Devicetree
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Louis-Alexis Eyraud" <louisalexis.eyraud@collabora.com>
Cc: linux-phy@lists.infradead.org, devicetree@vger.kernel.org,
	linux-scsi@vger.kernel.org, robh@kernel.org, conor+dt@kernel.org,
	vkoul@kernel.org, neil.armstrong@linaro.org, olteanv@gmail.com
Subject: Re: [PATCH v15 14/23] scsi: ufs: mediatek: Use the common PHY framework
Date: Thu, 19 Nov 2026 15:26:41 +0000	[thread overview]
Message-ID: <sashiko-outbox-223382@kernel.org> (raw)
In-Reply-To: <20261119-mt8196-ufs-v15-14-aae1eb0382e8@collabora.com>

Thank you for your contribution! Sashiko AI review found 3 potential
issues to consider:
- [High] scsi: ufs: mediatek: premature phy_exit() and power leak in
  ufs_mtk_exit()
- [High] scsi: ufs: mediatek: missing error rollback in
  ufs_mtk_runtime_suspend()
- [Medium] scsi: ufs: mediatek: PHY power leak and error in
  ufs_mtk_runtime_resume()
--

commit dd9f3229e4978dc4fda07cd42dec5399df0226ee
Author: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>

scsi: ufs: mediatek: Use the common PHY framework

This commit switches the MediaTek UFS driver from manual OF PHY device
handling in probe and runtime PM to acquiring and managing the PHY via
the common PHY framework in the ufshcd variant callbacks.

> diff --git a/drivers/ufs/host/ufs-mediatek.c b/drivers/ufs/host/ufs-mediatek.c
> index fec0fb24e7a2c..a78e015e329a5 100644
> --- a/drivers/ufs/host/ufs-mediatek.c
> +++ b/drivers/ufs/host/ufs-mediatek.c

[ ... ]

> @@ -1321,11 +1291,22 @@ static int ufs_mtk_init(struct ufs_hba *hba)
>  
>  	return 0;
>  
> +out_phy_exit:
> +	phy_exit(host->mphy);
>  out_variant_clear:
>  	ufshcd_set_variant(hba, NULL);
>  	return err;
>  }
>  
> +static void ufs_mtk_exit(struct ufs_hba *hba)
> +{
> +	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
> +
> +	ufs_mtk_mphy_power_on(hba, false);
> +
> +	phy_exit(host->mphy);
> +}

[Severity: High]
Does ufs_mtk_exit() call phy_exit(host->mphy) while a second
phy_power_on() reference is still held?

ufs_mtk_init() acquires two phy_power_on() references on host->mphy: one
in ufs_mtk_mphy_power_on(hba, true) and another in
ufs_mtk_setup_clocks(hba, true, POST_CHANGE) via
ufs_mtk_pwr_ctrl(hba, true).  During teardown, ufshcd_hba_exit() invokes
ufshcd_vops_exit(hba) before ufshcd_setup_clocks(hba, false):

drivers/ufs/core/ufshcd.c:ufshcd_hba_exit() {
	...
		ufshcd_exit_clk_gating(hba);
		if (hba->eh_wq)
			destroy_workqueue(hba->eh_wq);
		ufs_debugfs_hba_exit(hba);
		ufshcd_vops_exit(hba);
		ufshcd_setup_vreg(hba, false);
		ufshcd_setup_clocks(hba, false);
	...
}

In ufs_mtk_exit(), ufs_mtk_mphy_power_on(hba, false) drops only one
phy_power_on() reference (leaving mphy->power_count at 1) and turns off
the 0.9V supply before phy_exit(host->mphy) runs.

On driver removal, ufshcd_remove() leaves hba->uic_link_state as
UIC_LINK_ACTIVE_STATE, so the subsequent ufshcd_setup_clocks(hba, false)
skips ufs_mtk_pwr_ctrl(hba, false) and permanently leaks the remaining
phy_power_on() reference, PHY clocks, and runtime PM reference.  On an
early ufshcd_init() failure while the link is still off,
ufshcd_setup_clocks(hba, false) instead calls phy_power_off(host->mphy)
and accesses M-PHY registers after phy_exit() has run and the 0.9V supply
has been disabled.

[ ... ]

> @@ -2522,9 +2467,8 @@ static int ufs_mtk_system_resume(struct device *dev)
>  static int ufs_mtk_runtime_suspend(struct device *dev)
>  {
>  	struct ufs_hba *hba = dev_get_drvdata(dev);
> -	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
>  	struct arm_smccc_res res;
> -	int ret = 0;
> +	int ret;
>  
>  	ret = ufshcd_runtime_suspend(dev);
>  	if (ret)
> @@ -2535,8 +2479,13 @@ static int ufs_mtk_runtime_suspend(struct device *dev)
>  	if (ufs_mtk_is_rtff_mtcmos(hba))
>  		ufs_mtk_mtcmos_ctrl(false, res);
>  
> -	if (host->phy_dev)
> -		pm_runtime_put_sync(host->phy_dev);
> +	if (!ufshcd_is_link_active(hba)) {
> +		ret = ufs_mtk_mphy_power_on(hba, false);
> +		if (ret) {
> +			dev_err(dev, "Failed to power off PHY: %pe\n", ERR_PTR(ret));
> +			return ret;
> +		}
> +	}
>  
>  	return 0;
>  }

[Severity: High]
If ufs_mtk_mphy_power_on(hba, false) fails here, does returning an error
without rolling back leave the controller powered down while runtime PM
considers dev active?

At this point ufshcd_runtime_suspend(dev), ufs_mtk_dev_vreg_set_lpm(hba,
true), and ufs_mtk_mtcmos_ctrl(false, res) have already gated clocks,
disabled interrupts, placed device regulators into LPM, and turned off
MTCMOS.  When a runtime_suspend callback returns an error, rpm_suspend()
aborts the suspend and leaves the device in RPM_ACTIVE, so subsequent I/O
or register accesses will touch unpowered hardware without running
ufs_mtk_runtime_resume().  Also, powering off the PHY after
ufs_mtk_mtcmos_ctrl(false, res) runs in the same order as
ufs_mtk_runtime_resume() rather than the reverse order.

> @@ -2544,14 +2493,17 @@ static int ufs_mtk_runtime_suspend(struct device *dev)
>  static int ufs_mtk_runtime_resume(struct device *dev)
>  {
>  	struct ufs_hba *hba = dev_get_drvdata(dev);
> -	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
>  	struct arm_smccc_res res;
> +	int ret;
>  
>  	if (ufs_mtk_is_rtff_mtcmos(hba))
>  		ufs_mtk_mtcmos_ctrl(true, res);
>  
> -	if (host->phy_dev)
> -		pm_runtime_get_sync(host->phy_dev);
> +	ret = ufs_mtk_mphy_power_on(hba, true);
> +	if (ret) {
> +		dev_err(dev, "Failed to power on PHY: %pe\n", ERR_PTR(ret));
> +		return ret;
> +	}
>  
>  	ufs_mtk_dev_vreg_set_lpm(hba, false);
>  

[Severity: Medium]
Can powering on the M-PHY in ufs_mtk_runtime_resume() leave the PHY and
0.9V supply enabled across system suspend?

If hba->dev runtime-resumes while the ufs_device_wlun child remains
runtime-suspended, ufs_mtk_runtime_resume() sets host->mphy_powered_on
and enables the PHY.  If system suspend then runs before hba->dev
autosuspends (when ufshcd_rpm_ok_for_spm() is true), ufshcd_wl_suspend()
skips ufs_mtk_suspend(), and ufs_mtk_system_suspend() does not call
ufs_mtk_mphy_power_on(hba, false):

drivers/ufs/host/ufs-mediatek.c:ufs_mtk_system_suspend() {
	...
	ret = ufshcd_system_suspend(dev);
	if (ret)
		goto out;

	if (pm_runtime_suspended(hba->dev))
		goto out;

	ufs_mtk_dev_vreg_set_lpm(hba, true);

	if (ufs_mtk_is_rtff_mtcmos(hba))
		ufs_mtk_mtcmos_ctrl(false, res);
	...
}

This leaves the M-PHY and 0.9V regulator powered on during system sleep
while MTCMOS and clocks are off.  In addition, if
ufs_mtk_mphy_power_on(hba, true) or ufshcd_runtime_resume(dev) fails in
ufs_mtk_runtime_resume(), the preceding MTCMOS, PHY, and regulator state
changes are not unwound while dev remains RPM_SUSPENDED.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261119-mt8196-ufs-v15-0-aae1eb0382e8@collabora.com?part=14

  reply	other threads:[~2026-11-19 15:26 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-19 15:05 [PATCH v15 00/23] MediaTek UFS Cleanup and MT8196 Enablement Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 01/23] dt-bindings: phy: Add mediatek,mt8196-ufsphy variant Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 02/23] dt-bindings: ufs: mediatek,ufs: Complete the binding Louis-Alexis Eyraud
2026-11-19 15:25   ` sashiko-bot
2026-11-19 15:05 ` [PATCH v15 03/23] dt-bindings: ufs: mediatek,ufs: Add mt8196 variant Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 04/23] scsi: ufs: mediatek: Move MTK_SIP_UFS_CONTROL to mtk_sip_svc.h Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 05/23] phy: mediatek: ufs: Add support for resets Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 06/23] scsi: ufs: mediatek: Rework resets Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 07/23] scsi: ufs: mediatek: Rework 0.9V regulator Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 08/23] scsi: ufs: mediatek: Add dual 0.9V supply support Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 09/23] scsi: ufs: mediatek: Rework init function Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 10/23] scsi: ufs: mediatek: Rework the crypt-boost stuff Louis-Alexis Eyraud
2026-11-19 15:05 ` [PATCH v15 11/23] scsi: ufs: mediatek: Handle misc host voltage regulators Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 12/23] scsi: ufs: mediatek: Remove undocumented downstream reset cruft Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 13/23] scsi: ufs: mediatek: Remove vendor kernel quirks cruft Louis-Alexis Eyraud
2026-11-19 15:17   ` sashiko-bot
2026-11-19 15:06 ` [PATCH v15 14/23] scsi: ufs: mediatek: Use the common PHY framework Louis-Alexis Eyraud
2026-11-19 15:26   ` sashiko-bot [this message]
2026-11-19 15:06 ` [PATCH v15 15/23] scsi: ufs: mediatek: Remove mediatek,ufs-broken-rtc property Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 16/23] scsi: ufs: mediatek: Rework _ufs_mtk_clk_scale error paths Louis-Alexis Eyraud
2026-11-19 15:19   ` sashiko-bot
2026-11-19 15:06 ` [PATCH v15 17/23] scsi: ufs: mediatek: Clean up logging prints Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 18/23] scsi: ufs: mediatek: Rework ufs_mtk_wait_idle_state Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 19/23] scsi: ufs: mediatek: Don't acquire dvfsrc-vcore twice Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 20/23] scsi: ufs: mediatek: Rework hardware version reading Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 21/23] scsi: ufs: mediatek: Remove ret local from link_startup_notify Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 22/23] scsi: ufs: mediatek: Remove undocumented "clk-scale-up-vcore-min" Louis-Alexis Eyraud
2026-11-19 15:06 ` [PATCH v15 23/23] scsi: ufs: mediatek: Add MT8196 compatible, update copyright Louis-Alexis Eyraud

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-223382@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=linux-phy@lists.infradead.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=louisalexis.eyraud@collabora.com \
    --cc=neil.armstrong@linaro.org \
    --cc=olteanv@gmail.com \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    --cc=vkoul@kernel.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