mlxsw: spectrum_ptp: Initialize the clock to zero as part of initialization

As lately recommended in the mailing list[1], set the clock to zero time as
part of initialization.

The idea is that when the clock reads 'Jan 1, 1970', then it is clearly
wrong and user will not mistakenly think that the clock is set correctly.
If as part of initialization, the driver sets the clock, user might see
correct date and time (maybe with a small shift) and assume that there
is no need to sync the clock.

Fix the existing code of Spectrum-1 to set the 'timecounter' to zero.

[1]:
https://lore.kernel.org/netdev/20220201191041.GB7009@hoboy.vegasvil.org/

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c
index 3958667..eab3d63 100644 (file)
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c
@@ -267,8 +267,7 @@ mlxsw_sp1_ptp_clock_init(struct mlxsw_sp *mlxsw_sp, struct device *dev)
        clock->cycles.mask = CLOCKSOURCE_MASK(MLXSW_SP1_PTP_CLOCK_MASK);
        clock->core = mlxsw_sp->core;
 
-       timecounter_init(&clock->tc, &clock->cycles,
-                        ktime_to_ns(ktime_get_real()));
+       timecounter_init(&clock->tc, &clock->cycles, 0);
 
        /* Calculate period in seconds to call the overflow watchdog - to make
         * sure counter is checked at least twice every wrap around.