linux-kernel - Re: [PATCH] pinctrl: amd: Take suspend type into consideration which pins are non-wake

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <73486a76-d24d-40f0-ac32-792b80aca6d9@amd.com>
Date: Sun, 5 Jan 2025 21:18:53 -0600
From: Mario Limonciello <mario.limonciello@....com>
To: "Maciej S. Szmigiero" <mail@...iej.szmigiero.name>,
 Basavaraj Natikar <Basavaraj.Natikar@....com>,
 Shyam Sundar S K <Shyam-sundar.S-k@....com>
Cc: Linus Walleij <linus.walleij@...aro.org>, linux-gpio@...r.kernel.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH] pinctrl: amd: Take suspend type into consideration which
 pins are non-wake

On 12/29/2024 10:42, Maciej S. Szmigiero wrote:
> Some laptops have pins which are a wake source for S0i3/S3 but which
> aren't a wake source for S4/S5 and which cause issues when left unmasked
> during hibernation (S4).
> 
> For example HP EliteBook 855 G7 has pin #24 that causes instant wakeup
> (hibernation failure) if left unmasked (it is a wake source only for
> S0i3/S3).

On your machine do you know what pin 24 is connected to?

If not, can you run 
https://gitlab.freedesktop.org/drm/amd/-/blob/master/scripts/amd_s2idle.py 
and share the text report it generates to me?

I'll see if I can make sense in that report what it's most likely 
connected to.

Just want to make sure we're not papering over an issue in another 
component by making this change.

> 
> Fix this by considering a pin a wake source only if it is marked as one
> for the current suspend type (S0i3/S3 vs S4/S5).
> 
> Since I'm not sure if Z-wake pins should be included in either suspend
> category I excluded them from both, so pins with only the Z-wake flag set
> are treated as non-wake pins.

Z only makes sense at runtime.  As long as it's restored to previous 
value after exiting suspend or hibernate that should be totally fine.

> 
> Fixes: 2fff0b5e1a6b ("pinctrl: amd: Mask non-wake source pins with interrupt enabled at suspend")
> Signed-off-by: Maciej S. Szmigiero <mail@...iej.szmigiero.name>
> ---
>   drivers/pinctrl/pinctrl-amd.c | 27 +++++++++++++++++++++------
>   drivers/pinctrl/pinctrl-amd.h |  7 +++----
>   2 files changed, 24 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/pinctrl/pinctrl-amd.c b/drivers/pinctrl/pinctrl-amd.c
> index fff6d4209ad5..072d44b0fc8c 100644
> --- a/drivers/pinctrl/pinctrl-amd.c
> +++ b/drivers/pinctrl/pinctrl-amd.c
> @@ -908,12 +908,13 @@ static bool amd_gpio_should_save(struct amd_gpio *gpio_dev, unsigned int pin)
>   	return false;
>   }
>   
> -static int amd_gpio_suspend(struct device *dev)
> +static int amd_gpio_suspend_common(struct device *dev, bool is_s03)
>   {
>   	struct amd_gpio *gpio_dev = dev_get_drvdata(dev);
>   	struct pinctrl_desc *desc = gpio_dev->pctrl->desc;
>   	unsigned long flags;
>   	int i;
> +	u32 wake_mask = is_s03 ? WAKE_SOURCE_S03 : WAKE_SOURCE_S4;
>   
>   	for (i = 0; i < desc->npins; i++) {
>   		int pin = desc->pins[i].number;
> @@ -925,11 +926,11 @@ static int amd_gpio_suspend(struct device *dev)
>   		gpio_dev->saved_regs[i] = readl(gpio_dev->base + pin * 4) & ~PIN_IRQ_PENDING;
>   
>   		/* mask any interrupts not intended to be a wake source */
> -		if (!(gpio_dev->saved_regs[i] & WAKE_SOURCE)) {
> +		if (!(gpio_dev->saved_regs[i] & wake_mask)) {
>   			writel(gpio_dev->saved_regs[i] & ~BIT(INTERRUPT_MASK_OFF),
>   			       gpio_dev->base + pin * 4);
> -			pm_pr_dbg("Disabling GPIO #%d interrupt for suspend.\n",
> -				  pin);
> +			pm_pr_dbg("Disabling GPIO #%d interrupt for %s suspend.\n",
> +				  pin, is_s03 ? "S0idle3/S3" : "S4/S5");
>   		}
>   
>   		raw_spin_unlock_irqrestore(&gpio_dev->lock, flags);
> @@ -938,6 +939,16 @@ static int amd_gpio_suspend(struct device *dev)
>   	return 0;
>   }
>   
> +static int amd_gpio_suspend_s03(struct device *dev)
> +{
> +	return amd_gpio_suspend_common(dev, true);
> +}
> +
> +static int amd_gpio_suspend_s45(struct device *dev)
> +{
> +	return amd_gpio_suspend_common(dev, false);
> +}
> +
>   static int amd_gpio_resume(struct device *dev)
>   {
>   	struct amd_gpio *gpio_dev = dev_get_drvdata(dev);
> @@ -961,8 +972,12 @@ static int amd_gpio_resume(struct device *dev)
>   }
>   
>   static const struct dev_pm_ops amd_gpio_pm_ops = {
> -	SET_LATE_SYSTEM_SLEEP_PM_OPS(amd_gpio_suspend,
> -				     amd_gpio_resume)
> +	.suspend_late = amd_gpio_suspend_s03,
> +	.resume_early = amd_gpio_resume,
> +	.freeze_late = amd_gpio_suspend_s45,
> +	.thaw_early = amd_gpio_resume,
> +	.poweroff_late = amd_gpio_suspend_s45,
> +	.restore_early = amd_gpio_resume,
>   };
>   #endif
>   
> diff --git a/drivers/pinctrl/pinctrl-amd.h b/drivers/pinctrl/pinctrl-amd.h
> index 667be49c3f48..8bf9f410d7fb 100644
> --- a/drivers/pinctrl/pinctrl-amd.h
> +++ b/drivers/pinctrl/pinctrl-amd.h
> @@ -80,10 +80,9 @@
>   #define FUNCTION_MASK		GENMASK(1, 0)
>   #define FUNCTION_INVALID	GENMASK(7, 0)
>   
> -#define WAKE_SOURCE	(BIT(WAKE_CNTRL_OFF_S0I3) | \
> -			 BIT(WAKE_CNTRL_OFF_S3)   | \
> -			 BIT(WAKE_CNTRL_OFF_S4)   | \
> -			 BIT(WAKECNTRL_Z_OFF))
> +#define WAKE_SOURCE_S03 (BIT(WAKE_CNTRL_OFF_S0I3) | \
> +			 BIT(WAKE_CNTRL_OFF_S3))
> +#define WAKE_SOURCE_S4  BIT(WAKE_CNTRL_OFF_S4)

Since s03 doesn't make sense and s0i3 is wrong for s3 and s3 is wrong 
for s0i3 as a personal preference I would just call it

WAKE_SOURCE_SUSPEND
WAKE_SOURCE_HIBERNATE


>   
>   struct amd_function {
>   	const char *name;