On 18/02/2023 21:15, Rob Clark wrote:
> From: Rob Clark <[email protected]>
>
> Signed-off-by: Rob Clark <[email protected]>
> ---
>
> This should probably be re-written by someone who knows the i915
> request/timeline stuff better, to deal with non-immediate deadlines.
> But as-is I think this should be enough to handle the case where
> we want syncobj waits to trigger boost.

Yeah, there are endless possibilities. :) But I think it is effectively
similar enough to current waitboosting (when waits are done using the
i915 specific ioctl). So as a first step I'll try to organize some
internal power and performance testing, at least Chromebook focused, to
see if modern userspace (syncobj based) even benefits and does not by
some chance regress over the board.

Regards,

Tvrtko

>
> drivers/gpu/drm/i915/i915_driver.c | 2 +-
> drivers/gpu/drm/i915/i915_request.c | 20 ++++++++++++++++++++
> 2 files changed, 21 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_driver.c b/drivers/gpu/drm/i915/i915_driver.c
> index cf1c0970ecb4..bd40b7bcb38a 100644
> --- a/drivers/gpu/drm/i915/i915_driver.c
> +++ b/drivers/gpu/drm/i915/i915_driver.c
> @@ -1781,7 +1781,7 @@ static const struct drm_driver i915_drm_driver = {
> .driver_features =
> DRIVER_GEM |
> DRIVER_RENDER | DRIVER_MODESET | DRIVER_ATOMIC | DRIVER_SYNCOBJ |
> - DRIVER_SYNCOBJ_TIMELINE,
> + DRIVER_SYNCOBJ_TIMELINE | DRIVER_SYNCOBJ_DEADLINE,
> .release = i915_driver_release,
> .open = i915_driver_open,
> .lastclose = i915_driver_lastclose,
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index 7503dcb9043b..44491e7e214c 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -97,6 +97,25 @@ static bool i915_fence_enable_signaling(struct dma_fence *fence)
> return i915_request_enable_breadcrumb(to_request(fence));
> }
>
> +static void i915_fence_set_deadline(struct dma_fence *fence, ktime_t deadline)
> +{
> + struct i915_request *rq = to_request(fence);
> +
> + if (i915_request_completed(rq))
> + return;
> +
> + if (i915_request_started(rq))
> + return;
> +
> + /*
> + * TODO something more clever for deadlines that are in the
> + * future. I think probably track the nearest deadline in
> + * rq->timeline and set timer to trigger boost accordingly?
> + */
> +
> + intel_rps_boost(rq);
> +}
> +
> static signed long i915_fence_wait(struct dma_fence *fence,
> bool interruptible,
> signed long timeout)
> @@ -182,6 +201,7 @@ const struct dma_fence_ops i915_fence_ops = {
> .signaled = i915_fence_signaled,
> .wait = i915_fence_wait,
> .release = i915_fence_release,
> + .set_deadline = i915_fence_set_deadline,
> };
>
> static void irq_execute_cb(struct irq_work *wrk)

2023-02-20 15:55:56

by Rob Clark

[permalink] [raw]

Subject: Re: [PATCH v4 10/14] drm/vblank: Add helper to get next vblank time

On Mon, Feb 20, 2023 at 1:08 AM Pekka Paalanen <[email protected]> wrote:
>
> On Sat, 18 Feb 2023 13:15:53 -0800
> Rob Clark <[email protected]> wrote:
>
> > From: Rob Clark <[email protected]>
> >
> > Will be used in the next commit to set a deadline on fences that an
> > atomic update is waiting on.
> >
> > Signed-off-by: Rob Clark <[email protected]>
> > ---
> > drivers/gpu/drm/drm_vblank.c | 32 ++++++++++++++++++++++++++++++++
> > include/drm/drm_vblank.h | 1 +
> > 2 files changed, 33 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
> > index 2ff31717a3de..caf25ebb34c5 100644
> > --- a/drivers/gpu/drm/drm_vblank.c
> > +++ b/drivers/gpu/drm/drm_vblank.c
> > @@ -980,6 +980,38 @@ u64 drm_crtc_vblank_count_and_time(struct drm_crtc *crtc,
> > }
> > EXPORT_SYMBOL(drm_crtc_vblank_count_and_time);
> >
> > +/**
> > + * drm_crtc_next_vblank_time - calculate the time of the next vblank
> > + * @crtc: the crtc for which to calculate next vblank time
> > + * @vblanktime: pointer to time to receive the next vblank timestamp.
> > + *
> > + * Calculate the expected time of the next vblank based on time of previous
> > + * vblank and frame duration
>
> Hi,
>
> for VRR this targets the highest frame rate possible for the current
> VRR mode, right?
>

It is based on vblank->framedur_ns which is in turn based on
mode->crtc_clock. Presumably for VRR that ends up being a maximum?

BR,
-R

>
> Thanks,
> pq
>
> > + */
> > +int drm_crtc_next_vblank_time(struct drm_crtc *crtc, ktime_t *vblanktime)
> > +{
> > + unsigned int pipe = drm_crtc_index(crtc);
> > + struct drm_vblank_crtc *vblank = &crtc->dev->vblank[pipe];
> > + u64 count;
> > +
> > + if (!vblank->framedur_ns)
> > + return -EINVAL;
> > +
> > + count = drm_vblank_count_and_time(crtc->dev, pipe, vblanktime);
> > +
> > + /*
> > + * If we don't get a valid count, then we probably also don't
> > + * have a valid time:
> > + */
> > + if (!count)
> > + return -EINVAL;
> > +
> > + *vblanktime = ktime_add(*vblanktime, ns_to_ktime(vblank->framedur_ns));
> > +
> > + return 0;
> > +}
> > +EXPORT_SYMBOL(drm_crtc_next_vblank_time);
> > +
> > static void send_vblank_event(struct drm_device *dev,
> > struct drm_pending_vblank_event *e,
> > u64 seq, ktime_t now)
> > diff --git a/include/drm/drm_vblank.h b/include/drm/drm_vblank.h
> > index 733a3e2d1d10..a63bc2c92f3c 100644
> > --- a/include/drm/drm_vblank.h
> > +++ b/include/drm/drm_vblank.h
> > @@ -230,6 +230,7 @@ bool drm_dev_has_vblank(const struct drm_device *dev);
> > u64 drm_crtc_vblank_count(struct drm_crtc *crtc);
> > u64 drm_crtc_vblank_count_and_time(struct drm_crtc *crtc,
> > ktime_t *vblanktime);
> > +int drm_crtc_next_vblank_time(struct drm_crtc *crtc, ktime_t *vblanktime);
> > void drm_crtc_send_vblank_event(struct drm_crtc *crtc,
> > struct drm_pending_vblank_event *e);
> > void drm_crtc_arm_vblank_event(struct drm_crtc *crtc,
>

2023-02-20 16:09:23

by Rob Clark

[permalink] [raw]

Subject: Re: [PATCH v4 05/14] dma-buf/sync_file: Add SET_DEADLINE ioctl

On Mon, Feb 20, 2023 at 12:27 AM Christian König
<[email protected]> wrote:
>
> Am 18.02.23 um 22:15 schrieb Rob Clark:
> > From: Rob Clark <[email protected]>
> >
> > The initial purpose is for igt tests, but this would also be useful for
> > compositors that wait until close to vblank deadline to make decisions
> > about which frame to show.
> >
> > The igt tests can be found at:
> >
> > https://gitlab.freedesktop.org/robclark/igt-gpu-tools/-/commits/fence-deadline
> >
> > v2: Clarify the timebase, add link to igt tests
> >
> > Signed-off-by: Rob Clark <[email protected]>
> > ---
> > drivers/dma-buf/sync_file.c | 19 +++++++++++++++++++
> > include/uapi/linux/sync_file.h | 22 ++++++++++++++++++++++
> > 2 files changed, 41 insertions(+)
> >
> > diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c
> > index af57799c86ce..fb6ca1032885 100644
> > --- a/drivers/dma-buf/sync_file.c
> > +++ b/drivers/dma-buf/sync_file.c
> > @@ -350,6 +350,22 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> > return ret;
> > }
> >
> > +static int sync_file_ioctl_set_deadline(struct sync_file *sync_file,
> > + unsigned long arg)
> > +{
> > + struct sync_set_deadline ts;
> > +
> > + if (copy_from_user(&ts, (void __user *)arg, sizeof(ts)))
> > + return -EFAULT;
> > +
> > + if (ts.pad)
> > + return -EINVAL;
> > +
> > + dma_fence_set_deadline(sync_file->fence, ktime_set(ts.tv_sec, ts.tv_nsec));
> > +
> > + return 0;
> > +}
> > +
> > static long sync_file_ioctl(struct file *file, unsigned int cmd,
> > unsigned long arg)
> > {
> > @@ -362,6 +378,9 @@ static long sync_file_ioctl(struct file *file, unsigned int cmd,
> > case SYNC_IOC_FILE_INFO:
> > return sync_file_ioctl_fence_info(sync_file, arg);
> >
> > + case SYNC_IOC_SET_DEADLINE:
> > + return sync_file_ioctl_set_deadline(sync_file, arg);
> > +
> > default:
> > return -ENOTTY;
> > }
> > diff --git a/include/uapi/linux/sync_file.h b/include/uapi/linux/sync_file.h
> > index ee2dcfb3d660..c8666580816f 100644
> > --- a/include/uapi/linux/sync_file.h
> > +++ b/include/uapi/linux/sync_file.h
> > @@ -67,6 +67,20 @@ struct sync_file_info {
> > __u64 sync_fence_info;
> > };
> >
> > +/**
> > + * struct sync_set_deadline - set a deadline on a fence
> > + * @tv_sec: seconds elapsed since epoch
> > + * @tv_nsec: nanoseconds elapsed since the time given by the tv_sec
> > + * @pad: must be zero
> > + *
> > + * The timebase for the deadline is CLOCK_MONOTONIC (same as vblank)
> > + */
> > +struct sync_set_deadline {
> > + __s64 tv_sec;
> > + __s32 tv_nsec;
> > + __u32 pad;
>
> IIRC struct timespec defined this as time_t/long (which is horrible for
> an UAPI because of the sizeof(long) dependency), one possible
> alternative is to use 64bit nanoseconds from CLOCK_MONOTONIC (which is
> essentially ktime).
>
> Not 100% sure if there is any preferences documented, but I think the
> later might be better.

The original thought is that this maps directly to clock_gettime()
without extra conversion needed, and is similar to other pre-ktime_t
UAPI. But OTOH if userspace wants to add an offset, it is maybe
better to convert completely to ns in userspace and use a u64 (as that
is what ns_to_ktime() uses).. (and OFC whatever decision here also
applies to the syncobj wait ioctls)

I'm leaning towards u64 CLOCK_MONOTONIC ns if no one has a good
argument against that.

BR,
-R

> Either way the patch is Acked-by: Christian König
> <[email protected]> for this patch.
>
> Regards,
> Christian.
>
> > +};
> > +
> > #define SYNC_IOC_MAGIC '>'
> >
> > /**
> > @@ -95,4 +109,12 @@ struct sync_file_info {
> > */
> > #define SYNC_IOC_FILE_INFO _IOWR(SYNC_IOC_MAGIC, 4, struct sync_file_info)
> >
> > +
> > +/**
> > + * DOC: SYNC_IOC_SET_DEADLINE - set a deadline on a fence
> > + *
> > + * Allows userspace to set a deadline on a fence, see dma_fence_set_deadline()
> > + */
> > +#define SYNC_IOC_SET_DEADLINE _IOW(SYNC_IOC_MAGIC, 5, struct sync_set_deadline)
> > +
> > #endif /* _UAPI_LINUX_SYNC_H */
>

2023-02-20 16:15:03

by Rob Clark

[permalink] [raw]

Subject: Re: [PATCH v4 06/14] dma-buf/sync_file: Support (E)POLLPRI

On Mon, Feb 20, 2023 at 12:53 AM Pekka Paalanen <[email protected]> wrote:
>
> On Sat, 18 Feb 2023 13:15:49 -0800
> Rob Clark <[email protected]> wrote:
>
> > From: Rob Clark <[email protected]>
> >
> > Allow userspace to use the EPOLLPRI/POLLPRI flag to indicate an urgent
> > wait (as opposed to a "housekeeping" wait to know when to cleanup after
> > some work has completed). Usermode components of GPU driver stacks
> > often poll() on fence fd's to know when it is safe to do things like
> > free or reuse a buffer, but they can also poll() on a fence fd when
> > waiting to read back results from the GPU. The EPOLLPRI/POLLPRI flag
> > lets the kernel differentiate these two cases.
> >
> > Signed-off-by: Rob Clark <[email protected]>
>
> Hi,
>
> where would the UAPI documentation of this go?
> It seems to be missing.

Good question, I am not sure. The poll() man page has a description,
but my usage doesn't fit that _exactly_ (but OTOH the description is a
bit vague).

> If a Wayland compositor is polling application fences to know which
> client buffer to use in its rendering, should the compositor poll with
> PRI or not? If a compositor polls with PRI, then all fences from all
> applications would always be PRI. Would that be harmful somehow or
> would it be beneficial?

I think a compositor would rather use the deadline ioctl and then poll
without PRI. Otherwise you are giving an urgency signal to the fence
signaller which might not necessarily be needed.

The places where I expect PRI to be useful is more in mesa (things
like glFinish(), readpix, and other similar sorts of blocking APIs)

BR,
-R

>
>
> Thanks,
> pq
>
> > ---
> > drivers/dma-buf/sync_file.c | 8 ++++++++
> > 1 file changed, 8 insertions(+)
> >
> > diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c
> > index fb6ca1032885..c30b2085ee0a 100644
> > --- a/drivers/dma-buf/sync_file.c
> > +++ b/drivers/dma-buf/sync_file.c
> > @@ -192,6 +192,14 @@ static __poll_t sync_file_poll(struct file *file, poll_table *wait)
> > {
> > struct sync_file *sync_file = file->private_data;
> >
> > + /*
> > + * The POLLPRI/EPOLLPRI flag can be used to signal that
> > + * userspace wants the fence to signal ASAP, express this
> > + * as an immediate deadline.
> > + */
> > + if (poll_requested_events(wait) & EPOLLPRI)
> > + dma_fence_set_deadline(sync_file->fence, ktime_get());
> > +
> > poll_wait(file, &sync_file->wq, wait);
> >
> > if (list_empty(&sync_file->cb.node) &&
>

2023-02-20 16:20:20

by Rob Clark

On Tue, Feb 21, 2023 at 10:45:51AM +0200, Pekka Paalanen wrote:
> On Mon, 20 Feb 2023 07:55:41 -0800
> Rob Clark <[email protected]> wrote:
>
> > On Mon, Feb 20, 2023 at 1:08 AM Pekka Paalanen <[email protected]> wrote:
> > >
> > > On Sat, 18 Feb 2023 13:15:53 -0800
> > > Rob Clark <[email protected]> wrote:
> > >
> > > > From: Rob Clark <[email protected]>
> > > >
> > > > Will be used in the next commit to set a deadline on fences that an
> > > > atomic update is waiting on.
> > > >
> > > > Signed-off-by: Rob Clark <[email protected]>
> > > > ---
> > > > drivers/gpu/drm/drm_vblank.c | 32 ++++++++++++++++++++++++++++++++
> > > > include/drm/drm_vblank.h | 1 +
> > > > 2 files changed, 33 insertions(+)
> > > >
> > > > diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
> > > > index 2ff31717a3de..caf25ebb34c5 100644
> > > > --- a/drivers/gpu/drm/drm_vblank.c
> > > > +++ b/drivers/gpu/drm/drm_vblank.c
> > > > @@ -980,6 +980,38 @@ u64 drm_crtc_vblank_count_and_time(struct drm_crtc *crtc,
> > > > }
> > > > EXPORT_SYMBOL(drm_crtc_vblank_count_and_time);
> > > >
> > > > +/**
> > > > + * drm_crtc_next_vblank_time - calculate the time of the next vblank
> > > > + * @crtc: the crtc for which to calculate next vblank time
> > > > + * @vblanktime: pointer to time to receive the next vblank timestamp.
> > > > + *
> > > > + * Calculate the expected time of the next vblank based on time of previous
> > > > + * vblank and frame duration
> > >
> > > Hi,
> > >
> > > for VRR this targets the highest frame rate possible for the current
> > > VRR mode, right?
> > >
> >
> > It is based on vblank->framedur_ns which is in turn based on
> > mode->crtc_clock. Presumably for VRR that ends up being a maximum?
>
> I don't know. :-)

At least for i915 this will give you the maximum frame
duration.

Also this does not calculate the the start of vblank, it
calculates the start of active video.

>
> You need a number of clock cycles in addition to the clock frequency,
> and that could still be minimum, maximum, the last realized one, ...
>
> VRR works by adjusting the front porch length IIRC.
>
>
> Thanks,
> pq
>
> > BR,
> > -R
> >
> >
> > >
> > > Thanks,
> > > pq
> > >
> > > > + */
> > > > +int drm_crtc_next_vblank_time(struct drm_crtc *crtc, ktime_t *vblanktime)
> > > > +{
> > > > + unsigned int pipe = drm_crtc_index(crtc);
> > > > + struct drm_vblank_crtc *vblank = &crtc->dev->vblank[pipe];
> > > > + u64 count;
> > > > +
> > > > + if (!vblank->framedur_ns)
> > > > + return -EINVAL;
> > > > +
> > > > + count = drm_vblank_count_and_time(crtc->dev, pipe, vblanktime);
> > > > +
> > > > + /*
> > > > + * If we don't get a valid count, then we probably also don't
> > > > + * have a valid time:
> > > > + */
> > > > + if (!count)
> > > > + return -EINVAL;
> > > > +
> > > > + *vblanktime = ktime_add(*vblanktime, ns_to_ktime(vblank->framedur_ns));
> > > > +
> > > > + return 0;
> > > > +}
> > > > +EXPORT_SYMBOL(drm_crtc_next_vblank_time);
> > > > +
> > > > static void send_vblank_event(struct drm_device *dev,
> > > > struct drm_pending_vblank_event *e,
> > > > u64 seq, ktime_t now)
> > > > diff --git a/include/drm/drm_vblank.h b/include/drm/drm_vblank.h
> > > > index 733a3e2d1d10..a63bc2c92f3c 100644
> > > > --- a/include/drm/drm_vblank.h
> > > > +++ b/include/drm/drm_vblank.h
> > > > @@ -230,6 +230,7 @@ bool drm_dev_has_vblank(const struct drm_device *dev);
> > > > u64 drm_crtc_vblank_count(struct drm_crtc *crtc);
> > > > u64 drm_crtc_vblank_count_and_time(struct drm_crtc *crtc,
> > > > ktime_t *vblanktime);
> > > > +int drm_crtc_next_vblank_time(struct drm_crtc *crtc, ktime_t *vblanktime);
> > > > void drm_crtc_send_vblank_event(struct drm_crtc *crtc,
> > > > struct drm_pending_vblank_event *e);
> > > > void drm_crtc_arm_vblank_event(struct drm_crtc *crtc,
> > >
>

--
Ville Syrj?l?
Intel

2023-02-21 13:11:51

On Tue, 21 Feb 2023 09:53:56 -0800
Rob Clark <[email protected]> wrote:

> On Tue, Feb 21, 2023 at 8:48 AM Luben Tuikov <[email protected]> wrote:
> >
> > On 2023-02-20 11:14, Rob Clark wrote:
> > > On Mon, Feb 20, 2023 at 12:53 AM Pekka Paalanen <[email protected]> wrote:
> > >>
> > >> On Sat, 18 Feb 2023 13:15:49 -0800
> > >> Rob Clark <[email protected]> wrote:
> > >>
> > >>> From: Rob Clark <[email protected]>
> > >>>
> > >>> Allow userspace to use the EPOLLPRI/POLLPRI flag to indicate an urgent
> > >>> wait (as opposed to a "housekeeping" wait to know when to cleanup after
> > >>> some work has completed). Usermode components of GPU driver stacks
> > >>> often poll() on fence fd's to know when it is safe to do things like
> > >>> free or reuse a buffer, but they can also poll() on a fence fd when
> > >>> waiting to read back results from the GPU. The EPOLLPRI/POLLPRI flag
> > >>> lets the kernel differentiate these two cases.
> > >>>
> > >>> Signed-off-by: Rob Clark <[email protected]>
> > >>
> > >> Hi,
> > >>
> > >> where would the UAPI documentation of this go?
> > >> It seems to be missing.
> > >
> > > Good question, I am not sure. The poll() man page has a description,
> > > but my usage doesn't fit that _exactly_ (but OTOH the description is a
> > > bit vague).
> > >
> > >> If a Wayland compositor is polling application fences to know which
> > >> client buffer to use in its rendering, should the compositor poll with
> > >> PRI or not? If a compositor polls with PRI, then all fences from all
> > >> applications would always be PRI. Would that be harmful somehow or
> > >> would it be beneficial?
> > >
> > > I think a compositor would rather use the deadline ioctl and then poll
> > > without PRI. Otherwise you are giving an urgency signal to the fence
> > > signaller which might not necessarily be needed.
> > >
> > > The places where I expect PRI to be useful is more in mesa (things
> > > like glFinish(), readpix, and other similar sorts of blocking APIs)
> > Hi,
> >
> > Hmm, but then user-space could do the opposite, namely, submit work as usual--never
> > using the SET_DEADLINE ioctl, and then at the end, poll using (E)POLLPRI. That seems
> > like a possible usage pattern, unintended--maybe, but possible. Do we want to discourage
> > this? Wouldn't SET_DEADLINE be enough? I mean, one can call SET_DEADLINE with the current
> > time, and then wouldn't that be equivalent to (E)POLLPRI?
>
> Yeah, (E)POLLPRI isn't strictly needed if we have SET_DEADLINE. It is
> slightly more convenient if you want an immediate deadline (single
> syscall instead of two), but not strictly needed. OTOH it piggy-backs
> on existing UABI.

In that case, I would be conservative, and not add the POLLPRI
semantics. An UAPI addition that is not strictly needed and somewhat
unclear if it violates any design principles is best not done, until it
is proven to be beneficial.

Besides, a Wayland compositor does not necessary need to add the fd
to its main event loop for poll. It could just SET_DEADLINE, and then
when it renders simply check if the fence passed or not already. Not
polling means the compositor does not need to wake up at the moment the
fence signals to just record a flag.

On another matter, if the application uses SET_DEADLINE with one
timestamp, and the compositor uses SET_DEADLINE on the same thing with
another timestamp, what should happen?

Maybe it's a soft-realtime app whose primary goal is not display, and
it needs the result faster than the window server?

Maybe SET_DEADLINE should set the deadline only to an earlier timestamp
and never later?

Thanks,
pq

Attachments:

(No filename) (833.00 B)
OpenPGP digital signature

2023-02-22 09:57:45

On Wed, 22 Feb 2023 07:37:26 -0800
Rob Clark <[email protected]> wrote:

> On Wed, Feb 22, 2023 at 1:49 AM Pekka Paalanen <[email protected]> wrote:
> >
> > On Tue, 21 Feb 2023 09:53:56 -0800
> > Rob Clark <[email protected]> wrote:
> >
> > > On Tue, Feb 21, 2023 at 8:48 AM Luben Tuikov <[email protected]> wrote:
> > > >
> > > > On 2023-02-20 11:14, Rob Clark wrote:
> > > > > On Mon, Feb 20, 2023 at 12:53 AM Pekka Paalanen <[email protected]> wrote:
> > > > >>
> > > > >> On Sat, 18 Feb 2023 13:15:49 -0800
> > > > >> Rob Clark <[email protected]> wrote:
> > > > >>
> > > > >>> From: Rob Clark <[email protected]>
> > > > >>>
> > > > >>> Allow userspace to use the EPOLLPRI/POLLPRI flag to indicate an urgent
> > > > >>> wait (as opposed to a "housekeeping" wait to know when to cleanup after
> > > > >>> some work has completed). Usermode components of GPU driver stacks
> > > > >>> often poll() on fence fd's to know when it is safe to do things like
> > > > >>> free or reuse a buffer, but they can also poll() on a fence fd when
> > > > >>> waiting to read back results from the GPU. The EPOLLPRI/POLLPRI flag
> > > > >>> lets the kernel differentiate these two cases.
> > > > >>>
> > > > >>> Signed-off-by: Rob Clark <[email protected]>
> > > > >>
> > > > >> Hi,
> > > > >>
> > > > >> where would the UAPI documentation of this go?
> > > > >> It seems to be missing.
> > > > >
> > > > > Good question, I am not sure. The poll() man page has a description,
> > > > > but my usage doesn't fit that _exactly_ (but OTOH the description is a
> > > > > bit vague).
> > > > >
> > > > >> If a Wayland compositor is polling application fences to know which
> > > > >> client buffer to use in its rendering, should the compositor poll with
> > > > >> PRI or not? If a compositor polls with PRI, then all fences from all
> > > > >> applications would always be PRI. Would that be harmful somehow or
> > > > >> would it be beneficial?
> > > > >
> > > > > I think a compositor would rather use the deadline ioctl and then poll
> > > > > without PRI. Otherwise you are giving an urgency signal to the fence
> > > > > signaller which might not necessarily be needed.
> > > > >
> > > > > The places where I expect PRI to be useful is more in mesa (things
> > > > > like glFinish(), readpix, and other similar sorts of blocking APIs)
> > > > Hi,
> > > >
> > > > Hmm, but then user-space could do the opposite, namely, submit work as usual--never
> > > > using the SET_DEADLINE ioctl, and then at the end, poll using (E)POLLPRI. That seems
> > > > like a possible usage pattern, unintended--maybe, but possible. Do we want to discourage
> > > > this? Wouldn't SET_DEADLINE be enough? I mean, one can call SET_DEADLINE with the current
> > > > time, and then wouldn't that be equivalent to (E)POLLPRI?
> > >
> > > Yeah, (E)POLLPRI isn't strictly needed if we have SET_DEADLINE. It is
> > > slightly more convenient if you want an immediate deadline (single
> > > syscall instead of two), but not strictly needed. OTOH it piggy-backs
> > > on existing UABI.
> >
> > In that case, I would be conservative, and not add the POLLPRI
> > semantics. An UAPI addition that is not strictly needed and somewhat
> > unclear if it violates any design principles is best not done, until it
> > is proven to be beneficial.
> >
> > Besides, a Wayland compositor does not necessary need to add the fd
> > to its main event loop for poll. It could just SET_DEADLINE, and then
> > when it renders simply check if the fence passed or not already. Not
> > polling means the compositor does not need to wake up at the moment the
> > fence signals to just record a flag.
>
> poll(POLLPRI) isn't intended for wayland.. but is a thing I want in
> mesa for fence waits. I _could_ use SET_DEADLINE but it is two
> syscalls and correspondingly more code ;-)

But is it actually beneficial? "More code" seems quite irrelevant.

Would there be a hundred or more of those per frame? Or would it be
always limited to one or two? Or totally depend on what the application
is doing? Is it a significant impact?

> > On another matter, if the application uses SET_DEADLINE with one
> > timestamp, and the compositor uses SET_DEADLINE on the same thing with
> > another timestamp, what should happen?
>
> The expectation is that many deadline hints can be set on a fence.
> The fence signaller should track the soonest deadline.

You need to document that as UAPI, since it is observable to userspace.
It would be bad if drivers or subsystems would differ in behaviour.

Thanks,
pq

Attachments:

(No filename) (833.00 B)
OpenPGP digital signature

2023-02-23 18:52:06

by Rob Clark

[permalink] [raw]

Subject: Re: [PATCH v4 06/14] dma-buf/sync_file: Support (E)POLLPRI

On Thu, Feb 23, 2023 at 1:38 AM Pekka Paalanen <[email protected]> wrote:
>
> On Wed, 22 Feb 2023 07:37:26 -0800
> Rob Clark <[email protected]> wrote:
>
> > On Wed, Feb 22, 2023 at 1:49 AM Pekka Paalanen <[email protected]> wrote:
> > >
> > > On Tue, 21 Feb 2023 09:53:56 -0800
> > > Rob Clark <[email protected]> wrote:
> > >
> > > > On Tue, Feb 21, 2023 at 8:48 AM Luben Tuikov <[email protected]> wrote:
> > > > >
> > > > > On 2023-02-20 11:14, Rob Clark wrote:
> > > > > > On Mon, Feb 20, 2023 at 12:53 AM Pekka Paalanen <[email protected]> wrote:
> > > > > >>
> > > > > >> On Sat, 18 Feb 2023 13:15:49 -0800
> > > > > >> Rob Clark <[email protected]> wrote:
> > > > > >>
> > > > > >>> From: Rob Clark <[email protected]>
> > > > > >>>
> > > > > >>> Allow userspace to use the EPOLLPRI/POLLPRI flag to indicate an urgent
> > > > > >>> wait (as opposed to a "housekeeping" wait to know when to cleanup after
> > > > > >>> some work has completed). Usermode components of GPU driver stacks
> > > > > >>> often poll() on fence fd's to know when it is safe to do things like
> > > > > >>> free or reuse a buffer, but they can also poll() on a fence fd when
> > > > > >>> waiting to read back results from the GPU. The EPOLLPRI/POLLPRI flag
> > > > > >>> lets the kernel differentiate these two cases.
> > > > > >>>
> > > > > >>> Signed-off-by: Rob Clark <[email protected]>
> > > > > >>
> > > > > >> Hi,
> > > > > >>
> > > > > >> where would the UAPI documentation of this go?
> > > > > >> It seems to be missing.
> > > > > >
> > > > > > Good question, I am not sure. The poll() man page has a description,
> > > > > > but my usage doesn't fit that _exactly_ (but OTOH the description is a
> > > > > > bit vague).
> > > > > >
> > > > > >> If a Wayland compositor is polling application fences to know which
> > > > > >> client buffer to use in its rendering, should the compositor poll with
> > > > > >> PRI or not? If a compositor polls with PRI, then all fences from all
> > > > > >> applications would always be PRI. Would that be harmful somehow or
> > > > > >> would it be beneficial?
> > > > > >
> > > > > > I think a compositor would rather use the deadline ioctl and then poll
> > > > > > without PRI. Otherwise you are giving an urgency signal to the fence
> > > > > > signaller which might not necessarily be needed.
> > > > > >
> > > > > > The places where I expect PRI to be useful is more in mesa (things
> > > > > > like glFinish(), readpix, and other similar sorts of blocking APIs)
> > > > > Hi,
> > > > >
> > > > > Hmm, but then user-space could do the opposite, namely, submit work as usual--never
> > > > > using the SET_DEADLINE ioctl, and then at the end, poll using (E)POLLPRI. That seems
> > > > > like a possible usage pattern, unintended--maybe, but possible. Do we want to discourage
> > > > > this? Wouldn't SET_DEADLINE be enough? I mean, one can call SET_DEADLINE with the current
> > > > > time, and then wouldn't that be equivalent to (E)POLLPRI?
> > > >
> > > > Yeah, (E)POLLPRI isn't strictly needed if we have SET_DEADLINE. It is
> > > > slightly more convenient if you want an immediate deadline (single
> > > > syscall instead of two), but not strictly needed. OTOH it piggy-backs
> > > > on existing UABI.
> > >
> > > In that case, I would be conservative, and not add the POLLPRI
> > > semantics. An UAPI addition that is not strictly needed and somewhat
> > > unclear if it violates any design principles is best not done, until it
> > > is proven to be beneficial.
> > >
> > > Besides, a Wayland compositor does not necessary need to add the fd
> > > to its main event loop for poll. It could just SET_DEADLINE, and then
> > > when it renders simply check if the fence passed or not already. Not
> > > polling means the compositor does not need to wake up at the moment the
> > > fence signals to just record a flag.
> >
> > poll(POLLPRI) isn't intended for wayland.. but is a thing I want in
> > mesa for fence waits. I _could_ use SET_DEADLINE but it is two
> > syscalls and correspondingly more code ;-)
>
> But is it actually beneficial? "More code" seems quite irrelevant.
>
> Would there be a hundred or more of those per frame? Or would it be
> always limited to one or two? Or totally depend on what the application
> is doing? Is it a significant impact?

In general, any time the CPU is waiting on the GPU, you have already
lost. So I don't think the extra syscall is too much of a problem.
Just less convenient.

> > > On another matter, if the application uses SET_DEADLINE with one
> > > timestamp, and the compositor uses SET_DEADLINE on the same thing with
> > > another timestamp, what should happen?
> >
> > The expectation is that many deadline hints can be set on a fence.
> > The fence signaller should track the soonest deadline.
>
> You need to document that as UAPI, since it is observable to userspace.
> It would be bad if drivers or subsystems would differ in behaviour.
>

It is in the end a hint. It is about giving the driver more
information so that it can make better choices. But the driver is
even free to ignore it. So maybe "expectation" is too strong of a
word. Rather, any other behavior doesn't really make sense. But it
could end up being dictated by how the hw and/or fw works.

BR,
-R

>
> Thanks,
> pq

2023-02-24 09:27:00

On Fri, 24 Feb 2023 09:41:46 +0000
Tvrtko Ursulin <[email protected]> wrote:

> On 24/02/2023 09:26, Pekka Paalanen wrote:
> > On Thu, 23 Feb 2023 10:51:48 -0800
> > Rob Clark <[email protected]> wrote:
> >
> >> On Thu, Feb 23, 2023 at 1:38 AM Pekka Paalanen <[email protected]> wrote:
> >>>
> >>> On Wed, 22 Feb 2023 07:37:26 -0800
> >>> Rob Clark <[email protected]> wrote:
> >>>
> >>>> On Wed, Feb 22, 2023 at 1:49 AM Pekka Paalanen <[email protected]> wrote:
> >
> > ...
> >
> >>>>> On another matter, if the application uses SET_DEADLINE with one
> >>>>> timestamp, and the compositor uses SET_DEADLINE on the same thing with
> >>>>> another timestamp, what should happen?
> >>>>
> >>>> The expectation is that many deadline hints can be set on a fence.
> >>>> The fence signaller should track the soonest deadline.
> >>>
> >>> You need to document that as UAPI, since it is observable to userspace.
> >>> It would be bad if drivers or subsystems would differ in behaviour.
> >>>
> >>
> >> It is in the end a hint. It is about giving the driver more
> >> information so that it can make better choices. But the driver is
> >> even free to ignore it. So maybe "expectation" is too strong of a
> >> word. Rather, any other behavior doesn't really make sense. But it
> >> could end up being dictated by how the hw and/or fw works.
> >
> > It will stop being a hint once it has been implemented and used in the
> > wild long enough. The kernel userspace regression rules make sure of
> > that.
>
> Yeah, tricky and maybe a gray area in this case. I think we eluded
> elsewhere in the thread that renaming the thing might be an option.
>
> So maybe instead of deadline, which is a very strong word, use something
> along the lines of "present time hint", or "signalled time hint"? Maybe
> reads clumsy. Just throwing some ideas for a start.

You can try, but I fear that if it ever changes behaviour and
someone notices that, it's labelled as a kernel regression. I don't
think documentation has ever been the authoritative definition of UABI
in Linux, it just guides drivers and userspace towards a common
understanding and common usage patterns.

So even if the UABI contract is not documented (ugh), you need to be
prepared to set the UABI contract through kernel implementation.

If you do not document the UABI contract, then different drivers are
likely to implement it differently, leading to differing behaviour.
Also userspace will invent wild ways to abuse the UABI if there is no
documentation guiding it on proper use. If userspace or end users
observe different behaviour, that's bad even if it's not a regression.

I don't like the situation either, but it is what it is. UABI stability
trumps everything regardless of whether it was documented or not.

I bet userspace is going to use this as a "make it faster, make it
hotter" button. I would not be surprised if someone wrote a LD_PRELOAD
library that stamps any and all fences with an expired deadline to
just squeeze out a little more through some weird side-effect.

Well, that's hopefully overboard in scaring, but in the end, I would
like to see UABI documented so I can have a feeling of what it is for
and how it was intended to be used. That's all.

Thanks,
pq

Attachments:

(No filename) (833.00 B)
OpenPGP digital signature

2023-02-24 10:51:04

by Tvrtko Ursulin

[permalink] [raw]

Subject: Re: [PATCH v4 06/14] dma-buf/sync_file: Support (E)POLLPRI

On 24/02/2023 10:24, Pekka Paalanen wrote:
> On Fri, 24 Feb 2023 09:41:46 +0000
> Tvrtko Ursulin <[email protected]> wrote:
>
>> On 24/02/2023 09:26, Pekka Paalanen wrote:
>>> On Thu, 23 Feb 2023 10:51:48 -0800
>>> Rob Clark <[email protected]> wrote:
>>>
>>>> On Thu, Feb 23, 2023 at 1:38 AM Pekka Paalanen <[email protected]> wrote:
>>>>>
>>>>> On Wed, 22 Feb 2023 07:37:26 -0800
>>>>> Rob Clark <[email protected]> wrote:
>>>>>
>>>>>> On Wed, Feb 22, 2023 at 1:49 AM Pekka Paalanen <[email protected]> wrote:
>>>
>>> ...
>>>
>>>>>>> On another matter, if the application uses SET_DEADLINE with one
>>>>>>> timestamp, and the compositor uses SET_DEADLINE on the same thing with
>>>>>>> another timestamp, what should happen?
>>>>>>
>>>>>> The expectation is that many deadline hints can be set on a fence.
>>>>>> The fence signaller should track the soonest deadline.
>>>>>
>>>>> You need to document that as UAPI, since it is observable to userspace.
>>>>> It would be bad if drivers or subsystems would differ in behaviour.
>>>>>
>>>>
>>>> It is in the end a hint. It is about giving the driver more
>>>> information so that it can make better choices. But the driver is
>>>> even free to ignore it. So maybe "expectation" is too strong of a
>>>> word. Rather, any other behavior doesn't really make sense. But it
>>>> could end up being dictated by how the hw and/or fw works.
>>>
>>> It will stop being a hint once it has been implemented and used in the
>>> wild long enough. The kernel userspace regression rules make sure of
>>> that.
>>
>> Yeah, tricky and maybe a gray area in this case. I think we eluded
>> elsewhere in the thread that renaming the thing might be an option.
>>
>> So maybe instead of deadline, which is a very strong word, use something
>> along the lines of "present time hint", or "signalled time hint"? Maybe
>> reads clumsy. Just throwing some ideas for a start.
>
> You can try, but I fear that if it ever changes behaviour and
> someone notices that, it's labelled as a kernel regression. I don't
> think documentation has ever been the authoritative definition of UABI
> in Linux, it just guides drivers and userspace towards a common
> understanding and common usage patterns.
>
> So even if the UABI contract is not documented (ugh), you need to be
> prepared to set the UABI contract through kernel implementation.

To be the devil's advocate it probably wouldn't be an ABI regression but
just an regression. Same way as what nice(2) priorities mean hasn't
always been the same over the years, I don't think there is a strict
contract.

Having said that, it may be different with latency sensitive stuff such
as UIs though since it is very observable and can be very painful to users.

> If you do not document the UABI contract, then different drivers are
> likely to implement it differently, leading to differing behaviour.
> Also userspace will invent wild ways to abuse the UABI if there is no
> documentation guiding it on proper use. If userspace or end users
> observe different behaviour, that's bad even if it's not a regression.
>
> I don't like the situation either, but it is what it is. UABI stability
> trumps everything regardless of whether it was documented or not.
>
> I bet userspace is going to use this as a "make it faster, make it
> hotter" button. I would not be surprised if someone wrote a LD_PRELOAD
> library that stamps any and all fences with an expired deadline to
> just squeeze out a little more through some weird side-effect.
>
> Well, that's hopefully overboard in scaring, but in the end, I would
> like to see UABI documented so I can have a feeling of what it is for
> and how it was intended to be used. That's all.

We share the same concern. If you read elsewhere in these threads you
will notice I have been calling this an "arms race". If the ability to
make yourself go faster does not required additional privilege I also
worry everyone will do it at which point it becomes pointless. So yes, I
do share this concern about exposing any of this as an unprivileged uapi.

Is it possible to limit access to only compositors in some sane way?
Sounds tricky when dma-fence should be disconnected from DRM..

Regards,

Tvrtko

2023-02-24 11:03:17

On Mon, Feb 27, 2023 at 02:20:04PM -0800, Rob Clark wrote:
> On Mon, Feb 27, 2023 at 1:36 PM Rodrigo Vivi <[email protected]> wrote:
> >
> > On Fri, Feb 24, 2023 at 09:59:57AM -0800, Rob Clark wrote:
> > > On Fri, Feb 24, 2023 at 7:27 AM Luben Tuikov <[email protected]> wrote:
> > > >
> > > > On 2023-02-24 06:37, Tvrtko Ursulin wrote:
> > > > >
> > > > > On 24/02/2023 11:00, Pekka Paalanen wrote:
> > > > >> On Fri, 24 Feb 2023 10:50:51 +0000
> > > > >> Tvrtko Ursulin <[email protected]> wrote:
> > > > >>
> > > > >>> On 24/02/2023 10:24, Pekka Paalanen wrote:
> > > > >>>> On Fri, 24 Feb 2023 09:41:46 +0000
> > > > >>>> Tvrtko Ursulin <[email protected]> wrote:
> > > > >>>>
> > > > >>>>> On 24/02/2023 09:26, Pekka Paalanen wrote:
> > > > >>>>>> On Thu, 23 Feb 2023 10:51:48 -0800
> > > > >>>>>> Rob Clark <[email protected]> wrote:
> > > > >>>>>>
> > > > >>>>>>> On Thu, Feb 23, 2023 at 1:38 AM Pekka Paalanen <[email protected]> wrote:
> > > > >>>>>>>>
> > > > >>>>>>>> On Wed, 22 Feb 2023 07:37:26 -0800
> > > > >>>>>>>> Rob Clark <[email protected]> wrote:
> > > > >>>>>>>>
> > > > >>>>>>>>> On Wed, Feb 22, 2023 at 1:49 AM Pekka Paalanen <[email protected]> wrote:
> > > > >>>>>>
> > > > >>>>>> ...
> > > > >>>>>>
> > > > >>>>>>>>>> On another matter, if the application uses SET_DEADLINE with one
> > > > >>>>>>>>>> timestamp, and the compositor uses SET_DEADLINE on the same thing with
> > > > >>>>>>>>>> another timestamp, what should happen?
> > > > >>>>>>>>>
> > > > >>>>>>>>> The expectation is that many deadline hints can be set on a fence.
> > > > >>>>>>>>> The fence signaller should track the soonest deadline.
> > > > >>>>>>>>
> > > > >>>>>>>> You need to document that as UAPI, since it is observable to userspace.
> > > > >>>>>>>> It would be bad if drivers or subsystems would differ in behaviour.
> > > > >>>>>>>>
> > > > >>>>>>>
> > > > >>>>>>> It is in the end a hint. It is about giving the driver more
> > > > >>>>>>> information so that it can make better choices. But the driver is
> > > > >>>>>>> even free to ignore it. So maybe "expectation" is too strong of a
> > > > >>>>>>> word. Rather, any other behavior doesn't really make sense. But it
> > > > >>>>>>> could end up being dictated by how the hw and/or fw works.
> > > > >>>>>>
> > > > >>>>>> It will stop being a hint once it has been implemented and used in the
> > > > >>>>>> wild long enough. The kernel userspace regression rules make sure of
> > > > >>>>>> that.
> > > > >>>>>
> > > > >>>>> Yeah, tricky and maybe a gray area in this case. I think we eluded
> > > > >>>>> elsewhere in the thread that renaming the thing might be an option.
> > > > >>>>>
> > > > >>>>> So maybe instead of deadline, which is a very strong word, use something
> > > > >>>>> along the lines of "present time hint", or "signalled time hint"? Maybe
> > > > >>>>> reads clumsy. Just throwing some ideas for a start.
> > > > >>>>
> > > > >>>> You can try, but I fear that if it ever changes behaviour and
> > > > >>>> someone notices that, it's labelled as a kernel regression. I don't
> > > > >>>> think documentation has ever been the authoritative definition of UABI
> > > > >>>> in Linux, it just guides drivers and userspace towards a common
> > > > >>>> understanding and common usage patterns.
> > > > >>>>
> > > > >>>> So even if the UABI contract is not documented (ugh), you need to be
> > > > >>>> prepared to set the UABI contract through kernel implementation.
> > > > >>>
> > > > >>> To be the devil's advocate it probably wouldn't be an ABI regression but
> > > > >>> just an regression. Same way as what nice(2) priorities mean hasn't
> > > > >>> always been the same over the years, I don't think there is a strict
> > > > >>> contract.
> > > > >>>
> > > > >>> Having said that, it may be different with latency sensitive stuff such
> > > > >>> as UIs though since it is very observable and can be very painful to users.
> > > > >>>
> > > > >>>> If you do not document the UABI contract, then different drivers are
> > > > >>>> likely to implement it differently, leading to differing behaviour.
> > > > >>>> Also userspace will invent wild ways to abuse the UABI if there is no
> > > > >>>> documentation guiding it on proper use. If userspace or end users
> > > > >>>> observe different behaviour, that's bad even if it's not a regression.
> > > > >>>>
> > > > >>>> I don't like the situation either, but it is what it is. UABI stability
> > > > >>>> trumps everything regardless of whether it was documented or not.
> > > > >>>>
> > > > >>>> I bet userspace is going to use this as a "make it faster, make it
> > > > >>>> hotter" button. I would not be surprised if someone wrote a LD_PRELOAD
> > > > >>>> library that stamps any and all fences with an expired deadline to
> > > > >>>> just squeeze out a little more through some weird side-effect.
> > > > >>>>
> > > > >>>> Well, that's hopefully overboard in scaring, but in the end, I would
> > > > >>>> like to see UABI documented so I can have a feeling of what it is for
> > > > >>>> and how it was intended to be used. That's all.
> > > > >>>
> > > > >>> We share the same concern. If you read elsewhere in these threads you
> > > > >>> will notice I have been calling this an "arms race". If the ability to
> > > > >>> make yourself go faster does not required additional privilege I also
> > > > >>> worry everyone will do it at which point it becomes pointless. So yes, I
> > > > >>> do share this concern about exposing any of this as an unprivileged uapi.
> > > > >>>
> > > > >>> Is it possible to limit access to only compositors in some sane way?
> > > > >>> Sounds tricky when dma-fence should be disconnected from DRM..
> > > > >>
> > > > >> Maybe it's not that bad in this particular case, because we are talking
> > > > >> only about boosting GPU clocks which benefits everyone (except
> > > > >> battery life) and it does not penalize other programs like e.g.
> > > > >> job priorities do.
> > > > >
> > > > > Apart from efficiency that you mentioned, which does not always favor
> > > > > higher clocks, sometimes thermal budget is also shared between CPU and
> > > > > GPU. So more GPU clocks can mean fewer CPU clocks. It's really hard to
> > > > > make optimal choices without the full coordination between both schedulers.
> > > > >
> > > > > But that is even not the main point, which is that if everyone sets the
> > > > > immediate deadline then having the deadline API is a bit pointless. For
> > > > > instance there is a reason negative nice needs CAP_SYS_ADMIN.
> > > > >
> > > > > However Rob has also pointed out the existence of uclamp.min via
> > > > > sched_setattr which is unprivileged and can influence frequency
> > > > > selection in the CPU world, so I conceded on that point. If CPU world
> > > > > has accepted it so can we I guess.
> > > > >
> > > > > So IMO we are back to whether we can agree defining it is a hint is good
> > > > > enough, be in via the name of the ioctl/flag itself or via documentation.
> > > > >
> > > > >> Drivers are not going to use the deadline for scheduling priorities,
> > > > >> right? I don't recall seeing any mention of that.
> > > > >>
> > > > >> ...right?
> > > > >
> > > > > I wouldn't have thought it would be beneficial to preclude that, or
> > > > > assume what drivers would do with the info to begin with.
> > > > >
> > > > > For instance in i915 we almost had a deadline based scheduler which was
> > > > > much fairer than the current priority sorted fifo and in an ideal world
> > > > > we would either revive or re-implement that idea. In which case
> > > > > considering the fence deadline would naturally slot in and give true
> > > > > integration with compositor deadlines (not just boost clocks and pray it
> > > > > helps).
> > > > How is user-space to decide whether to use ioctl(SET_DEADLINE) or
> > > > poll(POLLPRI)?
> > >
> > > Implementation of blocking gl/vk/cl APIs, like glFinish() would use
> > > poll(POLLPRI). It could also set an immediate deadline and then call
> > > poll() without POLLPRI.
> > >
> > > Other than compositors which do frame-pacing I expect the main usage
> > > of either of these is mesa.
> >
> > Okay, so it looks like we already agreed that having a way to bump frequency
> > from userspace is acceptable. either because there are already other ways
> > that you can waste power and because this already acceptable in the CPU
> > world.
> >
> > But why we are doing this in hidden ways then?
> >
> > Why can't we have this hint per context that is getting executed?
> > (either with a boost-context flag or with some low/med/max or '-1' to '1'
> > value like the latency priority)?
> >
> > I don't like the waitboost because this heurisitic fails in some media cases.
> > I don't like the global setting because we might be alternating a top-priority
> > with low-priority cases...
> >
> > So, why not something per context in execution?
> >
>
> It needs to be finer granularity than per-context, because not all
> waits should trigger boosting. For example, virglrenderer ends up
> with a thread polling unsignaled fences to know when to signal an
> interrupt to the guest virtgpu. This alone shouldn't trigger
> boosting. (We also wouldn't want to completely disable boosting for
> virglrenderer.) Or the usermode driver could be waiting on a fence to
> know when to do some cleanup.
>
> That is not to say that there isn't room for per-context flags to
> disable/enable boosting for fences created by that context, meaning it
> could be an AND operation for i915 if it needs to be.

Right. It can be both ways I agree.

>
> BR,
> -R

2023-03-01 16:02:49

by Rob Clark

[permalink] [raw]

Subject: Re: [PATCH v4 06/14] dma-buf/sync_file: Support (E)POLLPRI

On Wed, Mar 1, 2023 at 7:31 AM Sebastian Wick <[email protected]> wrote:
>
> On Tue, Feb 28, 2023 at 11:52 PM Rob Clark <[email protected]> wrote:
> >
> > On Tue, Feb 28, 2023 at 6:30 AM Sebastian Wick
> > <[email protected]> wrote:
> > >
> > > On Tue, Feb 28, 2023 at 12:48 AM Rob Clark <[email protected]> wrote:
> > > >
> > > > On Mon, Feb 27, 2023 at 2:44 PM Sebastian Wick
> > > > <[email protected]> wrote:
> > > > >
> > > > > On Mon, Feb 27, 2023 at 11:20 PM Rob Clark <[email protected]> wrote:
> > > > > >
> > > > > > On Mon, Feb 27, 2023 at 1:36 PM Rodrigo Vivi <[email protected]> wrote:
> > > > > > >
> > > > > > > On Fri, Feb 24, 2023 at 09:59:57AM -0800, Rob Clark wrote:
> > > > > > > > On Fri, Feb 24, 2023 at 7:27 AM Luben Tuikov <[email protected]> wrote:
> > > > > > > > >
> > > > > > > > > On 2023-02-24 06:37, Tvrtko Ursulin wrote:
> > > > > > > > > >
> > > > > > > > > > On 24/02/2023 11:00, Pekka Paalanen wrote:
> > > > > > > > > >> On Fri, 24 Feb 2023 10:50:51 +0000
> > > > > > > > > >> Tvrtko Ursulin <[email protected]> wrote:
> > > > > > > > > >>
> > > > > > > > > >>> On 24/02/2023 10:24, Pekka Paalanen wrote:
> > > > > > > > > >>>> On Fri, 24 Feb 2023 09:41:46 +0000
> > > > > > > > > >>>> Tvrtko Ursulin <[email protected]> wrote:
> > > > > > > > > >>>>
> > > > > > > > > >>>>> On 24/02/2023 09:26, Pekka Paalanen wrote:
> > > > > > > > > >>>>>> On Thu, 23 Feb 2023 10:51:48 -0800
> > > > > > > > > >>>>>> Rob Clark <[email protected]> wrote:
> > > > > > > > > >>>>>>
> > > > > > > > > >>>>>>> On Thu, Feb 23, 2023 at 1:38 AM Pekka Paalanen <[email protected]> wrote:
> > > > > > > > > >>>>>>>>
> > > > > > > > > >>>>>>>> On Wed, 22 Feb 2023 07:37:26 -0800
> > > > > > > > > >>>>>>>> Rob Clark <[email protected]> wrote:
> > > > > > > > > >>>>>>>>
> > > > > > > > > >>>>>>>>> On Wed, Feb 22, 2023 at 1:49 AM Pekka Paalanen <[email protected]> wrote:
> > > > > > > > > >>>>>>
> > > > > > > > > >>>>>> ...
> > > > > > > > > >>>>>>
> > > > > > > > > >>>>>>>>>> On another matter, if the application uses SET_DEADLINE with one
> > > > > > > > > >>>>>>>>>> timestamp, and the compositor uses SET_DEADLINE on the same thing with
> > > > > > > > > >>>>>>>>>> another timestamp, what should happen?
> > > > > > > > > >>>>>>>>>
> > > > > > > > > >>>>>>>>> The expectation is that many deadline hints can be set on a fence.
> > > > > > > > > >>>>>>>>> The fence signaller should track the soonest deadline.
> > > > > > > > > >>>>>>>>
> > > > > > > > > >>>>>>>> You need to document that as UAPI, since it is observable to userspace.
> > > > > > > > > >>>>>>>> It would be bad if drivers or subsystems would differ in behaviour.
> > > > > > > > > >>>>>>>>
> > > > > > > > > >>>>>>>
> > > > > > > > > >>>>>>> It is in the end a hint. It is about giving the driver more
> > > > > > > > > >>>>>>> information so that it can make better choices. But the driver is
> > > > > > > > > >>>>>>> even free to ignore it. So maybe "expectation" is too strong of a
> > > > > > > > > >>>>>>> word. Rather, any other behavior doesn't really make sense. But it
> > > > > > > > > >>>>>>> could end up being dictated by how the hw and/or fw works.
> > > > > > > > > >>>>>>
> > > > > > > > > >>>>>> It will stop being a hint once it has been implemented and used in the
> > > > > > > > > >>>>>> wild long enough. The kernel userspace regression rules make sure of
> > > > > > > > > >>>>>> that.
> > > > > > > > > >>>>>
> > > > > > > > > >>>>> Yeah, tricky and maybe a gray area in this case. I think we eluded
> > > > > > > > > >>>>> elsewhere in the thread that renaming the thing might be an option.
> > > > > > > > > >>>>>
> > > > > > > > > >>>>> So maybe instead of deadline, which is a very strong word, use something
> > > > > > > > > >>>>> along the lines of "present time hint", or "signalled time hint"? Maybe
> > > > > > > > > >>>>> reads clumsy. Just throwing some ideas for a start.
> > > > > > > > > >>>>
> > > > > > > > > >>>> You can try, but I fear that if it ever changes behaviour and
> > > > > > > > > >>>> someone notices that, it's labelled as a kernel regression. I don't
> > > > > > > > > >>>> think documentation has ever been the authoritative definition of UABI
> > > > > > > > > >>>> in Linux, it just guides drivers and userspace towards a common
> > > > > > > > > >>>> understanding and common usage patterns.
> > > > > > > > > >>>>
> > > > > > > > > >>>> So even if the UABI contract is not documented (ugh), you need to be
> > > > > > > > > >>>> prepared to set the UABI contract through kernel implementation.
> > > > > > > > > >>>
> > > > > > > > > >>> To be the devil's advocate it probably wouldn't be an ABI regression but
> > > > > > > > > >>> just an regression. Same way as what nice(2) priorities mean hasn't
> > > > > > > > > >>> always been the same over the years, I don't think there is a strict
> > > > > > > > > >>> contract.
> > > > > > > > > >>>
> > > > > > > > > >>> Having said that, it may be different with latency sensitive stuff such
> > > > > > > > > >>> as UIs though since it is very observable and can be very painful to users.
> > > > > > > > > >>>
> > > > > > > > > >>>> If you do not document the UABI contract, then different drivers are
> > > > > > > > > >>>> likely to implement it differently, leading to differing behaviour.
> > > > > > > > > >>>> Also userspace will invent wild ways to abuse the UABI if there is no
> > > > > > > > > >>>> documentation guiding it on proper use. If userspace or end users
> > > > > > > > > >>>> observe different behaviour, that's bad even if it's not a regression.
> > > > > > > > > >>>>
> > > > > > > > > >>>> I don't like the situation either, but it is what it is. UABI stability
> > > > > > > > > >>>> trumps everything regardless of whether it was documented or not.
> > > > > > > > > >>>>
> > > > > > > > > >>>> I bet userspace is going to use this as a "make it faster, make it
> > > > > > > > > >>>> hotter" button. I would not be surprised if someone wrote a LD_PRELOAD
> > > > > > > > > >>>> library that stamps any and all fences with an expired deadline to
> > > > > > > > > >>>> just squeeze out a little more through some weird side-effect.
> > > > > > > > > >>>>
> > > > > > > > > >>>> Well, that's hopefully overboard in scaring, but in the end, I would
> > > > > > > > > >>>> like to see UABI documented so I can have a feeling of what it is for
> > > > > > > > > >>>> and how it was intended to be used. That's all.
> > > > > > > > > >>>
> > > > > > > > > >>> We share the same concern. If you read elsewhere in these threads you
> > > > > > > > > >>> will notice I have been calling this an "arms race". If the ability to
> > > > > > > > > >>> make yourself go faster does not required additional privilege I also
> > > > > > > > > >>> worry everyone will do it at which point it becomes pointless. So yes, I
> > > > > > > > > >>> do share this concern about exposing any of this as an unprivileged uapi.
> > > > > > > > > >>>
> > > > > > > > > >>> Is it possible to limit access to only compositors in some sane way?
> > > > > > > > > >>> Sounds tricky when dma-fence should be disconnected from DRM..
> > > > > > > > > >>
> > > > > > > > > >> Maybe it's not that bad in this particular case, because we are talking
> > > > > > > > > >> only about boosting GPU clocks which benefits everyone (except
> > > > > > > > > >> battery life) and it does not penalize other programs like e.g.
> > > > > > > > > >> job priorities do.
> > > > > > > > > >
> > > > > > > > > > Apart from efficiency that you mentioned, which does not always favor
> > > > > > > > > > higher clocks, sometimes thermal budget is also shared between CPU and
> > > > > > > > > > GPU. So more GPU clocks can mean fewer CPU clocks. It's really hard to
> > > > > > > > > > make optimal choices without the full coordination between both schedulers.
> > > > > > > > > >
> > > > > > > > > > But that is even not the main point, which is that if everyone sets the
> > > > > > > > > > immediate deadline then having the deadline API is a bit pointless. For
> > > > > > > > > > instance there is a reason negative nice needs CAP_SYS_ADMIN.
> > > > > > > > > >
> > > > > > > > > > However Rob has also pointed out the existence of uclamp.min via
> > > > > > > > > > sched_setattr which is unprivileged and can influence frequency
> > > > > > > > > > selection in the CPU world, so I conceded on that point. If CPU world
> > > > > > > > > > has accepted it so can we I guess.
> > > > > > > > > >
> > > > > > > > > > So IMO we are back to whether we can agree defining it is a hint is good
> > > > > > > > > > enough, be in via the name of the ioctl/flag itself or via documentation.
> > > > > > > > > >
> > > > > > > > > >> Drivers are not going to use the deadline for scheduling priorities,
> > > > > > > > > >> right? I don't recall seeing any mention of that.
> > > > > > > > > >>
> > > > > > > > > >> ...right?
> > > > > > > > > >
> > > > > > > > > > I wouldn't have thought it would be beneficial to preclude that, or
> > > > > > > > > > assume what drivers would do with the info to begin with.
> > > > > > > > > >
> > > > > > > > > > For instance in i915 we almost had a deadline based scheduler which was
> > > > > > > > > > much fairer than the current priority sorted fifo and in an ideal world
> > > > > > > > > > we would either revive or re-implement that idea. In which case
> > > > > > > > > > considering the fence deadline would naturally slot in and give true
> > > > > > > > > > integration with compositor deadlines (not just boost clocks and pray it
> > > > > > > > > > helps).
> > > > > > > > > How is user-space to decide whether to use ioctl(SET_DEADLINE) or
> > > > > > > > > poll(POLLPRI)?
> > > > > > > >
> > > > > > > > Implementation of blocking gl/vk/cl APIs, like glFinish() would use
> > > > > > > > poll(POLLPRI). It could also set an immediate deadline and then call
> > > > > > > > poll() without POLLPRI.
> > > > > > > >
> > > > > > > > Other than compositors which do frame-pacing I expect the main usage
> > > > > > > > of either of these is mesa.
> > > > > > >
> > > > > > > Okay, so it looks like we already agreed that having a way to bump frequency
> > > > > > > from userspace is acceptable. either because there are already other ways
> > > > > > > that you can waste power and because this already acceptable in the CPU
> > > > > > > world.
> > > > > > >
> > > > > > > But why we are doing this in hidden ways then?
> > > > > > >
> > > > > > > Why can't we have this hint per context that is getting executed?
> > > > > > > (either with a boost-context flag or with some low/med/max or '-1' to '1'
> > > > > > > value like the latency priority)?
> > > > > > >
> > > > > > > I don't like the waitboost because this heurisitic fails in some media cases.
> > > > > > > I don't like the global setting because we might be alternating a top-priority
> > > > > > > with low-priority cases...
> > > > > > >
> > > > > > > So, why not something per context in execution?
> > > > > > >
> > > > > >
> > > > > > It needs to be finer granularity than per-context, because not all
> > > > > > waits should trigger boosting. For example, virglrenderer ends up
> > > > > > with a thread polling unsignaled fences to know when to signal an
> > > > > > interrupt to the guest virtgpu. This alone shouldn't trigger
> > > > > > boosting. (We also wouldn't want to completely disable boosting for
> > > > > > virglrenderer.) Or the usermode driver could be waiting on a fence to
> > > > > > know when to do some cleanup.
> > > > > >
> > > > > > That is not to say that there isn't room for per-context flags to
> > > > > > disable/enable boosting for fences created by that context, meaning it
> > > > > > could be an AND operation for i915 if it needs to be.
> > > > >
> > > > > First of all, I believe that the fence deadline hint is a good idea.
> > > > > With that being said, I also don't think it is sufficient in a lot of
> > > > > cases.
> > > > >
> > > > > The one thing I was alluding to before and that Pekka mentioned as
> > > > > well is that mutter for example has a problem where we're missing the
> > > > > deadline consistently because the clocks don't ramp up fast enough and
> > > > > there is a MR which is just trying to keep the GPU busy to avoid this.
> > > >
> > > > the dynamic double/triple buffer thing?
> > >
> > > Yes
> > >
> > > > > It would be much better if the kernel could make sure the clocks are
> > > > > all ramped up when we start submitting work. In the compositor we
> > > > > actually have a lot of information that *should* influence clocks. We
> > > > > know when we're going to start submitting work and when the deadline
> > > > > for that work is beforehand. We know which windows are visible, and
> > > > > which one should have the highest priority.
> > > >
> > > > This sounds like something orthogonal.. something for cgroups? Ie.
> > > > android moves visible/foreground apps to a different cgroup to given
> > > > them higher priority. Tvrtko had a patchset to add drm cgroup
> > > > support..
> > >
> > > For the priority stuff, yes, probably. The visibility information on
> > > the other hand could be used to determine if we want to ramp up the
> > > GPU in the first place.
> >
> > Right, but I think that we could have multiple cgroup based knobs, one
> > that adjusts priority and one that limits/disables deadline based
> > boost? This way the compositor could setup different policies for
> > visible vs hidden apps influencing both how much time they get on the
> > GPU and boost.
>
> I'm not sure if a negative control really makes sense. There are
> probably timing sensitive use cases where the result doesn't show up
> on the local screen and penalizing them when they are not focused or
> hidden might also not be the best idea.

Policy decisions are up to the OS/distro.. we can only provide
controls that can be tuned, it is up to someone else to choose how to
use those controls, such as whether it wants to differentiate between
visible and non-visible apps. That is why cgroups and a negative
control are a good solution for controlling what the driver does with
the deadline boost hint.

> > > > > We know when there are
> > > > > input events which actually matter.
> > > >
> > > > This I see input as a different boost source for the driver. (Ie. one
> > > > boost signal is missing fence deadlines, another is input events,
> > > > etc.)
> > > >
> > > > We end up using downstream input-handlers on the kernel side for this.
> > > > Partially for the freq boost (but mostly not, UI interactive workloads
> > > > like touchscreen scrolling don't generally need high GPU freqs, they
> > > > are more memory bandwidth limited if they are limited by anything)..
> > > > really the reason here is to get a head-start on the ~2ms that it
> > > > takes to power up the GPU if it is suspended.
> > >
> > > Right, but one of my main points I want to make here is that we could
> > > get the head-start not only in response to input events but also for
> > > the GPU work the compositor submits and in the future also to GPU work
> > > that clients commit. Except that we don't have a way to tell the
> > > kernel about it.
> > >
> > > > But this is not quite perfect, since for example some keys should be
> > > > handled on key-down but others on key-up.
> > > >
> > > > But again, this is something different from fence deadlines. I'm
> > > > interested in proposals because we do need something for this. But I
> > > > think it is something is orthogonal to this series. For input, we
> > > > want the kernel to know long before userspace is ready to submit
> > > > rendering.
> > >
> > > We can do that in the compositor! Input events are really not
> > > something you should care about in the kernel. Input itself is also
> > > not the only indication of incoming animated content. Some network
> > > packets arriving could equally well result in the same situation.
> >
> > We do use input boost not just for GPU freq, but also for CPU freq.
> > It seems like triggering it from the kernel could happen somewhat
> > sooner than userspace. (But just speculation.)
>
> Technically it has to be sooner but I doubt it makes any difference.
> Getting a lot of false-positives on the other hand does make a
> difference.

Regardless of whether it is done in kernel or userspace, you do want a
cooldown period so you are not constantly boosting.

Generally false-positives aren't much of a problem.. ie. touch or
mouse events are not ambiguous. The exception is key events, because
ideally you don't want to hard-code in the kernel which keys are
modifier keys that should be handled on key-up instead of key-down.
But I think if we allowed userspace to configure this somehow, it
would be perfectly reasonable (and optimal) to handle input boost in
the kernel.

> > I'm not sure network packets count. Or at least compared to a touch
> > interface, the user won't perceive the lag nearly as much, compared to
> > whether or not the UI tracks their fingertips.
>
> Sure, stutter in interactive cases is especially awful but stutter is
> awful in general. If we can solve it in all cases we should do so.

Sure, but the cases where I've seen a need for input boost are all
about the transition from idle->active. Ie. the display panel has
gone to self-refresh, gpu is suspended, cpu and memory is downclocked.
And suddenly the user decides they want to start scrolling. We need
some help to adapt to the new state where we are all of a sudden busy
doing something. This is where input-boost comes in. If there is no
direct connection to user input, then there is no lag for the user to
percieve.

> > > > > We know when the deadline for
> > > > > client work is.
> > > > >
> > > > > In the future we also want to make sure clients know beforehand when
> > > > > they should start their work and when the deadline is but that's all
> > > > > very much WIP in both wayland and vulkan.
> > > > >
> > > > > There are two issues:
> > > > >
> > > > > 1. The compositor has no way to communicate any of that information to
> > > > > the kernel.
> > > > > 2. The only connection to client work the compositor has is a fence to
> > > > > the last bit of work that must be done before the deadline after a
> > > > > wl_surface.commit.
> > > >
> > > > If the client isn't using multiple GPUs, a single fence should be
> > > > sufficient. And even if it is, well we still have all the dependency
> > > > information on the kernel side. Ie. drm/sched knows what fences it is
> > > > waiting on if it is waiting to schedule the work associated with the
> > > > last fence. It would otherwise require drm/sched to be a bit more
> > > > tricky than it is so far in this series.
> > > >
> > > > But I think the normal dual-gpu case, the app is only dealing with a single GPU?
> > >
> > > We generally don't know which GPU a client uses though. We know which
> > > one we're using and tell the client that the buffer should be
> > > compatible with it but that's the extent of information we have, until
> > > we get a fence but that fence usually gets to the compositor pretty
> > > late. Way too late for the compositor to tell the kernel to ramp up
> > > the GPU and still have an impact.
> >
> > Are you sure about that? I'd have expected the app to hand off
> > fence+buffer to the compositor pretty quickly after rendering is
> > submitted. This is what I've seen elsewhere.
>
> After rendering is submitted it is already too late if the GPU takes
> 2ms to wake up. And if there is no rendering submitted there is no
> fence and if there is no fence it has no deadline. If we want to solve
> that we also need some kind of 'work will be submitted to the queue
> starting from this time' hint.

Right, which is why I view this as a different problem with a
different solution.

> > > It also seems like we're moving away from tracking execution
> > > dependencies with fences when we're switching to user mode fences.
> >
> > I suppose there are two cases..
> >
> > 1) dependent fences all from the same device.. no prob, the right
> > driver gets the signal that it needs to clk up, and figures out the
> > rest on it's own
>
> AFAIU with user mode fences it's impossible for the kernel to figure
> out what work depends on it and it might never signal. The whole
> deadline on fences thing breaks down with user mode fences.

Given that user-mode fences are not a solved problem (and I don't
think they'll ever completely replace kernel fences), I don't think
this is relevant.

BR,
-R

> > 2) dependent fences from different devices.. I suppose if device B is
> > waiting for a fence from device A before it can make forward progress,
> > why not express this as a deadline hint on A's fence? (But by the
> > time you have this problem, you are probably not really caring about
> > power consumption, so meh..)
> >
> > BR,
> > -R
> >
> > > > > So in both cases a fence is just not the right primitive for us. We
> > > > > need to be able to provide per-context/queue information for work that
> > > > > will happen in the future and we need a way to refer to a
> > > > > context/queue generically and over IPC to boost the clocks of the
> > > > > device that a client is actually using and maybe even give priority.
> > > > >
> > > > > But like I said, having a per-fence deadline is probably still a good
> > > > > idea and doesn't conflict with any of the more coarse information.
> > > >
> > > > Yeah, I think the thing is you need multiple things, and this is only
> > > > one of them ;-)
> > > >
> > > > BR,
> > > > -R
> > > >
> > >
> >
>