Ali Gholami Rudi wrote: > "Ronciak, John" <[email protected]> wrote: >> The enabling and/or disabling takes place during configuration which >> is based on the actual HW the driver is loading on. So I think the >> current code is correct. The HW just won't timestamp at all if it >> can't or can't be put into that mode. Did you read the datasheet on >> how the actual HW works regarding timestamping? While "all" can be > > I had a look at section 7.10 and 8.17 of OpenSDM_82576-DS-2p0.pdf, > but I might have overlooked something. > >> set for the 82576 HW it can't really do it as there need to be a >> register read per packet which can't happen in the time when another > > Does kernel igb driver support one of those hardwares? Jesse said > 82575/576 does not support HWTSTAMP_FILTER_ALL, and AFAICS igb module > handles those models, only. >
If you look in the code you will notice that the 82575 does not support any timestamping. The 82576 supports timesync but only has one register to store the value so as a result it makes it difficult to get timesync per packet. >> frame arrives. This will result in all packets arriving being >> timestamped with the same timestamp until the register is read. So >> all are being timestamped, just not the way you would expect. That >> is why it is clearly called out in the HW datasheet that >> timestamping is only recommended for PTP packets on the 82576 HW. >> Other Intel Ethernet HW can timestamp all packets. > > But I tried reading E1000_RXSTMP[LH] registers as soon as > E1000_RXD_STAT_TS is set with no luck; ie. the timestamp was activated > only for ptp packets (even tried reading those registers regardless of > .._STAT_TS without success). So as you said, that probably means > HWTSTAMP_FILTER_ALL is not working on my HW. Is there a way to find > out if HWTSTAMP_FILTER_ALL is supported by the HW so that the igb > driver return -ERANGE from the ioctl()? I think I see the problem. If TSYNCRXCTL.TYPE=100b, aka HWTSTAMP_FILTER_ALL, all the packets are time stamped; however, this bit is never set as the time stamp value is not locked. Thanks, Alex > Ali > > ------------------------------------------------------------------------------ > Join us December 9, 2009 for the Red Hat Virtual Experience, > a free event focused on virtualization and cloud computing. > Attend in-depth sessions from your desk. Your couch. Anywhere. > http://p.sf.net/sfu/redhat-sfdev2dev > _______________________________________________ > E1000-devel mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/e1000-devel ------------------------------------------------------------------------------ Join us December 9, 2009 for the Red Hat Virtual Experience, a free event focused on virtualization and cloud computing. Attend in-depth sessions from your desk. Your couch. Anywhere. http://p.sf.net/sfu/redhat-sfdev2dev _______________________________________________ E1000-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/e1000-devel
