It's only likely to paper over bugs.  Unlike the gpu, where we want to
keep things alive a bit longer in expectation of the next frame's
submit, when the display is shut down we can power off immediately.
Signed-off-by: NRob Clark <robdclark@gmail.com>
Acked-by: NArchit Taneja <architt@codeaurora.org>

Signed-off-by: NRob Clark <robdclark@gmail.com>

Note we need to move update_fences() to after msm_rd_dump_submit(),
otherwise the bo's referenced by the submit may no longer be valid.
Signed-off-by: NRob Clark <robdclark@gmail.com>

We need this if we want to dump the submit after cleanup (ie. from hang
or fault).  But in the backoff/unpin case we want to clear them.  So add
a flag so we can skip clearing the IOVAs in at cleanup.
Signed-off-by: NRob Clark <robdclark@gmail.com>

For faults or hangs, it is nice to be able to include a bit more
information.
Signed-off-by: NRob Clark <robdclark@gmail.com>

Split into two instances, the existing $debugfs/rd which continues to
dump all submits, and $debugfs/hangrd which will be used to dump just
submits that cause gpu hangs (and eventually faults, but that will
require some iommu framework enhancements).
Signed-off-by: NRob Clark <robdclark@gmail.com>

Prep work for adding a debugfs file that dumps just submits which
trigger hangs/faults.  In this case the bo may already be in the
MADV_DONTNEED state, but will be still on the active list (since
the submit hasn't completed yet).  So the normal check that the
bo is in the WILLNEED state does not apply.  (But of course the bo
should definitely not be in the PURGED state!)
Signed-off-by: NRob Clark <robdclark@gmail.com>

Now that freedreno gallium driver defaults to using submit_queue task
(render reordering), just showing task->comm is not so useful (ie. it is
always "flush_queue:0"), so also dump the cmdline.  This should also be
more useful for piglit/shader_runner.
Signed-off-by: NRob Clark <robdclark@gmail.com>

Currently the rd dump avoids any buffers marked as WRITE under
the assumption that the contents are not interesting.  While it
is true that the contents are uninteresting we should still print
the iova and size for all buffers so that any listening replay
tools can correctly construct the submission.

Print the header for all buffers but only dump the contents for
buffers marked as READ.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

Recent changes to locking have rendered struct_mutex_task
unused.

Unused since 0e08270a.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

Implement preemption for A5XX targets - this allows multiple
ringbuffers for different priorities with automatic preemption
of a lower priority ringbuffer if a higher one is ready.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

We use a global ringbuffer size and block size for all targets and
at least for 5XX preemption we need to know the value the RB_CNTL
in several locations so it makes sense to calculate it once and use
it everywhere.

The only monkey wrench is that we need to disable the RPTR shadow
for A430 targets but that only needs to be done once and doesn't
affect A5XX so we can or in the value at init time.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

Add a shadow pointer to track the current command being written into
the ring. Don't commit it as 'cur' until the command is submitted.
Because 'cur' is used to construct the software copy of the wptr this
ensures that somebody peeking in on the ring doesn't assume that a
command is inflight while it is being written. This isn't a huge deal
with a single ring (though technically the hangcheck could assume
the system is prematurely busy when it isn't) but it will be rather
important for preemption where the decision to preempt is based
on a non-empty ringbuffer. Without a shadow an aggressive preemption
scheme could assume that the ringbuffer is non empty and switch to it
before the CPU is done writing the command and boom.

Even though preemption won't be supported for all targets because of
the way the code is organized it is simpler to make this generic for
all targets. The extra load for non-preemption targets should be
minimal.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

In order to manage ringbuffer priority to its fullest userspace
should know how many ringbuffers it has to work with. Add a
parameter to return the number of active rings.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

Add the infrastructure to support the idea of multiple ringbuffers.
Assign each ringbuffer an id and use that as an index for the various
ring specific operations.

The biggest delta is to support legacy fences. Each fence gets its own
sequence number but the legacy functions expect to use a unique integer.
To handle this we return a unique identifier for each submission but
map it to a specific ring/sequence under the covers. Newer users use
a dma_fence pointer anyway so they don't care about the actual sequence
ID or ring.

The actual mechanics for multiple ringbuffers are very target specific
so this code just allows for the possibility but still only defines
one ringbuffer for each target family.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

When we move to multiple ringbuffers we're going to store the data
in the memptrs on a per-ring basis. In order to prepare for that
move the current memptrs from the adreno namespace into msm_gpu.
This is way cleaner and immediately lets us kill off some sub
functions so there is much less cost later when we do move to
per-ring structs.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

Currently the behavior of a command stream is provided by the user
application during submission and the application is expected to internally
maintain the settings for each 'context' or 'rendering queue' and specify
the correct ones.

This works okay for simple cases but as applications become more
complex we will want to set context specific flags and do various
permission checks to allow certain contexts to enable additional
privileges.

Add kernel-side submit queues to be analogous to 'contexts' or
'rendering queues' on the application side. Each file descriptor
instance will maintain its own list of queues. Queues cannot be
shared between file descriptors.

For backwards compatibility context id '0' is defined as a default
context specifying no priority and no special flags. This is
intended to be the usual configuration for 99% of applications so
that a garden variety application can function correctly without
creating a queue. Only those applications requiring the specific
benefit of different queues need create one.
Signed-off-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

Signed-off-by: NRob Clark <robdclark@gmail.com>

Signed-off-by: NRob Clark <robdclark@gmail.com>

We already have, as a result of upstreaming the gpu bindings,
msm_clk_get() which will try to get the clock both without and with a
"_clk" suffix.  Use this in HDMI code so we can drop the "_clk" suffix
in bindings while maintaing backwards compatibility.
Signed-off-by: NRob Clark <robdclark@gmail.com>
Reviewed-by: NSean Paul <seanpaul@chromium.org>

We already have, as a result of upstreaming the gpu bindings,
msm_clk_get() which will try to get the clock both without and with a
"_clk" suffix.  Use this in eDP code so we can drop the "_clk" suffix
in bindings while maintaing backwards compatibility.
Signed-off-by: NRob Clark <robdclark@gmail.com>
Reviewed-by: NSean Paul <seanpaul@chromium.org>

We already have, as a result of upstreaming the gpu bindings,
msm_clk_get() which will try to get the clock both without and with a
"_clk" suffix.  Use this in DSI code so we can drop the "_clk" suffix
in bindings while maintaing backwards compatibility.
Signed-off-by: NRob Clark <robdclark@gmail.com>
Reviewed-by: NSean Paul <seanpaul@chromium.org>

This is useful to see in the log, without requiring drm.debug.
Signed-off-by: NRob Clark <robdclark@gmail.com>

When firmware was added to linux-firmware, it was put in a qcom sub-
directory, unlike what we'd been using before.  For a300_pfp.fw and
a300_pm4.fw symlinks were created, but we'd prefer not to have to do
this in the future.  So add support to look in both places when
loading firmware.
Signed-off-by: NRob Clark <robdclark@gmail.com>

Prep work for the next patch.
Signed-off-by: NRob Clark <robdclark@gmail.com>

Previously, in an effort to defer initializing the gpu until firmware
was available (ie. rootfs mounted), the gpu was not loaded at when the
subdevice was bound.  Which resulted that clks/etc were requested in a
place that devm couldn't really help unwind if something failed.

Instead move request_firmware() to gpu->hw_init() and construct the gpu
earlier in adreno_bind().  To avoid the rest of the driver needing to
be aware of a gpu that hasn't managed to load firmware and hw_init()
yet, stash the gpu ptr in the adreno device's drvdata, and don't set
priv->gpu() until hw_init() succeeds.
Signed-off-by: NRob Clark <robdclark@gmail.com>

This was used as a placeholder. It was never really input to the MDSS/HDMI
clocks.
Signed-off-by: NArchit Taneja <architt@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

We need to call reservation_object_reserve_shared() in both cases, but
this wasn't happening in the _NO_IMPLICIT submit case.

Fixes: f0a42bb5 ("drm/msm: submit support for in-fences")
Reported-by: NJordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: NRob Clark <robdclark@gmail.com>

drm_mode_create_lease

	Creates a lease for a list of drm mode objects, returning an
	fd for the new drm_master and a 64-bit identifier for the lessee

drm_mode_list_lesees

	List the identifiers of the lessees for a master file

drm_mode_get_lease

	List the leased objects for a master file

drm_mode_revoke_lease

	Erase the set of objects managed by a lease.

This should suffice to at least create and query leases.

Changes for v2 as suggested by Daniel Vetter <daniel.vetter@ffwll.ch>:

 * query ioctls only query the master associated with
   the provided file.

 * 'mask_lease' value has been removed

 * change ioctl has been removed.

Changes for v3 suggested in part by Dave Airlie <airlied@gmail.com>

 * Add revoke ioctl.

Changes for v4 suggested by Dave Airlie <airlied@gmail.com>

 * Expand on the comment about the magic use of &drm_lease_idr_object
 * Pad lease ioctl structures to align on 64-bit boundaries

Changes for v5 suggested by Dave Airlie <airlied@gmail.com>

 * Check for non-negative object_id in create_lease to avoid debug
   output from the kernel.

Changes for v6 provided by Dave Airlie <airlied@gmail.com>

 * For non-universal planes add primary/cursor planes to lease

   If we aren't exposing universal planes to this userspace client,
   and it requests a lease on a crtc, we should implicitly export the
   primary and cursor planes for the crtc.

   If the lessee doesn't request universal planes, it will just see
   the crtc, but if it does request them it will then see the plane
   objects as well.

   This also moves the object look ups earlier as a side effect, so
   we'd exit the ioctl quicker for non-existant objects.

 * Restrict leases to crtc/connector/planes.

   This only allows leasing for objects we wish to allow.

Changes for v7 provided by Dave Airlie <airlied@gmail.com>

 * Check pad args are 0
 * Check create flags and object count are valid.
 * Check return from fd allocation
 * Refactor lease idr setup and add some simple validation
 * Use idr_mutex uniformly (Keith)
Signed-off-by: NKeith Packard <keithp@keithp.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Attempts to modify un-leased objects are rejected with an error.
Information returned about unleased objects is modified to make them
appear unusable and/or disconnected.

Changes for v2 as suggested by Daniel Vetter <daniel.vetter@ffwll.ch>:

 * With the change in the __drm_mode_object_find API to pass the
   file_priv along, we can now centralize most of the lease-based
   access checks in that function.

 * A few places skip that API and require in-line checks.

Changes for v3 provided by Dave Airlie <airlied@redhat.com>

 * remove support for leasing encoders.
 * add support for leasing planes.

Changes for v4

 * Only call drm_lease_held if DRIVER_MODESET.
Signed-off-by: NKeith Packard <keithp@keithp.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

This provides new data structures to hold "lease" information about
drm mode setting objects, and provides for creating new drm_masters
which have access to a subset of the available drm resources.

An 'owner' is a drm_master which is not leasing the objects from
another drm_master, and hence 'owns' them.

A 'lessee' is a drm_master which is leasing objects from some other
drm_master. Each lessee holds the set of objects which it is leasing
from the lessor.

A 'lessor' is a drm_master which is leasing objects to another
drm_master. This is the same as the owner in the current code.

The set of objects any drm_master 'controls' is limited to the set of
objects it leases (for lessees) or all objects (for owners).

Objects not controlled by a drm_master cannot be modified through the
various state manipulating ioctls, and any state reported back to user
space will be edited to make them appear idle and/or unusable. For
instance, connectors always report 'disconnected', while encoders
report no possible crtcs or clones.

The full list of lessees leasing objects from an owner (either
directly, or indirectly through another lessee), can be searched from
an idr in the drm_master of the owner.

Changes for v2 as suggested by Daniel Vetter <daniel.vetter@ffwll.ch>:

* Sub-leasing has been disabled.

* BUG_ON for lock checking replaced with lockdep_assert_held

* 'change' ioctl has been removed.

* Leased objects can always be controlled by the lessor; the
  'mask_lease' flag has been removed

* Checking for leased status has been simplified, replacing
  the drm_lease_check function with drm_lease_held.

Changes in v3, some suggested by Dave Airlie <airlied@gmail.com>

* Add revocation. This allows leases to be effectively revoked by
  removing all of the objects they have access to. The lease itself
  hangs around as it's hanging off a file.

* Free the leases IDR when the master is destroyed

* _drm_lease_held should look at lessees, not lessor

* Allow non-master files to check for lease status

Changes in v4, suggested by Dave Airlie <airlied@gmail.com>

* Formatting and whitespace changes

Changes in v5 (airlied)

* check DRIVER_MODESET before lease destroy call
* check DRIVER_MODESET for lease revoke (Chris)
* Use idr_mutex uniformly for all lease elements of struct drm_master. (Keith)
Signed-off-by: NKeith Packard <keithp@keithp.com>

Separate out lease debugging from the core.
Signed-off-by: NKeith Packard <keithp@keithp.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

These provide crtc-id based functions instead of pipe-number, while
also offering higher resolution time (ns) and wider frame count (64)
as required by the Vulkan API.

v2:

 * Check for DRIVER_MODESET in new crtc-based vblank ioctls

	Failing to check this will oops the driver.

 * Ensure vblank interupt is running in crtc_get_sequence ioctl

	The sequence and timing values are not correct while the
	interrupt is off, so make sure it's running before asking for
	them.

 * Short-circuit get_sequence if the counter is enabled and accurate

	Steal the idea from the code in wait_vblank to avoid the
	expense of drm_vblank_get/put

 * Return active state of crtc in crtc_get_sequence ioctl

	Might be useful for applications that aren't in charge of
	modesetting?

 * Use drm_crtc_vblank_get/put in new crtc-based vblank sequence ioctls

	Daniel Vetter prefers these over the old drm_vblank_put/get
	APIs.

 * Return s64 ns instead of u64 in new sequence event
Suggested-by: NDaniel Vetter <daniel@ffwll.ch>
Suggested-by: NVille Syrjälä <ville.syrjala@linux.intel.com>

v3:

 * Removed FIRST_PIXEL_OUT_FLAG
 * Document that the timestamp in the query and event are
   that of the first pixel leaving the display engine for
   the display (using the same wording as the Vulkan spec).
Suggested-by: NMichel Dänzer <michel@daenzer.net>
Acked-by: NDave Airlie <airlied@redhat.com>

[airlied: left->leaves (Michel)]
Signed-off-by: NKeith Packard <keithp@keithp.com>
Reviewed-by: NSean Paul <seanpaul@chromium.org>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The feature implementation isn't stable yet. Reject any attempt to use
the IOCTLs for now. This keeps most of the code in place, so we can stabilize
it in-tree, but keeps userspace from using the feature for now.
Signed-off-by: NLucas Stach <l.stach@pengutronix.de>

The performance monitoring feature isn't stable enough yet, so don't advertise
it to userspace yet.
Signed-off-by: NLucas Stach <l.stach@pengutronix.de>

Place drm_event_vblank in a new union that includes that and a bare
drm_event structure. This will allow new members of that union to be
added in the future without changing code related to the existing vbl
event type.

Assignments to the crtc_id field are now done when the event is
allocated, rather than when delievered. This way, delivery doesn't
need to have the crtc ID available.

v2:
 * Remove 'dev' argument from create_vblank_event

	It wasn't being used anyways, and if we need it in the future,
	we can always get it from crtc->dev.

 * Check for MODESETTING before looking for crtc in queue_vblank_event

	UMS drivers will oops if we try to get a crtc, so make sure
	we're modesetting before we try to find a crtc_id to fill into
	the event.
Signed-off-by: NKeith Packard <keithp@keithp.com>
Reviewed-by: NSean Paul <seanpaul@chromium.org>
Signed-off-by: NDave Airlie <airlied@redhat.com>
(cherry picked from commit dc695b85fde88eca3ef3b03fcd82f15b6bc6e462)

This modifies the datatypes used by the vblank code to provide 64 bits
of vblank count.

The driver interfaces have been left using 32 bits of vblank count;
all of the code necessary to widen that value for the user API was
already included to handle devices returning fewer than 32-bits.

This will provide the necessary datatypes for the Vulkan API.

v2:

 * Re-write wait_vblank ioctl to ABSOLUTE sequence

    When an application uses the WAIT_VBLANK ioctl with RELATIVE
    or NEXTONMISS bits set, the target vblank interval is updated
    within the kernel. We need to write that target back to the
    ioctl buffer and update the flags bits so that if the wait is
    interrupted by a signal, when it is re-started, it will target
    precisely the same vblank count as before.

 * Leave driver API with 32-bit vblank count

v3:

 * Rebase on top of Arnd Bergmann's patch which had
   the switch to ktime_t parts.

[airlied: fix conflict with Ville vblank change].
Suggested-by: NMichel Dänzer <michel@daenzer.net>
Suggested-by: NDaniel Vetter <daniel@ffwll.ch>
Signed-off-by: NKeith Packard <keithp@keithp.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>
(cherry picked from commit 2affbc16983e4fc90960bc7f70e7615f4228199b)

In order to support CEC the HDMI driver has to inform the CEC driver
whenever the physical address changes. So when the EDID is read the
CEC driver has to be informed and whenever the hotplug detect goes
away.

This is done through the cec-notifier framework.

The link between the HDMI driver and the CEC driver is done through
the hdmi-phandle property in the tegra-cec node in the device tree.
Signed-off-by: NHans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: NThierry Reding <treding@nvidia.com>

In order for the reset to be applied properly, the module clock must be
enabled during the assertion.
Signed-off-by: NThierry Reding <treding@nvidia.com>

When applying the PLL changes from the computed state object, make sure
to set the rate of the display controller module clock. Failing to do so
can yield to a situation where the parent will be set to the proper
pixel clock, but the module clock will be divided down to the rate that
is happened to be set to before the parent rate change.
Signed-off-by: NThierry Reding <treding@nvidia.com>