Due to the GPU preventing us from touching NV_PLTCG_LTCS_LTSS_CBC_BASE,
we cannot provide CBC/ZBC support without signed PMU firmware to handle
the task for us...
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
GP100 still supports the previous generations' page table layout, which
we will temporarily make use of here.
Proper support for the new MMU layout requires some rework to the common
MMU code, which is in progress.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
GFxxx/GM1xx support the selection of 64/128KiB big pages globally.
GM2xx supports the same, as well as another mode where the page size
can be selected per-instance.
We default to 128KiB pages (With per-instance for GM200, but the current
code selects 128KiB there already) as the MMU code isn't currently able
to handle otherwise.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Fixes the second DVI output on Quadro FX380.
Thanks to NVIDIA for providing the details on the full workaround.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Makes common the code that was previously used by the PMU table parsing,
as it appears other tables need this too.
Not much of an idea what this is all about...
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Some VBIOS have separate tables for each link of a given output path,
which means we have to specify the specific link we're using instead
of all possible links.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Corresponds with GT215. Don't rely on the lock test logic being
unconditionally enabled, and disable test logic when done (presumably
to save power).
v2: Remove warning, nvkm_msec already warns on time-out
Signed-off-by: Roy Spliet <nouveau@spliet.org>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Defer the loading of firmware files to the chip-specific part of secure
boot. This allows implementations to retry loading firmware if the first
attempt failed ; for the GM200 implementation, this happens when trying
to reset a falcon, typically in reaction to GR init.
Firmware loading may fail for a variety of reasons, such as the
filesystem where they reside not being ready at init time. This new
behavior allows GR to be initialized the next time we try to use it if
the firmware has become available.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Make it possible to call gm20x_secboot_prepare_blobs() several times
after either success or failure without re-building already existing
blobs. The function will now try to load firmware files that have
previously failed before returning success.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Handle and propagate secure boot errors. Failure to do so results in
Nouveau incorrectly believing init has succeeded and a completely
black display during boot. If we propagate the error, GR init will fail
and the user will at least have a working display.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Some members were documented in the wrong structure.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
We shouldn't set voltages below the min or above the max voltage the gpu is
able to set, so save the range for future lookups.
Signed-off-by: Karol Herbst <karolherbst@gmail.de>
Reviewed-by: Martin Peres <martin.peres@free.fr>
Tested-by: Pierre Moreau <pierre.morrow@free.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This patch adds support for advanced features supported by the
Noise-Aware PLL of Maxwell. Glitchless switch allows the PL field to be
updated without disabling the PLL first if the SYNC_MODE bit of the CFG
register is set.
More significantly, DFS allows the PLL to monitor the actual input
voltage and to dynamically lower the output frequency accordingly. This
allows the clock to be more tolerant of lower voltages.
These improvements are only supported for Tegra speedos >= 1.
Also add the voltage table that is suitable for GM20B's NAPLL. This
change needs to be done atomically for the right voltages to be used by
the clock driver.
v2. Fix build on non-Tegra platforms
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Strip the _ prefix off the gk20a clock constructor.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Split the MNP programming function into two functions for the cases
where we allow sliding or not, instead of making it take a parameter for
this. This results in less conditionals in the code and makes it easier
to read.
Also make the MNP programming functions take the PLL parameters as
arguments, and move bits of code to more relevant places (previous
programming tended to be just-in-time, which added more conditionnals in
the code).
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Use a dedicated function instead of always calculating n_lo on the fly.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Make functions manipulating PLL settings take them as an argument,
instead of assuming we want to work on the copy in the gk20a_clk
structure. This makes these functions more flexible, which we will need
in GM20B.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Add relevant functions to work with the gk20a_pll structure and use them
where they ought to be instead of directly manipulating registers.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Move variables declarations to their actual scope of use, and simplify
code a bit.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Slide setup needs to be performed only once, during init. Also
use the proper parameters for different clock speeds.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Chips may be characterized for a minimum voltage. Support this extra
parameter and select the appropriate minimum voltage for the detected
GPU speedo.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Strip the _ prefix off the gk20a volt constructor.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Give a name to this constant so we at least get an idea of what it is
for.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Nobody else is using these, so make them private.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The GPU speedo ID is required to select the right clk/volt parameters on
GM20B.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There are cases where subdevs need to perform additonal actions around
the master reset, so we want to expost the operations separately.
This commit also adds a flag to the NV_PMC_ENABLE bitfield definitions
which allow skipping the automatic reset() called from core/subdev.c.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
It isn't used and not waiting for the GPU after scheduling a move is
actually quite dangerous.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When we want to pipeline accelerated moves we need to wait in the fallback path.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Wait for idle before moving the BO in all drivers implementing
an accelerated move function.
This should keep the current behavior when removing the pre move wait.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fixes a regression caused by a stupid thinko from "disp/sor/gf119: both
links use the same training register".
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJXcHi9AAoJEHm+PkMAQRiGSJ0H/2o4t9VWYmhyPC1sdIHoCExJ
P4tBrcZYBmKcsOmIfnJDa5g/+IdhouEUM0v0fHPogS2UUWT9eRuJWYD3sY+HpEQ+
heKTli8X73gsFB25odeIbIt0jAoSiiMYWDrWqLNsuUV1tjEYVA8rH0SM94FiOC/5
7WVWXLTuH+Rm7JHP18BnKxmMMbzrTFmwisLMqFKyfZRRSlS+/ix7iLUNO9AFa39B
YHxNPihLrZ0oONyCOAQoHTIXXrw0cQbxV2utg3vnMcCZdme2xOn+iXMntTSKfZ39
iC9/T0vsO3R6OrRo2aDZAnCPUAniXnMEIhrKG37WMyXpj6cucZ/2QiNXcXviGV4=
=iLte
-----END PGP SIGNATURE-----
Back-merge tag 'v4.7-rc5' into drm-next
Linux 4.7-rc5
The fsl-dcu pull needs -rc3 so go to -rc5 for now.
nouveau leaks a runtime pm ref if at least one CRTC is enabled on
unload. The ref is taken by nouveau_crtc_set_config() and held as long
as a CRTC is in use.
nv04_display_destroy() should solve this by turning off all CRTCs, but
(1) nv50_display_destroy() doesn't do the same and
(2) it's broken since commit d6bf2f3707 ("drm/nouveau: run mode_config
destructor before destroying internal display state") because the
crtc structs are torn down by drm_mode_config_cleanup() before being
turned off. Also, there's no locking.
Move the code to turn off all CRTCs from nv04_display_destroy() to
nouveau_display_destroy() so that it's called for both nv04 and nv50
and before drm_mode_config_cleanup(). Use drm_crtc_force_disable_all()
helper to save on code and have proper locking.
Cc: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/66daa161322444bbde05d83cb0210b90a66988a4.1465392124.git.lukas@wunner.de
The PCI core calls pm_runtime_forbid() on device probe in pci_pm_init(),
making this the default state when nouveau is loaded. nouveau_drm_load()
therefore calls pm_runtime_allow(), but there's no pm_runtime_forbid()
in nouveau_drm_unload() to balance it. Add it so that we leave the
device in the same state that we found it.
This isn't a bug, it's just good housekeeping. When nouveau is first
loaded with runpm=1, then unloaded and loaded again with runpm=0,
pm_runtime_forbid() will be called from nouveau_pmops_runtime_idle() or
nouveau_pmops_runtime_suspend(), so the behaviour is correct. The nvidia
blob doesn't use runtime pm, but if it ever does, this commit avoids
that it has to clean up behind nouveau.
Cc: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Karol Herbst <karolherbst@gmail.com>
Tested-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/92cf96445088217a4d7d7081b90140f2d6f047da.1465392124.git.lukas@wunner.de
nouveau_drm_load() calls pm_runtime_put() if nouveau_runtime_pm != 0,
but nouveau_drm_unload() calls pm_runtime_get_sync() unconditionally.
We therefore leak a runtime pm ref whenever nouveau is loaded with
runpm=0 and then unloaded. The GPU will subsequently never runtime
suspend even if nouveau is loaded again with runpm=1.
Fix by taking the runtime pm ref under the same condition that it was
released on driver load.
Fixes: 5addcf0a5f ("nouveau: add runtime PM support (v0.9)")
Cc: Dave Airlie <airlied@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Reported-by: Karol Herbst <karolherbst@gmail.com>
Tested-by: Karol Herbst <karolherbst@gmail.com>
Tested-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1544b82007037601fbc510b1a50edc56c529e75f.1465392124.git.lukas@wunner.de
Hello,
after this commit:
commit f045f459d9
Author: Ben Skeggs <bskeggs@redhat.com>
Date: Thu Jun 2 12:23:31 2016 +1000
drm/nouveau/fbcon: fix out-of-bounds memory accesses
kernel started to oops when loading nouveau module when using GTX 780 Ti
video adapter. This patch fixes the problem.
Bug report: https://bugzilla.kernel.org/show_bug.cgi?id=120591
Signed-off-by: Dmitrii Tcvetkov <demfloro@demfloro.ru>
Suggested-by: Ilia Mirkin <imirkin@alum.mit.edu>
Fixes: f045f459d9 ("nouveau_fbcon_init()")
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
We already have a fallback in place to fill out the unique from
dev->unique, which is set to something reasonable in drm_dev_alloc.
Which means we only need to have a special set_busid for pci devices,
to be able to care the backwards compat code for drm 1.1 around, which
libdrm still needs.
While developing and testing this patch things blew up in really
interesting ways, and the code is rather confusing in naming things
between the kernel code, ioctl #defines and libdrm. For the next brave
dragon slayer, document all this madness properly in the userspace
interface section of gpu.tmpl.
v2: Make drm_dev_set_unique static and update kerneldoc.
v3: Entire rewrite, plus document what's going on for posterity in the
gpu docbook uapi section.
v4: Drop accidental amdgpu hunk (Emil).
v5: Drop accidental omapdrm vblank counter change (Emil).
v6: Rebase on top of the sphinx conversion.
Cc: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Tested-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> (virt_gpu)
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Replace the legacy drm_send_vblank_event(), drm_arm_vblank_event() and
drm_vblank_{get,put}() with the new helper functions.
v2: add crtc to nouveau_page_flip_state (comment from Mario Kleiner)
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1465308482-15104-1-git-send-email-gustavo@padovan.org
As promised, piles of prep work all around:
- drm_atomic_state rework, prep for nonblocking commit helpers
- fence patches from Gustavo and Christian to prep for atomic fences and
some cool work in ttm/amdgpu from Christian
- drm event prep for both nonblocking commit and atomic fences
- Gustavo seems on a crusade against the non-kms-native version of the
vblank functions.
- prep work from Boris to nuke all the silly ->best_encoder
implementations we have (we really only need that for truly dynamic
cases like dvi-i vs dvi-d or dp mst selecting the right transcoder on
intel)
- prep work from Laurent to rework the format handling functions
- and few small things all over
* tag 'topic/drm-misc-2016-06-07' of git://anongit.freedesktop.org/drm-intel: (47 commits)
drm/dsi: Implement set tear scanline
drm/fb_cma_helper: Implement fb_mmap callback
drm/qxl: Remove useless drm_fb_get_bpp_depth() call
drm/ast: Remove useless drm_fb_get_bpp_depth() call
drm/atomic: Fix remaining places where !funcs->best_encoder is valid
drm/core: Change declaration for gamma_set.
Documentation: add fence-array to kernel DocBook
drm/shmobile: use drm_crtc_vblank_{get,put}()
drm/radeon: use drm_crtc_vblank_{get,put}()
drm/qxl: use drm_crtc_vblank_{get,put}()
drm/atmel: use drm_crtc_vblank_{get,put}()
drm/armada: use drm_crtc_vblank_{get,put}()
drm/amdgpu: use drm_crtc_vblank_{get,put}()
drm/virtio: use drm_crtc_send_vblank_event()
drm/udl: use drm_crtc_send_vblank_event()
drm/qxl: use drm_crtc_send_vblank_event()
drm/atmel: use drm_crtc_send_vblank_event()
drm/armada: use drm_crtc_send_vblank_event()
drm/doc: Switch to sphinx/rst fixed-width quoting
drm/doc: Drop kerneldoc for static functions in drm_irq.c
...
lockless gem bo freeing patches (and the oddball related patch) for all
the drivers who's maintainers are asleep at the helm - includes you ;-)
I based this on top of drm-fixes to include Chris' fix for the cma issue.
* tag 'topic/lockless-gem-bo-freeing-2016-06-01' of git://anongit.freedesktop.org/drm-intel: (21 commits)
drm/arcpgu: Use lockless gem BO free callback
drm/sun4i: Use lockless gem BO free callback
drm/omapdrm: Nuke dummy fb->dirty callback
drm/msm: Nuke dummy fb->dirty callback
drm/rockchip: Use cma gem vm ops
drm/sti: Use lockless gem BO free callback
drm: sti: remove useless call to dev->struct_mutex
drm/virtio: Use lockless gem BO free callback
drm/tilcdc: Use lockless gem BO free callback
drm/shmob: Use lockless gem BO free callback
drm/rockchip: Use lockless gem BO free callback
drm/rcar-du: Use lockless gem BO free callback
drm/qxl: Use lockless gem BO free callback
drm/nouveau: Use lockless gem BO free callback
drm/mga200g: Use lockless gem BO free callback
drm/fls-dcu: Use lockless gem BO free callback
drm/cirrus: Use lockless gem BO free callback
drm/bochs: Use lockless gem BO free callback
drm/atmel: Use lockless gem BO free callback
drm/ast: Use lockless gem BO free callback
...
Change return value to int to propagate errors from gamma_set,
and remove start parameter. Updates always use the full size,
and some drivers even ignore the start parameter altogether.
This is needed for atomic drivers, where an atomic commit can
fail with -EINTR or -ENOMEM and should be restarted. This is already
and issue for drm_atomic_helper_legacy_set_gamma, which this patch
fixes up.
Changes since v1:
- Fix compiler warning. (Emil)
- Fix commit message (Daniel)
Cc: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Eric Anholt <eric@anholt.net>
Cc: VMware Graphics <linux-graphics-maintainer@vmware.com>
Cc: Mathieu Larouche <mathieu.larouche@matrox.com>
Cc: Thierry Reding <treding@nvidia.com>
Acked-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
[danvet: Improve commit message a bit more, mention that this fixes
the helper.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/742944bc-9f41-1acb-df4f-0fd4c8a10168@linux.intel.com
It appears that, for whatever reason, both link A and B use the same
register to control the training pattern. It's a little odd, as the
GPUs before this (Tesla/Fermi1) have per-link registers, as do newer
GPUs (Maxwell).
Fixes the third DP output on NVS 510 (GK107).
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
Now a drm_pending_event can either send a real drm_event or signal a
fence, or both. It allow us to signal via fences when the buffer is
displayed on the screen. Which in turn means that the previous buffer
is not in use anymore and can be freed or sent back to another driver
for processing.
v2: Comments from Daniel Vetter
- call fence_signal in drm_send_event_locked()
- remove unneeded !e->event check
v3: Remove drm_pending_event->destroy to fix a leak when e->file_priv
is not set.
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk> (v2)
[danvet: fix one e->destroy in arcpgu due to rebasing.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1464818821-5736-13-git-send-email-daniel.vetter@ffwll.ch
Ben Skeggs