[Intel-gfx] [PATCH i-g-t] tests/gem_softpin: New tests for softpin feature
Tvrtko Ursulin
tvrtko.ursulin at linux.intel.com
Thu Nov 26 08:09:13 PST 2015
Hi,
On 25/11/15 14:44, Vinay Belgaumkar wrote:
> These tests exercise the userptr ioctl to create shared buffers
> between CPU and GPU. They contain error and normal usage scenarios.
> They also contain a couple of stress tests which copy buffers between
> CPU and GPU. These tests rely on the softpin patch in order to pin buffers
> to a certain VA.
>
> Caveat: These tests were designed to run on 64-bit system. Future work
> includes adding logic to ensure these tests can run on 32-bit systems with
> PPGTT support. Some tests are currently disabled for 32-bit systems for that
> reason.
>
> v2: Added cc and signed-off-by fields
>
> v3: Fixed review comments, added helper functions. Removed userptr error
> scenarios covered by existing userptr tests. Modified stress test to have
> 100K buffers, it now runs for ~30 mins, checks every element has been written
> to correctly, and pins buffers at different VMAs.
>
> v4: Changed name to gem_softpin
>
> v5: More fixes. Removed the file based tests, will move them to userptr tests.
> Added a function that validates appropriate PPGTT support before running tests.
> Optimized stack space and memory footprint in stress test. Removed the eviction
> test, will add it back after verifying proper functionality.
>
> Cc: Michel Thierry <michel.thierry at intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar at intel.com>
> ---
> tests/.gitignore | 1 +
> tests/Makefile.sources | 1 +
> tests/gem_softpin.c | 972 +++++++++++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 974 insertions(+)
> create mode 100644 tests/gem_softpin.c
>
> diff --git a/tests/.gitignore b/tests/.gitignore
> index 80af9a7..424870b 100644
> --- a/tests/.gitignore
> +++ b/tests/.gitignore
> @@ -21,6 +21,7 @@ gem_bad_blit
> gem_bad_length
> gem_bad_reloc
> gem_basic
> +gem_softpin
> gem_caching
> gem_close_race
> gem_concurrent_all
> diff --git a/tests/Makefile.sources b/tests/Makefile.sources
> index 8fb2de8..2008d4a 100644
> --- a/tests/Makefile.sources
> +++ b/tests/Makefile.sources
> @@ -11,6 +11,7 @@ TESTS_progs_M = \
> drv_hangman \
> gem_bad_reloc \
> gem_basic \
> + gem_softpin \
> gem_caching \
> gem_close_race \
> gem_concurrent_blit \
> diff --git a/tests/gem_softpin.c b/tests/gem_softpin.c
> new file mode 100644
> index 0000000..95b664a
> --- /dev/null
> +++ b/tests/gem_softpin.c
> @@ -0,0 +1,972 @@
> +/*
> + * Copyright © 2015 Intel Corporation
> + *
> + * 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 (including the next
> + * paragraph) 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.
> + *
> + * Authors:
> + * Vinay Belgaumkar <vinay.belgaumkar at intel.com>
> + * Thomas Daniel <thomas.daniel at intel.com>
> + *
> + */
> +
> +#include <unistd.h>
> +#include <stdlib.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <string.h>
> +#include <fcntl.h>
> +#include <inttypes.h>
> +#include <errno.h>
> +#include <sys/stat.h>
> +#include <sys/ioctl.h>
> +#include <sys/time.h>
> +#include <malloc.h>
> +#include "drm.h"
> +#include "ioctl_wrappers.h"
> +#include "drmtest.h"
> +#include "intel_chipset.h"
> +#include "intel_io.h"
> +#include "i915_drm.h"
> +#include <assert.h>
> +#include <sys/wait.h>
> +#include <sys/ipc.h>
> +#include <sys/shm.h>
> +#include "igt_kms.h"
> +#include <inttypes.h>
> +#include <sys/types.h>
> +#include <sys/stat.h>
> +
> +#define BO_SIZE 4096
> +#define MULTIPAGE_BO_SIZE 2 * BO_SIZE
> +#define STORE_BATCH_BUFFER_SIZE 4
> +#define EXEC_OBJECT_PINNED (1<<4)
> +#define EXEC_OBJECT_SUPPORTS_48B_ADDRESS (1<<3)
> +#define SHARED_BUFFER_SIZE 4096
> +#define NUM_EXEC_OBJECTS 2
I would just hardcode 2 or 3 (or 1) in relevant places since some tests
use NUM_EXEC_OBJECTS + 1 anyway so it is not so useful to have a define
with a long name for this.
> +
> +typedef struct drm_i915_gem_userptr i915_gem_userptr;
> +
> +static uint32_t init_userptr(int fd, i915_gem_userptr*, void* ptr,
> + uint64_t size, bool read_only);
i915_gem_userptr *, void *ptr
> +static void *create_mem_buffer(uint64_t size);
> +static int gem_call_userptr_ioctl(int fd, i915_gem_userptr *userptr);
> +static void gem_basic_test(void);
> +static void gem_pin_invalid_vma_test(void);
> +static void gem_pin_overlap_test(void);
> +static void gem_pin_high_address_test(void);
> +
> +#define NO_PPGTT 0
> +#define ALIASING_PPGTT 1
> +#define FULL_32_BIT_PPGTT 2
> +#define FULL_48_BIT_PPGTT 3
> +/* uses_full_ppgtt
> + * Finds supported PPGTT details.
> + * @fd DRM fd
> + * @min can be
> + * 0 - No PPGTT
> + * 1 - Aliasing PPGTT
> + * 2 - Full PPGTT (32b)
> + * 3 - Full PPGTT (48b)
> + * RETURNS true/false if min support is present
> +*/
> +static bool uses_full_ppgtt(int fd, int min)
> +{
> + struct drm_i915_getparam gp;
> + int val = 0;
> +
> + memset(&gp, 0, sizeof(gp));
> + gp.param = 18; /* HAS_ALIASING_PPGTT */
> + gp.value = &val;
> +
> + if (drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp))
> + return 0;
> +
> + errno = 0;
> +
> + return val >= min;
> +}
> +
> +
> +/* gem_call_userptr_ioctl
> + * Helper to call ioctl - TODO: move to lib
> + * @fd - drm fd
> + * @userptr - pointer to initialised userptr
> + * RETURNS status of ioctl call
> +*/
> +static int gem_call_userptr_ioctl(int fd, i915_gem_userptr *userptr)
> +{
> + int ret;
> +
> + ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_USERPTR, userptr);
> +
> + if (ret)
> + ret = errno;
> +
> + return ret;
> +}
> +
> +/* init_userptr
> + * Helper that inits userptr an returns handle
> + * @fd - drm fd
> + * @userptr - pointer to empty userptr
> + * @ptr - buffer to be shared
> + * @size - size of buffer
> + * @ro - read only flag
> + * RETURNS handle to shared buffer
> +*/
> +static uint32_t init_userptr(int fd, i915_gem_userptr *userptr, void *ptr,
> + uint64_t size, bool ro)
I would lose the read-only argument if it is not interesting in this
test, just adds verbosity.
> +{
> + int ret;
> +
> + memset((void*)userptr, 0, sizeof(i915_gem_userptr));
> +
> + userptr->user_ptr = (uint64_t)ptr; /* Need the cast to overcome compiler warning */
> + userptr->user_size = size;
> + userptr->flags = I915_USERPTR_UNSYNCHRONIZED;
It would be better for test coverage to try the synchronized version if
available since that is what majority of users will use, given how
unsync is a root only feature.
> +
> + if (ro)
> + userptr->flags |= I915_USERPTR_READ_ONLY;
> +
> + ret = gem_call_userptr_ioctl(fd, userptr);
> + igt_assert_eq(ret, 0);
> +
> + return userptr->handle;
> +}
> +
> +/* create_mem_buffer
> + * Creates a 4K aligned CPU buffer
> + * @size - size of buffer
> + * RETURNS pointer to buffer of @size
> +*/
> +static void *create_mem_buffer(uint64_t size)
> +{
> + void *addr;
> + int ret;
> +
> + ret = posix_memalign(&addr, 4096, size);
> + igt_assert(ret == 0);
> +
> + return addr;
> +}
> +
> +/* setup_exec_obj
> + * populate exec object
> + * @exec - exec object
> + * @handle - handle to gem buffer
> + * @flags - any flags
> + * @offset - requested VMA
> +*/
> +static void setup_exec_obj(struct drm_i915_gem_exec_object2 *exec,
> + uint32_t handle, uint32_t flags,
> + uint64_t offset)
> +{
> + memset(exec, 0, sizeof(struct drm_i915_gem_exec_object2));
> + exec->handle = handle;
> + exec->flags = flags;
> + exec->offset = offset;
> +}
> +
> +/* gem_store_data_svm
> + * populate batch buffer with MI_STORE_DWORD_IMM command
> + * @fd: drm file descriptor
> + * @cmd_buf: batch buffer
> + * @vaddr: destination Virtual address
> + * @data: data to be store at destination
> + * @end: whether to end batch buffer or not
> +*/
> +static int gem_store_data_svm(int fd, uint32_t *cmd_buf, uint64_t vaddr,
> + uint32_t data, bool end)
> +{
> + int i = 0, len = 0;
Nitpick: Don't need to init len. On better look you don't even need that
variable? But still a nitpick..
> +
> + cmd_buf[i++] = MI_STORE_DWORD_IMM;
> + cmd_buf[i++] = vaddr & 0xFFFFFFFC;
> + cmd_buf[i++] = (vaddr >> 32) & 0xFFFF; /* bits 32:47 */
> +
> + cmd_buf[i++] = data;
> + if (end){
Missing space, "if (end) {", here and elsewhere.
> + cmd_buf[i++] = MI_BATCH_BUFFER_END;
> + cmd_buf[i++] = 0;
> + }
> +
> + len = i * sizeof(uint32_t);
> +
> + return len;
> +}
> +
> +/* gem_store_data
> + * populate batch buffer with MI_STORE_DWORD_IMM command
> + * This one fills up reloc buffer as well
> + * @fd: drm file descriptor
> + * @cmd_buf: batch buffer
> + * @data: data to be store at destination
> + * @reloc - relocation entry
> + * @end: whether to end batch buffer or not
> +*/
> +static int gem_store_data(int fd, uint32_t *cmd_buf,
> + uint32_t handle, uint32_t data,
> + struct drm_i915_gem_relocation_entry *reloc,
> + bool end)
More coding style (elsewhere in file as well) - please align parameters
to the first line.
> +{
> + int i = 0;
> +
> + cmd_buf[i++] = MI_STORE_DWORD_IMM;
> + cmd_buf[i++] = 0; /* lower 31 bits of 48 bit address - 0 reloc needed */
> + cmd_buf[i++] = 0; /* upper 15 bits of 48 bit address - 0 reloc needed */
> + reloc->offset = 1 * sizeof(uint32_t);
> + reloc->delta = 0;
> + reloc->target_handle = handle;
> + reloc->read_domains = I915_GEM_DOMAIN_RENDER;
> + reloc->write_domain = I915_GEM_DOMAIN_RENDER;
> + reloc->presumed_offset = 0;
> + cmd_buf[i++] = data;
> + if (end){
> + cmd_buf[i++] = MI_BATCH_BUFFER_END;
> + cmd_buf[i++] = 0;
> + }
> +
> + return (i * sizeof(uint32_t));
> +}
> +
> +/* setup_execbuffer
> + * helper for buffer execution
> + * @execbuf - pointer to execbuffer
> + * @exec_object - pointer to exec object2 struct
> + * @ring - ring to be used
> + * @buffer_count - how manu buffers to submit
> + * @batch_length - length of batch buffer
> +*/
> +static void setup_execbuffer(struct drm_i915_gem_execbuffer2 *execbuf,
> + struct drm_i915_gem_exec_object2 *exec_object,
> + int ring, int buffer_count, int batch_length)
> +{
> + execbuf->buffers_ptr = (uintptr_t)exec_object;
Bah uintptr_t.. I see that the code base is full of that inappropriate
thing so you are excused. :)
> + execbuf->buffer_count = buffer_count;
> + execbuf->batch_start_offset = 0;
> + execbuf->batch_len = batch_length;
> + execbuf->cliprects_ptr = 0;
> + execbuf->num_cliprects = 0;
> + execbuf->DR1 = 0;
> + execbuf->DR4 = 0;
> + execbuf->flags = ring;
> + i915_execbuffer2_set_context_id(*execbuf, 0);
> + execbuf->rsvd2 = 0;
> +}
> +
> +/* submit_and_sync
> + * Helper function for exec and sync functions
> + * @fd - drm fd
> + * @execbuf - pointer to execbuffer
> + * @batch_buf_handle - batch buffer handle
> +*/
> +static void submit_and_sync(int fd, struct drm_i915_gem_execbuffer2 *execbuf,
> + uint32_t batch_buf_handle)
> +{
> + gem_execbuf(fd, execbuf);
> + gem_sync(fd, batch_buf_handle);
> +}
> +
> +/* gem_userptr_sync
> + * helper for syncing to CPU domain - copy/paste from userblit
> + * @fd - drm fd
> + * @handle - buffer handle to sync
> +*/
> +static void gem_userptr_sync(int fd, uint32_t handle)
> +{
> + gem_set_domain(fd, handle, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
> +}
> +
> +
> +/* gem_basic_test
> + * This test will create a shared buffer, and create a command
> + * for GPU to write data in it
> + * CPU will read and make sure expected value is obtained
> + * @valid_shared_buffer - whether test with valid malloc'd buffer or not
This parameter is no more, including the test description below.
> + * if (valid_shared_buffer == true)
> + * Malloc a 4K buffer
> + * Share buffer with with GPU by using userptr ioctl
> + * Create batch buffer to write DATA to first dword of buffer
> + * Use 0x1000 address as destination address in batch buffer
> + * Set EXEC_OBJECT_PINNED flag in exec object
> + * Set 'offset' in exec object to 0x1000
> + * Submit execbuffer
> + * Verify value of first DWORD in shared buffer matches DATA
> +
> + * if (valid_shared_buffer == false)
> + * Declare null buffer
> + * Call Userptr ioctl with null buffer
> + * Run Basic Test
> + * Test should fail at submit execbuffer
> +*/
> +static void gem_basic_test(void)
> +{
> + i915_gem_userptr userptr;
> + int fd;
> + uint32_t* shared_buffer = NULL;
Coding style, plus don't need to initialise.
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS + 1];
> + uint32_t batch_buffer[STORE_BATCH_BUFFER_SIZE + 2];
> + uint32_t batch_buf_handle, shared_buf_handle, unshared_buf_handle;
> + struct drm_i915_gem_relocation_entry reloc[4];
> + int ring, len;
> + uint32_t value;
> + const uint32_t data = 0x12345678;
> + uint64_t pinning_offset = 0x1000;
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, ALIASING_PPGTT));
> + batch_buf_handle = gem_create(fd, BO_SIZE);
> +
> + /* create cpu buffer */
> + shared_buffer = create_mem_buffer(BO_SIZE);
> + unshared_buf_handle = gem_create(fd, BO_SIZE);
> +
> + /* share with GPU */
> + shared_buf_handle = init_userptr(fd, &userptr, shared_buffer,
> + BO_SIZE, false);
Weird alignment. Please check throughout.
> +
> + /* create command buffer with write command */
> + len = gem_store_data_svm(fd, batch_buffer, pinning_offset, data, false);
> + len += gem_store_data(fd, batch_buffer + (len/4), unshared_buf_handle, data,
> + reloc, true);
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> +
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[0], shared_buf_handle,
> + EXEC_OBJECT_PINNED, pinning_offset);
> + setup_exec_obj(&exec_object2[1], unshared_buf_handle, 0, 0);
> + setup_exec_obj(&exec_object2[2], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS + 1, len);
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + gem_userptr_sync(fd, shared_buf_handle);
> +
> + /* Check if driver pinned the buffer as requested */
> + igt_fail_on_f(exec_object2[0].offset != pinning_offset,
> + "\nFailed to pin at requested offset");
> + /* check on CPU to see if value changes */
> + igt_fail_on_f(shared_buffer[0] != data,
> + "\nCPU read does not match GPU write,\
> + expected: 0x%x, got: 0x%x\n",
> + data, shared_buffer[0]);
> + gem_read(fd, unshared_buf_handle, 0, (void*)&value, 4);
> + igt_assert(value == data);
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> + free(shared_buffer);
> +}
I though we agreed to have separate basic test case for a normal bo, and
a separate for userptr?
(And to exercise EXEC_OBJECT_PINNED in both cases.)
> +/* gem_multiple_process_test
> + * Run basic test simultaneously with multiple processes
> + * This will test pinning same VA separately in each process
> +
> + * fork();
> + * Execute basic test in parent/child processes
> +*/
> +#define MAX_NUM_PROCESSES 10
> +
> +static void gem_multiple_process_test(void)
> +{
> + igt_fork(child, MAX_NUM_PROCESSES) {
> + gem_basic_test();
> + }
> + igt_waitchildren();
> +}
> +
> +
> +/* gem_repin_test
> + * This test tries to repin a buffer at a previously pinned vma
> + * from a different execbuf.
> + * Malloc a 4K buffer
> + * Share buffer with with GPU by using userptr ioctl
> + * Create batch buffer to write DATA to first dword of buffer
> + * Use 0x1000 address as destination address in batch buffer
> + * Set EXEC_OBJECT_PINNED flag in exec object
> + * Set 'offset' in exec object to 0x1000 VMA
> + * Submit execbuffer
> + * Verify value of first DWORD in shared buffer matches DATA
> +
> + * Create second shared buffer
> + * Follow all steps above
> + * Execpt, for offset, use VMA of first buffer above
> + * Submit execbuffer
> + * Verify value of first DWORD in second shared buffer matches DATA
> +*/
> +static void gem_repin_test(void)
> +{
> + i915_gem_userptr userptr;
> + i915_gem_userptr userptr1;
> + int fd;
> + uint32_t* shared_buffer = NULL;
> + uint32_t* shared_buffer1 = NULL;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS];
> + uint32_t batch_buffer[STORE_BATCH_BUFFER_SIZE + 2];
> + uint32_t batch_buf_handle, shared_buf_handle, shared_buf_handle1;
> + int ring, len;
> + const uint32_t data = 0x12345678;
> + uint64_t pinning_offset = 0x1000;
> +
> + /* Create gem object */
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, ALIASING_PPGTT));
> +
> + batch_buf_handle = gem_create(fd, BO_SIZE);
> +
> + /* create cpu buffer, set first elements to 0x0 */
> + shared_buffer = create_mem_buffer(BO_SIZE);
> + shared_buffer1 = create_mem_buffer(BO_SIZE);
> + shared_buffer[0] = 0x0;
> + shared_buffer1[0] = 0x0;
> +
> + /* share with GPU and get handles */
> + shared_buf_handle = init_userptr(fd, &userptr, shared_buffer,
> + BO_SIZE, false);
> + shared_buf_handle1 = init_userptr(fd, &userptr1, shared_buffer1,
> + BO_SIZE, false);
> +
> + /* create command buffer with write command */
> + len = gem_store_data_svm(fd, batch_buffer, pinning_offset, data, true);
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> +
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[0], shared_buf_handle,
> + EXEC_OBJECT_PINNED, pinning_offset);
> + setup_exec_obj(&exec_object2[1], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS, len);
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + gem_userptr_sync(fd, shared_buf_handle);
> +
> + igt_assert(*shared_buffer == data);
I suggest always checking the offsets, not only data, whenever
EXEC_OBJECT_PINNED is asked for.
> +
> + /* Second buffer */
> + /* create command buffer with write command */
> + pinning_offset = exec_object2[0].offset;
> + len = gem_store_data_svm(fd, batch_buffer, pinning_offset, data, true);
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> +
> + /* submit command buffer */
> + /* Pin at shared_buffer, not shared_buffer1 */
> + /* We are requesting address where another buffer was pinned previously */
> + /* It does not exist there anymore, since execbuf was already executed */
It could still be there and will then need to be removed to make place
for this one.
> + setup_exec_obj(&exec_object2[0], shared_buf_handle1,
> + EXEC_OBJECT_PINNED, pinning_offset);
> + setup_exec_obj(&exec_object2[1], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS, len);
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + gem_userptr_sync(fd, shared_buf_handle1);
> +
> + igt_assert(*shared_buffer1 == data);
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> +
> + free(shared_buffer);
> + free(shared_buffer1);
> +}
> +
> +
> +/* gem_repin_overlap_test
> + * This test will attempt to pin two buffers at the same VMA as part of the same
> + execbuffer object
> +
> + * Malloc a 4K buffer
> + * Share buffer with with GPU by using userptr ioctl
> + * Create second shared buffer
> + * Create batch buffer to write DATA to first dword of each buffer
> + * Use same virtual address as destination addresses in batch buffer
> + * Set EXEC_OBJECT_PINNED flag in both exec objects
> + * Set 'offset' in both exec objects to same VMA
> + * Submit execbuffer
> + * Command should return EINVAL, since we are trying to pin to same VMA
> +*/
> +static void gem_pin_overlap_test(void)
> +{
> + i915_gem_userptr userptr;
> + i915_gem_userptr userptr1;
> + int fd, ret;
> + uint32_t* shared_buffer = NULL;
> + uint32_t* shared_buffer1 = NULL;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS + 1];
> + uint32_t batch_buffer[BO_SIZE];
> + uint32_t batch_buf_handle, shared_buf_handle, shared_buf_handle1;
> + int ring, len;
> + const uint32_t data = 0x12345678;
> + uint64_t pinning_offset = 0x1000;
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, ALIASING_PPGTT));
> +
> + batch_buf_handle = gem_create(fd, BO_SIZE);
> +
> + shared_buffer = create_mem_buffer(BO_SIZE);
> + shared_buffer1 = create_mem_buffer(BO_SIZE * 2);
> +
> + /* share with GPU */
> + shared_buf_handle = init_userptr(fd, &userptr, shared_buffer,
> + BO_SIZE, false);
> + shared_buf_handle1 = init_userptr(fd, &userptr1, shared_buffer1,
> + BO_SIZE * 2, false);
> +
> + len = gem_store_data_svm(fd, batch_buffer,
> + pinning_offset, data, false);
> + len += gem_store_data_svm(fd, (batch_buffer + len/4),
> + pinning_offset, data, true);
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
You can remove store dwords since you don't expect this test to ever
succeed. I think I suggested this in the previous round? Would make the
test shorter and easier to read.
Also, one thing worth adding would be different overlaps. Objects before
and after and overlap by page and vice-versa and similar.
This is really the core of the softpin functionality so it is worth
scrutinising a bit.
And then possibly the same but with something already at the VMA in
question, so it needs to move one and see that it cannot fit the other.
And then perhaps add an object which is immovable, like some long
running workload on another ring and try to pin at that address from
different ring while it is running to verify it fails.
Can add these as follow up tests cases.
> +
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[0], shared_buf_handle,
> + EXEC_OBJECT_PINNED, pinning_offset);
> + setup_exec_obj(&exec_object2[1], shared_buf_handle1,
> + EXEC_OBJECT_PINNED, pinning_offset);
> + setup_exec_obj(&exec_object2[2], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS+1, len);
> +
> + ret = drmIoctl(fd,
> + DRM_IOCTL_I915_GEM_EXECBUFFER2,
> + &execbuf);
> +
> + /* expect to fail */
> + igt_assert_neq(ret, 0);
> + igt_assert(errno == 22);
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> +
> + free(shared_buffer);
> + free(shared_buffer1);
> +}
> +
> +/* gem_softpin_stress_test
> + * Stress test which creates 10K buffers and shares with GPU
> + * Create 100K uint32 buffers of size 4K each
> + * Share with GPU using userptr ioctl
> + * Create batch buffer to write DATA in first element of each buffer
> + * Pin each buffer to varying addresses starting from 0x800000000000 going below
> + * Execute Batch Buffer on Blit ring STRESS_NUM_LOOPS times
> + * Validate every buffer has DATA in first element
> + * Rinse and Repeat on Render ring
> +*/
> +#define STRESS_NUM_BUFFERS 100000
> +#define STRESS_NUM_LOOPS 100
> +#define STRESS_STORE_COMMANDS 4 * STRESS_NUM_BUFFERS
> +
> +static void gem_softpin_stress_test(void)
> +{
> + i915_gem_userptr userptr;
> + int fd;
> + uint32_t **shared_buffer;
> + uint32_t *shared_handle;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 *exec_object2;
> + uint32_t *batch_buffer;
> + uint32_t batch_buf_handle;
> + int ring, len;
> + int buf, loop;
> + uint64_t pinning_offset = 0x800000000000;
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, FULL_48_BIT_PPGTT));
> +
> + /* Allocate blobs for all data structures */
> + shared_handle = calloc(STRESS_NUM_BUFFERS, sizeof(uint32_t));
> + shared_buffer = calloc(STRESS_NUM_BUFFERS, sizeof(uint32_t*));
> + exec_object2 = calloc(STRESS_NUM_BUFFERS + 1,
> + sizeof(struct drm_i915_gem_exec_object2));
> + /* 4 dwords per buffer + 2 for the end of batchbuffer */
> + batch_buffer = calloc(STRESS_STORE_COMMANDS + 2, sizeof(uint32_t));
> + batch_buf_handle = gem_create(fd, (STRESS_STORE_COMMANDS + 2)*4);
> +
> + /* create command buffer with write commands */
> + len = 0;
> + for(buf = 0; buf < STRESS_NUM_BUFFERS; buf++) {
> + shared_buffer[buf] = create_mem_buffer(BO_SIZE);
> + *shared_buffer[buf] = 0xFFFFFFFF;
> +
> + /* share with GPU */
> + shared_handle[buf] = init_userptr(fd, &userptr, shared_buffer[buf],
> + BO_SIZE, false);
> +
> + setup_exec_obj(&exec_object2[buf], shared_handle[buf],
> + EXEC_OBJECT_PINNED, pinning_offset);
> + len += gem_store_data_svm(fd, batch_buffer + (len/4), pinning_offset,
> + buf , (buf == STRESS_NUM_BUFFERS-1) ? true:false);
> +
> + /* decremental 4K aligned address */
> + pinning_offset = ALIGN(pinning_offset - 0x200000, 4096);
Oh I wasn't clear, my comment was about 0x200000, why hardcode that and
not just use
pinning_offset -= ALIGN(BO_SIZE, 4096)
?
> + }
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> +
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[STRESS_NUM_BUFFERS], batch_buf_handle, 0, 0);
> +
> + /* We want to run this on BLT ring if possible */
> + if (HAS_BLT_RING(intel_get_drm_devid(fd))){
> + ring = I915_EXEC_BLT;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring,
> + STRESS_NUM_BUFFERS + 1, len);
setup_execbuffer won't be called if there is not blt ring for the render
ring below.
> + for (loop = 0; loop < STRESS_NUM_LOOPS; loop++){
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + for(buf = 0; buf < STRESS_NUM_BUFFERS; buf++){
> + gem_userptr_sync(fd, shared_handle[buf]);
> + igt_fail_on_f(*shared_buffer[buf] != buf, \
> + "Mismatch in buffer %d, iteration %d: 0x%08X\n", \
> + buf, loop, *shared_buffer[buf]);
> + }
> + }
> + }
> +
> + /* Now Render Ring */
> + ring = I915_EXEC_RENDER;
> + execbuf.flags = ring;
> + for (loop = 0; loop < STRESS_NUM_LOOPS; loop++){
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + for(buf = 0; buf < STRESS_NUM_BUFFERS; buf++){
> + gem_userptr_sync(fd, shared_handle[buf]);
> + igt_fail_on_f(*shared_buffer[buf] != buf, \
> + "Mismatch in buffer %d, \
> + iteration %d: 0x%08X\n", buf, loop, *shared_buffer[buf]);
> + }
> + }
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> +
> + for(loop = 0; loop < STRESS_NUM_BUFFERS; loop++){
> + free(shared_buffer[loop]);
> + }
> +
> + free(shared_handle);
> + free(shared_buffer);
> + free(exec_object2);
> + free(batch_buffer);
> +}
> +
> +/* gem_write_multipage_buffer
> + * Create a buffer spanning multiple pages, and share with GPU.
> + * Write to every element of the buffer
> + * and verify correct contents.
> +
> + * Create 8K buffer
> + * Share with GPU using userptr ioctl
> + * Create batch buffer to write DATA in all elements of buffer
> + * Execute Batch Buffer
> + * Validate every element has DATA
> +*/
> +
> +#define DWORD_SIZE sizeof(uint32_t)
> +#define BB_SIZE ((MULTIPAGE_BO_SIZE / DWORD_SIZE) * STORE_BATCH_BUFFER_SIZE) + 2
> +static void gem_write_multipage_buffer_test(void)
> +{
> + i915_gem_userptr userptr;
> + int fd;
> + uint32_t* shared_buffer;
> + uint32_t shared_handle;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS];
> + uint32_t batch_buffer[BB_SIZE];
> + uint32_t batch_buf_handle;
> + int ring, len, j;
> + uint64_t pinning_offset=0x1000;
> + uint64_t vaddr;
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, ALIASING_PPGTT));
> +
> + batch_buf_handle = gem_create(fd, sizeof(batch_buffer));
> + shared_buffer = create_mem_buffer(MULTIPAGE_BO_SIZE);
> +
> + len = 0;
> + memset(batch_buffer, 0, sizeof(batch_buffer));
> + memset(shared_buffer, 0, MULTIPAGE_BO_SIZE);
> +
> + /* share with GPU */
> + shared_handle = init_userptr(fd, &userptr, shared_buffer,
> + MULTIPAGE_BO_SIZE, false);
> + setup_exec_obj(&exec_object2[0], shared_handle,
> + EXEC_OBJECT_PINNED, pinning_offset);
> +
> + /* create command buffer with write commands */
> + vaddr = pinning_offset;
> + for(j=0; j< (MULTIPAGE_BO_SIZE/sizeof(uint32_t)); j++){
> + /* BO_SIZE because it is 16K 4 byte entries */
Stale comment?
> + len += gem_store_data_svm(fd, batch_buffer + (len/4), vaddr,
> + j, (j == (((MULTIPAGE_BO_SIZE/sizeof(uint32_t))-1))) ?\
> + true:false);
> + vaddr += sizeof(shared_buffer[0]); /* 4 bytes */
> + }
> +
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> +
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[1], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS, len);
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + gem_userptr_sync(fd, shared_handle);
> +
> + for(j = 0; j < (MULTIPAGE_BO_SIZE/sizeof(uint32_t)); j++){
> + igt_fail_on_f(shared_buffer[j] != j,
> + "Mismatch in index %d: 0x%08X\n", j, shared_buffer[j]);
> + }
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> +
> + free(shared_buffer);
> +}
> +
> +/* gem_pin_invalid_vma_test
> + * This test will request to pin a shared buffer to an invalid
> + * VMA > 48-bit address
> + * Create shared buffer of size 4K
> + * Try and Pin object to address 0x9000000000000
> +*/
> +static void gem_pin_invalid_vma_test(void)
> +{
> + i915_gem_userptr userptr;
> + int fd, ret;
> + uint32_t* shared_buffer = NULL;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS - 1];
> + uint32_t shared_buf_handle;
> + int ring;
> + uint64_t invalid_address = 0x9000000000000; /* 52 bit address */
Would this fail with a >32-bit address as well? I thought I spotted
somewhere else that you need to pass in a special flag if you want to
use 48-bit offsets? Or I am confusing something?
Hm, yes, EXEC_OBJECT_SUPPORTS_48B_ADDRESS ?
So maybe two test cases, >32-bit offset without this flag should fail,
and >48-bit offset with it?
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, FULL_48_BIT_PPGTT));
> +
> + shared_buffer = create_mem_buffer(BO_SIZE);
> + *shared_buffer = 0xFFFFFFFF;
> +
> + /* share with GPU */
> + shared_buf_handle = init_userptr(fd, &userptr, shared_buffer, BO_SIZE, false);
> + setup_exec_obj(&exec_object2[0], shared_buf_handle,
> + EXEC_OBJECT_PINNED, invalid_address);
> +
> + ring = I915_EXEC_RENDER;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS - 1, 0);
> +
> + /* Expect execbuf to fail */
> + ret = drmIoctl(fd,
> + DRM_IOCTL_I915_GEM_EXECBUFFER2,
> + &execbuf);
> +
> + igt_assert_neq(ret, 0);
Do you also want to check the exact errno? Sounds like a good idea to
me, even though execbuf returns EINVAL for a ton of reasons..
> +
> + close(fd);
> + free(shared_buffer);
> +}
> +
> +
> +/* gem_pin_high_address_test
> + * This test will create a shared buffer, and create a command
> + * for GPU to write data in it. It will attempt to pin the buffer at address > 32 bits.
> + * CPU will read and make sure expected value is obtained
> +
> + * Malloc a 4K buffer
> + * Share buffer with with GPU by using userptr ioctl
> + * Create batch buffer to write DATA to first dword of buffer
> + * Use virtual address of buffer as 0x1100000000 (> 32 bit)
> + * Set EXEC_OBJECT_PINNED flag in exec object
> + * Set 'offset' in exec object to shared buffer VMA
> + * Submit execbuffer
> + * Verify value of first DWORD in shared buffer matches DATA
> +*/
> +
> +static void gem_pin_high_address_test(void)
> +{
> + i915_gem_userptr userptr;
> + int fd;
> + uint32_t* shared_buffer = NULL;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS];
> + uint32_t batch_buffer[BO_SIZE];
> + uint32_t batch_buf_handle, shared_buf_handle;
> + int ring, len;
> + const uint32_t data = 0x12345678;
> + uint64_t high_address = 0x1111FFFF000; /* 44 bit address */
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, FULL_48_BIT_PPGTT));
> +
> + batch_buf_handle = gem_create(fd, BO_SIZE);
> +
> + /* create cpu buffer, set to all 0xF's */
> + shared_buffer = create_mem_buffer(BO_SIZE);
> + *shared_buffer = 0xFFFFFFFF;
> +
> + /* share with GPU */
> + shared_buf_handle = init_userptr(fd, &userptr, shared_buffer, BO_SIZE, false);
> +
> + /* create command buffer with write command */
> + len = gem_store_data_svm(fd, batch_buffer, high_address, data, true);
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> +
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[0], shared_buf_handle,
> + EXEC_OBJECT_PINNED | EXEC_OBJECT_SUPPORTS_48B_ADDRESS, high_address);
> + setup_exec_obj(&exec_object2[1], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> +
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS, len);
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + gem_userptr_sync(fd, shared_buf_handle);
> +
> + /* check on CPU to see if value changes */
> + igt_fail_on_f(shared_buffer[0] != data,
> + "\nCPU read does not match GPU write, \
> + expected: 0x%x, got: 0x%x\n", data, shared_buffer[0]);
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> + free(shared_buffer);
> +}
> +
> +/* gem_pin_near_48Bit_test
> + * This test will create a shared buffer,
> + * and create a command for GPU to write data in it. It will attempt
> + * to pin the buffer at address > 47 bits <= 48-bit.
> + * CPU will read and make sure expected value is obtained
> +
> + * Malloc a 4K buffer
> + * Share buffer with with GPU by using userptr ioctl
> + * Create batch buffer to write DATA to first dword of buffer
> + * Use virtual address of buffer as range between 47-bit and 48-bit
> + * Set EXEC_OBJECT_PINNED flag in exec object
> + * Set 'offset' in exec object to shared buffer VMA
> + * Submit execbuffer
> + * Verify value of first DWORD in shared buffer matches DATA
> +*/
> +#define BEGIN_HIGH_ADDRESS 0x7FFFFFFFF000
> +#define END_HIGH_ADDRESS 0xFFFFFFFFC000
> +#define ADDRESS_INCREMENT 0x2000000000
> +static void gem_pin_near_48Bit_test(void)
> +{
> + i915_gem_userptr userptr;
> + int fd;
> + uint32_t* shared_buffer = NULL;
> + struct drm_i915_gem_execbuffer2 execbuf;
> + struct drm_i915_gem_exec_object2 exec_object2[NUM_EXEC_OBJECTS];
> + uint32_t batch_buffer[BO_SIZE];
> + uint32_t batch_buf_handle, shared_buf_handle;
> + int ring, len;
> + const uint32_t data = 0x12345678;
> + uint64_t high_address;
> +
> + fd = drm_open_driver(DRIVER_INTEL);
> + igt_require(uses_full_ppgtt(fd, FULL_48_BIT_PPGTT));
> +
> + batch_buf_handle = gem_create(fd, BO_SIZE);
> +
> + /* create cpu buffer, set to all 0xF's */
> + shared_buffer = create_mem_buffer(BO_SIZE);
> + *shared_buffer = 0xFFFFFFFF;
> +
> + /* share with GPU */
> + shared_buf_handle = init_userptr(fd, &userptr, shared_buffer, BO_SIZE, false);
> +
> + for (high_address = BEGIN_HIGH_ADDRESS; high_address <= END_HIGH_ADDRESS;
> + high_address+=ADDRESS_INCREMENT){
> + /* create command buffer with write command */
> + len = gem_store_data_svm(fd, batch_buffer, high_address,
> + data, true);
> + gem_write(fd, batch_buf_handle, 0, batch_buffer, len);
> + /* submit command buffer */
> + setup_exec_obj(&exec_object2[0], shared_buf_handle,
> + EXEC_OBJECT_PINNED | EXEC_OBJECT_SUPPORTS_48B_ADDRESS,
> + high_address);
> + setup_exec_obj(&exec_object2[1], batch_buf_handle, 0, 0);
> +
> + ring = I915_EXEC_RENDER;
> + setup_execbuffer(&execbuf, exec_object2, ring, NUM_EXEC_OBJECTS, len);
> + submit_and_sync(fd, &execbuf, batch_buf_handle);
> + gem_userptr_sync(fd, shared_buf_handle);
> +
> + /* check on CPU to see if value changes */
> + igt_fail_on_f(shared_buffer[0] != data,
> + "\nCPU read does not match GPU write, expected: 0x%x, \
> + got: 0x%x\n, 0x%"PRIx64"", data, shared_buffer[0], high_address);
> + }
> +
> + gem_close(fd, batch_buf_handle);
> + close(fd);
> + free(shared_buffer);
> +}
> +
> +
> +int main(int argc, char* argv[])
> +{
> + igt_subtest_init(argc, argv);
> + igt_skip_on_simulation();
> +
> + /* All tests need PPGTT support */
> + igt_subtest("gem_basic"){
> + gem_basic_test();
> + }
> + igt_subtest("gem_multiple_process"){
> + gem_multiple_process_test();
> + }
> + igt_subtest("gem_repin"){
> + gem_repin_test();
> + }
> + igt_subtest("gem_pin_overlap"){
> + gem_pin_overlap_test();
> + }
> + igt_subtest("gem_write_multipage_buffer"){
> + gem_write_multipage_buffer_test();
> + }
> +
> + /* Following tests need 48 Bit PPGTT support */
> + igt_subtest("gem_pin_invalid_vma"){
> + gem_pin_invalid_vma_test();
> + }
> + igt_subtest("gem_softpin_stress"){
> + gem_softpin_stress_test();
> + }
> + igt_subtest("gem_pin_high_address"){
> + gem_pin_high_address_test();
> + }
> + igt_subtest("gem_pin_near_48Bit"){
> + gem_pin_near_48Bit_test();
> + }
> +
> + igt_exit();
> +}
>
Regards,
Tvrtko
More information about the Intel-gfx
mailing list