This is a lot of prep-work for being able to send event notifications
back to userspace. Events now contain data, rather than a "something
just happened" signal.
Handler data is now embedded into a containing structure, rather than
being kmalloc()'d, and can optionally have the notify routine handled
in a workqueue.
Various races between suspend/unload with display HPD/DP IRQ handlers
automagically solved as a result.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Add support for reclocking on GK20A, using a statically-defined pstates
table. The algorithms for calculating the coefficients and setting the
clocks are directly taken from the ChromeOS kernel.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Make nouveau_clock_create() take new two optional arguments: an array
of pstates and its size. When these are specified,
nouveau_clock_create() will use the provided pstates instead of
probing them using the BIOS.
This is useful for platforms which do not provide a BIOS, like Tegra.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Allow the clock subsystem to operate even if voltage and thermal devices
are not set for the device (for people with watercooling! ;))
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The DMA API is the recommended way to map pages no matter what the
underlying bus is. Use the DMA functions for page mapping and remove
currently existing wrappers.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
GK20A's BAR is functionally identical to NVC0's, but do not support
being ioremapped write-combined. Create a BAR instance for GK20A that
reflect that state.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
As a side note.. It's a bit hard to figure out how to name this commit..
GK20A is NVEA, which is before NV108 (GK208).. Confusing.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
As documented at:
ftp://download.nvidia.com/open-gpu-doc/gk104-disable-graphics-power-gating/1/gk104-disable-graphics-power-gating.txt
NVIDIA were not able document the steps necessary to detect whether this
is required or not at this time. However, they did confirm that this
procedure is safe to perform unconditionally on GK104/6. GK107 does not
have the power gating feature, and it was recommended that we do not
perform these steps there as the effects were not verified.
The disable path is from observing the binary driver, and not
documented in the link above.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
When gcc 4.8 inlines this function, it eats up 16 bytes on the stack
every time. Eventually we hit warnings because our stack grew too
much:
ramnve0.c:1383:1: error: the frame size of 1496 bytes is larger than
1024 bytes
We fix this by preventing inlining for this function.
Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Fixes (at least) PTHERM accesses timing out at higher clock speeds.
Values and registers taken from what the binary driver does.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There's also provisions to allow a pad to be locked with a specific
routing, for an indefinite period of time. This will be used in
future patches.
The G94+ pad driver will now also power-down pads when not required.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This was a half-finished hack before, just enough to handle the shared
aux/i2c pad thing on G94 and up.
We got lucky with locking etc up until now, as this was (generally) all
protected by the DRM mode_config lock. It's about to become a lot more
likely to hit the races.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Re-uses the implementation's accessor functions rather than requiring
and init/fini implementation for each chipset.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Whenever a single nouveau_mc_intr() main gpu irq-handler invocation was
responsible for calling both, the vblank-irq handler (display engine irq)
and kms-pageflip completion handler (from fifo irq), the order of
invocation was wrong. nouveau_finish_flip() was called before
drm_handle_vblank() for the vblank of pageflip completion, so the
emitted pageflip event contained stale vblank count and timestamp
from previous vblank. This caused failure in userspace to timestamp
properly.
Reorder order of invocation of engine irq handlers: Put
NVDEV_ENGINE_DISP always on top, and thereby before NVDEV_ENGINE_FIFO,
so that drm_handle_vblank() gets called to update vblank timestamps
and count before potential pageflip events make use of that
information.
This works on nv-50 and later, where kms-pageflip completion triggers
an irq either after a separate vblank irq, or both pageflip and vblank
trigger one common irq invocation, but never before vblank irqs.
v2 (Ben):
- removed mods for nv04-nv40, it doesn't help there anyway
- this is considered a hack, and a better solution should be found
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: <stable@vger.kernel.org> # 3.13+
