From: Matthew Brost <matthew.brost@intel.com>
To: Francois Dugast <francois.dugast@intel.com>
Cc: <intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH v4 08/13] drm/xe/hw_engine_group: Ensure safe transition between execution modes
Date: Mon, 5 Aug 2024 20:29:41 +0000 [thread overview]
Message-ID: <ZrE2NR1nFWfSRlZ8@DUT025-TGLU.fm.intel.com> (raw)
In-Reply-To: <20240801125748.355078-9-francois.dugast@intel.com>
On Thu, Aug 01, 2024 at 02:56:49PM +0200, Francois Dugast wrote:
> Provide a way to safely transition execution modes of the hw engine group
> ahead of the actual execution. When necessary, either wait for running
> jobs to complete or preempt them, thus ensuring mutual exclusion between
> execution modes.
>
> Unlike a mutex, the rw_semaphore used in this context allows multiple
> submissions in the same mode.
>
> v2: Use lockdep_assert_held_write, add annotations (Matt Brost)
>
> Signed-off-by: Francois Dugast <francois.dugast@intel.com>
> ---
> drivers/gpu/drm/xe/xe_hw_engine_group.c | 71 +++++++++++++++++++++++++
> drivers/gpu/drm/xe/xe_hw_engine_group.h | 5 ++
> 2 files changed, 76 insertions(+)
>
> diff --git a/drivers/gpu/drm/xe/xe_hw_engine_group.c b/drivers/gpu/drm/xe/xe_hw_engine_group.c
> index de057c98d5a5..a5754e6de7c9 100644
> --- a/drivers/gpu/drm/xe/xe_hw_engine_group.c
> +++ b/drivers/gpu/drm/xe/xe_hw_engine_group.c
> @@ -208,3 +208,74 @@ static int xe_hw_engine_group_wait_for_dma_fence_jobs(struct xe_hw_engine_group
>
> return 0;
> }
> +
> +static int switch_mode(struct xe_hw_engine_group *group,
> + enum xe_hw_engine_group_execution_mode new_mode)
> +{
> + int err = 0;
> +
> + lockdep_assert_held_write(&group->mode_sem);
> +
> + if (group->cur_mode == EXEC_MODE_LR && new_mode == EXEC_MODE_DMA_FENCE)
This seems redunant, how about:
switch (group->cur_mode) {
case EXEC_MODE_LR:
xe_hw_engine_group_suspend_faulting_lr_jobs(group);
break;
case EXEC_MODE_DMA_FENCE:
err = xe_hw_engine_group_wait_for_dma_fence_jobs(group);
break;
}
Matt
> + xe_hw_engine_group_suspend_faulting_lr_jobs(group);
> + else if (group->cur_mode == EXEC_MODE_DMA_FENCE && new_mode == EXEC_MODE_LR)
> + err = xe_hw_engine_group_wait_for_dma_fence_jobs(group);
> + else
> + err = -EINVAL;
> +
> + if (err)
> + return err;
> +
> + group->cur_mode = new_mode;
> +
> + return 0;
> +}
> +
> +/**
> + * xe_hw_engine_group_get_mode() - Get the group to execute in the new mode
> + * @group: The hw engine group
> + * @mode: The new execution mode
> + * @previous_mode: Pointer to the previous mode provided for use by caller
> + *
> + * Return: 0 if successful, -EINTR if locking failed.
> + */
> +int xe_hw_engine_group_get_mode(struct xe_hw_engine_group *group,
> + enum xe_hw_engine_group_execution_mode new_mode,
> + enum xe_hw_engine_group_execution_mode *previous_mode)
> +__acquires(&group->mode_sem)
> +{
> + int err = down_read_interruptible(&group->mode_sem);
> +
> + if (err)
> + return err;
> +
> + *previous_mode = group->cur_mode;
> +
> + if (new_mode != group->cur_mode) {
> + up_read(&group->mode_sem);
> + err = down_write_killable(&group->mode_sem);
> + if (err)
> + return err;
> +
> + if (new_mode != group->cur_mode) {
> + err = switch_mode(group, new_mode);
> + if (err) {
> + up_write(&group->mode_sem);
> + return err;
> + }
> + }
> + downgrade_write(&group->mode_sem);
> + }
> +
> + return err;
> +}
> +
> +/**
> + * xe_hw_engine_group_put() - Put the group
> + * @group: The hw engine group
> + */
> +void xe_hw_engine_group_put(struct xe_hw_engine_group *group)
> +__releases(&group->mode_sem)
> +{
> + up_read(&group->mode_sem);
> +}
> diff --git a/drivers/gpu/drm/xe/xe_hw_engine_group.h b/drivers/gpu/drm/xe/xe_hw_engine_group.h
> index 857a83787504..e0deb7c7bb5b 100644
> --- a/drivers/gpu/drm/xe/xe_hw_engine_group.h
> +++ b/drivers/gpu/drm/xe/xe_hw_engine_group.h
> @@ -17,4 +17,9 @@ int xe_hw_engine_setup_groups(struct xe_gt *gt);
> int xe_hw_engine_group_add_exec_queue(struct xe_hw_engine_group *group, struct xe_exec_queue *q);
> void xe_hw_engine_group_del_exec_queue(struct xe_hw_engine_group *group, struct xe_exec_queue *q);
>
> +int xe_hw_engine_group_get_mode(struct xe_hw_engine_group *group,
> + enum xe_hw_engine_group_execution_mode new_mode,
> + enum xe_hw_engine_group_execution_mode *previous_mode);
> +void xe_hw_engine_group_put(struct xe_hw_engine_group *group);
> +
> #endif
> --
> 2.43.0
>
next prev parent reply other threads:[~2024-08-05 20:30 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-08-01 12:56 [PATCH v4 00/13] Parallel submission of dma fence jobs and LR jobs with shared hardware resources Francois Dugast
2024-08-01 12:56 ` [PATCH v4 01/13] drm/xe/hw_engine_group: Introduce xe_hw_engine_group Francois Dugast
2024-08-01 12:56 ` [PATCH v4 02/13] drm/xe/guc_submit: Make suspend_wait interruptible Francois Dugast
2024-08-01 12:56 ` [PATCH v4 03/13] drm/xe/hw_engine_group: Register hw engine group's exec queues Francois Dugast
2024-08-05 20:19 ` Matthew Brost
2024-08-01 12:56 ` [PATCH v4 04/13] drm/xe/hw_engine_group: Add helper to suspend faulting LR jobs Francois Dugast
2024-08-01 12:56 ` [PATCH v4 05/13] drm/xe/exec_queue: Remove duplicated code Francois Dugast
2024-08-01 12:56 ` [PATCH v4 06/13] drm/xe/exec_queue: Prepare last fence for hw engine group resume context Francois Dugast
2024-08-05 20:25 ` Matthew Brost
2024-08-06 9:46 ` Francois Dugast
2024-08-06 14:33 ` Matthew Brost
2024-08-01 12:56 ` [PATCH v4 07/13] drm/xe/hw_engine_group: Add helper to wait for dma fence jobs Francois Dugast
2024-08-01 12:56 ` [PATCH v4 08/13] drm/xe/hw_engine_group: Ensure safe transition between execution modes Francois Dugast
2024-08-05 20:29 ` Matthew Brost [this message]
2024-08-01 12:56 ` [PATCH v4 09/13] drm/xe/exec: Switch hw engine group execution mode upon job submission Francois Dugast
2024-08-01 12:56 ` [PATCH v4 10/13] drm/xe/guc_submit: Allow calling guc_exec_queue_resume with pending resume Francois Dugast
2024-08-01 12:56 ` [PATCH v4 11/13] drm/xe/hw_engine_group: Resume exec queues suspended by dma fence jobs Francois Dugast
2024-08-05 20:32 ` Matthew Brost
2024-08-01 12:56 ` [PATCH v4 12/13] drm/xe/vm: Remove restriction that all VMs must be faulting if one is Francois Dugast
2024-08-01 12:56 ` [PATCH v4 13/13] drm/xe/device: Remove unused xe_device::usm::num_vm_in_* Francois Dugast
2024-08-01 13:03 ` ✓ CI.Patch_applied: success for Parallel submission of dma fence jobs and LR jobs with shared hardware resources (rev4) Patchwork
2024-08-01 13:04 ` ✗ CI.checkpatch: warning " Patchwork
2024-08-01 13:05 ` ✓ CI.KUnit: success " Patchwork
2024-08-01 13:17 ` ✓ CI.Build: " Patchwork
2024-08-01 13:19 ` ✗ CI.Hooks: failure " Patchwork
2024-08-01 13:21 ` ✓ CI.checksparse: success " Patchwork
2024-08-01 13:41 ` ✗ CI.BAT: failure " Patchwork
2024-08-01 14:56 ` ✗ CI.FULL: " Patchwork
2024-08-01 16:27 ` ✓ CI.Patch_applied: success for Parallel submission of dma fence jobs and LR jobs with shared hardware resources (rev5) Patchwork
2024-08-01 16:27 ` ✗ CI.checkpatch: warning " Patchwork
2024-08-01 16:28 ` ✓ CI.KUnit: success " Patchwork
2024-08-01 16:40 ` ✓ CI.Build: " Patchwork
2024-08-01 16:42 ` ✗ CI.Hooks: failure " Patchwork
2024-08-01 16:44 ` ✓ CI.checksparse: success " Patchwork
2024-08-01 17:13 ` ✗ CI.BAT: failure " Patchwork
2024-08-01 20:32 ` ✗ CI.FULL: " Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ZrE2NR1nFWfSRlZ8@DUT025-TGLU.fm.intel.com \
--to=matthew.brost@intel.com \
--cc=francois.dugast@intel.com \
--cc=intel-xe@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox