Re: [PATCH v4 09/14] drm/syncobj: Add deadline support for syncobj waits

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Feb 20, 2023 at 1:05 AM Pekka Paalanen <ppaalanen@xxxxxxxxx> wrote:
>
> On Sat, 18 Feb 2023 13:15:52 -0800
> Rob Clark <robdclark@xxxxxxxxx> wrote:
>
> > From: Rob Clark <robdclark@xxxxxxxxxxxx>
> >
> > Add a new flag to let userspace provide a deadline as a hint for syncobj
> > and timeline waits.  This gives a hint to the driver signaling the
> > backing fences about how soon userspace needs it to compete work, so it
> > can addjust GPU frequency accordingly.  An immediate deadline can be
> > given to provide something equivalent to i915 "wait boost".
> >
> > Signed-off-by: Rob Clark <robdclark@xxxxxxxxxxxx>
> > ---
> >
> > I'm a bit on the fence about the addition of the DRM_CAP, but it seems
> > useful to give userspace a way to probe whether the kernel and driver
> > supports the new wait flag, especially since we have vk-common code
> > dealing with syncobjs.  But open to suggestions.
> >
> >  drivers/gpu/drm/drm_ioctl.c   |  3 ++
> >  drivers/gpu/drm/drm_syncobj.c | 59 ++++++++++++++++++++++++++++-------
> >  include/drm/drm_drv.h         |  6 ++++
> >  include/uapi/drm/drm.h        | 16 ++++++++--
> >  4 files changed, 71 insertions(+), 13 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c
> > index 7c9d66ee917d..1c5c942cf0f9 100644
> > --- a/drivers/gpu/drm/drm_ioctl.c
> > +++ b/drivers/gpu/drm/drm_ioctl.c
> > @@ -254,6 +254,9 @@ static int drm_getcap(struct drm_device *dev, void *data, struct drm_file *file_
> >       case DRM_CAP_SYNCOBJ_TIMELINE:
> >               req->value = drm_core_check_feature(dev, DRIVER_SYNCOBJ_TIMELINE);
> >               return 0;
> > +     case DRM_CAP_SYNCOBJ_DEADLINE:
> > +             req->value = drm_core_check_feature(dev, DRIVER_SYNCOBJ_TIMELINE);
>
> Hi,
>
> is that a typo for DRIVER_SYNCOBJ_DEADLINE?

Ahh, yes, that is a typo.. but I'm thinking of dropping the cap and
allowing count_handles==0 instead as a way for userspace to probe
whether the kernel supports the new ioctl flag/fields.

> > +             return 0;
> >       }
> >
> >       /* Other caps only work with KMS drivers */
> > diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c
> > index 0c2be8360525..61cf97972a60 100644
> > --- a/drivers/gpu/drm/drm_syncobj.c
> > +++ b/drivers/gpu/drm/drm_syncobj.c
> > @@ -973,7 +973,8 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs,
> >                                                 uint32_t count,
> >                                                 uint32_t flags,
> >                                                 signed long timeout,
> > -                                               uint32_t *idx)
> > +                                               uint32_t *idx,
> > +                                               ktime_t *deadline)
> >  {
> >       struct syncobj_wait_entry *entries;
> >       struct dma_fence *fence;
> > @@ -1053,6 +1054,15 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs,
> >                       drm_syncobj_fence_add_wait(syncobjs[i], &entries[i]);
> >       }
> >
> > +     if (deadline) {
> > +             for (i = 0; i < count; ++i) {
> > +                     fence = entries[i].fence;
> > +                     if (!fence)
> > +                             continue;
> > +                     dma_fence_set_deadline(fence, *deadline);
> > +             }
> > +     }
> > +
> >       do {
> >               set_current_state(TASK_INTERRUPTIBLE);
> >
> > @@ -1151,7 +1161,8 @@ static int drm_syncobj_array_wait(struct drm_device *dev,
> >                                 struct drm_file *file_private,
> >                                 struct drm_syncobj_wait *wait,
> >                                 struct drm_syncobj_timeline_wait *timeline_wait,
> > -                               struct drm_syncobj **syncobjs, bool timeline)
> > +                               struct drm_syncobj **syncobjs, bool timeline,
> > +                               ktime_t *deadline)
> >  {
> >       signed long timeout = 0;
> >       uint32_t first = ~0;
> > @@ -1162,7 +1173,8 @@ static int drm_syncobj_array_wait(struct drm_device *dev,
> >                                                        NULL,
> >                                                        wait->count_handles,
> >                                                        wait->flags,
> > -                                                      timeout, &first);
> > +                                                      timeout, &first,
> > +                                                      deadline);
> >               if (timeout < 0)
> >                       return timeout;
> >               wait->first_signaled = first;
> > @@ -1172,7 +1184,8 @@ static int drm_syncobj_array_wait(struct drm_device *dev,
> >                                                        u64_to_user_ptr(timeline_wait->points),
> >                                                        timeline_wait->count_handles,
> >                                                        timeline_wait->flags,
> > -                                                      timeout, &first);
> > +                                                      timeout, &first,
> > +                                                      deadline);
> >               if (timeout < 0)
> >                       return timeout;
> >               timeline_wait->first_signaled = first;
> > @@ -1243,13 +1256,20 @@ drm_syncobj_wait_ioctl(struct drm_device *dev, void *data,
> >  {
> >       struct drm_syncobj_wait *args = data;
> >       struct drm_syncobj **syncobjs;
> > +     unsigned possible_flags;
> > +     ktime_t t, *tp = NULL;
> >       int ret = 0;
> >
> >       if (!drm_core_check_feature(dev, DRIVER_SYNCOBJ))
> >               return -EOPNOTSUPP;
> >
> > -     if (args->flags & ~(DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL |
> > -                         DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT))
> > +     possible_flags = DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL |
> > +                      DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT;
> > +
> > +     if (drm_core_check_feature(dev, DRIVER_SYNCOBJ_DEADLINE))
> > +             possible_flags |= DRM_SYNCOBJ_WAIT_FLAGS_WAIT_DEADLINE;
> > +
> > +     if (args->flags & ~possible_flags)
> >               return -EINVAL;
> >
> >       if (args->count_handles == 0)
> > @@ -1262,8 +1282,13 @@ drm_syncobj_wait_ioctl(struct drm_device *dev, void *data,
> >       if (ret < 0)
> >               return ret;
> >
> > +     if (args->flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_DEADLINE) {
> > +             t = ktime_set(args->deadline_sec, args->deadline_nsec);
> > +             tp = &t;
> > +     }
> > +
> >       ret = drm_syncobj_array_wait(dev, file_private,
> > -                                  args, NULL, syncobjs, false);
> > +                                  args, NULL, syncobjs, false, tp);
> >
> >       drm_syncobj_array_free(syncobjs, args->count_handles);
> >
> > @@ -1276,14 +1301,21 @@ drm_syncobj_timeline_wait_ioctl(struct drm_device *dev, void *data,
> >  {
> >       struct drm_syncobj_timeline_wait *args = data;
> >       struct drm_syncobj **syncobjs;
> > +     unsigned possible_flags;
> > +     ktime_t t, *tp = NULL;
> >       int ret = 0;
> >
> >       if (!drm_core_check_feature(dev, DRIVER_SYNCOBJ_TIMELINE))
> >               return -EOPNOTSUPP;
> >
> > -     if (args->flags & ~(DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL |
> > -                         DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT |
> > -                         DRM_SYNCOBJ_WAIT_FLAGS_WAIT_AVAILABLE))
> > +     possible_flags = DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL |
> > +                      DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT |
> > +                      DRM_SYNCOBJ_WAIT_FLAGS_WAIT_AVAILABLE;
> > +
> > +     if (drm_core_check_feature(dev, DRIVER_SYNCOBJ_DEADLINE))
> > +             possible_flags |= DRM_SYNCOBJ_WAIT_FLAGS_WAIT_DEADLINE;
> > +
> > +     if (args->flags & ~possible_flags)
> >               return -EINVAL;
> >
> >       if (args->count_handles == 0)
> > @@ -1296,8 +1328,13 @@ drm_syncobj_timeline_wait_ioctl(struct drm_device *dev, void *data,
> >       if (ret < 0)
> >               return ret;
> >
> > +     if (args->flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_DEADLINE) {
> > +             t = ktime_set(args->deadline_sec, args->deadline_nsec);
> > +             tp = &t;
> > +     }
> > +
> >       ret = drm_syncobj_array_wait(dev, file_private,
> > -                                  NULL, args, syncobjs, true);
> > +                                  NULL, args, syncobjs, true, tp);
> >
> >       drm_syncobj_array_free(syncobjs, args->count_handles);
> >
> > diff --git a/include/drm/drm_drv.h b/include/drm/drm_drv.h
> > index 1d76d0686b03..9aa24f097e22 100644
> > --- a/include/drm/drm_drv.h
> > +++ b/include/drm/drm_drv.h
> > @@ -104,6 +104,12 @@ enum drm_driver_feature {
> >        * acceleration should be handled by two drivers that are connected using auxiliary bus.
> >        */
> >       DRIVER_COMPUTE_ACCEL            = BIT(7),
> > +     /**
> > +      * @DRIVER_SYNCOBJ_DEADLINE:
> > +      *
> > +      * Driver supports &dma_fence_ops.set_deadline
> > +      */
> > +     DRIVER_SYNCOBJ_DEADLINE         = BIT(8),
> >
> >       /* IMPORTANT: Below are all the legacy flags, add new ones above. */
> >
> > diff --git a/include/uapi/drm/drm.h b/include/uapi/drm/drm.h
> > index 642808520d92..c6b85bb13810 100644
> > --- a/include/uapi/drm/drm.h
> > +++ b/include/uapi/drm/drm.h
> > @@ -767,6 +767,13 @@ struct drm_gem_open {
> >   * Documentation/gpu/drm-mm.rst, section "DRM Sync Objects".
> >   */
> >  #define DRM_CAP_SYNCOBJ_TIMELINE     0x14
> > +/**
> > + * DRM_CAP_SYNCOBJ_DEADLINE
> > + *
> > + * If set to 1, the driver supports DRM_SYNCOBJ_WAIT_FLAGS_WAIT_DEADLINE flag
> > + * on the SYNCOBJ_TIMELINE_WAIT/SYNCOBJ_WAIT ioctls.
> > + */
> > +#define DRM_CAP_SYNCOBJ_DEADLINE     0x15
> >
> >  /* DRM_IOCTL_GET_CAP ioctl argument type */
> >  struct drm_get_cap {
> > @@ -887,6 +894,7 @@ struct drm_syncobj_transfer {
> >  #define DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL (1 << 0)
> >  #define DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT (1 << 1)
> >  #define DRM_SYNCOBJ_WAIT_FLAGS_WAIT_AVAILABLE (1 << 2) /* wait for time point to become available */
> > +#define DRM_SYNCOBJ_WAIT_FLAGS_WAIT_DEADLINE (1 << 3) /* set fence deadline based to deadline_nsec/sec */
>
> Where was the UAPI documentation explaining what a fence deadline is
> and what it does, again?
>
> >  struct drm_syncobj_wait {
> >       __u64 handles;
> >       /* absolute timeout */
> > @@ -894,7 +902,9 @@ struct drm_syncobj_wait {
> >       __u32 count_handles;
> >       __u32 flags;
> >       __u32 first_signaled; /* only valid when not waiting all */
> > -     __u32 pad;
> > +     /* Deadline to set on backing fence(s) in CLOCK_MONOTONIC: */
> > +     __u32 deadline_nsec;
> > +     __u64 deadline_sec;
> >  };
> >
> >  struct drm_syncobj_timeline_wait {
> > @@ -906,7 +916,9 @@ struct drm_syncobj_timeline_wait {
> >       __u32 count_handles;
> >       __u32 flags;
> >       __u32 first_signaled; /* only valid when not waiting all */
> > -     __u32 pad;
> > +     /* Deadline to set on backing fence(s) in CLOCK_MONOTONIC: */
> > +     __u32 deadline_nsec;
> > +     __u64 deadline_sec;
> >  };
>
> It seems inconsistent that these sec,nsec are here unsigned, when in
> other places they are signed. There is also the question if these need
> to meet clock_settime() requirements of valid values.
>

Yes, should have been signed.  But I think Christian has convinced me
to use 'u64 ns' (absolute monotonic) instead for the sync_file ioctl.
And it would make sense to use the same here.

BR,
-R

>
> Thanks,
> pq



[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux