Open source GPL/LGPL release

This commit is contained in:
svcmobrel-release
2025-12-19 15:25:44 -08:00
commit 9fc87a7ec7
2261 changed files with 576825 additions and 0 deletions

View File

@@ -0,0 +1,26 @@
# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
.SUFFIXES:
OBJS = page_allocator.o
MODULE = page_allocator
include ../../../Makefile.units

View File

@@ -0,0 +1,33 @@
################################### tell Emacs this is a -*- makefile-gmake -*-
#
# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
#
###############################################################################
NVGPU_UNIT_NAME=page_allocator
include $(NV_COMPONENT_DIR)/../../../Makefile.units.common.interface.tmk
# Local Variables:
# indent-tabs-mode: t
# tab-width: 8
# End:
# vi: set tabstop=8 noexpandtab:

View File

@@ -0,0 +1,33 @@
################################### tell Emacs this is a -*- makefile-gmake -*-
#
# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
#
###############################################################################
NVGPU_UNIT_NAME=page_allocator
include $(NV_COMPONENT_DIR)/../../../Makefile.units.common.tmk
# Local Variables:
# indent-tabs-mode: t
# tab-width: 8
# End:
# vi: set tabstop=8 noexpandtab:

View File

@@ -0,0 +1,644 @@
/*
* Copyright (c) 2019-2020, NVIDIA CORPORATION. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
#include <unit/io.h>
#include <unit/unit.h>
#include <nvgpu/gk20a.h>
#include <nvgpu/sizes.h>
#include <nvgpu/types.h>
#include <nvgpu/allocator.h>
#include <nvgpu/posix/kmem.h>
#include <nvgpu/posix/posix-fault-injection.h>
#include <nvgpu/page_allocator.h>
#include "page_allocator.h"
#ifdef CONFIG_NVGPU_DGPU
#define BA_DEFAULT_BASE SZ_4K
#define BA_DEFAULT_LENGTH SZ_1M
#define BA_DEFAULT_BLK_SIZE SZ_4K
#define SZ_2K (SZ_1K << 1)
#define SZ_8K (SZ_4K << 1)
#define SZ_16K (SZ_4K << 2)
#define SZ_32K (SZ_64K >> 1)
static struct nvgpu_allocator *na;
/*
* @fault_enable- Enable/Disable fault injection
* @fault_at - If fault is enabled,
fault_at contains fault_counter value
otherwise,
fault_at should be set to 0
* @base - Base address of allocation in case of fixed allocation
* @len - Length of memory to be allocated
* @flags - Additional flags to be enabled
* @ret_addr - Return address of allocation, this is used in free()
* @expected_zero - Expected result of test
* @error_msg - Message to be displayed if test fails
*/
struct test_parameters {
bool fault_enable;
u32 fault_at;
u64 base;
u64 len;
u64 flags;
u64 ret_addr;
bool expected_zero;
char *error_msg;
};
static struct test_parameters fault_at_alloc_cache = {
.fault_enable = true,
.fault_at = 0,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "alloced despite fault injection at alloc_cache",
};
static struct test_parameters fault_at_nvgpu_alloc = {
.fault_enable = true,
.fault_at = 1,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "alloced despite fault injection at nvgpu_alloc",
};
static struct test_parameters fault_at_sgl_alloc = {
.fault_enable = true,
.fault_at = 1,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "alloced despite fault injection at sgl alloc",
};
static struct test_parameters fault_at_page_cache = {
.fault_enable = true,
.fault_at = 2,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "alloced despite fault injection at page_cache",
};
static struct test_parameters first_simple_alloc_32K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "first instance of 32K alloc failed",
};
static struct test_parameters second_simple_alloc_32K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "second instance of 32K alloc failed",
};
static struct test_parameters third_simple_alloc_32K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "third instance of 32K alloc failed",
};
static struct test_parameters fourth_simple_alloc_32K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_32K,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "fourth instance of 32K alloc failed",
};
static struct test_parameters failing_alloc_16K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_16K,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "16K alloc is supposed to fail",
};
static struct test_parameters simple_alloc_8K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_8K,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "8K alloc failed",
};
static struct test_parameters failing_alloc_8K = {
.fault_enable = false,
.fault_at = 0,
.base = SZ_64K,
.len = SZ_8K,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "8K alloc supposed to fail",
};
static struct test_parameters alloc_no_scatter_gather = {
.fault_enable = false,
.fault_at = 0,
.base = SZ_64K,
.len = SZ_32K,
.flags = GPU_ALLOC_NO_SCATTER_GATHER,
.expected_zero = false,
.error_msg = "32K alloc failed with no_scatter_gather enabled",
};
static struct test_parameters simple_alloc_128K = {
.fault_enable = false,
.fault_at = 0,
.base = SZ_128K << 2,
.len = SZ_128K,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "128K alloc failed",
};
static struct test_parameters alloc_contiguous = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_128K << 2,
.flags = GPU_ALLOC_FORCE_CONTIG,
.expected_zero = true,
.error_msg = "contiguous alloc should have failed",
};
static struct test_parameters simple_alloc_512K = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_128K << 2,
.flags = 0ULL,
.expected_zero = false,
.error_msg = "8K alloc failed",
};
static struct test_parameters alloc_more_than_available = {
.fault_enable = false,
.fault_at = 0,
.base = 0ULL,
.len = SZ_1M,
.flags = 0ULL,
.expected_zero = true,
.error_msg = "Allocated more than available memory",
};
int test_page_alloc(struct unit_module *m, struct gk20a *g, void *args)
{
struct test_parameters *param = (struct test_parameters *) args;
struct nvgpu_page_allocator *pa = page_allocator(na);
struct nvgpu_posix_fault_inj *kmem_fi =
nvgpu_kmem_get_fault_injection();
pa->flags |= param->flags;
nvgpu_posix_enable_fault_injection(kmem_fi,
param->fault_enable, param->fault_at);
param->ret_addr = na->ops->alloc(na, param->len);
nvgpu_posix_enable_fault_injection(kmem_fi, false, 0);
pa->flags &= ~(param->flags);
if ((param->expected_zero && (param->ret_addr == 0)) ||
(!param->expected_zero && (param->ret_addr != 0))) {
return UNIT_SUCCESS;
} else {
unit_return_fail(m, "%s", param->error_msg);
}
}
int test_page_free(struct unit_module *m, struct gk20a *g, void *args)
{
struct test_parameters *param = (struct test_parameters *) args;
struct nvgpu_page_allocator *pa = page_allocator(na);
pa->flags |= param->flags;
na->ops->free_alloc(na, param->ret_addr);
pa->flags &= ~(param->flags);
return UNIT_SUCCESS;
}
int test_page_alloc_fixed(struct unit_module *m, struct gk20a *g, void *args)
{
struct test_parameters *param = (struct test_parameters *) args;
struct nvgpu_page_allocator *pa = page_allocator(na);
struct nvgpu_posix_fault_inj *kmem_fi =
nvgpu_kmem_get_fault_injection();
pa->flags |= param->flags;
nvgpu_posix_enable_fault_injection(kmem_fi,
param->fault_enable, param->fault_at);
/* page_size = SZ_4K ignored in the function */
param->ret_addr = na->ops->alloc_fixed(na,
param->base, param->len, SZ_4K);
nvgpu_posix_enable_fault_injection(kmem_fi, false, 0);
pa->flags &= ~(param->flags);
if ((param->expected_zero && (param->ret_addr == 0)) ||
(!param->expected_zero && (param->ret_addr != 0))) {
return UNIT_SUCCESS;
} else {
unit_return_fail(m, "%s", param->error_msg);
}
}
int test_page_free_fixed(struct unit_module *m, struct gk20a *g, void *args)
{
struct test_parameters *param = (struct test_parameters *) args;
struct nvgpu_page_allocator *pa = page_allocator(na);
pa->flags |= param->flags;
na->ops->free_fixed(na, param->ret_addr, param->len);
pa->flags &= ~(param->flags);
return UNIT_SUCCESS;
}
int test_page_allocator_init_slabs(struct unit_module *m,
struct gk20a *g, void *args)
{
u64 base = SZ_64K;
u64 length = SZ_128K;
u64 blk_size = SZ_64K;
u64 flags = GPU_ALLOC_4K_VIDMEM_PAGES;
struct nvgpu_posix_fault_inj *kmem_fi =
nvgpu_kmem_get_fault_injection();
na = (struct nvgpu_allocator *)
nvgpu_kzalloc(g, sizeof(struct nvgpu_allocator));
if (na == NULL) {
unit_return_fail(m, "Could not allocate nvgpu_allocator\n");
}
/* Fault injection at init_slabs */
nvgpu_posix_enable_fault_injection(kmem_fi, true, 3);
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m,
"pa with slabs inited despite fault injection\n");
}
nvgpu_posix_enable_fault_injection(kmem_fi, false, 0);
/*
* Expect to fail as blk_size is odd
*/
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
SZ_4K + 1ULL, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m,
"vidmem page allocator inited with odd blk_size\n");
}
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
SZ_4K, 0ULL, flags, PAGE_ALLOCATOR) != 0) {
unit_return_fail(m,
"vidmem page allocator inited with odd blk_size\n");
} else {
na->ops->fini(na);
}
/*
* Initialize page allocator
* This will be used for further tests.
*/
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) != 0) {
unit_return_fail(m, "init with slabs failed\n");
}
return UNIT_SUCCESS;
}
int test_page_allocator_sgt_ops(struct unit_module *m,
struct gk20a *g, void *args)
{
u64 addr;
void *sgl = NULL;
struct nvgpu_page_alloc *alloc = NULL;
addr = na->ops->alloc(na, SZ_32K);
if (addr == 0) {
unit_return_fail(m, "couldn't allocate 32K");
}
/* Test page allocator sgt ops */
alloc = (struct nvgpu_page_alloc *) addr;
sgl = alloc->sgt.sgl;
if (alloc->sgt.ops->sgl_next(sgl) != NULL) {
unit_return_fail(m, "sgl_next should be NULL\n");
}
if (alloc->sgt.ops->sgl_phys(g, sgl) != alloc->base) {
unit_return_fail(m, "sgl_phys != base address\n");
}
if (alloc->sgt.ops->sgl_ipa(g, sgl) != alloc->base) {
unit_return_fail(m, "sgl_ipa != base address\n");
}
if (alloc->sgt.ops->sgl_dma(sgl) != alloc->base) {
unit_return_fail(m, "sgl_dma != base address\n");
}
if (alloc->sgt.ops->sgl_gpu_addr(g, sgl, NULL) != alloc->base) {
unit_return_fail(m, "sgl_gpu_addr != base address\n");
}
if (alloc->sgt.ops->sgl_ipa_to_pa(g, sgl, SZ_4K, NULL) != SZ_4K) {
unit_return_fail(m, "sgl_ipa_to_pa != SZ_4K\n");
}
if (alloc->sgt.ops->sgl_length(sgl) != SZ_32K) {
unit_return_fail(m, "sgl_length != SZ_32K\n");
}
alloc->sgt.ops->sgt_free(g, &alloc->sgt);
na->ops->free_alloc(na, addr);
return UNIT_SUCCESS;
}
int test_nvgpu_page_allocator_ops(struct unit_module *m,
struct gk20a *g, void *args)
{
u64 addr;
int err;
struct nvgpu_alloc_carveout test_co =
NVGPU_CARVEOUT("test_co", 0ULL, 0ULL);
test_co.base = BA_DEFAULT_BASE;
test_co.length = SZ_8K;
if (!na->ops->inited(na)) {
unit_return_fail(m, "ops not inited\n");
}
addr = na->ops->base(na);
if (addr != BA_DEFAULT_BASE) {
unit_return_fail(m, "base incorrect\n");
}
addr = na->ops->length(na);
if (addr != BA_DEFAULT_LENGTH) {
unit_return_fail(m, "length incorrect\n");
}
addr = na->ops->end(na);
if (addr != (BA_DEFAULT_BASE + BA_DEFAULT_LENGTH)) {
unit_return_fail(m, "end incorrect\n");
}
addr = na->ops->space(na);
if (addr == 0) {
unit_return_fail(m, "zero space allocated\n");
}
err = na->ops->reserve_carveout(na, &test_co);
if (err < 0) {
unit_return_fail(m, "couldn't reserve 8K carveout\n");
}
na->ops->release_carveout(na, &test_co);
addr = na->ops->alloc(na, SZ_32K);
if (addr == 0) {
unit_return_fail(m, "couldn't allocate 32K");
}
err = na->ops->reserve_carveout(na, &test_co);
if (err == 0) {
unit_return_fail(m, "reserved carveout after alloc\n");
}
na->ops->free_alloc(na, addr);
return UNIT_SUCCESS;
}
int test_nvgpu_page_allocator_destroy(struct unit_module *m,
struct gk20a *g, void *args)
{
na->ops->fini(na);
if (na->priv != NULL) {
unit_return_fail(m, "page allocator destroy failed\n");
}
nvgpu_kfree(g, na);
return UNIT_SUCCESS;
}
int test_nvgpu_page_allocator_init(struct unit_module *m,
struct gk20a *g, void *args)
{
u64 base = BA_DEFAULT_BASE;
u64 length = BA_DEFAULT_LENGTH;
u64 blk_size = BA_DEFAULT_BLK_SIZE;
u64 flags = 0ULL;
struct nvgpu_posix_fault_inj *kmem_fi =
nvgpu_kmem_get_fault_injection();
na = (struct nvgpu_allocator *)
nvgpu_kzalloc(g, sizeof(struct nvgpu_allocator));
if (na == NULL) {
unit_return_fail(m, "Could not allocate nvgpu_allocator\n");
}
/*
* expect to fail as blk_size < SZ_4K
*/
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
0ULL, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m, "inited despite blk_size = 0\n");
}
/* Fault injection at nvgpu_page_allocator allocation */
nvgpu_posix_enable_fault_injection(kmem_fi, true, 0);
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m,
"inited despite fault injection at page_allocator\n");
}
nvgpu_posix_enable_fault_injection(kmem_fi, false, 0);
/* Fault injection at alloc_cache */
nvgpu_posix_enable_fault_injection(kmem_fi, true, 1);
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m,
"inited despite fault injection at alloc_cache\n");
}
nvgpu_posix_enable_fault_injection(kmem_fi, false, 0);
/* Fault injection at slab_page_cache */
nvgpu_posix_enable_fault_injection(kmem_fi, true, 2);
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m,
"inited despite fault injection at slab_page_cache\n");
}
nvgpu_posix_enable_fault_injection(kmem_fi, false, 0);
/*
* expect to fail as blk_size is odd
*/
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
SZ_4K + 3ULL, 0ULL, flags, PAGE_ALLOCATOR) == 0) {
unit_return_fail(m, "inited despite odd blk_size\n");
}
/* base = 0 */
if (nvgpu_allocator_init(g, na, NULL, "test_page", 0ULL, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) != 0) {
unit_return_fail(m, "init failed with base = 0\n");
} else {
na->ops->fini(na);
}
/*
* Initialize page allocator
* This will be used for further tests.
*/
if (nvgpu_allocator_init(g, na, NULL, "test_page", base, length,
blk_size, 0ULL, flags, PAGE_ALLOCATOR) != 0) {
unit_return_fail(m, "init failed\n");
}
return UNIT_SUCCESS;
}
#endif
struct unit_module_test page_allocator_tests[] = {
#ifdef CONFIG_NVGPU_DGPU
/* These tests create and evaluate page_allocator w/o 4K VIDMEM pages */
UNIT_TEST(init, test_nvgpu_page_allocator_init, NULL, 0),
UNIT_TEST(ops, test_nvgpu_page_allocator_ops, NULL, 0),
UNIT_TEST(sgt_ops, test_page_allocator_sgt_ops, NULL, 0),
/* Below tests examine page allocation */
/*
* NOTE: The test order should not be changed. Previous test develop
* memory allocation arrangement required for later tests.
*/
/* These tests check execution with fault injection at various locations */
UNIT_TEST(fixed_alloc_fault_at_alloc_cache, test_page_alloc_fixed, (void *) &fault_at_alloc_cache, 0),
UNIT_TEST(fixed_alloc_fault_at_sgl_alloc, test_page_alloc_fixed, (void *) &fault_at_sgl_alloc, 0),
UNIT_TEST(alloc_fault_at_alloc_cache, test_page_alloc, (void *) &fault_at_alloc_cache, 0),
UNIT_TEST(alloc_fault_at_nvgpu_alloc, test_page_alloc, (void *) &fault_at_nvgpu_alloc, 0),
/* Alloc some memory, this ensures fault injection at sgl alloc in next test */
UNIT_TEST(simple_32K_alloc, test_page_alloc, (void *) &first_simple_alloc_32K, 0),
UNIT_TEST(alloc_fault_at_sgl_alloc, test_page_alloc, (void *) &fault_at_sgl_alloc, 0),
/* Test different allocation scenarios using simple alloc function */
UNIT_TEST(alloc_no_scatter_gather, test_page_alloc, (void *) &alloc_no_scatter_gather, 0),
UNIT_TEST(free_no_scatter_gather, test_page_free, (void *) &alloc_no_scatter_gather, 0),
/* Second free call checks execution when address is NULL */
UNIT_TEST(free_no_scatter_gather_again, test_page_free, (void *) &alloc_no_scatter_gather, 0),
UNIT_TEST(free_32K_alloc, test_page_free, (void *) &first_simple_alloc_32K, 0),
UNIT_TEST(fixed_alloc_128K, test_page_alloc_fixed, (void *) &simple_alloc_128K, 0),
/* After previous allocations, contiguous 512K memory isn't available */
UNIT_TEST(contiguous_alloc_512K, test_page_alloc, (void *) &alloc_contiguous, 0),
UNIT_TEST(simple_alloc_512K, test_page_alloc, (void *) &simple_alloc_512K, 0),
UNIT_TEST(alloc_more_than_available, test_page_alloc, (void *) &alloc_more_than_available, 0),
UNIT_TEST(free_alloc_512K, test_page_free, (void *) &simple_alloc_512K, 0),
UNIT_TEST(alloc_fixed_no_scatter_gather, test_page_alloc_fixed, (void *) &alloc_no_scatter_gather, 0),
UNIT_TEST(free_fixed_no_scatter_gather, test_page_free_fixed, (void *) &alloc_no_scatter_gather, 0),
/* Second free call checks execution when address is NULL */
UNIT_TEST(free_fixed_no_scatter_gather_again, test_page_free_fixed, (void *) &alloc_no_scatter_gather, 0),
UNIT_TEST(free_fixed_128K, test_page_free_fixed, (void *) &simple_alloc_128K, 0),
UNIT_TEST(destroy, test_nvgpu_page_allocator_destroy, NULL, 0),
/* These tests create and evaluate page_allocator w/ 4K VIDMEM pages */
UNIT_TEST(init_slabs, test_page_allocator_init_slabs, NULL, 0),
/* Below tests examine slab allocation */
/*
* NOTE: The test order should not be changed. A test contructs
* required memory structure for later tests.
*/
/* These tests check execution with fault injection at various locations */
UNIT_TEST(slabs_fault_at_alloc_cache, test_page_alloc, (void *) &fault_at_alloc_cache, 0),
UNIT_TEST(slabs_fault_at_sgl_alloc, test_page_alloc, (void *) &fault_at_sgl_alloc, 0),
UNIT_TEST(slabs_fault_at_page_cache, test_page_alloc, (void *) &fault_at_page_cache, 0),
/* Test different allocation scenarios */
UNIT_TEST(add_partial_slab, test_page_alloc, (void *) &first_simple_alloc_32K, 0),
UNIT_TEST(add_full_slab, test_page_alloc, (void *) &second_simple_alloc_32K, 0),
UNIT_TEST(add_second_partial_slab, test_page_alloc, (void *) &third_simple_alloc_32K, 0),
UNIT_TEST(add_second_full_slab, test_page_alloc, (void *) &fourth_simple_alloc_32K, 0),
/* Note: No free memory available for allocation */
UNIT_TEST(fixed_alloc_8K, test_page_alloc_fixed, (void *) &failing_alloc_8K, 0),
/* Freeing allocated slabs, adds slabs to empty and free lists */
UNIT_TEST(revert_partial_slab, test_page_free, (void *) &fourth_simple_alloc_32K, 0),
UNIT_TEST(revert_second_partial_slab, test_page_free, (void *) &second_simple_alloc_32K, 0),
UNIT_TEST(add_empty_slab, test_page_free, (void *) &first_simple_alloc_32K, 0),
UNIT_TEST(free_slab, test_page_free, (void *) &third_simple_alloc_32K, 0),
UNIT_TEST(slabs_alloc_8K, test_page_alloc, (void *) &simple_alloc_8K, 0),
UNIT_TEST(slabs_alloc_32K, test_page_alloc, (void *) &first_simple_alloc_32K, 0),
/*
* Note: Page allocator has only 2 slabs.
* These are now allocated for 8K and 32K chunks
*/
UNIT_TEST(no_more_slabs, test_page_alloc, (void *) &failing_alloc_16K, 0),
UNIT_TEST(destroy_slabs, test_nvgpu_page_allocator_destroy, NULL, 0),
#endif
};
UNIT_MODULE(page_allocator, page_allocator_tests, UNIT_PRIO_NVGPU_TEST);

View File

@@ -0,0 +1,258 @@
/*
* Copyright (c) 2020, NVIDIA CORPORATION. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
#ifndef UNIT_PAGE_ALLOCATOR_H
#define UNIT_PAGE_ALLOCATOR_H
struct gk20a;
struct unit_module;
/** @addtogroup SWUTS-mm-allocators-page-allocator
* @{
*
* Software Unit Test Specification for mm.allocators.page_allocator
*/
/**
* Test specification for: test_nvgpu_page_allocator_init
*
* Description: Initialize page allocator.
*
* Test Type: Feature, Error injection
*
* Targets: nvgpu_allocator_init, nvgpu_page_allocator_init,
* nvgpu_allocator.ops.fini
*
* Input: None
*
* Steps:
* - Initialize page allocator with following characteristics.
* - 4K memory base address.
* - 1M length of memory.
* - 4K block size.
* - Check that page allocator initializations fails for scenarios such as
* odd value of block_size and fault injection for memory allocation.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_nvgpu_page_allocator_init(struct unit_module *m,
struct gk20a *g, void *args);
/**
* Test specification for: test_nvgpu_page_allocator_ops
*
* Description: Test page allocator ops
*
* Test Type: Feature
*
* Targets: nvgpu_allocator.ops.alloc, nvgpu_allocator.ops.free_alloc,
* nvgpu_allocator.ops.reserve_carveout,
* nvgpu_allocator.ops.release_carveout, nvgpu_allocator.ops.base,
* nvgpu_allocator.ops.end, nvgpu_allocator.ops.length,
* nvgpu_allocator.ops.inited, nvgpu_allocator.ops.space
*
* Input: test_nvgpu_page_allocator_init
*
* Steps:
* - Check page_allocator attributes using allocator ops.
* - Execute allocator ops to read attibute value.
* - Confirm that value is equal to the default values set during
* initialization.
* - Allocate carveout and confirm that allocation is successful. Check that
* carveout cannot be reserved after normal page allocation.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_nvgpu_page_allocator_ops(struct unit_module *m,
struct gk20a *g, void *args);
/**
* Test specification for: test_page_allocator_sgt_ops
*
* Description: Test page alloc sgt ops
*
* Test Type: Feature
*
* Targets: nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_next,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_phys,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_ipa,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_ipa_to_pa,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_dma,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_length,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgl_gpu_addr,
* nvgpu_page_alloc.nvgpu_sgt.nvgpu_sgt_ops.sgt_free
*
* Input: test_nvgpu_page_allocator_init
*
* Steps:
* - Check allocated page attributes using sgt ops
* - Confirm that allocation details are equal to values set during allocation.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_page_allocator_sgt_ops(struct unit_module *m,
struct gk20a *g, void *args);
/**
* Test specification for: test_page_alloc_fixed
*
* Description: Allocate memory at fixed address
*
* Test Type: Feature, Error injection
*
* Targets: nvgpu_allocator.ops.alloc_fixed
*
* Input: test_nvgpu_page_allocator_init or test_page_allocator_init_slabs,
* args (fault_at_alloc_cache, fault_at_sgl_alloc, simple_alloc_128K,
* alloc_no_scatter_gather, failing_alloc_8K)
*
* Steps:
* - Allocate chunk of memory at fixed address as per test_parameters input.
* - Check that result is equal to test_parameters expected output.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_page_alloc_fixed(struct unit_module *m, struct gk20a *g, void *args);
/**
* Test specification for: test_page_alloc
*
* Description: Allocate memory using page allocator
*
* Test Type: Feature, Error injection
*
* Targets: nvgpu_allocator.ops.alloc
*
* Input: test_nvgpu_page_allocator_init or test_page_allocator_init_slabs,
* args (fault_at_alloc_cache, fault_at_nvgpu_alloc,
* first_simple_alloc_32K, fault_at_sgl_alloc, alloc_no_scatter_gather,
* alloc_contiguous, simple_alloc_512K, alloc_more_than_available,
* fault_at_page_cache, second_simple_alloc_32K, third_simple_alloc_32K,
* fourth_simple_alloc_32K, simple_alloc_8K, failing_alloc_16K)
*
* Steps:
* - Allocate chunk of memory at any address as per test_parameters input.
* - Check that result is equal to test_parameters expected output.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_page_alloc(struct unit_module *m, struct gk20a *g, void *args);
/**
* Test specification for: test_page_free
*
* Description: Free allocated memory
*
* Test Type: Feature
*
* Targets: nvgpu_allocator.ops.free_alloc
*
* Input: test_nvgpu_page_allocator_init or test_page_allocator_init_slabs,
* args (alloc_no_scatter_gather, first_simple_alloc_32K,
* simple_alloc_512K, fourth_simple_alloc_32K, second_simple_alloc_32K,
* first_simple_alloc_32K, third_simple_alloc_32K)
*
* Steps:
* - Free allocated memory at given address as per test_parameters input.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_page_free(struct unit_module *m, struct gk20a *g, void *args);
/**
* Test specification for: test_page_free_fixed
*
* Description: Free allocated page at given address
*
* Test Type: Feature
*
* Targets: nvgpu_allocator.ops.free_fixed
*
* Input: test_nvgpu_page_allocator_init, args (alloc_no_scatter_gather,
* simple_alloc_128K)
*
* Steps:
* - Free allocated memory at fixed address as per test_parameters input.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_page_free_fixed(struct unit_module *m, struct gk20a *g, void *args);
/**
* Test specification for: test_page_allocator_init_slabs
*
* Description: Initialize page allocator with slabs.
*
* Test Type: Feature, Error injection
*
* Targets: nvgpu_allocator_init, nvgpu_page_allocator_init,
* nvgpu_page_alloc_init_slabs, nvgpu_allocator.ops.fini
*
* Input: None
*
* Steps:
* - Initialize page allocator with following characteristics.
* - 64K memory base address.
* - 128K length of memory.
* - 64K block size.
* - Flags set to GPU_ALLOC_4K_VIDMEM_PAGES to enable slabs.
* - Check that page allocator initializations fails for scenarios such as
* odd value of block_size and fault injection for memory allocation.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_page_allocator_init_slabs(struct unit_module *m,
struct gk20a *g, void *args);
/**
* Test specification for: test_nvgpu_page_allocator_destroy
*
* Description: Destroy page allocator structure
*
* Test Type: Feature
*
* Targets: nvgpu_allocator.ops.fini
*
* Input: test_nvgpu_page_allocator_init or test_page_allocator_init_slabs
*
* Steps:
* - De-initialize page allocator structure.
*
* Output: Returns SUCCESS if the steps above were executed successfully. FAIL
* otherwise.
*/
int test_nvgpu_page_allocator_destroy(struct unit_module *m,
struct gk20a *g, void *args);
/**
* @}
*/
#endif /* UNIT_PAGE_ALLOCATOR_H */