Re: Hung tasks due to a AB-BA deadlock between the leds_list_lock rwsem and the rtnl mutex

[Hans de Goede]

Hi,

On 5/31/24 3:29 PM, Andrew Lunn wrote:
>>> drivers/net/ethernet/realtek/r8169_leds.c:	led_cdev->hw_control_trigger = "netdev";
>>> drivers/net/ethernet/realtek/r8169_leds.c:	led_cdev->hw_control_trigger = "netdev";
>>> drivers/net/ethernet/intel/igc/igc_leds.c:	led_cdev->hw_control_trigger = "netdev";
>>> drivers/net/dsa/qca/qca8k-leds.c:		port_led->cdev.hw_control_trigger = "netdev";
>>> drivers/net/phy/phy_device.c:		cdev->hw_control_trigger = "netdev";
>>
>> Well those drivers combined, esp. with the generic phy_device in there
>> does mean that the ledtrig-netdev module now gets loaded on a whole lot
>> of x86 machines where before it would not.
> 
> phy_device will only do something if there is the needed Device Tree
> properties. Given that very few systems use DT on x86, that should not
> be an issue.

That is good to know.

> So only x86 systems with r8169 and igc should have the
> trigger module loaded because of this.

Those are very popular NICs though, so that is still a lot of
systems.

> It would be good to understand
> why other systems have the trigger loaded.

Actually my system has a RTL8168 ethernet NIC so the netdev trigger
getting loaded there is expected.

> However, as you say, this
> will not fix the underlying deadlock, it will just limit it to systems with r8169
> and igc...

Right, given on the above discussion I believe that it likely already
is limited to systems with Realtek r8169 or Intel i225 / i226 NICs.

Regards,

Hans