aboutsummaryrefslogtreecommitdiffstats
path: root/kernel/time
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-12-22 17:10:37 -0500
committerLinus Torvalds <torvalds@linux-foundation.org>2009-12-22 17:10:37 -0500
commit83f57a11d84460dfe2afdb5a8bc759953428e38b (patch)
treeb8506a6b61eb23e3d203fb248776c7eab9270c2d /kernel/time
parent0e2c8b8f55072a98b99e7bdad55c912084d6a526 (diff)
Revert "time: Remove xtime_cache"
This reverts commit 7bc7d637452383d56ba4368d4336b0dde1bb476d, as requested by John Stultz. Quoting John: "Petr Titěra reported an issue where he saw odd atime regressions with 2.6.33 where there were a full second worth of nanoseconds in the nanoseconds field. He also reviewed the time code and narrowed down the problem: unhandled overflow of the nanosecond field caused by rounding up the sub-nanosecond accumulated time. Details: * At the end of update_wall_time(), we currently round up the sub-nanosecond portion of accumulated time when storing it into xtime. This was added to avoid time inconsistencies caused when the sub-nanosecond portion was truncated when storing into xtime. Unfortunately we don't handle the possible second overflow caused by that rounding. * Previously the xtime_cache code hid this overflow by normalizing the xtime value when storing into the xtime_cache. * We could try to handle the second overflow after the rounding up, but since this affects the timekeeping's internal state, this would further complicate the next accumulation cycle, causing small errors in ntp steering. As much as I'd like to get rid of it, the xtime_cache code is known to work. * The correct fix is really to include the sub-nanosecond portion in the timekeeping accessor function, so we don't need to round up at during accumulation. This would greatly simplify the accumulation code. Unfortunately, we can't do this safely until the last three non-GENERIC_TIME arches (sparc32, arm, cris) are converted (those patches are in -mm) and we kill off the spots where arches set xtime directly. This is all 2.6.34 material, so I think reverting the xtime_cache change is the best approach for now. Many thanks to Petr for both reporting and finding the issue!" Reported-by: Petr Titěra <P.Titera@century.cz> Requested-by: john stultz <johnstul@us.ibm.com> Cc: Ingo Molnar <mingo@elte.hu> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'kernel/time')
-rw-r--r--kernel/time/timekeeping.c27
1 files changed, 23 insertions, 4 deletions
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index af4135f05825..7faaa32fbf4f 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -165,6 +165,13 @@ struct timespec raw_time;
165/* flag for if timekeeping is suspended */ 165/* flag for if timekeeping is suspended */
166int __read_mostly timekeeping_suspended; 166int __read_mostly timekeeping_suspended;
167 167
168static struct timespec xtime_cache __attribute__ ((aligned (16)));
169void update_xtime_cache(u64 nsec)
170{
171 xtime_cache = xtime;
172 timespec_add_ns(&xtime_cache, nsec);
173}
174
168/* must hold xtime_lock */ 175/* must hold xtime_lock */
169void timekeeping_leap_insert(int leapsecond) 176void timekeeping_leap_insert(int leapsecond)
170{ 177{
@@ -325,6 +332,8 @@ int do_settimeofday(struct timespec *tv)
325 332
326 xtime = *tv; 333 xtime = *tv;
327 334
335 update_xtime_cache(0);
336
328 timekeeper.ntp_error = 0; 337 timekeeper.ntp_error = 0;
329 ntp_clear(); 338 ntp_clear();
330 339
@@ -550,6 +559,7 @@ void __init timekeeping_init(void)
550 } 559 }
551 set_normalized_timespec(&wall_to_monotonic, 560 set_normalized_timespec(&wall_to_monotonic,
552 -boot.tv_sec, -boot.tv_nsec); 561 -boot.tv_sec, -boot.tv_nsec);
562 update_xtime_cache(0);
553 total_sleep_time.tv_sec = 0; 563 total_sleep_time.tv_sec = 0;
554 total_sleep_time.tv_nsec = 0; 564 total_sleep_time.tv_nsec = 0;
555 write_sequnlock_irqrestore(&xtime_lock, flags); 565 write_sequnlock_irqrestore(&xtime_lock, flags);
@@ -583,6 +593,7 @@ static int timekeeping_resume(struct sys_device *dev)
583 wall_to_monotonic = timespec_sub(wall_to_monotonic, ts); 593 wall_to_monotonic = timespec_sub(wall_to_monotonic, ts);
584 total_sleep_time = timespec_add_safe(total_sleep_time, ts); 594 total_sleep_time = timespec_add_safe(total_sleep_time, ts);
585 } 595 }
596 update_xtime_cache(0);
586 /* re-base the last cycle value */ 597 /* re-base the last cycle value */
587 timekeeper.clock->cycle_last = timekeeper.clock->read(timekeeper.clock); 598 timekeeper.clock->cycle_last = timekeeper.clock->read(timekeeper.clock);
588 timekeeper.ntp_error = 0; 599 timekeeper.ntp_error = 0;
@@ -722,6 +733,7 @@ static void timekeeping_adjust(s64 offset)
722 timekeeper.ntp_error_shift; 733 timekeeper.ntp_error_shift;
723} 734}
724 735
736
725/** 737/**
726 * logarithmic_accumulation - shifted accumulation of cycles 738 * logarithmic_accumulation - shifted accumulation of cycles
727 * 739 *
@@ -765,6 +777,7 @@ static cycle_t logarithmic_accumulation(cycle_t offset, int shift)
765 return offset; 777 return offset;
766} 778}
767 779
780
768/** 781/**
769 * update_wall_time - Uses the current clocksource to increment the wall time 782 * update_wall_time - Uses the current clocksource to increment the wall time
770 * 783 *
@@ -774,6 +787,7 @@ void update_wall_time(void)
774{ 787{
775 struct clocksource *clock; 788 struct clocksource *clock;
776 cycle_t offset; 789 cycle_t offset;
790 u64 nsecs;
777 int shift = 0, maxshift; 791 int shift = 0, maxshift;
778 792
779 /* Make sure we're fully resumed: */ 793 /* Make sure we're fully resumed: */
@@ -839,6 +853,9 @@ void update_wall_time(void)
839 timekeeper.ntp_error += timekeeper.xtime_nsec << 853 timekeeper.ntp_error += timekeeper.xtime_nsec <<
840 timekeeper.ntp_error_shift; 854 timekeeper.ntp_error_shift;
841 855
856 nsecs = clocksource_cyc2ns(offset, timekeeper.mult, timekeeper.shift);
857 update_xtime_cache(nsecs);
858
842 /* check to see if there is a new clocksource to use */ 859 /* check to see if there is a new clocksource to use */
843 update_vsyscall(&xtime, timekeeper.clock, timekeeper.mult); 860 update_vsyscall(&xtime, timekeeper.clock, timekeeper.mult);
844} 861}
@@ -875,13 +892,13 @@ void monotonic_to_bootbased(struct timespec *ts)
875 892
876unsigned long get_seconds(void) 893unsigned long get_seconds(void)
877{ 894{
878 return xtime.tv_sec; 895 return xtime_cache.tv_sec;
879} 896}
880EXPORT_SYMBOL(get_seconds); 897EXPORT_SYMBOL(get_seconds);
881 898
882struct timespec __current_kernel_time(void) 899struct timespec __current_kernel_time(void)
883{ 900{
884 return xtime; 901 return xtime_cache;
885} 902}
886 903
887struct timespec current_kernel_time(void) 904struct timespec current_kernel_time(void)
@@ -891,7 +908,8 @@ struct timespec current_kernel_time(void)
891 908
892 do { 909 do {
893 seq = read_seqbegin(&xtime_lock); 910 seq = read_seqbegin(&xtime_lock);
894 now = xtime; 911
912 now = xtime_cache;
895 } while (read_seqretry(&xtime_lock, seq)); 913 } while (read_seqretry(&xtime_lock, seq));
896 914
897 return now; 915 return now;
@@ -905,7 +923,8 @@ struct timespec get_monotonic_coarse(void)
905 923
906 do { 924 do {
907 seq = read_seqbegin(&xtime_lock); 925 seq = read_seqbegin(&xtime_lock);
908 now = xtime; 926
927 now = xtime_cache;
909 mono = wall_to_monotonic; 928 mono = wall_to_monotonic;
910 } while (read_seqretry(&xtime_lock, seq)); 929 } while (read_seqretry(&xtime_lock, seq));
911 930