channel_joblist_peek() returns NULL if the list is empty.
nvgpu_channel_joblist_is_empty() has been used only together with that
function; remove it and check against NULL to see whether there are jobs
in flight.
This removes some duplication, simplifies the call sites slightly, and
gets rid of a Coverity nag about a possible NULL pointer from peek that
really isn't (when the emptiness was already checked).
Jira NVGPU-5772
Change-Id: I814e9c510d99b88e59539359992fb44d4e7ce2ea
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2397394
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add a "priv" fence struct type and use that in the fence type to
emphasize that the inner data is not meant to be seen.
The fence unit needs to have an outside-visible fence type so that
fences can be allocated directly as a struct field in job metadata for
performance and simplicity, so hiding the type entirely wouldn't work.
A couple of places need to touch the priv data directly in channel code.
Those can be thought to be technically fence unit's code scattered
outside the fence files, but they mean that the architecture is not
perfect yet.
Jira NVGPU-5773
Change-Id: Ifa3c95757ae31eef0e32f2605293e23e210b065f
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2395071
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Split up nvgpu_channel_clean_up_jobs() on the clean_all parameter so
that there's one version for the asynchronous ("deferred") cleanup and
another for the synchronous deterministic cleanup that occurs in the
submit path.
Forking another version like this adds some repetition, but this lets us
look at both versions clearly in order to come up with a coherent plan.
For example, it might be feasible to have the light cleanup of pooled
items in also the nondeterministic path, and deferring heavy cleanup to
another, entirely separated job queue.
Jira NVGPU-5493
Change-Id: I5423fd474e5b8f7b273383f12302126f47076bd3
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2346065
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Move the sema-specific APIs to the sema-specific os fence header. Do the
same for syncpts. Stub out the os fence type and the initialized check
if os fence support is not enabled in build time. Guard the sema header
with CONFIG_NVGPU_SW_SEMAPHORE.
Jira NVGPU-5773
Change-Id: I838debd66a800b00cde76e65458b13eee367b55f
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2395070
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new APIs to bind/unbind PM resources to/from profiler objects:
nvgpu_profiler_bind_pm_resources()
nvgpu_profiler_unbind_pm_resources()
Implement support to bind/unbind SMPC/HWPM/HWPM_STREAMOUT in various
functions in common/profiler/profiler.c.
Unbind all the PM resources explicitly in
nvgpu_profiler_unbind_context() while closing the profiler object.
If resources are bound during a resource reservation request,
unbind the resources explicitly before reserving new resource.
It is responsibility of application to bind the PM resources again.
Bug 2510974
Jira NVGPU-5360
Change-Id: Ib2a0e017eaa23d0d376438771e8bf4e340865f03
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2389655
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Add two new functions to reserve/release PM resources :
nvgpu_prof_ioctl_reserve_pm_resource()
nvgpu_prof_ioctl_release_pm_resource()
Add ctxsw field to struct nvgpu_profiler_object to store per-resource
context switch enable flag.
Force resource reservation release while unbinding the context from
profiler object or while closing the profiler object. Add this code
in nvgpu_profiler_unbind_context() since both above paths will call
this function.
Bug 2510974
Jira NVGPU-5360
Change-Id: If334148e8df86360fba4162d1611187f3f04d01b
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2389654
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
The lockless allocator that spins in alloc and free ops using cmpxchg to
mitigate race conditions has only ever been used for the post fences in
preallocated job resources. Now each post fence has a clear owner (the
job struct which already is allocated well) and lifetime, so this
allocator has no longer a purpose. Delete it to avoid bitrot. (The
design of the job queue has always been such that there's minimal
contention in any case.)
Jira NVGPU-5773
Change-Id: Ied98d977c2c75bacfd3d010ce60c80fe709231e0
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2392705
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
As the submit job metadata has been simplified, the fence pool for job
tracking fences is now just complex code for very simple purposes, so
delete it. It's enough to hold the fence memory in the job struct itself
instead of having separately allocated objects with different lifetimes.
Each channel is using preallocated job arrays based on the prespecified
inflight job count. The fences are used for tracking job completion, and
a new job cannot be submitted before a previous wait has completed.
This means that even with a ringbuffer with space for only one job, the
previous job memory cannot get reclaimed by a new submit because the
submits are ordered.
Jira NVGPU-5773
Change-Id: I0c777df700aa7cfda6f971efa47aa72c5462b53a
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2392704
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Rework regops execution API to accomodate below updates for new
profiler design
- gops.regops.exec_regops() should accept TSG pointer instead of
channel pointer.
- Remove individual boolean parameters and add one flag field.
Below new flags are added to this API :
NVGPU_REG_OP_FLAG_MODE_ALL_OR_NONE
NVGPU_REG_OP_FLAG_MODE_CONTINUE_ON_ERROR
NVGPU_REG_OP_FLAG_ALL_PASSED
NVGPU_REG_OP_FLAG_DIRECT_OPS
Update other APIs, e.g. gr_gk20a_exec_ctx_ops() and validate_reg_ops()
as per new API changes.
Add new API gk20a_is_tsg_ctx_resident() to check context residency
from TSG pointer.
Convert gr_gk20a_ctx_patch_smpc() to a HAL gops.gr.ctx_patch_smpc().
Set this HAL only for gm20b since it is not required for later chips.
Also, remove subcontext code from this function since gm20b does not
support subcontext.
Remove stale comment about missing vGPU support in exec_regops_gk20a()
Bug 2510974
Jira NVGPU-5360
Change-Id: I3c25c34277b5ca88484da1e20d459118f15da102
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2389733
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Create new dev nodes for device and context profilers. Example of dev
nodes on iGPU
/dev/nvhost-prof-dev-gpu - device scope profiler
/dev/nvhost-prof-ctx-gpu - context scope profiler
Add below APIs to open/close above dev nodes :
nvgpu_prof_dev_fops_open()
nvgpu_prof_ctx_fops_open()
nvgpu_prof_fops_release()
Add common API nvgpu_prof_fops_ioctl() to handle IOCTL call on these
dev nodes. Add IOCTL NVGPU_PROFILER_IOCTL_BIND_CONTEXT to bind the TSG
to profiler objects.
Add nvgpu_tsg_get_from_file() to retrieve TSG struct pointer from
file descriptor. Also store profiler object pointer into TSG struct.
Enable NVGPU_SUPPORT_PROFILER_V2_DEVICE capability on gv11b and tu104.
Note that this is not yet enabled for vGPU.
Keep NVGPU_SUPPORT_PROFILER_V2_CONTEXT capabiity disabled since this
will take longer to support.
Add new IOCTL NVGPU_PROFILER_IOCTL_UNBIND_CONTEXT so that userspace can
explicitly unbind the context and release the resources before closing
the profiler descriptor.
Add context_init flag to profiler object for book keeping.
Bug 2510974
Jira NVGPU-5360
Change-Id: Ie07e0cfd5a9da9d80008f79c955c7ef93b4bc60f
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2384354
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
If recovery sequence profiling is enabled skip the debug dump that
happens during an MMU fault. This prevents the debug dump from
dominating the time spent by the recovery sequence. The debug dump
is severly limited in speed by the (lack of) UART bandwidth.
JIRA NVGPU-5606
Change-Id: Ifc7c326d33d9115d58b13c0fa42ec4bb7acb3075
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2382591
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
During recovery, we set ch->unserviceable at the end after we preempt
the TSG and reset the engines. It might be too late and user-space
might submit more work to the broken channel which is not desirable.
Move setting this unserviceable flag right at the start
of recovery sequence.
Another thread doing a submit can still read the unserviceable flag
just before it is set here, leaving that submit stuck if recovery
completes before the submit thread advances enough to set up a post
fence visible for other threads. This could be fixed with a big lock
or with a double check at the end of the submit code after the job
data has been made visible.
We still release the fences, semaphore and error notifier wait queues
at the end; so user-space would not trigger channel unbind while
channel is being recovered.
Also, change the handle_mmu_fault APIs to return void as the
debug_dump return value is not used in any of the caller APIs.
JIRA NVGPU-5843
Change-Id: Ib42c2816dd1dca542e4f630805411cab75fad90e
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2385256
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
During recovery, we preempt the faulty TSG from PBDMA and engines.
If the TSG preempt on PBDMA times out(timeout = 100ms), the PBDMA
might be hung state. We do not reset the HOST during recovery, so
stuck PBDMAs are unrecoverable.
Abort the recovery and trigger GPU to quiesce as there is no way
back.
Triggering Quiesce from recovery sequence should be fine as the only
redundant operation will be write to FIFO_RUNLIST_PREEMPT register.
The error notifiers will eventually be set by Quiesce thread.
Bug 2768005
JIRA NVGPU-4631
Change-Id: I914b9379aa8e48014e6ddace9abe47180a072863
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2368187
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
gk20a_perfbuf_map() allocates perfbuf VM, maps the user buffer into new
VM, and then triggers gops.perfbuf.perfbuf_enable(). This HAL then does
following :
- Allocate perfbuf instance block
- Initialize perfbuf instance block
- Reset stream buffer
- Program instance block address in PMA registers
- Program user buffer address into PMA registers
New profiler interface will have it's own API to setup PMA strem, and
it requires above setup to be done in two phases of perfbuf
initialization and then user buffer setup.
Split above functionalities into below functions
- nvgpu_perfbuf_init_vm()
- Allocate perfbuf VM
- Call gops.perfbuf.init_inst_block() to initialize perfbuf instance
block
- gops.perfbuf.init_inst_block()
- Allocate perfbuf instance block
- Initialize perfbuf instance block
- Program instance block address in PMA registers using
gops.perf.init_inst_block()
- In case of vGPU, trigger TEGRA_VGPU_CMD_PERFBUF_INST_BLOCK_MGT
command to gpu server
- gops.perf.init_inst_block()
- Reset stream buffer
- Program user buffer address into PMA registers
Also add corresponding cleanup functions as below :
gops.perf.deinit_inst_block()
gops.perfbuf.deinit_inst_block()
nvgpu_perfbuf_deinit_vm()
Bug 2510974
Jira NVGPU-5360
Change-Id: I486370f21012cbb7fea84fe46fb16db95bc16790
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2372984
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Currently the vGPU engine management rewrites a lot of the common
device agnostic engine management code.
With the new top HAL parsing one device at a time, it is now more
easily possible to tie the vGPU into the new common device framework
by implementing the top HAL but with the vGPU engine list backend.
This lets the vGPU inherit all the common engine and device
management code. By doing so the vGPU HAL need only implement a
trivial and simple HAL.
This also gets us a step closer to merging all of the CE init
code: logically it just iterates through all CE engines whatever
they may be. The only reason this differs between chips is because
of the swap from CE0-2 to LCEs in the Pascal generation. This could
be abstracted by the unit code easily enough.
Also, the pbdma_id for each engine has to be added to the device
struct. Eventually this was going to happen anyway, since the
device struct will soon replace the nvgpu_engine_info struct.
It's a little bit of an abuse but might be worth it long term. If
not, it should not be difficult to replace uses of dev->pbdma_id
with a proper lookup of PBDMA ID based on the device info.
JIRA NVGPU-5421
Change-Id: Ie8dcd3b0150184d58ca0f78940c2e7ca72994e64
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2351877
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add tu104 specific HAL tu104_gr_falcon_ctrl_ctxsw() that processes below
CTXSW methods to start/stop SMPC global mode :
NVGPU_GR_FALCON_METHOD_START_SMPC_GLOBAL_MODE
NVGPU_GR_FALCON_METHOD_STOP_SMPC_GLOBAL_MODE
Add new tu104 specific HAL tu104_gr_update_smpc_global_mode() to trigger
SMPC global mode start/stop using gops.gr.falcon.ctrl_ctxsw().
Update nvgpu_dbg_gpu_ioctl_smpc_ctxsw_mode() to enable/disable SMPC
global mode if channel is not bound to debug session.
Bug 2510974
Bug 2257799
Jira NVGPU-5360
Change-Id: I1f9d8f2a2d30a4738f291db3fc72c400d24f4048
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2368696
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Current PM resource reservation system is limited to HWPM resources
only. And reservation tracking is done using boolean variables.
New upcoming profiler support requires reservation for all the PM
resources like SMPC and PMA stream. Using boolean variables is
not scalable and confusing. Plus the variables have to be replicated
on gpu server in case of virtualization.
Remove flag tracking mechanism and use list based approach to track
all PM reservations. Also, current HALs are defined on debugger object.
Implement new HALs in new pm_reservation object since it is really an
independent functionality.
Add new source file common/profiler/pm_reservation.c which implements
functions to reserve/release resources and to check if any resource
is reserved or not.
Add common/vgpu/pm_reservation_vgpu.c for vGPU which simply forwards
the request to gpu server.
Define new HAL object gops.pm_reservation and assign above functions
to below respective HALs :
g->ops.pm_reservation.acquire()
g->ops.pm_reservation.release()
g->ops.pm_reservation.release_all_per_vmid()
Last HAL above is only used for gpu server cleanup of guest OS.
Add below new common profiler functions that act as APIs to reserve/
release resources for rest of the units in nvgpu.
nvgpu_profiler_pm_resource_reserve()
nvgpu_profiler_pm_resource_release()
Initialize the meta data required for reservtion system in
nvgpu_pm_reservation_init() and call it during nvgpu_finalize_poweron.
Clean up the meta data before releasing struct gk20a.
Delete below HALs :
g->ops.debugger.check_and_set_global_reservation()
g->ops.debugger.check_and_set_context_reservation()
g->ops.debugger.release_profiler_reservation()
Bug 2510974
Jira NVGPU-5360
Change-Id: I4d9f89c58c791b3b2e63099a8a603462e5319222
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2367224
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
1) In MIG mode, 2D, 3D, I2M and ZBC classes are not supported by
GR engine. NvGpu shall expose the HWCaps through
"struct nvgpu_gpu_characteristics".
2) NvGpu shall expose the following MIG related new caps through
"struct nvgpu_gpu_characteristics".
* mig_enabled - Flag to indicate whether MIG is enabled/disabled.
* gpu_instance_id - GPU instaces Id.
* gr_instance_id - graphics execution unit id.
* gr_sys_pipe_id - Sys pipe id of GR engine.
3) populate num_ppc_per_gpc - Pixel Processing cluster per GPC
4) populate max_veid_count_per_tsg - Maximum veid count per TSG
5) populate num_sub_partition_per_fbpa - Sub partition per FBPA.
JIRA NVGPU-5762
Change-Id: I06b5bcd3f568eb0b9c78c8fc6ce155b39aaeaba5
Signed-off-by: lm <lm@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2352100
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
TEGRA_VGPU_CMD_GET_ATTRIBUTE
TEGRA_VGPU_CMD_CHANNEL_FREE_GR_PATCH_CTX
TEGRA_VGPU_CMD_CHANNEL_UNMAP_GR_GLOBAL_CTX
TEGRA_VGPU_CMD_CHANNEL_SET_PRIORITY
TEGRA_VGPU_CMD_CHANNEL_SET_RUNLIST_INTERLEAVE
TEGRA_VGPU_CMD_CHANNEL_SET_TIMESLICE
TEGRA_VGPU_CMD_CHANNEL_FREE_HWPM_CTX
The above commands which are not used by clients anymore are being
removed.
Jira GVSCI-5155
Signed-off-by: Richard Zhao <rizhao@nvidia.com>
Change-Id: If5eef090308e6471a0e7aadf78878f1ad798ee9a
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2367556
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Unify the job metadata handling by deleting the parts that have handled
dynamically allocated job structs and fences. Now a channel can be in
one less mode than before which reduces branching in tricky places and
makes the submit/cleanup sequence easier to understand.
While preallocating all the resources upfront may increase average
memory consumption by some kilobytes, users of channels have to supply
the worst case numbers anyway and this preallocation has been already
done on deterministic channels.
Flip the channel_joblist_delete() call in nvgpu_channel_clean_up_jobs()
to be done after nvgpu_channel_free_job(). Deleting from the list (which
is a ringbuffer) makes it possible to reuse the job again, so the job
must be freed before that. The comment about using post_fence is no
longer valid; nvgpu_channel_abort() does not use fences.
This inverse order has not posed problems before because it's been buggy
only for deterministic channels, and such channels do not do the cleanup
asynchronously so no races are possible. With preallocated job list for
all channels, this would have become a problem.
Jira NVGPU-5492
Change-Id: I085066b0c9c2475e38be885a275d7be629725d64
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2346064
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Debarshi Dutta <ddutta@nvidia.com>
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Commit bd1ae5c9e1 ("gpu: nvgpu: fix MISRA 17.7 violations in mm") did
a seemingly harmless looking change in the cmpxchg() wrapper macro to
convert from atomic_compare_exchange_strong() to nvgpu_atomic_cmpxchg().
The latter is ultimately a wrapper for the former but the semantics are
different: the former takes old as a pointer and updates it for the read
value, while the latter takes it as a value and returns the read value.
The commit caused cmpxchg() to always return the old value, so a failing
compare has never been detected in a year and half.
This cmpxchg() is used only in the lockless allocator which is used only
in the fence code in deterministic kernel submits which hasn't been part
of safe code, so the broken code has been basically not used. (The
typecast from an integer pointer to an atomic pointer is a separate
concern.)
Jira NVGPU-5493
Change-Id: I932a69c6c185783c0e514e848e0ee6057ce74888
Signed-off-by: Konsta Hölttä <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2368118
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
On Linux, nvgpu mapping ioctl provides option to specify the access
type flags for the mapping. This support is not implemented for
other OS. For nvrm_gpu to know when to set these flags add new
enabled flag *_MAP_ACCESS_TYPE that is enabled only for Linux.
Bug 200621157
Change-Id: If1397bb0d5fdc5589458d92f24647afa586af1c2
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2363829
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Until now, all userspace buffers were mapped in the GMMU as Read & Write
(RW) by default. In order to enable the use cases which require the GPU
to only read the SYSMEM buffers and not inadvertently write to those,
map buffer ioctls need to provide interface to set the mapping access
type from the userspace.
Some of the use cases are:
1. A third party server process exposes shared memory that is
read-only to the client process, which does the GPU processing.
Registering this memory using cudaHostRegister API as read-only
in the client process will restict the access to Read Only type
from the GPU.
2. IO devices exposing streaming read-only data for processing by
the GPU.
3. For marking semantically read-only data as actually read-only
for the purposes of debugging data corruption.
This patch introduces new AS buffer mapping bitmask flag and
corresponding core VM mapping bitmask flag for representing
Read Only (RO) access type. By default, the access is set
as Read Write (RW).
Bug 200621157
Change-Id: I5ec9dec3ce089e577b86c43003d92b61eee4a90b
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2361750
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
