Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
To: <intel-xe@lists.freedesktop.org>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Subject: Re: [PATCH] drm/xe: improve Kconfig.profile help text for scheduler timeouts
Date: Fri, 5 Jun 2026 12:48:29 -0400	[thread overview]
Message-ID: <aiL93V-6K1Por7H4@intel.com> (raw)
In-Reply-To: <20260605140951.958172-2-rodrigo.vivi@intel.com>

On Fri, Jun 05, 2026 at 10:09:52AM -0400, Rodrigo Vivi wrote:
> The existing help texts for the JOB_TIMEOUT, TIMESLICE and PREEMPT_TIMEOUT
> configs were brief and did not make the role of each symbol clear:
> 
>  - _MIN / _MAX: hard bounds on the per-engine-class timeout. They are
>                 enforced unconditionally by the sysfs knobs, and (for
>                 TIMESLICE, the only one exposed via the SET_PROPERTY UAPI)
>                 they also bound CAP_SYS_NICE requests when
>                 DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT is enabled.
>  - PREEMPT_TIMEOUT: the boot-time default; the JOB_TIMEOUT and TIMESLICE
>                     defaults are hardcoded in the driver, not configured
>                     here.
> 
> Rewrite the help texts to reflect this, naming the relevant sysfs knobs
> and UAPI property explicitly.
> 
> v2: Adjusted commit message based on Sashiko's review.
> 
> Assisted-by: GitHub-Copilot:claude-sonnet-4.6
> Assisted-by: GitHub-Copilot:claude-opus-4.8 #v2

and if we want to be fair with all the tools and models
I guess I also need to add this:

Assisted-by: Sashiko-Review:gemini-3.1-pro-preview

> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> ---
>  drivers/gpu/drm/xe/Kconfig.profile | 71 +++++++++++++++++++-----------
>  1 file changed, 46 insertions(+), 25 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/Kconfig.profile b/drivers/gpu/drm/xe/Kconfig.profile
> index 7530df998148..e07517d120e0 100644
> --- a/drivers/gpu/drm/xe/Kconfig.profile
> +++ b/drivers/gpu/drm/xe/Kconfig.profile
> @@ -1,50 +1,71 @@
>  # SPDX-License-Identifier: GPL-2.0-only
>  config DRM_XE_JOB_TIMEOUT_MAX
> -	int "Default max job timeout (ms)"
> +	int "Hard upper limit for job timeout (ms)"
>  	default 10000 # milliseconds
>  	help
> -	  Configures the default max job timeout after which job will
> -	  be forcefully taken away from scheduler.
> +	  Absolute upper bound (in milliseconds) for the per-engine-class job
> +	  timeout. This is the maximum value that can be written to the sysfs
> +	  job_timeout_ms knob, regardless of privileges. To raise this ceiling,
> +	  increase this value and rebuild the kernel.
>  config DRM_XE_JOB_TIMEOUT_MIN
> -	int "Default min job timeout (ms)"
> +	int "Hard lower limit for job timeout (ms)"
>  	default 1 # milliseconds
>  	help
> -	  Configures the default min job timeout after which job will
> -	  be forcefully taken away from scheduler.
> +	  Absolute lower bound (in milliseconds) for the per-engine-class job
> +	  timeout. This is the minimum value that can be written to the sysfs
> +	  job_timeout_ms knob, regardless of privileges.
> +
> +	  Note: the job timeout default (5000 ms) is hardcoded in the driver
> +	  and is not configurable here. Use the sysfs job_timeout_ms knob at
> +	  runtime to change the engine-class default.
>  config DRM_XE_TIMESLICE_MAX
> -	int "Default max timeslice duration (us)"
> +	int "Hard upper limit for timeslice duration (us)"
>  	default 10000000 # microseconds
>  	help
> -	  Configures the default max timeslice duration between multiple
> -	  contexts by guc scheduling.
> +	  Absolute upper bound (in microseconds) for the timeslice duration.
> +	  This caps both the sysfs timeslice_duration_us knob and the value
> +	  accepted via the DRM_XE_EXEC_QUEUE_SET_PROPERTY_TIMESLICE UAPI for
> +	  processes with CAP_SYS_NICE when DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT
> +	  is enabled.
>  config DRM_XE_TIMESLICE_MIN
> -	int "Default min timeslice duration (us)"
> +	int "Hard lower limit for timeslice duration (us)"
>  	default 1 # microseconds
>  	help
> -	  Configures the default min timeslice duration between multiple
> -	  contexts by guc scheduling.
> +	  Absolute lower bound (in microseconds) for the timeslice duration.
> +	  This caps both the sysfs timeslice_duration_us knob and the value
> +	  accepted via the DRM_XE_EXEC_QUEUE_SET_PROPERTY_TIMESLICE UAPI for
> +	  processes with CAP_SYS_NICE when DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT
> +	  is enabled.
>  config DRM_XE_PREEMPT_TIMEOUT
> -	int "Preempt timeout (us, jiffy granularity)"
> +	int "Default preempt timeout (us, jiffy granularity)"
>  	default 640000 # microseconds
>  	help
> -	  How long to wait (in microseconds) for a preemption event to occur
> -	  when submitting a new context. If the current context does not hit
> -	  an arbitration point and yield to HW before the timer expires, the
> -	  HW will be reset to allow the more important context to execute.
> +	  Initial per-engine-class preemption timeout (in microseconds). This
> +	  is the value the driver programs at boot; it can be changed at
> +	  runtime via the sysfs preempt_timeout_us knob.
> +
> +	  This is how long the driver waits for the current context to reach
> +	  an arbitration point and yield the GPU voluntarily when a
> +	  higher-priority context becomes runnable. If the context does not
> +	  yield before the timer expires, the HW is reset to allow the
> +	  higher-priority context to execute.
> +
> +	  The range userspace may write via sysfs is bounded by
> +	  DRM_XE_PREEMPT_TIMEOUT_MIN and DRM_XE_PREEMPT_TIMEOUT_MAX.
>  config DRM_XE_PREEMPT_TIMEOUT_MAX
> -	int "Default max preempt timeout (us)"
> +	int "Hard upper limit for preempt timeout (us)"
>  	default 10000000 # microseconds
>  	help
> -	  Configures the default max preempt timeout after which context
> -	  will be forcefully taken away and higher priority context will
> -	  run.
> +	  Absolute upper bound (in microseconds) for the per-engine-class
> +	  preemption timeout. This is the maximum value that can be written to
> +	  the sysfs preempt_timeout_us knob, regardless of privileges.
>  config DRM_XE_PREEMPT_TIMEOUT_MIN
> -	int "Default min preempt timeout (us)"
> +	int "Hard lower limit for preempt timeout (us)"
>  	default 1 # microseconds
>  	help
> -	  Configures the default min preempt timeout after which context
> -	  will be forcefully taken away and higher priority context will
> -	  run.
> +	  Absolute lower bound (in microseconds) for the per-engine-class
> +	  preemption timeout. This is the minimum value that can be written to
> +	  the sysfs preempt_timeout_us knob, regardless of privileges.
>  config DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT
>  	bool "Default configuration of limitation on scheduler timeout"
>  	default y
> -- 
> 2.54.0
> 

  parent reply	other threads:[~2026-06-05 16:50 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-06-05 14:09 [PATCH] drm/xe: improve Kconfig.profile help text for scheduler timeouts Rodrigo Vivi
2026-06-05 15:32 ` ✓ CI.KUnit: success for drm/xe: improve Kconfig.profile help text for scheduler timeouts (rev2) Patchwork
2026-06-05 16:20 ` ✓ Xe.CI.BAT: " Patchwork
2026-06-05 16:48 ` Rodrigo Vivi [this message]
2026-06-06  4:11 ` ✓ Xe.CI.FULL: " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2026-05-28 20:24 [PATCH] drm/xe: improve Kconfig.profile help text for scheduler timeouts Rodrigo Vivi
2026-06-04 20:57 ` Zanoni, Paulo R

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=aiL93V-6K1Por7H4@intel.com \
    --to=rodrigo.vivi@intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=paulo.r.zanoni@intel.com \
    /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