ptp: get ready for 2038

PATCH Superseded Richard Cochran 6 participants 40 messages Mar 21, 2015 – Mar 31, 2015

Subsystems (15)

Networking Network drivers Arm STM32 Intel Ethernet +11 more subsystems

lore

Share & tools

lore

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #0 lore raw

v2 cover ptp: get ready for 2038

This series converts the core driver methods of the PTP Hardware Clock (PHC) subsystem to use the 64 bit version of the timespec structure, making the core API ready for the year 2038.

In addition, I reviewed how each driver and device represents the time value at the hardware register level. Most of the drivers are ready, but a few will need some work before the year 2038, as shown:

Patch Driver
------------------------------------------------

   12      drivers/net/ethernet/intel/igb/igb_ptp.c
   15 ?    drivers/net/ethernet/sfc/ptp.c
   16      drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c
   18 ?    drivers/net/ethernet/tile/tilegx.c
   19      drivers/net/phy/dp83640.c

The commit log message documents how each driver is ready or why it is not ready. For patches 15 and 18, I could not easily find out the hardware representation of the time value, so I would ask the maintainers for a review.

Thanks,
Richard

* ChangeLog
** V2
- use the new methods in the posix clock code right away (patch #3)

Richard Cochran (23):
ptp: introduce get/set time methods with explicit 64 bit seconds.
ptp: use the 64 bit gettime method for the SYS_OFFSET ioctl.
ptp: use the 64 bit get/set time methods for the posix clock.
ptp: blackfin: convert to the 64 bit get/set time methods.
ptp: xgbe: convert to the 64 bit get/set time methods.
ptp: bnx2x: convert to the 64 bit get/set time methods.
ptp: tg3: convert to the 64 bit get/set time methods.
ptp: fec: convert to the 64 bit get/set time methods.
ptp: gianfar: convert to the 64 bit get/set time methods.
ptp: e1000e: convert to the 64 bit get/set time methods.
ptp: fm10k: convert to the 64 bit get/set time methods.
ptp: i40e: convert to the 64 bit get/set time methods.
ptp: igb: convert to the 64 bit get/set time methods.
ptp: ixgbe: convert to the 64 bit get/set time methods.
ptp: mlx4: convert to the 64 bit get/set time methods.
ptp: sfc: convert to the 64 bit get/set time methods.
ptp: stmmac: convert to the 64 bit get/set time methods.
ptp: cpts: convert to the 64 bit get/set time methods.
ptp: tilegx: convert to the 64 bit get/set time methods.
ptp: dp83640: convert to the 64 bit get/set time methods.
ptp: ixp46x: convert to the 64 bit get/set time methods.
ptp: pch: convert to the 64 bit get/set time methods.
ptp: remove 32 bit get/set methods.

 drivers/net/ethernet/adi/bfin_mac.c              |    8 ++---
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c         |   13 +++----
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c |    8 ++---
 drivers/net/ethernet/broadcom/tg3.c              |   10 +++---
 drivers/net/ethernet/freescale/fec_ptp.c         |    8 ++---
 drivers/net/ethernet/freescale/gianfar_ptp.c     |   13 +++----
 drivers/net/ethernet/intel/e1000e/ptp.c          |   16 ++++-----
 drivers/net/ethernet/intel/fm10k/fm10k_ptp.c     |   12 +++----
 drivers/net/ethernet/intel/i40e/i40e_ptp.c       |   26 +++++++-------
 drivers/net/ethernet/intel/igb/igb_ptp.c         |   41 +++++++++++-----------
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c     |   14 ++++----
 drivers/net/ethernet/mellanox/mlx4/en_clock.c    |   11 +++---
 drivers/net/ethernet/sfc/ptp.c                   |   22 ++++++------
 drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c |    8 ++---
 drivers/net/ethernet/ti/cpts.c                   |   12 +++----
 drivers/net/ethernet/tile/tilegx.c               |   16 +++++----
 drivers/net/phy/dp83640.c                        |   12 ++++---
 drivers/ptp/ptp_chardev.c                        |    8 ++---
 drivers/ptp/ptp_clock.c                          |   12 +++++--
 drivers/ptp/ptp_ixp46x.c                         |    8 ++---
 drivers/ptp/ptp_pch.c                            |    8 ++---
 include/linux/ptp_clock_kernel.h                 |   12 +++----
 22 files changed, 158 insertions(+), 140 deletions(-)

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #1 lore raw

PATCH v2 1/23 ptp: introduce get/set time methods with explicit 64 bit seconds.

Converting the PHC drivers over to the new methods is one step along the way to making them ready for 2038. Once all the drivers are up to date, then the old methods will be removed.

Signed-off-by Richard Cochran <[email protected]>

---
 include/linux/ptp_clock_kernel.h |   12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

include/linux/ptp_clock_kernel.h +10 -2

index 0d8ff3f..7d6f8e6 100644
@@ -64,12 +64,18 @@ struct ptp_clock_request {
  * @adjtime:  Shifts the time of the hardware clock.
  *            parameter delta: Desired change in nanoseconds.
  *
- * @gettime:  Reads the current time from the hardware clock.
+ * @gettime:  Reads the current time from the hardware clock. (deprecated)
  *            parameter ts: Holds the result.
  *
- * @settime:  Set the current time on the hardware clock.
+ * @settime:  Set the current time on the hardware clock. (deprecated)
  *            parameter ts: Time value to set.
  *
+ * @gettime64:  Reads the current time from the hardware clock.
+ *              parameter ts: Holds the result.
+ *
+ * @settime64:  Set the current time on the hardware clock.
+ *              parameter ts: Time value to set.
+ *
  * @enable:   Request driver to enable or disable an ancillary feature.
  *            parameter request: Desired resource to enable or disable.
  *            parameter on: Caller passes one to enable or zero to disable.

@@ -106,6 +112,8 @@ struct ptp_clock_info {
 	int (*adjtime)(struct ptp_clock_info *ptp, s64 delta);
 	int (*gettime)(struct ptp_clock_info *ptp, struct timespec *ts);
 	int (*settime)(struct ptp_clock_info *ptp, const struct timespec *ts);
+	int (*gettime64)(struct ptp_clock_info *ptp, struct timespec64 *ts);
+	int (*settime64)(struct ptp_clock_info *p, const struct timespec64 *ts);
 	int (*enable)(struct ptp_clock_info *ptp,
 		      struct ptp_clock_request *request, int on);
 	int (*verify)(struct ptp_clock_info *ptp, unsigned int pin,

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #2 lore raw

PATCH v2 2/23 ptp: use the 64 bit gettime method for the SYS_OFFSET ioctl.

In reply to Richard Cochran

This patch changes the code to use the new method whenever implemented by the PHC driver.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/ptp/ptp_chardev.c |   14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

drivers/ptp/ptp_chardev.c +10 -4

index f8a7609..95bcf15 100644
@@ -124,7 +124,8 @@ long ptp_ioctl(struct posix_clock *pc, unsigned int cmd, unsigned long arg)
 	struct ptp_clock *ptp = container_of(pc, struct ptp_clock, clock);
 	struct ptp_clock_info *ops = ptp->info;
 	struct ptp_clock_time *pct;
-	struct timespec ts;
+	struct timespec64 ts;
+	struct timespec t2;
 	int enable, err = 0;
 	unsigned int i, pin_index;

@@ -197,16 +198,21 @@ long ptp_ioctl(struct posix_clock *pc, unsigned int cmd, unsigned long arg)
 		}
 		pct = &sysoff->ts[0];
 		for (i = 0; i < sysoff->n_samples; i++) {
-			getnstimeofday(&ts);
+			getnstimeofday64(&ts);
 			pct->sec = ts.tv_sec;
 			pct->nsec = ts.tv_nsec;
 			pct++;
-			ptp->info->gettime(ptp->info, &ts);
+			if (ptp->info->gettime64) {
+				ptp->info->gettime64(ptp->info, &ts);
+			} else {
+				ptp->info->gettime(ptp->info, &t2);
+				ts = timespec_to_timespec64(t2);
+			}
 			pct->sec = ts.tv_sec;
 			pct->nsec = ts.tv_nsec;
 			pct++;
 		}
-		getnstimeofday(&ts);
+		getnstimeofday64(&ts);
 		pct->sec = ts.tv_sec;
 		pct->nsec = ts.tv_nsec;
 		if (copy_to_user((void __user *)arg, sysoff, sizeof(*sysoff)))

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #3 lore raw

PATCH v2 3/23 ptp: use the 64 bit get/set time methods for the posix clock.

In reply to Richard Cochran

This patch changes the posix clock code to prefer the new methods whenever they are implemented by the PHC drivers.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/ptp/ptp_clock.c |   19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

drivers/ptp/ptp_clock.c +17 -2

index 296b0ec..df50d5e 100644
@@ -107,13 +107,28 @@ static int ptp_clock_getres(struct posix_clock *pc, struct timespec *tp)
 static int ptp_clock_settime(struct posix_clock *pc, const struct timespec *tp)
 {
 	struct ptp_clock *ptp = container_of(pc, struct ptp_clock, clock);
-	return ptp->info->settime(ptp->info, tp);
+	struct timespec64 ts = timespec_to_timespec64(*tp);
+
+	return  ptp->info->settime64 ?
+		ptp->info->settime64(ptp->info, &ts) :
+		ptp->info->settime(ptp->info, tp);
 }
 
 static int ptp_clock_gettime(struct posix_clock *pc, struct timespec *tp)
 {
 	struct ptp_clock *ptp = container_of(pc, struct ptp_clock, clock);
-	return ptp->info->gettime(ptp->info, tp);
+	struct timespec64 ts;
+	int err;
+
+	if (ptp->info->gettime64) {
+		err = ptp->info->gettime64(ptp->info, &ts);
+		if (!err)
+			*tp = timespec64_to_timespec(ts);
+	} else {
+		err = ptp->info->gettime(ptp->info, tp);
+	}
+
+	return err;
 }
 
 static int ptp_clock_adjtime(struct posix_clock *pc, struct timex *tx)

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #4 lore raw

PATCH v2 4/23 ptp: blackfin: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device uses 64 bit nanoseconds register, and so with this patch the driver is ready for the year 2038.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/adi/bfin_mac.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

drivers/net/ethernet/adi/bfin_mac.c +4 -4

index ec20611..0c37aef 100644
@@ -983,7 +983,7 @@ static int bfin_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int bfin_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int bfin_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	u64 ns;
 	u32 remainder;

@@ -1003,7 +1003,7 @@ static int bfin_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int bfin_ptp_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
 {
 	u64 ns;
 	unsigned long flags;

@@ -1039,8 +1039,8 @@ static struct ptp_clock_info bfin_ptp_caps = {
 	.pps		= 0,
 	.adjfreq	= bfin_ptp_adjfreq,
 	.adjtime	= bfin_ptp_adjtime,
-	.gettime	= bfin_ptp_gettime,
-	.settime	= bfin_ptp_settime,
+	.gettime64	= bfin_ptp_gettime,
+	.settime64	= bfin_ptp_settime,
 	.enable		= bfin_ptp_enable,
 };

Signature · 1 line

1.7.10.4

Arnd Bergmann

2015-03-22 02:29 UTC #5 lore raw

Re: [PATCH net-next V2 04/23] ptp: blackfin: convert to the 64 bit get/set time methods.

drivers/net/ethernet/adi/bfin_mac.c static int bfin_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta) · line 989

Quoted 5 lines · Richard Cochran

drivers/net/ethernet/adi/bfin_mac.c +1 -1

-static int bfin_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int bfin_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
        u64 ns;
        u32 remainder;

I think it would be good to replace the open-coded

        ns = ts->tv_sec * 1000000000ULL;
        ns += ts->tv_nsec;

here with a call to ns_to_timespec64(), here and in other drivers that you are already touching.

Arnd

Richard Cochran

author maintainer

2015-03-22 07:27 UTC #6 lore raw

Re: [PATCH net-next V2 04/23] ptp: blackfin: convert to the 64 bit get/set time methods.

Quoted 7 lines · Arnd Bergmann

I think it would be good to replace the open-coded
        ns = ts->tv_sec * 1000000000ULL;
        ns += ts->tv_nsec;
here with a call to ns_to_timespec64(), here and in other drivers that you are already touching.

Yes, I'll do that as a follow on series.

Thanks,
Richard

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #7 lore raw

PATCH v2 5/23 ptp: xgbe: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c |   13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

drivers/net/ethernet/amd/xgbe/xgbe-ptp.c +7 -6

index f326178..b03e4f5 100644
@@ -179,7 +179,7 @@ static int xgbe_adjtime(struct ptp_clock_info *info, s64 delta)
 	return 0;
 }
 
-static int xgbe_gettime(struct ptp_clock_info *info, struct timespec *ts)
+static int xgbe_gettime(struct ptp_clock_info *info, struct timespec64 *ts)
 {
 	struct xgbe_prv_data *pdata = container_of(info,
 						   struct xgbe_prv_data,

@@ -193,12 +193,13 @@ static int xgbe_gettime(struct ptp_clock_info *info, struct timespec *ts)
 
 	spin_unlock_irqrestore(&pdata->tstamp_lock, flags);
 
-	*ts = ns_to_timespec(nsec);
+	*ts = ns_to_timespec64(nsec);
 
 	return 0;
 }
 
-static int xgbe_settime(struct ptp_clock_info *info, const struct timespec *ts)
+static int xgbe_settime(struct ptp_clock_info *info,
+			const struct timespec64 *ts)
 {
 	struct xgbe_prv_data *pdata = container_of(info,
 						   struct xgbe_prv_data,

@@ -206,7 +207,7 @@ static int xgbe_settime(struct ptp_clock_info *info, const struct timespec *ts)
 	unsigned long flags;
 	u64 nsec;
 
-	nsec = timespec_to_ns(ts);
+	nsec = timespec64_to_ns(ts);
 
 	spin_lock_irqsave(&pdata->tstamp_lock, flags);

@@ -236,8 +237,8 @@ void xgbe_ptp_register(struct xgbe_prv_data *pdata)
 	info->max_adj = pdata->ptpclk_rate;
 	info->adjfreq = xgbe_adjfreq;
 	info->adjtime = xgbe_adjtime;
-	info->gettime = xgbe_gettime;
-	info->settime = xgbe_settime;
+	info->gettime64 = xgbe_gettime;
+	info->settime64 = xgbe_settime;
 	info->enable = xgbe_enable;
 
 	clock = ptp_clock_register(info, pdata->dev);

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #8 lore raw

PATCH v2 6/23 ptp: bnx2x: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c +4 -4

index 9677431..765667c 100644
@@ -13278,7 +13278,7 @@ static int bnx2x_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int bnx2x_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int bnx2x_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct bnx2x *bp = container_of(ptp, struct bnx2x, ptp_clock_info);
 	u64 ns;

@@ -13295,7 +13295,7 @@ static int bnx2x_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int bnx2x_ptp_settime(struct ptp_clock_info *ptp,
-			     const struct timespec *ts)
+			     const struct timespec64 *ts)
 {
 	struct bnx2x *bp = container_of(ptp, struct bnx2x, ptp_clock_info);
 	u64 ns;

@@ -13333,8 +13333,8 @@ static void bnx2x_register_phc(struct bnx2x *bp)
 	bp->ptp_clock_info.pps = 0;
 	bp->ptp_clock_info.adjfreq = bnx2x_ptp_adjfreq;
 	bp->ptp_clock_info.adjtime = bnx2x_ptp_adjtime;
-	bp->ptp_clock_info.gettime = bnx2x_ptp_gettime;
-	bp->ptp_clock_info.settime = bnx2x_ptp_settime;
+	bp->ptp_clock_info.gettime64 = bnx2x_ptp_gettime;
+	bp->ptp_clock_info.settime64 = bnx2x_ptp_settime;
 	bp->ptp_clock_info.enable = bnx2x_ptp_enable;
 
 	bp->ptp_clock = ptp_clock_register(&bp->ptp_clock_info, &bp->pdev->dev);

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #9 lore raw

PATCH v2 7/23 ptp: tg3: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device appears to use a 64 bit nanoseconds register, and so with this patch the driver should be ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/broadcom/tg3.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

drivers/net/ethernet/broadcom/tg3.c +5 -5

index 22b33da..d02d19b 100644
@@ -6217,7 +6217,7 @@ static int tg3_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int tg3_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int tg3_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	u64 ns;
 	u32 remainder;

@@ -6235,12 +6235,12 @@ static int tg3_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int tg3_ptp_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
 {
 	u64 ns;
 	struct tg3 *tp = container_of(ptp, struct tg3, ptp_info);
 
-	ns = timespec_to_ns(ts);
+	ns = timespec64_to_ns(ts);
 
 	tg3_full_lock(tp, 0);
 	tg3_refclk_write(tp, ns);

@@ -6320,8 +6320,8 @@ static const struct ptp_clock_info tg3_ptp_caps = {
 	.pps		= 0,
 	.adjfreq	= tg3_ptp_adjfreq,
 	.adjtime	= tg3_ptp_adjtime,
-	.gettime	= tg3_ptp_gettime,
-	.settime	= tg3_ptp_settime,
+	.gettime64	= tg3_ptp_gettime,
+	.settime64	= tg3_ptp_settime,
 	.enable		= tg3_ptp_enable,
 };

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #10 lore raw

PATCH v2 8/23 ptp: fec: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/freescale/fec_ptp.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

drivers/net/ethernet/freescale/fec_ptp.c +4 -4

index 4585895..b833993 100644
@@ -390,7 +390,7 @@ static int fec_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
  * read the timecounter and return the correct value on ns,
  * after converting it into a struct timespec.
  */
-static int fec_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int fec_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct fec_enet_private *adapter =
 	    container_of(ptp, struct fec_enet_private, ptp_caps);

@@ -417,7 +417,7 @@ static int fec_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
  * wall timer value.
  */
 static int fec_ptp_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
 {
 	struct fec_enet_private *fep =
 	    container_of(ptp, struct fec_enet_private, ptp_caps);

@@ -584,8 +584,8 @@ void fec_ptp_init(struct platform_device *pdev)
 	fep->ptp_caps.pps = 1;
 	fep->ptp_caps.adjfreq = fec_ptp_adjfreq;
 	fep->ptp_caps.adjtime = fec_ptp_adjtime;
-	fep->ptp_caps.gettime = fec_ptp_gettime;
-	fep->ptp_caps.settime = fec_ptp_settime;
+	fep->ptp_caps.gettime64 = fec_ptp_gettime;
+	fep->ptp_caps.settime64 = fec_ptp_settime;
 	fep->ptp_caps.enable = fec_ptp_enable;
 
 	fep->cycle_speed = clk_get_rate(fep->clk_ptp);

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #11 lore raw

PATCH v2 9/23 ptp: gianfar: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device features a 64 bit nanoseconds register, and so with this patch the driver is ready for the year 2038.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/freescale/gianfar_ptp.c |   13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

drivers/net/ethernet/freescale/gianfar_ptp.c +7 -6

index 77353366..04265a7 100644
@@ -322,7 +322,8 @@ static int ptp_gianfar_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int ptp_gianfar_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int ptp_gianfar_gettime(struct ptp_clock_info *ptp,
+			       struct timespec64 *ts)
 {
 	u64 ns;
 	u32 remainder;

@@ -341,7 +342,7 @@ static int ptp_gianfar_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int ptp_gianfar_settime(struct ptp_clock_info *ptp,
-			       const struct timespec *ts)
+			       const struct timespec64 *ts)
 {
 	u64 ns;
 	unsigned long flags;

@@ -418,8 +419,8 @@ static struct ptp_clock_info ptp_gianfar_caps = {
 	.pps		= 1,
 	.adjfreq	= ptp_gianfar_adjfreq,
 	.adjtime	= ptp_gianfar_adjtime,
-	.gettime	= ptp_gianfar_gettime,
-	.settime	= ptp_gianfar_settime,
+	.gettime64	= ptp_gianfar_gettime,
+	.settime64	= ptp_gianfar_settime,
 	.enable		= ptp_gianfar_enable,
 };

@@ -440,7 +441,7 @@ static int gianfar_ptp_probe(struct platform_device *dev)
 {
 	struct device_node *node = dev->dev.of_node;
 	struct etsects *etsects;
-	struct timespec now;
+	struct timespec64 now;
 	int err = -ENOMEM;
 	u32 tmr_ctrl;
 	unsigned long flags;

@@ -495,7 +496,7 @@ static int gianfar_ptp_probe(struct platform_device *dev)
 		pr_err("ioremap ptp registers failed\n");
 		goto no_ioremap;
 	}
-	getnstimeofday(&now);
+	getnstimeofday64(&now);
 	ptp_gianfar_settime(&etsects->caps, &now);
 
 	tmr_ctrl =

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #12 lore raw

PATCH v2 10/23 ptp: e1000e: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/intel/e1000e/ptp.c |   16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

drivers/net/ethernet/intel/e1000e/ptp.c +8 -8

index 1490f1e..b747a3e 100644
@@ -106,7 +106,7 @@ static int e1000e_phc_adjtime(struct ptp_clock_info *ptp, s64 delta)
  * Read the timecounter and return the correct value in ns after converting
  * it into a struct timespec.
  **/
-static int e1000e_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int e1000e_phc_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct e1000_adapter *adapter = container_of(ptp, struct e1000_adapter,
 						     ptp_clock_info);

@@ -133,14 +133,14 @@ static int e1000e_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
  * wall timer value.
  **/
 static int e1000e_phc_settime(struct ptp_clock_info *ptp,
-			      const struct timespec *ts)
+			      const struct timespec64 *ts)
 {
 	struct e1000_adapter *adapter = container_of(ptp, struct e1000_adapter,
 						     ptp_clock_info);
 	unsigned long flags;
 	u64 ns;
 
-	ns = timespec_to_ns(ts);
+	ns = timespec64_to_ns(ts);
 
 	/* reset the timecounter */
 	spin_lock_irqsave(&adapter->systim_lock, flags);

@@ -171,11 +171,11 @@ static void e1000e_systim_overflow_work(struct work_struct *work)
 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
 						     systim_overflow_work.work);
 	struct e1000_hw *hw = &adapter->hw;
-	struct timespec ts;
+	struct timespec64 ts;
 
-	adapter->ptp_clock_info.gettime(&adapter->ptp_clock_info, &ts);
+	adapter->ptp_clock_info.gettime64(&adapter->ptp_clock_info, &ts);
 
-	e_dbg("SYSTIM overflow check at %ld.%09lu\n", ts.tv_sec, ts.tv_nsec);
+	e_dbg("SYSTIM overflow check at %lld.%09lu\n", ts.tv_sec, ts.tv_nsec);
 
 	schedule_delayed_work(&adapter->systim_overflow_work,
 			      E1000_SYSTIM_OVERFLOW_PERIOD);

@@ -190,8 +190,8 @@ static const struct ptp_clock_info e1000e_ptp_clock_info = {
 	.pps		= 0,
 	.adjfreq	= e1000e_phc_adjfreq,
 	.adjtime	= e1000e_phc_adjtime,
-	.gettime	= e1000e_phc_gettime,
-	.settime	= e1000e_phc_settime,
+	.gettime64	= e1000e_phc_gettime,
+	.settime64	= e1000e_phc_settime,
 	.enable		= e1000e_phc_enable,
 };

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #13 lore raw

PATCH v2 11/23 ptp: fm10k: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device appears to use a 64 bit nanoseconds register, and so with this patch the driver should be ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/intel/fm10k/fm10k_ptp.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

drivers/net/ethernet/intel/fm10k/fm10k_ptp.c +6 -6

index d966044..02008e9 100644
@@ -285,7 +285,7 @@ static int fm10k_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int fm10k_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int fm10k_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct fm10k_intfc *interface;
 	unsigned long flags;

@@ -297,17 +297,17 @@ static int fm10k_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 	now = fm10k_systime_read(interface) + interface->ptp_adjust;
 	read_unlock_irqrestore(&interface->systime_lock, flags);
 
-	*ts = ns_to_timespec(now);
+	*ts = ns_to_timespec64(now);
 
 	return 0;
 }
 
 static int fm10k_ptp_settime(struct ptp_clock_info *ptp,
-			     const struct timespec *ts)
+			     const struct timespec64 *ts)
 {
 	struct fm10k_intfc *interface;
 	unsigned long flags;
-	u64 ns = timespec_to_ns(ts);
+	u64 ns = timespec64_to_ns(ts);
 
 	interface = container_of(ptp, struct fm10k_intfc, ptp_caps);

@@ -419,8 +419,8 @@ void fm10k_ptp_register(struct fm10k_intfc *interface)
 	ptp_caps->max_adj	= 976562;
 	ptp_caps->adjfreq	= fm10k_ptp_adjfreq;
 	ptp_caps->adjtime	= fm10k_ptp_adjtime;
-	ptp_caps->gettime	= fm10k_ptp_gettime;
-	ptp_caps->settime	= fm10k_ptp_settime;
+	ptp_caps->gettime64	= fm10k_ptp_gettime;
+	ptp_caps->settime64	= fm10k_ptp_settime;
 
 	/* provide pins if BAR4 is accessible */
 	if (interface->sw_addr) {

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #14 lore raw

PATCH v2 12/23 ptp: i40e: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device appears to use a 64 bit nanoseconds register, and so with this patch the driver should be ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/intel/i40e/i40e_ptp.c |   26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

drivers/net/ethernet/intel/i40e/i40e_ptp.c +13 ~~-13~~

index fabcfa1..a92b772 100644
@@ -57,7 +57,7 @@
  * timespec. However, since the registers are 64 bits of nanoseconds, we must
  * convert the result to a timespec before we can return.
  **/
-static void i40e_ptp_read(struct i40e_pf *pf, struct timespec *ts)
+static void i40e_ptp_read(struct i40e_pf *pf, struct timespec64 *ts)
 {
 	struct i40e_hw *hw = &pf->hw;
 	u32 hi, lo;

@@ -69,7 +69,7 @@ static void i40e_ptp_read(struct i40e_pf *pf, struct timespec *ts)
 
 	ns = (((u64)hi) << 32) | lo;
 
-	*ts = ns_to_timespec(ns);
+	*ts = ns_to_timespec64(ns);
 }
 
 /**

@@ -81,10 +81,10 @@ static void i40e_ptp_read(struct i40e_pf *pf, struct timespec *ts)
  * we receive a timespec from the stack, we must convert that timespec into
  * nanoseconds before programming the registers.
  **/
-static void i40e_ptp_write(struct i40e_pf *pf, const struct timespec *ts)
+static void i40e_ptp_write(struct i40e_pf *pf, const struct timespec64 *ts)
 {
 	struct i40e_hw *hw = &pf->hw;
-	u64 ns = timespec_to_ns(ts);
+	u64 ns = timespec64_to_ns(ts);
 
 	/* The timer will not update until the high register is written, so
 	 * write the low register first.

@@ -159,14 +159,14 @@ static int i40e_ptp_adjfreq(struct ptp_clock_info *ptp, s32 ppb)
 static int i40e_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 {
 	struct i40e_pf *pf = container_of(ptp, struct i40e_pf, ptp_caps);
-	struct timespec now, then = ns_to_timespec(delta);
+	struct timespec64 now, then = ns_to_timespec64(delta);
 	unsigned long flags;
 
 	spin_lock_irqsave(&pf->tmreg_lock, flags);
 
 	i40e_ptp_read(pf, &now);
-	now = timespec_add(now, then);
-	i40e_ptp_write(pf, (const struct timespec *)&now);
+	now = timespec64_add(now, then);
+	i40e_ptp_write(pf, (const struct timespec64 *)&now);
 
 	spin_unlock_irqrestore(&pf->tmreg_lock, flags);

@@ -181,7 +181,7 @@ static int i40e_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
  * Read the device clock and return the correct value on ns, after converting it
  * into a timespec struct.
  **/
-static int i40e_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int i40e_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct i40e_pf *pf = container_of(ptp, struct i40e_pf, ptp_caps);
 	unsigned long flags;

@@ -202,7 +202,7 @@ static int i40e_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
  * to ns happens in the write function.
  **/
 static int i40e_ptp_settime(struct ptp_clock_info *ptp,
-			    const struct timespec *ts)
+			    const struct timespec64 *ts)
 {
 	struct i40e_pf *pf = container_of(ptp, struct i40e_pf, ptp_caps);
 	unsigned long flags;

@@ -613,8 +613,8 @@ static long i40e_ptp_create_clock(struct i40e_pf *pf)
 	pf->ptp_caps.pps = 0;
 	pf->ptp_caps.adjfreq = i40e_ptp_adjfreq;
 	pf->ptp_caps.adjtime = i40e_ptp_adjtime;
-	pf->ptp_caps.gettime = i40e_ptp_gettime;
-	pf->ptp_caps.settime = i40e_ptp_settime;
+	pf->ptp_caps.gettime64 = i40e_ptp_gettime;
+	pf->ptp_caps.settime64 = i40e_ptp_settime;
 	pf->ptp_caps.enable = i40e_ptp_feature_enable;
 
 	/* Attempt to register the clock before enabling the hardware. */

@@ -673,7 +673,7 @@ void i40e_ptp_init(struct i40e_pf *pf)
 		dev_err(&pf->pdev->dev, "%s: ptp_clock_register failed\n",
 			__func__);
 	} else {
-		struct timespec ts;
+		struct timespec64 ts;
 		u32 regval;
 
 		dev_info(&pf->pdev->dev, "%s: added PHC on %s\n", __func__,

@@ -695,7 +695,7 @@ void i40e_ptp_init(struct i40e_pf *pf)
 		i40e_ptp_set_timestamp_mode(pf, &pf->tstamp_config);
 
 		/* Set the clock value. */
-		ts = ktime_to_timespec(ktime_get_real());
+		ts = ktime_to_timespec64(ktime_get_real());
 		i40e_ptp_settime(&pf->ptp_caps, &ts);
 	}
 }

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #15 lore raw

PATCH v2 13/23 ptp: igb: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

For the 82576, the driver's clock is implemented using a timecounter, and so with this patch that device is ready for the year 2038.

However, in the case of the i210, the device stores the number of seconds in a 32 bit register. Therefore, more work is needed on this driver before the year 2038 comes around.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/intel/igb/igb_ptp.c |   41 +++++++++++++++---------------
 1 file changed, 21 insertions(+), 20 deletions(-)

drivers/net/ethernet/intel/igb/igb_ptp.c +21 ~~-20~~

index 52d01b8..0348b7e 100644
@@ -116,7 +116,8 @@ static cycle_t igb_ptp_read_82580(const struct cyclecounter *cc)
 }
 
 /* SYSTIM read access for I210/I211 */
-static void igb_ptp_read_i210(struct igb_adapter *adapter, struct timespec *ts)
+static void igb_ptp_read_i210(struct igb_adapter *adapter,
+			      struct timespec64 *ts)
 {
 	struct e1000_hw *hw = &adapter->hw;
 	u32 sec, nsec;

@@ -134,7 +135,7 @@ static void igb_ptp_read_i210(struct igb_adapter *adapter, struct timespec *ts)
 }
 
 static void igb_ptp_write_i210(struct igb_adapter *adapter,
-			       const struct timespec *ts)
+			       const struct timespec64 *ts)
 {
 	struct e1000_hw *hw = &adapter->hw;

@@ -269,13 +270,13 @@ static int igb_ptp_adjtime_i210(struct ptp_clock_info *ptp, s64 delta)
 	struct igb_adapter *igb = container_of(ptp, struct igb_adapter,
 					       ptp_caps);
 	unsigned long flags;
-	struct timespec now, then = ns_to_timespec(delta);
+	struct timespec64 now, then = ns_to_timespec64(delta);
 
 	spin_lock_irqsave(&igb->tmreg_lock, flags);
 
 	igb_ptp_read_i210(igb, &now);
-	now = timespec_add(now, then);
-	igb_ptp_write_i210(igb, (const struct timespec *)&now);
+	now = timespec64_add(now, then);
+	igb_ptp_write_i210(igb, (const struct timespec64 *)&now);
 
 	spin_unlock_irqrestore(&igb->tmreg_lock, flags);

@@ -283,7 +284,7 @@ static int igb_ptp_adjtime_i210(struct ptp_clock_info *ptp, s64 delta)
 }
 
 static int igb_ptp_gettime_82576(struct ptp_clock_info *ptp,
-				 struct timespec *ts)
+				 struct timespec64 *ts)
 {
 	struct igb_adapter *igb = container_of(ptp, struct igb_adapter,
 					       ptp_caps);

@@ -304,7 +305,7 @@ static int igb_ptp_gettime_82576(struct ptp_clock_info *ptp,
 }
 
 static int igb_ptp_gettime_i210(struct ptp_clock_info *ptp,
-				struct timespec *ts)
+				struct timespec64 *ts)
 {
 	struct igb_adapter *igb = container_of(ptp, struct igb_adapter,
 					       ptp_caps);

@@ -320,7 +321,7 @@ static int igb_ptp_gettime_i210(struct ptp_clock_info *ptp,
 }
 
 static int igb_ptp_settime_82576(struct ptp_clock_info *ptp,
-				 const struct timespec *ts)
+				 const struct timespec64 *ts)
 {
 	struct igb_adapter *igb = container_of(ptp, struct igb_adapter,
 					       ptp_caps);

@@ -340,7 +341,7 @@ static int igb_ptp_settime_82576(struct ptp_clock_info *ptp,
 }
 
 static int igb_ptp_settime_i210(struct ptp_clock_info *ptp,
-				const struct timespec *ts)
+				const struct timespec64 *ts)
 {
 	struct igb_adapter *igb = container_of(ptp, struct igb_adapter,
 					       ptp_caps);

@@ -627,11 +628,11 @@ static void igb_ptp_overflow_check(struct work_struct *work)
 {
 	struct igb_adapter *igb =
 		container_of(work, struct igb_adapter, ptp_overflow_work.work);
-	struct timespec ts;
+	struct timespec64 ts;
 
-	igb->ptp_caps.gettime(&igb->ptp_caps, &ts);
+	igb->ptp_caps.gettime64(&igb->ptp_caps, &ts);
 
-	pr_debug("igb overflow check at %ld.%09lu\n", ts.tv_sec, ts.tv_nsec);
+	pr_debug("igb overflow check at %lld.%09lu\n", ts.tv_sec, ts.tv_nsec);
 
 	schedule_delayed_work(&igb->ptp_overflow_work,
 			      IGB_SYSTIM_OVERFLOW_PERIOD);

@@ -989,8 +990,8 @@ void igb_ptp_init(struct igb_adapter *adapter)
 		adapter->ptp_caps.pps = 0;
 		adapter->ptp_caps.adjfreq = igb_ptp_adjfreq_82576;
 		adapter->ptp_caps.adjtime = igb_ptp_adjtime_82576;
-		adapter->ptp_caps.gettime = igb_ptp_gettime_82576;
-		adapter->ptp_caps.settime = igb_ptp_settime_82576;
+		adapter->ptp_caps.gettime64 = igb_ptp_gettime_82576;
+		adapter->ptp_caps.settime64 = igb_ptp_settime_82576;
 		adapter->ptp_caps.enable = igb_ptp_feature_enable;
 		adapter->cc.read = igb_ptp_read_82576;
 		adapter->cc.mask = CYCLECOUNTER_MASK(64);

@@ -1009,8 +1010,8 @@ void igb_ptp_init(struct igb_adapter *adapter)
 		adapter->ptp_caps.pps = 0;
 		adapter->ptp_caps.adjfreq = igb_ptp_adjfreq_82580;
 		adapter->ptp_caps.adjtime = igb_ptp_adjtime_82576;
-		adapter->ptp_caps.gettime = igb_ptp_gettime_82576;
-		adapter->ptp_caps.settime = igb_ptp_settime_82576;
+		adapter->ptp_caps.gettime64 = igb_ptp_gettime_82576;
+		adapter->ptp_caps.settime64 = igb_ptp_settime_82576;
 		adapter->ptp_caps.enable = igb_ptp_feature_enable;
 		adapter->cc.read = igb_ptp_read_82580;
 		adapter->cc.mask = CYCLECOUNTER_MASK(IGB_NBITS_82580);

@@ -1038,8 +1039,8 @@ void igb_ptp_init(struct igb_adapter *adapter)
 		adapter->ptp_caps.pin_config = adapter->sdp_config;
 		adapter->ptp_caps.adjfreq = igb_ptp_adjfreq_82580;
 		adapter->ptp_caps.adjtime = igb_ptp_adjtime_i210;
-		adapter->ptp_caps.gettime = igb_ptp_gettime_i210;
-		adapter->ptp_caps.settime = igb_ptp_settime_i210;
+		adapter->ptp_caps.gettime64 = igb_ptp_gettime_i210;
+		adapter->ptp_caps.settime64 = igb_ptp_settime_i210;
 		adapter->ptp_caps.enable = igb_ptp_feature_enable_i210;
 		adapter->ptp_caps.verify = igb_ptp_verify_pin;
 		/* Enable the timer functions by clearing bit 31. */

@@ -1057,7 +1058,7 @@ void igb_ptp_init(struct igb_adapter *adapter)
 
 	/* Initialize the clock and overflow work for devices that need it. */
 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211)) {
-		struct timespec ts = ktime_to_timespec(ktime_get_real());
+		struct timespec64 ts = ktime_to_timespec64(ktime_get_real());
 
 		igb_ptp_settime_i210(&adapter->ptp_caps, &ts);
 	} else {

@@ -1171,7 +1172,7 @@ void igb_ptp_reset(struct igb_adapter *adapter)
 
 	/* Re-initialize the timer. */
 	if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211)) {
-		struct timespec ts = ktime_to_timespec(ktime_get_real());
+		struct timespec64 ts = ktime_to_timespec64(ktime_get_real());
 
 		igb_ptp_write_i210(adapter, &ts);
 	} else {

Signature · 1 line

1.7.10.4

Arnd Bergmann

2015-03-22 02:40 UTC #16 lore raw

Re: [PATCH net-next V2 13/23] ptp: igb: convert to the 64 bit get/set time methods.

Quoted 7 lines · Richard Cochran

drivers/net/ethernet/intel/igb/igb_ptp.c +1 -1

@@ -269,13 +270,13 @@ static int igb_ptp_adjtime_i210(struct ptp_clock_info *ptp, s64 delta)
        struct igb_adapter *igb = container_of(ptp, struct igb_adapter,
                                               ptp_caps);
        unsigned long flags;
-       struct timespec now, then = ns_to_timespec(delta);
+       struct timespec64 now, then = ns_to_timespec64(delta);

drivers/net/ethernet/intel/igb/igb_ptp.c static int igb_ptp_adjtime_i210(struct ptp_clock_info *ptp, s64 delta) · line 282

Quoted 10 lines

drivers/net/ethernet/intel/igb/igb_ptp.c +2 -2

        spin_lock_irqsave(&igb->tmreg_lock, flags);
 
        igb_ptp_read_i210(igb, &now);
-       now = timespec_add(now, then);
-       igb_ptp_write_i210(igb, (const struct timespec *)&now);
+       now = timespec64_add(now, then);
+       igb_ptp_write_i210(igb, (const struct timespec64 *)&now);
 
        spin_unlock_irqrestore(&igb->tmreg_lock, flags);

This is one of the functions that would actually become more efficient if the argument type gets changed to ktime_t.

Arnd

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #17 lore raw

PATCH v2 14/23 ptp: ixgbe: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c |   14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c +7 -7

index bd46f5d..75b9e93 100644
@@ -279,7 +279,7 @@ static int ixgbe_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
  * read the timecounter and return the correct value on ns,
  * after converting it into a struct timespec.
  */
-static int ixgbe_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int ixgbe_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct ixgbe_adapter *adapter =
 		container_of(ptp, struct ixgbe_adapter, ptp_caps);

@@ -306,7 +306,7 @@ static int ixgbe_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
  * wall timer value.
  */
 static int ixgbe_ptp_settime(struct ptp_clock_info *ptp,
-			     const struct timespec *ts)
+			     const struct timespec64 *ts)
 {
 	struct ixgbe_adapter *adapter =
 		container_of(ptp, struct ixgbe_adapter, ptp_caps);

@@ -407,7 +407,7 @@ void ixgbe_ptp_overflow_check(struct ixgbe_adapter *adapter)
 {
 	bool timeout = time_is_before_jiffies(adapter->last_overflow_check +
 					     IXGBE_OVERFLOW_PERIOD);
-	struct timespec ts;
+	struct timespec64 ts;
 
 	if (timeout) {
 		ixgbe_ptp_gettime(&adapter->ptp_caps, &ts);

@@ -874,8 +874,8 @@ static int ixgbe_ptp_create_clock(struct ixgbe_adapter *adapter)
 		adapter->ptp_caps.pps = 1;
 		adapter->ptp_caps.adjfreq = ixgbe_ptp_adjfreq;
 		adapter->ptp_caps.adjtime = ixgbe_ptp_adjtime;
-		adapter->ptp_caps.gettime = ixgbe_ptp_gettime;
-		adapter->ptp_caps.settime = ixgbe_ptp_settime;
+		adapter->ptp_caps.gettime64 = ixgbe_ptp_gettime;
+		adapter->ptp_caps.settime64 = ixgbe_ptp_settime;
 		adapter->ptp_caps.enable = ixgbe_ptp_feature_enable;
 		break;
 	case ixgbe_mac_82599EB:

@@ -890,8 +890,8 @@ static int ixgbe_ptp_create_clock(struct ixgbe_adapter *adapter)
 		adapter->ptp_caps.pps = 0;
 		adapter->ptp_caps.adjfreq = ixgbe_ptp_adjfreq;
 		adapter->ptp_caps.adjtime = ixgbe_ptp_adjtime;
-		adapter->ptp_caps.gettime = ixgbe_ptp_gettime;
-		adapter->ptp_caps.settime = ixgbe_ptp_settime;
+		adapter->ptp_caps.gettime64 = ixgbe_ptp_gettime;
+		adapter->ptp_caps.settime64 = ixgbe_ptp_settime;
 		adapter->ptp_caps.enable = ixgbe_ptp_feature_enable;
 		break;
 	default:

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #18 lore raw

PATCH v2 15/23 ptp: mlx4: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/mellanox/mlx4/en_clock.c |   11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

drivers/net/ethernet/mellanox/mlx4/en_clock.c +6 -5

index 90b5309..666e7bd 100644
@@ -164,7 +164,8 @@ static int mlx4_en_phc_adjtime(struct ptp_clock_info *ptp, s64 delta)
  * Read the timecounter and return the correct value in ns after converting
  * it into a struct timespec.
  **/
-static int mlx4_en_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int mlx4_en_phc_gettime(struct ptp_clock_info *ptp,
+			       struct timespec64 *ts)
 {
 	struct mlx4_en_dev *mdev = container_of(ptp, struct mlx4_en_dev,
 						ptp_clock_info);

@@ -191,11 +192,11 @@ static int mlx4_en_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
  * wall timer value.
  **/
 static int mlx4_en_phc_settime(struct ptp_clock_info *ptp,
-			       const struct timespec *ts)
+			       const struct timespec64 *ts)
 {
 	struct mlx4_en_dev *mdev = container_of(ptp, struct mlx4_en_dev,
 						ptp_clock_info);
-	u64 ns = timespec_to_ns(ts);
+	u64 ns = timespec64_to_ns(ts);
 	unsigned long flags;
 
 	/* reset the timecounter */

@@ -232,8 +233,8 @@ static const struct ptp_clock_info mlx4_en_ptp_clock_info = {
 	.pps		= 0,
 	.adjfreq	= mlx4_en_phc_adjfreq,
 	.adjtime	= mlx4_en_phc_adjtime,
-	.gettime	= mlx4_en_phc_gettime,
-	.settime	= mlx4_en_phc_settime,
+	.gettime64	= mlx4_en_phc_gettime,
+	.settime64	= mlx4_en_phc_settime,
 	.enable		= mlx4_en_phc_enable,
 };

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #19 lore raw

PATCH v2 16/23 ptp: sfc: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This patch changes the driver to use the newer API.

Depending on how the hardware represents a time value, this driver may or may not yet be ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/sfc/ptp.c |   22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

drivers/net/ethernet/sfc/ptp.c +11 ~~-11~~

index 6b861e3..a2e9aee 100644
@@ -323,9 +323,9 @@ struct efx_ptp_data {
 
 static int efx_phc_adjfreq(struct ptp_clock_info *ptp, s32 delta);
 static int efx_phc_adjtime(struct ptp_clock_info *ptp, s64 delta);
-static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts);
+static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts);
 static int efx_phc_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *e_ts);
+			   const struct timespec64 *e_ts);
 static int efx_phc_enable(struct ptp_clock_info *ptp,
 			  struct ptp_clock_request *request, int on);

@@ -1198,8 +1198,8 @@ static const struct ptp_clock_info efx_phc_clock_info = {
 	.pps		= 1,
 	.adjfreq	= efx_phc_adjfreq,
 	.adjtime	= efx_phc_adjtime,
-	.gettime	= efx_phc_gettime,
-	.settime	= efx_phc_settime,
+	.gettime64	= efx_phc_gettime,
+	.settime64	= efx_phc_settime,
 	.enable		= efx_phc_enable,
 };

@@ -1837,7 +1837,7 @@ static int efx_phc_adjtime(struct ptp_clock_info *ptp, s64 delta)
 			    NULL, 0, NULL);
 }
 
-static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct efx_ptp_data *ptp_data = container_of(ptp,
 						     struct efx_ptp_data,

@@ -1859,28 +1859,28 @@ static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 	kt = ptp_data->nic_to_kernel_time(
 		MCDI_DWORD(outbuf, PTP_OUT_READ_NIC_TIME_MAJOR),
 		MCDI_DWORD(outbuf, PTP_OUT_READ_NIC_TIME_MINOR), 0);
-	*ts = ktime_to_timespec(kt);
+	*ts = ktime_to_timespec64(kt);
 	return 0;
 }
 
 static int efx_phc_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *e_ts)
+			   const struct timespec64 *e_ts)
 {
 	/* Get the current NIC time, efx_phc_gettime.
 	 * Subtract from the desired time to get the offset
 	 * call efx_phc_adjtime with the offset
 	 */
 	int rc;
-	struct timespec time_now;
-	struct timespec delta;
+	struct timespec64 time_now;
+	struct timespec64 delta;
 
 	rc = efx_phc_gettime(ptp, &time_now);
 	if (rc != 0)
 		return rc;
 
-	delta = timespec_sub(*e_ts, time_now);
+	delta = timespec64_sub(*e_ts, time_now);
 
-	rc = efx_phc_adjtime(ptp, timespec_to_ns(&delta));
+	rc = efx_phc_adjtime(ptp, timespec64_to_ns(&delta));
 	if (rc != 0)
 		return rc;

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #20 lore raw

PATCH v2 17/23 ptp: stmmac: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This device stores the number of seconds in a 32 bit register. So more work is needed on this driver before the year 2038 comes around.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c +4 -4

index c5ee79d8..172f318 100644
@@ -105,7 +105,7 @@ static int stmmac_adjust_time(struct ptp_clock_info *ptp, s64 delta)
  * Description: this function will read the current time from the
  * hardware clock and store it in @ts.
  */
-static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec *ts)
+static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	struct stmmac_priv *priv =
 	    container_of(ptp, struct stmmac_priv, ptp_clock_ops);

@@ -135,7 +135,7 @@ static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec *ts)
  * hardware clock.
  */
 static int stmmac_set_time(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
 {
 	struct stmmac_priv *priv =
 	    container_of(ptp, struct stmmac_priv, ptp_clock_ops);

@@ -168,8 +168,8 @@ static struct ptp_clock_info stmmac_ptp_clock_ops = {
 	.pps = 0,
 	.adjfreq = stmmac_adjust_freq,
 	.adjtime = stmmac_adjust_time,
-	.gettime = stmmac_get_time,
-	.settime = stmmac_set_time,
+	.gettime64 = stmmac_get_time,
+	.settime64 = stmmac_set_time,
 	.enable = stmmac_enable,
 };

Signature · 1 line

1.7.10.4

Giuseppe CAVALLARO

2015-03-23 15:35 UTC #21 lore raw

Re: [PATCH net-next V2 17/23] ptp: stmmac: convert to the 64 bit get/set time methods.

Quoted 4 lines · Richard Cochran

This device stores the number of seconds in a 32 bit register. So more work is needed on this driver before the year 2038 comes around.
Compile tested only.

I cannot perform tests on my side, unfortunately. I have no setup at this moment. Added Rayagond on copy too

Peppe

Quoted 40 lines

Signed-off-by Richard Cochran <[email protected]>

---
  drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c |    8 ++++----
  1 file changed, 4 insertions(+), 4 deletions(-)

drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c +4 -4

index c5ee79d8..172f318 100644
@@ -105,7 +105,7 @@ static int stmmac_adjust_time(struct ptp_clock_info *ptp, s64 delta)
   * Description: this function will read the current time from the
   * hardware clock and store it in @ts.
   */
-static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec *ts)
+static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec64 *ts)
  {
  	struct stmmac_priv *priv =
  	    container_of(ptp, struct stmmac_priv, ptp_clock_ops);

@@ -135,7 +135,7 @@ static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec *ts)
   * hardware clock.
   */
  static int stmmac_set_time(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
  {
  	struct stmmac_priv *priv =
  	    container_of(ptp, struct stmmac_priv, ptp_clock_ops);

@@ -168,8 +168,8 @@ static struct ptp_clock_info stmmac_ptp_clock_ops = {
  	.pps = 0,
  	.adjfreq = stmmac_adjust_freq,
  	.adjtime = stmmac_adjust_time,
-	.gettime = stmmac_get_time,
-	.settime = stmmac_set_time,
+	.gettime64 = stmmac_get_time,
+	.settime64 = stmmac_set_time,
  	.enable = stmmac_enable,
  };

Rayagond Kokatanur

2015-03-31 13:22 UTC #22 lore raw

Re: [PATCH net-next V2 17/23] ptp: stmmac: convert to the 64 bit get/set time methods.

Sorry, I too don't have setup now.

--
Rayagond

Quoted 60 lines · Giuseppe CAVALLARO

Quoted 5 lines · Richard Cochran

This device stores the number of seconds in a 32 bit register. So more work is needed on this driver before the year 2038 comes around.
Compile tested only.

I cannot perform tests on my side, unfortunately. I have no setup at this moment. Added Rayagond on copy too

Peppe

Quoted 45 lines

Signed-off-by Richard Cochran <[email protected]>

---
  drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c |    8 ++++----
  1 file changed, 4 insertions(+), 4 deletions(-)

drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c +1 -1

index c5ee79d8..172f318 100644
@@ -105,7 +105,7 @@ static int stmmac_adjust_time(struct ptp_clock_info
 *ptp, s64 delta)
   * Description: this function will read the current time from the
   * hardware clock and store it in @ts.
   */
-static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec
 *ts)
+static int stmmac_get_time(struct ptp_clock_info *ptp, struct timespec64
 *ts)

  {
        struct stmmac_priv *priv =
            container_of(ptp, struct stmmac_priv, ptp_clock_ops);

drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c +1 -1

@@ -135,7 +135,7 @@ static int stmmac_get_time(struct ptp_clock_info *ptp,
 struct timespec *ts)
   * hardware clock.
   */
  static int stmmac_set_time(struct ptp_clock_info *ptp,
-                          const struct timespec *ts)
+                          const struct timespec64 *ts)
  {
        struct stmmac_priv *priv =

            container_of(ptp, struct stmmac_priv, ptp_clock_ops);

drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c +2 -2

@@ -168,8 +168,8 @@ static struct ptp_clock_info stmmac_ptp_clock_ops = {
        .pps = 0,
        .adjfreq = stmmac_adjust_freq,
        .adjtime = stmmac_adjust_time,
-       .gettime = stmmac_get_time,
-       .settime = stmmac_set_time,
+       .gettime64 = stmmac_get_time,
+       .settime64 = stmmac_set_time,
        .enable = stmmac_enable,
  };

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #23 lore raw

PATCH v2 18/23 ptp: cpts: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver's clock is implemented using a timecounter, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/ti/cpts.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

drivers/net/ethernet/ti/cpts.c +6 -6

index fbe42cb..074b636 100644
@@ -167,7 +167,7 @@ static int cpts_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int cpts_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int cpts_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	u64 ns;
 	u32 remainder;

@@ -185,7 +185,7 @@ static int cpts_ptp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int cpts_ptp_settime(struct ptp_clock_info *ptp,
-			    const struct timespec *ts)
+			    const struct timespec64 *ts)
 {
 	u64 ns;
 	unsigned long flags;

@@ -216,20 +216,20 @@ static struct ptp_clock_info cpts_info = {
 	.pps		= 0,
 	.adjfreq	= cpts_ptp_adjfreq,
 	.adjtime	= cpts_ptp_adjtime,
-	.gettime	= cpts_ptp_gettime,
-	.settime	= cpts_ptp_settime,
+	.gettime64	= cpts_ptp_gettime,
+	.settime64	= cpts_ptp_settime,
 	.enable		= cpts_ptp_enable,
 };
 
 static void cpts_overflow_check(struct work_struct *work)
 {
-	struct timespec ts;
+	struct timespec64 ts;
 	struct cpts *cpts = container_of(work, struct cpts, overflow_work.work);
 
 	cpts_write32(cpts, CPTS_EN, control);
 	cpts_write32(cpts, TS_PEND_EN, int_enable);
 	cpts_ptp_gettime(&cpts->info, &ts);
-	pr_debug("cpts overflow check at %ld.%09lu\n", ts.tv_sec, ts.tv_nsec);
+	pr_debug("cpts overflow check at %lld.%09lu\n", ts.tv_sec, ts.tv_nsec);
 	schedule_delayed_work(&cpts->overflow_work, CPTS_OVERFLOW_PERIOD);
 }

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #24 lore raw

PATCH v2 19/23 ptp: tilegx: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This driver calls code (via gxio_mpipe_get/set_timestamp) that makes the assumption that the tv_sec field is 64 bits wide. So apparently this driver is 64 bit only. So maybe this driver and device are ready for 2038, but maybe not.

Not even compile tested.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/ethernet/tile/tilegx.c |   16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

drivers/net/ethernet/tile/tilegx.c +10 -6

index bea8cd2..614ea83 100644
@@ -838,24 +838,28 @@ static int ptp_mpipe_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return ret;
 }
 
-static int ptp_mpipe_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int ptp_mpipe_gettime(struct ptp_clock_info *ptp,
+			     struct timespec64 *ts64)
 {
 	int ret = 0;
 	struct mpipe_data *md = container_of(ptp, struct mpipe_data, caps);
+	struct timespec ts;
 	mutex_lock(&md->ptp_lock);
-	if (gxio_mpipe_get_timestamp(&md->context, ts))
+	if (gxio_mpipe_get_timestamp(&md->context, &ts))
 		ret = -EBUSY;
 	mutex_unlock(&md->ptp_lock);
+	*ts64 = timespec_to_timespec64(ts);
 	return ret;
 }
 
 static int ptp_mpipe_settime(struct ptp_clock_info *ptp,
-			     const struct timespec *ts)
+			     const struct timespec64 *ts64)
 {
 	int ret = 0;
+	struct timespec ts = timespec64_to_timespec(*ts64);
 	struct mpipe_data *md = container_of(ptp, struct mpipe_data, caps);
 	mutex_lock(&md->ptp_lock);
-	if (gxio_mpipe_set_timestamp(&md->context, ts))
+	if (gxio_mpipe_set_timestamp(&md->context, &ts))
 		ret = -EBUSY;
 	mutex_unlock(&md->ptp_lock);
 	return ret;

@@ -876,8 +880,8 @@ static struct ptp_clock_info ptp_mpipe_caps = {
 	.pps		= 0,
 	.adjfreq	= ptp_mpipe_adjfreq,
 	.adjtime	= ptp_mpipe_adjtime,
-	.gettime	= ptp_mpipe_gettime,
-	.settime	= ptp_mpipe_settime,
+	.gettime64	= ptp_mpipe_gettime,
+	.settime64	= ptp_mpipe_settime,
 	.enable		= ptp_mpipe_enable,
 };

Signature · 1 line

1.7.10.4

Chris Metcalf

Acked-by on v2 19/23

2015-03-23 16:02 UTC #25 lore raw

Re: [PATCH net-next V2 19/23] ptp: tilegx: convert to the 64 bit get/set time methods.

Quoted 8 lines · Richard Cochran

This driver calls code (via gxio_mpipe_get/set_timestamp) that makes the assumption that the tv_sec field is 64 bits wide. So apparently this driver is 64 bit only. So maybe this driver and device are ready for 2038, but maybe not.
Not even compile tested.
Signed-off-by Richard Cochran<[email protected]>

This driver is 64-bit only. Would it make more sense to just change the accessors from gettime/settime to gettime/settime64 and nothing else, i.e. rely on the current behavior that timespec and timespec64 are the same type, and trust the compiler to fail it if somehow someone tried to build this driver into a 32-bit kernel?

In any case, if you think it's better as-is, you can certainly have my

Acked-by Chris Metcalf <[email protected]>

Signature · 2 lines

Chris Metcalf, EZChip Semiconductor
http://www.ezchip.com

Richard Cochran

author maintainer

2015-03-23 16:58 UTC #26 lore raw

Re: [PATCH net-next V2 19/23] ptp: tilegx: convert to the 64 bit get/set time methods.

Quoted 5 lines · Chris Metcalf

This driver is 64-bit only. Would it make more sense to just change the accessors from gettime/settime to gettime/settime64 and nothing else, i.e. rely on the current behavior that timespec and timespec64 are the same type, and trust the compiler to fail it if somehow someone tried to build this driver into a 32-bit kernel?

On the one hand, I think the best way would be for gxio_mpipe_get/set_timestamp to also take a timespec64, because it makes the width clear. With a plain old timespec, you have to remember whether you are 64 bit or not. However, but I couldn't tell if changing gxio_mpipe_get/set_timestamp would work for other callers or not.

If the driver is 64 bit only, shouldn't that be reflected in the Kconfig?

On the other hand, I wouldn't mind the change you suggest, if it were super obvious that the code is 64 bit only. It was not obvious to me.

Thanks,
Richard

Chris Metcalf

2015-03-23 17:26 UTC #27 lore raw

Re: [PATCH net-next V2 19/23] ptp: tilegx: convert to the 64 bit get/set time methods.

Quoted 6 lines · Richard Cochran

Quoted 5 lines · Chris Metcalf
Quoted 5 lines
This driver is 64-bit only. Would it make more sense to just change the accessors from gettime/settime to gettime/settime64 and nothing else, i.e. rely on the current behavior that timespec and timespec64 are the same type, and trust the compiler to fail it if somehow someone tried to build this driver into a 32-bit kernel?

Quoted 6 lines

On the one hand, I think the best way would be for gxio_mpipe_get/set_timestamp to also take a timespec64, because it makes the width clear. With a plain old timespec, you have to remember whether you are 64 bit or not. However, but I couldn't tell if changing gxio_mpipe_get/set_timestamp would work for other callers or not.

If you're suggesting changing the gxio_mpipe_xxx_timespec routines to take a timespec64, that sounds reasonable to me, particularly in conjunction with your new gettime64 etc API.

As far as I know, the tilegx driver is the only client of those gxio routines.

Quoted 2 lines

If the driver is 64 bit only, shouldn't that be reflected in the Kconfig?

Well, the tilegx driver is only supported for the tilegx architecture, which is only 64-bit for kernel-space. It may be a bit confusing because the actual driver is tile_net, which has separate implementations for tilegx (64-bit) and tilepro (32-bit). It's not immediately clear how to make that more obvious.

Quoted 2 lines

On the other hand, I wouldn't mind the change you suggest, if it were super obvious that the code is 64 bit only. It was not obvious to me.

I guess that brings us back to changing the gxio_xxx APIs.

Signature · 2 lines

Chris Metcalf, EZChip Semiconductor
http://www.ezchip.com

Richard Cochran

author maintainer

2015-03-23 20:06 UTC #28 lore raw

Re: [PATCH net-next V2 19/23] ptp: tilegx: convert to the 64 bit get/set time methods.

Quoted 2 lines · Chris Metcalf

As far as I know, the tilegx driver is the only client of those gxio routines.

...

Quoted 1 line

I guess that brings us back to changing the gxio_xxx APIs.

Ok, I'll do that for the next round.

Thanks,
Richard

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #29 lore raw

PATCH v2 20/23 ptp: dp83640: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

This device stores the number of seconds in a 32 bit register. So more work is needed on this driver before the year 2038 comes around.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/net/phy/dp83640.c |   12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

drivers/net/phy/dp83640.c +7 -5

index e22e602..1a88235 100644
@@ -427,7 +427,8 @@ static int ptp_dp83640_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return err;
 }
 
-static int ptp_dp83640_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int ptp_dp83640_gettime(struct ptp_clock_info *ptp,
+			       struct timespec64 *ts)
 {
 	struct dp83640_clock *clock =
 		container_of(ptp, struct dp83640_clock, caps);

@@ -452,16 +453,17 @@ static int ptp_dp83640_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int ptp_dp83640_settime(struct ptp_clock_info *ptp,
-			       const struct timespec *ts)
+			       const struct timespec64 *ts64)
 {
 	struct dp83640_clock *clock =
 		container_of(ptp, struct dp83640_clock, caps);
 	struct phy_device *phydev = clock->chosen->phydev;
+	struct timespec ts = timespec64_to_timespec(*ts64);
 	int err;
 
 	mutex_lock(&clock->extreg_lock);
 
-	err = tdr_write(1, phydev, ts, PTP_LOAD_CLK);
+	err = tdr_write(1, phydev, &ts, PTP_LOAD_CLK);
 
 	mutex_unlock(&clock->extreg_lock);

@@ -998,8 +1000,8 @@ static void dp83640_clock_init(struct dp83640_clock *clock, struct mii_bus *bus)
 	clock->caps.pps		= 0;
 	clock->caps.adjfreq	= ptp_dp83640_adjfreq;
 	clock->caps.adjtime	= ptp_dp83640_adjtime;
-	clock->caps.gettime	= ptp_dp83640_gettime;
-	clock->caps.settime	= ptp_dp83640_settime;
+	clock->caps.gettime64	= ptp_dp83640_gettime;
+	clock->caps.settime64	= ptp_dp83640_settime;
 	clock->caps.enable	= ptp_dp83640_enable;
 	clock->caps.verify	= ptp_dp83640_verify;
 	/*

Signature · 1 line

1.7.10.4

Arnd Bergmann

2015-03-22 02:36 UTC #30 lore raw

Re: [PATCH net-next V2 20/23] ptp: dp83640: convert to the 64 bit get/set time methods.

drivers/net/phy/dp83640.c static int ptp_dp83640_gettime(struct ptp_clock_info *ptp, struct timespec *ts) · line 468

Quoted 6 lines · Richard Cochran

        mutex_lock(&clock->extreg_lock);

drivers/net/phy/dp83640.c +1 -1

-       err = tdr_write(1, phydev, ts, PTP_LOAD_CLK);
+       err = tdr_write(1, phydev, &ts, PTP_LOAD_CLK);
 
        mutex_unlock(&clock->extreg_lock);

I don't see the change to the tdr_write() function that changes the argument from 'struct timespec ts' to 'struct timespec64 *', so this looks wrong to me.

Arnd

Richard Cochran

author maintainer

2015-03-22 07:29 UTC #31 lore raw

Re: [PATCH net-next V2 20/23] ptp: dp83640: convert to the 64 bit get/set time methods.

Quoted 11 lines · Arnd Bergmann

drivers/net/phy/dp83640.c static int ptp_dp83640_gettime(struct ptp_clock_info *ptp, struct timespec *ts) · line 468
Quoted 6 lines · Richard Cochran
        mutex_lock(&clock->extreg_lock);
drivers/net/phy/dp83640.c +1 -1
-       err = tdr_write(1, phydev, ts, PTP_LOAD_CLK);
+       err = tdr_write(1, phydev, &ts, PTP_LOAD_CLK);
 
        mutex_unlock(&clock->extreg_lock);
I don't see the change to the tdr_write() function that changes the argument from 'struct timespec ts' to 'struct timespec64 *', so this looks wrong to me.

There is a 'ts64' parameter, converted to 'ts' beforehand.

Thanks,
Richard

Arnd Bergmann

2015-03-22 17:48 UTC #32 lore raw

Re: [PATCH net-next V2 20/23] ptp: dp83640: convert to the 64 bit get/set time methods.

Quoted 7 lines · Richard Cochran

Quoted 6 lines · Arnd Bergmann
Quoted 5 lines · Richard Cochran
        mutex_lock(&clock->extreg_lock);
drivers/net/phy/dp83640.c +1 -1
-       err = tdr_write(1, phydev, ts, PTP_LOAD_CLK);
+       err = tdr_write(1, phydev, &ts, PTP_LOAD_CLK);

Quoted 8 lines

Quoted 5 lines
Quoted 1 line
        mutex_unlock(&clock->extreg_lock);
I don't see the change to the tdr_write() function that changes the argument from 'struct timespec ts' to 'struct timespec64 *', so this looks wrong to me.
There is a 'ts64' parameter, converted to 'ts' beforehand.

Ok, got it. The code looks correct then, though I'd like to see the use of 'timespec' pushed out as far as possible. How about changing the type for tdr_write() as well here?

tdr_write() itself should be fine until 2106, as it writes an unsigned number, but it's probably good to document that inside of that function.

Arnd

Richard Cochran

author maintainer

2015-03-23 08:10 UTC #33 lore raw

Re: [PATCH net-next V2 20/23] ptp: dp83640: convert to the 64 bit get/set time methods.

Quoted 6 lines · Arnd Bergmann

Ok, got it. The code looks correct then, though I'd like to see the use of 'timespec' pushed out as far as possible. How about changing the type for tdr_write() as well here?
tdr_write() itself should be fine until 2106, as it writes an unsigned number, but it's probably good to document that inside of that function.

Arnd, thanks for the careful review. I will work out follow up patches for this driver and the i210 (igb) to address the remaining issues. For this series, lets leave tdr_write as is now, but I will surely take another look for the fix.

Thanks,
Richard

Arnd Bergmann

2015-03-23 15:06 UTC #34 lore raw

Re: [PATCH net-next V2 20/23] ptp: dp83640: convert to the 64 bit get/set time methods.

Quoted 6 lines · Richard Cochran

Quoted 4 lines · Arnd Bergmann
Ok, got it. The code looks correct then, though I'd like to see the use of 'timespec' pushed out as far as possible. How about changing the type for tdr_write() as well here?

Quoted 8 lines

Quoted 2 lines
tdr_write() itself should be fine until 2106, as it writes an unsigned number, but it's probably good to document that inside of that function.
Arnd, thanks for the careful review. I will work out follow up patches for this driver and the i210 (igb) to address the remaining issues. For this series, lets leave tdr_write as is now, but I will surely take another look for the fix.

Ok, sounds good.

Arnd

Richard Cochran

author maintainer

2015-03-21 21:40 UTC #35 lore raw

PATCH v2 21/23 ptp: ixp46x: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device has a 64 bit clock register, where each clock tick is 16 nanoseconds, and so with this patch the driver is ready for the year 2038.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/ptp/ptp_ixp46x.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

drivers/ptp/ptp_ixp46x.c +4 -4

index 604d340..934c139 100644
@@ -175,7 +175,7 @@ static int ptp_ixp_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int ptp_ixp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int ptp_ixp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	u64 ns;
 	u32 remainder;

@@ -195,7 +195,7 @@ static int ptp_ixp_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int ptp_ixp_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
 {
 	u64 ns;
 	unsigned long flags;

@@ -248,8 +248,8 @@ static struct ptp_clock_info ptp_ixp_caps = {
 	.pps		= 0,
 	.adjfreq	= ptp_ixp_adjfreq,
 	.adjtime	= ptp_ixp_adjtime,
-	.gettime	= ptp_ixp_gettime,
-	.settime	= ptp_ixp_settime,
+	.gettime64	= ptp_ixp_gettime,
+	.settime64	= ptp_ixp_settime,
 	.enable		= ptp_ixp_enable,
 };

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #36 lore raw

PATCH v2 22/23 ptp: pch: convert to the 64 bit get/set time methods.

In reply to Richard Cochran

The device has a 64 bit clock register, where each clock tick is 32 nanoseconds, and so with this patch the driver is ready for the year 2038.

Compile tested only.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/ptp/ptp_pch.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

drivers/ptp/ptp_pch.c +4 -4

index 2554872..3aa22ae 100644
@@ -449,7 +449,7 @@ static int ptp_pch_adjtime(struct ptp_clock_info *ptp, s64 delta)
 	return 0;
 }
 
-static int ptp_pch_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+static int ptp_pch_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
 {
 	u64 ns;
 	u32 remainder;

@@ -467,7 +467,7 @@ static int ptp_pch_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
 }
 
 static int ptp_pch_settime(struct ptp_clock_info *ptp,
-			   const struct timespec *ts)
+			   const struct timespec64 *ts)
 {
 	u64 ns;
 	unsigned long flags;

@@ -518,8 +518,8 @@ static struct ptp_clock_info ptp_pch_caps = {
 	.pps		= 0,
 	.adjfreq	= ptp_pch_adjfreq,
 	.adjtime	= ptp_pch_adjtime,
-	.gettime	= ptp_pch_gettime,
-	.settime	= ptp_pch_settime,
+	.gettime64	= ptp_pch_gettime,
+	.settime64	= ptp_pch_settime,
 	.enable		= ptp_pch_enable,
 };

Signature · 1 line

1.7.10.4

Richard Cochran

author maintainer

2015-03-21 21:39 UTC #37 lore raw

PATCH v2 23/23 ptp: remove 32 bit get/set methods.

In reply to Richard Cochran

All of the PHC drivers have been converted to the new methods. This patch converts the three remaining callers within the core code and removes the older methods for good. As a result, the core PHC code is ready for the year 2038. However, some of the PHC drivers are not quite ready yet.

Signed-off-by Richard Cochran <[email protected]>

---
 drivers/ptp/ptp_chardev.c        |    8 +-------
 drivers/ptp/ptp_clock.c          |   15 ++++-----------
 include/linux/ptp_clock_kernel.h |    8 --------
 3 files changed, 5 insertions(+), 26 deletions(-)

drivers/ptp/ptp_chardev.c +1 -7

index 95bcf15..da7bae9 100644
@@ -125,7 +125,6 @@ long ptp_ioctl(struct posix_clock *pc, unsigned int cmd, unsigned long arg)
 	struct ptp_clock_info *ops = ptp->info;
 	struct ptp_clock_time *pct;
 	struct timespec64 ts;
-	struct timespec t2;
 	int enable, err = 0;
 	unsigned int i, pin_index;

@@ -202,12 +201,7 @@ long ptp_ioctl(struct posix_clock *pc, unsigned int cmd, unsigned long arg)
 			pct->sec = ts.tv_sec;
 			pct->nsec = ts.tv_nsec;
 			pct++;
-			if (ptp->info->gettime64) {
-				ptp->info->gettime64(ptp->info, &ts);
-			} else {
-				ptp->info->gettime(ptp->info, &t2);
-				ts = timespec_to_timespec64(t2);
-			}
+			ptp->info->gettime64(ptp->info, &ts);
 			pct->sec = ts.tv_sec;
 			pct->nsec = ts.tv_nsec;
 			pct++;

drivers/ptp/ptp_clock.c +4 ~~-11~~

index df50d5e..2e481b9 100644
@@ -109,9 +109,7 @@ static int ptp_clock_settime(struct posix_clock *pc, const struct timespec *tp)
 	struct ptp_clock *ptp = container_of(pc, struct ptp_clock, clock);
 	struct timespec64 ts = timespec_to_timespec64(*tp);
 
-	return  ptp->info->settime64 ?
-		ptp->info->settime64(ptp->info, &ts) :
-		ptp->info->settime(ptp->info, tp);
+	return  ptp->info->settime64(ptp->info, &ts);
 }
 
 static int ptp_clock_gettime(struct posix_clock *pc, struct timespec *tp)

@@ -120,14 +118,9 @@ static int ptp_clock_gettime(struct posix_clock *pc, struct timespec *tp)
 	struct timespec64 ts;
 	int err;
 
-	if (ptp->info->gettime64) {
-		err = ptp->info->gettime64(ptp->info, &ts);
-		if (!err)
-			*tp = timespec64_to_timespec(ts);
-	} else {
-		err = ptp->info->gettime(ptp->info, tp);
-	}
-
+	err = ptp->info->gettime64(ptp->info, &ts);
+	if (!err)
+		*tp = timespec64_to_timespec(ts);
 	return err;
 }

include/linux/ptp_clock_kernel.h +0 -8

index 7d6f8e6..b8b7306 100644
@@ -64,12 +64,6 @@ struct ptp_clock_request {
  * @adjtime:  Shifts the time of the hardware clock.
  *            parameter delta: Desired change in nanoseconds.
  *
- * @gettime:  Reads the current time from the hardware clock. (deprecated)
- *            parameter ts: Holds the result.
- *
- * @settime:  Set the current time on the hardware clock. (deprecated)
- *            parameter ts: Time value to set.
- *
  * @gettime64:  Reads the current time from the hardware clock.
  *              parameter ts: Holds the result.
  *

@@ -110,8 +104,6 @@ struct ptp_clock_info {
 	struct ptp_pin_desc *pin_config;
 	int (*adjfreq)(struct ptp_clock_info *ptp, s32 delta);
 	int (*adjtime)(struct ptp_clock_info *ptp, s64 delta);
-	int (*gettime)(struct ptp_clock_info *ptp, struct timespec *ts);
-	int (*settime)(struct ptp_clock_info *ptp, const struct timespec *ts);
 	int (*gettime64)(struct ptp_clock_info *ptp, struct timespec64 *ts);
 	int (*settime64)(struct ptp_clock_info *p, const struct timespec64 *ts);
 	int (*enable)(struct ptp_clock_info *ptp,

Signature · 1 line

1.7.10.4

Arnd Bergmann

2015-03-22 02:43 UTC #38 lore raw

In reply to Richard Cochran

Quoted 16 lines · Richard Cochran

This series converts the core driver methods of the PTP Hardware Clock (PHC) subsystem to use the 64 bit version of the timespec structure, making the core API ready for the year 2038.
In addition, I reviewed how each driver and device represents the time value at the hardware register level. Most of the drivers are ready, but a few will need some work before the year 2038, as shown:
Patch Driver
------------------------------------------------
   12      drivers/net/ethernet/intel/igb/igb_ptp.c
   15 ?    drivers/net/ethernet/sfc/ptp.c
   16      drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c
   18 ?    drivers/net/ethernet/tile/tilegx.c
   19      drivers/net/phy/dp83640.c

Quoted 4 lines

The commit log message documents how each driver is ready or why it is not ready. For patches 15 and 18, I could not easily find out the hardware representation of the time value, so I would ask the maintainers for a review.

Very nice, thanks for joining in with the effort.

I've commented on a few smaller issues on individual patches. Overall, I think we'd be slightly better off changing the type to ktime_t for the interface, the main downside of that being that we'd need to change a bit more code.

You'd still need to do the same kind of conversion, so the difference would be very minor in the end though, so I'm fine with whichever way you decide as the subsystem maintainer.

Arnd

Jeff Kirsher

Acked-by on the v2 series

2015-03-23 10:16 UTC #39 lore raw

In reply to Richard Cochran

Quoted 21 lines · Richard Cochran

This series converts the core driver methods of the PTP Hardware Clock (PHC) subsystem to use the 64 bit version of the timespec structure, making the core API ready for the year 2038.
In addition, I reviewed how each driver and device represents the time value at the hardware register level. Most of the drivers are ready, but a few will need some work before the year 2038, as shown:
Patch Driver
------------------------------------------------
   12      drivers/net/ethernet/intel/igb/igb_ptp.c
   15 ?    drivers/net/ethernet/sfc/ptp.c
   16      drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c
   18 ?    drivers/net/ethernet/tile/tilegx.c
   19      drivers/net/phy/dp83640.c
The commit log message documents how each driver is ready or why it is not ready. For patches 15 and 18, I could not easily find out the hardware representation of the time value, so I would ask the maintainers for a review.

Quoted 6 lines

Thanks,
Richard
* ChangeLog
** V2
- use the new methods in the posix clock code right away (patch #3)

The Intel driver changes look fine to me and as noted, more work appears to be needed for igb. It does not appear that Jake Jacobs got copied on the patches, so I will make sure he takes a look at the changes for our Intel drivers, since he did much of the work.

You get my ACK for the Intel changes.

Acked-by Jeff Kirsher <[email protected]>

Quoted 50 lines

Richard Cochran (23):
ptp: introduce get/set time methods with explicit 64 bit seconds.
ptp: use the 64 bit gettime method for the SYS_OFFSET ioctl.
ptp: use the 64 bit get/set time methods for the posix clock.
ptp: blackfin: convert to the 64 bit get/set time methods.
ptp: xgbe: convert to the 64 bit get/set time methods.
ptp: bnx2x: convert to the 64 bit get/set time methods.
ptp: tg3: convert to the 64 bit get/set time methods.
ptp: fec: convert to the 64 bit get/set time methods.
ptp: gianfar: convert to the 64 bit get/set time methods.
ptp: e1000e: convert to the 64 bit get/set time methods.
ptp: fm10k: convert to the 64 bit get/set time methods.
ptp: i40e: convert to the 64 bit get/set time methods.
ptp: igb: convert to the 64 bit get/set time methods.
ptp: ixgbe: convert to the 64 bit get/set time methods.
ptp: mlx4: convert to the 64 bit get/set time methods.
ptp: sfc: convert to the 64 bit get/set time methods.
ptp: stmmac: convert to the 64 bit get/set time methods.
ptp: cpts: convert to the 64 bit get/set time methods.
ptp: tilegx: convert to the 64 bit get/set time methods.
ptp: dp83640: convert to the 64 bit get/set time methods.
ptp: ixp46x: convert to the 64 bit get/set time methods.
ptp: pch: convert to the 64 bit get/set time methods.
ptp: remove 32 bit get/set methods.
 drivers/net/ethernet/adi/bfin_mac.c              |    8 ++---
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c         |   13 +++----
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c |    8 ++---
 drivers/net/ethernet/broadcom/tg3.c              |   10 +++---
 drivers/net/ethernet/freescale/fec_ptp.c         |    8 ++---
 drivers/net/ethernet/freescale/gianfar_ptp.c     |   13 +++----
 drivers/net/ethernet/intel/e1000e/ptp.c          |   16 ++++-----
 drivers/net/ethernet/intel/fm10k/fm10k_ptp.c     |   12 +++----
 drivers/net/ethernet/intel/i40e/i40e_ptp.c       |   26 +++++++-------
 drivers/net/ethernet/intel/igb/igb_ptp.c         |   41 +++++++++++-----------
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c     |   14 ++++----
 drivers/net/ethernet/mellanox/mlx4/en_clock.c    |   11 +++---
 drivers/net/ethernet/sfc/ptp.c                   |   22 ++++++------
 drivers/net/ethernet/stmicro/stmmac/stmmac_ptp.c |    8 ++---
 drivers/net/ethernet/ti/cpts.c                   |   12 +++----
 drivers/net/ethernet/tile/tilegx.c               |   16 +++++----
 drivers/net/phy/dp83640.c                        |   12 ++++---
 drivers/ptp/ptp_chardev.c                        |    8 ++---
 drivers/ptp/ptp_clock.c                          |   12 +++++--
 drivers/ptp/ptp_ixp46x.c                         |    8 ++---
 drivers/ptp/ptp_pch.c                            |    8 ++---
 include/linux/ptp_clock_kernel.h                 |   12 +++----
 22 files changed, 158 insertions(+), 140 deletions(-)