While I had thought I had fixed this issue in:
commit 342406e4fb ("drm/nouveau/i2c: Disable i2c bus access after
->fini()")
It turns out that while I did fix the error messages I was seeing on my
P50 when trying to access i2c busses with the GPU in runtime suspend, I
accidentally had missed one important detail that was mentioned on the
bug report this commit was supposed to fix: that the CPU would only lock
up when trying to access i2c busses _on connected devices_ _while the
GPU is not in runtime suspend_. Whoops. That definitely explains why I
was not able to get my machine to hang with i2c bus interactions until
now, as plugging my P50 into it's dock with an HDMI monitor connected
allowed me to finally reproduce this locally.
Now that I have managed to reproduce this issue properly, it looks like
the problem is much simpler then it looks. It turns out that some
connected devices, such as MST laptop docks, will actually ACK i2c reads
even if no data was actually read:
[ 275.063043] nouveau 0000:01:00.0: i2c: aux 000a: 1: 0000004c 1
[ 275.063447] nouveau 0000:01:00.0: i2c: aux 000a: 00 01101000 10040000
[ 275.063759] nouveau 0000:01:00.0: i2c: aux 000a: rd 00000001
[ 275.064024] nouveau 0000:01:00.0: i2c: aux 000a: rd 00000000
[ 275.064285] nouveau 0000:01:00.0: i2c: aux 000a: rd 00000000
[ 275.064594] nouveau 0000:01:00.0: i2c: aux 000a: rd 00000000
Because we don't handle the situation of i2c ack without any data, we
end up entering an infinite loop in nvkm_i2c_aux_i2c_xfer() since the
value of cnt always remains at 0. This finally properly explains how
this could result in a CPU hang like the ones observed in the
aforementioned commit.
So, fix this by retrying transactions if no data is written or received,
and give up and fail the transaction if we continue to not write or
receive any data after 32 retries.
Signed-off-by: Lyude Paul <lyude@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Appears to be fixed by "flcn/gp102-: improve implementation of
bind_context() on SEC2/GSP".
Tested on GP10[24678] and GV100.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The bulk SPDX addition made all these files into GPL-2.0 licensed files.
However the remainder of the project is MIT-licensed, these files
were simply missing the boiler plate and got caught up in the global update.
Fixes: 96ac6d4351 (treewide: Add SPDX license identifier - Kbuild)
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The bulk SPDX addition made all these files into GPL-2.0 licensed files.
However the remainder of the project is MIT-licensed, these files
(primarily header files) were simply missing the boiler plate and got
caught up in the global update.
Fixes: b24413180f (License cleanup: add SPDX GPL-2.0 license identifier to files with no license)
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
It turns out that while disabling i2c bus access from software when the
GPU is suspended was a step in the right direction with:
commit 342406e4fb ("drm/nouveau/i2c: Disable i2c bus access after
->fini()")
We also ended up accidentally breaking the vbios init scripts on some
older Tesla GPUs, as apparently said scripts can actually use the i2c
bus. Since these scripts are executed before initializing any
subdevices, we end up failing to acquire access to the i2c bus which has
left a number of cards with their fan controllers uninitialized. Luckily
this doesn't break hardware - it just means the fan gets stuck at 100%.
This also means that we've always been using our i2c busses before
initializing them during the init scripts for older GPUs, we just didn't
notice it until we started preventing them from being used until init.
It's pretty impressive this never caused us any issues before!
So, fix this by initializing our i2c pad and busses during subdev
pre-init. We skip initializing aux busses during pre-init, as those are
guaranteed to only ever be used by nouveau for DP aux transactions.
Signed-off-by: Lyude Paul <lyude@redhat.com>
Tested-by: Marc Meledandri <m.meledandri@gmail.com>
Fixes: 342406e4fb ("drm/nouveau/i2c: Disable i2c bus access after ->fini()")
Cc: stable@vger.kernel.org
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Modesetting only, still waiting on ACR/GR firmware from NVIDIA for Turing
graphics/compute bring-up.
Each subsystem was compared with traces, along with various tests to check
that things generally work as they should, and appears compatible enough
with the current TU117 code to enable support.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The fallthrough cases (pre-Fermi) would accidentally allow dual-link pixel
clocks even where they shouldn't be. This leads to a high resolution HDMI
displays, connected via a DVI->HDMI adapter, to fail on the original NV50.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Previously center scaling would get scaling applied to it (when it was
only supposed to center the image), and aspect-corrected scaling did not
always correctly pick whether to reduce width or height for a particular
combination of inputs/outputs.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110660
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Higher layers tend to add a lot of modes not actually in the EDID, such
as the standard DMT modes. Changing this would be extremely intrusive to
everyone, so just force the scaler more often. There are no practical
cases we're aware of where a LVDS/eDP panel has multiple resolutions
exposed, and i915 already does it this way.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110660
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This reverts commit 031e610a6a, reversing
changes made to 52d2d44eee.
The mm changes in there we premature and not fully ack or reviewed by core mm folks,
I dropped the ball by merging them via this tree, so lets take em all back out.
Signed-off-by: Dave Airlie <airlied@redhat.com>
mm/pgtable: drop pgtable_t variable from pte_fn_t functions
drops the token came in via the hmm tree, this caused lots of
conflicts, but applying this cleanup patch should reduce it
to something easier to handle. Just accept the token is unused
at this point.
Signed-off-by: Dave Airlie <airlied@redhat.com>
drm/imx: IPUv3 image converter improvements, enable scanout FIFO watermark
- Fix a saturation bit position in the colorspace converter
configuration memory.
- Fully describe colorspace conversions in the API to the imx-media driver.
- Add support for limited range and Rec.709 YUV encoding.
- Enable colorimetry configuration via the media-controller API.
- Enable the double write reduction feature for memory bandwidth savings
when the image converter writes YUV 4:2:0 output.
- Enable a scanout FIFO watermark feature that can increase priority of
scanout read transfers at the memory controller whenever the FIFO runs
low.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://patchwork.freedesktop.org/patch/msgid/1562326831.4291.8.camel@pengutronix.de
This RELEASE_MEM use has the Release semantic, which means we should write
back but not invalidate. Invalidations only make sense with the Acquire
semantic (ACQUIRE_MEM), or when RELEASE_MEM is used to do the combined
Acquire-Release semantic, which is a barrier, not a fence.
The undesirable side effect of doing invalidations for the Release semantic
is that it invalidates caches while shaders are running, because the Release
can execute in the middle of the next IB.
UMDs should use ACQUIRE_MEM at the beginning of IBs. Doing cache
invalidations for a fence (like in this case) doesn't do anything
for correctness.
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On 32-bit architectures, dividing a 64-bit integer in the kernel
leads to a link error:
ERROR: "__udivdi3" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!
ERROR: "__divdi3" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!
Change the two recently introduced instances to a multiply+shift
operation that is also much cheaper on 32-bit architectures.
We can do that here, since both of them are really 32-bit numbers
that change a few percent.
Fixes: bedbbe6af4 ("drm/amd/display: Move link functions from dc to dc_link")
Fixes: f18bc4e53a ("drm/amd/display: update calculated bounding box logic for NV")
Acked-by: Slava Abramov <slava.abramov@amd.com>
Tested-by: Slava Abramov <slava.abramov@amd.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lyude Paul
Ben Skeggs
Yongxin Liu
Ralph Campbell
Karol Herbst
Ilia Mirkin
Timo Wiren
Dave Airlie
Alex Deucher
Marek Olšák
Arnd Bergmann