[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20210701170712.GB24430@hoboy.vegasvil.org>
Date: Thu, 1 Jul 2021 10:07:12 -0700
From: Richard Cochran <richardcochran@...il.com>
To: Jonathan Lemon <jonathan.lemon@...il.com>
Cc: netdev@...r.kernel.org, kernel-team@...com
Subject: Re: [PATCH] ptp: Add PTP_CLOCK_EXTTSUSR internal ptp_event
On Thu, Jul 01, 2021 at 09:15:55AM -0700, Jonathan Lemon wrote:
> static void enqueue_external_timestamp(struct timestamp_event_queue *queue,
> struct ptp_clock_event *src)
> {
> struct ptp_extts_event *dst;
> unsigned long flags;
> s64 seconds;
> u32 remainder;
>
> seconds = div_u64_rem(src->timestamp, 1000000000, &remainder);
>
>
> It seems like there should be a way to use pps_times here instead
> of needing to convert back and forth.
You could re-factor that to have two callers, with the part that
enqueues in a shared helper function. The only reason the API has a
64 bit word instead of a timespec is that many, but not all drivers
use timecounter_cyc2time() or similar to calculate the time stamp.
But the ptp_clock_event is really meant to be polymorphic, with
pps_times only set for traditional NTP PPS events (activated by the
PTP_ENABLE_PPS ioctl).
* struct ptp_clock_event - decribes a PTP hardware clock event
*
* @type: One of the ptp_clock_events enumeration values.
* @index: Identifies the source of the event.
* @timestamp: When the event occurred (%PTP_CLOCK_EXTTS only).
* @pps_times: When the event occurred (%PTP_CLOCK_PPSUSR only).
The PTP_CLOCK_EXTTS is different. It is meant for generic time
stamping of external signals, activated by the PTP_EXTTS_REQUEST
ioctl.
I'm not sure which type is better suited to your HW.
Thanks,
Richard
Powered by blists - more mailing lists