summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/nouveau/nouveau_exec.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/nouveau/nouveau_exec.c')
-rw-r--r--drivers/gpu/drm/nouveau/nouveau_exec.c106
1 files changed, 45 insertions, 61 deletions
diff --git a/drivers/gpu/drm/nouveau/nouveau_exec.c b/drivers/gpu/drm/nouveau/nouveau_exec.c
index a90c4cd8cbb2..c4949e815eb3 100644
--- a/drivers/gpu/drm/nouveau/nouveau_exec.c
+++ b/drivers/gpu/drm/nouveau/nouveau_exec.c
@@ -1,7 +1,5 @@
// SPDX-License-Identifier: MIT
-#include <drm/drm_exec.h>
-
#include "nouveau_drv.h"
#include "nouveau_gem.h"
#include "nouveau_mem.h"
@@ -12,6 +10,8 @@
#include "nouveau_sched.h"
#include "nouveau_uvmm.h"
+#include <nvif/class.h>
+
/**
* DOC: Overview
*
@@ -60,14 +60,14 @@
* virtual address in the GPU's VA space there is no guarantee that the actual
* mappings are created in the GPU's MMU. If the given memory is swapped out
* at the time the bind operation is executed the kernel will stash the mapping
- * details into it's internal alloctor and create the actual MMU mappings once
+ * details into it's internal allocator and create the actual MMU mappings once
* the memory is swapped back in. While this is transparent for userspace, it is
* guaranteed that all the backing memory is swapped back in and all the memory
* mappings, as requested by userspace previously, are actually mapped once the
* DRM_NOUVEAU_EXEC ioctl is called to submit an exec job.
*
* A VM_BIND job can be executed either synchronously or asynchronously. If
- * exectued asynchronously, userspace may provide a list of syncobjs this job
+ * executed asynchronously, userspace may provide a list of syncobjs this job
* will wait for and/or a list of syncobj the kernel will signal once the
* VM_BIND job finished execution. If executed synchronously the ioctl will
* block until the bind job is finished. For synchronous jobs the kernel will
@@ -82,71 +82,47 @@
* Since VM_BIND jobs update the GPU's VA space on job submit, EXEC jobs do have
* an up to date view of the VA space. However, the actual mappings might still
* be pending. Hence, EXEC jobs require to have the particular fences - of
- * the corresponding VM_BIND jobs they depent on - attached to them.
+ * the corresponding VM_BIND jobs they depend on - attached to them.
*/
static int
-nouveau_exec_job_submit(struct nouveau_job *job)
+nouveau_exec_job_submit(struct nouveau_job *job,
+ struct drm_gpuvm_exec *vme)
{
struct nouveau_exec_job *exec_job = to_nouveau_exec_job(job);
struct nouveau_cli *cli = job->cli;
struct nouveau_uvmm *uvmm = nouveau_cli_uvmm(cli);
- struct drm_exec *exec = &job->exec;
- struct drm_gem_object *obj;
- unsigned long index;
int ret;
- ret = nouveau_fence_new(&exec_job->fence);
+ /* Create a new fence, but do not emit yet. */
+ ret = nouveau_fence_create(&exec_job->fence, exec_job->chan);
if (ret)
return ret;
nouveau_uvmm_lock(uvmm);
- drm_exec_init(exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
- DRM_EXEC_IGNORE_DUPLICATES);
- drm_exec_until_all_locked(exec) {
- struct drm_gpuva *va;
-
- drm_gpuva_for_each_va(va, &uvmm->umgr) {
- if (unlikely(va == &uvmm->umgr.kernel_alloc_node))
- continue;
-
- ret = drm_exec_prepare_obj(exec, va->gem.obj, 1);
- drm_exec_retry_on_contention(exec);
- if (ret)
- goto err_uvmm_unlock;
- }
+ ret = drm_gpuvm_exec_lock(vme);
+ if (ret) {
+ nouveau_uvmm_unlock(uvmm);
+ return ret;
}
nouveau_uvmm_unlock(uvmm);
- drm_exec_for_each_locked_object(exec, index, obj) {
- struct nouveau_bo *nvbo = nouveau_gem_object(obj);
-
- ret = nouveau_bo_validate(nvbo, true, false);
- if (ret)
- goto err_exec_fini;
+ ret = drm_gpuvm_exec_validate(vme);
+ if (ret) {
+ drm_gpuvm_exec_unlock(vme);
+ return ret;
}
return 0;
-
-err_uvmm_unlock:
- nouveau_uvmm_unlock(uvmm);
-err_exec_fini:
- drm_exec_fini(exec);
- return ret;
-
}
static void
-nouveau_exec_job_armed_submit(struct nouveau_job *job)
+nouveau_exec_job_armed_submit(struct nouveau_job *job,
+ struct drm_gpuvm_exec *vme)
{
- struct drm_exec *exec = &job->exec;
- struct drm_gem_object *obj;
- unsigned long index;
-
- drm_exec_for_each_locked_object(exec, index, obj)
- dma_resv_add_fence(obj->resv, job->done_fence, job->resv_usage);
-
- drm_exec_fini(exec);
+ drm_gpuvm_exec_resv_add_fence(vme, job->done_fence,
+ job->resv_usage, job->resv_usage);
+ drm_gpuvm_exec_unlock(vme);
}
static struct dma_fence *
@@ -157,7 +133,7 @@ nouveau_exec_job_run(struct nouveau_job *job)
struct nouveau_fence *fence = exec_job->fence;
int i, ret;
- ret = nouveau_dma_wait(chan, exec_job->push.count + 1, 16);
+ ret = nvif_chan_gpfifo_wait(&chan->chan, exec_job->push.count + 1, 16);
if (ret) {
NV_PRINTK(err, job->cli, "nv50cal_space: %d\n", ret);
return ERR_PTR(ret);
@@ -167,16 +143,22 @@ nouveau_exec_job_run(struct nouveau_job *job)
struct drm_nouveau_exec_push *p = &exec_job->push.s[i];
bool no_prefetch = p->flags & DRM_NOUVEAU_EXEC_PUSH_NO_PREFETCH;
- nv50_dma_push(chan, p->va, p->va_len, no_prefetch);
+ nvif_chan_gpfifo_push(&chan->chan, p->va, p->va_len, no_prefetch);
}
- ret = nouveau_fence_emit(fence, chan);
+ nvif_chan_gpfifo_post(&chan->chan);
+
+ ret = nouveau_fence_emit(fence);
if (ret) {
+ nouveau_fence_unref(&exec_job->fence);
NV_PRINTK(err, job->cli, "error fencing pushbuf: %d\n", ret);
WIND_RING(chan);
return ERR_PTR(ret);
}
+ /* The fence was emitted successfully, set the job's fence pointer to
+ * NULL in order to avoid freeing it up when the job is cleaned up.
+ */
exec_job->fence = NULL;
return &fence->base;
@@ -187,9 +169,10 @@ nouveau_exec_job_free(struct nouveau_job *job)
{
struct nouveau_exec_job *exec_job = to_nouveau_exec_job(job);
+ nouveau_job_done(job);
nouveau_job_free(job);
- nouveau_fence_unref(&exec_job->fence);
+ kfree(exec_job->fence);
kfree(exec_job->push.s);
kfree(exec_job);
}
@@ -206,12 +189,10 @@ nouveau_exec_job_timeout(struct nouveau_job *job)
NV_PRINTK(warn, job->cli, "job timeout, channel %d killed!\n",
chan->chid);
- nouveau_sched_entity_fini(job->entity);
-
- return DRM_GPU_SCHED_STAT_ENODEV;
+ return DRM_GPU_SCHED_STAT_RESET;
}
-static struct nouveau_job_ops nouveau_exec_job_ops = {
+static const struct nouveau_job_ops nouveau_exec_job_ops = {
.submit = nouveau_exec_job_submit,
.armed_submit = nouveau_exec_job_armed_submit,
.run = nouveau_exec_job_run,
@@ -254,10 +235,12 @@ nouveau_exec_job_init(struct nouveau_exec_job **pjob,
}
}
+ args.file_priv = __args->file_priv;
job->chan = __args->chan;
- args.sched_entity = __args->sched_entity;
- args.file_priv = __args->file_priv;
+ args.sched = __args->sched;
+ /* Plus one to account for the HW fence. */
+ args.credits = job->push.count + 1;
args.in_sync.count = __args->in_sync.count;
args.in_sync.s = __args->in_sync.s;
@@ -374,7 +357,7 @@ nouveau_exec_ioctl_exec(struct drm_device *dev,
struct nouveau_channel *chan = NULL;
struct nouveau_exec_job_args args = {};
struct drm_nouveau_exec *req = data;
- int ret = 0;
+ int push_max, ret = 0;
if (unlikely(!abi16))
return -ENOMEM;
@@ -396,12 +379,13 @@ nouveau_exec_ioctl_exec(struct drm_device *dev,
if (unlikely(atomic_read(&chan->killed)))
return nouveau_abi16_put(abi16, -ENODEV);
- if (!chan->dma.ib_max)
+ if (chan->user.oclass < NV50_CHANNEL_GPFIFO)
return nouveau_abi16_put(abi16, -ENOSYS);
- if (unlikely(req->push_count > NOUVEAU_GEM_MAX_PUSH)) {
+ push_max = nouveau_exec_push_max_from_ib_max(chan->chan.gpfifo.max);
+ if (unlikely(req->push_count > push_max)) {
NV_PRINTK(err, cli, "pushbuf push count exceeds limit: %d max %d\n",
- req->push_count, NOUVEAU_GEM_MAX_PUSH);
+ req->push_count, push_max);
return nouveau_abi16_put(abi16, -EINVAL);
}
@@ -409,7 +393,7 @@ nouveau_exec_ioctl_exec(struct drm_device *dev,
if (ret)
goto out;
- args.sched_entity = &chan16->sched_entity;
+ args.sched = chan16->sched;
args.file_priv = file_priv;
args.chan = chan;