On Sat, Mar 23 2024 at 01:38, Thomas Gleixner wrote:
> PTP_SYS_OFFSET_EXTENDED moves the outer sample points as close as
> possible to the actual PCH read and provides both outer samples to user
> space for analysis. It was introduced for a reason, no?
That said, it's a sad state of affairs that 16 drivers which did exist
before the introduction of the gettimex64() callback have not been
converted over to it within 4.5 years.
What's even worse is that 14 drivers have been merged _after_ the
gettimex64() callback got introduced without implementing it:
2019-02-12 drivers/net/ethernet/freescale/enetc/enetc_ptp.c
2019-10-24 drivers/net/ethernet/aquantia/atlantic/aq_ptp.c
2019-11-15 drivers/net/dsa/ocelot/felix_vsc9959.c
2020-01-12 drivers/net/ethernet/xscale/ptp_ixp46x.c
2020-06-20 drivers/net/ethernet/mscc/ocelot_vsc7514.c
2020-06-24 drivers/net/phy/mscc/mscc_ptp.c
2020-08-24 drivers/net/ethernet/marvell/octeontx2/nic/otx2_ptp.c
2020-11-05 drivers/net/dsa/hirschmann/hellcreek_ptp.c
2022-02-01 drivers/net/ethernet/microchip/lan966x/lan966x_ptp.c
2022-03-04 drivers/net/ethernet/microchip/sparx5/sparx5_ptp.c
2022-05-10 drivers/net/ethernet/sfc/siena/ptp.c
2022-11-02 drivers/net/ethernet/renesas/rcar_gen4_ptp.c
2023-01-13 drivers/net/dsa/microchip/ksz_ptp.c
2023-03-22 drivers/net/wireless/intel/iwlwifi/mvm/ptp.c
Not Sagi's fault at all, but it's telling and coherent with the approach
to solve the problem at hand.
See the previous reply for the observation on the letters P and C in PTP.
Oh well,
tglx
