* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
-*/
+ */
+
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <pthread.h>
#include <sched.h>
#include <sys/ioctl.h>
+#ifdef HAVE_ALLOCA_H
+# include <alloca.h>
+#endif
#include "xf86drm.h"
#include "amdgpu_drm.h"
#include "amdgpu_internal.h"
+static int amdgpu_cs_unreference_sem(amdgpu_semaphore_handle sem);
+static int amdgpu_cs_reset_sem(amdgpu_semaphore_handle sem);
+
/**
* Create command submission context
*
- * \param dev - \c [in] amdgpu device handle
- * \param context - \c [out] amdgpu context handle
+ * \param dev - \c [in] Device handle. See #amdgpu_device_initialize()
+ * \param priority - \c [in] Context creation flags. See AMDGPU_CTX_PRIORITY_*
+ * \param context - \c [out] GPU Context handle
*
* \return 0 on success otherwise POSIX Error code
*/
-int amdgpu_cs_ctx_create(amdgpu_device_handle dev,
- amdgpu_context_handle *context)
+drm_public int amdgpu_cs_ctx_create2(amdgpu_device_handle dev,
+ uint32_t priority,
+ amdgpu_context_handle *context)
{
- struct amdgpu_bo_alloc_request alloc_buffer = {};
- struct amdgpu_bo_alloc_result info = {};
struct amdgpu_context *gpu_context;
union drm_amdgpu_ctx args;
+ int i, j, k;
int r;
- if (NULL == dev)
- return -EINVAL;
- if (NULL == context)
+ if (!dev || !context)
return -EINVAL;
gpu_context = calloc(1, sizeof(struct amdgpu_context));
- if (NULL == gpu_context)
+ if (!gpu_context)
return -ENOMEM;
gpu_context->dev = dev;
r = pthread_mutex_init(&gpu_context->sequence_mutex, NULL);
if (r)
- goto error_mutex;
-
- /* Create the fence BO */
- alloc_buffer.alloc_size = 4 * 1024;
- alloc_buffer.phys_alignment = 4 * 1024;
- alloc_buffer.preferred_heap = AMDGPU_GEM_DOMAIN_GTT;
-
- r = amdgpu_bo_alloc(dev, &alloc_buffer, &info);
- if (r)
- goto error_fence_alloc;
- gpu_context->fence_bo = info.buf_handle;
-
- r = amdgpu_bo_cpu_map(gpu_context->fence_bo, &gpu_context->fence_cpu);
- if (r)
- goto error_fence_map;
+ goto error;
/* Create the context */
memset(&args, 0, sizeof(args));
args.in.op = AMDGPU_CTX_OP_ALLOC_CTX;
+ args.in.priority = priority;
+
r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_CTX, &args, sizeof(args));
if (r)
- goto error_kernel;
+ goto error;
gpu_context->id = args.out.alloc.ctx_id;
+ for (i = 0; i < AMDGPU_HW_IP_NUM; i++)
+ for (j = 0; j < AMDGPU_HW_IP_INSTANCE_MAX_COUNT; j++)
+ for (k = 0; k < AMDGPU_CS_MAX_RINGS; k++)
+ list_inithead(&gpu_context->sem_list[i][j][k]);
*context = (amdgpu_context_handle)gpu_context;
return 0;
-error_kernel:
- amdgpu_bo_cpu_unmap(gpu_context->fence_bo);
-
-error_fence_map:
- amdgpu_bo_free(gpu_context->fence_bo);
-
-error_fence_alloc:
+error:
pthread_mutex_destroy(&gpu_context->sequence_mutex);
-
-error_mutex:
free(gpu_context);
return r;
}
+drm_public int amdgpu_cs_ctx_create(amdgpu_device_handle dev,
+ amdgpu_context_handle *context)
+{
+ return amdgpu_cs_ctx_create2(dev, AMDGPU_CTX_PRIORITY_NORMAL, context);
+}
+
/**
* Release command submission context
*
*
* \return 0 on success otherwise POSIX Error code
*/
-int amdgpu_cs_ctx_free(amdgpu_context_handle context)
+drm_public int amdgpu_cs_ctx_free(amdgpu_context_handle context)
{
union drm_amdgpu_ctx args;
+ int i, j, k;
int r;
- if (NULL == context)
+ if (!context)
return -EINVAL;
- r = amdgpu_bo_cpu_unmap(context->fence_bo);
- if (r)
- return r;
-
- r = amdgpu_bo_free(context->fence_bo);
- if (r)
- return r;
-
pthread_mutex_destroy(&context->sequence_mutex);
/* now deal with kernel side */
args.in.ctx_id = context->id;
r = drmCommandWriteRead(context->dev->fd, DRM_AMDGPU_CTX,
&args, sizeof(args));
-
+ for (i = 0; i < AMDGPU_HW_IP_NUM; i++) {
+ for (j = 0; j < AMDGPU_HW_IP_INSTANCE_MAX_COUNT; j++) {
+ for (k = 0; k < AMDGPU_CS_MAX_RINGS; k++) {
+ amdgpu_semaphore_handle sem;
+ LIST_FOR_EACH_ENTRY(sem, &context->sem_list[i][j][k], list) {
+ list_del(&sem->list);
+ amdgpu_cs_reset_sem(sem);
+ amdgpu_cs_unreference_sem(sem);
+ }
+ }
+ }
+ }
free(context);
return r;
}
-int amdgpu_cs_query_reset_state(amdgpu_context_handle context,
- uint32_t *state, uint32_t *hangs)
+drm_public int amdgpu_cs_ctx_override_priority(amdgpu_device_handle dev,
+ amdgpu_context_handle context,
+ int master_fd,
+ unsigned priority)
+{
+ int r;
+
+ if (!dev || !context || master_fd < 0)
+ return -EINVAL;
+
+ union drm_amdgpu_sched args;
+ memset(&args, 0, sizeof(args));
+
+ args.in.op = AMDGPU_SCHED_OP_CONTEXT_PRIORITY_OVERRIDE;
+ args.in.fd = dev->fd;
+ args.in.priority = priority;
+ args.in.ctx_id = context->id;
+
+ r = drmCommandWrite(master_fd, DRM_AMDGPU_SCHED, &args, sizeof(args));
+ if (r)
+ return r;
+
+ return 0;
+}
+
+drm_public int amdgpu_cs_query_reset_state(amdgpu_context_handle context,
+ uint32_t *state, uint32_t *hangs)
{
union drm_amdgpu_ctx args;
int r;
return r;
}
-static uint32_t amdgpu_cs_fence_index(unsigned ip, unsigned ring)
-{
- return ip * AMDGPU_CS_MAX_RINGS + ring;
-}
-
/**
* Submit command to kernel DRM
* \param dev - \c [in] Device handle
* \sa amdgpu_cs_submit()
*/
static int amdgpu_cs_submit_one(amdgpu_context_handle context,
- struct amdgpu_cs_request *ibs_request,
- uint64_t *fence)
+ struct amdgpu_cs_request *ibs_request)
{
union drm_amdgpu_cs cs;
uint64_t *chunk_array;
struct drm_amdgpu_cs_chunk *chunks;
struct drm_amdgpu_cs_chunk_data *chunk_data;
struct drm_amdgpu_cs_chunk_dep *dependencies = NULL;
- uint32_t i, size;
+ struct drm_amdgpu_cs_chunk_dep *sem_dependencies = NULL;
+ struct list_head *sem_list;
+ amdgpu_semaphore_handle sem, tmp;
+ uint32_t i, size, sem_count = 0;
+ bool user_fence;
int r = 0;
if (ibs_request->ip_type >= AMDGPU_HW_IP_NUM)
return -EINVAL;
if (ibs_request->ring >= AMDGPU_CS_MAX_RINGS)
return -EINVAL;
- if (ibs_request->number_of_ibs > AMDGPU_CS_MAX_IBS_PER_SUBMIT)
- return -EINVAL;
+ if (ibs_request->number_of_ibs == 0) {
+ ibs_request->seq_no = AMDGPU_NULL_SUBMIT_SEQ;
+ return 0;
+ }
+ user_fence = (ibs_request->fence_info.handle != NULL);
- size = ibs_request->number_of_ibs + 2;
+ size = ibs_request->number_of_ibs + (user_fence ? 2 : 1) + 1;
chunk_array = alloca(sizeof(uint64_t) * size);
chunks = alloca(sizeof(struct drm_amdgpu_cs_chunk) * size);
- size = ibs_request->number_of_ibs + 1;
+ size = ibs_request->number_of_ibs + (user_fence ? 1 : 0);
+
chunk_data = alloca(sizeof(struct drm_amdgpu_cs_chunk_data) * size);
memset(&cs, 0, sizeof(cs));
pthread_mutex_lock(&context->sequence_mutex);
- if (ibs_request->ip_type != AMDGPU_HW_IP_UVD &&
- ibs_request->ip_type != AMDGPU_HW_IP_VCE) {
+ if (user_fence) {
i = cs.in.num_chunks++;
/* fence chunk */
chunks[i].chunk_data = (uint64_t)(uintptr_t)&chunk_data[i];
/* fence bo handle */
- chunk_data[i].fence_data.handle = context->fence_bo->handle;
+ chunk_data[i].fence_data.handle = ibs_request->fence_info.handle->handle;
/* offset */
- chunk_data[i].fence_data.offset = amdgpu_cs_fence_index(
- ibs_request->ip_type, ibs_request->ring);
- chunk_data[i].fence_data.offset *= sizeof(uint64_t);
+ chunk_data[i].fence_data.offset =
+ ibs_request->fence_info.offset * sizeof(uint64_t);
}
if (ibs_request->number_of_dependencies) {
}
for (i = 0; i < ibs_request->number_of_dependencies; ++i) {
- struct amdgpu_cs_dep_info *info = &ibs_request->dependencies[i];
+ struct amdgpu_cs_fence *info = &ibs_request->dependencies[i];
struct drm_amdgpu_cs_chunk_dep *dep = &dependencies[i];
dep->ip_type = info->ip_type;
dep->ip_instance = info->ip_instance;
chunks[i].chunk_data = (uint64_t)(uintptr_t)dependencies;
}
+ sem_list = &context->sem_list[ibs_request->ip_type][ibs_request->ip_instance][ibs_request->ring];
+ LIST_FOR_EACH_ENTRY(sem, sem_list, list)
+ sem_count++;
+ if (sem_count) {
+ sem_dependencies = malloc(sizeof(struct drm_amdgpu_cs_chunk_dep) * sem_count);
+ if (!sem_dependencies) {
+ r = -ENOMEM;
+ goto error_unlock;
+ }
+ sem_count = 0;
+ LIST_FOR_EACH_ENTRY_SAFE(sem, tmp, sem_list, list) {
+ struct amdgpu_cs_fence *info = &sem->signal_fence;
+ struct drm_amdgpu_cs_chunk_dep *dep = &sem_dependencies[sem_count++];
+ dep->ip_type = info->ip_type;
+ dep->ip_instance = info->ip_instance;
+ dep->ring = info->ring;
+ dep->ctx_id = info->context->id;
+ dep->handle = info->fence;
+
+ list_del(&sem->list);
+ amdgpu_cs_reset_sem(sem);
+ amdgpu_cs_unreference_sem(sem);
+ }
+ i = cs.in.num_chunks++;
+
+ /* dependencies chunk */
+ chunk_array[i] = (uint64_t)(uintptr_t)&chunks[i];
+ chunks[i].chunk_id = AMDGPU_CHUNK_ID_DEPENDENCIES;
+ chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_dep) / 4 * sem_count;
+ chunks[i].chunk_data = (uint64_t)(uintptr_t)sem_dependencies;
+ }
+
r = drmCommandWriteRead(context->dev->fd, DRM_AMDGPU_CS,
&cs, sizeof(cs));
if (r)
goto error_unlock;
- *fence = cs.out.handle;
-
+ ibs_request->seq_no = cs.out.handle;
+ context->last_seq[ibs_request->ip_type][ibs_request->ip_instance][ibs_request->ring] = ibs_request->seq_no;
error_unlock:
pthread_mutex_unlock(&context->sequence_mutex);
free(dependencies);
+ free(sem_dependencies);
return r;
}
-int amdgpu_cs_submit(amdgpu_context_handle context,
- uint64_t flags,
- struct amdgpu_cs_request *ibs_request,
- uint32_t number_of_requests,
- uint64_t *fences)
+drm_public int amdgpu_cs_submit(amdgpu_context_handle context,
+ uint64_t flags,
+ struct amdgpu_cs_request *ibs_request,
+ uint32_t number_of_requests)
{
uint32_t i;
int r;
- if (NULL == context)
- return -EINVAL;
- if (NULL == ibs_request)
- return -EINVAL;
- if (NULL == fences)
+ if (!context || !ibs_request)
return -EINVAL;
r = 0;
for (i = 0; i < number_of_requests; i++) {
- r = amdgpu_cs_submit_one(context, ibs_request, fences);
+ r = amdgpu_cs_submit_one(context, ibs_request);
if (r)
break;
- fences++;
ibs_request++;
}
*
* \return absolute timeout in nanoseconds
*/
-uint64_t amdgpu_cs_calculate_timeout(uint64_t timeout)
+drm_private uint64_t amdgpu_cs_calculate_timeout(uint64_t timeout)
{
int r;
if (timeout != AMDGPU_TIMEOUT_INFINITE) {
struct timespec current;
+ uint64_t current_ns;
r = clock_gettime(CLOCK_MONOTONIC, ¤t);
- if (r)
- return r;
+ if (r) {
+ fprintf(stderr, "clock_gettime() returned error (%d)!", errno);
+ return AMDGPU_TIMEOUT_INFINITE;
+ }
- timeout += ((uint64_t)current.tv_sec) * 1000000000ull;
- timeout += current.tv_nsec;
+ current_ns = ((uint64_t)current.tv_sec) * 1000000000ull;
+ current_ns += current.tv_nsec;
+ timeout += current_ns;
+ if (timeout < current_ns)
+ timeout = AMDGPU_TIMEOUT_INFINITE;
}
return timeout;
}
else
args.in.timeout = amdgpu_cs_calculate_timeout(timeout_ns);
- /* Handle errors manually here because of timeout */
- r = ioctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_CS, &args);
- if (r == -1 && (errno == EINTR || errno == EAGAIN)) {
- *busy = true;
- return 0;
- } else if (r)
+ r = drmIoctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_CS, &args);
+ if (r)
return -errno;
*busy = args.out.status;
return 0;
}
-int amdgpu_cs_query_fence_status(struct amdgpu_cs_query_fence *fence,
- uint32_t *expired)
+drm_public int amdgpu_cs_query_fence_status(struct amdgpu_cs_fence *fence,
+ uint64_t timeout_ns,
+ uint64_t flags,
+ uint32_t *expired)
{
- amdgpu_context_handle context;
- uint64_t *signaled_fence;
- uint64_t *expired_fence;
- unsigned ip_type, ip_instance;
- uint32_t ring;
bool busy = true;
int r;
- if (NULL == fence)
- return -EINVAL;
- if (NULL == expired)
- return -EINVAL;
- if (NULL == fence->context)
+ if (!fence || !expired || !fence->context)
return -EINVAL;
if (fence->ip_type >= AMDGPU_HW_IP_NUM)
return -EINVAL;
if (fence->ring >= AMDGPU_CS_MAX_RINGS)
return -EINVAL;
-
- context = fence->context;
- ip_type = fence->ip_type;
- ip_instance = fence->ip_instance;
- ring = fence->ring;
- signaled_fence = context->fence_cpu;
- signaled_fence += amdgpu_cs_fence_index(ip_type, ring);
- expired_fence = &context->expired_fences[ip_type][ip_instance][ring];
- *expired = false;
-
- pthread_mutex_lock(&context->sequence_mutex);
- if (fence->fence <= *expired_fence) {
- /* This fence value is expired already. */
- pthread_mutex_unlock(&context->sequence_mutex);
+ if (fence->fence == AMDGPU_NULL_SUBMIT_SEQ) {
*expired = true;
return 0;
}
- if (fence->fence <= *signaled_fence) {
- /* This fence value is signaled already. */
- *expired_fence = *signaled_fence;
- pthread_mutex_unlock(&context->sequence_mutex);
+ *expired = false;
+
+ r = amdgpu_ioctl_wait_cs(fence->context, fence->ip_type,
+ fence->ip_instance, fence->ring,
+ fence->fence, timeout_ns, flags, &busy);
+
+ if (!r && !busy)
*expired = true;
- return 0;
- }
- if (fence->timeout_ns == 0) {
- pthread_mutex_unlock(&context->sequence_mutex);
- return 0;
+ return r;
+}
+
+static int amdgpu_ioctl_wait_fences(struct amdgpu_cs_fence *fences,
+ uint32_t fence_count,
+ bool wait_all,
+ uint64_t timeout_ns,
+ uint32_t *status,
+ uint32_t *first)
+{
+ struct drm_amdgpu_fence *drm_fences;
+ amdgpu_device_handle dev = fences[0].context->dev;
+ union drm_amdgpu_wait_fences args;
+ int r;
+ uint32_t i;
+
+ drm_fences = alloca(sizeof(struct drm_amdgpu_fence) * fence_count);
+ for (i = 0; i < fence_count; i++) {
+ drm_fences[i].ctx_id = fences[i].context->id;
+ drm_fences[i].ip_type = fences[i].ip_type;
+ drm_fences[i].ip_instance = fences[i].ip_instance;
+ drm_fences[i].ring = fences[i].ring;
+ drm_fences[i].seq_no = fences[i].fence;
}
- pthread_mutex_unlock(&context->sequence_mutex);
+ memset(&args, 0, sizeof(args));
+ args.in.fences = (uint64_t)(uintptr_t)drm_fences;
+ args.in.fence_count = fence_count;
+ args.in.wait_all = wait_all;
+ args.in.timeout_ns = amdgpu_cs_calculate_timeout(timeout_ns);
- r = amdgpu_ioctl_wait_cs(context, ip_type, ip_instance, ring,
- fence->fence, fence->timeout_ns,
- fence->flags, &busy);
- if (!r && !busy) {
- *expired = true;
- pthread_mutex_lock(&context->sequence_mutex);
- /* The thread doesn't hold sequence_mutex. Other thread could
- update *expired_fence already. Check whether there is a
- newerly expired fence. */
- if (fence->fence > *expired_fence)
- *expired_fence = fence->fence;
- pthread_mutex_unlock(&context->sequence_mutex);
+ r = drmIoctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_FENCES, &args);
+ if (r)
+ return -errno;
+
+ *status = args.out.status;
+
+ if (first)
+ *first = args.out.first_signaled;
+
+ return 0;
+}
+
+drm_public int amdgpu_cs_wait_fences(struct amdgpu_cs_fence *fences,
+ uint32_t fence_count,
+ bool wait_all,
+ uint64_t timeout_ns,
+ uint32_t *status,
+ uint32_t *first)
+{
+ uint32_t i;
+
+ /* Sanity check */
+ if (!fences || !status || !fence_count)
+ return -EINVAL;
+
+ for (i = 0; i < fence_count; i++) {
+ if (NULL == fences[i].context)
+ return -EINVAL;
+ if (fences[i].ip_type >= AMDGPU_HW_IP_NUM)
+ return -EINVAL;
+ if (fences[i].ring >= AMDGPU_CS_MAX_RINGS)
+ return -EINVAL;
}
+ *status = 0;
+
+ return amdgpu_ioctl_wait_fences(fences, fence_count, wait_all,
+ timeout_ns, status, first);
+}
+
+drm_public int amdgpu_cs_create_semaphore(amdgpu_semaphore_handle *sem)
+{
+ struct amdgpu_semaphore *gpu_semaphore;
+
+ if (!sem)
+ return -EINVAL;
+
+ gpu_semaphore = calloc(1, sizeof(struct amdgpu_semaphore));
+ if (!gpu_semaphore)
+ return -ENOMEM;
+
+ atomic_set(&gpu_semaphore->refcount, 1);
+ *sem = gpu_semaphore;
+
+ return 0;
+}
+
+drm_public int amdgpu_cs_signal_semaphore(amdgpu_context_handle ctx,
+ uint32_t ip_type,
+ uint32_t ip_instance,
+ uint32_t ring,
+ amdgpu_semaphore_handle sem)
+{
+ if (!ctx || !sem)
+ return -EINVAL;
+ if (ip_type >= AMDGPU_HW_IP_NUM)
+ return -EINVAL;
+ if (ring >= AMDGPU_CS_MAX_RINGS)
+ return -EINVAL;
+ /* sem has been signaled */
+ if (sem->signal_fence.context)
+ return -EINVAL;
+ pthread_mutex_lock(&ctx->sequence_mutex);
+ sem->signal_fence.context = ctx;
+ sem->signal_fence.ip_type = ip_type;
+ sem->signal_fence.ip_instance = ip_instance;
+ sem->signal_fence.ring = ring;
+ sem->signal_fence.fence = ctx->last_seq[ip_type][ip_instance][ring];
+ update_references(NULL, &sem->refcount);
+ pthread_mutex_unlock(&ctx->sequence_mutex);
+ return 0;
+}
+
+drm_public int amdgpu_cs_wait_semaphore(amdgpu_context_handle ctx,
+ uint32_t ip_type,
+ uint32_t ip_instance,
+ uint32_t ring,
+ amdgpu_semaphore_handle sem)
+{
+ if (!ctx || !sem)
+ return -EINVAL;
+ if (ip_type >= AMDGPU_HW_IP_NUM)
+ return -EINVAL;
+ if (ring >= AMDGPU_CS_MAX_RINGS)
+ return -EINVAL;
+ /* must signal first */
+ if (!sem->signal_fence.context)
+ return -EINVAL;
+
+ pthread_mutex_lock(&ctx->sequence_mutex);
+ list_add(&sem->list, &ctx->sem_list[ip_type][ip_instance][ring]);
+ pthread_mutex_unlock(&ctx->sequence_mutex);
+ return 0;
+}
+
+static int amdgpu_cs_reset_sem(amdgpu_semaphore_handle sem)
+{
+ if (!sem || !sem->signal_fence.context)
+ return -EINVAL;
+
+ sem->signal_fence.context = NULL;
+ sem->signal_fence.ip_type = 0;
+ sem->signal_fence.ip_instance = 0;
+ sem->signal_fence.ring = 0;
+ sem->signal_fence.fence = 0;
+
+ return 0;
+}
+
+static int amdgpu_cs_unreference_sem(amdgpu_semaphore_handle sem)
+{
+ if (!sem)
+ return -EINVAL;
+
+ if (update_references(&sem->refcount, NULL))
+ free(sem);
+ return 0;
+}
+
+drm_public int amdgpu_cs_destroy_semaphore(amdgpu_semaphore_handle sem)
+{
+ return amdgpu_cs_unreference_sem(sem);
+}
+
+drm_public int amdgpu_cs_create_syncobj2(amdgpu_device_handle dev,
+ uint32_t flags,
+ uint32_t *handle)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjCreate(dev->fd, flags, handle);
+}
+
+drm_public int amdgpu_cs_create_syncobj(amdgpu_device_handle dev,
+ uint32_t *handle)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjCreate(dev->fd, 0, handle);
+}
+
+drm_public int amdgpu_cs_destroy_syncobj(amdgpu_device_handle dev,
+ uint32_t handle)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjDestroy(dev->fd, handle);
+}
+
+drm_public int amdgpu_cs_syncobj_reset(amdgpu_device_handle dev,
+ const uint32_t *syncobjs,
+ uint32_t syncobj_count)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjReset(dev->fd, syncobjs, syncobj_count);
+}
+
+drm_public int amdgpu_cs_syncobj_signal(amdgpu_device_handle dev,
+ const uint32_t *syncobjs,
+ uint32_t syncobj_count)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjSignal(dev->fd, syncobjs, syncobj_count);
+}
+
+drm_public int amdgpu_cs_syncobj_timeline_signal(amdgpu_device_handle dev,
+ const uint32_t *syncobjs,
+ uint64_t *points,
+ uint32_t syncobj_count)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjTimelineSignal(dev->fd, syncobjs,
+ points, syncobj_count);
+}
+
+drm_public int amdgpu_cs_syncobj_wait(amdgpu_device_handle dev,
+ uint32_t *handles, unsigned num_handles,
+ int64_t timeout_nsec, unsigned flags,
+ uint32_t *first_signaled)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjWait(dev->fd, handles, num_handles, timeout_nsec,
+ flags, first_signaled);
+}
+
+drm_public int amdgpu_cs_syncobj_timeline_wait(amdgpu_device_handle dev,
+ uint32_t *handles, uint64_t *points,
+ unsigned num_handles,
+ int64_t timeout_nsec, unsigned flags,
+ uint32_t *first_signaled)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjTimelineWait(dev->fd, handles, points, num_handles,
+ timeout_nsec, flags, first_signaled);
+}
+
+drm_public int amdgpu_cs_syncobj_query(amdgpu_device_handle dev,
+ uint32_t *handles, uint64_t *points,
+ unsigned num_handles)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjQuery(dev->fd, handles, points, num_handles);
+}
+
+drm_public int amdgpu_cs_export_syncobj(amdgpu_device_handle dev,
+ uint32_t handle,
+ int *shared_fd)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjHandleToFD(dev->fd, handle, shared_fd);
+}
+
+drm_public int amdgpu_cs_import_syncobj(amdgpu_device_handle dev,
+ int shared_fd,
+ uint32_t *handle)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjFDToHandle(dev->fd, shared_fd, handle);
+}
+
+drm_public int amdgpu_cs_syncobj_export_sync_file(amdgpu_device_handle dev,
+ uint32_t syncobj,
+ int *sync_file_fd)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjExportSyncFile(dev->fd, syncobj, sync_file_fd);
+}
+
+drm_public int amdgpu_cs_syncobj_import_sync_file(amdgpu_device_handle dev,
+ uint32_t syncobj,
+ int sync_file_fd)
+{
+ if (NULL == dev)
+ return -EINVAL;
+
+ return drmSyncobjImportSyncFile(dev->fd, syncobj, sync_file_fd);
+}
+
+drm_public int amdgpu_cs_syncobj_export_sync_file2(amdgpu_device_handle dev,
+ uint32_t syncobj,
+ uint64_t point,
+ uint32_t flags,
+ int *sync_file_fd)
+{
+ uint32_t binary_handle;
+ int ret;
+
+ if (NULL == dev)
+ return -EINVAL;
+
+ if (!point)
+ return drmSyncobjExportSyncFile(dev->fd, syncobj, sync_file_fd);
+
+ ret = drmSyncobjCreate(dev->fd, 0, &binary_handle);
+ if (ret)
+ return ret;
+
+ ret = drmSyncobjTransfer(dev->fd, binary_handle, 0,
+ syncobj, point, flags);
+ if (ret)
+ goto out;
+ ret = drmSyncobjExportSyncFile(dev->fd, binary_handle, sync_file_fd);
+out:
+ drmSyncobjDestroy(dev->fd, binary_handle);
+ return ret;
+}
+
+drm_public int amdgpu_cs_syncobj_import_sync_file2(amdgpu_device_handle dev,
+ uint32_t syncobj,
+ uint64_t point,
+ int sync_file_fd)
+{
+ uint32_t binary_handle;
+ int ret;
+
+ if (NULL == dev)
+ return -EINVAL;
+
+ if (!point)
+ return drmSyncobjImportSyncFile(dev->fd, syncobj, sync_file_fd);
+
+ ret = drmSyncobjCreate(dev->fd, 0, &binary_handle);
+ if (ret)
+ return ret;
+ ret = drmSyncobjImportSyncFile(dev->fd, binary_handle, sync_file_fd);
+ if (ret)
+ goto out;
+ ret = drmSyncobjTransfer(dev->fd, syncobj, point,
+ binary_handle, 0, 0);
+out:
+ drmSyncobjDestroy(dev->fd, binary_handle);
+ return ret;
+}
+
+drm_public int amdgpu_cs_submit_raw(amdgpu_device_handle dev,
+ amdgpu_context_handle context,
+ amdgpu_bo_list_handle bo_list_handle,
+ int num_chunks,
+ struct drm_amdgpu_cs_chunk *chunks,
+ uint64_t *seq_no)
+{
+ union drm_amdgpu_cs cs;
+ uint64_t *chunk_array;
+ int i, r;
+ if (num_chunks == 0)
+ return -EINVAL;
+
+ memset(&cs, 0, sizeof(cs));
+ chunk_array = alloca(sizeof(uint64_t) * num_chunks);
+ for (i = 0; i < num_chunks; i++)
+ chunk_array[i] = (uint64_t)(uintptr_t)&chunks[i];
+ cs.in.chunks = (uint64_t)(uintptr_t)chunk_array;
+ cs.in.ctx_id = context->id;
+ cs.in.bo_list_handle = bo_list_handle ? bo_list_handle->handle : 0;
+ cs.in.num_chunks = num_chunks;
+ r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_CS,
+ &cs, sizeof(cs));
+ if (r)
+ return r;
+
+ if (seq_no)
+ *seq_no = cs.out.handle;
+ return 0;
+}
+
+drm_public int amdgpu_cs_submit_raw2(amdgpu_device_handle dev,
+ amdgpu_context_handle context,
+ uint32_t bo_list_handle,
+ int num_chunks,
+ struct drm_amdgpu_cs_chunk *chunks,
+ uint64_t *seq_no)
+{
+ union drm_amdgpu_cs cs;
+ uint64_t *chunk_array;
+ int i, r;
+
+ memset(&cs, 0, sizeof(cs));
+ chunk_array = alloca(sizeof(uint64_t) * num_chunks);
+ for (i = 0; i < num_chunks; i++)
+ chunk_array[i] = (uint64_t)(uintptr_t)&chunks[i];
+ cs.in.chunks = (uint64_t)(uintptr_t)chunk_array;
+ cs.in.ctx_id = context->id;
+ cs.in.bo_list_handle = bo_list_handle;
+ cs.in.num_chunks = num_chunks;
+ r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_CS,
+ &cs, sizeof(cs));
+ if (!r && seq_no)
+ *seq_no = cs.out.handle;
return r;
}
+drm_public void amdgpu_cs_chunk_fence_info_to_data(struct amdgpu_cs_fence_info *fence_info,
+ struct drm_amdgpu_cs_chunk_data *data)
+{
+ data->fence_data.handle = fence_info->handle->handle;
+ data->fence_data.offset = fence_info->offset * sizeof(uint64_t);
+}
+
+drm_public void amdgpu_cs_chunk_fence_to_dep(struct amdgpu_cs_fence *fence,
+ struct drm_amdgpu_cs_chunk_dep *dep)
+{
+ dep->ip_type = fence->ip_type;
+ dep->ip_instance = fence->ip_instance;
+ dep->ring = fence->ring;
+ dep->ctx_id = fence->context->id;
+ dep->handle = fence->fence;
+}
+
+drm_public int amdgpu_cs_fence_to_handle(amdgpu_device_handle dev,
+ struct amdgpu_cs_fence *fence,
+ uint32_t what,
+ uint32_t *out_handle)
+{
+ union drm_amdgpu_fence_to_handle fth;
+ int r;
+
+ memset(&fth, 0, sizeof(fth));
+ fth.in.fence.ctx_id = fence->context->id;
+ fth.in.fence.ip_type = fence->ip_type;
+ fth.in.fence.ip_instance = fence->ip_instance;
+ fth.in.fence.ring = fence->ring;
+ fth.in.fence.seq_no = fence->fence;
+ fth.in.what = what;
+
+ r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_FENCE_TO_HANDLE,
+ &fth, sizeof(fth));
+ if (r == 0)
+ *out_handle = fth.out.handle;
+ return r;
+}