[Intel-gfx] [igt-dev] [PATCH i-g-t 1/3] lib: Launch spinners from inside userptr
Mika Kuoppala
mika.kuoppala at linux.intel.com
Fri Oct 9 16:35:13 UTC 2020
Chris Wilson <chris at chris-wilson.co.uk> writes:
Needs a commit message like:
Add support for dummyload to be userptr.
Reviewed-by: Mika Kuoppala <mika.kuoppala at linux.intel.com>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
> lib/igt_dummyload.c | 87 ++++++++++++++++++++++++-------------
> lib/igt_dummyload.h | 13 ++++--
> tests/i915/gem_spin_batch.c | 23 ++++++----
> 3 files changed, 80 insertions(+), 43 deletions(-)
>
> diff --git a/lib/igt_dummyload.c b/lib/igt_dummyload.c
> index 26ea154ac..d58f73108 100644
> --- a/lib/igt_dummyload.c
> +++ b/lib/igt_dummyload.c
> @@ -68,6 +68,24 @@ static const int LOOP_START_OFFSET = 64;
> static IGT_LIST_HEAD(spin_list);
> static pthread_mutex_t list_lock = PTHREAD_MUTEX_INITIALIZER;
>
> +static uint32_t
> +handle_create(int fd, size_t sz, unsigned long flags, uint32_t **mem)
> +{
> + *mem = NULL;
> +
> + if (flags & IGT_SPIN_USERPTR) {
> + uint32_t handle;
> +
> + *mem = mmap(NULL, sz, PROT_WRITE, MAP_SHARED | MAP_ANON, -1, 0);
> + igt_assert(*mem != (uint32_t *)-1);
> + gem_userptr(fd, *mem, sz, 0, 0, &handle);
> +
> + return handle;
> + }
> +
> + return gem_create(fd, sz);
> +}
> +
> static int
> emit_recursive_batch(igt_spin_t *spin,
> int fd, const struct igt_spin_factory *opts)
> @@ -81,8 +99,8 @@ emit_recursive_batch(igt_spin_t *spin,
> unsigned int flags[GEM_MAX_ENGINES];
> unsigned int nengine;
> int fence_fd = -1;
> - uint32_t *cs, *batch;
> uint64_t addr;
> + uint32_t *cs;
> int i;
>
> /*
> @@ -126,13 +144,16 @@ emit_recursive_batch(igt_spin_t *spin,
> execbuf->flags = I915_EXEC_NO_RELOC;
> obj = memset(spin->obj, 0, sizeof(spin->obj));
>
> - obj[BATCH].handle = gem_create(fd, BATCH_SIZE);
> - batch = gem_mmap__device_coherent(fd, obj[BATCH].handle,
> - 0, BATCH_SIZE, PROT_WRITE);
> - gem_set_domain(fd, obj[BATCH].handle,
> - I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
> + obj[BATCH].handle =
> + handle_create(fd, BATCH_SIZE, opts->flags, &spin->batch);
> + if (!spin->batch) {
> + spin->batch = gem_mmap__device_coherent(fd, obj[BATCH].handle,
> + 0, BATCH_SIZE, PROT_WRITE);
> + gem_set_domain(fd, obj[BATCH].handle,
> + I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
> + }
> execbuf->buffer_count++;
> - cs = batch;
> + cs = spin->batch;
>
> obj[BATCH].offset = addr;
> addr += BATCH_SIZE;
> @@ -165,19 +186,22 @@ emit_recursive_batch(igt_spin_t *spin,
> igt_require(__igt_device_set_master(fd) == 0);
> }
>
> - spin->poll_handle = gem_create(fd, 4096);
> + spin->poll_handle =
> + handle_create(fd, 4096, opts->flags, &spin->poll);
> obj[SCRATCH].handle = spin->poll_handle;
>
> - if (__gem_set_caching(fd, spin->poll_handle,
> - I915_CACHING_CACHED) == 0)
> - spin->poll = gem_mmap__cpu(fd, spin->poll_handle,
> - 0, 4096,
> - PROT_READ | PROT_WRITE);
> - else
> - spin->poll = gem_mmap__device_coherent(fd,
> - spin->poll_handle,
> - 0, 4096,
> - PROT_READ | PROT_WRITE);
> + if (!spin->poll) {
> + if (__gem_set_caching(fd, spin->poll_handle,
> + I915_CACHING_CACHED) == 0)
> + spin->poll = gem_mmap__cpu(fd, spin->poll_handle,
> + 0, 4096,
> + PROT_READ | PROT_WRITE);
> + else
> + spin->poll = gem_mmap__device_coherent(fd,
> + spin->poll_handle,
> + 0, 4096,
> + PROT_READ | PROT_WRITE);
> + }
> addr += 4096; /* guard page */
> obj[SCRATCH].offset = addr;
> addr += 4096;
> @@ -210,8 +234,8 @@ emit_recursive_batch(igt_spin_t *spin,
>
> spin->handle = obj[BATCH].handle;
>
> - igt_assert_lt(cs - batch, LOOP_START_OFFSET / sizeof(*cs));
> - spin->condition = batch + LOOP_START_OFFSET / sizeof(*cs);
> + igt_assert_lt(cs - spin->batch, LOOP_START_OFFSET / sizeof(*cs));
> + spin->condition = spin->batch + LOOP_START_OFFSET / sizeof(*cs);
> cs = spin->condition;
>
> /* Allow ourselves to be preempted */
> @@ -255,15 +279,15 @@ emit_recursive_batch(igt_spin_t *spin,
> * (using 5 << 12).
> * For simplicity, we try to stick to a one-size fits all.
> */
> - spin->condition = batch + BATCH_SIZE / sizeof(*batch) - 2;
> + spin->condition = spin->batch + BATCH_SIZE / sizeof(*spin->batch) - 2;
> spin->condition[0] = 0xffffffff;
> spin->condition[1] = 0xffffffff;
>
> r->presumed_offset = obj[BATCH].offset;
> r->target_handle = obj[BATCH].handle;
> - r->offset = (cs + 2 - batch) * sizeof(*cs);
> + r->offset = (cs + 2 - spin->batch) * sizeof(*cs);
> r->read_domains = I915_GEM_DOMAIN_COMMAND;
> - r->delta = (spin->condition - batch) * sizeof(*cs);
> + r->delta = (spin->condition - spin->batch) * sizeof(*cs);
>
> *cs++ = MI_COND_BATCH_BUFFER_END | MI_DO_COMPARE | 2;
> *cs++ = MI_BATCH_BUFFER_END;
> @@ -275,7 +299,7 @@ emit_recursive_batch(igt_spin_t *spin,
> r = &relocs[obj[BATCH].relocation_count++];
> r->target_handle = obj[BATCH].handle;
> r->presumed_offset = obj[BATCH].offset;
> - r->offset = (cs + 1 - batch) * sizeof(*cs);
> + r->offset = (cs + 1 - spin->batch) * sizeof(*cs);
> r->read_domains = I915_GEM_DOMAIN_COMMAND;
> r->delta = LOOP_START_OFFSET;
> if (gen >= 8) {
> @@ -294,8 +318,8 @@ emit_recursive_batch(igt_spin_t *spin,
> }
> obj[BATCH].relocs_ptr = to_user_pointer(relocs);
>
> - execbuf->buffers_ptr = to_user_pointer(obj +
> - (2 - execbuf->buffer_count));
> + execbuf->buffers_ptr =
> + to_user_pointer(obj + (2 - execbuf->buffer_count));
> execbuf->rsvd1 = opts->ctx;
>
> if (opts->flags & IGT_SPIN_FENCE_OUT)
> @@ -329,7 +353,7 @@ emit_recursive_batch(igt_spin_t *spin,
> }
> }
>
> - igt_assert_lt(cs - batch, BATCH_SIZE / sizeof(*cs));
> + igt_assert_lt(cs - spin->batch, BATCH_SIZE / sizeof(*cs));
>
> /* Make it easier for callers to resubmit. */
> for (i = 0; i < ARRAY_SIZE(spin->obj); i++) {
> @@ -532,13 +556,14 @@ void igt_spin_free(int fd, igt_spin_t *spin)
> }
>
> igt_spin_end(spin);
> - gem_munmap((void *)((unsigned long)spin->condition & (~4095UL)),
> - BATCH_SIZE);
>
> - if (spin->poll) {
> + if (spin->poll)
> gem_munmap(spin->poll, 4096);
> + if (spin->batch)
> + gem_munmap(spin->batch, BATCH_SIZE);
> +
> + if (spin->poll_handle)
> gem_close(fd, spin->poll_handle);
> - }
>
> if (spin->handle)
> gem_close(fd, spin->handle);
> diff --git a/lib/igt_dummyload.h b/lib/igt_dummyload.h
> index aac0c83a9..6d3e65ce2 100644
> --- a/lib/igt_dummyload.h
> +++ b/lib/igt_dummyload.h
> @@ -33,12 +33,19 @@
> #include "i915_drm.h"
>
> typedef struct igt_spin {
> - unsigned int handle;
> struct igt_list_head link;
>
> + uint32_t handle;
> + uint32_t poll_handle;
> +
> + uint32_t *batch;
> +
> uint32_t *condition;
> uint32_t cmd_precondition;
>
> + uint32_t *poll;
> +#define SPIN_POLL_START_IDX 0
> +
> struct timespec last_signal;
> pthread_t timer_thread;
> int timerfd;
> @@ -47,9 +54,6 @@ typedef struct igt_spin {
> struct drm_i915_gem_exec_object2 obj[2];
> #define IGT_SPIN_BATCH 1
> struct drm_i915_gem_execbuffer2 execbuf;
> - uint32_t poll_handle;
> - uint32_t *poll;
> -#define SPIN_POLL_START_IDX 0
> } igt_spin_t;
>
> struct igt_spin_factory {
> @@ -66,6 +70,7 @@ struct igt_spin_factory {
> #define IGT_SPIN_FAST (1 << 3)
> #define IGT_SPIN_NO_PREEMPTION (1 << 4)
> #define IGT_SPIN_INVALID_CS (1 << 5)
> +#define IGT_SPIN_USERPTR (1 << 6)
>
> igt_spin_t *
> __igt_spin_factory(int fd, const struct igt_spin_factory *opts);
> diff --git a/tests/i915/gem_spin_batch.c b/tests/i915/gem_spin_batch.c
> index e7dd58ec2..19bc4638d 100644
> --- a/tests/i915/gem_spin_batch.c
> +++ b/tests/i915/gem_spin_batch.c
> @@ -33,7 +33,9 @@
> "'%s' != '%s' (%lld not within %d%% tolerance of %lld)\n",\
> #x, #ref, (long long)x, tolerance, (long long)ref)
>
> -static void spin(int fd, const struct intel_execution_engine2 *e2,
> +static void spin(int fd,
> + const struct intel_execution_engine2 *e2,
> + unsigned int flags,
> unsigned int timeout_sec)
> {
> const uint64_t timeout_100ms = 100000000LL;
> @@ -43,9 +45,10 @@ static void spin(int fd, const struct intel_execution_engine2 *e2,
> struct timespec itv = { };
> uint64_t elapsed;
>
> - spin = __igt_spin_new(fd, .engine = e2->flags);
> + spin = __igt_spin_new(fd, .engine = e2->flags, .flags = flags);
> while ((elapsed = igt_nsec_elapsed(&tv)) >> 30 < timeout_sec) {
> - igt_spin_t *next = __igt_spin_new(fd, .engine = e2->flags);
> + igt_spin_t *next =
> + __igt_spin_new(fd, .engine = e2->flags, .flags = flags);
>
> igt_spin_set_timeout(spin,
> timeout_100ms - igt_nsec_elapsed(&itv));
> @@ -120,14 +123,15 @@ static void spin_exit_handler(int sig)
> igt_terminate_spins();
> }
>
> -static void spin_on_all_engines(int fd, unsigned int timeout_sec)
> +static void
> +spin_on_all_engines(int fd, unsigned long flags, unsigned int timeout_sec)
> {
> const struct intel_execution_engine2 *e2;
>
> __for_each_physical_engine(fd, e2) {
> igt_fork(child, 1) {
> igt_install_exit_handler(spin_exit_handler);
> - spin(fd, e2, timeout_sec);
> + spin(fd, e2, flags, timeout_sec);
> }
> }
>
> @@ -186,7 +190,7 @@ igt_main
> e2 = &e2__;
>
> igt_subtest_f("legacy-%s", e->name)
> - spin(fd, e2, 3);
> + spin(fd, e2, 0, 3);
>
> igt_subtest_f("legacy-resubmit-%s", e->name)
> spin_resubmit(fd, e2, 0);
> @@ -202,7 +206,7 @@ igt_main
>
> __for_each_physical_engine(fd, e2) {
> igt_subtest_f("%s", e2->name)
> - spin(fd, e2, 3);
> + spin(fd, e2, 0, 3);
>
> igt_subtest_f("resubmit-%s", e2->name)
> spin_resubmit(fd, e2, 0);
> @@ -220,7 +224,10 @@ igt_main
> }
>
> igt_subtest("spin-each")
> - spin_on_all_engines(fd, 3);
> + spin_on_all_engines(fd, 0, 3);
> +
> + igt_subtest("user-each")
> + spin_on_all_engines(fd, IGT_SPIN_USERPTR, 3);
>
> igt_fixture {
> igt_stop_hang_detector();
> --
> 2.28.0
>
> _______________________________________________
> igt-dev mailing list
> igt-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/igt-dev
More information about the Intel-gfx
mailing list