[net-next PATCH 3/3] octeontx2-af: cn10k: add workaround for ptp errata

[Rakesh Babu Saladi]

From: Naveen Mamindlapalli <naveenm@xxxxxxxxxxx>

This patch adds workaround for PTP errata given below.

1. At the time of 1 sec rollover of nano-second counter,
   the nano-second counter is set to 0. However, it should
   be set to (existing counter_value - 10^9). This leads to
   an accumulating error in the timestamp value with each sec
   rollover.
2. Additionally, the nano-second counter currently is rolling
   over at 'h3B9A_C9FF. It should roll over at 'h3B9A_CA00.

The workaround for issue #1 is to speed up the ptp clock by
adjusting PTP_CLOCK_COMP register to the desired value to
compensate for the nanoseconds lost per each second.

The workaround for issue #2 is to slow down the ptp clock
such that the rollover occurs at ~1sec.

Signed-off-by: Naveen Mamindlapalli <naveenm@xxxxxxxxxxx>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@xxxxxxxxxxx>
Signed-off-by: Rakesh Babu Saladi <rsaladi2@xxxxxxxxxxx>
---
 .../net/ethernet/marvell/octeontx2/af/ptp.c   | 87 +++++++++++++++++--
 1 file changed, 80 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/marvell/octeontx2/af/ptp.c b/drivers/net/ethernet/marvell/octeontx2/af/ptp.c
index 6f5e1a5d957f..faf8f34421f3 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/ptp.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/ptp.c
@@ -51,9 +51,19 @@
 #define PTP_TIMESTAMP				0xF20ULL
 #define PTP_CLOCK_SEC				0xFD0ULL
 
+#define CYCLE_MULT				1000
+
 static struct ptp *first_ptp_block;
 static const struct pci_device_id ptp_id_table[];
 
+static bool cn10k_ptp_errata(struct ptp *ptp)
+{
+	if (ptp->pdev->subsystem_device == PCI_SUBSYS_DEVID_CN10K_A_PTP ||
+	    ptp->pdev->subsystem_device == PCI_SUBSYS_DEVID_CNF10K_A_PTP)
+		return true;
+	return false;
+}
+
 static bool is_ptp_tsfmt_sec_nsec(struct ptp *ptp)
 {
 	if (ptp->pdev->subsystem_device == PCI_SUBSYS_DEVID_CN10K_A_PTP ||
@@ -86,6 +96,58 @@ static u64 read_ptp_tstmp_nsec(struct ptp *ptp)
 	return readq(ptp->reg_base + PTP_CLOCK_HI);
 }
 
+static u64 ptp_calc_adjusted_comp(u64 ptp_clock_freq)
+{
+	u64 comp, adj = 0, cycles_per_sec, ns_drift = 0;
+	u32 ptp_clock_nsec, cycle_time;
+	int cycle;
+
+	/* Errata:
+	 * Issue #1: At the time of 1 sec rollover of the nano-second counter,
+	 * the nano-second counter is set to 0. However, it should be set to
+	 * (existing counter_value - 10^9).
+	 *
+	 * Issue #2: The nano-second counter rolls over at 0x3B9A_C9FF.
+	 * It should roll over at 0x3B9A_CA00.
+	 */
+
+	/* calculate ptp_clock_comp value */
+	comp = ((u64)1000000000ULL << 32) / ptp_clock_freq;
+	/* use CYCLE_MULT to avoid accuracy loss due to integer arithmetic */
+	cycle_time = NSEC_PER_SEC * CYCLE_MULT / ptp_clock_freq;
+	/* cycles per sec */
+	cycles_per_sec = ptp_clock_freq;
+
+	/* check whether ptp nanosecond counter rolls over early */
+	cycle = cycles_per_sec - 1;
+	ptp_clock_nsec = (cycle * comp) >> 32;
+	while (ptp_clock_nsec < NSEC_PER_SEC) {
+		if (ptp_clock_nsec == 0x3B9AC9FF)
+			goto calc_adj_comp;
+		cycle++;
+		ptp_clock_nsec = (cycle * comp) >> 32;
+	}
+	/* compute nanoseconds lost per second when nsec counter rolls over */
+	ns_drift = ptp_clock_nsec - NSEC_PER_SEC;
+	/* calculate ptp_clock_comp adjustment */
+	if (ns_drift > 0) {
+		adj = comp * ns_drift;
+		adj = adj / 1000000000ULL;
+	}
+	/* speed up the ptp clock to account for nanoseconds lost */
+	comp += adj;
+	return comp;
+
+calc_adj_comp:
+	/* slow down the ptp clock to not rollover early */
+	adj = comp * cycle_time;
+	adj = adj / 1000000000ULL;
+	adj = adj / CYCLE_MULT;
+	comp -= adj;
+
+	return comp;
+}
+
 struct ptp *ptp_get(void)
 {
 	struct ptp *ptp = first_ptp_block;
@@ -113,8 +175,8 @@ void ptp_put(struct ptp *ptp)
 static int ptp_adjfine(struct ptp *ptp, long scaled_ppm)
 {
 	bool neg_adj = false;
-	u64 comp;
-	u64 adj;
+	u32 freq, freq_adj;
+	u64 comp, adj;
 	s64 ppb;
 
 	if (scaled_ppm < 0) {
@@ -136,15 +198,22 @@ static int ptp_adjfine(struct ptp *ptp, long scaled_ppm)
 	 * where tbase is the basic compensation value calculated
 	 * initialy in the probe function.
 	 */
-	comp = ((u64)1000000000ull << 32) / ptp->clock_rate;
 	/* convert scaled_ppm to ppb */
 	ppb = 1 + scaled_ppm;
 	ppb *= 125;
 	ppb >>= 13;
-	adj = comp * ppb;
-	adj = div_u64(adj, 1000000000ull);
-	comp = neg_adj ? comp - adj : comp + adj;
 
+	if (cn10k_ptp_errata(ptp)) {
+		/* calculate the new frequency based on ppb */
+		freq_adj = (ptp->clock_rate * ppb) / 1000000000ULL;
+		freq = neg_adj ? ptp->clock_rate + freq_adj : ptp->clock_rate - freq_adj;
+		comp = ptp_calc_adjusted_comp(freq);
+	} else {
+		comp = ((u64)1000000000ull << 32) / ptp->clock_rate;
+		adj = comp * ppb;
+		adj = div_u64(adj, 1000000000ull);
+		comp = neg_adj ? comp - adj : comp + adj;
+	}
 	writeq(comp, ptp->reg_base + PTP_CLOCK_COMP);
 
 	return 0;
@@ -220,7 +289,11 @@ void ptp_start(struct ptp *ptp, u64 sclk, u32 ext_clk_freq, u32 extts)
 	writeq(0x1dcd650000000000, ptp->reg_base + PTP_PPS_HI_INCR);
 	writeq(0x1dcd650000000000, ptp->reg_base + PTP_PPS_LO_INCR);
 
-	clock_comp = ((u64)1000000000ull << 32) / ptp->clock_rate;
+	if (cn10k_ptp_errata(ptp))
+		clock_comp = ptp_calc_adjusted_comp(ptp->clock_rate);
+	else
+		clock_comp = ((u64)1000000000ull << 32) / ptp->clock_rate;
+
 	/* Initial compensation value to start the nanosecs counter */
 	writeq(clock_comp, ptp->reg_base + PTP_CLOCK_COMP);
 }
-- 
2.17.1