The spsc_queue_peek function is accessing queue->head which belongs to
the consumer thread and shouldn't be accessed by the producer
This is fixing a rare race condition when destroying entities.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Monk.liu@amd.com
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There is a spelling mistake in a pr_err error message. Fix it. Also
add a space after a comma to clean up a checkpatch warning.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
the direct send message to smc to set hard clokc will failed
when smc clock dpm feature is disabled.
so use function of smu_set_hard_freq_range to replace it.
the function will check feature enablement.
eg: when uclk (mclk) dpm feature is disabled on navi10
[ 300.675901] amdgpu: [powerplay] failed send message: SetHardMinByFreq(28)
param: 0x00020064 response 0xfffffffb
Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use dma_set_mask_and_coherent to set both masks in one go, and remove
the no longer required fallback, as the kernel now always accepts
larger than required DMA masks. Fail the driver probe if we can't
set the DMA mask, as that means the system can only support a larger
mask.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use dma_set_mask_and_coherent to set both masks in one go, and remove
the no longer required fallback, as the kernel now always accepts
larger than required DMA masks. Fail the driver probe if we can't
set the DMA mask, as that means the system can only support a larger
mask.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
amdgpu uses a need_dma32 flag to indicate to the drm core that some
allocations need to be done using GFP_DMA32, but it only checks the
device addressing capabilities to make that decision. Unfortunately
PCIe root ports that have limited addressing exist as well. Use the
dma_addressing_limited instead to also take those into account.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
radeon uses a need_dma32 flag to indicate to the drm core that some
allocations need to be done using GFP_DMA32, but it only checks the
device addressing capabilities to make that decision. Unfortunately
PCIe root ports that have limited addressing exist as well. Use the
dma_addressing_limited instead to also take those into account.
Reviewed-by: Christian König <christian.koenig@amd.com>
Reported-by: Atish Patra <Atish.Patra@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
With softpin we allow the userspace to take control over the GPU virtual
address space. The new capability is relected by a bump of the minor DRM
version. There are a few restrictions for userspace to take into
account:
1. The kernel reserves a bit of the address space to implement zero page
faulting and mapping of the kernel internal ring buffer. Userspace can
query the kernel for the first usable GPU VM address via
ETNAVIV_PARAM_SOFTPIN_START_ADDR.
2. We only allow softpin on GPUs, which implement proper process
separation via PPAS. If softpin is not available the softpin start
address will be set to ~0.
3. Softpin is all or nothing. A submit using softpin must not use any
address fixups via relocs.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
Allow the mapping code to request a specific virtual address for the gem
mapping. If the virtual address is zero we fall back to the old mode of
allocating a virtual address for the mapping.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
With per-process address spaces in place, a rogue process submitting
bogus command streams can only hurt itself. There is no need to
validate the command stream before execution anymore.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
This builds on top of the MMU contexts introduced earlier. Instead of having
one context per GPU core, each GPU client receives its own context.
On MMUv1 this still means a single shared pagetable set is used by all
clients, but on MMUv2 there is now a distinct set of pagetables for each
client. As the command fetch is also translated via the MMU on MMUv2 the
kernel command ringbuffer is mapped into each of the client pagetables.
As the MMU context switch is a bit of a heavy operation, due to the needed
cache and TLB flushing, this patch implements a lazy way of switching the
MMU context. The kernel does not have its own MMU context, but reuses the
last client context for all of its operations. This has some visible impact,
as the GPU can now only be started once a client has submitted some work and
we got the client MMU context assigned. Also the MMU context has a different
lifetime than the general client context, as the GPU might still execute the
kernel command buffer in the context of a client even after the client has
completed all GPU work and has been terminated. Only when the GPU is runtime
suspended or switches to another clients MMU context is the old context
freed up.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
In preparation to having a context per process, etnaviv_gem_mapping_get
should not use the current GPU context, but needs to be told which
context to use.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
Move buffer setup and starting of the FE loop in the kernel ringbuffer
into a separate function. This is a preparation to start the FE later
in the submit process.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
This reworks the MMU handling to make it possible to have multiple MMU contexts.
A context is basically one instance of GPU page tables. Currently we have one
set of page tables per GPU, which isn't all that clever, as it has the
following two consequences:
1. All GPU clients (aka processes) are sharing the same pagetables, which means
there is no isolation between clients, but only between GPU assigned memory
spaces and the rest of the system. Better than nothing, but also not great.
2. Clients operating on the same set of buffers with different etnaviv GPU
cores, e.g. a workload using both the 2D and 3D GPU, need to map the used
buffers into the pagetable sets of each used GPU.
This patch reworks all the MMU handling to introduce the abstraction of the
MMU context. A context can be shared across different GPU cores, as long as
they have compatible MMU implementations, which is the case for all systems
with Vivante GPUs seen in the wild.
As MMUv1 is not able to change pagetables on the fly, without a
"stop the world" operation, which stops GPU, changes pagetables via CPU
interaction, restarts GPU, the implementation introduces a shared context on
MMUv1, which is returned whenever there is a request for a new context.
This patch assigns a MMU context to each GPU, so on MMUv2 systems there is
still one set of pagetables per GPU, but due to the shared context MMUv1
systems see a change in behavior as now a single pagetable set is used
across all GPU cores.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
If a MMU is shared between multiple GPUs, all of them need to flush their
TLBs, so a single marker that gets reset on the first flush won't do.
Replace the flush marker with a sequence number, so that it's possible to
check if the TLB is in sync with the current page table state for each GPU.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
There is no need for each GPU to have it's own cmdbuf suballocation
region. Only allocate a single one for the the etnaviv virtual device
and share it across all GPUs.
As the suballoc space is now potentially shared by more hardware jobs
running in parallel, double its size to 512KB to avoid contention.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
This allows to decouple the cmdbuf suballocator create and mapping
the region into the GPU address space. Allowing multiple AS to share
a single cmdbuf suballoc.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
Remember if the GPU has been sucessfully initialized. Only in that case
do we need to clean up various structures in the unbind path. If the
GPU hasn't been sucessfully initialized all the cleanups should happen
in the failure paths of the init function.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Guido Günther <agx@sigxcpu.org>
Due to the tracking provided by the scheduler we know exactly which
submit is failing. Only dump this single submit and the required
auxiliary information. This cuts down the size of the devcoredumps
by only including relevant information.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
TTM provides a means to assign eviction priorities to buffer object. This
means that all buffer objects with a lower priority will be evicted first
on memory pressure.
Use this to make sure surfaces and in particular non-dirty surfaces are
evicted first. Evicting in particular shaders, cotables and contexts imply
a significant performance hit on vmwgfx, so make sure these resources are
evicted last.
Some buffer objects are sub-allocated in user-space which means we can have
many resources attached to a single buffer object or resource. In that case
the buffer object is given the highest priority of the attached resources.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
Drop use of the deprecated drmP.h file from the
remaining files.
In several cases the drmP.h include could be removed without
furter fixes. Other files required a few header files to be added.
In all files divided includes files in blocks and sort them.
v2:
- fix warning in i386 build wiht HIGHMEM disabled
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Reported-by: kbuild test robot <lkp@intel.com> [warning in i386 build]
Cc: VMware Graphics <linux-graphics-maintainer@vmware.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
Signed-off-by: Deepak Rawat <drawat@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
To facilitate removal of drmP.h in the .c
files remove the use from header files first.
Fix fallout in the other files.
Sorted include files in blocks and sorted files
within each block in alphabetical order.
This revealed a dependency from an uapi header to a header
located below drivers/gpu/drm/vmwgfx/.
Added FIXME to remind someone to fix this.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: VMware Graphics <linux-graphics-maintainer@vmware.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
Signed-off-by: Deepak Rawat <drawat@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
At one point, the GPU command verifier and user-space handle manager
couldn't properly protect GPU clients from accessing each other's data.
Instead there was an elaborate mechanism to make sure only the active
master's primary clients could render. The other clients were either
put to sleep or even killed (if the master had exited). VRAM was
evicted on master switch. With the advent of render-node functionality,
we relaxed the VRAM eviction, but the other mechanisms stayed in place.
Now that the GPU command verifier and ttm object manager properly
isolates primary clients from different master realms we can remove the
master switch related code and drop those legacy features.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
I -thought- I had fixed this entirely, but it looks like that I didn't
test this thoroughly enough as we apparently still make one big mistake
with nv50_msto_atomic_check() - we don't handle the following scenario:
* CRTC #1 has n VCPI allocated to it, is attached to connector DP-4
which is attached to encoder #1. enabled=y active=n
* CRTC #1 is changed from DP-4 to DP-5, causing:
* DP-4 crtc=#1→NULL (VCPI n→0)
* DP-5 crtc=NULL→#1
* CRTC #1 steals encoder #1 back from DP-4 and gives it to DP-5
* CRTC #1 maintains the same mode as before, just with a different
connector
* mode_changed=n connectors_changed=y
(we _SHOULD_ do VCPI 0→n here, but don't)
Once the above scenario is repeated once, we'll attempt freeing VCPI
from the connector that we didn't allocate due to the connectors
changing, but the mode staying the same. Sigh.
Since nv50_msto_atomic_check() has broken a few times now, let's rethink
things a bit to be more careful: limit both VCPI/PBN allocations to
mode_changed || connectors_changed, since neither VCPI or PBN should
ever need to change outside of routing and mode changes.
Changes since v1:
* Fix accidental reversal of clock and bpp arguments in
drm_dp_calc_pbn_mode() - William Lewis
Signed-off-by: Lyude Paul <lyude@redhat.com>
Reported-by: Bohdan Milar <bmilar@redhat.com>
Tested-by: Bohdan Milar <bmilar@redhat.com>
Fixes: 232c9eec41 ("drm/nouveau: Use atomic VCPI helpers for MST")
References: 412e85b605 ("drm/nouveau: Only release VCPI slots on mode changes")
Cc: Lyude Paul <lyude@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Airlie <airlied@redhat.com>
Cc: Jerry Zuo <Jerry.Zuo@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Juston Li <juston.li@intel.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Karol Herbst <karolherbst@gmail.com>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: <stable@vger.kernel.org> # v5.1+
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190809005307.18391-1-lyude@redhat.com
This driver requires imported PRIME buffers to appear contiguously in
its IO address space. Make sure this is the case by setting the maximum
DMA segment size to a more suitable value than the default 64KB.
Signed-off-by: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Tomasz Figa <tfiga@chromium.org>
Signed-off-by: CK Hu <ck.hu@mediatek.com>
PRIME buffers should be imported using the DMA device. To this end, use
a custom import function that mimics drm_gem_prime_import_dev(), but
passes the correct device.
Fixes: 119f517362 ("drm/mediatek: Add DRM Driver for Mediatek SoC MT8173.")
Signed-off-by: Alexandre Courbot <acourbot@chromium.org>
Signed-off-by: CK Hu <ck.hu@mediatek.com>
The spsc_queue_peek function is accessing queue->head which belongs to
the consumer thread and shouldn't be accessed by the producer
This is fixing a rare race condition when destroying entities.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Monk.liu@amd.com
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
