| Message ID | 20221114213701.815132-1-jacob.e.keller@intel.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | d82303df06481235fe7cbaf605075e0c2c87e99b |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next] mlxsw: update adjfine to use adjust_by_scaled_ppm | expand |
On Mon, Nov 14, 2022 at 01:37:01PM -0800, Jacob Keller wrote: > The mlxsw adjfine implementation in the spectrum_ptp.c file converts > scaled_ppm into ppb before updating a cyclecounter multiplier using the > standard "base * ppb / 1billion" calculation. > > This can be re-written to use adjust_by_scaled_ppm, directly using the > scaled parts per million and reducing the amount of code required to > express this calculation. > > We still calculate the parts per billion for passing into > mlxsw_sp_ptp_phc_adjfreq because this function requires the input to be in > parts per billion. > > Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> > Cc: Amit Cohen <amcohen@nvidia.com> > Cc: Ido Schimmel <idosch@nvidia.com> > Cc: Petr Machata <petrm@nvidia.com> Thanks for the patch, code looks good to me. Petr, please apply this patch to our tree for testing.
Ido Schimmel <idosch@nvidia.com> writes: > On Mon, Nov 14, 2022 at 01:37:01PM -0800, Jacob Keller wrote: >> The mlxsw adjfine implementation in the spectrum_ptp.c file converts >> scaled_ppm into ppb before updating a cyclecounter multiplier using the >> standard "base * ppb / 1billion" calculation. >> >> This can be re-written to use adjust_by_scaled_ppm, directly using the >> scaled parts per million and reducing the amount of code required to >> express this calculation. >> >> We still calculate the parts per billion for passing into >> mlxsw_sp_ptp_phc_adjfreq because this function requires the input to be in >> parts per billion. >> >> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> >> Cc: Amit Cohen <amcohen@nvidia.com> >> Cc: Ido Schimmel <idosch@nvidia.com> >> Cc: Petr Machata <petrm@nvidia.com> > > Thanks for the patch, code looks good to me. > > Petr, please apply this patch to our tree for testing. Applied. Jacob, we'll let you know tomorrow whether it exploded.
On Mon, Nov 14, 2022 at 01:37:01PM -0800, Jacob Keller wrote: > The mlxsw adjfine implementation in the spectrum_ptp.c file converts > scaled_ppm into ppb before updating a cyclecounter multiplier using the > standard "base * ppb / 1billion" calculation. > > This can be re-written to use adjust_by_scaled_ppm, directly using the > scaled parts per million and reducing the amount of code required to > express this calculation. > > We still calculate the parts per billion for passing into > mlxsw_sp_ptp_phc_adjfreq because this function requires the input to be in > parts per billion. > > Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Ido Schimmel <idosch@nvidia.com>
On Wed, 16 Nov 2022 10:01:18 +0200 Ido Schimmel wrote: > On Mon, Nov 14, 2022 at 01:37:01PM -0800, Jacob Keller wrote: > > The mlxsw adjfine implementation in the spectrum_ptp.c file converts > > scaled_ppm into ppb before updating a cyclecounter multiplier using the > > standard "base * ppb / 1billion" calculation. > > > > This can be re-written to use adjust_by_scaled_ppm, directly using the > > scaled parts per million and reducing the amount of code required to > > express this calculation. > > > > We still calculate the parts per billion for passing into > > mlxsw_sp_ptp_phc_adjfreq because this function requires the input to be in > > parts per billion. > > > > Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> > > Reviewed-by: Ido Schimmel <idosch@nvidia.com> > Tested-by: Ido Schimmel <idosch@nvidia.com> FTR this patch got marked as 'not applicable' in pw but I'm not sure why, so applying...
Hello: This patch was applied to netdev/net-next.git (master) by Jakub Kicinski <kuba@kernel.org>: On Mon, 14 Nov 2022 13:37:01 -0800 you wrote: > The mlxsw adjfine implementation in the spectrum_ptp.c file converts > scaled_ppm into ppb before updating a cyclecounter multiplier using the > standard "base * ppb / 1billion" calculation. > > This can be re-written to use adjust_by_scaled_ppm, directly using the > scaled parts per million and reducing the amount of code required to > express this calculation. > > [...] Here is the summary with links: - [net-next] mlxsw: update adjfine to use adjust_by_scaled_ppm https://git.kernel.org/netdev/net-next/c/d82303df0648 You are awesome, thank you!
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c index 7b01b9c20722..cbb6c75a6620 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c @@ -189,29 +189,17 @@ mlxsw_sp1_ptp_phc_settime(struct mlxsw_sp1_ptp_clock *clock, u64 nsec) static int mlxsw_sp1_ptp_adjfine(struct ptp_clock_info *ptp, long scaled_ppm) { struct mlxsw_sp1_ptp_clock *clock = mlxsw_sp1_ptp_clock(ptp); - int neg_adj = 0; - u32 diff; - u64 adj; s32 ppb; ppb = scaled_ppm_to_ppb(scaled_ppm); - if (ppb < 0) { - neg_adj = 1; - ppb = -ppb; - } - - adj = clock->nominal_c_mult; - adj *= ppb; - diff = div_u64(adj, NSEC_PER_SEC); - spin_lock_bh(&clock->lock); timecounter_read(&clock->tc); - clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff : - clock->nominal_c_mult + diff; + clock->cycles.mult = adjust_by_scaled_ppm(clock->nominal_c_mult, + scaled_ppm); spin_unlock_bh(&clock->lock); - return mlxsw_sp_ptp_phc_adjfreq(&clock->common, neg_adj ? -ppb : ppb); + return mlxsw_sp_ptp_phc_adjfreq(&clock->common, ppb); } static int mlxsw_sp1_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
The mlxsw adjfine implementation in the spectrum_ptp.c file converts scaled_ppm into ppb before updating a cyclecounter multiplier using the standard "base * ppb / 1billion" calculation. This can be re-written to use adjust_by_scaled_ppm, directly using the scaled parts per million and reducing the amount of code required to express this calculation. We still calculate the parts per billion for passing into mlxsw_sp_ptp_phc_adjfreq because this function requires the input to be in parts per billion. Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Cc: Amit Cohen <amcohen@nvidia.com> Cc: Ido Schimmel <idosch@nvidia.com> Cc: Petr Machata <petrm@nvidia.com> --- Noticed this while investigating conversion of max_adj to scaled PPM format. This was missed in the previous round of updates that modified drivers to use the adjust_by_scaled_ppm interface. .../net/ethernet/mellanox/mlxsw/spectrum_ptp.c | 18 +++--------------- 1 file changed, 3 insertions(+), 15 deletions(-) base-commit: f12ed9c04804eec4f1819097a0fd0b4800adac2f