When calling smmu mapping, if iova is specified directly by user, not
allocated dynamically in dma-mapping.c, smmu driver needs to provide
support for this. This is needed in early display case. In this scenario,
LK has set physical memory to display hardware for fetching, so if iova
is not explicitly specified in kernel, but instead dynamically produced
by "alloc_iova" in dma-mapping.c, display hardware has no chance to know
this new iova, then smmu fault will happen if enabling the iommu stage-1
translation.
To fix this smmu fault problem, add re-routing to the right path when
iova specified by user is not 0 in smmu map/unmap function.
Change-Id: I062b04d7eec65af1c106a5caa09ec787b5d26d0d
Signed-off-by: Guchun Chen <guchunc@codeaurora.org>
When HDMI resolution is bigger than 2560 pixel of width, driver
needs to use two hardware pipes. Use virtual plane to support this feature.
Change-Id: I19e3bb32aa2a16c83393b0e3c6bec3db03827eca
Signed-off-by: wyun <wyun@codeaurora.org>
At a higher level, all objects are created with definite size i.e. 0 is
illegal. In forthcoming patches, this assumption is dependent upon in
the drm_mm range manager, i.e. trying to create a drm_mm node with size
0 will have undefined behaviour. Add a couple of WARNs upon creating the
drm_mm node to prevent later bugs.
Change-Id: Ieb81a8ef5a05a32dd37dfc040ea3f78243b59161
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470248788-30873-1-git-send-email-chris@chris-wilson.co.uk
Git-commit: aafdcfd3f9aa3c9f77ae4e9385f21bf9ae120d3e
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
As we always add this to the drm_mm->hole_stack as our first operation,
we do not need to initialise the list node.
Change-Id: If2e9e06141bdb9a9411ef875cb83e41a48e5f1a9
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470236651-678-3-git-send-email-chris@chris-wilson.co.uk
Git-commit: a787900564ffc6eebbcb7086b45e04a0bc3370f0
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
In addition to the last-in/first-out stack for accessing drm_mm nodes,
we occasionally and in the future often want to find a drm_mm_node by an
address. To do so efficiently we need to track the nodes in an interval
tree - lookups for a particular address will then be O(lg(N)), where N
is the number of nodes in the range manager as opposed to O(N).
Insertion however gains an extra O(lg(N)) step for all nodes
irrespective of whether the interval tree is in use. For future i915
patches, eliminating the linear walk is a significant improvement.
v2: Use generic interval-tree template for u64 and faster insertion.
Change-Id: Iddcb7891480ca7d6e0469c8d394fcd1962ed4583
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470236651-678-1-git-send-email-chris@chris-wilson.co.uk
Git-commit: 202b52b7fbf70858609ec20829c7d69a13ffa351
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Buffer object specific resources like pages, domains, sg list
need not be protected with struct_mutex. They can be protected
with a buffer object level lock. This simplifies locking and
makes it easier to avoid potential recursive locking scenarios
for SVM involving mmap_sem and struct_mutex. This also removes
unnecessary serialization when creating buffer objects, and also
between buffer object creation and GPU command submission.
Change-Id: I40cb437d0186c3d9aac365c9baba0aa4792f0aa1
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Define a 36-bit address space for TTBR1 which is used for
kernel side GPU buffer objects.
Change-Id: I1c4eaee0fd92236793621c7d3dba1700e56fefd2
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
get_pages doesn't keep a reference of the pages allocated
when it fails later in the code path. This can lead to
a memory leak. Keep reference of the allocated pages so
that it can be freed when msm_gem_free_object gets called
later during cleanup.
Change-Id: I44dea20e10c056c1c5e80d01bab0a274ff768b1c
Signed-off-by: Prakash Kamliya <pkamliya@codeaurora.org>
In function submit_create, if nr_cmds or nr_bos is assigned with
negative value, the allocated buffer may be small than intended.
Using this buffer will lead to buffer overflow issue.
Change-Id: I0b61cccffd836e2dd3c859446470af4b6451b9ed
Signed-off-by: Kasin Li <donglil@codeaurora.org>
Return an appropriate error code if the default pagetable is NULL.
Change-Id: Ic88b066c40a8f840d95fd3fbc9ee9274c428b66a
Signed-off-by: Lynus Vaz <lvaz@codeaurora.org>
For any cache operation, the current code tries to map all pages to the
kernel using vmap in case sg table is not available and then performs
the requested cache operation. If vmap fails because of memory crunch
ioctl just returns failure.
This change avoids using vmap and performs per page cache operation
even when sg table is not available. This is done to avoid failures
because of vmap especially on 32 bit systems.
Change-Id: I123b46e6a55a62cbf934ab6a2a49dcd1f0d4c7d4
Signed-off-by: Deepak Kumar <dkumar@codeaurora.org>
commit e345da82bd6bdfa8492f80b3ce4370acfd868d95 upstream.
The builtin eDP panel in the HP zBook 17 G2 supports 10 bpc,
as advertised by the Laptops product specs and verified via
injecting a fixed edid + photometer measurements, but edid
reports unknown depth, so drivers fall back to 6 bpc.
Add a quirk to get the full 10 bpc.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1492787108-23959-1-git-send-email-mario.kleiner.de@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1b0f84380b10ee97f7d2dd191294de9017e94d1d upstream.
If the time to the next alarm is short enough, we could race with HW and
end up with an ~4 second delay until it triggers.
Fix this by checking again after we update HW.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 330bdf62fe6a6c5b99a647f7bf7157107c9348b3 upstream.
The idea here was to avoid having to "manually" program the HW if there's
a new earliest alarm. This was lazy and bad, as it leads to loads of fun
races between inter-related callers (ie. therm).
Turns out, it's not so difficult after all. Go figure ;)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9fc64667ee48c9a25e7dca1a6bcb6906fec5bcc5 upstream.
At least therm/fantog "attempts" to work around this issue, which could
lead to corruption of the pending alarm list.
Fix it properly by not updating the timestamp without the lock held, or
trying to add an already pending alarm to the pending alarm list....
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3733bd8b407211739e72d051e5f30ad82a52c4bc upstream.
Fixes a race where we can miss an alarm that triggers while we're already
processing previous alarms.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e4311ee51d1e2676001b2d8fcefd92bdd79aad85 upstream.
These were ineffective due to touching the list without the alarm lock,
but should no longer be required.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit d63c277dc672e0c568481af043359420fa9d4736 upstream.
Avoid big roundoff errors in scanline/hactive durations for
high pixel clocks, especially for >= 500 Mhz, and thereby
program more accurate display fifo watermarks.
Implemented here for DCE 6,8,10,11.
Successfully tested on DCE 10 with AMD R9 380 Tonga.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e190ed1ea7458e446230de4113cc5d53b8dc4ec8 upstream.
At dot clocks > approx. 250 Mhz, some of these calcs will overflow and
cause miscalculation of latency watermarks, and for some overflows also
divide-by-zero driver crash ("divide error: 0000 [#1] PREEMPT SMP" in
"dce_v10_0_latency_watermark+0x12d/0x190").
This zero-divide happened, e.g., on AMD Tonga Pro under DCE-10,
on a Displayport panel when trying to set a video mode of 2560x1440
at 165 Hz vrefresh with a dot clock of 635.540 Mhz.
Refine calculations to avoid the overflows.
Tested for DCE-10 with R9 380 Tonga + ASUS ROG PG279 panel.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Make HDMI device as the interrupt parent of CEC device, so add
the new IRQ domain in HDMI driver dealing with the necessary
IRQ mapping.
Change-Id: Id935da1d1e488ccee01b831b9f085a83d67268f2
Signed-off-by: Ray Zhang <rayz@codeaurora.org>
Update QoS settings for A508 VBIF based on recommendation.
VBIF_GATE_OFF_WRREQ_EN register needs to be programmed by SW.
Change-Id: I7d41c8350ad09c595f288bd2a2b45fc2abef15f8
Signed-off-by: Rajesh Kemisetti <rajeshk@codeaurora.org>
Return mementry from kgsl_sharedmem_find only if pending_free
is not set for that mementry. This is necessary to avoid use
of a mementry after it is already marked for free.
Change-Id: I23111e9c82a88ccbda2ab259074c38d91f9ff5cb
Signed-off-by: Deepak Kumar <dkumar@codeaurora.org>
If we delete uninitialized timer on CONFIG_DEBUG_OBJECTS
disabled build del_timer_sync() will block for ever. For
all A3xx targets preemption timer is not initialized, but
dispatcher fault handler is trying to delete the
uninitialized preemption timer. Fix this issue by adding
a preemption check before we delete it.
CRs-Fixed: 2023690
Change-Id: I2c51a0b2286b82bf2eb5ee68d923dd9585f07f00
Signed-off-by: Wenbin Wang <wwenbin@codeaurora.org>
Signed-off-by: Abhilash Kumar <krabhi@codeaurora.org>
In async commit case, driver needs to always wait for input fence
before triggering the complete_commit path. Otherwise, there could
be tearing since GPU hasn't finished the composition rendering.
Change-Id: I73a54f5811fdcf8639618ce3cacf4cbaa00b406c
Signed-off-by: Felix Xiong <xayang@codeaurora.org>
Signed-off-by: Jin Li <jinl@codeaurora.org>
Signed-off-by: Yunyun Cao <yunyunc@codeaurora.org>
The SPLIT related registers are only for DSI interfaces. Without
checking the interface type, they could be overwrote by
configurations through HDMI path.
CRs-Fixed: 1085586
Change-Id: I7ace9fd8dfe5ee99cb750b2723e8f22701039552
Signed-off-by: Jin Li <jinl@codeaurora.org>
Signed-off-by: Yunyun Cao <yunyunc@codeaurora.org>
The h/v polarity should always be set from the panel configuration.
For HDMI display, it's from the EDID information. For DSI display,
it's from the panel settings in the dtsi.
CRs-Fixed: 1085021
Change-Id: I3776603d7055e69eb2c8e5003ab83bc0483ab7c8
Signed-off-by: Jin Li <jinl@codeaurora.org>
Signed-off-by: Yunyun Cao <yunyunc@codeaurora.org>
The initial version of the patch save the command submit_time and
queue_time in seconds, but its desired by the users of this profiling
API to return the time in nanoseconds resolution.
Change-Id: I3a56e3ffd3ebe86f51a00a12b7c3e7c4b4c9a956
Signed-off-by: Sharat Masetty <smasetty@codeaurora.org>
At this point, there is nothing left to fail. And submit already has a
fence assigned and is added to the submit_list. Any problems from here
on out are asynchronous (ie. hangcheck/recovery).
Change-Id: Ib6b6bf00099137972649c97cc6cd8c4fe25ce7c3
Signed-off-by: Rob Clark <robdclark@gmail.com>
Git-commit: 1193c3bcb581807d58dd7df90528ec744af387a9
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
[smasetty@codeaurora.org: fixed merge conflict issues; made corresponding
changes to A5XX submit function.]
Signed-off-by: Sharat Masetty <smasetty@codeaurora.org>
commit 3089c1df10e2931b1d72d2ffa7d86431084c86b3 upstream.
The vm fault handler relies on the fact that the VMA owns a reference
to the BO. However, once mmap_sem is released, other tasks are free to
destroy the VMA, which can lead to the BO being freed. Fix two code
paths where that can happen, both related to vm fault retries.
Found via a lock debugging warning which flagged &bo->wu_mutex as
locked while being destroyed.
Fixes: cbe12e74ee ("drm/ttm: Allow vm fault retries")
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Any memory free ioctl doesn't need to be blocked till the
corresponding mementry is destroyed. This change defers
the mementry put to unblock all memory free ioctls immediately.
This is done to reduce the time spent by user applications in
waiting for memory to be freed.
Change-Id: Iaa37ac5dbdedc3d02c41886c2bdf7f3d016176ac
Signed-off-by: Deepak Kumar <dkumar@codeaurora.org>
