Use nvgpu_cond for waiting for all channel accesses to finalize
before closing a channel, and for signalling for the same event.
JIRA NVGPU-14
Change-Id: Ifac14ad9afe5c44d4443b4a4a94a4d0ad2ea7053
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1469764
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Lakshmanan M <lm@nvidia.com>
GVS: Gerrit_Virtual_Submit
Introduce a priv ring HAL and define ISR as the only function in it.
Introduce a gp10b version of the ISR that writes error message to
UART for every priv ring error, and leave the old chips with silent
error handling.
Bug 1846641
Change-Id: I73e38396205ac7bb7b8488b7fbca3ff67a3db3bb
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1473696
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
In nvgpu_request_firmware(), if fw->data is NULL, we jump
to label "err" and leak the storage pointed by linux_fw
Fix this by releasing firmware in error path if
fw->data is NULL
Coverity id : 2513066
Bug 200291879
Change-Id: Ieb5e22137cebb4cd02415b805941969b95a38668
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/1479884
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
We have this recursion in headers as pointed
by Coverity
gk20a/gk20a.h" includes itself: gk20a.h -> clk.h -> clk_vin.h ->
boardobjgrp_e32.h -> gk20a.h
Fix this by removing gk20a.h header from boardobjgrp_e32.h
Also remove unused header gk20a.h from boardobjgrp_e255.h too
Coverity id : 2513065
Bug 200291879
Change-Id: I4bee04f064dfb8b02026de9b5a07429b1f48d0a0
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/1479883
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
The __nvgpu_mem_create_from_pages() API takes a number of pages from
which to create a buffer. In the ACR code only 1 page was passed. If
the WPR region (what the ACR code creates an nvgpu_mem from) is larger
than 1 page then the current code will truncate that region down to
only a single page.
Change-Id: I11833619c4f35a076f72071cc5b98ae2d192c5d0
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: http://git-master/r/1477672
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Move the function pointer for VPR page allocation to struct gk20a
and use it from there. At the same time remove secure_page_alloc
pointer and add a direct call to it in probe.
Move platform_tegra.h as tegra/linux/platform_gk20a_tegra.h, as it's
only declaring functions defined in platform_gk20a_tegra.c to other
files in the same directory.
JIRA NVGPU-16
Change-Id: I19ac9ee0b2f6734203ae32a1f51d67fd51aced9f
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1473706
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Begin moving (and renaming) the GMMU code into common/mm/gmmu.c. This
block of code will be responsible for handling the platform/OS
independent GMMU operations.
JIRA NVGPU-12
JIRA NVGPU-30
Change-Id: Ide761bab75e5d84be3dcb977c4842ae4b3a7c1b3
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: http://git-master/r/1464083
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Since ch_wdt_timeout_ms was moved to struct gk20a, vgpu needs to init
it with default platform value.
Jira VFND-3796
Change-Id: I61aca06989ae6b9ad5b3a264ce21d885ca769349
Signed-off-by: Richard Zhao <rizhao@nvidia.com>
Reviewed-on: http://git-master/r/1478670
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
CBC clean timeout is set to 1ms. We're seeing longer times taken
in 3dtex_deep_image, so bump the timeout significantly.
Change-Id: I16febbe663f130236c455d169c6eab47e4f73b52
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1478334
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Added function pointers to check chip specific valid
gfx class and compute class. Also added function pointer
to update ctx header with preemption buffer pointers.
Bug 200292090
Change-Id: I8119ee082e2abb67186a8ac07088f8db7f410ba1
Signed-off-by: seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-on: http://git-master/r/1293502
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
mark functions local to the file as static
fixing errors in volt and flcn modules.
Bug 200299572
Change-Id: Ibacbd83649fee3066a90694a3df90bb909b24aa5
Signed-off-by: Vijayakumar <vsubbu@nvidia.com>
Reviewed-on: http://git-master/r/1475357
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
Add these bits in the gpu characteristics flags:
NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_SUBMIT_NO_JOBTRACKING - fast
submits with no in-kernel job tracking are supported.
NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_SUBMIT_FULL - deterministic
submits also with job tracking and num_inflight_jobs set are supported.
Either of these may get disabled if the particular channel or submit
still requires features that block these.
Make gk20a_channel_sync_needs_sync_framework() take a gk20a pointer
instead of a channel pointer so that it can be called without a channel.
It does not need any per-channel data.
Bug 200291300
Change-Id: I5f82510b6d39b53bcf6f1006dd83bdd9053963a0
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: http://git-master/r/1456845
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Fixed following sparse warning by including relevant header:
$TOP/kernel/nvgpu/drivers/gpu/nvgpu/gp10b/therm_gp10b.c:127:6: warning:
symbol 'gp10b_init_therm_ops' was not declared. Should it be static?
Bug 200299572
Change-Id: I2caff721e98739f0c5978cca54d374cd959ac94a
Signed-off-by: seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-on: http://git-master/r/1474004
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Clock gating list for CE was programmed at GR init, but at that time
CE has not yet been brought out of reset. This causes a priv ring
error and the clock gating setting does not take place. Move
programming of CE clock gating list to CE initialization.
Bug 1846641
Change-Id: Ibc9fe2487408358304f80cd679d3b1ecac7cebe8
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1473301
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Clock gating list contains a register which does not exist on gm20b.
Remove it from the list.
Bug 1846641
Change-Id: Ifb6f5c46482bcaad626bc875e9858d486d3dfa1c
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1473223
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: David Martinez Nieto <dmartineznie@nvidia.com>
KASAN reports below slab-out-of-bounds error when accessing
gr->map_tiles since gr->map_tiles is allocated with wrong(small) size.
fixing it by passing correct size to nvgpu_kzalloc.
BUG: KASAN: slab-out-of-bounds in gr_gk20a_init_map_tiles+0x6f0/0x7b0 ...
....
BUG: KASAN: slab-out-of-bounds in gr_gk20a_zcull_init_hw+0x184/0x848 ...
...
BUG: KASAN: slab-out-of-bounds in gr_gk20a_setup_rop_mapping+0x108/0x1740 ...
...
BUG: KASAN: slab-out-of-bounds in gr_gk20a_setup_rop_mapping+0x108/0x1740 ...
Bug 1918671
Change-Id: I667ac80b20a3d8539ed3eaae6e0f98e91f917819
Signed-off-by: David Pu <dpu@nvidia.com>
Reviewed-on: http://git-master/r/1472491
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
- struct nvgpu_falcon to hold properties of falcon controller
- falcon controller interface layer which establish access
to required falcon controller HAL based on struct nvgpu_falcon member
flcn_id & flcn_base parameter.
- each falcon nvgpu_falcon struct initialized during init
with id, base-address along with other properties at HAL.
- Added defines related to flacon controller.
Change-Id: Ia7777c01ecc542150ddd72f8603b7b4475522b58
Signed-off-by: Mahantesh Kumbar <mkumbar@nvidia.com>
Reviewed-on: http://git-master/r/1467523
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Remove an unused variable that missed compilation checks because
it gets assigned to by do_div(). This same issue was propagated
to multiple places.
Change-Id: Ica2f675abbb3c08107ea4e6bc19044c0537a7484
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: http://git-master/r/1472365
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
In the Linux kernel many bitmap operations are included in bitmap.h
and not in bitops.h. In nvgpu it seems we assume that these two sets
of functions are contained in the same header. This patch ensures
then when including bitops.h we get both bitmap declarations and bit
ops declarations.
Change-Id: I9fea75d6c920e1a2992922f927f2d91bbdbdedd3
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: http://git-master/r/1472364
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
use memset to fill structures with zero instead of
assigning zero.
mark functions local to the file as static
fixing errors in clk, perf and therm modules.
Bug 200299572
Change-Id: I0470298803c35b6faed2edc2a0c1dbf0e47e842e
Signed-off-by: Vijayakumar <vsubbu@nvidia.com>
Reviewed-on: http://git-master/r/1472940
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Fix issues related with wrong storage type for
64 bit variables.
(1) Fixed width of HZ_TO_MHZ constant
(2) changed fence_wait timeout to store unsigned
long
bug 200299572
Change-Id: Ie8f2386b738f3aafce75fc2440947e36befac273
Signed-off-by: David Nieto <dmartineznie@nvidia.com>
Reviewed-on: http://git-master/r/1471611
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
We access ROM control register with xve_writel, but we also
add the base register address. This leads to adding the base
address twice, and the access goes to wrong register.
Bug 1846641
Change-Id: I46ef277aac661a08049935b08505120cad1a5e76
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1471505
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Write an error to UART on priv ring error. This uncovers any
accesses to missing registers or illegal accesses to registers.
Bug 1846641
Change-Id: Ic1e5ecadcd95777f2b3f7bd77accf98ddce97282
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1294683
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Use an nvgpu_mem created from a page in the ACR code instead of
a scatter gather table created by the Linux SGT APIs. The ACR code
needs to have the GPU map a physical page queried from an external
API (other than the regular DMA API).
Note that this code used to explicitly mark the SGT it makes as
bypassing the SMMU. However, that is now taken care of implicitly
by the __nvgpu_mem_create_from_pages() function.
JIRA NVGPU-12
JIRA NVGPU-30
Change-Id: Ie40152a7611e985e1b97ac2ddc7e27664b71917c
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: http://git-master/r/1464082
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
The gfp_t argument for dma_alloc_attrs() cannot take DMA_MEMORY_NOMAP -
it's interpreted as __GFP_RECLAIMABLE which has the same integer value.
Use GFP_KERNEL for the flag argument and set DMA_ATTR_NO_KERNEL_MAPPING
for dma attrs which the code was trying to do with the flag that is
meant for the coherent allocation API.
Bug 200299572
Change-Id: Ie4d988fbeeb954f6f7ccd4f9fb438968d76f0c6c
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: http://git-master/r/1468315
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
