Since commit <31ade30692dc9680bfc95700d794818fa3f754ac>, timekeeping_init()
checks for presence of persistent clock by attempting to read a non-zero
time value from real-time clock. This is an issue on platforms where
persistent_clock (instead of a RTC) is implemented as a free-running counter
starting from zero on each boot and running during suspend. Examples are some
ARM platforms (e.g. PandaBoard). An attempt to read such a clock during
timekeeping_init() may return zero value and falsely declare persistent clock
as missing. Additionally, in the above case suspend times may be accounted
twice (once from timekeeping_resume() and once from rtc_resume()), resulting
in a gradual drift of system time.
This patch does a run-time correction of the issue by doing the same check
during timekeeping_suspend().
A better long-term solution would have to return error when trying to read
non-existing clock and zero when trying to read an uninitialized clock, but
that would require changing all persistent_clock implementations.
This patch addresses the immediate breakage, for now.
Cc: John Stultz <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: [email protected]
Signed-off-by: Zoran Markovic <[email protected]>
---
kernel/time/timekeeping.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index 98cd470..baeeb5c 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -975,6 +975,14 @@ static int timekeeping_suspend(void)
read_persistent_clock(&timekeeping_suspend_time);
+ /*
+ * On some systems the persistent_clock can not be detected at
+ * timekeeping_init by its return value, so if we see a valid
+ * value returned, update the persistent_clock_exists flag.
+ */
+ if (timekeeping_suspend_time.tv_sec || timekeeping_suspend_time.tv_nsec)
+ persistent_clock_exist = true;
+
raw_spin_lock_irqsave(&timekeeper_lock, flags);
write_seqcount_begin(&timekeeper_seq);
timekeeping_forward_now(tk);
--
1.7.9.5
On 05/17/2013 11:24 AM, Zoran Markovic wrote:
> Since commit <31ade30692dc9680bfc95700d794818fa3f754ac>, timekeeping_init()
> checks for presence of persistent clock by attempting to read a non-zero
> time value from real-time clock. This is an issue on platforms where
> persistent_clock (instead of a RTC) is implemented as a free-running counter
> starting from zero on each boot and running during suspend. Examples are some
> ARM platforms (e.g. PandaBoard). An attempt to read such a clock during
> timekeeping_init() may return zero value and falsely declare persistent clock
> as missing. Additionally, in the above case suspend times may be accounted
> twice (once from timekeeping_resume() and once from rtc_resume()), resulting
> in a gradual drift of system time.
>
> This patch does a run-time correction of the issue by doing the same check
> during timekeeping_suspend().
>
> A better long-term solution would have to return error when trying to read
> non-existing clock and zero when trying to read an uninitialized clock, but
> that would require changing all persistent_clock implementations.
>
> This patch addresses the immediate breakage, for now.
>
> Cc: John Stultz <[email protected]>
> Cc: Thomas Gleixner <[email protected]>
> Cc: Feng Tang <[email protected]>
> Cc: [email protected]
> Signed-off-by: Zoran Markovic <[email protected]>
Thanks for finding and sending this out! I'll queue this for 3.10.
thanks again
-john
Hi Zoran,
Thanks for the patch! (This reply may be toooo late :))
One question just for curiosity: for the counter_32K timer, it's running
at 32K Hz and has one 32b counter. I understand it is only for suspend
time calculation use, but the wrap time for it is about
4G/32K ~= 128K seconds ~= 35 hours
What if one suspend time is longer than that?
- Feng
On Fri, May 17, 2013 at 11:24:05AM -0700, Zoran Markovic wrote:
> Since commit <31ade30692dc9680bfc95700d794818fa3f754ac>, timekeeping_init()
> checks for presence of persistent clock by attempting to read a non-zero
> time value from real-time clock. This is an issue on platforms where
> persistent_clock (instead of a RTC) is implemented as a free-running counter
> starting from zero on each boot and running during suspend. Examples are some
> ARM platforms (e.g. PandaBoard). An attempt to read such a clock during
> timekeeping_init() may return zero value and falsely declare persistent clock
> as missing. Additionally, in the above case suspend times may be accounted
> twice (once from timekeeping_resume() and once from rtc_resume()), resulting
> in a gradual drift of system time.
>
> This patch does a run-time correction of the issue by doing the same check
> during timekeeping_suspend().
>
> A better long-term solution would have to return error when trying to read
> non-existing clock and zero when trying to read an uninitialized clock, but
> that would require changing all persistent_clock implementations.
>
> This patch addresses the immediate breakage, for now.
>
> Cc: John Stultz <[email protected]>
> Cc: Thomas Gleixner <[email protected]>
> Cc: Feng Tang <[email protected]>
> Cc: [email protected]
> Signed-off-by: Zoran Markovic <[email protected]>
> ---
> kernel/time/timekeeping.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
> index 98cd470..baeeb5c 100644
> --- a/kernel/time/timekeeping.c
> +++ b/kernel/time/timekeeping.c
> @@ -975,6 +975,14 @@ static int timekeeping_suspend(void)
>
> read_persistent_clock(&timekeeping_suspend_time);
>
> + /*
> + * On some systems the persistent_clock can not be detected at
> + * timekeeping_init by its return value, so if we see a valid
> + * value returned, update the persistent_clock_exists flag.
> + */
> + if (timekeeping_suspend_time.tv_sec || timekeeping_suspend_time.tv_nsec)
> + persistent_clock_exist = true;
> +
> raw_spin_lock_irqsave(&timekeeper_lock, flags);
> write_seqcount_begin(&timekeeper_seq);
> timekeeping_forward_now(tk);
> --
> 1.7.9.5
Hi Feng,
Looking at the OMAP implementation, persistent_clock is updated on
every read of the 32K counter. If the read doesn't happen often enough
to accurately update persistent_clock, then the 32K counter would fail
the definition of a persistent clock and some other timekeeping source
should be used.
Regards, Zoran
On 12 October 2013 00:48, Feng Tang <[email protected]> wrote:
> Hi Zoran,
>
> Thanks for the patch! (This reply may be toooo late :))
>
> One question just for curiosity: for the counter_32K timer, it's running
> at 32K Hz and has one 32b counter. I understand it is only for suspend
> time calculation use, but the wrap time for it is about
> 4G/32K ~= 128K seconds ~= 35 hours
> What if one suspend time is longer than that?
>
> - Feng
>
> On Fri, May 17, 2013 at 11:24:05AM -0700, Zoran Markovic wrote:
>> Since commit <31ade30692dc9680bfc95700d794818fa3f754ac>, timekeeping_init()
>> checks for presence of persistent clock by attempting to read a non-zero
>> time value from real-time clock. This is an issue on platforms where
>> persistent_clock (instead of a RTC) is implemented as a free-running counter
>> starting from zero on each boot and running during suspend. Examples are some
>> ARM platforms (e.g. PandaBoard). An attempt to read such a clock during
>> timekeeping_init() may return zero value and falsely declare persistent clock
>> as missing. Additionally, in the above case suspend times may be accounted
>> twice (once from timekeeping_resume() and once from rtc_resume()), resulting
>> in a gradual drift of system time.
>>
>> This patch does a run-time correction of the issue by doing the same check
>> during timekeeping_suspend().
>>
>> A better long-term solution would have to return error when trying to read
>> non-existing clock and zero when trying to read an uninitialized clock, but
>> that would require changing all persistent_clock implementations.
>>
>> This patch addresses the immediate breakage, for now.
>>
>> Cc: John Stultz <[email protected]>
>> Cc: Thomas Gleixner <[email protected]>
>> Cc: Feng Tang <[email protected]>
>> Cc: [email protected]
>> Signed-off-by: Zoran Markovic <[email protected]>
>> ---
>> kernel/time/timekeeping.c | 8 ++++++++
>> 1 file changed, 8 insertions(+)
>>
>> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
>> index 98cd470..baeeb5c 100644
>> --- a/kernel/time/timekeeping.c
>> +++ b/kernel/time/timekeeping.c
>> @@ -975,6 +975,14 @@ static int timekeeping_suspend(void)
>>
>> read_persistent_clock(&timekeeping_suspend_time);
>>
>> + /*
>> + * On some systems the persistent_clock can not be detected at
>> + * timekeeping_init by its return value, so if we see a valid
>> + * value returned, update the persistent_clock_exists flag.
>> + */
>> + if (timekeeping_suspend_time.tv_sec || timekeeping_suspend_time.tv_nsec)
>> + persistent_clock_exist = true;
>> +
>> raw_spin_lock_irqsave(&timekeeper_lock, flags);
>> write_seqcount_begin(&timekeeper_seq);
>> timekeeping_forward_now(tk);
>> --
>> 1.7.9.5