[v6,0/4] media: pisp-be: Split jobs creation and scheduling

Message

Jacopo Mondi Sept. 30, 2024, 8:32 a.m. UTC

v5->v6:
- Make the driver depend on PM
  - Simplify the probe() routine by using pm_runtime_
  - Remove suspend call from remove()

v4->v5:
- Use appropriate locking constructs:
  - spin_lock_irq() for pispbe_prepare_job() called from non irq context
  - spin_lock_irqsave() for pispbe_schedule() called from irq context
  - Remove hw_lock from ready_queue accesses in stop_streaming and
    start_streaming
  - Fix trivial indentation mistake in 4/4

v3->v4:
- Expand commit message in 2/4 to explain why removing validation in schedule()
  is safe
- Drop ready_lock spinlock
- Use non _irqsave version of safe_guard(spinlock
- Support !CONFIG_PM in 4/4 by calling the enable/disable routines directly
  and adjust pm_runtime usage as suggested by Laurent

v2->v3:
- Mark pispbe_runtime_resume() as __maybe_unused
- Add fixes tags where appropriate

v1->v2:
- Add two patches to address Laurent's comments separately
- use scoped_guard() when possible
- Add patch to fix runtime_pm imbalance

Currently the 'pispbe_schedule()' function does two things:

1) Tries to assemble a job by inspecting all the video node queues
   to make sure all the required buffers are available
2) Submit the job to the hardware

The pispbe_schedule() function is called at:

- video device start_streaming() time
- video device qbuf() time
- irq handler

As assembling a job requires inspecting all queues, it is a rather
time consuming operation which is better not run in IRQ context.

To avoid executing the time consuming job creation in interrupt
context, split the job creation and job scheduling in two distinct
operations. When a well-formed job is created, append it to the
newly introduced 'pispbe->job_queue' where it will be dequeued from
by the scheduling routine.

At start_streaming() and qbuf() time immediately try to schedule a job
if one has been created as the irq handler routine is only called when
a job has completed, and we can't solely rely on it for scheduling new
jobs.

Signed-off-by: Jacopo Mondi <[e-mail]>
---
Jacopo Mondi (4):
      media: pisp_be: Drop reference to non-existing function
      media: pisp_be: Remove config validation from schedule()
      media: pisp_be: Split jobs creation and scheduling
      media: pisp_be: Fix pm_runtime underrun in probe

 drivers/media/platform/raspberrypi/pisp_be/Kconfig |   1 +
 .../media/platform/raspberrypi/pisp_be/pisp_be.c   | 183 ++++++++++-----------
 2 files changed, 88 insertions(+), 96 deletions(-)
---
base-commit: 81ee62e8d09ee3c7107d11c8bbfd64073ab601ad
change-id: 20240930-pispbe-mainline-split-jobs-handling-v6-15dc16e11e3a

Best regards,

Comments

Jacopo Mondi Oct. 14, 2024, 6:55 a.m. UTC | #1

Hi,
  gentle ping as I wish this to go in for v6.13

On Mon, Sep 30, 2024 at 10:32:56AM GMT, Jacopo Mondi wrote:
> v5->v6:
> - Make the driver depend on PM
>   - Simplify the probe() routine by using pm_runtime_
>   - Remove suspend call from remove()
>
> v4->v5:
> - Use appropriate locking constructs:
>   - spin_lock_irq() for pispbe_prepare_job() called from non irq context
>   - spin_lock_irqsave() for pispbe_schedule() called from irq context
>   - Remove hw_lock from ready_queue accesses in stop_streaming and
>     start_streaming
>   - Fix trivial indentation mistake in 4/4
>
> v3->v4:
> - Expand commit message in 2/4 to explain why removing validation in schedule()
>   is safe
> - Drop ready_lock spinlock
> - Use non _irqsave version of safe_guard(spinlock
> - Support !CONFIG_PM in 4/4 by calling the enable/disable routines directly
>   and adjust pm_runtime usage as suggested by Laurent
>
> v2->v3:
> - Mark pispbe_runtime_resume() as __maybe_unused
> - Add fixes tags where appropriate
>
> v1->v2:
> - Add two patches to address Laurent's comments separately
> - use scoped_guard() when possible
> - Add patch to fix runtime_pm imbalance
>
> Currently the 'pispbe_schedule()' function does two things:
>
> 1) Tries to assemble a job by inspecting all the video node queues
>    to make sure all the required buffers are available
> 2) Submit the job to the hardware
>
> The pispbe_schedule() function is called at:
>
> - video device start_streaming() time
> - video device qbuf() time
> - irq handler
>
> As assembling a job requires inspecting all queues, it is a rather
> time consuming operation which is better not run in IRQ context.
>
> To avoid executing the time consuming job creation in interrupt
> context, split the job creation and job scheduling in two distinct
> operations. When a well-formed job is created, append it to the
> newly introduced 'pispbe->job_queue' where it will be dequeued from
> by the scheduling routine.
>
> At start_streaming() and qbuf() time immediately try to schedule a job
> if one has been created as the irq handler routine is only called when
> a job has completed, and we can't solely rely on it for scheduling new
> jobs.
>
> Signed-off-by: Jacopo Mondi <[e-mail]>
> ---
> Jacopo Mondi (4):
>       media: pisp_be: Drop reference to non-existing function
>       media: pisp_be: Remove config validation from schedule()
>       media: pisp_be: Split jobs creation and scheduling
>       media: pisp_be: Fix pm_runtime underrun in probe
>
>  drivers/media/platform/raspberrypi/pisp_be/Kconfig |   1 +
>  .../media/platform/raspberrypi/pisp_be/pisp_be.c   | 183 ++++++++++-----------
>  2 files changed, 88 insertions(+), 96 deletions(-)
> ---
> base-commit: 81ee62e8d09ee3c7107d11c8bbfd64073ab601ad
> change-id: 20240930-pispbe-mainline-split-jobs-handling-v6-15dc16e11e3a
>
> Best regards,
> --
> Jacopo Mondi <[e-mail]>
>
>