summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/msm/msm_gem_vma.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/msm/msm_gem_vma.c')
-rw-r--r--drivers/gpu/drm/msm/msm_gem_vma.c127
1 files changed, 98 insertions, 29 deletions
diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
index 3cd8562a5109..8316af1723c2 100644
--- a/drivers/gpu/drm/msm/msm_gem_vma.c
+++ b/drivers/gpu/drm/msm/msm_gem_vma.c
@@ -319,13 +319,10 @@ msm_gem_vma_map(struct drm_gpuva *vma, int prot, struct sg_table *sgt)
mutex_lock(&vm->mmu_lock);
/*
- * NOTE: iommu/io-pgtable can allocate pages, so we cannot hold
+ * NOTE: if not using pgtable preallocation, we cannot hold
* a lock across map/unmap which is also used in the job_run()
* path, as this can cause deadlock in job_run() vs shrinker/
* reclaim.
- *
- * Revisit this if we can come up with a scheme to pre-alloc pages
- * for the pgtable in map/unmap ops.
*/
ret = vm_map_op(vm, &(struct msm_vm_map_op){
.iova = vma->va.addr,
@@ -399,7 +396,14 @@ msm_gem_vma_new(struct drm_gpuvm *gpuvm, struct drm_gem_object *obj,
if (obj)
GEM_WARN_ON((range_end - range_start) > obj->size);
- drm_gpuva_init(&vma->base, range_start, range_end - range_start, obj, offset);
+ struct drm_gpuva_op_map op_map = {
+ .va.addr = range_start,
+ .va.range = range_end - range_start,
+ .gem.obj = obj,
+ .gem.offset = offset,
+ };
+
+ drm_gpuva_init_from_op(&vma->base, &op_map);
vma->mapped = false;
ret = drm_gpuva_insert(&vm->base, &vma->base);
@@ -454,6 +458,8 @@ msm_gem_vm_bo_validate(struct drm_gpuvm_bo *vm_bo, struct drm_exec *exec)
struct op_arg {
unsigned flags;
struct msm_vm_bind_job *job;
+ const struct msm_vm_bind_op *op;
+ bool kept;
};
static void
@@ -475,14 +481,18 @@ vma_from_op(struct op_arg *arg, struct drm_gpuva_op_map *op)
}
static int
-msm_gem_vm_sm_step_map(struct drm_gpuva_op *op, void *arg)
+msm_gem_vm_sm_step_map(struct drm_gpuva_op *op, void *_arg)
{
- struct msm_vm_bind_job *job = ((struct op_arg *)arg)->job;
+ struct op_arg *arg = _arg;
+ struct msm_vm_bind_job *job = arg->job;
struct drm_gem_object *obj = op->map.gem.obj;
struct drm_gpuva *vma;
struct sg_table *sgt;
unsigned prot;
+ if (arg->kept)
+ return 0;
+
vma = vma_from_op(arg, &op->map);
if (WARN_ON(IS_ERR(vma)))
return PTR_ERR(vma);
@@ -602,15 +612,41 @@ msm_gem_vm_sm_step_remap(struct drm_gpuva_op *op, void *arg)
}
static int
-msm_gem_vm_sm_step_unmap(struct drm_gpuva_op *op, void *arg)
+msm_gem_vm_sm_step_unmap(struct drm_gpuva_op *op, void *_arg)
{
- struct msm_vm_bind_job *job = ((struct op_arg *)arg)->job;
+ struct op_arg *arg = _arg;
+ struct msm_vm_bind_job *job = arg->job;
struct drm_gpuva *vma = op->unmap.va;
struct msm_gem_vma *msm_vma = to_msm_vma(vma);
vm_dbg("%p:%p:%p: %016llx %016llx", vma->vm, vma, vma->gem.obj,
vma->va.addr, vma->va.range);
+ /*
+ * Detect in-place remap. Turnip does this to change the vma flags,
+ * in particular MSM_VMA_DUMP. In this case we want to avoid actually
+ * touching the page tables, as that would require synchronization
+ * against SUBMIT jobs running on the GPU.
+ */
+ if (op->unmap.keep &&
+ (arg->op->op == MSM_VM_BIND_OP_MAP) &&
+ (vma->gem.obj == arg->op->obj) &&
+ (vma->gem.offset == arg->op->obj_offset) &&
+ (vma->va.addr == arg->op->iova) &&
+ (vma->va.range == arg->op->range)) {
+ /* We are only expecting a single in-place unmap+map cb pair: */
+ WARN_ON(arg->kept);
+
+ /* Leave the existing VMA in place, but signal that to the map cb: */
+ arg->kept = true;
+
+ /* Only flags are changing, so update that in-place: */
+ unsigned orig_flags = vma->flags & (DRM_GPUVA_USERBITS - 1);
+ vma->flags = orig_flags | arg->flags;
+
+ return 0;
+ }
+
if (!msm_vma->mapped)
goto out_close;
@@ -994,6 +1030,7 @@ vm_bind_job_lookup_ops(struct msm_vm_bind_job *job, struct drm_msm_vm_bind *args
struct drm_device *dev = job->vm->drm;
int ret = 0;
int cnt = 0;
+ int i = -1;
if (args->nr_ops == 1) {
/* Single op case, the op is inlined: */
@@ -1027,11 +1064,12 @@ vm_bind_job_lookup_ops(struct msm_vm_bind_job *job, struct drm_msm_vm_bind *args
spin_lock(&file->table_lock);
- for (unsigned i = 0; i < args->nr_ops; i++) {
+ for (i = 0; i < args->nr_ops; i++) {
+ struct msm_vm_bind_op *op = &job->ops[i];
struct drm_gem_object *obj;
- if (!job->ops[i].handle) {
- job->ops[i].obj = NULL;
+ if (!op->handle) {
+ op->obj = NULL;
continue;
}
@@ -1039,16 +1077,22 @@ vm_bind_job_lookup_ops(struct msm_vm_bind_job *job, struct drm_msm_vm_bind *args
* normally use drm_gem_object_lookup(), but for bulk lookup
* all under single table_lock just hit object_idr directly:
*/
- obj = idr_find(&file->object_idr, job->ops[i].handle);
+ obj = idr_find(&file->object_idr, op->handle);
if (!obj) {
- ret = UERR(EINVAL, dev, "invalid handle %u at index %u\n", job->ops[i].handle, i);
+ ret = UERR(EINVAL, dev, "invalid handle %u at index %u\n", op->handle, i);
goto out_unlock;
}
drm_gem_object_get(obj);
- job->ops[i].obj = obj;
+ op->obj = obj;
cnt++;
+
+ if ((op->range + op->obj_offset) > obj->size) {
+ ret = UERR(EINVAL, dev, "invalid range: %016llx + %016llx > %016zx\n",
+ op->range, op->obj_offset, obj->size);
+ goto out_unlock;
+ }
}
*nr_bos = cnt;
@@ -1056,6 +1100,17 @@ vm_bind_job_lookup_ops(struct msm_vm_bind_job *job, struct drm_msm_vm_bind *args
out_unlock:
spin_unlock(&file->table_lock);
+ if (ret) {
+ for (; i >= 0; i--) {
+ struct msm_vm_bind_op *op = &job->ops[i];
+
+ if (!op->obj)
+ continue;
+
+ drm_gem_object_put(op->obj);
+ op->obj = NULL;
+ }
+ }
out:
return ret;
}
@@ -1171,11 +1226,17 @@ vm_bind_job_lock_objects(struct msm_vm_bind_job *job, struct drm_exec *exec)
op->obj_offset);
break;
case MSM_VM_BIND_OP_MAP:
- case MSM_VM_BIND_OP_MAP_NULL:
- ret = drm_gpuvm_sm_map_exec_lock(job->vm, exec, 1,
- op->iova, op->range,
- op->obj, op->obj_offset);
+ case MSM_VM_BIND_OP_MAP_NULL: {
+ struct drm_gpuvm_map_req map_req = {
+ .map.va.addr = op->iova,
+ .map.va.range = op->range,
+ .map.gem.obj = op->obj,
+ .map.gem.offset = op->obj_offset,
+ };
+
+ ret = drm_gpuvm_sm_map_exec_lock(job->vm, exec, 1, &map_req);
break;
+ }
default:
/*
* lookup_op() should have already thrown an error for
@@ -1271,6 +1332,7 @@ vm_bind_job_prepare(struct msm_vm_bind_job *job)
const struct msm_vm_bind_op *op = &job->ops[i];
struct op_arg arg = {
.job = job,
+ .op = op,
};
switch (op->op) {
@@ -1282,10 +1344,17 @@ vm_bind_job_prepare(struct msm_vm_bind_job *job)
if (op->flags & MSM_VM_BIND_OP_DUMP)
arg.flags |= MSM_VMA_DUMP;
fallthrough;
- case MSM_VM_BIND_OP_MAP_NULL:
- ret = drm_gpuvm_sm_map(job->vm, &arg, op->iova,
- op->range, op->obj, op->obj_offset);
+ case MSM_VM_BIND_OP_MAP_NULL: {
+ struct drm_gpuvm_map_req map_req = {
+ .map.va.addr = op->iova,
+ .map.va.range = op->range,
+ .map.gem.obj = op->obj,
+ .map.gem.offset = op->obj_offset,
+ };
+
+ ret = drm_gpuvm_sm_map(job->vm, &arg, &map_req);
break;
+ }
default:
/*
* lookup_op() should have already thrown an error for
@@ -1460,12 +1529,8 @@ msm_ioctl_vm_bind(struct drm_device *dev, void *data, struct drm_file *file)
if (args->flags & MSM_VM_BIND_FENCE_FD_OUT) {
sync_file = sync_file_create(job->fence);
- if (!sync_file) {
+ if (!sync_file)
ret = -ENOMEM;
- } else {
- fd_install(out_fence_fd, sync_file->file);
- args->fence_fd = out_fence_fd;
- }
}
if (ret)
@@ -1494,10 +1559,14 @@ out:
out_unlock:
mutex_unlock(&queue->lock);
out_post_unlock:
- if (ret && (out_fence_fd >= 0)) {
- put_unused_fd(out_fence_fd);
+ if (ret) {
+ if (out_fence_fd >= 0)
+ put_unused_fd(out_fence_fd);
if (sync_file)
fput(sync_file->file);
+ } else if (sync_file) {
+ fd_install(out_fence_fd, sync_file->file);
+ args->fence_fd = out_fence_fd;
}
if (!IS_ERR_OR_NULL(job)) {
generated by cgit v1.2.3 (git 2.46.0) at 2025-10-03 17:32:34 +0000