Re: [PATCH net RESEND 1/2] net: macb: Preserve one-step mode on rejected timestamp requests

[kimwooseok]

Hi Paolo,

Thanks for forwarding the review. Taking the items in turn:

TX_OFF / OSSMODE: this is pre-existing and independent of what this
patch fixes. The original code also left OSSMODE untouched on
HWTSTAMP_TX_OFF; this patch only addresses rejected (-ERANGE)
requests changing hardware state, while TX_OFF is an accepted request.

I propose handling TX_OFF in a separate follow-up patch, keeping this
series focused on the two reported bugs. It is a distinct defect and
should have its own changelog and Fixes: tag for stable backporting.
I will verify the TX_OFF fix on the Pi 5 and post it as a follow-up
once this series lands, since it touches the same lines.

NCR read-modify-write locking, the gem_get_ts_info() rx_filters bitmap,
and re-applying tstamp_config on resume: as the review itself notes,
these predate the series and are not made worse by it. I'd handle
them separately (net-next) rather than widen these Fixes patches.

SRTSM: NCR.SRTSM only makes the MAC replace the received FCS with the RX
timestamp nanoseconds in memory; with DRFCS set it has no effect on the
descriptor-based timestamps this driver uses, so leaving it set does not
affect timestamping. It is also unrelated to the two PTPv1 filters
remapped in patch 2.

Thanks,
Wooseok