Re: [PATCH v7 09/11] leds: trigger: netdev: add additional hardware only triggers

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Dec 15, 2022 at 05:35:47PM +0000, Russell King (Oracle) wrote:
> On Thu, Dec 15, 2022 at 12:54:36AM +0100, Christian Marangi wrote:
> > Add additional hardware only triggers commonly supported by switch LEDs.
> > 
> > Additional modes:
> > link_10: LED on with link up AND speed 10mbps
> > link_100: LED on with link up AND speed 100mbps
> > link_1000: LED on with link up AND speed 1000mbps
> > half_duplex: LED on with link up AND half_duplex mode
> > full_duplex: LED on with link up AND full duplex mode
> 
> Looking at Marvell 88e151x, I don't think this is usable there.
> We have the option of supporting link_1000 on one of the LEDs,
> link_100 on another, and link_10 on the other. It's rather rare
> for all three leds to be wired though.

The 88e151x will need to enumerate what it actually supports from the
above list, per LED. I also think we can carefully expand the list
above, adding a few more modes. We just need to ensure what is added
is reasonably generic, modes we expect multiple PHY to support. What
we need to avoid is adding every single mode a PHY supports, but no
other PHY has.

> This is also a PHY where "activity" mode is supported (illuminated
> or blinking if any traffic is transmitted or received) but may not
> support individual directional traffic in hardware. However, it
> does support forcing the LED on or off, so software mode can handle
> those until the user selects a combination of modes that are
> supported in the hardware.
> 
> > Additional blink interval modes:
> > blink_2hz: LED blink on any even at 2Hz (250ms)
> > blink_4hz: LED blink on any even at 4Hz (125ms)
> > blink_8hz: LED blink on any even at 8Hz (62ms)
> 
> This seems too restrictive. For example, Marvell 88e151x supports
> none of these, but does support 42, 84, 170, 340, 670ms.

I would actually drop this whole idea of being able to configure the
blink period. It seems like it is going to cause problems. I expect
most PHYs actual share the period across multiple LEDs, which you
cannot easily model here.

So i would have the driver hard coded to pick a frequency at thats' it.

   Andrew



[Index of Archives]     [Kernel Newbies]     [Security]     [Netfilter]     [Bugtraq]     [Linux FS]     [Yosemite Forum]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Video 4 Linux]     [Device Mapper]     [Linux Resources]

  Powered by Linux