> -----Original Message----- > From: Petr Machata [mailto:pe...@mellanox.com] > Sent: Monday, June 10, 2019 7:09 AM > To: linuxptp-devel@lists.sourceforge.net > Subject: [Linuxptp-devel] [PATCH] sk: Recognize HWTSTAMP_FILTER_SOME > > struct hwtstamp_config.rx_filter passed to SIOCSHWTSTAMP can be updated by > the kernel with the value of HWTSTAMP_FILTER_SOME. That indicates that all > requested packets will be timestamped, and some others as well. > > Update hwts_init() to recognize this as a valid response, instead of > rejecting it as mismatch. > > Signed-off-by: Petr Machata <pe...@mellanox.com> > --- > sk.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/sk.c b/sk.c > index 93ba77a..416d784 100644 > --- a/sk.c > +++ b/sk.c > @@ -105,7 +105,8 @@ static int hwts_init(int fd, const char *device, int > rx_filter, > if (cfg.tx_type != tx_type || > (cfg.rx_filter != rx_filter && > cfg.rx_filter != rx_filter2 && > - cfg.rx_filter != HWTSTAMP_FILTER_ALL)) { > + cfg.rx_filter != HWTSTAMP_FILTER_ALL && > + cfg.rx_filter != HWTSTAMP_FILTER_SOME)) { > pr_debug("tx_type %d not %d", cfg.tx_type, tx_type); > pr_debug("rx_filter %d not %d or %d", cfg.rx_filter, rx_filter, > rx_filter2);
You can't accept HWTSTAMP_FILTER_SOME from the "get hwtstamp config" ioctl, because it's unclear what packets "SOME" actually refers to. I.e. one caller could set it to filter V1 packets, the driver could report SOME, and then ptp4l could "accept" this filter even though it's not valid. You can accept it in response to the "set hwtstamp config", I suppose, but it is a tad confusing. I never really liked that return value, and just wish the API had been "you may timestamp more than requested without reporting it if there is no valid filter matching what you timestamp". Thanks, Jake _______________________________________________ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
