MediaTek MT8188 SoC has an ARM Mali-G57 MC3 GPU (Valhall-JM), which
constantly faults with the current panfrost support.
For instance, running `glmark2-es2-drm` benchmark test:
```
[ 79.617461] panfrost 13000000.gpu: js fault, js=1, status=JOB_BUS_FAULT, head=0xaadc380, tail=0xaadc380
[ 80.119811] panfrost 13000000.gpu: gpu sched timeout, js=0, config=0x7300, status=0x58, head=0xaaca180, tail=0xaaca180, sched_job=000000002fd03ccc
[ 80.129083] panfrost 13000000.gpu: Unhandled Page fault in AS0 at VA 0x0000000000000000
[ 80.129083] Reason: TODO
[ 80.129083] raw fault status: 0x1C2
[ 80.129083] decoded fault status: SLAVE FAULT
[ 80.129083] exception type 0xC2: TRANSLATION_FAULT_2
[ 80.129083] access type 0x1: EXECUTE
[ 80.129083] source id 0x0
```
Note that current panfrost mode (Mali LPAE - LEGACY) only allows to
specify write-cache or implementation-defined as the caching policy,
probably not matching the right configuration. As depicted in the source
code:
drivers/iommu/io-pgtable-arm.c:
```
* MEMATTR: Mali has no actual notion of a non-cacheable type, so the
* best we can do is mimic the out-of-tree driver and hope that the
* "implementation-defined caching policy" is good enough...
```
Now that Panfrost supports AARCH64_4K page table format, let's enable it
on Mediatek MT8188 and configure the cache/shareability policies
properly.
Signed-off-by: Ariel D'Alessandro <ariel.dalessandro@collabora.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://lore.kernel.org/r/20250324185801.168664-6-ariel.dalessandro@collabora.com
The v6.13-rc2 release included a bunch of breaking changes,
specifically the MODULE_IMPORT_NS commit.
Backmerge in order to fix them before the next pull-request.
Include the fix from Stephen Roswell.
Caused by commit
25c3fd1183 ("drm/virtio: Add a helper to map and note the dma addrs and lengths")
Interacting with commit
cdd30ebb1b ("module: Convert symbol namespace to string literal")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Link: https://patchwork.freedesktop.org/patch/msgid/20241209121717.2abe8026@canb.auug.org.au
Signed-off-by: Maarten Lankhorst <dev@lankhorst.se>
The continual trickle of small conversion patches is grating on me, and
is really not helping. Just get rid of the 'remove_new' member
function, which is just an alias for the plain 'remove', and had a
comment to that effect:
/*
* .remove_new() is a relic from a prototype conversion of .remove().
* New drivers are supposed to implement .remove(). Once all drivers are
* converted to not use .remove_new any more, it will be dropped.
*/
This was just a tree-wide 'sed' script that replaced '.remove_new' with
'.remove', with some care taken to turn a subsequent tab into two tabs
to make things line up.
I did do some minimal manual whitespace adjustment for places that used
spaces to line things up.
Then I just removed the old (sic) .remove_new member function, and this
is the end result. No more unnecessary conversion noise.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current implementation of drm_sched_start uses a hardcoded
-ECANCELED to dispose of a job when the parent/hw fence is NULL.
This results in drm_sched_job_done being called with -ECANCELED for
each job with a NULL parent in the pending list, making it difficult
to distinguish between recovery methods, whether a queue reset or a
full GPU reset was used.
To improve this, we first try a soft recovery for timeout jobs and
use the error code -ENODATA. If soft recovery fails, we proceed with
a queue reset, where the error code remains -ENODATA for the job.
Finally, for a full GPU reset, we use error codes -ECANCELED or
-ETIME. This patch adds an error code parameter to drm_sched_start,
allowing us to differentiate between queue reset and GPU reset
failures. This enables user mode and test applications to validate
the expected correctness of the requested operation. After a
successful queue reset, the only way to continue normal operation is
to call drm_sched_job_done with the specific error code -ENODATA.
v1: Initial implementation by Jesse utilized amdgpu_device_lock_reset_domain
and amdgpu_device_unlock_reset_domain to allow user mode to track
the queue reset status and distinguish between queue reset and
GPU reset.
v2: Christian suggested using the error codes -ENODATA for queue reset
and -ECANCELED or -ETIME for GPU reset, returned to
amdgpu_cs_wait_ioctl.
v3: To meet the requirements, we introduce a new function
drm_sched_start_ex with an additional parameter to set
dma_fence_set_error, allowing us to handle the specific error
codes appropriately and dispose of bad jobs with the selected
error code depending on whether it was a queue reset or GPU reset.
v4: Alex suggested using a new name, drm_sched_start_with_recovery_error,
which more accurately describes the function's purpose.
Additionally, it was recommended to add documentation details
about the new method.
v5: Fixed declaration of new function drm_sched_start_with_recovery_error.(Alex)
v6 (chk): rebase on upstream changes, cleanup the commit message,
drop the new function again and update all callers,
apply the errno also to scheduler fences with hw fences
v7 (chk): rebased
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240826122541.85663-1-christian.koenig@amd.com
Extend the uAPI with a new job requirement flag for cycle
counters. This requirement is used by userland to indicate that a job
requires cycle counters or system timestamp to be propagated. (for use
with write value timestamp jobs)
We cannot enable cycle counters unconditionally as this would result in
an increase of GPU power consumption. As a result, they should be left
off unless required by the application.
If a job requires cycle counters or system timestamps propagation, we
must enable cycle counting before issuing a job and disable it right
after the job completes.
Since this extends the uAPI and because userland needs a way to advertise
features like VK_KHR_shader_clock conditionally, we bumps the driver
minor version.
v2:
- Rework commit message
- Squash uAPI changes and implementation in this commit
- Simplify changes based on Steven Price comments
v3:
- Add Steven Price r-b
- Fix a codestyle issue
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Tested-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240819080224.24914-3-mary.guillemard@collabora.com
Expose system timestamp and frequency supported by the GPU.
Mali uses an external timer as GPU system time. On ARM, this is wired to
the generic arch timer so we wire cntfrq_el0 as device frequency.
This new uAPI will be used in Mesa to implement timestamp queries and
VK_KHR_calibrated_timestamps.
v2:
- Rewrote to use GPU timestamp register
- Add missing include for arch_timer_get_cntfrq
- Rework commit message
v3:
- Move panfrost_cycle_counter_get and panfrost_cycle_counter_put to
panfrost_ioctl_query_timestamp
- Handle possible overflow in panfrost_timestamp_read
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Tested-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240819080224.24914-2-mary.guillemard@collabora.com
Panfrost DRM driver uses devfreq to perform DVFS, while using simple_ondemand
devfreq governor by default. This causes driver initialization to fail on
boot when simple_ondemand governor isn't built into the kernel statically,
as a result of the missing module dependency and, consequently, the required
governor module not being included in the initial ramdisk. Thus, let's mark
simple_ondemand governor as a softdep for Panfrost, to have its kernel module
included in the initial ramdisk.
This is a rather longstanding issue that has forced distributions to build
devfreq governors statically into their kernels, [1][2] or has forced users
to introduce some unnecessary workarounds. [3]
For future reference, not having support for the simple_ondemand governor in
the initial ramdisk produces errors in the kernel log similar to these below,
which were taken from a Pine64 RockPro64:
panfrost ff9a0000.gpu: [drm:panfrost_devfreq_init [panfrost]] *ERROR* Couldn't initialize GPU devfreq
panfrost ff9a0000.gpu: Fatal error during GPU init
panfrost: probe of ff9a0000.gpu failed with error -22
Having simple_ondemand marked as a softdep for Panfrost may not resolve this
issue for all Linux distributions. In particular, it will remain unresolved
for the distributions whose utilities for the initial ramdisk generation do
not handle the available softdep information [4] properly yet. However, some
Linux distributions already handle softdeps properly while generating their
initial ramdisks, [5] and this is a prerequisite step in the right direction
for the distributions that don't handle them properly yet.
[1] https://gitlab.manjaro.org/manjaro-arm/packages/core/linux/-/blob/linux61/config?ref_type=heads#L8180
[2] https://salsa.debian.org/kernel-team/linux/-/merge_requests/1066
[3] https://forum.pine64.org/showthread.php?tid=15458
[4] https://git.kernel.org/pub/scm/utils/kernel/kmod/kmod.git/commit/?id=49d8e0b59052999de577ab732b719cfbeb89504d
[5] 97ac4d37aa
Cc: Diederik de Haas <didi.debian@cknow.org>
Cc: Furkan Kardame <f.kardame@manjaro.org>
Cc: stable@vger.kernel.org
Fixes: f3ba91228e ("drm/panfrost: Add initial panfrost driver")
Signed-off-by: Dragan Simic <dsimic@manjaro.org>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/4e1e00422a14db4e2a80870afb704405da16fd1b.1718655077.git.dsimic@manjaro.org
When Panfrost must pin an object that is being prepared a dma-buf
attachment for on behalf of another driver, the core drm gem object pinning
code already takes a lock on the object's dma reservation.
However, Panfrost GEM object's pinning callback would eventually try taking
the lock on the same dma reservation when delegating pinning of the object
onto the shmem subsystem, which led to a deadlock.
This can be shown by enabling CONFIG_DEBUG_WW_MUTEX_SLOWPATH, which throws
the following recursive locking situation:
weston/3440 is trying to acquire lock:
ffff000000e235a0 (reservation_ww_class_mutex){+.+.}-{3:3}, at: drm_gem_shmem_pin+0x34/0xb8 [drm_shmem_helper]
but task is already holding lock:
ffff000000e235a0 (reservation_ww_class_mutex){+.+.}-{3:3}, at: drm_gem_pin+0x2c/0x80 [drm]
Fix it by replacing drm_gem_shmem_pin with its locked version, as the lock
had already been taken by drm_gem_pin().
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Cc: Boris Brezillon <boris.brezillon@collabora.com>
Cc: Steven Price <steven.price@arm.com>
Fixes: a780278472 ("drm/gem: Acquire reservation lock in drm_gem_{pin/unpin}()")
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240523113236.432585-2-adrian.larumbe@collabora.com
Linux 6.9-rc5
I've had a persistent msm failure on clang, and the fix is in fixes
so just pull it back to fix that.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Increase the timeout value to prevent system logs on Amlogic boards flooding
with power transition warnings:
[ 13.047638] panfrost ffe40000.gpu: shader power transition timeout
[ 13.048674] panfrost ffe40000.gpu: l2 power transition timeout
[ 13.937324] panfrost ffe40000.gpu: shader power transition timeout
[ 13.938351] panfrost ffe40000.gpu: l2 power transition timeout
...
[39829.506904] panfrost ffe40000.gpu: shader power transition timeout
[39829.507938] panfrost ffe40000.gpu: l2 power transition timeout
[39949.508369] panfrost ffe40000.gpu: shader power transition timeout
[39949.509405] panfrost ffe40000.gpu: l2 power transition timeout
The 2000 value has been found through trial and error testing with devices
using G52 and G31 GPUs.
Fixes: 22aa1a2090 ("drm/panfrost: Really power off GPU cores in panfrost_gpu_power_off()")
Signed-off-by: Christian Hewitt <christianshewitt@gmail.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240322164525.2617508-1-christianshewitt@gmail.com
To make sure that we don't unintentionally perform any unclocked and/or
unpowered R/W operation on GPU registers, before turning off clocks and
regulators we must make sure that no GPU, JOB or MMU ISR execution is
pending: doing that requires to add a mechanism to synchronize the
interrupts on suspend.
Add functions panfrost_{gpu,job,mmu}_suspend_irq() which will perform
interrupts masking and ISR execution synchronization, and then call
those in the panfrost_device_runtime_suspend() handler in the exact
sequence of job (may require mmu!) -> mmu -> gpu.
As a side note, JOB and MMU suspend_irq functions needed some special
treatment: as their interrupt handlers will unmask interrupts, it was
necessary to add an `is_suspended` bitmap which is used to address the
possible corner case of unintentional IRQ unmasking because of ISR
execution after a call to synchronize_irq().
At resume, clear each is_suspended bit in the reset path of JOB/MMU
to allow unmasking the interrupts.
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231204114215.54575-4-angelogioacchino.delregno@collabora.com
Some SoCs may be equipped with a GPU containing two core groups
and this is exactly the case of Samsung's Exynos 5422 featuring
an ARM Mali-T628 MP6 GPU: the support for this GPU in Panfrost
is partial, as this driver currently supports using only one
core group and that's reflected on all parts of it, including
the power on (and power off, previously to this patch) function.
The issue with this is that even though executing the soft reset
operation should power off all cores unconditionally, on at least
one platform we're seeing a crash that seems to be happening due
to an interrupt firing which may be because we are calling power
transition only on the first core group, leaving the second one
unchanged, or because ISR execution was pending before entering
the panfrost_gpu_power_off() function and executed after powering
off the GPU cores, or all of the above.
Finally, solve this by:
- Avoid to enable the power transition interrupt on reset; and
- Ignoring the core_mask and ask the GPU to poweroff both core groups
Fixes: 22aa1a2090 ("drm/panfrost: Really power off GPU cores in panfrost_gpu_power_off()")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231204114215.54575-2-angelogioacchino.delregno@collabora.com
It was noticed when setting the Panfrost's DVFS device to the performant
governor, GPU frequency as reported by fdinfo had dropped to 0 permamently.
There are two separate issues causing this behaviour:
- Not initialising the device's current_frequency variable to its original
value during device probe().
- Updating said variable in Panfrost devfreq's get_dev_status() rather
than after the new OPP's frequency had been retrieved in target(), which
meant the old frequency would be assigned instead.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Fixes: f11b0417ee ("drm/panfrost: Add fdinfo support GPU load metrics")
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231125205438.375407-3-adrian.larumbe@collabora.com
All of the MediaTek SoCs supported by Panfrost can completely cut power
to the GPU during full system sleep without any user-noticeable delay
in the resume operation, as shown by measurements taken on multiple
MediaTek SoCs (MT8183/86/92/95).
As an example, for MT8195 - a "before" with only runtime PM operations
(so, without turning on/off regulators), and an "after" executing both
the system sleep .resume() handler and .runtime_resume() (so the time
refers to T_Resume + T_Runtime_Resume):
Average Panfrost-only system sleep resume time, before: ~33500ns
Average Panfrost-only system sleep resume time, after: ~336200ns
Keep in mind that this additional ~308200 nanoseconds delay happens only
in resume from a full system suspend, and not in runtime PM operations,
hence it is acceptable.
Measurements were also taken on MT8186, showing a delay of ~312000 ns.
Testing of this happened on all of the aforementioned MediaTek SoCs, but:
MT8183 got tested only by KernelCI with <=10 suspend/resume cycles
MT8186, MT8192, MT8195 were tested manually with over 100 suspend/resume
cycles with GNOME DE (Mutter + Wayland).
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.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/20231109102543.42971-7-angelogioacchino.delregno@collabora.com
All of the MediaTek SoCs supported by Panfrost can switch the clocks
off and on during system sleep to save some power without any user
experience penalty.
Measurements taken on multiple MediaTek SoCs (MT8183/8186/8192/8195)
show that adding this will not prolong the time that is required to
resume the system in any meaningful way.
As an example, for MT8195 - a "before" with only runtime PM operations
(so, without turning on/off GPU clocks), and an "after" executing both
the system sleep .resume() handler and .runtime_resume() (so the time
refers to T_Resume + T_Runtime_Resume):
Average Panfrost-only system sleep resume time, before: ~28000ns
Average Panfrost-only system sleep resume time, after: ~33500ns
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.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/20231109102543.42971-5-angelogioacchino.delregno@collabora.com
Currently, the GPU is being internally powered off for runtime suspend
and turned back on for runtime resume through commands sent to it, but
note that the GPU doesn't need to be clocked during the poweroff state,
hence it is possible to save some power on selected platforms.
Add suspend and resume handlers for full system sleep and then add
a new panfrost_gpu_pm enumeration and a pm_features variable in the
panfrost_compatible structure: BIT(GPU_PM_CLK_DIS) will be used to
enable this power saving technique only on SoCs that are able to
safely use it.
Note that this was implemented only for the system sleep case and not
for runtime PM because testing on one of my MediaTek platforms showed
issues when turning on and off clocks aggressively (in PM runtime)
resulting in a full system lockup.
Doing this only for full system sleep never showed issues during my
testing by suspending and resuming the system continuously for more
than 100 cycles.
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.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/20231109102543.42971-4-angelogioacchino.delregno@collabora.com
In many cases, soft reset takes more than 1 microsecond, but definitely
less than 10; moreover in the poweron flow, tilers, shaders and l2 will
become ready (each) in less than 10 microseconds as well.
Even in the cases (at least on my platforms, rarely) in which those take
more than 10 microseconds, it's very unlikely to see both soft reset and
poweron to take more than 70 microseconds.
Shorten the polling delay to 10 microseconds to consistently reduce the
runtime resume time of the GPU.
As an indicative example, measurements taken on a MediaTek MT8195 SoC
Average runtime resume time in nanoseconds before this commit:
GDM, user selection up/down: 88435ns
GDM, Text Entry (typing user/password): 91489ns
GNOME Desktop, idling, GKRELLM running: 73200ns
After this commit:
GDM: user selection up/down: 26690ns
GDM: Text Entry (typing user/password): 27917ns
GNOME Desktop, idling, GKRELLM running: 25304ns
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.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/20231109102543.42971-3-angelogioacchino.delregno@collabora.com
Even though soft reset should ideally never fail, during development of
some power management features I managed to get some bits wrong: this
resulted in GPU soft reset failures, where the GPU was never able to
recover, not even after suspend/resume cycles, meaning that the only
way to get functionality back was to reboot the machine.
Perform a hard reset after a soft reset failure to be able to recover
the GPU during runtime (so, without any machine reboot).
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.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/20231109102543.42971-2-angelogioacchino.delregno@collabora.com
The layout of the registers {TILER,SHADER,L2}_PWROFF_LO, used to request
powering off cores, is the same as the {TILER,SHADER,L2}_PWRON_LO ones:
this means that in order to request poweroff of cores, we are supposed
to write a bitmask of cores that should be powered off!
This means that the panfrost_gpu_power_off() function has always been
doing nothing.
Fix powering off the GPU by writing a bitmask of the cores to poweroff
to the relevant PWROFF_LO registers and then check that the transition
(from ON to OFF) has finished by polling the relevant PWRTRANS_LO
registers.
While at it, in order to avoid code duplication, move the core mask
logic from panfrost_gpu_power_on() to a new panfrost_get_core_mask()
function, used in both poweron and poweroff.
Fixes: f3ba91228e ("drm/panfrost: Add initial panfrost driver")
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.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/20231102141507.73481-1-angelogioacchino.delregno@collabora.com
