[PATCH 5/8] drm/xe/oa: Signal output fences
Cavitt, Jonathan
jonathan.cavitt at intel.com
Thu Aug 8 21:16:24 UTC 2024
-----Original Message-----
From: Intel-xe <intel-xe-bounces at lists.freedesktop.org> On Behalf Of Ashutosh Dixit
Sent: Thursday, August 8, 2024 10:42 AM
To: intel-xe at lists.freedesktop.org
Cc: Nerlige Ramappa, Umesh <umesh.nerlige.ramappa at intel.com>; Souza, Jose <jose.souza at intel.com>; Landwerlin, Lionel G <lionel.g.landwerlin at intel.com>
Subject: [PATCH 5/8] drm/xe/oa: Signal output fences
>
> Complete 'struct xe_oa_fence' to include the dma_fence used to signal
> output fences in the xe_sync array. The fences are signalled
> asynchronously. When there are no output fences to signal, the OA
> configuration wait is synchronously re-introduced into the ioctl.
s/signalled/signaled
Also, it might be best to elaborate on why the dma_fence base
and spinlock_t lock weren't added as a part of patch 2, when
the xe_oa_fence was initially declared.
Otherwise:
Reviewed-by: Jonathan Cavitt <jonathan.cavitt at intel.com>
-Jonathan Cavitt
>
> Signed-off-by: Ashutosh Dixit <ashutosh.dixit at intel.com>
> ---
> drivers/gpu/drm/xe/xe_oa.c | 46 +++++++++++++++++++++++++++++++++++---
> 1 file changed, 43 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c
> index 416e031ac454b..bc421cd0af6ba 100644
> --- a/drivers/gpu/drm/xe/xe_oa.c
> +++ b/drivers/gpu/drm/xe/xe_oa.c
> @@ -96,6 +96,10 @@ struct xe_oa_config_bo {
> };
>
> struct xe_oa_fence {
> + /* @base: dma fence base */
> + struct dma_fence base;
> + /* @lock: lock for the fence */
> + spinlock_t lock;
> /* @xe: pointer to xe device */
> struct xe_device *xe;
> /* @work: work to signal that OA configuration is applied */
> @@ -953,9 +957,26 @@ static void xe_oa_fence_work_fn(struct work_struct *w)
> /* Additional empirical delay needed for NOA programming after registers are written */
> usleep_range(us, 2 * us);
>
> - kfree(ofence);
> + /* Now signal fence to indicate new OA configuration is active */
> + dma_fence_signal(&ofence->base);
> + dma_fence_put(&ofence->base);
> }
>
> +static const char *xe_oa_get_driver_name(struct dma_fence *fence)
> +{
> + return "xe_oa";
> +}
> +
> +static const char *xe_oa_get_timeline_name(struct dma_fence *fence)
> +{
> + return "unbound";
> +}
> +
> +static const struct dma_fence_ops xe_oa_fence_ops = {
> + .get_driver_name = xe_oa_get_driver_name,
> + .get_timeline_name = xe_oa_get_timeline_name,
> +};
> +
> static struct xe_oa_fence *xe_oa_fence_init(struct xe_device *xe, struct dma_fence *config_fence)
> {
> struct xe_oa_fence *ofence;
> @@ -967,6 +988,8 @@ static struct xe_oa_fence *xe_oa_fence_init(struct xe_device *xe, struct dma_fen
> ofence->xe = xe;
> INIT_WORK(&ofence->work, xe_oa_fence_work_fn);
> ofence->config_fence = config_fence;
> + spin_lock_init(&ofence->lock);
> + dma_fence_init(&ofence->base, &xe_oa_fence_ops, &ofence->lock, 0, 0);
>
> return ofence;
> }
> @@ -975,8 +998,8 @@ static int xe_oa_emit_oa_config(struct xe_oa_stream *stream, struct xe_oa_config
> {
> struct xe_oa_config_bo *oa_bo;
> struct xe_oa_fence *ofence;
> + int i, err, num_signal = 0;
> struct dma_fence *fence;
> - int err;
>
> /* Emit OA configuration batch */
> oa_bo = xe_oa_alloc_config_buffer(stream, config);
> @@ -989,13 +1012,30 @@ static int xe_oa_emit_oa_config(struct xe_oa_stream *stream, struct xe_oa_config
> if (err)
> goto exit;
>
> + /* Initialize and set fence to signal */
> ofence = xe_oa_fence_init(stream->oa->xe, fence);
> if (IS_ERR(ofence)) {
> err = PTR_ERR(ofence);
> goto put_fence;
> }
>
> - xe_oa_fence_work_fn(&ofence->work);
> + for (i = 0; i < stream->num_syncs; i++)
> + xe_sync_entry_signal(&stream->syncs[i], &ofence->base);
> +
> + /* Schedule work to signal the fence */
> + queue_work(system_unbound_wq, &ofence->work);
> +
> + /* If nothing needs to be signaled we wait synchronously */
> + for (i = 0; i < stream->num_syncs; i++)
> + if (stream->syncs[i].flags & DRM_XE_SYNC_FLAG_SIGNAL)
> + num_signal++;
> + if (!num_signal)
> + flush_work(&ofence->work);
> +
> + /* Done with syncs */
> + for (i = 0; i < stream->num_syncs; i++)
> + xe_sync_entry_cleanup(&stream->syncs[i]);
> + kfree(stream->syncs);
>
> return 0;
> put_fence:
> --
> 2.41.0
>
>
More information about the Intel-xe
mailing list