Re: [PATCH net-next 5/5] igc: Export LEDs

From: Kurt Kanzenbach <kurt@linutronix.de>
To: Andrew Lunn <andrew@lunn.ch>
Cc: Tony Nguyen <anthony.l.nguyen@intel.com>,
	davem@davemloft.net, kuba@kernel.org, netdev@vger.kernel.org,
	sasha.neftin@intel.com, vitaly.lifshits@intel.com,
	vinicius.gomes@intel.com,
	Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
	Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Subject: Re: [PATCH net-next 5/5] igc: Export LEDs
Date: Tue, 20 Jul 2021 10:54:24 +0200	[thread overview]
Message-ID: <87y2a19xhb.fsf@kurt> (raw)
In-Reply-To: <YPV6+PQq1fvH8aSy@lunn.ch>

[-- Attachment #1: Type: text/plain, Size: 2019 bytes --]

Hi Andrew,

On Mon Jul 19 2021, Andrew Lunn wrote:
> On Mon, Jul 19, 2021 at 08:06:41AM +0200, Kurt Kanzenbach wrote:
>> There are different modes such as ON, OFF, LINK established, LINK
>> activity, PAUSED ... Blinking is controlled by a different register.
>> 
>> Are there better ways to export this?
>
> As i said in another email, using LED triggers. For simple link speed
> indication, take a look at CONFIG_LED_TRIGGER_PHY. This is purely
> software, and probably not what you want.

Here's my use case/reasoning behind this patch: Upon reception of a
certain Ethernet frame, the LEDs should blink for a certain period of
time. Afterwards the default behavior should be restored. The blinking
can be done in hardware, but only for a fixed period. I needed a
different period.

Therefore, I've exported these as regular LEDs to toggle the brightness
from user space directly.

> The more complex offload of to LED control to hardware in the PHY
> subsystem it still going around and around. The basic idea is agreed,
> just not the implementation.  However, most of the implementation is
> of no help to you, since Intel drivers ignore the kernel PHY drivers
> and do their own. But the basic idea and style of user space API
> should be kept the same. So take a look at the work Marek Behún has
> been doing, e.g.
>
> https://lwn.net/Articles/830947/
>
> and a more recent version
>
> https://lore.kernel.org/netdev/20210602144439.4d20b295@dellmb/T/#m4e97c9016597fb40849c104c7c0e3bc10d5c1ff5

Thanks for the pointer.

>
> Looking at the basic idea of using LED triggers and offloading
> them. Please also try to make us of generic names for these triggers,
> so the PHY subsystem might also use the same or similar names when it
> eventually gets something merged.
>
> Please also Cc: The LED maintainers and LED list. Doing that from the
> start would of avoided this revert, since you would of get earlier
> feedback about this.

Yeah, noted.

Thanks,
Kurt

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 861 bytes --]