raspberrypi inclusion
category: feature
bugzilla: 50432

--------------------------------

The pm_runtime was blocking changelist submission, so delete it as a
temporary workaround.
Signed-off-by: NPhil Elwell <phil@raspberrypi.org>
Signed-off-by: NFang Yafen <yafen@iscas.ac.cn>
Signed-off-by: NZheng Zengkai <zhengzengkai@huawei.com>

raspberrypi inclusion
category: feature
bugzilla: 50432

--------------------------------

My various attempts at re-enabling runtime PM have failed, so just
crank the clock down when V3D is idle to reduce power consumption.
Signed-off-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NFang Yafen <yafen@iscas.ac.cn>
Signed-off-by: NZheng Zengkai <zhengzengkai@huawei.com>

Originally this error path used to leak "bin" but then we accidentally
applied two separate commits to fix it and ended up with a double free.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NMaxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20201026094905.GA1634423@mwanda

Spelling out _unlocked for each and every driver is a annoying.
Especially if we consider how many drivers, do not know (or need to)
about the horror stories involving struct_mutex.

Just drop the suffix. It makes the API cleaner.

Done via the following script:

__from=drm_gem_object_put_unlocked
__to=drm_gem_object_put
for __file in $(git grep --name-only $__from); do
  sed -i  "s/$__from/$__to/g" $__file;
done

Cc: Eric Anholt <eric@anholt.net>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: NEmil Velikov <emil.velikov@collabora.com>
Acked-by: NSam Ravnborg <sam@ravnborg.org>
Acked-by: NThomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20200515095118.2743122-33-emil.l.velikov@gmail.com

We already have it in v3d_dev->drm.dev with zero additional pointer
chasing. Personally I don't like duplicated pointers like this
because:
- reviewers need to check whether the pointer is for the same or
  different objects if there's multiple
- compilers have an easier time too

But also a bit a bikeshed, so feel free to ignore.
Acked-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Cc: Eric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20200415074034.175360-10-daniel.vetter@ffwll.ch

In the impelementation of v3d_submit_cl_ioctl() there are two memory
leaks. One is when allocation for bin fails, and the other is when bin
initialization fails. If kcalloc fails to allocate memory for bin then
render->base should be put. Also, if v3d_job_init() fails to initialize
bin->base then allocated memory for bin should be released.

Fixes: a783a09e ("drm/v3d: Refactor job management.")
Signed-off-by: NNavid Emamdoost <navid.emamdoost@gmail.com>
Reviewed-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20191021185250.26130-1-navid.emamdoost@gmail.com

Extends the user space ioctl for CL submissions so it can include a request
to flush the cache once the CL execution has completed. Fixes memory
write violation messages reported by the kernel in workloads involving
shader memory writes (SSBOs, shader images, scratch, etc) which sometimes
also lead to GPU resets during Piglit and CTS workloads.

v2: if v3d_job_init() fails we need to kfree() the job instead of
    v3d_job_put() it (Eric Anholt).

v3 (Eric Anholt):
  - Drop _FLAG suffix from the new flag name.
  - Add a new param so userspace can tell whether cache flushing is
    implemented in the kernel.
Signed-off-by: NIago Toral Quiroga <itoral@igalia.com>
Reviewed-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190919071016.4578-1-itoral@igalia.com

If the initialization of the job fails we need to kfree() it
before returning.
Signed-off-by: NIago Toral Quiroga <itoral@igalia.com>
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190916071125.5255-1-itoral@igalia.com
Fixes: a783a09e ("drm/v3d: Refactor job management.")
Reviewed-by: NEric Anholt <eric@anholt.net>

Be more consistent with the naming of the other DMA-buf objects.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/323401/

Drop use of the deprecated drmP.h header file.
Made v3d_drv.h self-contained with only sufficient
include files.
Fixed fallout in remaining files.
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>
Acked-by: NEmil Velikov <emil.velikov@collabora.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: NEric Anholt <eric@anholt.net>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20190716064220.18157-3-sam@ravnborg.org

Looking at a hang recently, I noticed these registers that might tell
me if something obvious was wrong.  They didn't help in this case, but
keep it around for the future.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190419001014.23579-3-eric@anholt.netReviewed-by: NPaul Kocialkowski <paul.kocialkowski@bootlin.com>

It is the expectation of existing userspace (X11 + Mesa, in
particular) that jobs submitted to the kernel against a shared BO will
get implicitly synchronized by their submission order.  If we want to
allow clever userspace to disable implicit synchronization, we should
do that under its own submit flag (as amdgpu and lima do).

Note that we currently only implicitly sync for the rendering pass,
not binning -- if you texture-from-pixmap in the binning vertex shader
(vertex coordinate generation), you'll miss out on synchronization.

Fixes flickering when multiple clients are running in parallel,
particularly GL apps and compositors.

v2: Fix a missing refcount on the CSD done fence for L2 cleaning.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190416225856.20264-6-eric@anholt.netAcked-by: NRob Clark <robdclark@gmail.com>

We only set the excl (possible-writing) fence pointer and never add a
shared (read-only) fence.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190416225856.20264-5-eric@anholt.netAcked-by: NRob Clark <robdclark@gmail.com>

The compute shader dispatch interface is pretty simple -- just pass in
the regs that userspace has passed us, with no CLs to run.  However,
with no CL to run it means that we need to do manual cache flushing of
the L2 after the HW execution completes (for SSBO, atomic, and
image_load_store writes that are the output of compute shaders).

This doesn't yet expose the L2 cache's ability to have a region of the
address space not write back to memory (which could be used for
shared_var storage).

So far, the Mesa side has been tested on V3D v4.2 simpenrose (passing
the ES31 tests), and on the kernel side on 7278 (failing atomic
compswap tests in a way that doesn't reproduce on simpenrose).

v2: Fix excessive allocation for the clean_job (reported by Dan
    Carpenter).  Keep refs on jobs until clean_job is finished, to
    avoid spurious MMU errors if the output BOs are freed by userspace
    before L2 cleaning is finished.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190416225856.20264-4-eric@anholt.netAcked-by: NRob Clark <robdclark@gmail.com>

The CL submission had two jobs embedded in an exec struct.  When I
added TFU support, I had to replicate some of the exec stuff and some
of the job stuff.  As I went to add CSD, it became clear that actually
what was in exec should just be in the two CL jobs, and it would let
us share a lot more code between the 4 queues.

v2: Fix missing error path in TFU ioctl's bo[] allocation.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190416225856.20264-3-eric@anholt.netAcked-by: NRob Clark <robdclark@gmail.com>

All consumers wanted drm_gem_object * now.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190416225856.20264-2-eric@anholt.netAcked-by: NRob Clark <robdclark@gmail.com>

We have another thing called the "done fence" that tracks when the
scheduler considers the job done, and having the shared name was
confusing.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190313235211.28995-2-eric@anholt.netReviewed-by: NDave Emett <david.emett@broadcom.com>

The new shmem helpers from Noralf and Rob abstract out a bunch of our
BO creation and mapping code.

v2: Use the new sgt getter, and flag pages as dirty before freeing.
v3: Remove the mismatched put_pages.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190314163451.13431-1-eric@anholt.net
Reviewed-by: Rob Herring <robh@kernel.org> (v2)

Now that we have core helpers, this gets rid of a lot of boilerplate.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190308161716.2466-3-eric@anholt.netAcked-by: NRob Herring <robh@kernel.org>

No compatible string for it yet, just the version-dependent changes.
They've now tied the hub and the core interrupt lines into a single
interrupt line coming out of the block.  It also turns out I made a
mistake in modeling the V3D v3.3 and v4.1 bridge as a part of V3D
itself -- the bridge is going away in favor of an external reset
controller in a larger HW module.

v2: Use consistent checks for whether we're on 4.2, and fix a leak in
    an error path.
v3: Use more general means of determining if the current 4.2 changes
    are in place, as apparently other platforms may switch back (noted
    by Dave).  Update the binding doc.
v4: Improve error handling for IRQ init.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190308174336.7866-2-eric@anholt.netReviewed-by: NDave Emett <david.emett@broadcom.com>

The old field is gone and the register now has a different field,
QRMAXCNT for how many TMU requests get serviced before thread switch.
We were accidentally reducing it from its default of 0x3 (4 requests)
to 0x0 (1).

v2: Skip setting the reg at all on 4.x, instead of trying to update
    only the old field.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190220233658.986-2-eric@anholt.netReviewed-by: NDave Emett <david.emett@broadcom.com>

Now that the base struct drm_gem_object has a reservation_object, use it
and remove the private BO one.

Cc: Eric Anholt <eric@anholt.net>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: David Airlie <airlied@linux.ie>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: NRob Herring <robh@kernel.org>
Acked-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20190202154158.10443-5-robh@kernel.orgSigned-off-by: NMaxime Ripard <maxime.ripard@bootlin.com>

This would be a fairly obscure race, but let's make sure we don't ever
lose it.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20181203222438.25417-6-eric@anholt.netReviewed-by: NDave Emett <david.emett@broadcom.com>

This cache was replaced with the slice accessing the L2T in the newer
generations.  Noted by Dave during review.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20181203222438.25417-5-eric@anholt.netReviewed-by: NDave Emett <david.emett@broadcom.com>

According to Dave, once you've started an L2T flush, all L2T accesses
will be blocked until the flush completes.  This fixes a consistent
3-4ms stall between the ioctl and running the job, and 3DMMES Taiji
goes from 27fps to 110fps.

v2: Leave a note about why we don't need to wait for completion.
Signed-off-by: NEric Anholt <eric@anholt.net>
Fixes: 57692c94 ("drm/v3d: Introduce a new DRM driver for Broadcom V3D V3.x+")
Reviewed-by: NDave Emett <david.emett@broadcom.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181203222438.25417-4-eric@anholt.net

This is the write combiner for TMU writes.  You're supposed to flush
that at job end if you had dirtied any cachelines.  Flushing it at job
start then doesn't make any sense.
Signed-off-by: NEric Anholt <eric@anholt.net>
Fixes: 57692c94 ("drm/v3d: Introduce a new DRM driver for Broadcom V3D V3.x+")
Reviewed-by: NDave Emett <david.emett@broadcom.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181203222438.25417-3-eric@anholt.net

Now that I've specified how the end-of-pipeline flushing should work,
we're never going to use this function.
Signed-off-by: NEric Anholt <eric@anholt.net>
Reviewed-by: NDave Emett <david.emett@broadcom.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181203222438.25417-2-eric@anholt.net

I missed one case during the recent revert of the replace_fence
interface change.

Fixes: 0b258ed1 drm: revert "expand replace_fence to support timeline point v2"
Signed-off-by: NChristian König <christian.koenig@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/266134/

This reverts commit 9a09a423.

The whole interface isn't thought through. Since this function can't
fail we actually can't allocate an object to store the sync point.

Sorry, I should have taken the lead on this from the very beginning and
reviewed it more thoughtfully. Going to propose a new interface as a
follow up change.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NChunming Zhou <david1.zhou@amd.com>
Link: https://patchwork.freedesktop.org/patch/265580/

The core scheduler tells us when the job is pushed to the scheduler's
queue, and I had the job_run functions saying when they actually queue
the job to the hardware.  By adding tracepoints for the very top of
the ioctls and the IRQs signaling job completion, "perf record -a -e
v3d:.\* -e gpu_scheduler:.\* <job>; perf script" gets you a pretty
decent timeline.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20181201005759.28093-5-eric@anholt.netReviewed-by: NDave Emett <david.emett@broadcom.com>

They were unused, as Dave Emett noticed in TFU review.
Signed-off-by: NEric Anholt <eric@anholt.net>
Cc: Dave Emett <david.emett@broadcom.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181128230927.10951-2-eric@anholt.netReviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The TFU can copy from raster, UIF, and SAND input images to UIF output
images, with optional mipmap generation.  This will certainly be
useful for media EGL image input, but is also useful immediately for
mipmap generation without bogging the V3D core down.

For now we only run the queue 1 job deep, and don't have any hang
recovery (though I don't think we should need it, with TFU).  Queuing
multiple jobs in the HW will require synchronizing the YUV coefficient
regs updates since they don't get FIFOed with the job.

v2: Change the ioctl to IOW instead of IOWR, always set COEF0, explain
    why TFU is AUTH, clarify the syncing docs, drop the unused TFU
    interrupt regs (you're expected to use the hub's), don't take
    &bo->base for NULL bos.
v3: Fix a little whitespace alignment (noticed by checkpatch), rebase
    on drm_sched_job_cleanup() changes.
Signed-off-by: NEric Anholt <eric@anholt.net>
Reviewed-by: Dave Emett <david.emett@broadcom.com> (v2)
Link: https://patchwork.freedesktop.org/patch/264607/

The extra to_v3d_bo() calls came from copying this from the vc4
driver, which stored the cma gem object in the structs.

v2: Fix an unused var warning
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20181108161654.19888-4-eric@anholt.net
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com> (v1)

Let's support simultaneous submissions to multiple engines.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NJunwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: NHuang Rui <ray.huang@amd.com>
Link: https://patchwork.kernel.org/patch/10626149/

flags can be used by driver to decide whether need to block wait submission.
Signed-off-by: NChunming Zhou <david1.zhou@amd.com>
SIgned-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.kernel.org/patch/10641339/

Once we push the job, the scheduler could run it and free it.  So, if
we want to reference their fences, we need to grab them before then.
I haven't seen this happen in many days of conformance test runtime,
but let's still close the race.
Signed-off-by: NEric Anholt <eric@anholt.net>
Fixes: 57692c94 ("drm/v3d: Introduce a new DRM driver for Broadcom V3D V3.x+")
Link: https://patchwork.freedesktop.org/patch/254119/Reviewed-by: NBoris Brezillon <boris.brezillon@bootlin.com>

we can place a fence to a timeline point after expanded.
v2: change func parameter order
Signed-off-by: NChunming Zhou <david1.zhou@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/246543/

we can fetch timeline point fence after expanded.
v2: The parameter fence is the result of the function and should come last.
Signed-off-by: NChunming Zhou <david1.zhou@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/246541/

entity has a scheduler field and we don't need the sched argument
in any of the functions where entity is provided.
Signed-off-by: NNayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Since our seqno value comes from a counter associated with the GPU
ring, not the entity (aka client), they'll be completed out of order.
There's actually no need for this code at all, since we don't have
enable_signaling() and thus DMA_FENCE_SIGNALED_BIT will be set before
we could be called.
Signed-off-by: NEric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20180605190302.18279-2-eric@anholt.netReviewed-by: NLucas Stach <l.stach@pengutronix.de>