Merge branch 'net-dsa-mv88e6xxx-fix-mv88e6393x-phc-frequency-on-internal-clock'

Shenghao Yang says:

====================
net: dsa: mv88e6xxx: fix MV88E6393X PHC frequency on internal clock

The MV88E6393X family of switches can additionally run their cycle
counters using a 250MHz internal clock instead of the usual 125MHz
external clock [1].

The driver currently assumes all designs utilize that external clock,
but MikroTik's RB5009 uses the internal source - causing the PHC to be
seen running at 2x real time in userspace, making synchronization
with ptp4l impossible.

This series adds support for reading off the cycle counter frequency
known to the hardware in the TAI_CLOCK_PERIOD register and picking an
appropriate set of scaling coefficients instead of using a fixed set
for each switch family.

Patch 1 groups those cycle counter coefficients into a new structure to
make it easier to pass them around.

Patch 2 modifies PTP initialization to probe TAI_CLOCK_PERIOD and
use an appropriate set of coefficients.

Patch 3 adds support for 4000ps cycle counter periods.

Changes since v2 [2]:

- Patch 1: "net: dsa: mv88e6xxx: group cycle counter coefficients"
  - Moved declaration of mv88e6xxx_cc_coeffs to avoid moving that in
    Patch 2.

- Patch 2: "net: dsa: mv88e6xxx: read cycle counter period from hardware"
  - Removed move of mv88e6xxx_cc_coeffs declaration.

- Patch 3: "net: dsa: mv88e6xxx: support 4000ps cycle counter periods"
  - No change.

[1] https://lore.kernel.org/netdev/d6622575-bf1b-445a-b08f-2739e3642aae@lunn.ch/
[2] https://lore.kernel.org/netdev/20241006145951.719162-1-me@shenghaoyang.info/
====================

Link: https://patch.msgid.link/20241020063833.5425-1-me@shenghaoyang.info
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

2 files changed, 77 insertions, 37 deletions

diff --git a/drivers/net/dsa/mv88e6xxx/chip.h b/drivers/net/dsa/mv88e6xxx/chip.h
index c34caf9815c5..a54682240839 100644
--- a/drivers/net/dsa/mv88e6xxx/chip.h
+++ b/drivers/net/dsa/mv88e6xxx/chip.h
@@ -206,6 +206,7 @@ struct mv88e6xxx_gpio_ops;
 struct mv88e6xxx_avb_ops;
 struct mv88e6xxx_ptp_ops;
 struct mv88e6xxx_pcs_ops;
+struct mv88e6xxx_cc_coeffs;
 
 struct mv88e6xxx_irq {
 	u16 masked;
@@ -408,6 +409,7 @@ struct mv88e6xxx_chip {
 	struct cyclecounter	tstamp_cc;
 	struct timecounter	tstamp_tc;
 	struct delayed_work	overflow_work;
+	const struct mv88e6xxx_cc_coeffs *cc_coeffs;
 
 	struct ptp_clock	*ptp_clock;
 	struct ptp_clock_info	ptp_clock_info;
@@ -731,10 +733,6 @@ struct mv88e6xxx_ptp_ops {
 	int arr1_sts_reg;
 	int dep_sts_reg;
 	u32 rx_filters;
-	u32 cc_shift;
-	u32 cc_mult;
-	u32 cc_mult_num;
-	u32 cc_mult_dem;
 };
 
 struct mv88e6xxx_pcs_ops {
diff --git a/drivers/net/dsa/mv88e6xxx/ptp.c b/drivers/net/dsa/mv88e6xxx/ptp.c
index 56391e09b325..aed4a4b07f34 100644
--- a/drivers/net/dsa/mv88e6xxx/ptp.c
+++ b/drivers/net/dsa/mv88e6xxx/ptp.c
@@ -18,6 +18,13 @@
 
 #define MV88E6XXX_MAX_ADJ_PPB	1000000
 
+struct mv88e6xxx_cc_coeffs {
+	u32 cc_shift;
+	u32 cc_mult;
+	u32 cc_mult_num;
+	u32 cc_mult_dem;
+};
+
 /* Family MV88E6250:
  * Raw timestamps are in units of 10-ns clock periods.
  *
@@ -25,22 +32,43 @@
  * simplifies to
  * clkadj = scaled_ppm * 2^7 / 5^5
  */
-#define MV88E6250_CC_SHIFT	28
-#define MV88E6250_CC_MULT	(10 << MV88E6250_CC_SHIFT)
-#define MV88E6250_CC_MULT_NUM	(1 << 7)
-#define MV88E6250_CC_MULT_DEM	3125ULL
+#define MV88E6XXX_CC_10NS_SHIFT 28
+static const struct mv88e6xxx_cc_coeffs mv88e6xxx_cc_10ns_coeffs = {
+	.cc_shift = MV88E6XXX_CC_10NS_SHIFT,
+	.cc_mult = 10 << MV88E6XXX_CC_10NS_SHIFT,
+	.cc_mult_num = 1 << 7,
+	.cc_mult_dem = 3125ULL,
+};
 
-/* Other families:
+/* Other families except MV88E6393X in internal clock mode:
  * Raw timestamps are in units of 8-ns clock periods.
  *
  * clkadj = scaled_ppm * 8*2^28 / (10^6 * 2^16)
  * simplifies to
  * clkadj = scaled_ppm * 2^9 / 5^6
  */
-#define MV88E6XXX_CC_SHIFT	28
-#define MV88E6XXX_CC_MULT	(8 << MV88E6XXX_CC_SHIFT)
-#define MV88E6XXX_CC_MULT_NUM	(1 << 9)
-#define MV88E6XXX_CC_MULT_DEM	15625ULL
+#define MV88E6XXX_CC_8NS_SHIFT 28
+static const struct mv88e6xxx_cc_coeffs mv88e6xxx_cc_8ns_coeffs = {
+	.cc_shift = MV88E6XXX_CC_8NS_SHIFT,
+	.cc_mult = 8 << MV88E6XXX_CC_8NS_SHIFT,
+	.cc_mult_num = 1 << 9,
+	.cc_mult_dem = 15625ULL
+};
+
+/* Family MV88E6393X using internal clock:
+ * Raw timestamps are in units of 4-ns clock periods.
+ *
+ * clkadj = scaled_ppm * 4*2^28 / (10^6 * 2^16)
+ * simplifies to
+ * clkadj = scaled_ppm * 2^8 / 5^6
+ */
+#define MV88E6XXX_CC_4NS_SHIFT 28
+static const struct mv88e6xxx_cc_coeffs mv88e6xxx_cc_4ns_coeffs = {
+	.cc_shift = MV88E6XXX_CC_4NS_SHIFT,
+	.cc_mult = 4 << MV88E6XXX_CC_4NS_SHIFT,
+	.cc_mult_num = 1 << 8,
+	.cc_mult_dem = 15625ULL
+};
 
 #define TAI_EVENT_WORK_INTERVAL msecs_to_jiffies(100)
 
@@ -83,6 +111,33 @@ static int mv88e6352_set_gpio_func(struct mv88e6xxx_chip *chip, int pin,
 	return chip->info->ops->gpio_ops->set_pctl(chip, pin, func);
 }
 
+static const struct mv88e6xxx_cc_coeffs *
+mv88e6xxx_cc_coeff_get(struct mv88e6xxx_chip *chip)
+{
+	u16 period_ps;
+	int err;
+
+	err = mv88e6xxx_tai_read(chip, MV88E6XXX_TAI_CLOCK_PERIOD, &period_ps, 1);
+	if (err) {
+		dev_err(chip->dev, "failed to read cycle counter period: %d\n",
+			err);
+		return ERR_PTR(err);
+	}
+
+	switch (period_ps) {
+	case 4000:
+		return &mv88e6xxx_cc_4ns_coeffs;
+	case 8000:
+		return &mv88e6xxx_cc_8ns_coeffs;
+	case 10000:
+		return &mv88e6xxx_cc_10ns_coeffs;
+	default:
+		dev_err(chip->dev, "unexpected cycle counter period of %u ps\n",
+			period_ps);
+		return ERR_PTR(-ENODEV);
+	}
+}
+
 static u64 mv88e6352_ptp_clock_read(const struct cyclecounter *cc)
 {
 	struct mv88e6xxx_chip *chip = cc_to_chip(cc);
@@ -204,7 +259,6 @@ out:
 static int mv88e6xxx_ptp_adjfine(struct ptp_clock_info *ptp, long scaled_ppm)
 {
 	struct mv88e6xxx_chip *chip = ptp_to_chip(ptp);
-	const struct mv88e6xxx_ptp_ops *ptp_ops = chip->info->ops->ptp_ops;
 	int neg_adj = 0;
 	u32 diff, mult;
 	u64 adj;
@@ -214,10 +268,10 @@ static int mv88e6xxx_ptp_adjfine(struct ptp_clock_info *ptp, long scaled_ppm)
 		scaled_ppm = -scaled_ppm;
 	}
 
-	mult = ptp_ops->cc_mult;
-	adj = ptp_ops->cc_mult_num;
+	mult = chip->cc_coeffs->cc_mult;
+	adj = chip->cc_coeffs->cc_mult_num;
 	adj *= scaled_ppm;
-	diff = div_u64(adj, ptp_ops->cc_mult_dem);
+	diff = div_u64(adj, chip->cc_coeffs->cc_mult_dem);
 
 	mv88e6xxx_reg_lock(chip);
 
@@ -364,10 +418,6 @@ const struct mv88e6xxx_ptp_ops mv88e6165_ptp_ops = {
 		(1 << HWTSTAMP_FILTER_PTP_V2_EVENT) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_SYNC) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_DELAY_REQ),
-	.cc_shift = MV88E6XXX_CC_SHIFT,
-	.cc_mult = MV88E6XXX_CC_MULT,
-	.cc_mult_num = MV88E6XXX_CC_MULT_NUM,
-	.cc_mult_dem = MV88E6XXX_CC_MULT_DEM,
 };
 
 const struct mv88e6xxx_ptp_ops mv88e6250_ptp_ops = {
@@ -391,10 +441,6 @@ const struct mv88e6xxx_ptp_ops mv88e6250_ptp_ops = {
 		(1 << HWTSTAMP_FILTER_PTP_V2_EVENT) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_SYNC) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_DELAY_REQ),
-	.cc_shift = MV88E6250_CC_SHIFT,
-	.cc_mult = MV88E6250_CC_MULT,
-	.cc_mult_num = MV88E6250_CC_MULT_NUM,
-	.cc_mult_dem = MV88E6250_CC_MULT_DEM,
 };
 
 const struct mv88e6xxx_ptp_ops mv88e6352_ptp_ops = {
@@ -418,10 +464,6 @@ const struct mv88e6xxx_ptp_ops mv88e6352_ptp_ops = {
 		(1 << HWTSTAMP_FILTER_PTP_V2_EVENT) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_SYNC) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_DELAY_REQ),
-	.cc_shift = MV88E6XXX_CC_SHIFT,
-	.cc_mult = MV88E6XXX_CC_MULT,
-	.cc_mult_num = MV88E6XXX_CC_MULT_NUM,
-	.cc_mult_dem = MV88E6XXX_CC_MULT_DEM,
 };
 
 const struct mv88e6xxx_ptp_ops mv88e6390_ptp_ops = {
@@ -446,10 +488,6 @@ const struct mv88e6xxx_ptp_ops mv88e6390_ptp_ops = {
 		(1 << HWTSTAMP_FILTER_PTP_V2_EVENT) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_SYNC) |
 		(1 << HWTSTAMP_FILTER_PTP_V2_DELAY_REQ),
-	.cc_shift = MV88E6XXX_CC_SHIFT,
-	.cc_mult = MV88E6XXX_CC_MULT,
-	.cc_mult_num = MV88E6XXX_CC_MULT_NUM,
-	.cc_mult_dem = MV88E6XXX_CC_MULT_DEM,
 };
 
 static u64 mv88e6xxx_ptp_clock_read(const struct cyclecounter *cc)
@@ -462,10 +500,10 @@ static u64 mv88e6xxx_ptp_clock_read(const struct cyclecounter *cc)
 	return 0;
 }
 
-/* With a 125MHz input clock, the 32-bit timestamp counter overflows in ~34.3
+/* With a 250MHz input clock, the 32-bit timestamp counter overflows in ~17.2
  * seconds; this task forces periodic reads so that we don't miss any.
  */
-#define MV88E6XXX_TAI_OVERFLOW_PERIOD (HZ * 16)
+#define MV88E6XXX_TAI_OVERFLOW_PERIOD (HZ * 8)
 static void mv88e6xxx_ptp_overflow_check(struct work_struct *work)
 {
 	struct delayed_work *dw = to_delayed_work(work);
@@ -484,11 +522,15 @@ int mv88e6xxx_ptp_setup(struct mv88e6xxx_chip *chip)
 	int i;
 
 	/* Set up the cycle counter */
+	chip->cc_coeffs = mv88e6xxx_cc_coeff_get(chip);
+	if (IS_ERR(chip->cc_coeffs))
+		return PTR_ERR(chip->cc_coeffs);
+
 	memset(&chip->tstamp_cc, 0, sizeof(chip->tstamp_cc));
 	chip->tstamp_cc.read	= mv88e6xxx_ptp_clock_read;
 	chip->tstamp_cc.mask	= CYCLECOUNTER_MASK(32);
-	chip->tstamp_cc.mult	= ptp_ops->cc_mult;
-	chip->tstamp_cc.shift	= ptp_ops->cc_shift;
+	chip->tstamp_cc.mult	= chip->cc_coeffs->cc_mult;
+	chip->tstamp_cc.shift	= chip->cc_coeffs->cc_shift;
 
 	timecounter_init(&chip->tstamp_tc, &chip->tstamp_cc,
 			 ktime_to_ns(ktime_get_real()));