Re: [PATCH 0/4] RFC CPSW switchdev mode
From: Grygorii Strashko <grygorii.strashko@ti.com>
Date: 2018-06-05 21:42:24
From: Grygorii Strashko <grygorii.strashko@ti.com>
Date: 2018-06-05 21:42:24
On 06/05/2018 04:28 PM, Andrew Lunn wrote:
quoted
I hope you are right - question is always in number of available options and which one to select - and, most important, explain it to the end user :(The end customer being ptp4linux? At least for Marvell switches, it is happy about everything except that the switch is a bit slow, so we need to modify some of the time outs in the configuration file.quoted
For example: phc_index is returned as part of .get_ts_info() = cpsw_get_ts_info(), so which intf should return phc_index?It is not a 1:1 relationship. See: https://elixir.bootlin.com/linux/latest/source/drivers/net/dsa/mv88e6xxx/hwtstamp.c#L61 All interfaces return the same index. In fact, for a switch, having a PHC per port would be odd. That would mean you need to sync the PHCs in order to act as a boundary clock.
PHC only one, but hw timestamping blocks are per port. -- regards, -grygorii