Re: [PATCH v2 net-next 0/2] net: dsa: mv88e6xxx: Improve indirect addressing performance

[Tobias Waldekranz <tobias@waldekranz.com>]

On Fri, Jan 28, 2022 at 17:10, Andrew Lunn <andrew@lunn.ch> wrote:
> On Fri, Jan 28, 2022 at 04:58:02PM +0100, Tobias Waldekranz wrote:
>> On Fri, Jan 28, 2022 at 14:10, David Laight <David.Laight@ACULAB.COM> wrote:
>> > From: Tobias Waldekranz
>> >> Sent: 28 January 2022 10:50
>> >> 
>> >> The individual patches have all the details. This work was triggered
>> >> by recent work on a platform that took 16s (sic) to load the mv88e6xxx
>> >> module.
>> >> 
>> >> The first patch gets rid of most of that time by replacing a very long
>> >> delay with a tighter poll loop to wait for the busy bit to clear.
>> >> 
>> >> The second patch shaves off some more time by avoiding redundant
>> >> busy-bit-checks, saving 1 out of 4 MDIO operations for every register
>> >> read/write in the optimal case.
>> >
>> > I don't think you should fast-poll for the entire timeout period.
>> > Much better to drop to a usleep_range() after the first 2 (or 3)
>> > reads fail.
>> 
>> You could, I suppose. Andrew, do you want a v3?
>
> You have i available, so it would be a simple change. So yes please.

Alright, v3 coming up.

> But saying that, it seems like if the switch does not complete within
> 2 polls, it is likely to be dead and we are about to start a cascade
> of failures. We probably don't care about a bit of CPU usage when the
> devices purpose in being has just stopped working.

Yeah, that's pretty much where my mind went as well.