Allow userspace to control directly the PTE kind for the mappings by
supplying NVGPU_AS_MAP_BUFFER_FLAGS_DIRECT_KIND_CTRL for MAP_BUFFER_EX.
In particular, in this mode, the userspace will tell the kernel
whether the kind is compressible, and if so, what is the
incompressible fallback kind. By supplying only the compressible kind,
the userspace can require that the map kind will not be demoted to the
incompressible fallback kind in case of comptag allocation failure.
Add also a GPU characteristics flag
NVGPU_GPU_FLAGS_SUPPORT_MAP_DIRECT_KIND_CTRL to signal whether direct
kind control is supported.
Fix indentation of nvgpu_as_map_buffer_ex_args header comment.
Bug 1705731
Change-Id: I317ab474ae53b78eb8fdd31bd6bca0541fcba9a4
Signed-off-by: Sami Kiminki <skiminki@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1543462
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
We right now remove a channel from TSG list and disable all the channels in
TSG while removing a channel from TSG
With this sequence if any one channel in TSG is closed, rest of the channels
are set as timed out and cannot be used anymore
We need to fix this sequence as below to allow removing a channel from active
TSG so that rest of the channels can still be used
- disable all channels of TSG
- preempt TSG
- check if CTX_RELOAD is set if support is available
if CTX_RELOAD is set on channel, it should be moved to some other channel
- check if FAULTED is set if support is available
- if NEXT is set on channel then it means channel is still active
print out an error in this case for the time being until properly handled
- remove the channel from runlist
- remove channel from TSG list
- re-enable rest of the channels in TSG
- clean up the channel (same as regular channels)
Add below fifo operations to support checking channel status
g->ops.fifo.tsg_verify_status_ctx_reload
g->ops.fifo.tsg_verify_status_faulted
Define ops.fifo.tsg_verify_status_ctx_reload operation for gm20b/gp10b/gp106
as gm20b_fifo_tsg_verify_status_ctx_reload()
This API will check if channel to be released has CTX_RELOAD set, if yes
CTX_RELOAD needs to be moved to some other channel in TSG
Remove static from channel_gk20a_update_runlist() and export it
Bug 200327095
Change-Id: I0dd4be7c7e0b9b759389ec12c5a148a4b919d3e2
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1560637
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Due to a h/w bug in Maxwell and Pascal we first need to enable all channels with
NEXT and CTX_RELOAD set in a TSG, and then rest of the channels should be
enabled
Add this sequence to gk20a_tsg_enable()
Add new APIs to enable/disable scheduling of TSG runlist
gk20a_fifo_enable_tsg_sched()
gk20a_fifo_disble_tsg_sched()
Add new APIs to check if channel has NEXT or CTX_RELOAD set
gk20a_fifo_channel_status_is_next()
gk20a_fifo_channel_status_is_ctx_reload()
Bug 1739362
Change-Id: I4891cbd7f22ebc1e0bf32c52801002cdc259dbe1
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1560636
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Add platform specific operations to enable/disable a TSG and use them instead
of directly calling enable/disable APIs
For gm20b/gp106/gp10b we continue to use gk20a_enable_tsg() and
gk20a_disable_tsg() as platform specific operations
Bug 1739362
Change-Id: I2dd0f38c8303757e8c7a47d8da0e30a790e514f0
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1560635
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
The following changes are part of the porting of the bind/unbind
functionality.
These changes reuse the shutdown codepaths in iGPU and dGPU and fix a locking
issue with in gk20a_busy() where the usage count can lead to a deadlock during
the driver shutdown. It fixes a racing condition with the gr/mm code by
invalidating the sw ready flag while holding the busy lock
JIRA: EVLR-1739
Change-Id: I62ce47378436b21f447f4cd93388759ed3f9bad1
Signed-off-by: David Nieto <dmartineznie@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1554959
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Explicitly page align DMA memory allocations. Non-page aligned
DMA allocs can lead to nvgpu_mem structs that have a size that's
not page aligned. Those allocs, in some cases, can cause vGPU
maps to return an error.
More generally DMA allocs in Linux are never going to not be
page aligned both in size and address. So might as well page align
the alloc size to make code else where in the driver more simple.
To imlpement this an aligned_size field has been added to struct
nvgpu_mem. This field has the real page aligned size of the
allocation. The original size is still saved in size.
Change-Id: Ie08cfc4f39d5f97db84a54b8e314ad1fa53b72be
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1547902
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Fix a race condition in gk20a_get_channel_from_file() that returns a
channel pointer from an fd: take a reference to the channel before
putting the file ref back. Now the caller is responsible of releasing
the channel reference eventually.
Also document why dbg_session_channel_data has to hold a ref to the
channel file instead of just the channel: that might deadlock if the fds
were closed in "wrong" order.
Change-Id: I8e91b809f5f7b1cb0c1487bd955ad6d643727a53
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1549290
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
In the Linux VM code the kind attribute is assigned to a u8 from an
int. The code is safe due to the checks in place, however, coverity
does not like this so explicitly cast to u8 before assigning kind
to kind_v.
Coverity ID: 2567919
Bug 200291879
Change-Id: I3860faa99da8ac1a24bd0c3d77c7fbc72207614a
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1548712
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Use the error code generated when the XVE clock change times out.
Before this the error code was being set and the while loop that
was polling the clock change status was broken out of. But the
error code was only used later on in the while loop.
Now just set the error code and only break from the loop later on
when that error code is actually checked.
Coverity ID: 2624553
Bug 200291879
Change-Id: If8f6046efcc9c4ad0eb13715735a57870be26110
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1548711
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Remove the kernel feature to map compbit backing store areas to GPU VA.
The feature is not used, and the relevant user space code is getting
removed, too.
Change-Id: I94f8bb9145da872694fdc5d3eb3c1365b2f47945
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1547898
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: Sami Kiminki <skiminki@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
The header <nvgpu/linux/atomic.h> in file include/nvgpu/atomic need
to be included inside __kernel__
Change-Id: Iac65ec570cdf6ba3bdf0149e4648f9f64cd23b00
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1547537
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
NVGPU_SUBMIT_GPFIFO_FLAGS_RESCHEDULE_RUNLIST causes host to expire
current timeslice and reschedule from front of runlist.
This can be used with NVGPU_RUNLIST_INTERLEAVE_LEVEL_HIGH to make a
channel start sooner after submit rather than waiting for natural
timeslice expiration or block/finish of currently running channel.
Bug 1968813
Change-Id: I632e87c5f583a09ec8bf521dc73f595150abebb0
Signed-off-by: David Li <davli@nvidia.com>
Reviewed-on: http://git-master/r/#/c/1537198
Reviewed-on: https://git-master.nvidia.com/r/1537198
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Currently nvgpu common header files include the
Linux specific headers protected by __KERNEL__
check.
The patch includes the rmos specific headers when
the common headers (and common nvgpu files, by
extension) are compiled under QNX, so that QNX
specific data structures / type definitions can
be retrieved.
Change-Id: Icb03fdc90f6ef4cff8666a6fd0f0f06b5ef62c97
Signed-off-by: Sourab Gupta <sourabg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1547658
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
The vGPU mapping code in gp10b gives very little debugging info when
there's a failure. This change adds much more detail to the error printing
so that thee failures can more easily be debugged without needing to
run the virtual guest locally.
Change-Id: Ibb4412fd4ab322b366f0e08eaa399b7b90ea22c7
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1544506
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Remove debugging features that did not really get used and make
the debugging code use the nvgpu_log() functionality. This ties
the allocator debugging into the larger nvgpu debug framework.
Also modify many of the places CONFIG_DEBUG_FS was used to
conditionally compile allocator debug code to use __KERNEL__
instead. This is because that debug code can still be called even
when debugfs is not present in Linux.
Change-Id: I112ebe1cae22d6f8db96d023993498093e18d74a
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1544439
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
When compiling the kernel with an O= option (which stores built files
outside the source tree) the kernel adds various extra source paths to the
system include path. However, this doesn't happen when building in-tree.
Adjust the main nvgpu Makefile to ensure all required paths are part of
the system include path so that all headers can be found.
Bug 1978388
Change-Id: I51ffc78b3863b89ebb5f051c963a8016258534a3
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1544320
(cherry picked from commit 572c509152 in rel-28)
Reviewed-on: https://git-master.nvidia.com/r/1545675
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Bo Yan <byan@nvidia.com>
GM20B GPCPLL dvfs register cannot be accessed through sys registers
bank (as other PLL registers), instead bcast bank must be used. This
limitation was already taken into account for production access, but
dbugfs access has incorrectly used sys bank. Fixed bank access in
this commit.
Change-Id: Ic7ca640c586addea3aaae4f10a98af8497d6f3cb
Signed-off-by: Alex Frid <afrid@nvidia.com>
Reviewed-on: http://git-master/r/1504241
Reviewed-on: https://git-master.nvidia.com/r/1543854
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: Krishna Reddy <vdumpa@nvidia.com>
Prints addresses of device-specific HAL functions to debugfs file
hal/gops.
The list of functions is produced by dumping the contents of the
gpu_ops substruct of the gk20a struct. This interface makes the
assumption that there are only function pointers in gpu_ops.
Companion Python script nvgpu_debug_hal.py analyzes gk20a.h to
determine operation counts and prettyify debugfs interface's
output.
Jira NVGPU-107
Change-Id: I0910e86638d144979e8630bbc5b330bccfd3ad94
Signed-off-by: Sunny He <suhe@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1542990
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reorganize HAL initialization to remove inheritance and construct
the gpu_ops struct at compile time. This patch only covers the
gr sub-module of the gpu_ops struct.
Perform HAL function assignments in hal_gxxxx.c through the
population of a chip-specific copy of gpu_ops.
Jira NVGPU-74
Change-Id: Ie37638f442fd68aca8a7ade5f297118447bdc91e
Signed-off-by: Sunny He <suhe@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1542989
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com>
Limit the maximum job size of CBC ctrl clear to 16 klines. This avoids
timeouts and excessive lock hold duration when clearing comptags for
huge surface. 16 klines corresponds to a 1-GB surface for 64-kB
compression page size.
If the requested CBC ctrl job is larger than 16 klines, split it to
at most 16-kline chunks.
Bug 1860962
Bug 200334740
Change-Id: Ibc69adc8bf59527b1acec5b2097b5aefa2169960
Signed-off-by: Sami Kiminki <skiminki@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1540432
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Although igpu does not have an FBPA unit, the hardware reports one, and
the ucode leaves space for one in the HWPM context save buffer. So
let NUM_FBPAS reflect this, so that registers that follow this section
in the context buffer are offset properly
JIRA EVLR-1716
Change-Id: I40105cf2ee7f92a44d1c7cc2d23a17276bbb9e84
Signed-off-by: Peter Daifuku <pdaifuku@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1535273
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
gm206_bios_init() uses nvgpu_os_linux only for accessing debugfs.
That's the only reason os_linux.h is included. Make the #include
conditional and enabled only when debugfs is enabled.
Change-Id: I2939d51a25865ab634c92676daa9a7c10551fee5
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1541511
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Sourab Gupta <sourabg@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reorganize HAL initialization to remove inheritance and construct
the gpu_ops struct at compile time. This patch only covers the
pmu sub-module of the gpu_ops struct.
Perform HAL function assignments in hal_gxxxx.c through the
population of a chip-specific copy of gpu_ops.
Jira NVGPU-74
Change-Id: I8839ac99e87153637005e23b3013237f57275c54
Signed-off-by: Sunny He <suhe@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1530982
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Move non-function pointer members out of the pmu and pmu_ver
substructs of gpu_ops. Ideally gpu_ops will have only function
ponters, better matching its intended purpose and improving
readability.
- g.ops.pmu_ver.cmd_id_zbc_table_update has been changed to
g.pmu_ver_cmd_id_zbc_table_update
- g.ops.pmu.lspmuwprinitdone has been changed to
g.pmu_lsf_pmu_wpr_init_done
- g.ops.pmu.lsfloadedfalconid has been changed to
g.pmu_lsf_loaded_falcon_id
Boolean flags have been implemented using the enabled.h API
- g.ops.pmu_ver.is_pmu_zbc_save_supported moved to
common flag NVGPU_PMU_ZBC_SAVE
- g.ops.pmu.fecsbootstrapdone moved to
common flag NVGPU_PMU_FECS_BOOTSTRAP_DONE
Jira NVGPU-74
Change-Id: I08fb20f8f382277f2c579f06d561914c000ea6e0
Signed-off-by: Sunny He <suhe@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1530981
Reviewed-by: svccoveritychecker <svccoveritychecker@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
