Mapping of large buffers to GMMU end up needing many
pages for the PTE tables. Allocating these one by one
can end up being a performance bottleneck, particularly
in the virtualized case.
This is adding the following changes:
- As the TLB invalidation doesn't have access to mem_off,
allow top-level allocation by alloc_cache_direct().
- Define NVGPU_PD_CACHE_SIZE, the allocation size for a new slab
for the PD cache, effectively set to 64K bytes
- Use the PD cache for any allocation < NVGPU_PD_CACHE_SIZE
When freeing up cached entries, avoid prefetch errors by
invalidating the entry (memset to 0).
- Try to fall back to direct allocation of smaller chunk for
contiguous allocation failures.
- Unit test changes.
Bug 200649243
Change-Id: I0a667af0ba01d9147c703e64fc970880e52a8fbc
Signed-off-by: dt <dt@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2404371
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Not sure if there's an actual bug or JIRA filed for this, but the
change here fixes a long standing bug in the MM code for unit tests.
Te GMMU programming code verifies that the CPU _physical_ address
programmed into the GMMU PDE0 is a valid Tegra SoC CPU physical
address. That means that it's not too large a value.
The POSIX imlementation of the nvgpu_mem related code used the CPU
virtual address as the "phys" address. Obviously, in userspace,
there's no access to physical addresses, so in some sense it's a
meaningless function. But the GMMU code does care, as described
above, about the format of the address.
The fix is simple enough: since the nvgpu_mem_get_addr() and
nvgpu_mem_get_phys_addr() values shouldn't actually be accessed by
the driver anyway (they could be vidmem addresses or IOVA addresses
in real life) ANDing them with 0xffffffff (e.g 32 bits) truncates
the potentially problematic CPU virtual address bits returned by
malloc() in the POSIX environment.
With this, a run of the unit test framework passes for me locally
on my Ubuntu 18 machine.
Also, clean up a few whitespace issues I noticed while I debugged
this and fix another long standing bug where the
NVGPU_DEFAULT_DBG_MASK was not being copied to g->log_mask during
gk20a struct init.
Change-Id: Ie92d3bd26240d194183b4376973d4d32cb6f9b8f
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2395953
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
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: Vijayakumar Subbu <vsubbu@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
MISRA Advisory Rule 2.5 states that a project should not contain unused
macro declarations.
While most of the violations in the nvgpu driver are due to unused
macros from hw headers, devinit-related headers, etc. there is a small
number that are due to things like:
* macros not being used when they could/should be
* macros in C files that are really not referenced
* CPP build flag mismatches
This change eliminates such violations from the following:
* replace constants with existing macros in timeout conversion code
* wrap nvgpu_gmmu_dbg macro #defines in #ifdef CONFIG_NVGPU_TRACE/#endif
* wrap MAX_MC_INTR_REGS #define in #ifdef CONFIG_NVGPU_NON_FUSA/#endif
* remove unused FECS_MAILBOX_0_ACK_RESTORE from runlist code
* wrap BACKTRACE_MAXSIZE macro with #ifndef _QNX_SOURCE/#endif
Jira NVGPU-3178
Change-Id: I2bc72f706d7af3f8e7b062126e8543d0dc8ac250
Signed-off-by: Scott Long <scottl@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2284419
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: Vedashree Vidwans <vvidwans@nvidia.com>
Reviewed-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-by: Adeel Raza <araza@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
MISRA Advisory Directive 4.5 states that identifiers in the same
name space with overlapping visibility should be typographically
unambiguous.
In both the nvgpu_locate_pte_last_level() and gp10b_get_pde0_pgz()
routines this violation is raised because a variable used as
a for-loop index ('i') is ambiguous with a parameter that points
to a struct gk20a_mmu_level ('l').
To fix these violations the loop index 'i' is renamed to 'idx'.
Jira NVGPU-3178
Change-Id: I2cec904201075b48ab6ccfbd0ff6d7e9dcac2867
Signed-off-by: Scott Long <scottl@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2271456
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: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Adeel Raza <araza@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
MISRA rule 11.2 doesn't allow conversions of a pointer from or to an
incomplete type. These type of conversions may result in a pointer
aligned incorrectly and may further result in undefined behavior.
This patch addresses rule 11.2 violations related to pointers to and
from struct nvgpu_sgl. This patch replaces struct nvgpu_sgl pointers by
void pointers.
Jira NVGPU-3736
Change-Id: I8fd5766eacace596f2761b308bce79f22f2cb207
Signed-off-by: Vedashree Vidwans <vvidwans@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2267876
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Whitelist 2 MISRA Rule 17.2 violations in MM that were approved as
deviations in TID-278. These two violations are for recursive functions
that handle page table descriptors in the GMMU page table. Both cases
are tightly controlled recursion by limiting the recursion depth to the
number of possible page table levels in the hardware. For current
hardware that is a maximum recursion depth of 5 which is easily an
acceptable depth and should cause no stack issues.
JIRA NVGPU-3489
JIRA NVGPU-3492
JIRA TID-278
Change-Id: I5b801ff77f66bb8698f1d6adcd41ebbad3f86f92
Signed-off-by: Philip Elcan <pelcan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2230077
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: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
The two MM functions nvgpu_set_pd_level() and nvgpu_vm_do_free_entries()
are simple recursive functions for processing page descriptors (PDs) by
traversing the levels in the PDs. MISRA Rule 17.2 prohibits functions
calling themselves because "unless recursion is tightly controlled, it
is not possible to determine before execution what the worst-case stack
usage could be." So, this change limits the recursion depth of each of
these functions to the maximum number of page table levels by checking
the level against the MM HAL max_page_table_levels().
This also required configuring this HAL in various unit tests as they
were no previously using this HAL.
JIRA NVGPU-3489
Change-Id: Iadee3fa5ba9f45cd643ac6c202e9296d75d51880
Signed-off-by: Philip Elcan <pelcan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2224450
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add macros for whitelisting coverity violations. These macros use pragma
directives. The pragma directives and whitelisting macros are only
enabled when a coverity scan is being run.
The whitelisting macros have been added to a new header called
static_analysis.h. The contents of safe_ops.h (CERT C safe ops) have
been moved into static_analysis.h because this will be the new header
for static analysis related macros/defines/etc.
JIRA NVGPU-3820
Change-Id: I9c63f20f670880b420415535738034619314b7c3
Signed-off-by: Adeel Raza <araza@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2180600
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
MISRA Advisory Rule 10.5 states that the value of an expression should
not be cast to an inappropriate essential type.
This change removes five violations of this rule that involve casting
boolean results to unsigned values:
* pass 1/0 (instead of true/false) to nvgpu_atomic operations
* fix skip_mask handling in gm20b_gr_init_pd_skip_table_gpc()
* fix l3_alloc flags check in nvgpu_gmmu_map_locked()
This change also eliminates several MISRA Advisory Rule 4.6 violations.
JIRA NVGPU-3798
Change-Id: I707da8a812bfb32eaeb2200463885c0961b197b3
Signed-off-by: Scott Long <scottl@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2153070
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
CERT-C INT32 states to ensure that operations on signed integers do not
overflow. In page_table.c, the parameter lvl was triggering INT32
violations when adding to the value.
To address these violations, do two things. First, since this is
really an unsigned value, make it a u32. And rather than just make the
INT32 violations INT30 violations, use the safe operations.
JIRA NVGPU-3515
Change-Id: Iabd5e239813c695638988143ee901b1c19a0df5d
Signed-off-by: Philip Elcan <pelcan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2125028
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Vinod Gopalakrishnakurup <vinodg@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
There are many miscellaneous HALs for various MM related functionality.
This patch aims to migrate all the remaining MM code from the <chip>/
mm_<chip>.[ch] files in HAL files under hal/.
Much of this is fairly straightforward copy/paste and updates to the
HAL init files.
The exception to that is the move of the left over gv11b MMU fault
handling code in mm_gv11b.c. Having both a hal/mm/mm/mm_gv11b.c and
a gv11b/mm_gv11b.c file causes tmake to choke so the gv11b/mm_gv11b.c
file was moved to gv11b/mmu_fault_gv11b.c. This will be cleaned up in
a subsequent patch.
JIRA NVGPU-2042
Change-Id: I12896de865d890a61afbcb71159cff486119ffb8
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2109050
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Make a hal/mm/gmmu sub-unit for the GMMU HAL code. Also move the
gk20a specific HAL code there. gp10b will happen in the next patch.
This change also updates all the GMMU related HAL usage, of which
there is quite a bit. Generally the only change is a .gmmu needs to
be inserted into the HAL path. Each HAL init was also updated.
JIRA NVGPU-2042
Change-Id: I6c46bdfddb8e021f56103d9457fb3e2a226f8947
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2099693
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Rename the two native GPU GMMU map/unmap functions and update the
HAL initializations to reflect this:
gk20a_locked_gmmu_map -> nvgpu_gmmu_map_locked
gk20a_locked_gmmu_unmap -> nvgpu_gmmu_unmap_locked
This matches what other units do for handling vGPU "HAL" indirection.
Also move the function declarations to <nvgpu/gmmu.h> since these are
shared among all non-vGPU chips. But since these are still technically
HAL operations they should never be called directly. This is a bit of
an organixational issue that I have not thought through hwo to solve
yet.
Ideally they would go into a "hal/mm/gmmu/" include somewhere, but
that A) doesn't yet exist, and B) those are chip specific; these
functions are native specific. Ugh.
JIRA NVGPU-2042
Change-Id: Ibc614f2928630d12eafcec6ce73019628b44ad94
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2099692
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add a new MM HAL directory to contain all MM related HAL units.
As part of this change add cache unit to the MM HAL. This contains
several related fixes:
1. Move the cache code in gk20a/mm_gk20a.c and gv11b/mm_gv11b.c to
the new cache HAL. Update makefiles and header includes to take
this into account. Also rename gk20a_{read,write}l() to their
nvgpu_ variants.
2. Update the MM gops: move the cache related functions to the new
cache HAL and update all calls to this HAL to reflect the new
name.
3. Update some direct calls to gk20a MM cache ops to pass through
the HAL instead.
4. Update the unit tests for various MM related things to use the
new MM HAL locations.
This change accomplishes two architecture design goals. Firstly it
removes a multiple HW include from mm_gk20a.c (the flush HW header).
Secondly it moves code from the gk20a/ and gv11b/ directories into
more proper locations under hal/.
JIRA NVGPU-2042
Change-Id: I91e4bdca4341be4dbb46fabd72622b917769f4a6
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2095749
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
The fb APIs compression_page_size() and compression_align_mask() were
returning u32s, but the users all really need u64s.
This also eliminates MISRA Rule 10.3 violations for implicit casts to
smaller size objects.
JIRA NVGPU-2954
Change-Id: I8dc2b434d9564c89c0e8a1b19c4acbe167e339c1
Signed-off-by: Philip Elcan <pelcan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2075595
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
MISRA Rule-17.7 requires the return value of all functions to be used.
Fix is either to use the return value or change the function to return
void. This patch ensures that WARN and WARN_ON always return void; and
introduces a new nvgpu_do_assert construct to trigger the equivalent
of WARN_ON(true) so that stack can be dumped (depends on OS support)
JIRA NVGPU-677
Change-Id: Ie2312c5588ceb5b1db825d15a096149b63b69af4
Signed-off-by: Nicolas Benech <nbenech@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2018706
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new variable in nvgpu_as_map_buffer_ex_args for app
to specify the platform atomic support for the page.
When platform atomic attribute flag is set, pte memory
aperture is set to be coherent type.
renamed nvgpu_aperture_mask_coh -> nvgpu_aperture_mask_raw
function.
bug 200473147
Change-Id: I18266724dafdc8dfd96a0711f23cf08e23682afc
Signed-off-by: Vinod G <vinodg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2012679
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
gv11b_mm_l2_flush was not checking error codes from the various
functions it was calling. MISRA Rule-17.7 requires the return value
of all functions to be used. This patch now checks return values and
propagates the error upstream.
JIRA NVGPU-677
Change-Id: I9005c6d3a406f9665d318014d21a1da34f87ca30
Signed-off-by: Nicolas Benech <nbenech@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1998809
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
MISRA Rule 10.4 only allows the usage of arithmetic operations on
operands of the same essential type category.
Adding "U" at the end of the integer literals or casting operands
to have same type of operands when an arithmetic operation is
performed.
This fixes violations where an arithmetic operation is performed on
signed and unsigned int types.
JIRA NVGPU-992
Change-Id: I27e3e59c3559c377b4bd3cbcfced90fdf90350f2
Signed-off-by: Sai Nikhil <snikhil@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1921459
Reviewed-by: svc-misra-checker <svc-misra-checker@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Adeel Raza <araza@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
The pd_cache header declarations were oriignally part of the
gmmu.h header. This is not good from a unit isolation perspective
so this patch moves all the pd_cache specifics over to a new
header file: <nvgpu/pd_cache.h>.
Also a couple of static inlines that were possible when the code
was part of gmmu.h were turned into real, first class functions.
This allowed the pd_cache.h header to not include the gmmu.h
header file.
Also fix an issue in the nvgpu_pd_write() function where the data
was being passed as a size_t for some reason. This has now been
changed to a u32.
JIRA NVGPU-1444
Change-Id: Ib9e9e5a54544de403bfcd8e11c30de05721ddbcc
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1966352
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
This reverts commit 15603b9fd5.
Causes a build break in the PD cache unit test. Not sure how this
passed GVS - must have been a race or something? Unclear.
Change-Id: Ia484a801d098d69441326fa1dd40a1c86e2e23ce
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1966335
