Under certain circumstances the trapped address will contain subc 7,
which GK104 GR doesn't have anymore.

Notice this case to avoid causing additional priv ring faults.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

No idea if "3" is a constant or derived from something else, but the
value is unchanged in the limited traces of gm107/gm204 I have here.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Make static a few functions and structures that should be.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

A "return 0" found its way in the middle of the error path of
nouveau_platform_probe(), remove it as it will make the kernel crash if
we try to unload the module afterwards.

While we are at it, also remove the IOMMU domain if it has been created,
as we should.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

nvkm_mm_fini() was not called when exiting the driver, resulting in a
memory leak. Fix this.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

On some of these chipsets, reading NV_PGRAPH_GPC_GPM_PD_PES_TPC_ID_MASK
can trigger a PRI fault and return an error code instead of a TPC mask,
unless PGOB has been disabled first.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Before we moved gk110's implementation of this to pmu, the functions were
identical.  This commit just switches GK208 to use the new (more complete)
implementation of the power-up sequence.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Turns out the PTHERM part of this dance is bracketed by the same PMU
fiddling that occurs on GK104/6, let's assume it's also PGOB.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

So we can actually use the full 512 byte code space
Signed-off-by: NRoy Spliet <rspliet@eclipso.eu>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NRoy Spliet <rspliet@eclipso.eu>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

If a memory allocation fails when using the DMA allocator,
gk20a_instobj_dtor_dma() will be called on the failed instmem object.
At this time, node->handle might not be NULL despite the call to
dma_alloc_attrs() having failed. node->cpuaddr is the right member to
check for such a failure, so use it instead.
Reported-by: NVince Hsu <vinceh@nvidia.com>
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Reviewed-by: NThierry Reding <treding@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

User-space use mappable BOs notably for fences, and expects that a
value update by the GPU will be immediatly visible through the
user-space mapping.

ARM has a property that may prevent this from happening though: memory
can be mapped multiple times only if the different mappings share the
same caching properties. However all the lowmem memory is already
identity-mapped into the kernel with cache enabled, so when user-space
requests an uncached mapping, we actually get an "undefined caching
policy" one and this has strange side-effects described on Freedesktop
bug 86690.

To prevent this from happening, allow user-space to explicitly specify
which objects should be coherent, and create such objects with the
TTM_PL_FLAG_UNCACHED flag. This will make TTM allocate memory using the
DMA API, which will fix the identify mapping and allow us to safely map
the objects to user-space uncached.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Reviewed-by: NLucas Stach <dev@lynxeye.de>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Let GK20A's instmem take advantage of the IOMMU if it is present. Having
an IOMMU means that instmem is no longer allocated using the DMA API,
but instead obtained through page_alloc and made contiguous to the GPU
by IOMMU mappings.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Tegra SoCs have an IOMMU that can be used to present non-contiguous
physical memory as contiguous to the GPU and maximize the use of large
pages in the GPU MMU, leading to performance gains. This patch adds
support for probing such a IOMMU if present and make its properties
available in the nouveau_platform_gpu structure so subsystems can take
advantage of it.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

instmem for GK20A is allocated using dma_alloc_coherent(), which
provides us with a coherent CPU mapping that we never use because
instmem objects are accessed through PRAMIN. Switch to
dma_alloc_attrs() which gives us the option to dismiss that CPU mapping
and free up some CPU virtual space.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Now that Nouveau can operate even when there is no RAM device, remove
the dummy one used by GK20A.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

GK20A does not have dedicated RAM, thus having a RAM device for it does
not make sense. Move the contiguous physical memory allocation to
instmem.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Having a RAM device does not make sense for chips like GK20A which have
no dedicated video memory. The dummy RAM device that we used so far
works as a temporary band-aid, but in the longer term it is desirable
for the driver to be able to work without any kind of VRAM.

This patch adds a few conditionals in places where a RAM device was
assumed to be present and allows some more objects to be allocated from
the TT domain, allowing Nouveau to handle GPUs for which
pfb->ram == NULL.
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Notify interrupt is only used for cyclestats. We can just clear it and
avoid an "unknown stat" error that gets printed to dmesg otherwise.
Signed-off-by: NLauri Peltonen <lpeltonen@nvidia.com>
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Other methods in this file suggest this is the correct way to retrieve
the engine pointer.
Signed-off-by: NLauri Peltonen <lpeltonen@nvidia.com>
Signed-off-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

We use -1 to mean "not read from hw yet"
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

This if statement is correct but it wasn't indented, so it looked like
some code was missing.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Spotted by coccinelle:
drivers/gpu/drm/nouveau/core/subdev/fuse/gm107.c:50:5-8: WARNING: end returns can be simpified
Signed-off-by: NMartin Peres <martin.peres@free.fr>
Reviewed-by: NTobias Klausmann <tobias.johannes.klausmann@mni.thm.de>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Code before looked only at bit 31 to decide if a port is unused.
However dcb 4.1 spec says 0x1F in bits 31-27 and 26-22 means unused.

This fixed hdmi monitor detection on GM206.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Enough to get VGA monitor on DVI-I output have output.
HDMI output not yet working
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

fdo#88868
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

fdo#83992
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Complete bong hit (and not the last...), the hardware will reassert the
interrupt to PMC if it's necessary.

Also potentially harmful in the face of interrupts such as the non-stall
interrupt, which remain active in NV_PFIFO_INTR even when we don't care
about servicing it.

It appears (hopefully, fdo#87244), that under certain loads, the methods
may pass quickly enough to hit the "100 spins and kill PFIFO" thing that
we had going on.  Not ideal ;)
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

We need to store device offsets in 64 bit as the device
address space may be larger than the CPU's.

Fixes GPU init failures on radeons with 4GB or more of
vram on 32 bit kernels.  We put vram at the start of the
GPU's address space so the gart aperture starts at 4 GB
causing all GPU addresses in the gart aperture to get
truncated.

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=89072

[airlied: fix warning on nouveau build]
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: thellstrom@vmware.com
Acked-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The namespace of NVKM is being changed to nvkm_ instead of nouveau_,
which will be used for the DRM part of the driver.  This is being
done in order to make it very clear as to what part of the driver a
given symbol belongs to, and as a minor step towards splitting the
DRM driver out to be able to stand on its own (for virt).

Because there's already a large amount of churn here anyway, this is
as good a time as any to also switch to NVIDIA's device and chipset
naming to ease collaboration with them.

A comparison of objdump disassemblies proves no code changes.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

The namespace of NVKM is being changed to nvkm_ instead of nouveau_,
which will be used for the DRM part of the driver.  This is being
done in order to make it very clear as to what part of the driver a
given symbol belongs to, and as a minor step towards splitting the
DRM driver out to be able to stand on its own (for virt).

Because there's already a large amount of churn here anyway, this is
as good a time as any to also switch to NVIDIA's device and chipset
naming to ease collaboration with them.

A comparison of objdump disassemblies proves no code changes.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

The namespace of NVKM is being changed to nvkm_ instead of nouveau_,
which will be used for the DRM part of the driver.  This is being
done in order to make it very clear as to what part of the driver a
given symbol belongs to, and as a minor step towards splitting the
DRM driver out to be able to stand on its own (for virt).

Because there's already a large amount of churn here anyway, this is
as good a time as any to also switch to NVIDIA's device and chipset
naming to ease collaboration with them.

A comparison of objdump disassemblies proves no code changes.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>