.
Do not modify it in jadard_unprepare() as it is already properly
asserted with "1", which seems to be the intended behavior.
Fixes: 6b818c533d ("drm: panel: Add Jadard JD9365DA-H3 DSI panel")
Cc: stable@vger.kernel.org
Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20240927135306.857617-1-hugo@hugovil.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20240927135306.857617-1-hugo@hugovil.com
Without the DP helper code, the newly added displayport support
causes a link failure:
x86_64-linux-ld: drivers/gpu/drm/hisilicon/hibmc/dp/dp_aux.o: in function `hibmc_dp_aux_init':
dp_aux.c:(.text+0x37e): undefined reference to `drm_dp_aux_init'
x86_64-linux-ld: drivers/gpu/drm/hisilicon/hibmc/dp/dp_link.o: in function `hibmc_dp_link_set_pattern':
dp_link.c:(.text+0xae): undefined reference to `drm_dp_dpcd_write'
x86_64-linux-ld: drivers/gpu/drm/hisilicon/hibmc/dp/dp_link.o: in function `hibmc_dp_link_get_adjust_train':
dp_link.c:(.text+0x121): undefined reference to `drm_dp_get_adjust_request_voltage'
x86_64-linux-ld: dp_link.c:(.text+0x12e): undefined reference to `drm_dp_get_adjust_request_pre_emphasis'
x86_64-linux-ld: drivers/gpu/drm/hisilicon/hibmc/dp/dp_link.o: in function `hibmc_dp_link_training':
dp_link.c:(.text+0x2b0): undefined reference to `drm_dp_dpcd_write'
x86_64-linux-ld: dp_link.c:(.text+0x2e3): undefined reference to `drm_dp_dpcd_write'
Add both DRM_DISPLAY_DP_HELPER and DRM_DISPLAY_HELPER, which is
in turn required by the former.
Fixes: 0ab6ea261c ("drm/hisilicon/hibmc: add dp module in hibmc")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20250127071059.617567-1-arnd@kernel.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
The find_preferred_mode() functions takes the mode_config mutex, but due
to the order most tests have, is called with the crtc_ww_class_mutex
taken. This raises a warning for a circular dependency when running the
tests with lockdep.
Reorder the tests to call find_preferred_mode before the acquire context
has been created to avoid the issue.
Reviewed-by: Simona Vetter <simona.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20250129-test-kunit-v2-4-fe59c43805d5@kernel.org
Signed-off-by: Maxime Ripard <mripard@kernel.org>
