...as this will allow split BTF modules with a base BTF
representation (rather than the full vmlinux BTF at time of
BTF encoding) to resolve their references to kernel types in a
way that is more resilient to small changes in kernel types.
This will allow modules that are not built every time the kernel
is to provide more resilient BTF, rather than have it invalidated
every time BTF ids for core kernel types change.
Fields are ordered to avoid holes in struct module.
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20240620091733.1967885-3-alan.maguire@oracle.com
For almost 2 decades, the max allowed requests were limited to 512KB because of
SDMA's max 512KiB boundary limit.
ADMA2 and ADMA3 do not have such limits and were effectively made so any
kind of block count would not impose interrupt and managing stress to the host.
By limiting that to 512KiB, it effectively downgrades these DMA modes to SDMA.
Fix that by actually following the spec:
When ADMA is selected tuning mode is advised.
On lesser modes 4MiB transfers is selected as max, so re-tuning if timer trigger
or if requested by host interrupt, can be done in time.
Otherwise, the only limit is the variable size of types used.
In this implementation, 16MiB is used as maximum since tests showed that after
that point, there are diminishing returns.
Also 16MiB in worst case scenarios, when card is eMMC and its max speed is a
generous 350MiB/s, will generate interrupts every 45ms on huge data transfers.
For example, on local tests with rigorous CPU/GPU burn-in tests and abrupt
cut-offs to generate huge temperature changes (upwards/downwards) to the card,
tested host was fine up to 128MB/s transfers on slow cards that used SDR104
bus timing without re-tuning.
In that case the 4MiB limit was overridden with a more than safe 8MiB value.
In all testing cases and boards, that change brought the following:
Depending on bus timing and eMMC/SD specs:
* Max Read throughput increased by 2-20%
* Max Write throughput increased by 50-200%
Depending on CPU frequency and transfer sizes:
* Reduced mmcqd cpu core usage by 4-50%
Signed-off-by: CTCaer <ctcaer@gmail.com>
Lets isolate scheduler internals from drivers such as pvr which currently
walks the dependency array to look for fences.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Danilo Krummrich <dakr@redhat.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Philipp Stanner <pstanner@redhat.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Acked-by: Danilo Krummrich <dakr@kernel.org>
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20250113103341.43914-1-tvrtko.ursulin@igalia.com
There is no need to check the boolean in the work item's prologues since
the boolean can be set at any later time anyway.
The helper which pauses submission sets it and synchronously cancels the
work and helpers which queue the work check for the flag so all should be
good.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Danilo Krummrich <dakr@redhat.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20250114105942.64832-1-tvrtko.ursulin@igalia.com
No driver is using the update_job_credits() schduler vfunc
so lets remove it.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Danilo Krummrich <dakr@redhat.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Philipp Stanner <pstanner@redhat.com>
Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Matt Coster <matt.coster@imgtec.com>
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20250110111301.76909-1-tvrtko.ursulin@igalia.com
Commit baf4afc5831438 ("drm/sched: Improve teardown documentation")
documents problems of drm_sched_fini() in form of a list. The checklist
triggers htmldocs warning (but renders correctly in htmldocs output):
Documentation/gpu/drm-mm:571: ./drivers/gpu/drm/scheduler/sched_main.c:1359: ERROR: Unexpected indentation.
Separate the list from the preceding paragraph by a blank line to fix
the warning. While at it, also end the aforementioned paragraph by a
colon.
Fixes: baf4afc58314 ("drm/sched: Improve teardown documentation")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Closes: https://lore.kernel.org/r/20241108175655.6d3fcfb7@canb.auug.org.au/
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
[phasta: Adjust commit message]
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241217034915.62594-1-bagasdotme@gmail.com
If a reset is scheduled when the suspend happens, we drop the
reset-pending info on the floor assuming the resume will fix things,
but the resume logic might try a fast reset. If we're lucky, the
fast reset fails and we fallback to a slow reset, but if the FW was
corrupted in a way that makes it partially functional (it boots but
doesn't quite do what it's expected to do), we won't notice immediately
that things are not working correctly, leading to a new reset further
down the road.
Fixes: 5fe909cae118 ("drm/panthor: Add the device logical block")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241217092457.1582053-1-boris.brezillon@collabora.com
Use the correct format for all kernel-doc comments.
Use structname.membername for named structs.
Don't precede function names in kernel-doc with '@' sign.
Use the correct function parameter names in kernel-doc comments.
This fixes around 80 kernel-doc warnings.
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Boris Brezillon <boris.brezillon@collabora.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Liviu Dudau <liviu.dudau@arm.com>
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: Simona Vetter <simona@ffwll.ch>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250111062832.910495-1-rdunlap@infradead.org
Groups can be killed during a reset even though they did nothing wrong.
That usually happens when the FW is put in a bad state by other groups,
resulting in group suspension failures when the reset happens.
If we end up in that situation, flag the group innocent and report
innocence through a new DRM_PANTHOR_GROUP_STATE flag.
Bump the minor driver version to reflect the uAPI change.
Changes in v4:
- Add an entry to the driver version changelog
- Add R-bs
Changes in v3:
- Actually report innocence to userspace
Changes in v2:
- New patch
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211080500.2349505-1-boris.brezillon@collabora.com
If we do a GPU soft-reset, that's no longer fast reset. This also means
the slow reset fallback doesn't work because the MCU state is only reset
after a GPU soft-reset.
Let's move the retry logic to panthor_device_resume() to issue a
soft-reset between the fast and slow attempts, and patch
panthor_gpu_suspend() to only power-off the L2 when a fast reset is
requested.
v3:
- No changes
v2:
- Add R-b
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211075419.2333731-6-boris.brezillon@collabora.com
When the runtime PM resume callback returns an error, it puts the device
in a state where it can't be resumed anymore. Make sure we can recover
from such transient failures by calling pm_runtime_set_suspended()
explicitly after a pm_runtime_resume_and_get() failure.
v3:
- Add R-b/A-b
v2:
- Add a comment explaining potential races in
panthor_device_resume_and_get()
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Adrian Larumbe <adrian.larumbe@collabora.com>
Acked-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211075419.2333731-5-boris.brezillon@collabora.com
devfreq_{resume,suspend}_device() don't bother undoing the suspend_count
modifications if something fails, so either it assumes failures are
harmless, or it's super fragile/buggy. In either case it's not something
we can address at the driver level, so let's just assume failures are
harmless for now, like is done in panfrost.
v3:
- Add R-b
v2:
- Add R-b
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Adrian Larumbe <adrian.larumbe@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211075419.2333731-4-boris.brezillon@collabora.com
The runtime PM resume operation is not guaranteed to succeed, but if it
fails, the device should be in a suspended state. Make sure we're robust
to resume failures in the unplug path.
v3:
- Fix typo
- Add R-bs
v2:
- Move the bit that belonged in the next commit
- Drop the panthor_device_unplug() changes
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Adrian Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211075419.2333731-3-boris.brezillon@collabora.com
WARN() will return true if the condition is true, false otherwise.
If we store the return of drm_WARN_ON() in ret, we lose the actual
error code.
v3:
- Add R-b
v2:
- Add R-b
Fixes: 5fe909cae118 ("drm/panthor: Add the device logical block")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Adrian Larumbe <adrian.larumbe@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241211075419.2333731-2-boris.brezillon@collabora.com
Rockchip extends the base driver with vendor-specific requirements:
- Add private ioctl call RKMODULE_GET_MODULE_INFO
- Implement specific V4L2 callbacks: g_input_status, get_mbus_config
- Read vendor-specific camera module information from device tree
- Hardcoded number of MIPI-CSI2 lanes to 2
- Use old version of probe instead of probe_new
`ROCKCHIP_THUNDER_BOOT_DEFER_FREE_MEMBLOCK` should depend
on `ROCKCHIP_THUNDER_BOOT`, and it creates an artificial
threads-max limit that is not suitable for generic linux.
Disable it by default to avoid it being accidentally enabled.
Signed-off-by: nyanmisaka <nst799610810@gmail.com>
Realtek changed the format of the firmware file as v2. The driver
should implement the patch to extract the firmware data from the
firmware file. The future chips must apply this patch for firmware loading.
This patch is compatible with the both previous format and v2 as well.
Signed-off-by: Allen Chen <allen_chen@realsil.com.cn>
Signed-off-by: Alex Lu <alex_lu@realsil.com.cn>
Tested-by: Hilda Wu <hildawu@realtek.com>
Signed-off-by: Max Chou <max.chou@realtek.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
(cherry picked from commit 9a24ce5e29b15c4c6b0c89c04f9df6ce14addefa)
Signed-off-by: Stephen Chen <stephen@radxa.com>
