Age | Commit message (Collapse) | Author |
|
When the dGPU is entering GCOFF, the link should first be placed into
L2/L3 as appropriate for the design, then when exiting, the link should
be placed back into L0. This patch fixes that oversight.
BUG=b:239719056
TEST=build
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: Ia3bdfe5641216675e06ebe82ffe58bf8c049b26b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66200
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Ivy Jian <ivy.jian@quanta.corp-partner.google.com>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
|
|
After staring at lots of scope shots, the EE has determined that a few
modifications to the GCOFF sequence can be made:
- Remove delay between PERST# assertion and GPU_ALLRAILS_PG deassertion
- Remove delay after ramping down FBVDD
This patch implements these minor changes.
BUG=b:240199017
TEST=verified by EE
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I7d492b3e65a231bc5f64fe9c3add60b5e72eb072
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66199
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Ivy Jian <ivy.jian@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
|
|
The full dGPU power-on sequence, when executed from ACPI, is taking
roughly 15ms or so, which puts it close to the maximum of 20ms required
from the Nvidia spec. Changing the polling period to 100 us instead of 1
ms drastically reduces the time required for this sequence, now taking
typically 7 ms or so. This gives a lot more margin during the power on
sequence.
BUG=b:238466724
TEST=Sequence verified by EE on a scope
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I3ba676c5fac983a0c1ad1d60c3863d06ed33fa27
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66020
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
The agah EC code includes a driver to keep track of the current D Notify
level that the GPU should be at. When it changes, it will send a host
event to the ACPI FW, which will then pass that Notify on to the kernel
driver. This patch adds support for that feature, which is described in
the Nvidia Software Design Guide.
BUG=b:229405562
TEST=add Printf() calls to the ACPI, and work through the various
scenarios on the EC that will cause D Notify levels to change; this
will cause the Printfs() to show up in the kernel log.
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I5cd8bd7d177ea10a165613ed0726a6d6fd86c226
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65486
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Robert Zieba <robertzieba@google.com>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
GPIO_GPU_NVVDD_EN is incorrectly (duplicately) assigned to GPP_A19 in
power.asl, but a double check of the schematic shows that the actual pad
is GPP_A17, so this patch fixes that.
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I4432b50c737508b7e0d595423d614a723d2499c4
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65580
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
I misread my notes when writing the code for the GC6I/GC6O Methods, and
accidentally included NV_33 in the GC6 sequence, which is incorrect
(confirmed in the Hardware Design Guide). This patch removes the code
that brings NV_33 up and down during the GC6 sequences.
BUG=b:236676400
TEST=build
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: Iaa6c5ef3d7b1edbe13257f99013ab0e4382bdbf5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65565
Reviewed-by: Robert Zieba <robertzieba@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
The _ON and _OFF methods for the root port's power resource were
calling the _ON and _OFF in the PEGP namespace, which was the
incorrect method, it should have been NPON/NPOF, so this patch
updates that.
BUG=b:236676400
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: Ia3653996329473f133e3f0d53306882dc3213b6b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65487
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Ivy Jian <ivy.jian@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
The preferred way of polling in ACPI I've seen is usually to just
divide the sleep into N chunks, and ignore the time taken in between.
This works in practice (validated with Timer calls before and after).
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I4a2cd82cea05c539eff30b9b9d6ef18550d17686
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65484
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Robert Zieba <robertzieba@google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
The NBCI "get callbacks" _DSM subfunction should utilize the same "get
callbacks" subfunction from the GPS _DSM subfunction; this patch adds
that Method call into the ACPI code.
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: Idf2f148b5a95acccb02f47cba1ef33a9fc16bcd9
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65483
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Robert Zieba <robertzieba@google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
To avoid extraneous calls from the kernel to _ON or _OFF, keep track
of the power state of the GPU in an integer and exit _ON and _OFF
routines early when attempting to enter the current state.
BUG=b:236676400
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: Ie874fcdc7022c4fde6f557d1ee06e8392ae3d850
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65482
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Robert Zieba <robertzieba@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
The EEs have observed the ramp down delay on this signal in more detail
and 40 ms can still meet the sequencing requirements.
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I49ef801f7a3fd7945ded63da1399eaf57fd6aef0
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65581
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
|
|
We have clarified the powerdown sequence with Nvidia, and the EEs have
come up with this modified sequence which still meets the requirements
from the hardware design guide.
BUG=b:233959099
TEST=Verified by ODM and EE
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I37715165ab488f994c825fb9ff532ebf8d7f4cb0
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65182
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Robert Zieba <robertzieba@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
|
|
The _DSM subfunction for the Nvidia GN20 supports 1 additional
subfunction, known as GPS, which is required to support GPU Boost. This
implementation is minimal, essentially letting the GPU manage its own
temperature.
BUG=b:214581372
TEST=abuild
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I21331bd811a13212f3825bda44be44d1b5ae7c74
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64995
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
Now that the power sequencing for the GPU is in a better shape, ensure
that the ACPI code that performs power sequencing matches the C code
that does the same.
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I797ee99f22a7a6aaacfe54862595674d4ada06ee
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64994
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
The Nvidia GPU kernel driver supports another _DSM subfunction which
is known as NVPCF (Nvidia Platform and Control Framework). The
subfunction informs the kernel driver about Dynamic Boost parameters,
which is done at init time, but can also be changed dynamically.
BUG=b:214581372
TEST=build
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I7887bfc2e8e1cae606e12502a9eda3a7954c8d7a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64535
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
The Nvidia GPU supports another function named NBCI (NoteBook Common
Interface), which has some subfunctions which are required for the
Nvidia kernel driver to consume. The specification for this function
comes from the Nvidia GN20 Software Design Guide.
BUG=b:214581763
TEST=build
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I19eb9417923d297a084d6f5329682e91cd506a9e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64008
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
|
|
Some brya variants will use a GN20 series Nvidia GPU, which requires
quite a bit of ACPI support code to be written for it. This patch
lands a decent bit of the initial code for it on the brya platform,
including:
1) PEG RTD3 methods
2) DGPU power operations (RTD3 and GCOFF, NVJT _DSM and other Methods)
3) NVOP _DSM method
There will be more support to come later, this is all written to
specifications from the Nvidia Software Design Guide for GN20.
BUG=b:214581763
TEST=build patch train
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: Ifce1610210e9636e87dda4b55c8287334adfcc42
Reviewed-on: https://review.coreboot.org/c/coreboot/+/62931
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
|