linux-kernel - Re: [PATCH v2] net: phy: marvell: Honor phy LED set by system firmware on a Dell hardware

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <ceead395-a641-7413-8d9f-77879dcaa36e@gmail.com>
Date:   Fri, 21 Jan 2022 08:49:39 +0100
From:   Heiner Kallweit <hkallweit1@...il.com>
To:     Andrew Lunn <andrew@...n.ch>,
        Kai-Heng Feng <kai.heng.feng@...onical.com>
Cc:     linux@...linux.org.uk, "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] net: phy: marvell: Honor phy LED set by system
 firmware on a Dell hardware

On 20.01.2022 14:46, Andrew Lunn wrote:
> On Thu, Jan 20, 2022 at 01:19:29PM +0800, Kai-Heng Feng wrote:
>> BIOS on Dell Edge Gateway 3200 already makes its own phy LED setting, so
>> instead of setting another value, keep it untouched and restore the saved
>> value on system resume.
> 
> Please split this patch into two:
> 
> Don't touch the LEDs
> 
> Save and restore the LED configuration over suspend/resume.
> 
>> -static void marvell_config_led(struct phy_device *phydev)
>> +static int marvell_find_led_config(struct phy_device *phydev)
>>  {
>> -	u16 def_config;
>> -	int err;
>> +	int def_config;
>> +
>> +	if (phydev->dev_flags & PHY_USE_FIRMWARE_LED) {
>> +		def_config = phy_read_paged(phydev, MII_MARVELL_LED_PAGE, MII_PHY_LED_CTRL);
>> +		return def_config < 0 ? -1 : def_config;
> 
> What about the other two registers which configure the LEDs?
> 
> Since you talked about suspend/resume, does this machine support WoL?
> Is the BIOS configuring LED2 to be used as an interrupt when WoL is
> enabled in the BIOS? Do you need to save/restore that configuration
> over suspend/review? And prevent the driver from changing the
> configuration?
> 
>> +static const struct dmi_system_id platform_flags[] = {
>> +	{
>> +		.matches = {
>> +			DMI_MATCH(DMI_SYS_VENDOR, "Dell EMC"),
>> +			DMI_MATCH(DMI_PRODUCT_NAME, "Edge Gateway 3200"),
>> +		},
>> +		.driver_data = (void *)PHY_USE_FIRMWARE_LED,
>> +	},
> 
> This needs a big fat warning, that it will affect all LEDs for PHYs
> which linux is driving, on that machine. So PHYs on USB dongles, PHYs
> in SFPs, PHYs on plugin PCIe card etc.
> 
> Have you talked with Dells Product Manager and do they understand the
> implications of this? 
> 
>> +	{}
>> +};
>> +
>>  /**
>>   * phy_attach_direct - attach a network device to a given PHY device pointer
>>   * @dev: network device to attach
>> @@ -1363,6 +1379,7 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
>>  	struct mii_bus *bus = phydev->mdio.bus;
>>  	struct device *d = &phydev->mdio.dev;
>>  	struct module *ndev_owner = NULL;
>> +	const struct dmi_system_id *dmi;
>>  	bool using_genphy = false;
>>  	int err;
>>  
>> @@ -1443,6 +1460,10 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
>>  			phydev_err(phydev, "error creating 'phy_standalone' sysfs entry\n");
>>  	}
>>  
>> +	dmi = dmi_first_match(platform_flags);
>> +	if (dmi)
>> +		phydev->dev_flags |= (u32)dmi->driver_data;
> 
> Please us your new flag directly. We don't want this abused to pass
> any old flag to the PHY.
> 
>> +
>>  /**
>>   * struct phy_device - An instance of a PHY
>>   *
>> @@ -663,6 +665,7 @@ struct phy_device {
>>  
>>  	struct phy_led_trigger *led_link_trigger;
>>  #endif
>> +	int led_config;
> 
> You cannot put this here because you don't know how many registers are
> used to hold the configuration. Marvell has 3, other drivers can have
> other numbers. The information needs to be saved into the drivers on
> priv structure.
> 
>>  
>>  	/*
>>  	 * Interrupt number for this PHY
>> @@ -776,6 +779,12 @@ struct phy_driver {
>>  	 */
>>  	int (*config_init)(struct phy_device *phydev);
>>  
>> +	/**
>> +	 * @config_led: Called to config the PHY LED,
>> +	 * Use the resume flag to indicate init or resume
>> +	 */
>> +	void (*config_led)(struct phy_device *phydev, bool resume);
> 
> I don't see any need for this.
> 
>   Andrew

I had a look at the history of LED settings in the Marvell PHY driver:

In marvell_config_aneg() we do
phy_write(phydev, MII_M1111_PHY_LED_CONTROL, MII_M1111_PHY_LED_DIRECT);

This originates from 2007: 76884679c644 ("phylib: Add support for Marvell 88e1111S and 88e1145")
and sets the LED control to the reset default (for no obvious reason).
Especially strange is that this is done in config_aneg.
Only non-LED bit here is bit 11 that forces the interrupt pin to assert.

Simply wrong is that register MII_M1111_PHY_LED_CONTROL (reg 24, page 0)
is written also on other chip versions (like 88E1112) where it's not
defined and marked as reserved.
I think we should remove this code.

Then we set some LED defaults in marvell_config_led().
This also originates from > 10yrs ago:
140bc9290328 ("phylib: Basic support for the M88E1121R Marvell chip")
Again there's no obvious reason.

Last but not least we have a93f7fe13454 ("net: phy: marvell: add new default led configure for m88e151x")
This adds a flag to set the PHY LED mode from hns3 MAC driver.
Intention of this patch is to set the LED mode for specific boards.
The chosen approach doesn't seem to be the best. As it's meant to be
board-specific maybe better the reg-init DT property would have been
used.

I'd say we can remove all LED config code and accept whatever boot
loader or BIOS set.

Heiner