Skip to contentSkip to conversation

pps: several fixes and improvements

PATCH Superseded Alexander Gordeev 5 participants 33 messages Nov 24, 2010 – Dec 16, 2010

Subsystems (5)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v5 cover pps: several fixes and improvements

This patchset contains several changes that improve an overall design/performance of PPS subsystem. I'd like these patches to be merged mainline if no one objects.

Patches 1-4 are bugfixes.
Patches 5-12 are other improvements to PPS subsystem. Patches 13-15 add kernel consumer support. Patch 16 adds parallel port PPS client.
Patch 17 adds parallel port PPS generator.

You can find description for my previous patchset (it describes patches 13-17 in more detailed) here: http://lkml.org/lkml/2010/2/24/189

This patchset is tested against the vanilla 2.6.36 kernel. But we are actually using it on 2.6.33.7-rt29 rt-preempt kernel most of the time. Those who are interested in other versions of the patchset can find them in my git repository:
git://github.com/ago/linux-2.6.git

There is one problem however: kernel consumer works bad (if enabled) when CONFIG_NO_HZ is enabled. The reason for this is commit a092ff0f90cae22b2ac8028ecd2c6f6c1a9e4601. Without it hardpps() is able to sync to 1us precision in about 10 seconds. With CONFIG_NO_HZ it is not syncing at all. This only affects patches 13-15, others are ok.

Changelog
v4 -> v5:
* drop patch that adds PPS workqueue, it needs some more thinking
* replace patch that stops disabling interrupts with a lighter version
* add patch 5 which ensures that dcd_change() will never be called before tty_ldisc_open() or after tty_ldisc_close()
* fix patch 7: convert all remaining prints
* use BUG_ON only in pps_event
* use "canonical" error-handling style in pps_tty_open

Alexander Gordeev (17):
pps: trivial fixes
pps: declare variables where they are used in switch
pps: fix race in PPS_FETCH handler
pps: unify timestamp gathering
tty: don't allow ldisc dcd_change() after ldisc halt
pps: access pps device by direct pointer
pps: convert printk/pr_* to dev_*
pps: move idr stuff to pps.c
pps: do not disable interrupts for idr operations
pps: use BUG_ON for kernel API safety checks
pps: simplify conditions a bit
pps: timestamp is always passed to dcd_change()
ntp: add hardpps implementation
pps: capture MONOTONIC_RAW timestamps as well
pps: add kernel consumer support
pps: add parallel port PPS client
pps: add parallel port PPS signal generator

 Documentation/ioctl/ioctl-number.txt     |    2 +-
 Documentation/pps/pps.txt                |   46 ++++
 Documentation/serial/tty.txt             |    2 +-
 drivers/char/tty_io.c                    |    1 +
 drivers/char/tty_ldisc.c                 |    7 +
 drivers/pps/Kconfig                      |   10 +
 drivers/pps/Makefile                     |    2 +-
 drivers/pps/clients/Kconfig              |    7 +
 drivers/pps/clients/Makefile             |    1 +
 drivers/pps/clients/pps-ktimer.c         |   44 ++--
 drivers/pps/clients/pps-ldisc.c          |   54 ++---
 drivers/pps/clients/pps_parport.c        |  247 +++++++++++++++++
 drivers/pps/generators/Kconfig           |   17 ++
 drivers/pps/generators/Makefile          |    9 +
 drivers/pps/generators/pps_gen_parport.c |  275 +++++++++++++++++++
 drivers/pps/kapi.c                       |  221 +++++-----------
 drivers/pps/pps.c                        |  177 ++++++++++---
 include/linux/pps.h                      |    7 +
 include/linux/pps_kernel.h               |   54 +++-
 include/linux/serial_core.h              |   23 +-
 include/linux/time.h                     |    2 +
 include/linux/timex.h                    |    1 +
 include/linux/tty.h                      |    1 +
 include/linux/tty_ldisc.h                |    7 +-
 kernel/time/ntp.c                        |  425 ++++++++++++++++++++++++++++-
 kernel/time/timekeeping.c                |   38 +++
 26 files changed, 1387 insertions(+), 293 deletions(-)
 create mode 100644 drivers/pps/clients/pps_parport.c
 create mode 100644 drivers/pps/generators/Kconfig
 create mode 100644 drivers/pps/generators/Makefile
 create mode 100644 drivers/pps/generators/pps_gen_parport.c
Signature · 1 line
1.7.2.3
#1 lore raw

PATCH v5 1/17 pps: trivial fixes

Here are some very trivial fixes combined:
* add macro definitions to protect header file from including several times
* remove declaration for an unexistent array
* fix typos

---
 drivers/pps/kapi.c         |    2 +-
 include/linux/pps_kernel.h |    7 ++++++-
 2 files changed, 7 insertions(+), 2 deletions(-)
drivers/pps/kapi.c +1 -1
index 1aa02db..55f3961 100644
@@ -324,7 +324,7 @@ void pps_event(int source, struct pps_ktime *ts, int event, void *data)
 		captured = ~0;
 	}
 
-	/* Wake up iif captured somthing */
+	/* Wake up if captured something */
 	if (captured) {
 		pps->go = ~0;
 		wake_up_interruptible(&pps->queue);
include/linux/pps_kernel.h +6 -1
index e0a193f..c930d11 100644
@@ -18,6 +18,9 @@
  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#ifndef LINUX_PPS_KERNEL_H
+#define LINUX_PPS_KERNEL_H
+
 #include <linux/pps.h>
 
 #include <linux/cdev.h>
@@ -71,7 +74,6 @@ struct pps_device {
 
 extern spinlock_t pps_idr_lock;
 extern struct idr pps_idr;
-extern struct timespec pps_irq_ts[];
 
 extern struct device_attribute pps_attrs[];
 
@@ -87,3 +89,6 @@ extern void pps_unregister_source(int source);
 extern int pps_register_cdev(struct pps_device *pps);
 extern void pps_unregister_cdev(struct pps_device *pps);
 extern void pps_event(int source, struct pps_ktime *ts, int event, void *data);
+
+#endif /* LINUX_PPS_KERNEL_H */
+
Signature · 1 line
1.7.2.3
#2 lore raw

PATCH v5 2/17 pps: declare variables where they are used in switch

In reply to Alexander Gordeev

Move variable declarations where they are used in pps_cdev_ioctl.

---
 drivers/pps/pps.c |   10 ++++++----
 1 files changed, 6 insertions(+), 4 deletions(-)
drivers/pps/pps.c +6 -4
index ca5183b..c76afb9 100644
@@ -61,8 +61,6 @@ static long pps_cdev_ioctl(struct file *file,
 {
 	struct pps_device *pps = file->private_data;
 	struct pps_kparams params;
-	struct pps_fdata fdata;
-	unsigned long ticks;
 	void __user *uarg = (void __user *) arg;
 	int __user *iuarg = (int __user *) arg;
 	int err;
@@ -136,7 +134,9 @@ static long pps_cdev_ioctl(struct file *file,
 
 		break;
 
-	case PPS_FETCH:
+	case PPS_FETCH: {
+		struct pps_fdata fdata;
+
 		pr_debug("PPS_FETCH: source %d\n", pps->id);
 
 		err = copy_from_user(&fdata, uarg, sizeof(struct pps_fdata));
@@ -149,6 +149,8 @@ static long pps_cdev_ioctl(struct file *file,
 		if (fdata.timeout.flags & PPS_TIME_INVALID)
 			err = wait_event_interruptible(pps->queue, pps->go);
 		else {
+			unsigned long ticks;
+
 			pr_debug("timeout %lld.%09d\n",
 					(long long) fdata.timeout.sec,
 					fdata.timeout.nsec);
@@ -185,7 +187,7 @@ static long pps_cdev_ioctl(struct file *file,
 			return -EFAULT;
 
 		break;
-
+	}
 	default:
 		return -ENOTTY;
 		break;
Signature · 1 line
1.7.2.3
#3 lore raw

PATCH v5 3/17 pps: fix race in PPS_FETCH handler

In reply to Alexander Gordeev

There was a race in PPS_FETCH ioctl handler when several processes want to obtain PPS data simultaneously using sleeping PPS_FETCH. They all sleep most of the time in the system call.
With the old approach when the first process waiting on the pps queue is waken up it makes new system call right away and zeroes pps->go. So other processes continue to sleep. This is a clear race condition because of the global 'go' variable.
With the new approach pps->last_ev holds some value increasing at each PPS event. PPS_FETCH ioctl handler saves current value to the local variable at the very beginning so it can safely check that there is a new event by just comparing both variables.

---
 drivers/pps/kapi.c         |    4 ++--
 drivers/pps/pps.c          |   10 +++++++---
 include/linux/pps_kernel.h |    2 +-
 3 files changed, 10 insertions(+), 6 deletions(-)
drivers/pps/kapi.c +2 -2
index 55f3961..3f89f5eb 100644
@@ -326,8 +326,8 @@ void pps_event(int source, struct pps_ktime *ts, int event, void *data)
 
 	/* Wake up if captured something */
 	if (captured) {
-		pps->go = ~0;
-		wake_up_interruptible(&pps->queue);
+		pps->last_ev++;
+		wake_up_interruptible_all(&pps->queue);
 
 		kill_fasync(&pps->async_queue, SIGIO, POLL_IN);
 	}
drivers/pps/pps.c +7 -3
index c76afb9..dc7e66c 100644
@@ -136,6 +136,7 @@ static long pps_cdev_ioctl(struct file *file,
 
 	case PPS_FETCH: {
 		struct pps_fdata fdata;
+		unsigned int ev;
 
 		pr_debug("PPS_FETCH: source %d\n", pps->id);
 
@@ -143,11 +144,12 @@ static long pps_cdev_ioctl(struct file *file,
 		if (err)
 			return -EFAULT;
 
-		pps->go = 0;
+		ev = pps->last_ev;
 
 		/* Manage the timeout */
 		if (fdata.timeout.flags & PPS_TIME_INVALID)
-			err = wait_event_interruptible(pps->queue, pps->go);
+			err = wait_event_interruptible(pps->queue,
+					ev != pps->last_ev);
 		else {
 			unsigned long ticks;
 
@@ -159,7 +161,9 @@ static long pps_cdev_ioctl(struct file *file,
 
 			if (ticks != 0) {
 				err = wait_event_interruptible_timeout(
-						pps->queue, pps->go, ticks);
+						pps->queue,
+						ev != pps->last_ev,
+						ticks);
 				if (err == 0)
 					return -ETIMEDOUT;
 			}
include/linux/pps_kernel.h +1 -1
index c930d11..c3aed4b 100644
@@ -55,7 +55,7 @@ struct pps_device {
 	struct pps_ktime clear_tu;
 	int current_mode;			/* PPS mode at event time */
 
-	int go;					/* PPS event is arrived? */
+	unsigned int last_ev;			/* last PPS event id */
 	wait_queue_head_t queue;		/* PPS event queue */
 
 	unsigned int id;			/* PPS source unique ID */
Signature · 1 line
1.7.2.3
#4 lore raw

PATCH v5 4/17 pps: unify timestamp gathering

In reply to Alexander Gordeev

Add a helper function to gather timestamps. This way clients don't have to duplicate it.

---
 drivers/pps/clients/pps-ktimer.c |    9 ++-------
 drivers/pps/clients/pps-ldisc.c  |   18 ++++++------------
 drivers/pps/kapi.c               |   19 ++++++++++++-------
 include/linux/pps_kernel.h       |   19 ++++++++++++++++++-
 include/linux/serial_core.h      |    5 +++--
 include/linux/tty_ldisc.h        |    5 +++--
 6 files changed, 44 insertions(+), 31 deletions(-)
drivers/pps/clients/pps-ktimer.c +2 -7
index e7ef5b8..e1bdd8b 100644
@@ -40,18 +40,13 @@ static struct timer_list ktimer;
 
 static void pps_ktimer_event(unsigned long ptr)
 {
-	struct timespec __ts;
-	struct pps_ktime ts;
+	struct pps_event_time ts;
 
 	/* First of all we get the time stamp... */
-	getnstimeofday(&__ts);
+	pps_get_ts(&ts);
 
 	pr_info("PPS event at %lu\n", jiffies);
 
-	/* ... and translate it to PPS time data struct */
-	ts.sec = __ts.tv_sec;
-	ts.nsec = __ts.tv_nsec;
-
 	pps_event(source, &ts, PPS_CAPTUREASSERT, NULL);
 
 	mod_timer(&ktimer, jiffies + HZ);
drivers/pps/clients/pps-ldisc.c +6 -12
index 8e1932d..20fc9f7 100644
@@ -27,26 +27,20 @@
 #define PPS_TTY_MAGIC		0x0001
 
 static void pps_tty_dcd_change(struct tty_struct *tty, unsigned int status,
-				struct timespec *ts)
+				struct pps_event_time *ts)
 {
 	int id = (long)tty->disc_data;
-	struct timespec __ts;
-	struct pps_ktime pps_ts;
+	struct pps_event_time __ts;
 
 	/* First of all we get the time stamp... */
-	getnstimeofday(&__ts);
+	pps_get_ts(&__ts);
 
 	/* Does caller give us a timestamp? */
-	if (ts) {	/* Yes. Let's use it! */
-		pps_ts.sec = ts->tv_sec;
-		pps_ts.nsec = ts->tv_nsec;
-	} else {	/* No. Do it ourself! */
-		pps_ts.sec = __ts.tv_sec;
-		pps_ts.nsec = __ts.tv_nsec;
-	}
+	if (!ts)	/* No. Do it ourself! */
+		ts = &__ts;
 
 	/* Now do the PPS event report */
-	pps_event(id, &pps_ts, status ? PPS_CAPTUREASSERT : PPS_CAPTURECLEAR,
+	pps_event(id, ts, status ? PPS_CAPTUREASSERT : PPS_CAPTURECLEAR,
 			NULL);
 
 	pr_debug("PPS %s at %lu on source #%d\n",
drivers/pps/kapi.c +12 -7
index 3f89f5eb..b431d83 100644
@@ -268,11 +268,12 @@ EXPORT_SYMBOL(pps_unregister_source);
  *	pps->info.echo(source, event, data);
  */
 
-void pps_event(int source, struct pps_ktime *ts, int event, void *data)
+void pps_event(int source, struct pps_event_time *ts, int event, void *data)
 {
 	struct pps_device *pps;
 	unsigned long flags;
 	int captured = 0;
+	struct pps_ktime ts_real;
 
 	if ((event & (PPS_CAPTUREASSERT | PPS_CAPTURECLEAR)) == 0) {
 		printk(KERN_ERR "pps: unknown event (%x) for source %d\n",
@@ -284,8 +285,10 @@ void pps_event(int source, struct pps_ktime *ts, int event, void *data)
 	if (!pps)
 		return;
 
-	pr_debug("PPS event on source %d at %llu.%06u\n",
-			pps->id, (unsigned long long) ts->sec, ts->nsec);
+	pr_debug("PPS event on source %d at %ld.%09ld\n",
+			pps->id, ts->ts_real.tv_sec, ts->ts_real.tv_nsec);
+
+	timespec_to_pps_ktime(&ts_real, ts->ts_real);
 
 	spin_lock_irqsave(&pps->lock, flags);
 
@@ -299,10 +302,11 @@ void pps_event(int source, struct pps_ktime *ts, int event, void *data)
 			(pps->params.mode & PPS_CAPTUREASSERT)) {
 		/* We have to add an offset? */
 		if (pps->params.mode & PPS_OFFSETASSERT)
-			pps_add_offset(ts, &pps->params.assert_off_tu);
+			pps_add_offset(&ts_real,
+					&pps->params.assert_off_tu);
 
 		/* Save the time stamp */
-		pps->assert_tu = *ts;
+		pps->assert_tu = ts_real;
 		pps->assert_sequence++;
 		pr_debug("capture assert seq #%u for source %d\n",
 			pps->assert_sequence, source);
@@ -313,10 +317,11 @@ void pps_event(int source, struct pps_ktime *ts, int event, void *data)
 			(pps->params.mode & PPS_CAPTURECLEAR)) {
 		/* We have to add an offset? */
 		if (pps->params.mode & PPS_OFFSETCLEAR)
-			pps_add_offset(ts, &pps->params.clear_off_tu);
+			pps_add_offset(&ts_real,
+					&pps->params.clear_off_tu);
 
 		/* Save the time stamp */
-		pps->clear_tu = *ts;
+		pps->clear_tu = ts_real;
 		pps->clear_sequence++;
 		pr_debug("capture clear seq #%u for source %d\n",
 			pps->clear_sequence, source);
include/linux/pps_kernel.h +18 -1
index c3aed4b..32aa676 100644
@@ -43,6 +43,10 @@ struct pps_source_info {
 	struct device *dev;
 };
 
+struct pps_event_time {
+	struct timespec ts_real;
+};
+
 /* The main struct */
 struct pps_device {
 	struct pps_source_info info;		/* PSS source info */
@@ -88,7 +92,20 @@ extern int pps_register_source(struct pps_source_info *info,
 extern void pps_unregister_source(int source);
 extern int pps_register_cdev(struct pps_device *pps);
 extern void pps_unregister_cdev(struct pps_device *pps);
-extern void pps_event(int source, struct pps_ktime *ts, int event, void *data);
+extern void pps_event(int source, struct pps_event_time *ts, int event,
+		void *data);
+
+static inline void timespec_to_pps_ktime(struct pps_ktime *kt,
+		struct timespec ts)
+{
+	kt->sec = ts.tv_sec;
+	kt->nsec = ts.tv_nsec;
+}
+
+static inline void pps_get_ts(struct pps_event_time *ts)
+{
+	getnstimeofday(&ts->ts_real);
+}
 
 #endif /* LINUX_PPS_KERNEL_H */
 
include/linux/serial_core.h +3 -2
index 563e234..55c8192 100644
@@ -206,6 +206,7 @@
 #include <linux/tty.h>
 #include <linux/mutex.h>
 #include <linux/sysrq.h>
+#include <linux/pps_kernel.h>
 
 struct uart_port;
 struct serial_struct;
@@ -510,10 +511,10 @@ uart_handle_dcd_change(struct uart_port *uport, unsigned int status)
 	struct uart_state *state = uport->state;
 	struct tty_port *port = &state->port;
 	struct tty_ldisc *ld = tty_ldisc_ref(port->tty);
-	struct timespec ts;
+	struct pps_event_time ts;
 
 	if (ld && ld->ops->dcd_change)
-		getnstimeofday(&ts);
+		pps_get_ts(&ts);
 
 	uport->icount.dcd++;
 #ifdef CONFIG_HARD_PPS
include/linux/tty_ldisc.h +3 -2
index 526d66f..763e061 100644
@@ -101,7 +101,7 @@
  *	any pending driver I/O is completed.
  *
  * void (*dcd_change)(struct tty_struct *tty, unsigned int status,
- * 			struct timespec *ts)
+ *			struct pps_event_time *ts)
  *
  *	Tells the discipline that the DCD pin has changed its status and
  *	the relative timestamp. Pointer ts can be NULL.
@@ -109,6 +109,7 @@
 
 #include <linux/fs.h>
 #include <linux/wait.h>
+#include <linux/pps_kernel.h>
 
 struct tty_ldisc_ops {
 	int	magic;
@@ -143,7 +144,7 @@ struct tty_ldisc_ops {
 			       char *fp, int count);
 	void	(*write_wakeup)(struct tty_struct *);
 	void	(*dcd_change)(struct tty_struct *, unsigned int,
-				struct timespec *);
+				struct pps_event_time *);
 
 	struct  module *owner;
 	
Signature · 1 line
1.7.2.3
#5 lore raw

PATCH v5 5/17 tty: don't allow ldisc dcd_change() after ldisc halt

In reply to Alexander Gordeev

There was a possibility that uart_handle_dcd_change() could obtain a reference to ldisc while running in parallel with tty_set_ldisc() on different CPU but call dcd_change() operation after tty_ldisc_close() which is incorrect.

Treat this situation specially by locking the whole uart_handle_dcd_change() with spinlock and adding a "barrier" to tty_ldisc_halt() which ensures that there are no active ldisc references in uart_handle_dcd_change() after tty_ldisc_halt().

---
 drivers/char/tty_io.c       |    1 +
 drivers/char/tty_ldisc.c    |    7 +++++++
 include/linux/serial_core.h |   18 ++++++++++++------
 include/linux/tty.h         |    1 +
 4 files changed, 21 insertions(+), 6 deletions(-)
drivers/char/tty_io.c +1 -0
index 613c852..18576d4 100644
@@ -2816,6 +2816,7 @@ void initialize_tty_struct(struct tty_struct *tty,
 	mutex_init(&tty->echo_lock);
 	spin_lock_init(&tty->read_lock);
 	spin_lock_init(&tty->ctrl_lock);
+	spin_lock_init(&tty->dcd_change_lock);
 	INIT_LIST_HEAD(&tty->tty_files);
 	INIT_WORK(&tty->SAK_work, do_SAK_work);
 
drivers/char/tty_ldisc.c +7 -0
index 412f977..27fadb0 100644
@@ -522,11 +522,18 @@ static void tty_ldisc_restore(struct tty_struct *tty, struct tty_ldisc *old)
  *	You need to do a 'flush_scheduled_work()' (outside the ldisc_mutex)
  *	in order to make sure any currently executing ldisc work is also
  *	flushed.
+ *
+ *	dcd_change() doesn't use workqueues so it needs a special
+ *	"barrier", which ensures that there are no active ldisc references
+ *	in dcd_change().
  */
 
 static int tty_ldisc_halt(struct tty_struct *tty)
 {
+	spin_lock_irq(&tty->dcd_change_lock);
 	clear_bit(TTY_LDISC, &tty->flags);
+	spin_unlock_irq(&tty->dcd_change_lock);
+
 	return cancel_delayed_work_sync(&tty->buf.work);
 }
 
include/linux/serial_core.h +12 -6
index 55c8192..62835b6 100644
@@ -508,11 +508,15 @@ static inline int uart_handle_break(struct uart_port *port)
 static inline void
 uart_handle_dcd_change(struct uart_port *uport, unsigned int status)
 {
-	struct uart_state *state = uport->state;
-	struct tty_port *port = &state->port;
-	struct tty_ldisc *ld = tty_ldisc_ref(port->tty);
+	struct tty_port *port = &uport->state->port;
+	struct tty_struct *tty = port->tty;
 	struct pps_event_time ts;
+	struct tty_ldisc *ld;
+	unsigned long flags;
 
+	spin_lock_irqsave(&tty->dcd_change_lock, flags);
+
+	ld = tty_ldisc_ref(tty);
 	if (ld && ld->ops->dcd_change)
 		pps_get_ts(&ts);
 
@@ -525,14 +529,16 @@ uart_handle_dcd_change(struct uart_port *uport, unsigned int status)
 	if (port->flags & ASYNC_CHECK_CD) {
 		if (status)
 			wake_up_interruptible(&port->open_wait);
-		else if (port->tty)
-			tty_hangup(port->tty);
+		else if (tty)
+			tty_hangup(tty);
 	}
 
 	if (ld && ld->ops->dcd_change)
-		ld->ops->dcd_change(port->tty, status, &ts);
+		ld->ops->dcd_change(tty, status, &ts);
 	if (ld)
 		tty_ldisc_deref(ld);
+
+	spin_unlock_irqrestore(&tty->dcd_change_lock, flags);
 }
 
 /**
include/linux/tty.h +1 -0
index 67d64e6..506fe1c 100644
@@ -327,6 +327,7 @@ struct tty_struct {
 	/* If the tty has a pending do_SAK, queue it here - akpm */
 	struct work_struct SAK_work;
 	struct tty_port *port;
+	spinlock_t dcd_change_lock;
 };
 
 /* Each of a tty's open files has private_data pointing to tty_file_private */
Signature · 1 line
1.7.2.3
#6 lore raw

Re: [PATCHv5 05/17] tty: don't allow ldisc dcd_change() after ldisc halt

Quoted 4 lines · On Wed, 24 Nov 2010 19:15:43 +0300 Alexander Gordeev <[email protected]>…

There was a possibility that uart_handle_dcd_change() could obtain a reference to ldisc while running in parallel with tty_set_ldisc() on different CPU but call dcd_change() operation after tty_ldisc_close() which is incorrect.

How can this occur ?

include/linux/serial_core.h static inline int uart_handle_break(struct uart_port *port) · line 519

Quoted 3 lines
include/linux/serial_core.h +3 -0
+	spin_lock_irqsave(&tty->dcd_change_lock, flags);
+
+	ld = tty_ldisc_ref(tty);

What is the expecting lock ordering rule here ?

I don't see why this patch is needed. You've got an ldisc ref from tty_ldisc_ref, until you drop that ldisc ref you are fine. If for some reason that is not the case then there is a bug in the ldisc code.

Alan

#7 lore raw

Re: [PATCHv5 05/17] tty: don't allow ldisc dcd_change() after ldisc halt

В Wed, 24 Nov 2010 16:43:29 +0000
Alan Cox <[email protected]> пишет:

Quoted 19 lines
Quoted 4 lines · On Wed, 24 Nov 2010 19:15:43 +0300 Alexander Gordeev <[email protected]>…

There was a possibility that uart_handle_dcd_change() could obtain a reference to ldisc while running in parallel with tty_set_ldisc() on different CPU but call dcd_change() operation after tty_ldisc_close() which is incorrect.

How can this occur ?

include/linux/serial_core.h static inline int uart_handle_break(struct uart_port *port) · line 519

Quoted 3 lines
include/linux/serial_core.h +3 -0
+	spin_lock_irqsave(&tty->dcd_change_lock, flags);
+
+	ld = tty_ldisc_ref(tty);

What is the expecting lock ordering rule here ?

Quoted 3 lines

I don't see why this patch is needed. You've got an ldisc ref from tty_ldisc_ref, until you drop that ldisc ref you are fine. If for some reason that is not the case then there is a bug in the ldisc code.

Yes, indeed, it's a bug. Please consider the following example:

CPU1 CPU2
=========================================================
uart_handle_dcd_change() { tty_set_ldisc() {
ld = tty_ldisc_ref(...) ...

  ...					  tty_ldisc_halt(...)
  ...					  ...
  ...					  tty_ldisc_close(...)
  if (ld && ld->ops->dcd_change)	  ...
    ld->ops->dcd_change(...);		  ...

... tty_ldisc_open(...)
} }

I think that semantically ldisc ops should never be called before open or after close. This situation is possible because tty_ldisc_halt() only ensures that no more references are taken. This is ok for everything except dcd_change() because it cleans up workqueue and doesn't accept any more data. dcd_change() is a different story because it doesn't use workqueues.

I think tty code is exactly the right place to fix this bug; this is what my patch is for.

Signature · 1 line
  Alexander
#8 lore raw

Re: [PATCHv5 05/17] tty: don't allow ldisc dcd_change() after ldisc halt

Quoted 6 lines

Yes, indeed, it's a bug. Please consider the following example:

CPU1 CPU2
=========================================================
uart_handle_dcd_change() { tty_set_ldisc() {
ld = tty_ldisc_ref(...) ...

[We have a reference]

Quoted 1 line

... tty_ldisc_halt(...)

[Should block]

Quoted 2 lines

I think tty code is exactly the right place to fix this bug; this is what my patch is for.

More special case magic on top of the current crap isn't the right fix here, tty_ldisc_halt needs to wait for the references to hit zero.

Alan

#9 lore raw

Re: [PATCHv5 05/17] tty: don't allow ldisc dcd_change() after ldisc halt

В Wed, 24 Nov 2010 22:36:00 +0000
Alan Cox <[email protected]> пишет:

Quoted 8 lines
Quoted 6 lines

Yes, indeed, it's a bug. Please consider the following example:

CPU1 CPU2
=========================================================
uart_handle_dcd_change() { tty_set_ldisc() {
ld = tty_ldisc_ref(...) ...

[We have a reference]

Quoted 8 lines
Quoted 1 line

... tty_ldisc_halt(...)

[Should block]

Quoted 2 lines

I think tty code is exactly the right place to fix this bug; this is what my patch is for.

More special case magic on top of the current crap isn't the right fix here, tty_ldisc_halt needs to wait for the references to hit zero.

Didn't know that current design is crap. :)
Ok, I think I'll add a new waitqueue and a new bit (TTY_LDISC_NOREF) that halt will wait for. Is it good?

Signature · 1 line
  Alexander
#10 lore raw

Re: [PATCHv5 05/17] tty: don't allow ldisc dcd_change() after ldisc halt

Quoted 1 line

Didn't know that current design is crap. :)

It's based on the old big kernel lock and it has lots of horrors when used from interrupt paths as a result. A big chunk of it can in time I think move to just having a per tty ldisc lock that covers all changes etc, but we aren't there yet.

Quoted 1 line

Ok, I think I'll add a new waitqueue and a new bit (TTY_LDISC_NOREF)

We've got the wait queue already

My first guess would be something like

static struct tty_ldisc *tty_ldisc_irqget(struct tty_struct *tty)

{
        unsigned long flags;
        struct tty_ldisc *ld;
   
        spin_lock_irqsave(&tty_ldisc_lock, flags);
        ld = NULL;
        if (test_bit(TTY_LDISC, &tty->flags)) {
                ld = get_ldisc(tty->ldisc);
		WARN_ON(test_and_set_bit(TTY_LDISC_IRQ, &tty->flags));
	}
        spin_unlock_irqrestore(&tty_ldisc_lock, flags);
        return ld;
}

static struct tty_ldisc *tty_ldisc_irqput(struct tty_ldisc *ld)

{
	clear_bit(TTY_LDISC_IRQ, &tty->flags);
	tty_ldisc_put(ld);
}

(and I think you can then wait on tty_ldisc_idle for the flag to go clear after clearing TTY_LDISC)

It might be better to make it a count of course in case there are later multiple users of this ?

Alan

#11 lore raw

Re: [PATCHv5 05/17] tty: don't allow ldisc dcd_change() after ldisc halt

Hi,

Very sorry for the long delay.

В Thu, 25 Nov 2010 14:03:54 +0000
Alan Cox <[email protected]> пишет:

Quoted 34 lines
Quoted 1 line

Didn't know that current design is crap. :)

It's based on the old big kernel lock and it has lots of horrors when used from interrupt paths as a result. A big chunk of it can in time I think move to just having a per tty ldisc lock that covers all changes etc, but we aren't there yet.

Quoted 1 line

Ok, I think I'll add a new waitqueue and a new bit (TTY_LDISC_NOREF)

We've got the wait queue already

My first guess would be something like

static struct tty_ldisc *tty_ldisc_irqget(struct tty_struct *tty)

{
        unsigned long flags;
        struct tty_ldisc *ld;
   
        spin_lock_irqsave(&tty_ldisc_lock, flags);
        ld = NULL;
        if (test_bit(TTY_LDISC, &tty->flags)) {
                ld = get_ldisc(tty->ldisc);
		WARN_ON(test_and_set_bit(TTY_LDISC_IRQ, &tty->flags));
	}
        spin_unlock_irqrestore(&tty_ldisc_lock, flags);
        return ld;
}

static struct tty_ldisc *tty_ldisc_irqput(struct tty_ldisc *ld)

{
	clear_bit(TTY_LDISC_IRQ, &tty->flags);
	tty_ldisc_put(ld);
}
Quoted 5 lines

(and I think you can then wait on tty_ldisc_idle for the flag to go clear after clearing TTY_LDISC)

It might be better to make it a count of course in case there are later multiple users of this ?

I've just checked the code from v2.6.37 release candidates and found that the problem is already completely fixed there with the revoking of tty_ldisc_idle and tty_ldisc_wait_idle() in 100eeae2. :)

Sorry for the fuzz. The discussed patched is not needed any more and will be dropped.

Signature · 1 line
  Alexander
#12 lore raw

PATCH v5 6/17 pps: access pps device by direct pointer

In reply to Alexander Gordeev

Using device index as a pointer needs some unnecessary work to be done every time the pointer is needed (in irq handler for example). Using a direct pointer is much more easy (and safe as well).

---
 drivers/pps/clients/pps-ktimer.c |   30 ++++-----
 drivers/pps/clients/pps-ldisc.c  |   36 ++++++-----
 drivers/pps/kapi.c               |  124 +++++++++-----------------------------
 drivers/pps/pps.c                |   22 ++-----
 include/linux/pps_kernel.h       |   23 +++----
 5 files changed, 75 insertions(+), 160 deletions(-)
drivers/pps/clients/pps-ktimer.c +13 -17
index e1bdd8b..966ead1 100644
@@ -31,7 +31,7 @@
  * Global variables
  */
 
-static int source;
+static struct pps_device *pps;
 static struct timer_list ktimer;
 
 /*
@@ -47,7 +47,7 @@ static void pps_ktimer_event(unsigned long ptr)
 
 	pr_info("PPS event at %lu\n", jiffies);
 
-	pps_event(source, &ts, PPS_CAPTUREASSERT, NULL);
+	pps_event(pps, &ts, PPS_CAPTUREASSERT, NULL);
 
 	mod_timer(&ktimer, jiffies + HZ);
 }
@@ -56,12 +56,11 @@ static void pps_ktimer_event(unsigned long ptr)
  * The echo function
  */
 
-static void pps_ktimer_echo(int source, int event, void *data)
+static void pps_ktimer_echo(struct pps_device *pps, int event, void *data)
 {
-	pr_info("echo %s %s for source %d\n",
+	dev_info(pps->dev, "echo %s %s\n",
 		event & PPS_CAPTUREASSERT ? "assert" : "",
-		event & PPS_CAPTURECLEAR ? "clear" : "",
-		source);
+		event & PPS_CAPTURECLEAR ? "clear" : "");
 }
 
 /*
@@ -84,30 +83,27 @@ static struct pps_source_info pps_ktimer_info = {
 
 static void __exit pps_ktimer_exit(void)
 {
-	del_timer_sync(&ktimer);
-	pps_unregister_source(source);
+	dev_info(pps->dev, "ktimer PPS source unregistered\n");
 
-	pr_info("ktimer PPS source unregistered\n");
+	del_timer_sync(&ktimer);
+	pps_unregister_source(pps);
 }
 
 static int __init pps_ktimer_init(void)
 {
-	int ret;
-
-	ret = pps_register_source(&pps_ktimer_info,
+	pps = pps_register_source(&pps_ktimer_info,
 				PPS_CAPTUREASSERT | PPS_OFFSETASSERT);
-	if (ret < 0) {
+	if (pps == NULL) {
 		printk(KERN_ERR "cannot register ktimer source\n");
-		return ret;
+		return -ENOMEM;
 	}
-	source = ret;
 
 	setup_timer(&ktimer, pps_ktimer_event, 0);
 	mod_timer(&ktimer, jiffies + HZ);
 
-	pr_info("ktimer PPS source registered at %d\n", source);
+	dev_info(pps->dev, "ktimer PPS source registered\n");
 
-	return  0;
+	return 0;
 }
 
 module_init(pps_ktimer_init);
drivers/pps/clients/pps-ldisc.c +19 -17
index 20fc9f7..f0ba65e 100644
@@ -29,7 +29,7 @@
 static void pps_tty_dcd_change(struct tty_struct *tty, unsigned int status,
 				struct pps_event_time *ts)
 {
-	int id = (long)tty->disc_data;
+	struct pps_device *pps = (struct pps_device *)tty->disc_data;
 	struct pps_event_time __ts;
 
 	/* First of all we get the time stamp... */
@@ -39,12 +39,14 @@ static void pps_tty_dcd_change(struct tty_struct *tty, unsigned int status,
 	if (!ts)	/* No. Do it ourself! */
 		ts = &__ts;
 
+	BUG_ON(pps == NULL);
+
 	/* Now do the PPS event report */
-	pps_event(id, ts, status ? PPS_CAPTUREASSERT : PPS_CAPTURECLEAR,
-			NULL);
+	pps_event(pps, ts, status ? PPS_CAPTUREASSERT :
+			PPS_CAPTURECLEAR, NULL);
 
-	pr_debug("PPS %s at %lu on source #%d\n",
-			status ? "assert" : "clear", jiffies, id);
+	dev_dbg(pps->dev, "PPS %s at %lu\n",
+			status ? "assert" : "clear", jiffies);
 }
 
 static int (*alias_n_tty_open)(struct tty_struct *tty);
@@ -54,7 +56,7 @@ static int pps_tty_open(struct tty_struct *tty)
 	struct pps_source_info info;
 	struct tty_driver *drv = tty->driver;
 	int index = tty->index + drv->name_base;
-	int ret;
+	struct pps_device *pps;
 
 	info.owner = THIS_MODULE;
 	info.dev = NULL;
@@ -64,20 +66,19 @@ static int pps_tty_open(struct tty_struct *tty)
 			PPS_OFFSETASSERT | PPS_OFFSETCLEAR | \
 			PPS_CANWAIT | PPS_TSFMT_TSPEC;
 
-	ret = pps_register_source(&info, PPS_CAPTUREBOTH | \
+	pps = pps_register_source(&info, PPS_CAPTUREBOTH | \
 				PPS_OFFSETASSERT | PPS_OFFSETCLEAR);
-	if (ret < 0) {
+	if (pps == NULL) {
 		pr_err("cannot register PPS source \"%s\"\n", info.path);
-		return ret;
+		return -ENOMEM;
 	}
-	tty->disc_data = (void *)(long)ret;
+	tty->disc_data = pps;
 
 	/* Should open N_TTY ldisc too */
-	ret = alias_n_tty_open(tty);
-	if (ret < 0)
-		pps_unregister_source((long)tty->disc_data);
+	if (alias_n_tty_open(tty) < 0)
+		pps_unregister_source(pps);
 
-	pr_info("PPS source #%d \"%s\" added\n", ret, info.path);
+	dev_info(pps->dev, "source \"%s\" added\n", info.path);
 
 	return 0;
 }
@@ -86,12 +87,13 @@ static void (*alias_n_tty_close)(struct tty_struct *tty);
 
 static void pps_tty_close(struct tty_struct *tty)
 {
-	int id = (long)tty->disc_data;
+	struct pps_device *pps = (struct pps_device *)tty->disc_data;
 
-	pps_unregister_source(id);
 	alias_n_tty_close(tty);
 
-	pr_info("PPS source #%d removed\n", id);
+	tty->disc_data = NULL;
+	dev_info(pps->dev, "removed\n");
+	pps_unregister_source(pps);
 }
 
 static struct tty_ldisc_ops pps_ldisc_ops;
drivers/pps/kapi.c +28 -96

222 lines not on this page · raw message

drivers/pps/pps.c +6 -16
index dc7e66c..1922f27 100644
@@ -204,12 +204,6 @@ static int pps_cdev_open(struct inode *inode, struct file *file)
 {
 	struct pps_device *pps = container_of(inode->i_cdev,
 						struct pps_device, cdev);
-	int found;
-
-	found = pps_get_source(pps->id) != 0;
-	if (!found)
-		return -ENODEV;
-
 	file->private_data = pps;
 
 	return 0;
@@ -217,11 +211,6 @@ static int pps_cdev_open(struct inode *inode, struct file *file)
 
 static int pps_cdev_release(struct inode *inode, struct file *file)
 {
-	struct pps_device *pps = file->private_data;
-
-	/* Free the PPS source and wake up (possible) deregistration */
-	pps_put_source(pps);
-
 	return 0;
 }
 
@@ -242,22 +231,23 @@ static const struct file_operations pps_cdev_fops = {
 int pps_register_cdev(struct pps_device *pps)
 {
 	int err;
+	dev_t devt;
+
+	devt = MKDEV(MAJOR(pps_devt), pps->id);
 
-	pps->devno = MKDEV(MAJOR(pps_devt), pps->id);
 	cdev_init(&pps->cdev, &pps_cdev_fops);
 	pps->cdev.owner = pps->info.owner;
 
-	err = cdev_add(&pps->cdev, pps->devno, 1);
+	err = cdev_add(&pps->cdev, devt, 1);
 	if (err) {
 		printk(KERN_ERR "pps: %s: failed to add char device %d:%d\n",
 				pps->info.name, MAJOR(pps_devt), pps->id);
 		return err;
 	}
-	pps->dev = device_create(pps_class, pps->info.dev, pps->devno, NULL,
+	pps->dev = device_create(pps_class, pps->info.dev, devt, pps,
 							"pps%d", pps->id);
 	if (IS_ERR(pps->dev))
 		goto del_cdev;
-	dev_set_drvdata(pps->dev, pps);
 
 	pr_debug("source %s got cdev (%d:%d)\n", pps->info.name,
 			MAJOR(pps_devt), pps->id);
@@ -272,7 +262,7 @@ del_cdev:
 
 void pps_unregister_cdev(struct pps_device *pps)
 {
-	device_destroy(pps_class, pps->devno);
+	device_destroy(pps_class, pps->dev->devt);
 	cdev_del(&pps->cdev);
 }
 
include/linux/pps_kernel.h +9 -14
index 32aa676..1aedf50 100644
@@ -31,13 +31,16 @@
  * Global defines
  */
 
+struct pps_device;
+
 /* The specific PPS source info */
 struct pps_source_info {
 	char name[PPS_MAX_NAME_LEN];		/* simbolic name */
 	char path[PPS_MAX_NAME_LEN];		/* path of connected device */
 	int mode;				/* PPS's allowed mode */
 
-	void (*echo)(int source, int event, void *data); /* PPS echo function */
+	void (*echo)(struct pps_device *pps,
+			int event, void *data);	/* PPS echo function */
 
 	struct module *owner;
 	struct device *dev;
@@ -65,35 +68,27 @@ struct pps_device {
 	unsigned int id;			/* PPS source unique ID */
 	struct cdev cdev;
 	struct device *dev;
-	int devno;
 	struct fasync_struct *async_queue;	/* fasync method */
 	spinlock_t lock;
-
-	atomic_t usage;				/* usage count */
 };
 
 /*
  * Global variables
  */
 
-extern spinlock_t pps_idr_lock;
-extern struct idr pps_idr;
-
 extern struct device_attribute pps_attrs[];
 
 /*
  * Exported functions
  */
 
-struct pps_device *pps_get_source(int source);
-extern void pps_put_source(struct pps_device *pps);
-extern int pps_register_source(struct pps_source_info *info,
-				int default_params);
-extern void pps_unregister_source(int source);
+extern struct pps_device *pps_register_source(
+		struct pps_source_info *info, int default_params);
+extern void pps_unregister_source(struct pps_device *pps);
 extern int pps_register_cdev(struct pps_device *pps);
 extern void pps_unregister_cdev(struct pps_device *pps);
-extern void pps_event(int source, struct pps_event_time *ts, int event,
-		void *data);
+extern void pps_event(struct pps_device *pps,
+		struct pps_event_time *ts, int event, void *data);
 
 static inline void timespec_to_pps_ktime(struct pps_ktime *kt,
 		struct timespec ts)
Signature · 1 line
1.7.2.3
maintainer Acked-by on v5 6/17
#13 lore raw

Re: [PATCHv5 06/17] pps: access pps device by direct pointer

Quoted 5 lines · Alexander Gordeev

Using device index as a pointer needs some unnecessary work to be done every time the pointer is needed (in irq handler for example). Using a direct pointer is much more easy (and safe as well).

Signature · 5 lines
GNU/Linux Solutions                  e-mail: [email protected]
Linux Device Driver                          [email protected]
Embedded Systems                     phone:  +39 349 2432127
UNIX programming                     skype:  rodolfo.giometti
Freelance ICT Italia - Consulente ICT Italia - www.consulenti-ict.it
#14 lore raw

Re: [PATCHv5 06/17] pps: access pps device by direct pointer

В Thu, 16 Dec 2010 16:52:38 +0100
Rodolfo Giometti <[email protected]> пишет:

Quoted 8 lines · Alexander Gordeev
Quoted 5 lines

Using device index as a pointer needs some unnecessary work to be done every time the pointer is needed (in irq handler for example). Using a direct pointer is much more easy (and safe as well).

Thanks, Rodolfo!
I'll resend the patchset now with all your acks and other changes.

Signature · 1 line
  Alexander
#15 lore raw

PATCH v5 7/17 pps: convert printk/pr_* to dev_*

In reply to Alexander Gordeev

Since we now have direct pointers to struct pps_device everywhere it's easy to use dev_* functions to print messages instead of plain printks. Where dev_* cannot be used printks are converted to pr_*.

---
 drivers/pps/clients/pps-ktimer.c |    5 +++--
 drivers/pps/clients/pps-ldisc.c  |    2 ++
 drivers/pps/kapi.c               |   15 ++++++++-------
 drivers/pps/pps.c                |   25 +++++++++++++------------
 4 files changed, 26 insertions(+), 21 deletions(-)
drivers/pps/clients/pps-ktimer.c +3 -2
index 966ead1..2728469 100644
@@ -19,6 +19,7 @@
  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -45,7 +46,7 @@ static void pps_ktimer_event(unsigned long ptr)
 	/* First of all we get the time stamp... */
 	pps_get_ts(&ts);
 
-	pr_info("PPS event at %lu\n", jiffies);
+	dev_info(pps->dev, "PPS event at %lu\n", jiffies);
 
 	pps_event(pps, &ts, PPS_CAPTUREASSERT, NULL);
 
@@ -94,7 +95,7 @@ static int __init pps_ktimer_init(void)
 	pps = pps_register_source(&pps_ktimer_info,
 				PPS_CAPTUREASSERT | PPS_OFFSETASSERT);
 	if (pps == NULL) {
-		printk(KERN_ERR "cannot register ktimer source\n");
+		pr_err("cannot register PPS source\n");
 		return -ENOMEM;
 	}
 
drivers/pps/clients/pps-ldisc.c +2 -0
index f0ba65e..199a83b 100644
@@ -19,6 +19,8 @@
  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
 #include <linux/module.h>
 #include <linux/serial_core.h>
 #include <linux/tty.h>
drivers/pps/kapi.c +8 -7
index 98d4012..e8847c1 100644
@@ -19,6 +19,7 @@
  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -80,20 +81,20 @@ struct pps_device *pps_register_source(struct pps_source_info *info,
 
 	/* Sanity checks */
 	if ((info->mode & default_params) != default_params) {
-		printk(KERN_ERR "pps: %s: unsupported default parameters\n",
+		pr_err("%s: unsupported default parameters\n",
 					info->name);
 		err = -EINVAL;
 		goto pps_register_source_exit;
 	}
 	if ((info->mode & (PPS_ECHOASSERT | PPS_ECHOCLEAR)) != 0 &&
 			info->echo == NULL) {
-		printk(KERN_ERR "pps: %s: echo function is not defined\n",
+		pr_err("%s: echo function is not defined\n",
 					info->name);
 		err = -EINVAL;
 		goto pps_register_source_exit;
 	}
 	if ((info->mode & (PPS_TSFMT_TSPEC | PPS_TSFMT_NTPFP)) == 0) {
-		printk(KERN_ERR "pps: %s: unspecified time format\n",
+		pr_err("%s: unspecified time format\n",
 					info->name);
 		err = -EINVAL;
 		goto pps_register_source_exit;
@@ -138,7 +139,7 @@ struct pps_device *pps_register_source(struct pps_source_info *info,
 	if (id >= PPS_MAX_SOURCES) {
 		spin_unlock_irq(&pps_idr_lock);
 
-		printk(KERN_ERR "pps: %s: too many PPS sources in the system\n",
+		pr_err("%s: too many PPS sources in the system\n",
 					info->name);
 		err = -EBUSY;
 		goto free_idr;
@@ -150,12 +151,12 @@ struct pps_device *pps_register_source(struct pps_source_info *info,
 	/* Create the char device */
 	err = pps_register_cdev(pps);
 	if (err < 0) {
-		printk(KERN_ERR "pps: %s: unable to create char device\n",
+		pr_err("%s: unable to create char device\n",
 					info->name);
 		goto free_idr;
 	}
 
-	pr_info("new PPS source %s at ID %d\n", info->name, id);
+	dev_info(pps->dev, "new PPS source %s\n", info->name);
 
 	return pps;
 
@@ -168,7 +169,7 @@ kfree_pps:
 	kfree(pps);
 
 pps_register_source_exit:
-	printk(KERN_ERR "pps: %s: unable to register source\n", info->name);
+	pr_err("%s: unable to register source\n", info->name);
 
 	return NULL;
 }
drivers/pps/pps.c +13 -12
index 1922f27..9f7c2e8 100644
@@ -19,6 +19,7 @@
  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -67,7 +68,7 @@ static long pps_cdev_ioctl(struct file *file,
 
 	switch (cmd) {
 	case PPS_GETPARAMS:
-		pr_debug("PPS_GETPARAMS: source %d\n", pps->id);
+		dev_dbg(pps->dev, "PPS_GETPARAMS\n");
 
 		spin_lock_irq(&pps->lock);
 
@@ -83,7 +84,7 @@ static long pps_cdev_ioctl(struct file *file,
 		break;
 
 	case PPS_SETPARAMS:
-		pr_debug("PPS_SETPARAMS: source %d\n", pps->id);
+		dev_dbg(pps->dev, "PPS_SETPARAMS\n");
 
 		/* Check the capabilities */
 		if (!capable(CAP_SYS_TIME))
@@ -93,14 +94,14 @@ static long pps_cdev_ioctl(struct file *file,
 		if (err)
 			return -EFAULT;
 		if (!(params.mode & (PPS_CAPTUREASSERT | PPS_CAPTURECLEAR))) {
-			pr_debug("capture mode unspecified (%x)\n",
+			dev_dbg(pps->dev, "capture mode unspecified (%x)\n",
 								params.mode);
 			return -EINVAL;
 		}
 
 		/* Check for supported capabilities */
 		if ((params.mode & ~pps->info.mode) != 0) {
-			pr_debug("unsupported capabilities (%x)\n",
+			dev_dbg(pps->dev, "unsupported capabilities (%x)\n",
 								params.mode);
 			return -EINVAL;
 		}
@@ -113,7 +114,7 @@ static long pps_cdev_ioctl(struct file *file,
 		/* Restore the read only parameters */
 		if ((params.mode & (PPS_TSFMT_TSPEC | PPS_TSFMT_NTPFP)) == 0) {
 			/* section 3.3 of RFC 2783 interpreted */
-			pr_debug("time format unspecified (%x)\n",
+			dev_dbg(pps->dev, "time format unspecified (%x)\n",
 								params.mode);
 			pps->params.mode |= PPS_TSFMT_TSPEC;
 		}
@@ -126,7 +127,7 @@ static long pps_cdev_ioctl(struct file *file,
 		break;
 
 	case PPS_GETCAP:
-		pr_debug("PPS_GETCAP: source %d\n", pps->id);
+		dev_dbg(pps->dev, "PPS_GETCAP\n");
 
 		err = put_user(pps->info.mode, iuarg);
 		if (err)
@@ -138,7 +139,7 @@ static long pps_cdev_ioctl(struct file *file,
 		struct pps_fdata fdata;
 		unsigned int ev;
 
-		pr_debug("PPS_FETCH: source %d\n", pps->id);
+		dev_dbg(pps->dev, "PPS_FETCH\n");
 
 		err = copy_from_user(&fdata, uarg, sizeof(struct pps_fdata));
 		if (err)
@@ -153,7 +154,7 @@ static long pps_cdev_ioctl(struct file *file,
 		else {
 			unsigned long ticks;
 
-			pr_debug("timeout %lld.%09d\n",
+			dev_dbg(pps->dev, "timeout %lld.%09d\n",
 					(long long) fdata.timeout.sec,
 					fdata.timeout.nsec);
 			ticks = fdata.timeout.sec * HZ;
@@ -171,7 +172,7 @@ static long pps_cdev_ioctl(struct file *file,
 
 		/* Check for pending signals */
 		if (err == -ERESTARTSYS) {
-			pr_debug("pending signal caught\n");
+			dev_dbg(pps->dev, "pending signal caught\n");
 			return -EINTR;
 		}
 
@@ -240,7 +241,7 @@ int pps_register_cdev(struct pps_device *pps)
 
 	err = cdev_add(&pps->cdev, devt, 1);
 	if (err) {
-		printk(KERN_ERR "pps: %s: failed to add char device %d:%d\n",
+		pr_err("%s: failed to add char device %d:%d\n",
 				pps->info.name, MAJOR(pps_devt), pps->id);
 		return err;
 	}
@@ -282,14 +283,14 @@ static int __init pps_init(void)
 
 	pps_class = class_create(THIS_MODULE, "pps");
 	if (!pps_class) {
-		printk(KERN_ERR "pps: failed to allocate class\n");
+		pr_err("failed to allocate class\n");
 		return -ENOMEM;
 	}
 	pps_class->dev_attrs = pps_attrs;
 
 	err = alloc_chrdev_region(&pps_devt, 0, PPS_MAX_SOURCES, "pps");
 	if (err < 0) {
-		printk(KERN_ERR "pps: failed to allocate char device region\n");
+		pr_err("failed to allocate char device region\n");
 		goto remove_class;
 	}
 
Signature · 1 line
1.7.2.3
#16 lore raw

PATCH v5 8/17 pps: move idr stuff to pps.c

In reply to Alexander Gordeev

Since now idr is only used to manage char device id's and not used in kernel API anymore it should be moved to pps.c. This also makes it possible to release id only at actual device freeing so nobody can register a pps device with the same id while our device is not freed yet.

---
 drivers/pps/kapi.c |   56 ++-------------------------------------------------
 drivers/pps/pps.c  |   50 +++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 52 insertions(+), 54 deletions(-)
drivers/pps/kapi.c +3 -53
index e8847c1..c42d3cb 100644
@@ -27,19 +27,11 @@
 #include <linux/sched.h>
 #include <linux/time.h>
 #include <linux/spinlock.h>
-#include <linux/idr.h>
 #include <linux/fs.h>
 #include <linux/pps_kernel.h>
 #include <linux/slab.h>
 
 /*
- * Local variables
- */
-
-static DEFINE_SPINLOCK(pps_idr_lock);
-static DEFINE_IDR(pps_idr);
-
-/*
  * Local functions
  */
 
@@ -76,7 +68,6 @@ struct pps_device *pps_register_source(struct pps_source_info *info,
 		int default_params)
 {
 	struct pps_device *pps;
-	int id;
 	int err;
 
 	/* Sanity checks */
@@ -117,54 +108,18 @@ struct pps_device *pps_register_source(struct pps_source_info *info,
 	init_waitqueue_head(&pps->queue);
 	spin_lock_init(&pps->lock);
 
-	/* Get new ID for the new PPS source */
-	if (idr_pre_get(&pps_idr, GFP_KERNEL) == 0) {
-		err = -ENOMEM;
-		goto kfree_pps;
-	}
-
-	spin_lock_irq(&pps_idr_lock);
-
-	/* Now really allocate the PPS source.
-	 * After idr_get_new() calling the new source will be freely available
-	 * into the kernel.
-	 */
-	err = idr_get_new(&pps_idr, pps, &id);
-	if (err < 0) {
-		spin_unlock_irq(&pps_idr_lock);
-		goto kfree_pps;
-	}
-
-	id = id & MAX_ID_MASK;
-	if (id >= PPS_MAX_SOURCES) {
-		spin_unlock_irq(&pps_idr_lock);
-
-		pr_err("%s: too many PPS sources in the system\n",
-					info->name);
-		err = -EBUSY;
-		goto free_idr;
-	}
-	pps->id = id;
-
-	spin_unlock_irq(&pps_idr_lock);
-
 	/* Create the char device */
 	err = pps_register_cdev(pps);
 	if (err < 0) {
 		pr_err("%s: unable to create char device\n",
 					info->name);
-		goto free_idr;
+		goto kfree_pps;
 	}
 
 	dev_info(pps->dev, "new PPS source %s\n", info->name);
 
 	return pps;
 
-free_idr:
-	spin_lock_irq(&pps_idr_lock);
-	idr_remove(&pps_idr, id);
-	spin_unlock_irq(&pps_idr_lock);
-
 kfree_pps:
 	kfree(pps);
 
@@ -184,15 +139,10 @@ EXPORT_SYMBOL(pps_register_source);
 
 void pps_unregister_source(struct pps_device *pps)
 {
-	unsigned int id = pps->id;
-
 	pps_unregister_cdev(pps);
 
-	spin_lock_irq(&pps_idr_lock);
-	idr_remove(&pps_idr, pps->id);
-	spin_unlock_irq(&pps_idr_lock);
-
-	kfree(pps);
+	/* don't have to kfree(pps) here because it will be done on
+	 * device destruction */
 }
 EXPORT_SYMBOL(pps_unregister_source);
 
drivers/pps/pps.c +49 -1

91 lines not on this page · raw message

Signature · 1 line
1.7.2.3
maintainer Acked-by on v5 8/17
#17 lore raw

Re: [PATCHv5 08/17] pps: move idr stuff to pps.c

Quoted 7 lines · Alexander Gordeev

Since now idr is only used to manage char device id's and not used in kernel API anymore it should be moved to pps.c. This also makes it possible to release id only at actual device freeing so nobody can register a pps device with the same id while our device is not freed yet.

Signature · 5 lines
GNU/Linux Solutions                  e-mail: [email protected]
Linux Device Driver                          [email protected]
Embedded Systems                     phone:  +39 349 2432127
UNIX programming                     skype:  rodolfo.giometti
Freelance ICT Italia - Consulente ICT Italia - www.consulenti-ict.it
#18 lore raw

PATCH v5 9/17 pps: do not disable interrupts for idr operations

In reply to Alexander Gordeev

Now pps_idr_lock is never used in interrupt context so replace spin_lock_irq/spin_unlock_irq with plain spin_lock/spin_unlock.

---
 drivers/pps/pps.c |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)
drivers/pps/pps.c +6 -6
index 79b4455..54964a5 100644
@@ -239,9 +239,9 @@ static void pps_device_destruct(struct device *dev)
 
 	/* release id here to protect others from using it while it's
 	 * still in use */
-	spin_lock_irq(&pps_idr_lock);
+	spin_lock(&pps_idr_lock);
 	idr_remove(&pps_idr, pps->id);
-	spin_unlock_irq(&pps_idr_lock);
+	spin_unlock(&pps_idr_lock);
 
 	kfree(dev);
 	kfree(pps);
@@ -260,9 +260,9 @@ int pps_register_cdev(struct pps_device *pps)
 	 * After idr_get_new() calling the new source will be freely available
 	 * into the kernel.
 	 */
-	spin_lock_irq(&pps_idr_lock);
+	spin_lock(&pps_idr_lock);
 	err = idr_get_new(&pps_idr, pps, &pps->id);
-	spin_unlock_irq(&pps_idr_lock);
+	spin_unlock(&pps_idr_lock);
 
 	if (err < 0)
 		return err;
@@ -302,9 +302,9 @@ del_cdev:
 	cdev_del(&pps->cdev);
 
 free_idr:
-	spin_lock_irq(&pps_idr_lock);
+	spin_lock(&pps_idr_lock);
 	idr_remove(&pps_idr, pps->id);
-	spin_unlock_irq(&pps_idr_lock);
+	spin_unlock(&pps_idr_lock);
 
 	return err;
 }
Signature · 1 line
1.7.2.3
maintainer Acked-by on v5 9/17
#19 lore raw

Re: [PATCHv5 09/17] pps: do not disable interrupts for idr operations

Quoted 4 lines · Alexander Gordeev

Now pps_idr_lock is never used in interrupt context so replace spin_lock_irq/spin_unlock_irq with plain spin_lock/spin_unlock.

Signature · 5 lines
GNU/Linux Solutions                  e-mail: [email protected]
Linux Device Driver                          [email protected]
Embedded Systems                     phone:  +39 349 2432127
UNIX programming                     skype:  rodolfo.giometti
Freelance ICT Italia - Consulente ICT Italia - www.consulenti-ict.it
#20 lore raw

PATCH v5 10/17 pps: use BUG_ON for kernel API safety checks

In reply to Alexander Gordeev

This way less overhead is involved when running production kernel. If you want to debug a pps client module please define DEBUG to enable the checks.

---
 drivers/pps/kapi.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
drivers/pps/kapi.c +2 -4
index c42d3cb..3e8eb3f 100644
@@ -166,10 +166,8 @@ void pps_event(struct pps_device *pps, struct pps_event_time *ts, int event,
 	int captured = 0;
 	struct pps_ktime ts_real;
 
-	if ((event & (PPS_CAPTUREASSERT | PPS_CAPTURECLEAR)) == 0) {
-		dev_err(pps->dev, "unknown event (%x)\n", event);
-		return;
-	}
+	/* check event type */
+	BUG_ON((event & (PPS_CAPTUREASSERT | PPS_CAPTURECLEAR)) == 0);
 
 	dev_dbg(pps->dev, "PPS event at %ld.%09ld\n",
 			ts->ts_real.tv_sec, ts->ts_real.tv_nsec);
Signature · 1 line
1.7.2.3
#21 lore raw

PATCH v5 11/17 pps: simplify conditions a bit

In reply to Alexander Gordeev

Bitwise conjunction is distributive so we can simplify some conditions.

---
 drivers/pps/kapi.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
drivers/pps/kapi.c +2 -4
index 3e8eb3f..ebf3374 100644
@@ -182,8 +182,7 @@ void pps_event(struct pps_device *pps, struct pps_event_time *ts, int event,
 
 	/* Check the event */
 	pps->current_mode = pps->params.mode;
-	if ((event & PPS_CAPTUREASSERT) &
-			(pps->params.mode & PPS_CAPTUREASSERT)) {
+	if (event & pps->params.mode & PPS_CAPTUREASSERT) {
 		/* We have to add an offset? */
 		if (pps->params.mode & PPS_OFFSETASSERT)
 			pps_add_offset(&ts_real,
@@ -197,8 +196,7 @@ void pps_event(struct pps_device *pps, struct pps_event_time *ts, int event,
 
 		captured = ~0;
 	}
-	if ((event & PPS_CAPTURECLEAR) &
-			(pps->params.mode & PPS_CAPTURECLEAR)) {
+	if (event & pps->params.mode & PPS_CAPTURECLEAR) {
 		/* We have to add an offset? */
 		if (pps->params.mode & PPS_OFFSETCLEAR)
 			pps_add_offset(&ts_real,
Signature · 1 line
1.7.2.3
#22 lore raw

PATCH v5 12/17 pps: timestamp is always passed to dcd_change()

In reply to Alexander Gordeev

Remove the code that gatheres timestamp in pps_tty_dcd_change() in case passed ts parameter is NULL because it never happens in the current code. Fix comments as well.

---
 Documentation/serial/tty.txt    |    2 +-
 drivers/pps/clients/pps-ldisc.c |    8 --------
 include/linux/tty_ldisc.h       |    2 +-
 3 files changed, 2 insertions(+), 10 deletions(-)
Documentation/serial/tty.txt +1 -1
index 7c90050..540db41 100644
@@ -107,7 +107,7 @@ write_wakeup()	-	May be called at any point between open and close.
 
 dcd_change()	-	Report to the tty line the current DCD pin status
 			changes and the relative timestamp. The timestamp
-			can be NULL.
+			cannot be NULL.
 
 
 Driver Access
drivers/pps/clients/pps-ldisc.c +0 -8
index 199a83b..19b3477 100644
@@ -32,14 +32,6 @@ static void pps_tty_dcd_change(struct tty_struct *tty, unsigned int status,
 				struct pps_event_time *ts)
 {
 	struct pps_device *pps = (struct pps_device *)tty->disc_data;
-	struct pps_event_time __ts;
-
-	/* First of all we get the time stamp... */
-	pps_get_ts(&__ts);
-
-	/* Does caller give us a timestamp? */
-	if (!ts)	/* No. Do it ourself! */
-		ts = &__ts;
 
 	BUG_ON(pps == NULL);
 
include/linux/tty_ldisc.h +1 -1
index 763e061..ff7dc08 100644
@@ -104,7 +104,7 @@
  *			struct pps_event_time *ts)
  *
  *	Tells the discipline that the DCD pin has changed its status and
- *	the relative timestamp. Pointer ts can be NULL.
+ *	the relative timestamp. Pointer ts cannot be NULL.
  */
 
 #include <linux/fs.h>
Signature · 1 line
1.7.2.3
#23 lore raw

PATCH v5 13/17 ntp: add hardpps implementation

In reply to Alexander Gordeev

This commit adds hardpps() implementation based upon the original one from the NTPv4 reference kernel code from David Mills. However, it is highly optimized towards very fast syncronization and maximum stickness to PPS signal. The typical error is less then a microsecond. To make it sync faster I had to throw away exponential phase filter so that the full phase offset is corrected immediately. Then I also had to throw away median phase filter because it gives a bigger error itself if used without exponential filter.
Maybe we will find an appropriate filtering scheme in the future but it's not necessary if the signal quality is ok.

---
 drivers/pps/Kconfig   |    7 +
 include/linux/timex.h |    1 +
 kernel/time/ntp.c     |  425 +++++++++++++++++++++++++++++++++++++++++++++++--
 3 files changed, 418 insertions(+), 15 deletions(-)
drivers/pps/Kconfig +7 -0
index 1afe4e0..33c9126 100644
@@ -30,6 +30,13 @@ config PPS_DEBUG
 	  messages to the system log.  Select this if you are having a
 	  problem with PPS support and want to see more of what is going on.
 
+config NTP_PPS
+	bool "PPS kernel consumer support"
+	depends on PPS
+	help
+	  This option adds support for direct in-kernel time
+	  syncronization using an external PPS signal.
+
 source drivers/pps/clients/Kconfig
 
 endmenu
include/linux/timex.h +1 -0
index 32d852f..d23999f 100644
@@ -268,6 +268,7 @@ extern u64 tick_length;
 extern void second_overflow(void);
 extern void update_ntp_one_tick(void);
 extern int do_adjtimex(struct timex *);
+extern void hardpps(const struct timespec *, const struct timespec *);
 
 int read_current_timer(unsigned long *timer_val);
 
kernel/time/ntp.c +410 -15

484 lines not on this page · raw message

Signature · 1 line
1.7.2.3
#24 lore raw

PATCH v5 14/17 pps: capture MONOTONIC_RAW timestamps as well

In reply to Alexander Gordeev

MONOTONIC_RAW clock timestamps are ideally suited for frequency calculation and also fit well into the original NTP hardpps design. Now phase and frequency can be adjusted separately: the former based on REALTIME clock and the latter based on MONOTONIC_RAW clock. A new function getnstime_raw_and_real is added to timekeeping subsystem to capture both timestamps at the same time and atomically.

---
 include/linux/pps_kernel.h |    3 ++-
 include/linux/time.h       |    2 ++
 kernel/time/timekeeping.c  |   38 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 42 insertions(+), 1 deletions(-)
include/linux/pps_kernel.h +2 -1
index 1aedf50..6ae8898 100644
@@ -47,6 +47,7 @@ struct pps_source_info {
 };
 
 struct pps_event_time {
+	struct timespec ts_raw;
 	struct timespec ts_real;
 };
 
@@ -99,7 +100,7 @@ static inline void timespec_to_pps_ktime(struct pps_ktime *kt,
 
 static inline void pps_get_ts(struct pps_event_time *ts)
 {
-	getnstimeofday(&ts->ts_real);
+	getnstime_raw_and_real(&ts->ts_raw, &ts->ts_real);
 }
 
 #endif /* LINUX_PPS_KERNEL_H */
include/linux/time.h +2 -0
index 9f15ac7..1e6d3b5 100644
@@ -158,6 +158,8 @@ extern unsigned int alarm_setitimer(unsigned int seconds);
 extern int do_getitimer(int which, struct itimerval *value);
 extern void getnstimeofday(struct timespec *tv);
 extern void getrawmonotonic(struct timespec *ts);
+extern void getnstime_raw_and_real(struct timespec *ts_raw,
+		struct timespec *ts_real);
 extern void getboottime(struct timespec *ts);
 extern void monotonic_to_bootbased(struct timespec *ts);
 
kernel/time/timekeeping.c +38 -0
index 49010d8..947c7dc 100644
@@ -285,6 +285,44 @@ void ktime_get_ts(struct timespec *ts)
 EXPORT_SYMBOL_GPL(ktime_get_ts);
 
 /**
+ * getnstime_raw_and_real - Returns both the time of day an raw
+ * monotonic time in a timespec format
+ * @ts_mono_raw:	pointer to the timespec to be set to raw
+ *			monotonic time
+ * @ts_real:		pointer to the timespec to be set to the time
+ *			of day
+ */
+void getnstime_raw_and_real(struct timespec *ts_raw, struct timespec *ts_real)
+{
+	unsigned long seq;
+	s64 nsecs_raw, nsecs_real;
+
+	WARN_ON_ONCE(timekeeping_suspended);
+
+	do {
+		u32 arch_offset;
+
+		seq = read_seqbegin(&xtime_lock);
+
+		*ts_raw = raw_time;
+		*ts_real = xtime;
+
+		nsecs_raw = timekeeping_get_ns_raw();
+		nsecs_real = timekeeping_get_ns();
+
+		/* If arch requires, add in gettimeoffset() */
+		arch_offset = arch_gettimeoffset();
+		nsecs_raw += arch_offset;
+		nsecs_real += arch_offset;
+
+	} while (read_seqretry(&xtime_lock, seq));
+
+	timespec_add_ns(ts_raw, nsecs_raw);
+	timespec_add_ns(ts_real, nsecs_real);
+}
+EXPORT_SYMBOL(getnstime_raw_and_real);
+
+/**
  * do_gettimeofday - Returns the time of day in a timeval
  * @tv:		pointer to the timeval to be set
  *
Signature · 1 line
1.7.2.3
#25 lore raw

Re: [PATCHv5 14/17] pps: capture MONOTONIC_RAW timestamps as well

Quoted 3 lines · Alexander Gordeev

/**
+ * getnstime_raw_and_real - Returns both the time of day an raw
+ * monotonic time in a timespec format

IIRC then kerneldoc does not handle multiline comments for the function name

kernel/time/timekeeping.c void ktime_get_ts(struct timespec *ts) · line 300

Quoted 11 lines

+ * @ts_mono_raw: pointer to the timespec to be set to raw
+ * monotonic time
+ * @ts_real: pointer to the timespec to be set to the time
+ * of day
+ */

kernel/time/timekeeping.c +6 -0
+void getnstime_raw_and_real(struct timespec *ts_raw, struct timespec *ts_real)
+{
+	unsigned long seq;
+	s64 nsecs_raw, nsecs_real;
+
+	WARN_ON_ONCE(timekeeping_suspended);

Shouldn't this just return an error code instead of reading some stale or random value ?

Quoted 12 lines
kernel/time/timekeeping.c +12 -0
+	do {
+		u32 arch_offset;
+
+		seq = read_seqbegin(&xtime_lock);
+
+		*ts_raw = raw_time;
+		*ts_real = xtime;
+
+		nsecs_raw = timekeeping_get_ns_raw();
+		nsecs_real = timekeeping_get_ns();
+
+		/* If arch requires, add in gettimeoffset() */

kernel/time/timekeeping.c void ktime_get_ts(struct timespec *ts) · line 323

Quoted 10 lines
kernel/time/timekeeping.c +10 -0
+		arch_offset = arch_gettimeoffset();
+		nsecs_raw += arch_offset;
+		nsecs_real += arch_offset;
+
+	} while (read_seqretry(&xtime_lock, seq));
+
+	timespec_add_ns(ts_raw, nsecs_raw);
+	timespec_add_ns(ts_real, nsecs_real);
+}
+EXPORT_SYMBOL(getnstime_raw_and_real);

Otherwise, I'm fine with that.

Thanks,

tglx

#26 lore raw

Re: [PATCHv5 14/17] pps: capture MONOTONIC_RAW timestamps as well

В Wed, 24 Nov 2010 17:49:44 +0100 (CET)
Thomas Gleixner <[email protected]> пишет:

Quoted 7 lines · Alexander Gordeev
Quoted 3 lines

/**
+ * getnstime_raw_and_real - Returns both the time of day an raw
+ * monotonic time in a timespec format

IIRC then kerneldoc does not handle multiline comments for the function name

Ah, I see, fixed that.

Quoted 6 lines
Quoted 6 lines

+ * @ts_mono_raw: pointer to the timespec to be set to raw
+ * monotonic time
+ * @ts_real: pointer to the timespec to be set to the time
+ * of day
+ */
+void getnstime_raw_and_real(struct timespec *ts_raw, struct timespec *ts_real)

Quoted 8 lines

kernel/time/timekeeping.c void ktime_get_ts(struct timespec *ts) · line 300

Quoted 5 lines
kernel/time/timekeeping.c +5 -0
+{
+	unsigned long seq;
+	s64 nsecs_raw, nsecs_real;
+
+	WARN_ON_ONCE(timekeeping_suspended);

Shouldn't this just return an error code instead of reading some stale or random value ?

Sorry, I don't know if it should, just copied it from getnstimofday(). BTW what is the bad value for timespec or you mean returning int instead of void? I think this function should be handled the same way as getnstimeofday() and no one currently expects that the latter can return some bad value.

Quoted 28 lines
Quoted 12 lines
kernel/time/timekeeping.c +12 -0
+	do {
+		u32 arch_offset;
+
+		seq = read_seqbegin(&xtime_lock);
+
+		*ts_raw = raw_time;
+		*ts_real = xtime;
+
+		nsecs_raw = timekeeping_get_ns_raw();
+		nsecs_real = timekeeping_get_ns();
+
+		/* If arch requires, add in gettimeoffset() */

kernel/time/timekeeping.c void ktime_get_ts(struct timespec *ts) · line 323

Quoted 10 lines
kernel/time/timekeeping.c +10 -0
+		arch_offset = arch_gettimeoffset();
+		nsecs_raw += arch_offset;
+		nsecs_real += arch_offset;
+
+	} while (read_seqretry(&xtime_lock, seq));
+
+	timespec_add_ns(ts_raw, nsecs_raw);
+	timespec_add_ns(ts_real, nsecs_real);
+}
+EXPORT_SYMBOL(getnstime_raw_and_real);

Otherwise, I'm fine with that.

Thanks,

tglx

Signature · 1 line
  Alexander
#27 lore raw

PATCH v5 15/17 pps: add kernel consumer support

In reply to Alexander Gordeev

Add an optional feature of PPSAPI, kernel consumer support, which uses the added hardpps() function.

---
 Documentation/ioctl/ioctl-number.txt |    2 +-
 drivers/pps/Kconfig                  |    1 +
 drivers/pps/kapi.c                   |   27 +++++++++++++++
 drivers/pps/pps.c                    |   62 +++++++++++++++++++++++++++++++++-
 include/linux/pps.h                  |    7 ++++
 include/linux/pps_kernel.h           |    6 +++
 6 files changed, 103 insertions(+), 2 deletions(-)
Documentation/ioctl/ioctl-number.txt +1 -1
index 33223ff..de2b113 100644
@@ -249,7 +249,7 @@ Code  Seq#(hex)	Include File		Comments
 'p'	40-7F	linux/nvram.h
 'p'	80-9F	linux/ppdev.h		user-space parport
 					<mailto:[email protected]>
-'p'	A1-A4	linux/pps.h		LinuxPPS
+'p'	A1-A5	linux/pps.h		LinuxPPS
 					<mailto:[email protected]>
 'q'	00-1F	linux/serio.h
 'q'	80-FF	linux/telephony.h	Internet PhoneJACK, Internet LineJACK
drivers/pps/Kconfig +1 -0
index 33c9126..4823e47 100644
@@ -7,6 +7,7 @@ menu "PPS support"
 config PPS
 	tristate "PPS support"
 	depends on EXPERIMENTAL
+	select NTP_PPS
 	---help---
 	  PPS (Pulse Per Second) is a special pulse provided by some GPS
 	  antennae. Userland can use it to get a high-precision time
drivers/pps/kapi.c +27 -0
index ebf3374..ed78f6d 100644
@@ -26,12 +26,23 @@
 #include <linux/init.h>
 #include <linux/sched.h>
 #include <linux/time.h>
+#include <linux/timex.h>
 #include <linux/spinlock.h>
 #include <linux/fs.h>
 #include <linux/pps_kernel.h>
 #include <linux/slab.h>
 
 /*
+ * Global variables
+ */
+
+/* state variables to bind kernel consumer */
+/* PPS API (RFC 2783): current source and mode for ``kernel consumer'' */
+DEFINE_SPINLOCK(pps_kc_hardpps_lock);
+void	*pps_kc_hardpps_dev;	/* some unique pointer to device */
+int	pps_kc_hardpps_mode;	/* mode bits for kernel consumer */
+
+/*
  * Local functions
  */
 
@@ -139,6 +150,16 @@ EXPORT_SYMBOL(pps_register_source);
 
 void pps_unregister_source(struct pps_device *pps)
 {
+	spin_lock_irq(&pps_kc_hardpps_lock);
+	if (pps == pps_kc_hardpps_dev) {
+		pps_kc_hardpps_mode = 0;
+		pps_kc_hardpps_dev = NULL;
+		spin_unlock_irq(&pps_kc_hardpps_lock);
+		dev_info(pps->dev, "unbound kernel consumer"
+				" on device removal\n");
+	} else
+		spin_unlock_irq(&pps_kc_hardpps_lock);
+
 	pps_unregister_cdev(pps);
 
 	/* don't have to kfree(pps) here because it will be done on
@@ -211,6 +232,12 @@ void pps_event(struct pps_device *pps, struct pps_event_time *ts, int event,
 		captured = ~0;
 	}
 
+	/* Pass some events to kernel consumer if activated */
+	spin_lock(&pps_kc_hardpps_lock);
+	if (pps == pps_kc_hardpps_dev && event & pps_kc_hardpps_mode)
+		hardpps(&ts->ts_real, &ts->ts_raw);
+	spin_unlock(&pps_kc_hardpps_lock);
+
 	/* Wake up if captured something */
 	if (captured) {
 		pps->last_ev++;
drivers/pps/pps.c +61 -1

70 lines not on this page · raw message

include/linux/pps.h +7 -0
index 0194ab0..a9bb1d9 100644
@@ -114,11 +114,18 @@ struct pps_fdata {
 	struct pps_ktime timeout;
 };
 
+struct pps_bind_args {
+	int tsformat;	/* format of time stamps */
+	int edge;	/* selected event type */
+	int consumer;	/* selected kernel consumer */
+};
+
 #include <linux/ioctl.h>
 
 #define PPS_GETPARAMS		_IOR('p', 0xa1, struct pps_kparams *)
 #define PPS_SETPARAMS		_IOW('p', 0xa2, struct pps_kparams *)
 #define PPS_GETCAP		_IOR('p', 0xa3, int *)
 #define PPS_FETCH		_IOWR('p', 0xa4, struct pps_fdata *)
+#define PPS_KC_BIND		_IOW('p', 0xa5, struct pps_bind_args *)
 
 #endif /* _PPS_H_ */
include/linux/pps_kernel.h +6 -0
index 6ae8898..2302274 100644
@@ -79,6 +79,12 @@ struct pps_device {
 
 extern struct device_attribute pps_attrs[];
 
+/* state variables to bind kernel consumer */
+/* PPS API (RFC 2783): current source and mode for ``kernel consumer'' */
+extern spinlock_t pps_kc_hardpps_lock;
+extern void *pps_kc_hardpps_dev;	/* some unique pointer to device */
+extern int pps_kc_hardpps_mode;		/* mode bits for kernel consumer */
+
 /*
  * Exported functions
  */
Signature · 1 line
1.7.2.3
maintainer Acked-by on v5 15/17
#28 lore raw

Re: [PATCHv5 15/17] pps: add kernel consumer support

Quoted 4 lines · Alexander Gordeev

Add an optional feature of PPSAPI, kernel consumer support, which uses the added hardpps() function.

Signature · 5 lines
GNU/Linux Solutions                  e-mail: [email protected]
Linux Device Driver                          [email protected]
Embedded Systems                     phone:  +39 349 2432127
UNIX programming                     skype:  rodolfo.giometti
Freelance ICT Italia - Consulente ICT Italia - www.consulenti-ict.it
#29 lore raw

PATCH v5 16/17 pps: add parallel port PPS client

In reply to Alexander Gordeev

Add parallel port PPS client. It uses a standard method for capturing timestamps for assert edge transitions: getting a timestamp soon after an interrupt has happened. This is not a very precise source of time information due to interrupt handling delays. However, timestamps for clear edge transitions are much more precise because the interrupt handler continuously polls hardware port until the transition is done. Hardware port operations require only about 1us so the maximum error should not exceed this value. This was my primary goal when developing this client.
Clear edge capture could be disabled using clear_wait parameter.

---
 drivers/pps/clients/Kconfig       |    7 +
 drivers/pps/clients/Makefile      |    1 +
 drivers/pps/clients/pps_parport.c |  247 +++++++++++++++++++++++++++++++++++++
 3 files changed, 255 insertions(+), 0 deletions(-)
 create mode 100644 drivers/pps/clients/pps_parport.c
drivers/pps/clients/Kconfig +7 -0
index 4e801bd..8520a7f 100644
@@ -22,4 +22,11 @@ config PPS_CLIENT_LDISC
 	  If you say yes here you get support for a PPS source connected
 	  with the CD (Carrier Detect) pin of your serial port.
 
+config PPS_CLIENT_PARPORT
+	tristate "Parallel port PPS client"
+	depends on PPS && PARPORT
+	help
+	  If you say yes here you get support for a PPS source connected
+	  with the interrupt pin of your parallel port.
+
 endif
drivers/pps/clients/Makefile +1 -0
index 812c9b1..42517da 100644
@@ -4,6 +4,7 @@
 
 obj-$(CONFIG_PPS_CLIENT_KTIMER)	+= pps-ktimer.o
 obj-$(CONFIG_PPS_CLIENT_LDISC)	+= pps-ldisc.o
+obj-$(CONFIG_PPS_CLIENT_PARPORT) += pps_parport.o
 
 ifeq ($(CONFIG_PPS_DEBUG),y)
 EXTRA_CFLAGS += -DDEBUG
drivers/pps/clients/pps_parport.c new file +247 -0

247 lines not on this page · raw message

Signature · 1 line
1.7.2.3
maintainer Acked-by on v5 16/17
#30 lore raw

Re: [PATCHv5 16/17] pps: add parallel port PPS client

Quoted 11 lines · Alexander Gordeev

Add parallel port PPS client. It uses a standard method for capturing timestamps for assert edge transitions: getting a timestamp soon after an interrupt has happened. This is not a very precise source of time information due to interrupt handling delays. However, timestamps for clear edge transitions are much more precise because the interrupt handler continuously polls hardware port until the transition is done. Hardware port operations require only about 1us so the maximum error should not exceed this value. This was my primary goal when developing this client.
Clear edge capture could be disabled using clear_wait parameter.

Quoted 1 line
Signature · 5 lines
GNU/Linux Solutions                  e-mail: [email protected]
Linux Device Driver                          [email protected]
Embedded Systems                     phone:  +39 349 2432127
UNIX programming                     skype:  rodolfo.giometti
Freelance ICT Italia - Consulente ICT Italia - www.consulenti-ict.it
#31 lore raw

PATCH v5 17/17 pps: add parallel port PPS signal generator

In reply to Alexander Gordeev

Add PPS signal generator which utilizes STROBE pin of a parallel port to send PPS signals. It uses parport abstraction layer and hrtimers to precisely control the signal.

---
 Documentation/pps/pps.txt                |   46 +++++
 drivers/pps/Kconfig                      |    2 +
 drivers/pps/Makefile                     |    2 +-
 drivers/pps/generators/Kconfig           |   17 ++
 drivers/pps/generators/Makefile          |    9 +
 drivers/pps/generators/pps_gen_parport.c |  275 ++++++++++++++++++++++++++++++
 6 files changed, 350 insertions(+), 1 deletions(-)
 create mode 100644 drivers/pps/generators/Kconfig
 create mode 100644 drivers/pps/generators/Makefile
 create mode 100644 drivers/pps/generators/pps_gen_parport.c
Documentation/pps/pps.txt +46 -0
index 125f4ab..d35dcdd 100644
@@ -170,3 +170,49 @@ and the run ppstest as follow:
 
 Please, note that to compile userland programs you need the file timepps.h
 (see Documentation/pps/).
+
+
+Generators
+----------
+
+Sometimes one needs to be able not only to catch PPS signals but to produce
+them also. For example, running a distributed simulation, which requires
+computers' clock to be synchronized very tightly. One way to do this is to
+invent some complicated hardware solutions but it may be neither necessary
+nor affordable. The cheap way is to load a PPS generator on one of the
+computers (master) and PPS clients on others (slaves), and use very simple
+cables to deliver signals using parallel ports, for example.
+
+Parallel port cable pinout:
+pin	name	master      slave
+1	STROBE	  *------     *
+2	D0	  *     |     *
+3	D1	  *     |     *
+4	D2	  *     |     *
+5	D3	  *     |     *
+6	D4	  *     |     *
+7	D5	  *     |     *
+8	D6	  *     |     *
+9	D7	  *     |     *
+10	ACK	  *     ------*
+11	BUSY	  *           *
+12	PE	  *           *
+13	SEL	  *           *
+14	AUTOFD	  *           *
+15	ERROR	  *           *
+16	INIT	  *           *
+17	SELIN	  *           *
+18-25	GND	  *-----------*
+
+Please note that parallel port interrupt occurs only on high->low transition,
+so it is used for PPS assert edge. PPS clear edge can be determined only
+using polling in the interrupt handler which actually can be done way more
+precisely because interrupt handling delays can be quite big and random. So
+current parport PPS generator implementation (pps_gen_parport module) is
+geared towards using the clear edge for time synchronization.
+
+Clear edge polling is done with disabled interrupts so it's better to select
+delay between assert and clear edge as small as possible to reduce system
+latencies. But if it is too small slave won't be able to capture clear edge
+transition. The default of 30us should be good enough in most situations.
+The delay can be selected using 'delay' pps_gen_parport module parameter.
drivers/pps/Kconfig +2 -0
index 4823e47..79bda60 100644
@@ -40,4 +40,6 @@ config NTP_PPS
 
 source drivers/pps/clients/Kconfig
 
+source drivers/pps/generators/Kconfig
+
 endmenu
drivers/pps/Makefile +1 -1
index 98960dd..1906eb70 100644
@@ -4,6 +4,6 @@
 
 pps_core-y			:= pps.o kapi.o sysfs.o
 obj-$(CONFIG_PPS)		:= pps_core.o
-obj-y				+= clients/
+obj-y				+= clients/ generators/
 
 ccflags-$(CONFIG_PPS_DEBUG) := -DDEBUG
drivers/pps/generators/Kconfig new file +17 -0
new file mode 100644
index 0000000..5fbd614
@@ -0,0 +1,17 @@
+#
+# PPS generators configuration
+#
+
+if PPS
+
+comment "PPS generators support"
+
+config PPS_GENERATOR_PARPORT
+	tristate "Parallel port PPS signal generator"
+	depends on PARPORT != n && GENERIC_TIME
+	help
+	  If you say yes here you get support for a PPS signal generator which
+	  utilizes STROBE pin of a parallel port to send PPS signals. It uses
+	  parport abstraction layer and hrtimers to precisely control the signal.
+
+endif
drivers/pps/generators/Makefile new file +9 -0
new file mode 100644
index 0000000..303304a
@@ -0,0 +1,9 @@
+#
+# Makefile for PPS generators.
+#
+
+obj-$(CONFIG_PPS_GENERATOR_PARPORT) += pps_gen_parport.o
+
+ifeq ($(CONFIG_PPS_DEBUG),y)
+EXTRA_CFLAGS += -DDEBUG
+endif
drivers/pps/generators/pps_gen_parport.c new file +275 -0

275 lines not on this page · raw message

Signature · 1 line
1.7.2.3
maintainer Acked-by on v5 17/17
#32 lore raw

Re: [PATCHv5 17/17] pps: add parallel port PPS signal generator

Quoted 5 lines · Alexander Gordeev

Add PPS signal generator which utilizes STROBE pin of a parallel port to send PPS signals. It uses parport abstraction layer and hrtimers to precisely control the signal.

Signature · 5 lines
GNU/Linux Solutions                  e-mail: [email protected]
Linux Device Driver                          [email protected]
Embedded Systems                     phone:  +39 349 2432127
UNIX programming                     skype:  rodolfo.giometti
Freelance ICT Italia - Consulente ICT Italia - www.consulenti-ict.it