Skip to contentSkip to conversation

3.2-stable timekeeping fixes merged in 3.6

PATCH In review John Stultz 3 participants 8 messages Sep 11, 2012 – Sep 27, 2012 Timekeeping, clocksource core, NTP, alarmtimer

lore
Share & tools
lore
#0 lore raw

cover 3.2-stable timekeeping fixes merged in 3.6

Just wanted to send out a few timekeeping fixes that were merged in 3.6 which are appropriate for -stable.

This queue backports the following fixes:
-----------------------------------------

cee58483cf56e0ba355fdd97ff5e8925329aa936    time: Move ktime_t overflow checking into timespec_valid_strict
bf2ac312195155511a0f79325515cbb61929898a    time: Avoid making adjustments if we haven't accumulated anything
4e8b14526ca7fb046a81c94002c1c43b6fdf0e9b    time: Improve sanity checking of timekeeping inputs

I've run these through my timetest suite w/ kvm on both i386 & x86_64. But more testing would be of course appreciated.
https://github.com/johnstultz-work/timetests

I also have patch queues for all the -stable trees that I'll be sending out as my testing completes for those trees.

Cc Prarit Bhargava <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Linux Kernel <[email protected]>

John Stultz (3):
3.2.y: time: Improve sanity checking of timekeeping inputs 3.2.y: time: Avoid making adjustments if we haven't accumulated
anything
3.2.y: time: Move ktime_t overflow checking into
timespec_valid_strict

 include/linux/ktime.h     |    7 -------
 include/linux/time.h      |   29 +++++++++++++++++++++++++++--
 kernel/time/timekeeping.c |   32 +++++++++++++++++++++++++++++---
 3 files changed, 56 insertions(+), 12 deletions(-)
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH 1/3 3.2.y: time: Improve sanity checking of timekeeping inputs

This is a -stable backport of 4e8b14526ca7fb046a81c94002c1c43b6fdf0e9b

Unexpected behavior could occur if the time is set to a value large enough to overflow a 64bit ktime_t (which is something larger then the year 2262).

Also unexpected behavior could occur if large negative offsets are injected via adjtimex.

So this patch improves the sanity check timekeeping inputs by improving the timespec_valid() check, and then makes better use of timespec_valid() to make sure we don't set the time to an invalid negative value or one that overflows ktime_t.

Note: This does not protect from setting the time close to overflowing ktime_t and then letting natural accumulation cause the overflow.

---
 include/linux/ktime.h     |    7 -------
 include/linux/time.h      |   22 ++++++++++++++++++++--
 kernel/time/timekeeping.c |   26 ++++++++++++++++++++++++--
 3 files changed, 44 insertions(+), 11 deletions(-)
include/linux/ktime.h +0 -7
index 603bec2..06177ba10 100644
@@ -58,13 +58,6 @@ union ktime {
 
 typedef union ktime ktime_t;		/* Kill this */
 
-#define KTIME_MAX			((s64)~((u64)1 << 63))
-#if (BITS_PER_LONG == 64)
-# define KTIME_SEC_MAX			(KTIME_MAX / NSEC_PER_SEC)
-#else
-# define KTIME_SEC_MAX			LONG_MAX
-#endif
-
 /*
  * ktime_t definitions when using the 64-bit scalar representation:
  */
include/linux/time.h +20 -2
index b306178..c8f7233 100644
@@ -107,11 +107,29 @@ static inline struct timespec timespec_sub(struct timespec lhs,
 	return ts_delta;
 }
 
+#define KTIME_MAX			((s64)~((u64)1 << 63))
+#if (BITS_PER_LONG == 64)
+# define KTIME_SEC_MAX			(KTIME_MAX / NSEC_PER_SEC)
+#else
+# define KTIME_SEC_MAX			LONG_MAX
+#endif
+
 /*
  * Returns true if the timespec is norm, false if denorm:
  */
-#define timespec_valid(ts) \
-	(((ts)->tv_sec >= 0) && (((unsigned long) (ts)->tv_nsec) < NSEC_PER_SEC))
+static inline bool timespec_valid(const struct timespec *ts)
+{
+	/* Dates before 1970 are bogus */
+	if (ts->tv_sec < 0)
+		return false;
+	/* Can't have more nanoseconds then a second */
+	if ((unsigned long)ts->tv_nsec >= NSEC_PER_SEC)
+		return false;
+	/* Disallow values that could overflow ktime_t */
+	if ((unsigned long long)ts->tv_sec >= KTIME_SEC_MAX)
+		return false;
+	return true;
+}
 
 extern void read_persistent_clock(struct timespec *ts);
 extern void read_boot_clock(struct timespec *ts);
kernel/time/timekeeping.c +24 -2
index 03e67d4..36440c88 100644
@@ -382,7 +382,7 @@ int do_settimeofday(const struct timespec *tv)
 	struct timespec ts_delta;
 	unsigned long flags;
 
-	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
+	if (!timespec_valid(tv))
 		return -EINVAL;
 
 	write_seqlock_irqsave(&xtime_lock, flags);
@@ -417,6 +417,8 @@ EXPORT_SYMBOL(do_settimeofday);
 int timekeeping_inject_offset(struct timespec *ts)
 {
 	unsigned long flags;
+	struct timespec tmp;
+	int ret = 0;
 
 	if ((unsigned long)ts->tv_nsec >= NSEC_PER_SEC)
 		return -EINVAL;
@@ -425,9 +427,16 @@ int timekeeping_inject_offset(struct timespec *ts)
 
 	timekeeping_forward_now();
 
+	tmp = timespec_add(xtime,  *ts);
+	if (!timespec_valid(&tmp)) {
+		ret = -EINVAL;
+		goto error;
+	}
+
 	xtime = timespec_add(xtime, *ts);
 	wall_to_monotonic = timespec_sub(wall_to_monotonic, *ts);
 
+error: /* even if we error out, we forwarded the time, so call update */
 	timekeeping_update(true);
 
 	write_sequnlock_irqrestore(&xtime_lock, flags);
@@ -435,7 +444,7 @@ int timekeeping_inject_offset(struct timespec *ts)
 	/* signal hrtimers about time change */
 	clock_was_set();
 
-	return 0;
+	return ret;
 }
 EXPORT_SYMBOL(timekeeping_inject_offset);
 
@@ -582,7 +591,20 @@ void __init timekeeping_init(void)
 	struct timespec now, boot;
 
 	read_persistent_clock(&now);
+	if (!timespec_valid(&now)) {
+		pr_warn("WARNING: Persistent clock returned invalid value!\n"
+			"         Check your CMOS/BIOS settings.\n");
+		now.tv_sec = 0;
+		now.tv_nsec = 0;
+	}
+
 	read_boot_clock(&boot);
+	if (!timespec_valid(&boot)) {
+		pr_warn("WARNING: Boot clock returned invalid value!\n"
+			"         Check your CMOS/BIOS settings.\n");
+		boot.tv_sec = 0;
+		boot.tv_nsec = 0;
+	}
 
 	write_seqlock_irqsave(&xtime_lock, flags);
 
Signature · 1 line
1.7.9.5
#2 lore raw

PATCH 2/3 3.2.y: time: Avoid making adjustments if we haven't accumulated anything

In reply to John Stultz

This is a -stable backport of bf2ac312195155511a0f79325515cbb61929898a

If update_wall_time() is called and the current offset isn't large enough to accumulate, avoid re-calling timekeeping_adjust which may change the clock freq and can cause 1ns inconsistencies with CLOCK_REALTIME_COARSE/CLOCK_MONOTONIC_COARSE.

---
 kernel/time/timekeeping.c |    4 ++++
 1 file changed, 4 insertions(+)
kernel/time/timekeeping.c +4 -0
index 36440c88..59544dc 100644
@@ -1033,6 +1033,10 @@ static void update_wall_time(void)
 #else
 	offset = (clock->read(clock) - clock->cycle_last) & clock->mask;
 #endif
+	/* Check if there's really nothing to do */
+	if (offset < timekeeper.cycle_interval)
+		return;
+
 	timekeeper.xtime_nsec = (s64)xtime.tv_nsec << timekeeper.shift;
 
 	/*
Signature · 1 line
1.7.9.5
#3 lore raw

PATCH 3/3 3.2.y: time: Move ktime_t overflow checking into timespec_valid_strict

In reply to John Stultz

This is a -stable backport of cee58483cf56e0ba355fdd97ff5e8925329aa936

Andreas Bombe reported that the added ktime_t overflow checking added to timespec_valid in commit 4e8b14526ca7 ("time: Improve sanity checking of timekeeping inputs") was causing problems with X.org because it caused timeouts larger then KTIME_T to be invalid.

Previously, these large timeouts would be clamped to KTIME_MAX and would never expire, which is valid.

This patch splits the ktime_t overflow checking into a new timespec_valid_strict function, and converts the timekeeping codes internal checking to use this more strict function.

Reported-and-tested-by Andreas Bombe <[email protected]>
Cc Zhouping Liu <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Prarit Bhargava <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Linux Kernel <[email protected]>
---
 include/linux/time.h      |    7 +++++++
 kernel/time/timekeeping.c |   10 +++++-----
 2 files changed, 12 insertions(+), 5 deletions(-)
include/linux/time.h +7 -0
index c8f7233..8c0216e 100644
@@ -125,6 +125,13 @@ static inline bool timespec_valid(const struct timespec *ts)
 	/* Can't have more nanoseconds then a second */
 	if ((unsigned long)ts->tv_nsec >= NSEC_PER_SEC)
 		return false;
+	return true;
+}
+
+static inline bool timespec_valid_strict(const struct timespec *ts)
+{
+	if (!timespec_valid(ts))
+		return false;
 	/* Disallow values that could overflow ktime_t */
 	if ((unsigned long long)ts->tv_sec >= KTIME_SEC_MAX)
 		return false;
kernel/time/timekeeping.c +5 -5
index 59544dc..5ee1ac0 100644
@@ -382,7 +382,7 @@ int do_settimeofday(const struct timespec *tv)
 	struct timespec ts_delta;
 	unsigned long flags;
 
-	if (!timespec_valid(tv))
+	if (!timespec_valid_strict(tv))
 		return -EINVAL;
 
 	write_seqlock_irqsave(&xtime_lock, flags);
@@ -428,7 +428,7 @@ int timekeeping_inject_offset(struct timespec *ts)
 	timekeeping_forward_now();
 
 	tmp = timespec_add(xtime,  *ts);
-	if (!timespec_valid(&tmp)) {
+	if (!timespec_valid_strict(&tmp)) {
 		ret = -EINVAL;
 		goto error;
 	}
@@ -591,7 +591,7 @@ void __init timekeeping_init(void)
 	struct timespec now, boot;
 
 	read_persistent_clock(&now);
-	if (!timespec_valid(&now)) {
+	if (!timespec_valid_strict(&now)) {
 		pr_warn("WARNING: Persistent clock returned invalid value!\n"
 			"         Check your CMOS/BIOS settings.\n");
 		now.tv_sec = 0;
@@ -599,7 +599,7 @@ void __init timekeeping_init(void)
 	}
 
 	read_boot_clock(&boot);
-	if (!timespec_valid(&boot)) {
+	if (!timespec_valid_strict(&boot)) {
 		pr_warn("WARNING: Boot clock returned invalid value!\n"
 			"         Check your CMOS/BIOS settings.\n");
 		boot.tv_sec = 0;
@@ -649,7 +649,7 @@ static void update_sleep_time(struct timespec t)
  */
 static void __timekeeping_inject_sleeptime(struct timespec *delta)
 {
-	if (!timespec_valid(delta)) {
+	if (!timespec_valid_strict(delta)) {
 		printk(KERN_WARNING "__timekeeping_inject_sleeptime: Invalid "
 					"sleep delta value!\n");
 		return;
Signature · 1 line
1.7.9.5
#4 lore raw
In reply to John Stultz
Quoted 8 lines · John Stultz

Just wanted to send out a few timekeeping fixes that were merged in 3.6 which are appropriate for -stable.

This queue backports the following fixes:
-----------------------------------------

cee58483cf56e0ba355fdd97ff5e8925329aa936    time: Move ktime_t overflow checking into timespec_valid_strict
bf2ac312195155511a0f79325515cbb61929898a    time: Avoid making adjustments if we haven't accumulated anything
4e8b14526ca7fb046a81c94002c1c43b6fdf0e9b    time: Improve sanity checking of timekeeping inputs

[...]

All queued up for 3.2, thanks.

Ben.

Signature · 2 lines
Ben Hutchings
Experience is what causes a person to make new mistakes instead of old ones.
#5 lore raw
In reply to John Stultz
Quoted 9 lines · John Stultz

Just wanted to send out a few timekeeping fixes that were merged in 3.6 which are appropriate for -stable.

This queue backports the following fixes:
-----------------------------------------

cee58483cf56e0ba355fdd97ff5e8925329aa936    time: Move ktime_t overflow checking into timespec_valid_strict
bf2ac312195155511a0f79325515cbb61929898a    time: Avoid making adjustments if we haven't accumulated anything
4e8b14526ca7fb046a81c94002c1c43b6fdf0e9b    time: Improve sanity checking of timekeeping inputs
Quoted 6 lines

I've run these through my timetest suite w/ kvm on both i386 & x86_64. But more testing would be of course appreciated.
https://github.com/johnstultz-work/timetests

I also have patch queues for all the -stable trees that I'll be sending out as my testing completes for those trees.

Did you also send these out for the 3.0.y tree and I just missed them? Should I just take this 3.2.y series and see how they well apply?

thanks,

greg k-h

#6 lore raw
Quoted 10 lines · Greg KH
Quoted 9 lines · John Stultz

Just wanted to send out a few timekeeping fixes that were merged in 3.6 which are appropriate for -stable.

This queue backports the following fixes:
-----------------------------------------

cee58483cf56e0ba355fdd97ff5e8925329aa936    time: Move ktime_t overflow checking into timespec_valid_strict
bf2ac312195155511a0f79325515cbb61929898a    time: Avoid making adjustments if we haven't accumulated anything
4e8b14526ca7fb046a81c94002c1c43b6fdf0e9b    time: Improve sanity checking of timekeeping inputs
Quoted 8 lines
Quoted 6 lines

I've run these through my timetest suite w/ kvm on both i386 & x86_64. But more testing would be of course appreciated.
https://github.com/johnstultz-work/timetests

I also have patch queues for all the -stable trees that I'll be sending out as my testing completes for those trees.

Did you also send these out for the 3.0.y tree and I just missed them?

I missed them, sorry for the noise, just found them...

greg k-h

#7 lore raw
In reply to Greg KH
Quoted 10 lines · Greg KH
Quoted 9 lines · John Stultz

Just wanted to send out a few timekeeping fixes that were merged in 3.6 which are appropriate for -stable.

This queue backports the following fixes:
-----------------------------------------

cee58483cf56e0ba355fdd97ff5e8925329aa936    time: Move ktime_t overflow checking into timespec_valid_strict
bf2ac312195155511a0f79325515cbb61929898a    time: Avoid making adjustments if we haven't accumulated anything
4e8b14526ca7fb046a81c94002c1c43b6fdf0e9b    time: Improve sanity checking of timekeeping inputs
Quoted 8 lines
Quoted 6 lines

I've run these through my timetest suite w/ kvm on both i386 & x86_64. But more testing would be of course appreciated.
https://github.com/johnstultz-work/timetests

I also have patch queues for all the -stable trees that I'll be sending out as my testing completes for those trees.

Did you also send these out for the 3.0.y tree and I just missed them? Should I just take this 3.2.y series and see how they well apply?

I did send them out for 3.0, on Sep 11.

Let me know if you can't find them and I'll resend.

thanks
-john