[PATCH mlx5-next 0/6] mlx5 real time clock

[PATCH mlx5-next 0/6] mlx5 real time clock

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

Hi, 

This series adds the support for mlx5 clock real time translation mode.
In case of no objection the series will be applied to mlx5-next branch.

Device timestamp can be in real time mode (cycles to time translation is
offloaded into the Hardware). With real time mode, HW provides timestamp
which is already translated into nanoseconds.

For more information about real time mode please see last patch in the
series.

Thanks,
Saeed.

---

Aharon Landau (1):
  net/mlx5: Add new timestamp mode bits

Aya Levin (1):
  net/mlx5: Add cyc2time HW translation mode support

Eran Ben Elisha (4):
  net/mlx5: Add register layout to support real-time time-stamp
  net/mlx5: Refactor init clock function
  net/mlx5: Move all internal timer metadata into a dedicated struct
  net/mlx5: Move some PPS logic into helper functions

 drivers/net/ethernet/mellanox/mlx5/core/en.h  |   3 +
 .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |   7 +-
 .../net/ethernet/mellanox/mlx5/core/en/txrx.h |   7 +
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  18 +-
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  11 +-
 .../net/ethernet/mellanox/mlx5/core/en_tx.c   |   3 +-
 .../ethernet/mellanox/mlx5/core/lib/clock.c   | 431 +++++++++++++-----
 .../ethernet/mellanox/mlx5/core/lib/clock.h   |  36 +-
 include/linux/mlx5/device.h                   |   5 +-
 include/linux/mlx5/driver.h                   |  13 +-
 include/linux/mlx5/mlx5_ifc.h                 |  84 +++-
 11 files changed, 489 insertions(+), 129 deletions(-)

-- 
2.29.2

[PATCH mlx5-next 1/6] net/mlx5: Add new timestamp mode bits

[Saeed Mahameed]

From: Aharon Landau <aharonl@nvidia.com>

These fields declare which timestamp mode is supported by the device
per RQ/SQ/QP.

In addition add the ts_format field to the select the mode for
RQ/SQ/QP.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 include/linux/mlx5/mlx5_ifc.h | 54 +++++++++++++++++++++++++++++++----
 1 file changed, 49 insertions(+), 5 deletions(-)

diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index cf692fc17f41..436d6f421dfd 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -932,11 +932,18 @@ struct mlx5_ifc_per_protocol_networking_offload_caps_bits {
 	u8         reserved_at_200[0x600];
 };
 
+enum {
+	MLX5_QP_TIMESTAMP_FORMAT_CAP_FREE_RUNNING               = 0x0,
+	MLX5_QP_TIMESTAMP_FORMAT_CAP_REAL_TIME                  = 0x1,
+	MLX5_QP_TIMESTAMP_FORMAT_CAP_FREE_RUNNING_AND_REAL_TIME = 0x2,
+};
+
 struct mlx5_ifc_roce_cap_bits {
 	u8         roce_apm[0x1];
 	u8         reserved_at_1[0x3];
 	u8         sw_r_roce_src_udp_port[0x1];
-	u8         reserved_at_5[0x1b];
+	u8         reserved_at_5[0x19];
+	u8	   qp_ts_format[0x2];
 
 	u8         reserved_at_20[0x60];
 
@@ -1253,6 +1260,18 @@ enum {
 	MLX5_STEERING_FORMAT_CONNECTX_6DX = 1,
 };
 
+enum {
+	MLX5_SQ_TIMESTAMP_FORMAT_CAP_FREE_RUNNING               = 0x0,
+	MLX5_SQ_TIMESTAMP_FORMAT_CAP_REAL_TIME                  = 0x1,
+	MLX5_SQ_TIMESTAMP_FORMAT_CAP_FREE_RUNNING_AND_REAL_TIME = 0x2,
+};
+
+enum {
+	MLX5_RQ_TIMESTAMP_FORMAT_CAP_FREE_RUNNING               = 0x0,
+	MLX5_RQ_TIMESTAMP_FORMAT_CAP_REAL_TIME                  = 0x1,
+	MLX5_RQ_TIMESTAMP_FORMAT_CAP_FREE_RUNNING_AND_REAL_TIME = 0x2,
+};
+
 struct mlx5_ifc_cmd_hca_cap_bits {
 	u8         reserved_at_0[0x1f];
 	u8         vhca_resource_manager[0x1];
@@ -1564,7 +1583,8 @@ struct mlx5_ifc_cmd_hca_cap_bits {
 
 	u8         general_obj_types[0x40];
 
-	u8         reserved_at_440[0x4];
+	u8         sq_ts_format[0x2];
+	u8         rq_ts_format[0x2];
 	u8         steering_format_version[0x4];
 	u8         create_qp_start_hint[0x18];
 
@@ -2868,6 +2888,12 @@ enum {
 	MLX5_QPC_CS_RES_UP_TO_64B  = 0x2,
 };
 
+enum {
+	MLX5_QPC_TIMESTAMP_FORMAT_FREE_RUNNING = 0x0,
+	MLX5_QPC_TIMESTAMP_FORMAT_DEFAULT      = 0x1,
+	MLX5_QPC_TIMESTAMP_FORMAT_REAL_TIME    = 0x2,
+};
+
 struct mlx5_ifc_qpc_bits {
 	u8         state[0x4];
 	u8         lag_tx_port_affinity[0x4];
@@ -2896,7 +2922,9 @@ struct mlx5_ifc_qpc_bits {
 	u8         log_rq_stride[0x3];
 	u8         no_sq[0x1];
 	u8         log_sq_size[0x4];
-	u8         reserved_at_55[0x6];
+	u8         reserved_at_55[0x3];
+	u8	   ts_format[0x2];
+	u8         reserved_at_5a[0x1];
 	u8         rlky[0x1];
 	u8         ulp_stateless_offload_mode[0x4];
 
@@ -3312,6 +3340,12 @@ enum {
 	MLX5_SQC_STATE_ERR  = 0x3,
 };
 
+enum {
+	MLX5_SQC_TIMESTAMP_FORMAT_FREE_RUNNING = 0x0,
+	MLX5_SQC_TIMESTAMP_FORMAT_DEFAULT      = 0x1,
+	MLX5_SQC_TIMESTAMP_FORMAT_REAL_TIME    = 0x2,
+};
+
 struct mlx5_ifc_sqc_bits {
 	u8         rlky[0x1];
 	u8         cd_master[0x1];
@@ -3323,7 +3357,9 @@ struct mlx5_ifc_sqc_bits {
 	u8         reg_umr[0x1];
 	u8         allow_swp[0x1];
 	u8         hairpin[0x1];
-	u8         reserved_at_f[0x11];
+	u8         reserved_at_f[0xb];
+	u8	   ts_format[0x2];
+	u8	   reserved_at_1c[0x4];
 
 	u8         reserved_at_20[0x8];
 	u8         user_index[0x18];
@@ -3414,6 +3450,12 @@ enum {
 	MLX5_RQC_STATE_ERR  = 0x3,
 };
 
+enum {
+	MLX5_RQC_TIMESTAMP_FORMAT_FREE_RUNNING = 0x0,
+	MLX5_RQC_TIMESTAMP_FORMAT_DEFAULT      = 0x1,
+	MLX5_RQC_TIMESTAMP_FORMAT_REAL_TIME    = 0x2,
+};
+
 struct mlx5_ifc_rqc_bits {
 	u8         rlky[0x1];
 	u8	   delay_drop_en[0x1];
@@ -3424,7 +3466,9 @@ struct mlx5_ifc_rqc_bits {
 	u8         reserved_at_c[0x1];
 	u8         flush_in_error_en[0x1];
 	u8         hairpin[0x1];
-	u8         reserved_at_f[0x11];
+	u8         reserved_at_f[0xb];
+	u8	   ts_format[0x2];
+	u8	   reserved_at_1c[0x4];
 
 	u8         reserved_at_20[0x8];
 	u8         user_index[0x18];
-- 
2.29.2

Re: [PATCH mlx5-next 1/6] net/mlx5: Add new timestamp mode bits

[Jason Gunthorpe]

On Fri, Feb 12, 2021 at 02:30:37PM -0800, Saeed Mahameed wrote:
> From: Aharon Landau <aharonl@nvidia.com>
> 
> These fields declare which timestamp mode is supported by the device
> per RQ/SQ/QP.
> 
> In addition add the ts_format field to the select the mode for
> RQ/SQ/QP.
> 
> Signed-off-by: Aharon Landau <aharonl@nvidia.com>
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
> ---
>  include/linux/mlx5/mlx5_ifc.h | 54 +++++++++++++++++++++++++++++++----
>  1 file changed, 49 insertions(+), 5 deletions(-)

This is a commit in the shared branch now, so this series will have to
go as a pull request if it wants to go before the next rc1

Jason

Re: [PATCH mlx5-next 1/6] net/mlx5: Add new timestamp mode bits

[Saeed Mahameed]

On Tue, 2021-02-16 at 16:03 -0400, Jason Gunthorpe wrote:
> On Fri, Feb 12, 2021 at 02:30:37PM -0800, Saeed Mahameed wrote:
> > From: Aharon Landau <aharonl@nvidia.com>
> > 
> > These fields declare which timestamp mode is supported by the
> > device
> > per RQ/SQ/QP.
> > 
> > In addition add the ts_format field to the select the mode for
> > RQ/SQ/QP.
> > 
> > Signed-off-by: Aharon Landau <aharonl@nvidia.com>
> > Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
> > ---
> >  include/linux/mlx5/mlx5_ifc.h | 54
> > +++++++++++++++++++++++++++++++----
> >  1 file changed, 49 insertions(+), 5 deletions(-)
> 
> This is a commit in the shared branch now, so this series will have
> to
> go as a pull request if it wants to go before the next rc1
> 
> Jason

Series applied to mlx5-next, i will send the PR shortly.

[PATCH mlx5-next 2/6] net/mlx5: Add register layout to support real-time time-stamp

[Saeed Mahameed]

From: Eran Ben Elisha <eranbe@mellanox.com>

Add needed structure layouts and defines for MTUTC (Management UTC)
register. MTUTC will be used for cyc2time HW translation.

In addition, add cyc2time modify capability bit and init segment HCA
real time address.

Finally, add capability bits indicating which time-stamping format is
supported per SQ and RQ. Add ts_format in the queue's context layout to
allow configuration.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 include/linux/mlx5/device.h   |  5 ++++-
 include/linux/mlx5/driver.h   |  1 +
 include/linux/mlx5/mlx5_ifc.h | 30 ++++++++++++++++++++++++++++--
 3 files changed, 33 insertions(+), 3 deletions(-)

diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h
index f1de49d64a98..fd694add6ff0 100644
--- a/include/linux/mlx5/device.h
+++ b/include/linux/mlx5/device.h
@@ -578,7 +578,10 @@ struct mlx5_init_seg {
 	__be32			internal_timer_l;
 	__be32			rsvd3[2];
 	__be32			health_counter;
-	__be32			rsvd4[1019];
+	__be32			rsvd4[11];
+	__be32			real_time_h;
+	__be32			real_time_l;
+	__be32			rsvd5[1006];
 	__be64			ieee1588_clk;
 	__be32			ieee1588_clk_type;
 	__be32			clr_intx;
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index f93bfe7473aa..2eb7755143d7 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -143,6 +143,7 @@ enum {
 	MLX5_REG_MPCNT		 = 0x9051,
 	MLX5_REG_MTPPS		 = 0x9053,
 	MLX5_REG_MTPPSE		 = 0x9054,
+	MLX5_REG_MTUTC		 = 0x9055,
 	MLX5_REG_MPEGC		 = 0x9056,
 	MLX5_REG_MCQS		 = 0x9060,
 	MLX5_REG_MCQI		 = 0x9061,
diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index 436d6f421dfd..7059fcdf54a3 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -9150,6 +9150,28 @@ struct mlx5_ifc_mpegc_reg_bits {
 	u8         reserved_at_60[0x100];
 };
 
+enum {
+	MLX5_MTUTC_OPERATION_SET_TIME_IMMEDIATE   = 0x1,
+	MLX5_MTUTC_OPERATION_ADJUST_TIME          = 0x2,
+	MLX5_MTUTC_OPERATION_ADJUST_FREQ_UTC      = 0x3,
+};
+
+struct mlx5_ifc_mtutc_reg_bits {
+	u8         reserved_at_0[0x1c];
+	u8         operation[0x4];
+
+	u8         freq_adjustment[0x20];
+
+	u8         reserved_at_40[0x40];
+
+	u8         utc_sec[0x20];
+
+	u8         reserved_at_a0[0x2];
+	u8         utc_nsec[0x1e];
+
+	u8         time_adjustment[0x20];
+};
+
 struct mlx5_ifc_pcam_enhanced_features_bits {
 	u8         reserved_at_0[0x68];
 	u8         fec_50G_per_lane_in_pplm[0x1];
@@ -9208,7 +9230,9 @@ struct mlx5_ifc_pcam_reg_bits {
 };
 
 struct mlx5_ifc_mcam_enhanced_features_bits {
-	u8         reserved_at_0[0x6e];
+	u8         reserved_at_0[0x6b];
+	u8         ptpcyc2realtime_modify[0x1];
+	u8         reserved_at_6c[0x2];
 	u8         pci_status_and_power[0x1];
 	u8         reserved_at_6f[0x5];
 	u8         mark_tx_action_cnp[0x1];
@@ -9231,7 +9255,8 @@ struct mlx5_ifc_mcam_access_reg_bits {
 
 	u8         regs_95_to_87[0x9];
 	u8         mpegc[0x1];
-	u8         regs_85_to_68[0x12];
+	u8         mtutc[0x1];
+	u8         regs_84_to_68[0x11];
 	u8         tracer_registers[0x4];
 
 	u8         regs_63_to_32[0x20];
@@ -9964,6 +9989,7 @@ union mlx5_ifc_ports_control_registers_document_bits {
 	struct mlx5_ifc_mcda_reg_bits mcda_reg;
 	struct mlx5_ifc_mirc_reg_bits mirc_reg;
 	struct mlx5_ifc_mfrl_reg_bits mfrl_reg;
+	struct mlx5_ifc_mtutc_reg_bits mtutc_reg;
 	u8         reserved_at_0[0x60e0];
 };
 
-- 
2.29.2

[PATCH mlx5-next 3/6] net/mlx5: Refactor init clock function

[Saeed Mahameed]

From: Eran Ben Elisha <eranbe@mellanox.com>

Function mlx5_init_clock() is responsible for internal PTP related metadata
initializations. Break mlx5_init_clock() to sub functions, each takes care
of its own logic.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../ethernet/mellanox/mlx5/core/lib/clock.c   | 76 +++++++++++++------
 1 file changed, 53 insertions(+), 23 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index c70c1f0ca0c1..aaf7d837a967 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -591,20 +591,12 @@ static int mlx5_pps_event(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
-void mlx5_init_clock(struct mlx5_core_dev *mdev)
+static void mlx5_timecounter_init(struct mlx5_core_dev *mdev)
 {
 	struct mlx5_clock *clock = &mdev->clock;
-	u64 overflow_cycles;
-	u64 ns;
-	u64 frac = 0;
 	u32 dev_freq;
 
 	dev_freq = MLX5_CAP_GEN(mdev, device_frequency_khz);
-	if (!dev_freq) {
-		mlx5_core_warn(mdev, "invalid device_frequency_khz, aborting HW clock init\n");
-		return;
-	}
-	seqlock_init(&clock->lock);
 	clock->cycles.read = read_internal_timer;
 	clock->cycles.shift = MLX5_CYCLES_SHIFT;
 	clock->cycles.mult = clocksource_khz2mult(dev_freq,
@@ -614,6 +606,15 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
 
 	timecounter_init(&clock->tc, &clock->cycles,
 			 ktime_to_ns(ktime_get_real()));
+}
+
+static void mlx5_init_overflow_period(struct mlx5_clock *clock)
+{
+	struct mlx5_core_dev *mdev = container_of(clock, struct mlx5_core_dev, clock);
+	struct mlx5_ib_clock_info *clock_info = mdev->clock_info;
+	u64 overflow_cycles;
+	u64 frac = 0;
+	u64 ns;
 
 	/* Calculate period in seconds to call the overflow watchdog - to make
 	 * sure counter is checked at least twice every wrap around.
@@ -630,24 +631,53 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
 	do_div(ns, NSEC_PER_SEC / HZ);
 	clock->overflow_period = ns;
 
-	mdev->clock_info =
-		(struct mlx5_ib_clock_info *)get_zeroed_page(GFP_KERNEL);
-	if (mdev->clock_info) {
-		mdev->clock_info->nsec = clock->tc.nsec;
-		mdev->clock_info->cycles = clock->tc.cycle_last;
-		mdev->clock_info->mask = clock->cycles.mask;
-		mdev->clock_info->mult = clock->nominal_c_mult;
-		mdev->clock_info->shift = clock->cycles.shift;
-		mdev->clock_info->frac = clock->tc.frac;
-		mdev->clock_info->overflow_period = clock->overflow_period;
-	}
-
-	INIT_WORK(&clock->pps_info.out_work, mlx5_pps_out);
 	INIT_DELAYED_WORK(&clock->overflow_work, mlx5_timestamp_overflow);
 	if (clock->overflow_period)
 		schedule_delayed_work(&clock->overflow_work, 0);
 	else
-		mlx5_core_warn(mdev, "invalid overflow period, overflow_work is not scheduled\n");
+		mlx5_core_warn(mdev,
+			       "invalid overflow period, overflow_work is not scheduled\n");
+
+	if (clock_info)
+		clock_info->overflow_period = clock->overflow_period;
+}
+
+static void mlx5_init_clock_info(struct mlx5_core_dev *mdev)
+{
+	struct mlx5_clock *clock = &mdev->clock;
+	struct mlx5_ib_clock_info *info;
+
+	mdev->clock_info = (struct mlx5_ib_clock_info *)get_zeroed_page(GFP_KERNEL);
+	if (!mdev->clock_info) {
+		mlx5_core_warn(mdev, "Failed to allocate IB clock info page\n");
+		return;
+	}
+
+	info = mdev->clock_info;
+
+	info->nsec = clock->tc.nsec;
+	info->cycles = clock->tc.cycle_last;
+	info->mask = clock->cycles.mask;
+	info->mult = clock->nominal_c_mult;
+	info->shift = clock->cycles.shift;
+	info->frac = clock->tc.frac;
+}
+
+void mlx5_init_clock(struct mlx5_core_dev *mdev)
+{
+	struct mlx5_clock *clock = &mdev->clock;
+
+	if (!MLX5_CAP_GEN(mdev, device_frequency_khz)) {
+		mlx5_core_warn(mdev, "invalid device_frequency_khz, aborting HW clock init\n");
+		return;
+	}
+
+	seqlock_init(&clock->lock);
+
+	mlx5_timecounter_init(mdev);
+	mlx5_init_clock_info(mdev);
+	mlx5_init_overflow_period(clock);
+	INIT_WORK(&clock->pps_info.out_work, mlx5_pps_out);
 
 	/* Configure the PHC */
 	clock->ptp_info = mlx5_ptp_clock_info;
-- 
2.29.2

[PATCH mlx5-next 4/6] net/mlx5: Move all internal timer metadata into a dedicated struct

[Saeed Mahameed]

From: Eran Ben Elisha <eranbe@mellanox.com>

Internal timer mode (SW clock) requires some PTP clock related metadata
structs. Real time mode (HW clock) will not need these metadata structs.
This separation emphasize the different interfaces for HW clock and SW
clock.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../ethernet/mellanox/mlx5/core/lib/clock.c   | 107 ++++++++++--------
 .../ethernet/mellanox/mlx5/core/lib/clock.h   |   3 +-
 include/linux/mlx5/driver.h                   |  12 +-
 3 files changed, 71 insertions(+), 51 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index aaf7d837a967..dcacaa451b53 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -89,7 +89,8 @@ static u64 mlx5_read_internal_timer(struct mlx5_core_dev *dev,
 
 static u64 read_internal_timer(const struct cyclecounter *cc)
 {
-	struct mlx5_clock *clock = container_of(cc, struct mlx5_clock, cycles);
+	struct mlx5_timer *timer = container_of(cc, struct mlx5_timer, cycles);
+	struct mlx5_clock *clock = container_of(timer, struct mlx5_clock, timer);
 	struct mlx5_core_dev *mdev = container_of(clock, struct mlx5_core_dev,
 						  clock);
 
@@ -100,6 +101,7 @@ static void mlx5_update_clock_info_page(struct mlx5_core_dev *mdev)
 {
 	struct mlx5_ib_clock_info *clock_info = mdev->clock_info;
 	struct mlx5_clock *clock = &mdev->clock;
+	struct mlx5_timer *timer;
 	u32 sign;
 
 	if (!clock_info)
@@ -109,10 +111,11 @@ static void mlx5_update_clock_info_page(struct mlx5_core_dev *mdev)
 	smp_store_mb(clock_info->sign,
 		     sign | MLX5_IB_CLOCK_INFO_KERNEL_UPDATING);
 
-	clock_info->cycles = clock->tc.cycle_last;
-	clock_info->mult   = clock->cycles.mult;
-	clock_info->nsec   = clock->tc.nsec;
-	clock_info->frac   = clock->tc.frac;
+	timer = &clock->timer;
+	clock_info->cycles = timer->tc.cycle_last;
+	clock_info->mult   = timer->cycles.mult;
+	clock_info->nsec   = timer->tc.nsec;
+	clock_info->frac   = timer->tc.frac;
 
 	smp_store_release(&clock_info->sign,
 			  sign + MLX5_IB_CLOCK_INFO_KERNEL_UPDATING * 2);
@@ -151,28 +154,32 @@ static void mlx5_timestamp_overflow(struct work_struct *work)
 {
 	struct delayed_work *dwork = to_delayed_work(work);
 	struct mlx5_core_dev *mdev;
+	struct mlx5_timer *timer;
 	struct mlx5_clock *clock;
 	unsigned long flags;
 
-	clock = container_of(dwork, struct mlx5_clock, overflow_work);
+	timer = container_of(dwork, struct mlx5_timer, overflow_work);
+	clock = container_of(timer, struct mlx5_clock, timer);
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
+
 	write_seqlock_irqsave(&clock->lock, flags);
-	timecounter_read(&clock->tc);
+	timecounter_read(&timer->tc);
 	mlx5_update_clock_info_page(mdev);
 	write_sequnlock_irqrestore(&clock->lock, flags);
-	schedule_delayed_work(&clock->overflow_work, clock->overflow_period);
+	schedule_delayed_work(&timer->overflow_work, timer->overflow_period);
 }
 
 static int mlx5_ptp_settime(struct ptp_clock_info *ptp, const struct timespec64 *ts)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
+	struct mlx5_timer *timer = &clock->timer;
 	u64 ns = timespec64_to_ns(ts);
 	struct mlx5_core_dev *mdev;
 	unsigned long flags;
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
 	write_seqlock_irqsave(&clock->lock, flags);
-	timecounter_init(&clock->tc, &clock->cycles, ns);
+	timecounter_init(&timer->tc, &timer->cycles, ns);
 	mlx5_update_clock_info_page(mdev);
 	write_sequnlock_irqrestore(&clock->lock, flags);
 
@@ -183,6 +190,7 @@ static int mlx5_ptp_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,
 			     struct ptp_system_timestamp *sts)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
+	struct mlx5_timer *timer = &clock->timer;
 	struct mlx5_core_dev *mdev;
 	unsigned long flags;
 	u64 cycles, ns;
@@ -190,7 +198,7 @@ static int mlx5_ptp_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
 	write_seqlock_irqsave(&clock->lock, flags);
 	cycles = mlx5_read_internal_timer(mdev, sts);
-	ns = timecounter_cyc2time(&clock->tc, cycles);
+	ns = timecounter_cyc2time(&timer->tc, cycles);
 	write_sequnlock_irqrestore(&clock->lock, flags);
 
 	*ts = ns_to_timespec64(ns);
@@ -201,12 +209,13 @@ static int mlx5_ptp_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,
 static int mlx5_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
+	struct mlx5_timer *timer = &clock->timer;
 	struct mlx5_core_dev *mdev;
 	unsigned long flags;
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
 	write_seqlock_irqsave(&clock->lock, flags);
-	timecounter_adjtime(&clock->tc, delta);
+	timecounter_adjtime(&timer->tc, delta);
 	mlx5_update_clock_info_page(mdev);
 	write_sequnlock_irqrestore(&clock->lock, flags);
 
@@ -216,27 +225,27 @@ static int mlx5_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 static int mlx5_ptp_adjfreq(struct ptp_clock_info *ptp, s32 delta)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
+	struct mlx5_timer *timer = &clock->timer;
 	struct mlx5_core_dev *mdev;
 	unsigned long flags;
 	int neg_adj = 0;
 	u32 diff;
 	u64 adj;
 
-
 	if (delta < 0) {
 		neg_adj = 1;
 		delta = -delta;
 	}
 
-	adj = clock->nominal_c_mult;
+	adj = timer->nominal_c_mult;
 	adj *= delta;
 	diff = div_u64(adj, 1000000000ULL);
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
 	write_seqlock_irqsave(&clock->lock, flags);
-	timecounter_read(&clock->tc);
-	clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff :
-				       clock->nominal_c_mult + diff;
+	timecounter_read(&timer->tc);
+	timer->cycles.mult = neg_adj ? timer->nominal_c_mult - diff :
+				       timer->nominal_c_mult + diff;
 	mlx5_update_clock_info_page(mdev);
 	write_sequnlock_irqrestore(&clock->lock, flags);
 
@@ -313,6 +322,7 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 			container_of(ptp, struct mlx5_clock, ptp_info);
 	struct mlx5_core_dev *mdev =
 			container_of(clock, struct mlx5_core_dev, clock);
+	struct mlx5_timer *timer = &clock->timer;
 	u32 in[MLX5_ST_SZ_DW(mtpps_reg)] = {0};
 	u64 nsec_now, nsec_delta, time_stamp = 0;
 	u64 cycles_now, cycles_delta;
@@ -355,10 +365,10 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 		ns = timespec64_to_ns(&ts);
 		cycles_now = mlx5_read_internal_timer(mdev, NULL);
 		write_seqlock_irqsave(&clock->lock, flags);
-		nsec_now = timecounter_cyc2time(&clock->tc, cycles_now);
+		nsec_now = timecounter_cyc2time(&timer->tc, cycles_now);
 		nsec_delta = ns - nsec_now;
-		cycles_delta = div64_u64(nsec_delta << clock->cycles.shift,
-					 clock->cycles.mult);
+		cycles_delta = div64_u64(nsec_delta << timer->cycles.shift,
+					 timer->cycles.mult);
 		write_sequnlock_irqrestore(&clock->lock, flags);
 		time_stamp = cycles_now + cycles_delta;
 		field_select = MLX5_MTPPS_FS_PIN_MODE |
@@ -541,6 +551,7 @@ static int mlx5_pps_event(struct notifier_block *nb,
 			  unsigned long type, void *data)
 {
 	struct mlx5_clock *clock = mlx5_nb_cof(nb, struct mlx5_clock, pps_nb);
+	struct mlx5_timer *timer = &clock->timer;
 	struct ptp_clock_event ptp_event;
 	u64 cycles_now, cycles_delta;
 	u64 nsec_now, nsec_delta, ns;
@@ -575,10 +586,10 @@ static int mlx5_pps_event(struct notifier_block *nb,
 		ts.tv_nsec = 0;
 		ns = timespec64_to_ns(&ts);
 		write_seqlock_irqsave(&clock->lock, flags);
-		nsec_now = timecounter_cyc2time(&clock->tc, cycles_now);
+		nsec_now = timecounter_cyc2time(&timer->tc, cycles_now);
 		nsec_delta = ns - nsec_now;
-		cycles_delta = div64_u64(nsec_delta << clock->cycles.shift,
-					 clock->cycles.mult);
+		cycles_delta = div64_u64(nsec_delta << timer->cycles.shift,
+					 timer->cycles.mult);
 		clock->pps_info.start[pin] = cycles_now + cycles_delta;
 		write_sequnlock_irqrestore(&clock->lock, flags);
 		schedule_work(&clock->pps_info.out_work);
@@ -594,17 +605,18 @@ static int mlx5_pps_event(struct notifier_block *nb,
 static void mlx5_timecounter_init(struct mlx5_core_dev *mdev)
 {
 	struct mlx5_clock *clock = &mdev->clock;
+	struct mlx5_timer *timer = &clock->timer;
 	u32 dev_freq;
 
 	dev_freq = MLX5_CAP_GEN(mdev, device_frequency_khz);
-	clock->cycles.read = read_internal_timer;
-	clock->cycles.shift = MLX5_CYCLES_SHIFT;
-	clock->cycles.mult = clocksource_khz2mult(dev_freq,
-						  clock->cycles.shift);
-	clock->nominal_c_mult = clock->cycles.mult;
-	clock->cycles.mask = CLOCKSOURCE_MASK(41);
-
-	timecounter_init(&clock->tc, &clock->cycles,
+	timer->cycles.read = read_internal_timer;
+	timer->cycles.shift = MLX5_CYCLES_SHIFT;
+	timer->cycles.mult = clocksource_khz2mult(dev_freq,
+						  timer->cycles.shift);
+	timer->nominal_c_mult = timer->cycles.mult;
+	timer->cycles.mask = CLOCKSOURCE_MASK(41);
+
+	timecounter_init(&timer->tc, &timer->cycles,
 			 ktime_to_ns(ktime_get_real()));
 }
 
@@ -612,6 +624,7 @@ static void mlx5_init_overflow_period(struct mlx5_clock *clock)
 {
 	struct mlx5_core_dev *mdev = container_of(clock, struct mlx5_core_dev, clock);
 	struct mlx5_ib_clock_info *clock_info = mdev->clock_info;
+	struct mlx5_timer *timer = &clock->timer;
 	u64 overflow_cycles;
 	u64 frac = 0;
 	u64 ns;
@@ -623,29 +636,30 @@ static void mlx5_init_overflow_period(struct mlx5_clock *clock)
 	 * multiplied by clock multiplier where the result doesn't exceed
 	 * 64bits.
 	 */
-	overflow_cycles = div64_u64(~0ULL >> 1, clock->cycles.mult);
-	overflow_cycles = min(overflow_cycles, div_u64(clock->cycles.mask, 3));
+	overflow_cycles = div64_u64(~0ULL >> 1, timer->cycles.mult);
+	overflow_cycles = min(overflow_cycles, div_u64(timer->cycles.mask, 3));
 
-	ns = cyclecounter_cyc2ns(&clock->cycles, overflow_cycles,
+	ns = cyclecounter_cyc2ns(&timer->cycles, overflow_cycles,
 				 frac, &frac);
 	do_div(ns, NSEC_PER_SEC / HZ);
-	clock->overflow_period = ns;
+	timer->overflow_period = ns;
 
-	INIT_DELAYED_WORK(&clock->overflow_work, mlx5_timestamp_overflow);
-	if (clock->overflow_period)
-		schedule_delayed_work(&clock->overflow_work, 0);
+	INIT_DELAYED_WORK(&timer->overflow_work, mlx5_timestamp_overflow);
+	if (timer->overflow_period)
+		schedule_delayed_work(&timer->overflow_work, 0);
 	else
 		mlx5_core_warn(mdev,
 			       "invalid overflow period, overflow_work is not scheduled\n");
 
 	if (clock_info)
-		clock_info->overflow_period = clock->overflow_period;
+		clock_info->overflow_period = timer->overflow_period;
 }
 
 static void mlx5_init_clock_info(struct mlx5_core_dev *mdev)
 {
 	struct mlx5_clock *clock = &mdev->clock;
 	struct mlx5_ib_clock_info *info;
+	struct mlx5_timer *timer;
 
 	mdev->clock_info = (struct mlx5_ib_clock_info *)get_zeroed_page(GFP_KERNEL);
 	if (!mdev->clock_info) {
@@ -654,13 +668,14 @@ static void mlx5_init_clock_info(struct mlx5_core_dev *mdev)
 	}
 
 	info = mdev->clock_info;
-
-	info->nsec = clock->tc.nsec;
-	info->cycles = clock->tc.cycle_last;
-	info->mask = clock->cycles.mask;
-	info->mult = clock->nominal_c_mult;
-	info->shift = clock->cycles.shift;
-	info->frac = clock->tc.frac;
+	timer = &clock->timer;
+
+	info->nsec = timer->tc.nsec;
+	info->cycles = timer->tc.cycle_last;
+	info->mask = timer->cycles.mask;
+	info->mult = timer->nominal_c_mult;
+	info->shift = timer->cycles.shift;
+	info->frac = timer->tc.frac;
 }
 
 void mlx5_init_clock(struct mlx5_core_dev *mdev)
@@ -714,7 +729,7 @@ void mlx5_cleanup_clock(struct mlx5_core_dev *mdev)
 	}
 
 	cancel_work_sync(&clock->pps_info.out_work);
-	cancel_delayed_work_sync(&clock->overflow_work);
+	cancel_delayed_work_sync(&clock->timer.overflow_work);
 
 	if (mdev->clock_info) {
 		free_page((unsigned long)mdev->clock_info);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
index 31600924bdc3..6e8804ebc773 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
@@ -45,12 +45,13 @@ static inline int mlx5_clock_get_ptp_index(struct mlx5_core_dev *mdev)
 static inline ktime_t mlx5_timecounter_cyc2time(struct mlx5_clock *clock,
 						u64 timestamp)
 {
+	struct mlx5_timer *timer = &clock->timer;
 	unsigned int seq;
 	u64 nsec;
 
 	do {
 		seq = read_seqbegin(&clock->lock);
-		nsec = timecounter_cyc2time(&clock->tc, timestamp);
+		nsec = timecounter_cyc2time(&timer->tc, timestamp);
 	} while (read_seqretry(&clock->lock, seq));
 
 	return ns_to_ktime(nsec);
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 2eb7755143d7..2f4d6182df1b 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -662,18 +662,22 @@ struct mlx5_pps {
 	u8                         enabled;
 };
 
-struct mlx5_clock {
-	struct mlx5_nb             pps_nb;
-	seqlock_t                  lock;
+struct mlx5_timer {
 	struct cyclecounter        cycles;
 	struct timecounter         tc;
-	struct hwtstamp_config     hwtstamp_config;
 	u32                        nominal_c_mult;
 	unsigned long              overflow_period;
 	struct delayed_work        overflow_work;
+};
+
+struct mlx5_clock {
+	struct mlx5_nb             pps_nb;
+	seqlock_t                  lock;
+	struct hwtstamp_config     hwtstamp_config;
 	struct ptp_clock          *ptp;
 	struct ptp_clock_info      ptp_info;
 	struct mlx5_pps            pps_info;
+	struct mlx5_timer          timer;
 };
 
 struct mlx5_dm;
-- 
2.29.2

[PATCH mlx5-next 5/6] net/mlx5: Move some PPS logic into helper functions

[Saeed Mahameed]

From: Eran Ben Elisha <eranbe@mellanox.com>

Some of PPS logic (timestamp calculations) fits only internal timer
timestamp mode. Move these logics into helper functions. Later in the
patchset cyc2time HW translation mode will expose its own PPS timestamp
calculations.

With this change, main flow will only hold calling PPS logic based on run
time mode.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../ethernet/mellanox/mlx5/core/lib/clock.c   | 113 +++++++++++-------
 1 file changed, 73 insertions(+), 40 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index dcacaa451b53..dbb6138db096 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -314,6 +314,40 @@ static int mlx5_extts_configure(struct ptp_clock_info *ptp,
 			       MLX5_EVENT_MODE_REPETETIVE & on);
 }
 
+static u64 find_target_cycles(struct mlx5_core_dev *mdev, s64 target_ns)
+{
+	struct mlx5_clock *clock = &mdev->clock;
+	u64 cycles_now, cycles_delta;
+	u64 nsec_now, nsec_delta;
+	struct mlx5_timer *timer;
+	unsigned long flags;
+
+	timer = &clock->timer;
+
+	cycles_now = mlx5_read_internal_timer(mdev, NULL);
+	write_seqlock_irqsave(&clock->lock, flags);
+	nsec_now = timecounter_cyc2time(&timer->tc, cycles_now);
+	nsec_delta = target_ns - nsec_now;
+	cycles_delta = div64_u64(nsec_delta << timer->cycles.shift,
+				 timer->cycles.mult);
+	write_sequnlock_irqrestore(&clock->lock, flags);
+
+	return cycles_now + cycles_delta;
+}
+
+static u64 perout_conf_internal_timer(struct mlx5_core_dev *mdev,
+				      s64 sec, u32 nsec)
+{
+	struct timespec64 ts;
+	s64 target_ns;
+
+	ts.tv_sec = sec;
+	ts.tv_nsec = nsec;
+	target_ns = timespec64_to_ns(&ts);
+
+	return find_target_cycles(mdev, target_ns);
+}
+
 static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 				 struct ptp_clock_request *rq,
 				 int on)
@@ -322,13 +356,10 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 			container_of(ptp, struct mlx5_clock, ptp_info);
 	struct mlx5_core_dev *mdev =
 			container_of(clock, struct mlx5_core_dev, clock);
-	struct mlx5_timer *timer = &clock->timer;
 	u32 in[MLX5_ST_SZ_DW(mtpps_reg)] = {0};
-	u64 nsec_now, nsec_delta, time_stamp = 0;
-	u64 cycles_now, cycles_delta;
 	struct timespec64 ts;
-	unsigned long flags;
 	u32 field_select = 0;
+	u64 time_stamp = 0;
 	u8 pin_mode = 0;
 	u8 pattern = 0;
 	int pin = -1;
@@ -345,12 +376,15 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 	if (rq->perout.index >= clock->ptp_info.n_pins)
 		return -EINVAL;
 
-	pin = ptp_find_pin(clock->ptp, PTP_PF_PEROUT,
-			   rq->perout.index);
-	if (pin < 0)
-		return -EBUSY;
-
+	field_select = MLX5_MTPPS_FS_ENABLE;
 	if (on) {
+		u32 nsec;
+		s64 sec;
+
+		pin = ptp_find_pin(clock->ptp, PTP_PF_PEROUT, rq->perout.index);
+		if (pin < 0)
+			return -EBUSY;
+
 		pin_mode = MLX5_PIN_MODE_OUT;
 		pattern = MLX5_OUT_PATTERN_PERIODIC;
 		ts.tv_sec = rq->perout.period.sec;
@@ -360,23 +394,14 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 		if ((ns >> 1) != 500000000LL)
 			return -EINVAL;
 
-		ts.tv_sec = rq->perout.start.sec;
-		ts.tv_nsec = rq->perout.start.nsec;
-		ns = timespec64_to_ns(&ts);
-		cycles_now = mlx5_read_internal_timer(mdev, NULL);
-		write_seqlock_irqsave(&clock->lock, flags);
-		nsec_now = timecounter_cyc2time(&timer->tc, cycles_now);
-		nsec_delta = ns - nsec_now;
-		cycles_delta = div64_u64(nsec_delta << timer->cycles.shift,
-					 timer->cycles.mult);
-		write_sequnlock_irqrestore(&clock->lock, flags);
-		time_stamp = cycles_now + cycles_delta;
-		field_select = MLX5_MTPPS_FS_PIN_MODE |
-			       MLX5_MTPPS_FS_PATTERN |
-			       MLX5_MTPPS_FS_ENABLE |
-			       MLX5_MTPPS_FS_TIME_STAMP;
-	} else {
-		field_select = MLX5_MTPPS_FS_ENABLE;
+		nsec = rq->perout.start.nsec;
+		sec = rq->perout.start.sec;
+
+		time_stamp = perout_conf_internal_timer(mdev, sec, nsec);
+
+		field_select |= MLX5_MTPPS_FS_PIN_MODE |
+				MLX5_MTPPS_FS_PATTERN |
+				MLX5_MTPPS_FS_TIME_STAMP;
 	}
 
 	MLX5_SET(mtpps_reg, in, pin, pin);
@@ -547,19 +572,35 @@ static void mlx5_get_pps_caps(struct mlx5_core_dev *mdev)
 	clock->pps_info.pin_caps[7] = MLX5_GET(mtpps_reg, out, cap_pin_7_mode);
 }
 
+static void ts_next_sec(struct timespec64 *ts)
+{
+	ts->tv_sec += 1;
+	ts->tv_nsec = 0;
+}
+
+static u64 perout_conf_next_event_inernal_timer(struct mlx5_core_dev *mdev,
+						struct mlx5_clock *clock)
+{
+	struct timespec64 ts;
+	s64 target_ns;
+
+	mlx5_ptp_gettimex(&clock->ptp_info, &ts, NULL);
+	ts_next_sec(&ts);
+	target_ns = timespec64_to_ns(&ts);
+
+	return find_target_cycles(mdev, target_ns);
+}
+
 static int mlx5_pps_event(struct notifier_block *nb,
 			  unsigned long type, void *data)
 {
 	struct mlx5_clock *clock = mlx5_nb_cof(nb, struct mlx5_clock, pps_nb);
-	struct mlx5_timer *timer = &clock->timer;
 	struct ptp_clock_event ptp_event;
-	u64 cycles_now, cycles_delta;
-	u64 nsec_now, nsec_delta, ns;
 	struct mlx5_eqe *eqe = data;
 	int pin = eqe->data.pps.pin;
 	struct mlx5_core_dev *mdev;
-	struct timespec64 ts;
 	unsigned long flags;
+	u64 ns;
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
 
@@ -580,17 +621,9 @@ static int mlx5_pps_event(struct notifier_block *nb,
 		ptp_clock_event(clock->ptp, &ptp_event);
 		break;
 	case PTP_PF_PEROUT:
-		mlx5_ptp_gettimex(&clock->ptp_info, &ts, NULL);
-		cycles_now = mlx5_read_internal_timer(mdev, NULL);
-		ts.tv_sec += 1;
-		ts.tv_nsec = 0;
-		ns = timespec64_to_ns(&ts);
+		ns = perout_conf_next_event_inernal_timer(mdev, clock);
 		write_seqlock_irqsave(&clock->lock, flags);
-		nsec_now = timecounter_cyc2time(&timer->tc, cycles_now);
-		nsec_delta = ns - nsec_now;
-		cycles_delta = div64_u64(nsec_delta << timer->cycles.shift,
-					 timer->cycles.mult);
-		clock->pps_info.start[pin] = cycles_now + cycles_delta;
+		clock->pps_info.start[pin] = ns;
 		write_sequnlock_irqrestore(&clock->lock, flags);
 		schedule_work(&clock->pps_info.out_work);
 		break;
-- 
2.29.2

[PATCH mlx5-next 6/6] net/mlx5: Add cyc2time HW translation mode support

[Saeed Mahameed]

From: Aya Levin <ayal@nvidia.com>

Device timestamp can be in real time mode (cycles to time translation is
offloaded into the Hardware). With real time mode, HW provides timestamp
which is already translated into nanoseconds.

With this mode, driver adjusts both the HW and timecounter (to keep
clock_info_page updated) using callbacks: adjfreq, adjtime and settime.
HW clock modifications are done via MTUTC access reg commands. Driver is
allowed to modify HW real time clock only if MCAM ptpcyc2realtime_modify
capability is set.

Add MTUTC set function to be used for configuring the HW real time
clock. Modify existing code to support both internal timer (with
conversion via timecounter_cyc2time() and real time (no conversions).

Align the signatures of the helpers converting from timestamp to
nanoseconds. With that, when allocating a queue assign the corresponding
callback with respect to the capability.

Adjust 1PPS timestamp calculation flows based on the timestamp mode.

Cyc2time offload brings two major advantages:
- Improve MTAE (Max Time Absolute Error) for HW TS by up to 160 ns over a
  100% loaded CPU.
- Faster data-path timestamp to nanoseconds, as translation is
  lock-less and done in HW.

On real time mode, timestamp format is 32 high bits of seconds and 32
low bits of nanoseconds. On some flows, driver shall convert this format
into nanoseconds wall-clock with REAL_TIME_TO_NS macro.

HW supports a single clock, and it is shared by all functions on a
device. In case real time clock is used, it is recommended to use
a single GM to all device's functions.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |   3 +
 .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |   7 +-
 .../net/ethernet/mellanox/mlx5/core/en/txrx.h |   7 +
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  18 +-
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  11 +-
 .../net/ethernet/mellanox/mlx5/core/en_tx.c   |   3 +-
 .../ethernet/mellanox/mlx5/core/lib/clock.c   | 197 +++++++++++++++---
 .../ethernet/mellanox/mlx5/core/lib/clock.h   |  33 +++
 8 files changed, 241 insertions(+), 38 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 055baf3b6cb1..096e44d2ab4b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -56,6 +56,7 @@
 #include "en/dcbnl.h"
 #include "en/fs.h"
 #include "lib/hv_vhca.h"
+#include "lib/clock.h"
 
 extern const struct net_device_ops mlx5e_netdev_ops;
 struct page_pool;
@@ -389,6 +390,7 @@ struct mlx5e_txqsq {
 	u32                        rate_limit;
 	struct work_struct         recover_work;
 	struct mlx5e_ptpsq        *ptpsq;
+	cqe_ts_to_ns               ptp_cyc2time;
 } ____cacheline_aligned_in_smp;
 
 struct mlx5e_dma_info {
@@ -650,6 +652,7 @@ struct mlx5e_rq {
 
 	/* XDP read-mostly */
 	struct xdp_rxq_info    xdp_rxq;
+	cqe_ts_to_ns           ptp_cyc2time;
 } ____cacheline_aligned_in_smp;
 
 enum mlx5e_channel_state {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c
index 2a2bac30daaa..b1a2b1418803 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c
@@ -3,7 +3,6 @@
 
 #include "en/ptp.h"
 #include "en/txrx.h"
-#include "lib/clock.h"
 
 struct mlx5e_skb_cb_hwtstamp {
 	ktime_t cqe_hwtstamp;
@@ -70,6 +69,7 @@ static void mlx5e_ptp_handle_ts_cqe(struct mlx5e_ptpsq *ptpsq,
 				    int budget)
 {
 	struct sk_buff *skb = mlx5e_skb_fifo_pop(&ptpsq->skb_fifo);
+	struct mlx5e_txqsq *sq = &ptpsq->txqsq;
 	ktime_t hwtstamp;
 
 	if (unlikely(MLX5E_RX_ERR_CQE(cqe))) {
@@ -77,7 +77,7 @@ static void mlx5e_ptp_handle_ts_cqe(struct mlx5e_ptpsq *ptpsq,
 		goto out;
 	}
 
-	hwtstamp = mlx5_timecounter_cyc2time(ptpsq->txqsq.clock, get_cqe_ts(cqe));
+	hwtstamp = mlx5e_cqe_ts_to_ns(sq->ptp_cyc2time, sq->clock, get_cqe_ts(cqe));
 	mlx5e_skb_cb_hwtstamp_handler(skb, MLX5E_SKB_CB_PORT_HWTSTAMP,
 				      hwtstamp, ptpsq->cq_stats);
 	ptpsq->cq_stats->cqe++;
@@ -183,6 +183,9 @@ static int mlx5e_ptp_alloc_txqsq(struct mlx5e_port_ptp *c, int txq_ix,
 	if (!MLX5_CAP_ETH(mdev, wqe_vlan_insert))
 		set_bit(MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE, &sq->state);
 	sq->stop_room = param->stop_room;
+	sq->ptp_cyc2time = mlx5_is_real_time_sq(mdev) ?
+			   mlx5_real_time_cyc2time :
+			   mlx5_timecounter_cyc2time;
 
 	node = dev_to_node(mlx5_core_dma_dev(mdev));
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
index 4880f2179273..2371b83dad9c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
@@ -26,6 +26,13 @@
 
 #define MLX5E_RX_ERR_CQE(cqe) (get_cqe_opcode(cqe) != MLX5_CQE_RESP_SEND)
 
+static inline
+ktime_t mlx5e_cqe_ts_to_ns(cqe_ts_to_ns func, struct mlx5_clock *clock, u64 cqe_ts)
+{
+	return INDIRECT_CALL_2(func, mlx5_real_time_cyc2time, mlx5_timecounter_cyc2time,
+			       clock, cqe_ts);
+}
+
 enum mlx5e_icosq_wqe_type {
 	MLX5E_ICOSQ_WQE_NOP,
 	MLX5E_ICOSQ_WQE_UMR_RX,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 6a852b4901aa..e40dec25cf6b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -422,6 +422,9 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	rq->hw_mtu  = MLX5E_SW2HW_MTU(params, params->sw_mtu);
 	rq->xdpsq   = &c->rq_xdpsq;
 	rq->xsk_pool = xsk_pool;
+	rq->ptp_cyc2time = mlx5_is_real_time_rq(mdev) ?
+			   mlx5_real_time_cyc2time :
+			   mlx5_timecounter_cyc2time;
 
 	if (rq->xsk_pool)
 		rq->stats = &c->priv->channel_stats[c->ix].xskrq;
@@ -654,7 +657,7 @@ static int mlx5e_create_rq(struct mlx5e_rq *rq,
 			   struct mlx5e_rq_param *param)
 {
 	struct mlx5_core_dev *mdev = rq->mdev;
-
+	u8 ts_format;
 	void *in;
 	void *rqc;
 	void *wq;
@@ -667,6 +670,9 @@ static int mlx5e_create_rq(struct mlx5e_rq *rq,
 	if (!in)
 		return -ENOMEM;
 
+	ts_format = mlx5_is_real_time_rq(mdev) ?
+		    MLX5_RQC_TIMESTAMP_FORMAT_REAL_TIME :
+		    MLX5_RQC_TIMESTAMP_FORMAT_FREE_RUNNING;
 	rqc = MLX5_ADDR_OF(create_rq_in, in, ctx);
 	wq  = MLX5_ADDR_OF(rqc, rqc, wq);
 
@@ -674,6 +680,7 @@ static int mlx5e_create_rq(struct mlx5e_rq *rq,
 
 	MLX5_SET(rqc,  rqc, cqn,		rq->cq.mcq.cqn);
 	MLX5_SET(rqc,  rqc, state,		MLX5_RQC_STATE_RST);
+	MLX5_SET(rqc,  rqc, ts_format,		ts_format);
 	MLX5_SET(wq,   wq,  log_wq_pg_sz,	rq->wq_ctrl.buf.page_shift -
 						MLX5_ADAPTER_PAGE_SHIFT);
 	MLX5_SET64(wq, wq,  dbr_addr,		rq->wq_ctrl.db.dma);
@@ -1154,6 +1161,9 @@ static int mlx5e_alloc_txqsq(struct mlx5e_channel *c,
 	if (param->is_mpw)
 		set_bit(MLX5E_SQ_STATE_MPWQE, &sq->state);
 	sq->stop_room = param->stop_room;
+	sq->ptp_cyc2time = mlx5_is_real_time_sq(mdev) ?
+			   mlx5_real_time_cyc2time :
+			   mlx5_timecounter_cyc2time;
 
 	param->wq.db_numa_node = cpu_to_node(c->cpu);
 	err = mlx5_wq_cyc_create(mdev, &param->wq, sqc_wq, wq, &sq->wq_ctrl);
@@ -1187,6 +1197,7 @@ static int mlx5e_create_sq(struct mlx5_core_dev *mdev,
 			   struct mlx5e_create_sq_param *csp,
 			   u32 *sqn)
 {
+	u8 ts_format;
 	void *in;
 	void *sqc;
 	void *wq;
@@ -1199,6 +1210,9 @@ static int mlx5e_create_sq(struct mlx5_core_dev *mdev,
 	if (!in)
 		return -ENOMEM;
 
+	ts_format = mlx5_is_real_time_sq(mdev) ?
+		    MLX5_SQC_TIMESTAMP_FORMAT_REAL_TIME :
+		    MLX5_SQC_TIMESTAMP_FORMAT_FREE_RUNNING;
 	sqc = MLX5_ADDR_OF(create_sq_in, in, ctx);
 	wq = MLX5_ADDR_OF(sqc, sqc, wq);
 
@@ -1207,6 +1221,8 @@ static int mlx5e_create_sq(struct mlx5_core_dev *mdev,
 	MLX5_SET(sqc,  sqc, tis_num_0, csp->tisn);
 	MLX5_SET(sqc,  sqc, cqn, csp->cqn);
 	MLX5_SET(sqc,  sqc, ts_cqe_to_dest_cqn, csp->ts_cqe_to_dest_cqn);
+	MLX5_SET(sqc,  sqc, ts_format, ts_format);
+
 
 	if (MLX5_CAP_ETH(mdev, wqe_inline_mode) == MLX5_CAP_INLINE_MODE_VPORT_CONTEXT)
 		MLX5_SET(sqc,  sqc, min_wqe_inline_mode, csp->min_inline_mode);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index 7f5851c61218..25e5ce7ec6f2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -47,7 +47,6 @@
 #include "fpga/ipsec.h"
 #include "en_accel/ipsec_rxtx.h"
 #include "en_accel/tls_rxtx.h"
-#include "lib/clock.h"
 #include "en/xdp.h"
 #include "en/xsk/rx.h"
 #include "en/health.h"
@@ -1066,9 +1065,8 @@ static inline void mlx5e_build_rx_skb(struct mlx5_cqe64 *cqe,
 	}
 
 	if (unlikely(mlx5e_rx_hw_stamp(rq->tstamp)))
-		skb_hwtstamps(skb)->hwtstamp =
-				mlx5_timecounter_cyc2time(rq->clock, get_cqe_ts(cqe));
-
+		skb_hwtstamps(skb)->hwtstamp = mlx5e_cqe_ts_to_ns(rq->ptp_cyc2time,
+								  rq->clock, get_cqe_ts(cqe));
 	skb_record_rx_queue(skb, rq->ix);
 
 	if (likely(netdev->features & NETIF_F_RXHASH))
@@ -1667,9 +1665,8 @@ static inline void mlx5i_complete_rx_cqe(struct mlx5e_rq *rq,
 	}
 
 	if (unlikely(mlx5e_rx_hw_stamp(tstamp)))
-		skb_hwtstamps(skb)->hwtstamp =
-				mlx5_timecounter_cyc2time(rq->clock, get_cqe_ts(cqe));
-
+		skb_hwtstamps(skb)->hwtstamp = mlx5e_cqe_ts_to_ns(rq->ptp_cyc2time,
+								  rq->clock, get_cqe_ts(cqe));
 	skb_record_rx_queue(skb, rq->ix);
 
 	if (likely(netdev->features & NETIF_F_RXHASH))
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
index 61ed671fe741..31ecce5c3163 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
@@ -39,7 +39,6 @@
 #include "en/txrx.h"
 #include "ipoib/ipoib.h"
 #include "en_accel/en_accel.h"
-#include "lib/clock.h"
 #include "en/ptp.h"
 
 static void mlx5e_dma_unmap_wqe_err(struct mlx5e_txqsq *sq, u8 num_dma)
@@ -774,7 +773,7 @@ static void mlx5e_consume_skb(struct mlx5e_txqsq *sq, struct sk_buff *skb,
 		struct skb_shared_hwtstamps hwts = {};
 		u64 ts = get_cqe_ts(cqe);
 
-		hwts.hwtstamp = mlx5_timecounter_cyc2time(sq->clock, ts);
+		hwts.hwtstamp = mlx5e_cqe_ts_to_ns(sq->ptp_cyc2time, sq->clock, ts);
 		if (sq->ptpsq)
 			mlx5e_skb_cb_hwtstamp_handler(skb, MLX5E_SKB_CB_CQE_HWTSTAMP,
 						      hwts.hwtstamp, sq->ptpsq->cq_stats);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index dbb6138db096..b0e129d0f6d8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -67,24 +67,51 @@ enum {
 	MLX5_MTPPS_FS_ENH_OUT_PER_ADJ		= BIT(0x7),
 };
 
-static u64 mlx5_read_internal_timer(struct mlx5_core_dev *dev,
-				    struct ptp_system_timestamp *sts)
+static bool mlx5_real_time_mode(struct mlx5_core_dev *mdev)
+{
+	return (mlx5_is_real_time_rq(mdev) || mlx5_is_real_time_sq(mdev));
+}
+
+static bool mlx5_modify_mtutc_allowed(struct mlx5_core_dev *mdev)
+{
+	return MLX5_CAP_MCAM_FEATURE(mdev, ptpcyc2realtime_modify);
+}
+
+static int mlx5_set_mtutc(struct mlx5_core_dev *dev, u32 *mtutc, u32 size)
+{
+	u32 out[MLX5_ST_SZ_DW(mtutc_reg)] = {};
+
+	if (!MLX5_CAP_MCAM_REG(dev, mtutc))
+		return -EOPNOTSUPP;
+
+	return mlx5_core_access_reg(dev, mtutc, size, out, sizeof(out),
+				    MLX5_REG_MTUTC, 0, 1);
+}
+
+static u64 mlx5_read_time(struct mlx5_core_dev *dev,
+			  struct ptp_system_timestamp *sts,
+			  bool real_time)
 {
 	u32 timer_h, timer_h1, timer_l;
 
-	timer_h = ioread32be(&dev->iseg->internal_timer_h);
+	timer_h = ioread32be(real_time ? &dev->iseg->real_time_h :
+			     &dev->iseg->internal_timer_h);
 	ptp_read_system_prets(sts);
-	timer_l = ioread32be(&dev->iseg->internal_timer_l);
+	timer_l = ioread32be(real_time ? &dev->iseg->real_time_l :
+			     &dev->iseg->internal_timer_l);
 	ptp_read_system_postts(sts);
-	timer_h1 = ioread32be(&dev->iseg->internal_timer_h);
+	timer_h1 = ioread32be(real_time ? &dev->iseg->real_time_h :
+			      &dev->iseg->internal_timer_h);
 	if (timer_h != timer_h1) {
 		/* wrap around */
 		ptp_read_system_prets(sts);
-		timer_l = ioread32be(&dev->iseg->internal_timer_l);
+		timer_l = ioread32be(real_time ? &dev->iseg->real_time_l :
+				     &dev->iseg->internal_timer_l);
 		ptp_read_system_postts(sts);
 	}
 
-	return (u64)timer_l | (u64)timer_h1 << 32;
+	return real_time ? REAL_TIME_TO_NS(timer_h1, timer_l) :
+			   (u64)timer_l | (u64)timer_h1 << 32;
 }
 
 static u64 read_internal_timer(const struct cyclecounter *cc)
@@ -94,7 +121,7 @@ static u64 read_internal_timer(const struct cyclecounter *cc)
 	struct mlx5_core_dev *mdev = container_of(clock, struct mlx5_core_dev,
 						  clock);
 
-	return mlx5_read_internal_timer(mdev, NULL) & cc->mask;
+	return mlx5_read_time(mdev, NULL, false) & cc->mask;
 }
 
 static void mlx5_update_clock_info_page(struct mlx5_core_dev *mdev)
@@ -169,23 +196,58 @@ static void mlx5_timestamp_overflow(struct work_struct *work)
 	schedule_delayed_work(&timer->overflow_work, timer->overflow_period);
 }
 
+static int mlx5_ptp_settime_real_time(struct mlx5_core_dev *mdev,
+				      const struct timespec64 *ts)
+{
+	u32 in[MLX5_ST_SZ_DW(mtutc_reg)] = {};
+
+	if (!mlx5_modify_mtutc_allowed(mdev))
+		return 0;
+
+	if (ts->tv_sec < 0 || ts->tv_sec > U32_MAX ||
+	    ts->tv_nsec < 0 || ts->tv_nsec > NSEC_PER_SEC)
+		return -EINVAL;
+
+	MLX5_SET(mtutc_reg, in, operation, MLX5_MTUTC_OPERATION_SET_TIME_IMMEDIATE);
+	MLX5_SET(mtutc_reg, in, utc_sec, ts->tv_sec);
+	MLX5_SET(mtutc_reg, in, utc_nsec, ts->tv_nsec);
+
+	return mlx5_set_mtutc(mdev, in, sizeof(in));
+}
+
 static int mlx5_ptp_settime(struct ptp_clock_info *ptp, const struct timespec64 *ts)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
 	struct mlx5_timer *timer = &clock->timer;
-	u64 ns = timespec64_to_ns(ts);
 	struct mlx5_core_dev *mdev;
 	unsigned long flags;
+	int err;
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
+	err = mlx5_ptp_settime_real_time(mdev, ts);
+	if (err)
+		return err;
+
 	write_seqlock_irqsave(&clock->lock, flags);
-	timecounter_init(&timer->tc, &timer->cycles, ns);
+	timecounter_init(&timer->tc, &timer->cycles, timespec64_to_ns(ts));
 	mlx5_update_clock_info_page(mdev);
 	write_sequnlock_irqrestore(&clock->lock, flags);
 
 	return 0;
 }
 
+static
+struct timespec64 mlx5_ptp_gettimex_real_time(struct mlx5_core_dev *mdev,
+					      struct ptp_system_timestamp *sts)
+{
+	struct timespec64 ts;
+	u64 time;
+
+	time = mlx5_read_time(mdev, sts, true);
+	ts = ns_to_timespec64(time);
+	return ts;
+}
+
 static int mlx5_ptp_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,
 			     struct ptp_system_timestamp *sts)
 {
@@ -196,24 +258,57 @@ static int mlx5_ptp_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,
 	u64 cycles, ns;
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
+	if (mlx5_real_time_mode(mdev)) {
+		*ts = mlx5_ptp_gettimex_real_time(mdev, sts);
+		goto out;
+	}
+
 	write_seqlock_irqsave(&clock->lock, flags);
-	cycles = mlx5_read_internal_timer(mdev, sts);
+	cycles = mlx5_read_time(mdev, sts, false);
 	ns = timecounter_cyc2time(&timer->tc, cycles);
 	write_sequnlock_irqrestore(&clock->lock, flags);
-
 	*ts = ns_to_timespec64(ns);
-
+out:
 	return 0;
 }
 
+static int mlx5_ptp_adjtime_real_time(struct mlx5_core_dev *mdev, s64 delta)
+{
+	u32 in[MLX5_ST_SZ_DW(mtutc_reg)] = {};
+
+	if (!mlx5_modify_mtutc_allowed(mdev))
+		return 0;
+
+	/* HW time adjustment range is s16. If out of range, settime instead */
+	if (delta < S16_MIN || delta > S16_MAX) {
+		struct timespec64 ts;
+		s64 ns;
+
+		ts = mlx5_ptp_gettimex_real_time(mdev, NULL);
+		ns = timespec64_to_ns(&ts) + delta;
+		ts = ns_to_timespec64(ns);
+		return mlx5_ptp_settime_real_time(mdev, &ts);
+	}
+
+	MLX5_SET(mtutc_reg, in, operation, MLX5_MTUTC_OPERATION_ADJUST_TIME);
+	MLX5_SET(mtutc_reg, in, time_adjustment, delta);
+
+	return mlx5_set_mtutc(mdev, in, sizeof(in));
+}
+
 static int mlx5_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
 	struct mlx5_timer *timer = &clock->timer;
 	struct mlx5_core_dev *mdev;
 	unsigned long flags;
+	int err;
 
 	mdev = container_of(clock, struct mlx5_core_dev, clock);
+
+	err = mlx5_ptp_adjtime_real_time(mdev, delta);
+	if (err)
+		return err;
 	write_seqlock_irqsave(&clock->lock, flags);
 	timecounter_adjtime(&timer->tc, delta);
 	mlx5_update_clock_info_page(mdev);
@@ -222,6 +317,19 @@ static int mlx5_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
+static int mlx5_ptp_adjfreq_real_time(struct mlx5_core_dev *mdev, s32 freq)
+{
+	u32 in[MLX5_ST_SZ_DW(mtutc_reg)] = {};
+
+	if (!mlx5_modify_mtutc_allowed(mdev))
+		return 0;
+
+	MLX5_SET(mtutc_reg, in, operation, MLX5_MTUTC_OPERATION_ADJUST_FREQ_UTC);
+	MLX5_SET(mtutc_reg, in, freq_adjustment, freq);
+
+	return mlx5_set_mtutc(mdev, in, sizeof(in));
+}
+
 static int mlx5_ptp_adjfreq(struct ptp_clock_info *ptp, s32 delta)
 {
 	struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);
@@ -231,6 +339,12 @@ static int mlx5_ptp_adjfreq(struct ptp_clock_info *ptp, s32 delta)
 	int neg_adj = 0;
 	u32 diff;
 	u64 adj;
+	int err;
+
+	mdev = container_of(clock, struct mlx5_core_dev, clock);
+	err = mlx5_ptp_adjfreq_real_time(mdev, delta);
+	if (err)
+		return err;
 
 	if (delta < 0) {
 		neg_adj = 1;
@@ -241,7 +355,6 @@ static int mlx5_ptp_adjfreq(struct ptp_clock_info *ptp, s32 delta)
 	adj *= delta;
 	diff = div_u64(adj, 1000000000ULL);
 
-	mdev = container_of(clock, struct mlx5_core_dev, clock);
 	write_seqlock_irqsave(&clock->lock, flags);
 	timecounter_read(&timer->tc);
 	timer->cycles.mult = neg_adj ? timer->nominal_c_mult - diff :
@@ -324,7 +437,7 @@ static u64 find_target_cycles(struct mlx5_core_dev *mdev, s64 target_ns)
 
 	timer = &clock->timer;
 
-	cycles_now = mlx5_read_internal_timer(mdev, NULL);
+	cycles_now = mlx5_read_time(mdev, NULL, false);
 	write_seqlock_irqsave(&clock->lock, flags);
 	nsec_now = timecounter_cyc2time(&timer->tc, cycles_now);
 	nsec_delta = target_ns - nsec_now;
@@ -348,6 +461,11 @@ static u64 perout_conf_internal_timer(struct mlx5_core_dev *mdev,
 	return find_target_cycles(mdev, target_ns);
 }
 
+static u64 perout_conf_real_time(s64 sec, u32 nsec)
+{
+	return (u64)nsec | (u64)sec << 32;
+}
+
 static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 				 struct ptp_clock_request *rq,
 				 int on)
@@ -378,6 +496,7 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 
 	field_select = MLX5_MTPPS_FS_ENABLE;
 	if (on) {
+		bool rt_mode = mlx5_real_time_mode(mdev);
 		u32 nsec;
 		s64 sec;
 
@@ -397,7 +516,11 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
 		nsec = rq->perout.start.nsec;
 		sec = rq->perout.start.sec;
 
-		time_stamp = perout_conf_internal_timer(mdev, sec, nsec);
+		if (rt_mode && sec > U32_MAX)
+			return -EINVAL;
+
+		time_stamp = rt_mode ? perout_conf_real_time(sec, nsec) :
+				       perout_conf_internal_timer(mdev, sec, nsec);
 
 		field_select |= MLX5_MTPPS_FS_PIN_MODE |
 				MLX5_MTPPS_FS_PATTERN |
@@ -578,17 +701,23 @@ static void ts_next_sec(struct timespec64 *ts)
 	ts->tv_nsec = 0;
 }
 
-static u64 perout_conf_next_event_inernal_timer(struct mlx5_core_dev *mdev,
-						struct mlx5_clock *clock)
+static u64 perout_conf_next_event_timer(struct mlx5_core_dev *mdev,
+					struct mlx5_clock *clock)
 {
+	bool rt_mode = mlx5_real_time_mode(mdev);
 	struct timespec64 ts;
 	s64 target_ns;
 
-	mlx5_ptp_gettimex(&clock->ptp_info, &ts, NULL);
+	if (rt_mode)
+		ts = mlx5_ptp_gettimex_real_time(mdev, NULL);
+	else
+		mlx5_ptp_gettimex(&clock->ptp_info, &ts, NULL);
+
 	ts_next_sec(&ts);
 	target_ns = timespec64_to_ns(&ts);
 
-	return find_target_cycles(mdev, target_ns);
+	return rt_mode ? perout_conf_real_time(ts.tv_sec, ts.tv_nsec) :
+			 find_target_cycles(mdev, target_ns);
 }
 
 static int mlx5_pps_event(struct notifier_block *nb,
@@ -607,7 +736,9 @@ static int mlx5_pps_event(struct notifier_block *nb,
 	switch (clock->ptp_info.pin_config[pin].func) {
 	case PTP_PF_EXTTS:
 		ptp_event.index = pin;
-		ptp_event.timestamp =
+		ptp_event.timestamp = mlx5_real_time_mode(mdev) ?
+			mlx5_real_time_cyc2time(clock,
+						be64_to_cpu(eqe->data.pps.time_stamp)) :
 			mlx5_timecounter_cyc2time(clock,
 						  be64_to_cpu(eqe->data.pps.time_stamp));
 		if (clock->pps_info.enabled) {
@@ -621,7 +752,7 @@ static int mlx5_pps_event(struct notifier_block *nb,
 		ptp_clock_event(clock->ptp, &ptp_event);
 		break;
 	case PTP_PF_PEROUT:
-		ns = perout_conf_next_event_inernal_timer(mdev, clock);
+		ns = perout_conf_next_event_timer(mdev, clock);
 		write_seqlock_irqsave(&clock->lock, flags);
 		clock->pps_info.start[pin] = ns;
 		write_sequnlock_irqrestore(&clock->lock, flags);
@@ -711,6 +842,23 @@ static void mlx5_init_clock_info(struct mlx5_core_dev *mdev)
 	info->frac = timer->tc.frac;
 }
 
+static void mlx5_init_timer_clock(struct mlx5_core_dev *mdev)
+{
+	struct mlx5_clock *clock = &mdev->clock;
+
+	mlx5_timecounter_init(mdev);
+	mlx5_init_clock_info(mdev);
+	mlx5_init_overflow_period(clock);
+	clock->ptp_info = mlx5_ptp_clock_info;
+
+	if (mlx5_real_time_mode(mdev)) {
+		struct timespec64 ts;
+
+		ktime_get_real_ts64(&ts);
+		mlx5_ptp_settime(&clock->ptp_info, &ts);
+	}
+}
+
 void mlx5_init_clock(struct mlx5_core_dev *mdev)
 {
 	struct mlx5_clock *clock = &mdev->clock;
@@ -721,10 +869,7 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
 	}
 
 	seqlock_init(&clock->lock);
-
-	mlx5_timecounter_init(mdev);
-	mlx5_init_clock_info(mdev);
-	mlx5_init_overflow_period(clock);
+	mlx5_init_timer_clock(mdev);
 	INIT_WORK(&clock->pps_info.out_work, mlx5_pps_out);
 
 	/* Configure the PHC */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
index 6e8804ebc773..a12c7da618a7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
@@ -33,6 +33,24 @@
 #ifndef __LIB_CLOCK_H__
 #define __LIB_CLOCK_H__
 
+static inline bool mlx5_is_real_time_rq(struct mlx5_core_dev *mdev)
+{
+	u8 rq_ts_format_cap = MLX5_CAP_GEN(mdev, rq_ts_format);
+
+	return (rq_ts_format_cap == MLX5_RQ_TIMESTAMP_FORMAT_CAP_REAL_TIME  ||
+		rq_ts_format_cap == MLX5_RQ_TIMESTAMP_FORMAT_CAP_FREE_RUNNING_AND_REAL_TIME);
+}
+
+static inline bool mlx5_is_real_time_sq(struct mlx5_core_dev *mdev)
+{
+	u8 sq_ts_format_cap = MLX5_CAP_GEN(mdev, sq_ts_format);
+
+	return (sq_ts_format_cap == MLX5_SQ_TIMESTAMP_FORMAT_CAP_REAL_TIME  ||
+		sq_ts_format_cap == MLX5_SQ_TIMESTAMP_FORMAT_CAP_FREE_RUNNING_AND_REAL_TIME);
+}
+
+typedef ktime_t (*cqe_ts_to_ns)(struct mlx5_clock *, u64);
+
 #if IS_ENABLED(CONFIG_PTP_1588_CLOCK)
 void mlx5_init_clock(struct mlx5_core_dev *mdev);
 void mlx5_cleanup_clock(struct mlx5_core_dev *mdev);
@@ -57,6 +75,15 @@ static inline ktime_t mlx5_timecounter_cyc2time(struct mlx5_clock *clock,
 	return ns_to_ktime(nsec);
 }
 
+#define REAL_TIME_TO_NS(hi, low) (((u64)hi) * NSEC_PER_SEC + ((u64)low))
+
+static inline ktime_t mlx5_real_time_cyc2time(struct mlx5_clock *clock,
+					      u64 timestamp)
+{
+	u64 time = REAL_TIME_TO_NS(timestamp >> 32, timestamp & 0xFFFFFFFF);
+
+	return ns_to_ktime(time);
+}
 #else
 static inline void mlx5_init_clock(struct mlx5_core_dev *mdev) {}
 static inline void mlx5_cleanup_clock(struct mlx5_core_dev *mdev) {}
@@ -70,6 +97,12 @@ static inline ktime_t mlx5_timecounter_cyc2time(struct mlx5_clock *clock,
 {
 	return 0;
 }
+
+static inline ktime_t mlx5_real_time_cyc2time(struct mlx5_clock *clock,
+					      u64 timestamp)
+{
+	return 0;
+}
 #endif
 
 #endif
-- 
2.29.2