[PATCH v2 1/4] drm/vc4: Use DRM Execution Contexts
Melissa Wen
mwen at igalia.com
Tue Dec 17 12:50:16 UTC 2024
On 12/17, Christian König wrote:
> Am 16.12.24 um 20:08 schrieb Melissa Wen:
> > On 12/12, Maíra Canal wrote:
> > > VC4 has internal copies of `drm_gem_lock_reservations()` and
> > > `drm_gem_unlock_reservations()` inside the driver and ideally, we should
> > > move those hard-coded functions to the generic functions provided by DRM.
> > > But, instead of using the DRM GEM functions to (un)lock reservations, move
> > > the new DRM Execution Contexts API.
> > >
> > > Signed-off-by: Maíra Canal <mcanal at igalia.com>
> > > ---
> > > drivers/gpu/drm/vc4/Kconfig | 1 +
> > > drivers/gpu/drm/vc4/vc4_gem.c | 99 ++++++++---------------------------
> > > 2 files changed, 22 insertions(+), 78 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/vc4/Kconfig b/drivers/gpu/drm/vc4/Kconfig
> > > index c5f30b317698..0627e826fda4 100644
> > > --- a/drivers/gpu/drm/vc4/Kconfig
> > > +++ b/drivers/gpu/drm/vc4/Kconfig
> > > @@ -13,6 +13,7 @@ config DRM_VC4
> > > select DRM_DISPLAY_HDMI_HELPER
> > > select DRM_DISPLAY_HDMI_STATE_HELPER
> > > select DRM_DISPLAY_HELPER
> > > + select DRM_EXEC
> > > select DRM_KMS_HELPER
> > > select DRM_GEM_DMA_HELPER
> > > select DRM_PANEL_BRIDGE
> > > diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c
> > > index 22bccd69eb62..1021f45cb53c 100644
> > > --- a/drivers/gpu/drm/vc4/vc4_gem.c
> > > +++ b/drivers/gpu/drm/vc4/vc4_gem.c
> > > @@ -29,6 +29,7 @@
> > > #include <linux/sched/signal.h>
> > > #include <linux/dma-fence-array.h>
> > > +#include <drm/drm_exec.h>
> > > #include <drm/drm_syncobj.h>
> > > #include "uapi/drm/vc4_drm.h"
> > > @@ -578,19 +579,6 @@ vc4_update_bo_seqnos(struct vc4_exec_info *exec, uint64_t seqno)
> > > }
> > > }
> > > -static void
> > > -vc4_unlock_bo_reservations(struct drm_device *dev,
> > > - struct vc4_exec_info *exec,
> > > - struct ww_acquire_ctx *acquire_ctx)
> > > -{
> > > - int i;
> > > -
> > > - for (i = 0; i < exec->bo_count; i++)
> > > - dma_resv_unlock(exec->bo[i]->resv);
> > > -
> > > - ww_acquire_fini(acquire_ctx);
> > > -}
> > > -
> > > /* Takes the reservation lock on all the BOs being referenced, so that
> > > * at queue submit time we can update the reservations.
> > > *
> > > @@ -599,70 +587,23 @@ vc4_unlock_bo_reservations(struct drm_device *dev,
> > > * to vc4, so we don't attach dma-buf fences to them.
> > > */
> > > static int
> > > -vc4_lock_bo_reservations(struct drm_device *dev,
> > > - struct vc4_exec_info *exec,
> > > - struct ww_acquire_ctx *acquire_ctx)
> > > +vc4_lock_bo_reservations(struct vc4_exec_info *exec,
> > > + struct drm_exec *exec_ctx)
> > > {
> > > - int contended_lock = -1;
> > > - int i, ret;
> > > - struct drm_gem_object *bo;
> > > -
> > > - ww_acquire_init(acquire_ctx, &reservation_ww_class);
> > > -
> > > -retry:
> > > - if (contended_lock != -1) {
> > > - bo = exec->bo[contended_lock];
> > > - ret = dma_resv_lock_slow_interruptible(bo->resv, acquire_ctx);
> > > - if (ret) {
> > > - ww_acquire_done(acquire_ctx);
> > > - return ret;
> > > - }
> > > - }
> > > -
> > > - for (i = 0; i < exec->bo_count; i++) {
> > > - if (i == contended_lock)
> > > - continue;
> > > -
> > > - bo = exec->bo[i];
> > > -
> > > - ret = dma_resv_lock_interruptible(bo->resv, acquire_ctx);
> > > - if (ret) {
> > > - int j;
> > > -
> > > - for (j = 0; j < i; j++) {
> > > - bo = exec->bo[j];
> > > - dma_resv_unlock(bo->resv);
> > > - }
> > > -
> > > - if (contended_lock != -1 && contended_lock >= i) {
> > > - bo = exec->bo[contended_lock];
> > > -
> > > - dma_resv_unlock(bo->resv);
> > > - }
> > > -
> > > - if (ret == -EDEADLK) {
> > > - contended_lock = i;
> > > - goto retry;
> > > - }
> > > -
> > > - ww_acquire_done(acquire_ctx);
> > > - return ret;
> > > - }
> > > - }
> > > -
> > > - ww_acquire_done(acquire_ctx);
> > > + int ret;
> > > /* Reserve space for our shared (read-only) fence references,
> > > * before we commit the CL to the hardware.
> > > */
> > > - for (i = 0; i < exec->bo_count; i++) {
> > > - bo = exec->bo[i];
> > > + drm_exec_init(exec_ctx, DRM_EXEC_INTERRUPTIBLE_WAIT, exec->bo_count);
> > > + drm_exec_until_all_locked(exec_ctx) {
> > > + ret = drm_exec_prepare_array(exec_ctx, exec->bo,
> > > + exec->bo_count, 1);
> > Hi Maíra,
> >
> > So, I'm not familiar too drm_exec, but the original implementation of
> > vc4_lock_bo_reservations() has a retry of locks from the contention and
> > I don't see it inside the drm_exec_prepare_array(), why don't use the
> > loop drm_exec_prepare_obj() plus drm_exec_retry_on_contention() (similar
> > to the typical usage documented for drm_exec)?
>
> The way how drm_exec and drm_exec_prepare_array is used seems to be correct
> here.
>
> drm_exec_prepare_array() basically just loops over all the GEM BOs in the
> array and calls drm_exec_prepare_obj() on them, so no need to open code
> that.
>
> drm_exec_retry_on_contention() is only needed if you have multiple calls to
> drm_exec_prepare_array() or drm_exec_prepare_obj(), so that the loop is
> restarted in between the calls.
Current behavior of vc4 is to retry locking the contended lock and then
all others, instead of just return on error, and I don't see this being
covered by drm_exec_prepare_array(). vc4 behavior seems closer to what
is done by MSM, a loop of drm_exec_prepare_obj() plus
drm_exec_retry_on_contention().
Moreover, I don't see any DRM driver using drm_exec_prepare_array(), so
maybe updating this function would make it more useful and meet vc4
current behavior too. Something as below:
```
int drm_exec_prepare_array(struct drm_exec *exec,
struct drm_gem_object **objects,
unsigned int num_objects,
unsigned int num_fences)
{
int ret;
for (unsigned int i = 0; i < num_objects; ++i) {
ret = drm_exec_prepare_obj(exec, objects[i], num_fences);
drm_exec_retry_on_contention(exec); <----- here
if (unlikely(ret))
return ret;
}
return 0;
}
```
WDYT?
Melissa
>
> Regards,
> Christian.
>
> >
> > Also, probably you already considered that: we can extend this update to
> > v3d, right?
> >
> > Melissa
> >
> > > + }
> > > - ret = dma_resv_reserve_fences(bo->resv, 1);
> > > - if (ret) {
> > > - vc4_unlock_bo_reservations(dev, exec, acquire_ctx);
> > > - return ret;
> > > - }
> > > + if (ret) {
> > > + drm_exec_fini(exec_ctx);
> > > + return ret;
> > > }
> > > return 0;
> > > @@ -679,7 +620,7 @@ vc4_lock_bo_reservations(struct drm_device *dev,
> > > */
> > > static int
> > > vc4_queue_submit(struct drm_device *dev, struct vc4_exec_info *exec,
> > > - struct ww_acquire_ctx *acquire_ctx,
> > > + struct drm_exec *exec_ctx,
> > > struct drm_syncobj *out_sync)
> > > {
> > > struct vc4_dev *vc4 = to_vc4_dev(dev);
> > > @@ -708,7 +649,7 @@ vc4_queue_submit(struct drm_device *dev, struct vc4_exec_info *exec,
> > > vc4_update_bo_seqnos(exec, seqno);
> > > - vc4_unlock_bo_reservations(dev, exec, acquire_ctx);
> > > + drm_exec_fini(exec_ctx);
> > > list_add_tail(&exec->head, &vc4->bin_job_list);
> > > @@ -1123,7 +1064,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
> > > struct drm_vc4_submit_cl *args = data;
> > > struct drm_syncobj *out_sync = NULL;
> > > struct vc4_exec_info *exec;
> > > - struct ww_acquire_ctx acquire_ctx;
> > > + struct drm_exec exec_ctx;
> > > struct dma_fence *in_fence;
> > > int ret = 0;
> > > @@ -1216,7 +1157,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
> > > if (ret)
> > > goto fail;
> > > - ret = vc4_lock_bo_reservations(dev, exec, &acquire_ctx);
> > > + ret = vc4_lock_bo_reservations(exec, &exec_ctx);
> > > if (ret)
> > > goto fail;
> > > @@ -1224,7 +1165,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
> > > out_sync = drm_syncobj_find(file_priv, args->out_sync);
> > > if (!out_sync) {
> > > ret = -EINVAL;
> > > - goto fail;
> > > + goto fail_unreserve;
> > > }
> > > /* We replace the fence in out_sync in vc4_queue_submit since
> > > @@ -1239,7 +1180,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
> > > */
> > > exec->args = NULL;
> > > - ret = vc4_queue_submit(dev, exec, &acquire_ctx, out_sync);
> > > + ret = vc4_queue_submit(dev, exec, &exec_ctx, out_sync);
> > > /* The syncobj isn't part of the exec data and we need to free our
> > > * reference even if job submission failed.
> > > @@ -1248,13 +1189,15 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
> > > drm_syncobj_put(out_sync);
> > > if (ret)
> > > - goto fail;
> > > + goto fail_unreserve;
> > > /* Return the seqno for our job. */
> > > args->seqno = vc4->emit_seqno;
> > > return 0;
> > > +fail_unreserve:
> > > + drm_exec_fini(&exec_ctx);
> > > fail:
> > > vc4_complete_exec(&vc4->base, exec);
> > > --
> > > 2.47.1
> > >
>
More information about the dri-devel
mailing list