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. On qca8k it's just anarchy. You can apply the same rule table to whatever led you want. They are all the same... OEM decide what to do (add white led, amber, green...) Most common configuration is 2 leds, one react with port 1000 and the other with port 100. But each led can be set to be powered on to any speed by enabling 10 100 and 1000 rule. Rejecting a configuration falls in the driver returning error on configure. > > 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. > Eh this is really bad, it was an idea to support blink internal for each event but I expected other switch had something different... But the generic function is still there... Wonder if we should consider adding generic naming like SLOW, NORMAL(?), FAST. And just assign them from the driver side? Permit to have some kind of configuration while also keeping it generic enough? Just as reference... The blink mode already had a compromise since qca8k can support a way to blink differently based on the link speed. -- Ansuel
