time: Fix timeekeping_get_ns overflow on 32bit systems
authorJohn Stultz <john.stultz@linaro.org>
Tue, 11 Sep 2012 23:26:03 +0000 (19:26 -0400)
committerIngo Molnar <mingo@kernel.org>
Thu, 13 Sep 2012 15:39:14 +0000 (17:39 +0200)
commitec145babe754f9ea1079034a108104b6001e001c
tree2b65fdc6916989cacdc2bbcc94a48ef847eae099
parent0bd1189e239c76eb3a50e458548fbe7e4a5dfff1
time: Fix timeekeping_get_ns overflow on 32bit systems

Daniel Lezcano reported seeing multi-second stalls from
keyboard input on his T61 laptop when NOHZ and CPU_IDLE
were enabled on a 32bit kernel.

He bisected the problem down to commit
1e75fa8be9fb6 ("time: Condense timekeeper.xtime into xtime_sec").

After reproducing this issue, I narrowed the problem down
to the fact that timekeeping_get_ns() returns a 64bit
nsec value that hasn't been accumulated. In some cases
this value was being then stored in timespec.tv_nsec
(which is a long).

On 32bit systems, with idle times larger then 4 seconds
(or less, depending on the value of xtime_nsec), the
returned nsec value would overflow 32bits. This limited
kept time from increasing, causing timers to not expire.

The fix is to make sure we don't directly store the
result of timekeeping_get_ns() into a tv_nsec field,
instead using a 64bit nsec value which can then be
added into the timespec via timespec_add_ns().

Reported-and-bisected-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Tested-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Acked-by: Prarit Bhargava <prarit@redhat.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Link: http://lkml.kernel.org/r/1347405963-35715-1-git-send-email-john.stultz@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/time/timekeeping.c