All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Vinay Belgaumkar" <vinay.belgaumkar@intel.com>,
	intel-xe@lists.freedesktop.org
Subject: Re: [PATCH v3 1/2] drm/xe: Disallow freq changes while mert cap is active
Date: Wed, 18 Nov 2026 02:01:45 +0000	[thread overview]
Message-ID: <sashiko-outbox-219921@kernel.org> (raw)
In-Reply-To: <20261118012453.626087-3-vinay.belgaumkar@intel.com>

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Medium] drm/xe: mert_freq_cap_active left set if async gsc_upload()
  fails
--

commit 122f6e08f3dbfe345a69d449458dd10c810ca28c
Author: Vinay Belgaumkar <vinay.belgaumkar@intel.com>

drm/xe: Disallow freq changes while mert cap is active

This commit adds a mert_freq_cap_active flag to defer user min/max GT
frequency changes while the Wa_22019338487 frequency cap is in force, and
applies any deferred requests when restoring stashed frequencies.

> diff --git a/drivers/gpu/drm/xe/xe_guc_pc.c b/drivers/gpu/drm/xe/xe_guc_pc.c
> index 9558d7196a51a..ea47b0ded83a4 100644
> --- a/drivers/gpu/drm/xe/xe_guc_pc.c
> +++ b/drivers/gpu/drm/xe/xe_guc_pc.c

[ ... ]

> @@ -638,17 +639,26 @@ int xe_guc_pc_get_min_freq(struct xe_guc_pc *pc, u32 *freq)
>  
>  static int xe_guc_pc_set_min_freq_locked(struct xe_guc_pc *pc, u32 freq)
>  {
> -	int ret;
> -
>  	lockdep_assert_held(&pc->freq_lock);
>  
>  	/* Might be in the middle of a gt reset */
>  	if (!pc->freq_ready)
>  		return -EAGAIN;
>  
> -	ret = pc_set_min_freq(pc, freq);
> -	if (ret)
> -		return ret;
> +	/*
> +	 * While the Wa_22019338487 cap is active, defer applying the new min
> +	 * to GuC; xe_guc_pc_restore_stashed_freq() will apply it once the
> +	 * cap is removed.
> +	 */
> +	if (pc->mert_freq_cap_active) {
> +		if (!pc_freq_in_range(pc, freq))
> +			return -EINVAL;
> +	} else {
> +		int ret = pc_set_min_freq(pc, freq);
> +
> +		if (ret)
> +			return ret;
> +	}
>  
>  	pc->user_requested_min = freq;
>  
> @@ -707,17 +717,26 @@ int xe_guc_pc_get_max_freq(struct xe_guc_pc *pc, u32 *freq)
>  
>  static int xe_guc_pc_set_max_freq_locked(struct xe_guc_pc *pc, u32 freq)
>  {
> -	int ret;
> -
>  	lockdep_assert_held(&pc->freq_lock);
>  
>  	/* Might be in the middle of a gt reset */
>  	if (!pc->freq_ready)
>  		return -EAGAIN;
>  
> -	ret = pc_set_max_freq(pc, freq);
> -	if (ret)
> -		return ret;
> +	/*
> +	 * While the Wa_22019338487 cap is active, defer applying the new max
> +	 * to GuC; xe_guc_pc_restore_stashed_freq() will apply it once the
> +	 * cap is lifted.
> +	 */
> +	if (pc->mert_freq_cap_active) {
> +		if (!pc_freq_in_range(pc, freq))
> +			return -EINVAL;
> +	} else {
> +		int ret = pc_set_max_freq(pc, freq);
> +
> +		if (ret)
> +			return ret;
> +	}
>  
>  	pc->user_requested_max = freq;
>  
> @@ -1094,6 +1113,9 @@ static int pc_set_mert_freq_cap(struct xe_guc_pc *pc)
>  	if (!ret)
>  		ret = pc_set_max_freq(pc, min(pc->rp0_freq, pc_max_freq_cap(pc)));
>  
> +	if (!ret)
> +		pc->mert_freq_cap_active = true;
> +
>  	return ret;
>  }

[Severity: Medium]
Can pc->mert_freq_cap_active remain stuck at true if asynchronous GSC
firmware upload fails?

During xe_device_probe(), xe_gt_sanitize_freq() is a no-op while gsc_work()
is still uploading the GSC firmware (status XE_UC_FIRMWARE_LOADABLE).  If
gsc_upload() subsequently fails, gsc_upload_and_init() returns early before
calling xe_gt_sanitize_freq(), and gsc_work() transitions the firmware to
XE_UC_FIRMWARE_LOAD_FAIL without restoring the stashed frequencies:

drivers/gpu/drm/xe/xe_gsc.c:gsc_upload_and_init() {
	...
	ret = gsc_upload(gsc);
	...
	if (ret)
		return ret;

	xe_uc_fw_change_status(&gsc->fw, XE_UC_FIRMWARE_TRANSFERRED);

	/* GSC load is done, restore expected GT frequencies */
	xe_gt_sanitize_freq(gt);
	...
}

drivers/gpu/drm/xe/xe_gsc.c:gsc_work() {
	...
	if (actions & GSC_ACTION_FW_LOAD) {
		ret = gsc_upload_and_init(gsc);
		if (ret && ret != -EEXIST)
			xe_uc_fw_change_status(&gsc->fw, XE_UC_FIRMWARE_LOAD_FAIL);
	...
}

Because xe_guc_pc_restore_stashed_freq() is never called on that error
path, pc->mert_freq_cap_active stays true and all subsequent sysfs writes
to min_freq and max_freq only update pc->user_requested_min and
pc->user_requested_max without sending the new limits to GuC, leaving GT
frequency locked at the cap.  Calling xe_gt_sanitize_freq(gt) in gsc_work()
after setting XE_UC_FIRMWARE_LOAD_FAIL would lift the cap on load failure.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261118012453.626087-3-vinay.belgaumkar@intel.com?part=1

  parent reply	other threads:[~2026-11-18  2:01 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18  1:24 [PATCH v3 1/2] drm/xe: Disallow freq changes while mert cap is active Vinay Belgaumkar
2026-11-18  1:24 ` [PATCH v3 2/2] drm/xe: Add sysfs to disable RPe Vinay Belgaumkar
2026-11-18  1:50   ` sashiko-bot
2026-11-18  1:34 ` ✓ CI.KUnit: success for series starting with [v3,1/2] drm/xe: Disallow freq changes while mert cap is active Patchwork
2026-11-18  2:01 ` sashiko-bot [this message]
2026-11-18  2:10 ` ✗ Xe.CI.BAT: failure " 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=sashiko-outbox-219921@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=sashiko-reviews@lists.linux.dev \
    --cc=vinay.belgaumkar@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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.