- During nvgpu module unload, the poweroff sequence
will not wait for the ACK from the PMU for the RPCs sent.
- Due to this, rpc_payload struct info will be present
in pmu seq struct.
- This can lead to memory corruption during unload path.
- To avoid this, return a different value for driver shutting
down scenario from fw ack function and based on this return value
free the RPC payload and release the respective pmu sequence struct.
Bug 3789998
Change-Id: I25104828d836ae37e127b40c88209da81754ffb8
Signed-off-by: Divya <dsinghatwari@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2839968
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Mahantesh Kumbar <mkumbar@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
When a user-domain gets removed, tsg belongs to this also gets removed
and runlist update happens accordingly. If same tsg was submitted to gpu
then updated runlist also needs to re-submit. This works fine with the
existing legacy cases but if GPU is running the shadow domain submitted
by manual mode scheduler and domain belongs to this gets removed then
updated runlist is not being submitted to GPU. This runlist buffer
inconsistency causes mmu fault later.
This change adds a "remove" field in the runlist domain which gets set
to true when runlist update happens for the channel removal. Later
worker thread submit the updated runlist if this flag set to true.
Bug 3884011
Change-Id: I3ce08a5a281e20661915746e70ac0dcd711f3f38
Signed-off-by: prsethi <prsethi@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2838808
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
This patch introduces the flag NVGPU_DISABLE_STATIC_POWERGATE. Further,
static pg related initialization APIs have been updated to check whether
this newly added flag is enabled. If this flag is enabled, then static
pg init gets skipped.
JIRA NVGPU-9350
Change-Id: I450b6dd2c541d31fc40fb5540e557b5db730fee2
Signed-off-by: Rajesh Devaraj <rdevaraj@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2834021
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
EPL lib constructor is replaced by NvEplInit() by safety services.
NvGPU, being an EPL user, needs to call NvEplInit before using EPL
API to report errors.
NvEplInit() usage is limited to user space in Linux, so NvGPU is not
expected to call NvEplInit on Linux.
Bug 3863536
Change-Id: I7fc9b33d3443bd39a6367a0c691bddb80b9edb68
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2817245
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
The memory leak issue was due to nvgpu_profiler_unbind_context() calling
nvgpu_profiler_pm_resource_release() for all resources which clears the
flag required by nvgpu_profiler_free_pma_stream() to release the memory
for perf_buf instance block.
Fixing this issue by splitting nvgpu_profiler_unbind_context() to release
all the pm resources at a later time separately.
Bug 3510455
Change-Id: Ibab8d071693e600c46f7e7f16575e36e6f62af3c
Signed-off-by: atanand <atanand@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2825013
Reviewed-by: svcacv <svcacv@nvidia.com>
Reviewed-by: Dinesh T <dt@nvidia.com>
Reviewed-by: Vaibhav Kachore <vkachore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
This patch is required for
- enabling NVGPU driver on OOT by enabling various
configs required.
- replacing new APIs for some deprecated APIs by
guarding with linux version for soc.c.
- CONFIG_TEGRA_HV_MANAGER is enabled by default in OOT kernel,
removing CONFIG_TEGRA_HV_MANAGER check from various places.
Bug 3812973
Change-Id: I07f0b738ca95d4a3996e7f3ee5e895463db0626b
Signed-off-by: Dinesh T <dt@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2822434
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Changes
1. Remove dGPU flag dependency on calling gsp sw init on tot.
2. Created Enable flag for gsp scheduler to enable them on ga10b
platforms.
3. Engine config flag is only enabled for dGPU enabled platforms, as gsp
is using engine functions it need to be enabled for all gsp sched
enabled builds.
4. Changes in gsp_sequence_init/de_init where on qnx we are seeing
issues.
NVGPU-9297
Change-Id: Ia4bce85ae8fd2794da1553e9ea418c76845a10ac
Signed-off-by: rmylavarapu <rmylavarapu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2822537
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
- As part of mmap call handler mmap_sem is being acquired which is
causing an BUG: scheduling while atomic:. This is happening because
mmap callback gets called with mmap_sem held. The mmap file handler
is called via call_mmap() via mmap_region() via do_mmap(), and the
caller of do_mmap() must hold down_write(¤t->mm->mmap_sem).
To fix this issue removing the mmap_sem locking from nvs mmap handler.
- If remap_vmalloc_range() is used to map the pages to userspace then
allocated pages should be mapped into virtually contiguous space by
passing VM_USERMAP flag to vmap(). Passing VM_USERMAP to vmap() call
if API nvgpu_dma_alloc_flags_sys() is called with flag
NVGPU_DMA_VM_USERMAP_ADDRESS.
Bug 3884011
Change-Id: I64264f6f6e0dd75b1f828dc58355d740e6ef5ccb
Signed-off-by: prsethi <prsethi@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2820781
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add API to query all memhandles used for pde and pte.
- Some direct pde/pte allocation should also add entry to the pd-cache
full list.
- Add OS API for querying MemServ handle from nvgpu_mem.
- Traverse through all pd-cache partial and full lists to get memhandles
for all pde/pte buffers.
Jira NVGPU-8284
Change-Id: I8e7adf1be1409264d24e17501eb7c32a81950728
Signed-off-by: Shashank Singh <shashsingh@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2735657
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Vaibhav Kachore <vkachore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
Now that we are planning to enable CTRL_FIFO support with NVS,
there is no need for a separate enabled flag for the same.
CTRL_FIFO support is instead determined by the presence of
NVGPU_SUPPORT_NVS enable flag alone.
For non-auto platforms, Control-Fifo can be disabled by restricting
access to /dev/nvsched_ctrl_fifo.
Jira NVGPU-8619
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Change-Id: I9dbec60e5668f38e1460c43800584e88b16a2550
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2814435
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Use CONFIG_KMD_SCHEDULING_WORKER_THREAD instead of
CONFIG_NVS_KMD_BACKEND to remove confusion about the CPU based
KMD scheduling worker thread.
The KMD based scheduling worker thread caters to both Manual Mode
CPU based scheduler as well as Automatic Round Robin CPU based
scheduler.
For the traditional submit path, add correct handling of the
CONFIG_NVS_PRESENT. CPU based worker thread should be part of
CONFIG_NVS_PRESENT. Eventually, when DCONFIG_KMD_SCHEDULING_WORKER_THREAD
is removed, the application must switch to GSP.
Jira NVGPU-8619
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Change-Id: I0886ef3b2e0124b6fe22c2bf0bf7d1fa98039d00
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2810217
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Any corrected or uncorrected error reported by gpu hw
will be seen by nvgpu-mon. nvgpu-mon will raise a devctl call
to notify nvgpu-rm if its a fatal error interrupt.
nvgpu_cic_mon_handle_fatal_intr is the corresponding handler which will
walk through the entire tree structure of interrupts for all the subunits
and enter quiesce state.
Change-Id: I3c00c61a7f2c52ae1920f84ee7dfb65cba6b683d
Signed-off-by: Kishan <kpalankar@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2801693
Reviewed-by: svcacv <svcacv@nvidia.com>
Reviewed-by: Vaibhav Kachore <vkachore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
- When we enable AELPG, we send maximum idle threshold value
as one of the parameters.
- This was set to 70 msec whereas the ELPG IDLE threshold
was set to 15 msec.
- Thus, when AELPG is enabled, the threshold is getting modified
to a much higher value. So sometimes ELPG is not getting engaged
and this is leading to timeout in ELPG MODS 101 test on GVS.
Bug 3737783
Change-Id: Iac94f053e19cea5898b962c3d02369d556b8518f
Signed-off-by: Divya <dsinghatwari@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2786749
(cherry picked from commit 5010eaffda2ecaf6c9cc5f0fed68498cb2947071)
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2809154
Reviewed-by: svcacv <svcacv@nvidia.com>
Reviewed-by: Mahantesh Kumbar <mkumbar@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
In order to share the TSG across different devices securely, device
instance IDs are to be exchanged for endpoint identification. Add
device instance ID field to gk20a_ctrl_priv which is generated
from gk20a level device instance id value.
Share this ID to userspace via gpu characteristics.
Bug 3677982
JIRA NVGPU-8681
Change-Id: I79d92a81c02272c52e24f5b12c452c8993137037
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2792079
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Scott Long <scottl@nvidia.com>
Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
Currently, the registeration with error injection utility is done
only for GA10b using HAL. But HALs are not initialized during the
probe stage when we try to register the error injection utility.
So, the callback registration does not happen HAL is set to NULL.
Move the callback registration from probe to poweron stage when HAL
is initialized.
Update the nvgpu_cic_mon_init_lut() API name as it is no longer
doing only LUT initialization.
Bug 3828050
Change-Id: Ide718029e9317124749b4a51c423ae70dc8227c8
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2790269
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Vaibhav Kachore <vkachore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
Implement the ioctls NVGPU_TSG_IOCTL_CREATE_SUBCONTEXT and
NVGPU_TSG_IOCTL_DELETE_SUBCONTEXT. These will allocate and
free the VEID numbers.
Address space association with the VEIDs is verified to
ensure that channels association with VEIDs and address
space remains consistent.
Bug 3677982
JIRA NVGPU-8681
Change-Id: I2d913baf61a6bdeec412c58270c0024b80ca15c6
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2766765
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
Added infrastructure for enabling parsing Control-Fifo's
ring buffers(i.e. send/receive).
Initialization of these buffers are handled as part of
nvgpu_nvs_buffer_alloc() call itself.
A follow-up change shall implement the methods defined here
as part of the existing NVS worker thread.
The changes adhered to the design laid out in the header
nvsched/include/nvs/nvs-control-interface.h.
Jira NVGPU-8619
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Change-Id: I2050e6fb681eba80e01cf547ada37a955e58315a
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2764518
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Use CONFIG_NVS_KMD_BACKEND to enclose all NVS KMD based scheduling
code.
Current configuration contains all the scheduling code managed within
CONFIG_NVS_PRESENT. Eventually, scheduling code shall only use GSP.
Hence, isolate KMD based scheduling code to a config
CONFIG_NVS_KMD_BACKEND. This shall make it easier to remove this code
later.
Jira NVGPU-8619
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Change-Id: I9dc668e0fa3e7706c111fda7a5e2415e1fc0dd03
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2769465
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
The error injection code was enabled only when CONFIG_NVGPU_DGPU = n
so that the dGPUs do not attempt any error injection callback
function registration. But, this introduced dependency on DGPU
config when needs to be explicitly set to n for error injection to
be enabled.
Remove the dependency by moving the error injection callback
registration and deregistration to a HAL which is enabled only
on GA10b.
Bug 3819160
Change-Id: I4f4eb99189b1af3502d719536a91cc5e5d866bce
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2787202
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Changes
- Initialize virtual memory for gsp. This space is used for creating
queues for ctrl fifo. Also can be used to ro map sync-pt to this
instance where gsp firmware can poll the sync-pt with sync-pt id.
- Enabled gsp context interface and written the instance block pointer
to nxtctx register for the gsp firmware to access created virtual memory.
- Added required gsp registers for this feature.
NVGPU-8730
Bug 3770916
Change-Id: If538f615eca3f9b7840ffe2787826528b4808886
Signed-off-by: rmylavarapu <rmylavarapu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2764649
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
- Perfmon cmds are non-blocking calls and response
may/may-not come during railgate sequence for the
perfmon command sent as part of nvgpu_pmu_destroy call.
- if response is missed then payload allocated will not be
freed and allocation info will be present as part seq data
structure.
- This will be carried forward for multiple railgate/
rail-ungate sequence and that will cause the memleak
when new allocation request is made for same seq-id.
- Cleanup the sequence data struct as part of nvgpu_pmu_destroy
call by freeing the memory if cb_params is not NULL.
Bug 3747586
Bug 3722721
Change-Id: I1a0f192197769acec12993ae575277e38c9ca9ca
Signed-off-by: Mahantesh Kumbar <mkumbar@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2763054
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Divya Singhatwaria <dsinghatwari@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
Tested-by: Divya Singhatwaria <dsinghatwari@nvidia.com>
Only certain combination of channels of GFX/Compute object classes can
be assigned to particular pbdma and/or VEID. CILP can be enabled only
in certain configs. Implement checks for the configurations verified
during alloc_obj_ctx and/or setting preemption mode.
Bug 3677982
Change-Id: Ie7026cbb240819c1727b3736ed34044d7138d3cd
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2719995
Reviewed-by: Ankur Kishore <ankkishore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
Subcontext PDBs and valid mask in the instance blocks of the channels
in various subcontexts has to be updated when new subcontext is
created or a subcontext is removed.
Replayable fault state is cached in the channel structure. Replayable
fault state for subcontext is set based on first channel's bind
parameter. It was earlier programmed in function channel_setup_ramfc.
init_inst_block_core is updated to setup TSG level pdb map and mask.
Added new hal gv11b_channel_bind to enable the subcontext on channel
bind.
Bug 3677982
Change-Id: I58156c5b3ab6309b6a4b8e72b0e798d6a39c1bee
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2719994
Reviewed-by: Ankur Kishore <ankkishore@nvidia.com>
GVS: Gerrit_Virtual_Submit <buildbot_gerritrpt@nvidia.com>
