Early 4.3 versions of gcc apparently aggressively optimize the raw
time accumulation loop, replacing it with a divide.
On 32bit systems, this causes the following link errors:
undefined reference to `__umoddi3'
undefined reference to `__udivdi3'
The gcc issue has been fixed in 4.4 and greater.
This patch replaces the accumulation loop with a do_div, as suggested
by Linus.
Signed-off-by: John Stultz <[email protected]>
CC: Jason Wessel <[email protected]>
CC: Larry Finger <[email protected]>
CC: Ingo Molnar <[email protected]>
CC: Linus Torvalds <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
To properly handle clocksources that change frequencies
at the clocksource->enable() point, this patch adds
a method that will update the clocksource's mult/shift and
max_idle_ns values.
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This converts the most common of the x86 clocksources over to use
clocksource_register_hz/khz.
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch makes xtime and wall_to_monotonic static, as planned in
Documentation/feature-removal-schedule.txt. This will allow for
further cleanups to the timekeeping core.
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Provides an accessor function to replace hrtimer.c's
direct access of wall_to_monotonic.
This will allow wall_to_monotonic to be made static as
planned in Documentation/feature-removal-schedule.txt
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the um arch to use read_persistent_clock().
This allows it to avoid accessing xtime and wall_to_monotonic
directly.
Signed-off-by: John Stultz <[email protected]>
Cc: Jeff Dike <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
update_vsyscall() did not provide the wall_to_monotoinc offset,
so arch specific implementations tend to reference wall_to_monotonic
directly. This limits future cleanups in the timekeeping core, so
this patch fixes the update_vsyscall interface to provide
wall_to_monotonic, allowing wall_to_monotonic to be made static
as planned in Documentation/feature-removal-schedule.txt
Signed-off-by: John Stultz <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: Anton Blanchard <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Tony Luck <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Currently powerpc's update_vsyscall calls an inline update_gtod.
However, both are straightforward, and there are no other users,
so this patch merges update_gtod into update_vsyscall.
Signed-off-by: John Stultz <[email protected]>
Cc: Anton Blanchard <[email protected]>
Cc: Paul Mackerras <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Now that all arches have been converted over to use generic time via
clocksources or arch_gettimeoffset(), we can remove the GENERIC_TIME
config option and simplify the generic code.
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
After accidentally misusing timespec_add_safe, I wanted to make sure
we don't accidently trip over that issue again, so I created a simple
timespec_add() function which we can use to replace the instances
of timespec_add_safe() that don't want the overflow detection.
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Due to vtime calling vgettimeofday(), its possible that an application
could call time();create("stuff",O_RDRW); only to see the file's
creation timestamp to be before the value returned by time.
A similar way to reproduce the issue is to compare the vsyscall time()
with the syscall time(), and observe ordering issues.
The modified test case from Oleg Nesterov below can illustrate this:
int main(void)
{
time_t sec1,sec2;
do {
sec1 = time(&sec2);
sec2 = syscall(__NR_time, NULL);
} while (sec1 <= sec2);
printf("vtime: %d.000000\n", sec1);
printf("time: %d.000000\n", sec2);
return 0;
}
The proper fix is to make vtime use the same time value as
current_kernel_time() (which is exported via update_vsyscall) instead of
vgettime().
Thanks to Jiri Olsa for bringing up the issue and catching bugs in
earlier verisons of this fix.
Signed-off-by: John Stultz <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Oleg Nesterov <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Alpha has a tsc like rpcc counter that it uses to manage time.
This can be converted to an actual clocksource instead of utilizing
the arch_gettimeoffset method that is really only there for legacy
systems with no continuous counter.
Further cleanups could be made if alpha converted to the clockevent
model.
CC: Thomas Gleixner <[email protected]>
CC: Richard Henderson <[email protected]>
Acked-by: Ivan Kokshaysky <[email protected]>
Tested-by: Ivan Kokshaysky <[email protected]>
Signed-off-by: Matt Turner <[email protected]>
Signed-off-by: John Stultz <[email protected]>
How to pick good mult/shift pairs has always been difficult to
describe to folks writing clocksource drivers, since it requires
careful tradeoffs in adjustment accuracy vs overflow limits.
Now, with the clocks_calc_mult_shift function, its much
easier. However, not many clocksources have converted to using that
function, and there is still the issue of the max interval length
assumption being made by each clocksource driver independently.
So this patch simplifies the registration process by having
clocksources be registered with a hz/khz value and the registration
function taking care of setting mult/shift.
This should take most of the confusion out of writing a clocksource
driver.
Additionally it also keeps the shift size tradeoff (more accuracy vs
longer possible nohz times) centralized so the timekeeping core can
keep track of the assumptions being made.
[ tglx: Coding style and comments fixed ]
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
With the earlier logarithmic time accumulation patch, xtime will now
always be within one "tick" of the current time, instead of possibly
half a second off.
This removes the need for the xtime_cache value, which always stored the
time at the last interrupt, so this patch cleans that up removing the
xtime_cache related code.
This patch also addresses an issue with an earlier version of this change,
where xtime_cache was normalizing xtime, which could in some cases be
not valid (ie: tv_nsec == NSEC_PER_SEC). This is fixed by handling
the edge case in update_wall_time().
Signed-off-by: John Stultz <[email protected]>
Cc: Petr Titěra <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Convert arm to use GENERIC_TIME via the arch_getoffset() infrastructure,
reducing the amount of arch specific code we need to maintain.
The arm architecture is the last arch that need to be converted.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Russell King <[email protected]>
Cc: Andrew Morton <[email protected]>
Now that no arches are accessing time_adjust directly,
make it static.
Signed-off-by: John Stultz <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
The logarithmic accumulation done in the timekeeping has some overflow
protection that limits the max shift value. That means it will take
more then shift loops to accumulate all of the cycles. This causes
the shift decrement to underflow, which causes the loop to never exit.
The simplest fix would be simply to do a:
if (shift)
shift--;
However that is not optimal, as we know the cycle offset is larger
then the interval << shift, the above would make shift drop to zero,
then we would be spinning for quite awhile accumulating at interval
chunks at a time.
Instead, this patch only decreases shift if the offset is smaller
then cycle_interval << shift. This makes sure we accumulate using
the largest chunks possible without overflowing tick_length, and limits
the number of iterations through the loop.
This issue was found and reported by Sonic Zhang, who also tested the fix.
Many thanks your explanation and testing!
Reported-by: Sonic Zhang <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Tested-by: Sonic Zhang <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the sparc architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
[ davem: compile fix: Here's a version that compiles, you have to get
rid of the now unused variably last_rtc_update since we build with
-Werror ]
Signed-off-by: John Stultz <[email protected]>
Acked-by: David S. Miller <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: [email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the cris architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Mikael Starvik <[email protected]>
Cc: Jesper Nilsson <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the m68k architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Geert Uytterhoeven <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the m32r architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Hirokazu Takata <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the blackfin architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Mike Frysinger <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the ia64 architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Tony Luck <[email protected]>
Cc: Fenghua Yu <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the avr32 architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Haavard Skinnemoen <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the h8300 architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: Yoshinori Sato <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the frv architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: David Howells <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the mn10300 architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Cc: David Howells <[email protected]>
Cc: Koichi Yasutake <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This patch converts the alpha architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
igned-off-by: John Stultz <[email protected]>
Cc: Richard Henderson <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
xtensa supports read_persisitent_clock(), so there is no reason for
the arch specific code to be setting xtime. Setting xtime will be
done by the generic timekeeping code.
Signed-off-by: John Stultz <[email protected]>
Cc: Chris Zankel <[email protected]>
Cc: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
They aren't really features, but fair warning to out of tree
driver folks who might be accessing these variables.
Signed-off-by: John Stultz <[email protected]>
CC: Ingo Molnar <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
warp_clock() currently accesses timekeeping internal state directly, which
is unnecessary. Convert it to use the proper timekeeping interfaces.
Signed-off-by: John Stultz <[email protected]>
Cc: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
This replaces direct xtime usage in the s390 arch with timekeeping accessors,
so we can further clean up the timekeeping core.
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
This patch converts the parisc architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
I have not built or tested this patch, so help from arch maintainers
would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Helge Deller <[email protected]>
Signed-off-by: Kyle McMartin <[email protected]>
Convert cris to use GENERIC_TIME via the arch_getoffset() infrastructure,
reducing the amount of arch specific code we need to maintain.
Signed-off-by: John Stultz <[email protected]>
Cc: Mikael Starvik <[email protected]>
Cc: Jesper Nilsson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This patch converts the sh architecture to use the generic
read_persistent_clock and update_persistent_clock interfaces, reducing
the amount of arch specific code we have to maintain, and allowing for
further cleanups in the future.
Cc: Thomas Gleixner <[email protected]>
Cc: Andrew Morton <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Paul Mundt <[email protected]>
Aaro Koskinen reported an issue in kernel.org bugzilla #15366, where
on non-GENERIC_TIME systems, accessing
/sys/devices/system/clocksource/clocksource0/current_clocksource
results in an oops.
It seems the timekeeper/clocksource rework missed initializing the
curr_clocksource value in the !GENERIC_TIME case.
Thanks to Aaro for reporting and diagnosing the issue as well as
testing the fix!
Reported-by: Aaro Koskinen <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: [email protected]
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
ntp.c doesn't need to access timekeeping internals directly, so change
xtime references to use the get_seconds() timekeeping interface.
Signed-off-by: John Stultz <[email protected]>
Cc: [email protected]
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
The m68knommu coldfire pit clocksource looks like it was incorrectly
marked as a continuous clocksource. Running with it marked as a
continuous clocksource could cause hangs when the system switches to
highres mode or enables nohz.
This patch removes the CLOCK_SOURCE_IS_CONTINUOUS flag on the coldfire pit
clocksource. This will disallow systems using this clocksource from
entering oneshot mode (disabling highres timers and nohz).
Signed-off-by: John Stultz <[email protected]>
Acked-by: Greg Ungerer <[email protected]>
Cc: Steven King <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This patch converts sparc (specifically sparc32) to use GENERIC_TIME via
the arch_getoffset() infrastructure, reducing the amount of arch
specific code we need to maintain.
The sparc architecture is one of the last 3 arches that need to be
converted.
This patch applies on top of Linus' current -git tree
I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Setting a thread's comm to be something unique is a very useful ability
and is helpful for debugging complicated threaded applications. However
currently the only way to set a thread name is for the thread to name
itself via the PR_SET_NAME prctl.
However, there may be situations where it would be advantageous for a
thread dispatcher to be naming the threads its managing, rather then
having the threads self-describe themselves. This sort of behavior is
available on other systems via the pthread_setname_np() interface.
This patch exports a task's comm via proc/pid/comm and
proc/pid/task/tid/comm interfaces, and allows thread siblings to write to
these values.
[[email protected]: cleanups]
Signed-off-by: John Stultz <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Arjan van de Ven <[email protected]>
Cc: Mike Fulton <[email protected]>
Cc: Sean Foley <[email protected]>
Cc: Darren Hart <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Convert Blackfin to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.
I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Mike Frysinger <[email protected]>
With the prior logarithmic time accumulation patch, xtime will now
always be within one "tick" of the current time, instead of
possibly half a second off.
This removes the need for the xtime_cache value, which always
stored the time at the last interrupt, so this patch cleans that up
removing the xtime_cache related code.
This is a bit simpler, but still could use some wider testing.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Reviewed-by: John Kacur <[email protected]>
Cc: Clark Williams <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Accumulating one tick at a time works well unless we're using NOHZ.
Then it can be an issue, since we may have to run through the loop
a few thousand times, which can increase timer interrupt caused
latency.
The current solution was to accumulate in half-second intervals
with NOHZ. This kept the number of loops down, however it did
slightly change how we make NTP adjustments. While not an issue
with NTPd users, as NTPd makes adjustments over a longer period of
time, other adjtimex() users have noticed the half-second
granularity with which we can apply frequency changes to the clock.
For instance, if a application tries to apply a 100ppm frequency
correction for 20ms to correct a 2us offset, with NOHZ they either
get no correction, or a 50us correction.
Now, there will always be some granularity error for applying
frequency corrections. However with users sensitive to this error
have seen a 50-500x increase with NOHZ compared to running without
NOHZ.
So I figured I'd try another approach then just simply increasing
the interval. My approach is to consume the time interval
logarithmically. This reduces the number of times through the loop
needed keeping latency down, while still preserving the original
granularity error for adjtimex() changes.
Further, this change allows us to remove the xtime_cache code
(patch to follow), as xtime is always within one tick of the
current time, instead of the half-second updates it saw before.
An earlier version of this patch has been shipping to x86 users in
the RedHat MRG releases for awhile without issue, but I've reworked
this version to be even more careful about avoiding possible
overflows if the shift value gets too large.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Reviewed-by: John Kacur <[email protected]>
Cc: Clark Williams <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Convert m68k to use GENERIC_TIME via the arch_getoffset() infrastructure,
reducing the amount of arch specific code we need to maintain.
I've taken my best swing at converting this, but I'm not 100% confident
I got it right. My cross-compiler is now out of date (gcc4.2) so I
wasn't able to check if it compiled. Any assistance from arch
maintainers or testers to get this merged would be great.
Signed-off-by: John Stultz <[email protected]>
Cc: Geert Uytterhoeven <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Convert m32r to use GENERIC_TIME via the arch_getoffset() infrastructure,
reducing the amount of arch specific code we need to maintain.
I also noted that m32r doesn't seem to be taking the xtime write lock
before calling do_timer()! That looks like a pretty bad bug to me. If
folks agree, let me know and I can move the lock grab to the correct spot.
Signed-off-by: John Stultz <[email protected]>
Cc: Hirokazu Takata <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Converts alpha to use GENERIC_TIME via the arch_getoffset()
infrastructure, reducing the amount of arch specific code we need to
maintain.
I suspect the alpha arch could even be further improved to provide and
rpcc() based clocksource, but not having the hardware, I don't feel
comfortable attempting the more complicated conversion (but I'd be glad to
help if anyone else is interested).
[[email protected]: fix build]
Signed-off-by: John Stultz <[email protected]>
Cc: Richard Henderson <[email protected]>
Cc: Ivan Kokshaysky <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
MOD_NANO, ADJ_NANO, MOD_NANO, ADJ_NANO!
Lets call the whole thing off!
But oh! If we call the whole thing off,
Then we must part.
And oh! If we ever part,
Then that might break my heart^H^H^H^Hclock!
So, if you like MOD_NANO and I like ADJ_NANO,
I'll include MOD_NANO and give up ADJ_NANO (not really!).
For we know we need each other,
So we better call the calling off off.
Let's call the whole thing off!
The tumultuous NTP and Linux relationship has hit another snag: Ends
up NTPd still uses the "xntp 3.4 compatability names" and when the
STA_NANO value was added (along with ADJ_NANO), NTPd expected MOD_NANO
to be added and has apparently hit some build errors.
Report to ntp hackers:
https://lists.ntp.org/pipermail/hackers/2009-August/004455.html
Related Bugs:
https://support.ntp.org/bugs/show_bug.cgi?id=1219https://bugzilla.redhat.com/show_bug.cgi?id=505566
So in an effort to make peace, here's a patch to help get things
building again. I also have updated the comment to make sure folks
don't think the MOD_* values are just legacy constants.
Of course, NTPd really uses the glibc-headers, so those will need to
be similarly updated before things are working again (the RH bug above
should probably cover that).
Thanks to Michael Tatarinov and Hal Murray for finding and reporting
the issue!
Signed-off-by: John Stultz <[email protected]>
Cc: Miroslav Lichvar <[email protected]>
Cc: [email protected]
Cc: Ulrich Drepper <[email protected]>
Cc: Michael Tatarinov <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
After talking with some application writers who want very fast, but not
fine-grained timestamps, I decided to try to implement new clock_ids
to clock_gettime(): CLOCK_REALTIME_COARSE and CLOCK_MONOTONIC_COARSE
which returns the time at the last tick. This is very fast as we don't
have to access any hardware (which can be very painful if you're using
something like the acpi_pm clocksource), and we can even use the vdso
clock_gettime() method to avoid the syscall. The only trade off is you
only get low-res tick grained time resolution.
This isn't a new idea, I know Ingo has a patch in the -rt tree that made
the vsyscall gettimeofday() return coarse grained time when the
vsyscall64 sysctrl was set to 2. However this affects all applications
on a system.
With this method, applications can choose the proper speed/granularity
trade-off for themselves.
Signed-off-by: John Stultz <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: [email protected]
Cc: Darren Hart <[email protected]>
Cc: [email protected]
Cc: [email protected]
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Move the adjustment of xtime, wall_to_monotonic and the update of the
vsyscall variables to the timekeeping code.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Thomas, Andrew and Ingo pointed out that we don't have any safety checks
in the clocksource sysfs entries to make sure sysadmins don't try to
change the clocksource to a non high-res timer capable clocksource (such
as jiffies) when high-res timers (HRT) is enabled. Doing so will likely
hang a system.
Correct this by filtering non HRT clocksources from available_clocksources
and not accepting non HRT clocksources with HRT enabled.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Bernhard Schiffner noticed I had a few comment typos in this patch,
(note: to save embarrassment, when making typos, avoid copying and
pasting them) so this patch corrects them.
[ Impact: cleanup ]
Reported-by: Bernhard Schiffner <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Cc: [email protected]
Cc: [email protected]
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
The conversion to the ntpv4 reference model
f199239373 ("ntp: convert to the NTP4
reference model") in 2.6.19 added nanosecond resolution the adjtimex
interface, but also changed the "stiffness" of the frequency adjustments,
causing NTP convergence time to greatly increase.
SHIFT_PLL, which reduces the stiffness of the freq adjustments, was
designed to be inversely linked to HZ, and the reference value of 4 was
designed for Unix systems using HZ=100. However Linux's clock steering
code mostly independent of HZ.
So this patch reduces the SHIFT_PLL value from 4 to 2, which causes NTPd
behavior to match kernels prior to 2.6.19, greatly reducing convergence
times, and improving close synchronization through environmental thermal
changes.
The patch also changes some l's to L's in nearby code to avoid misreading
50l as 501.
[ Impact: tweak NTP algorithm for faster convergence ]
Signed-off-by: John Stultz <[email protected]>
Acked-by: Rik van Riel <[email protected]>
Cc: [email protected]
Signed-off-by: Andrew Morton <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Convert sh to use GENERIC_TIME via the arch_getoffset() infrastructure.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Paul Mundt <[email protected]>
Some arches don't supply their own clocksource. This is mainly the
case in architectures that get their inter-tick times by reading the
counter on their interval timer. Since these timers wrap every tick,
they're not really useful as clocksources. Wrapping them to act like
one is possible but not very efficient. So we provide a callout these
arches can implement for use with the jiffies clocksource to provide
finer then tick granular time.
[ Impact: ease the migration to generic time keeping ]
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
tick_handle_periodic() can lock up hard when a one shot clock event
device is used in combination with jiffies clocksource.
Avoid an endless loop issue by requiring that a highres valid
clocksource be installed before we call tick_periodic() in a loop when
using ONESHOT mode. The result is we will only increment jiffies once
per interrupt until a continuous hardware clocksource is available.
Without this, we can run into a endless loop, where each cycle through
the loop, jiffies is updated which increments time by tick_period or
more (due to clock steering), which can cause the event programming to
think the next event was before the newly incremented time and fail
causing tick_periodic() to be called again and the whole process loops
forever.
[ Impact: prevent hard lock up ]
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: [email protected]
The time_status conditional was accidentally placed right after we clear
the checked time_status bits, which causes us to take the conditional
every time through. This fixes it by moving the conditional to before we
clear the time_status bits.
Signed-off-by: John Stultz <[email protected]>
Cc: Clark Williams <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Since the GENERIC_TIME changes landed, the adjtimex behavior changed
for struct timex.tick and .freq changed. When the tick or freq value
is set, we adjust the tick_length_base in ntp_update_frequency().
However, this new value doesn't get applied to tick_length until the
next second (via second_overflow).
This means some applications that do quick time tweaking do not see the
requested change made as quickly as expected.
I've run a few tests with this change, and ntpd still functions fine.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Between 2.6.23 and 2.6.24-rc1 a change was made that broke IBM LS21
systems that had the HPET enabled in the BIOS, resulting in boot hangs
for x86_64.
Specifically commit b8ce335906, which
merges the i386 and x86_64 HPET code.
Prior to this commit, when we setup the HPET timers in x86_64, we did
the following:
hpet_writel(HPET_TN_ENABLE | HPET_TN_PERIODIC | HPET_TN_SETVAL |
HPET_TN_32BIT, HPET_T0_CFG);
However after the i386/x86_64 HPET merge, we do the following:
cfg = hpet_readl(HPET_Tn_CFG(timer));
cfg |= HPET_TN_ENABLE | HPET_TN_PERIODIC |
HPET_TN_SETVAL | HPET_TN_32BIT;
hpet_writel(cfg, HPET_Tn_CFG(timer));
However on LS21s with HPET enabled in the BIOS, the HPET_T0_CFG register
boots with Level triggered interrupts (HPET_TN_LEVEL) enabled. This
causes the periodic interrupt to be not so periodic, and that results in
the boot time hang I reported earlier in the delay calibration.
My fix: Always disable HPET_TN_LEVEL when setting up periodic mode.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Impact: fix time warp bug
Alex Shi, along with Yanmin Zhang have been noticing occasional time
inconsistencies recently. Through their great diagnosis, they found that
the xtime_nsec value used in update_wall_time was occasionally going
negative. After looking through the code for awhile, I realized we have
the possibility for an underflow when three conditions are met in
update_wall_time():
1) We have accumulated a second's worth of nanoseconds, so we
incremented xtime.tv_sec and appropriately decrement xtime_nsec.
(This doesn't cause xtime_nsec to go negative, but it can cause it
to be small).
2) The remaining offset value is large, but just slightly less then
cycle_interval.
3) clocksource_adjust() is speeding up the clock, causing a
corrective amount (compensating for the increase in the multiplier
being multiplied against the unaccumulated offset value) to be
subtracted from xtime_nsec.
This can cause xtime_nsec to underflow.
Unfortunately, since we notify the NTP subsystem via second_overflow()
whenever we accumulate a full second, and this effects the error
accumulation that has already occured, we cannot simply revert the
accumulated second from xtime nor move the second accumulation to after
the clocksource_adjust call without a change in behavior.
This leaves us with (at least) two options:
1) Simply return from clocksource_adjust() without making a change if we
notice the adjustment would cause xtime_nsec to go negative.
This would work, but I'm concerned that if a large adjustment was needed
(due to the error being large), it may be possible to get stuck with an
ever increasing error that becomes too large to correct (since it may
always force xtime_nsec negative). This may just be paranoia on my part.
2) Catch xtime_nsec if it is negative, then add back the amount its
negative to both xtime_nsec and the error.
This second method is consistent with how we've handled earlier rounding
issues, and also has the benefit that the error being added is always in
the oposite direction also always equal or smaller then the correction
being applied. So the risk of a corner case where things get out of
control is lessened.
This patch fixes bug 11970, as tested by Yanmin Zhang
http://bugzilla.kernel.org/show_bug.cgi?id=11970
Reported-by: [email protected]
Signed-off-by: John Stultz <[email protected]>
Acked-by: "Zhang, Yanmin" <[email protected]>
Tested-by: "Zhang, Yanmin" <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
In talking with Josip Loncaric, and his work on clock synchronization (see
btime.sf.net), he mentioned that for really close synchronization, it is
useful to have access to "hardware time", that is a notion of time that is
not in any way adjusted by the clock slewing done to keep close time sync.
Part of the issue is if we are using the kernel's ntp adjusted
representation of time in order to measure how we should correct time, we
can run into what Paul McKenney aptly described as "Painting a road using
the lines we're painting as the guide".
I had been thinking of a similar problem, and was trying to come up with a
way to give users access to a purely hardware based time representation
that avoided users having to know the underlying frequency and mask values
needed to deal with the wide variety of possible underlying hardware
counters.
My solution is to introduce CLOCK_MONOTONIC_RAW. This exposes a
nanosecond based time value, that increments starting at bootup and has no
frequency adjustments made to it what so ever.
The time is accessed from userspace via the posix_clock_gettime() syscall,
passing CLOCK_MONOTONIC_RAW as the clock_id.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
The clocksource frequency is represented by
clocksource->mult/2^(clocksource->shift). Currently, when NTP makes
adjustments to the clock frequency, they are made directly to the mult
value.
This has the drawback that once changed, we cannot know what the orignal
mult value was, or how much adjustment has been applied.
This property causes problems in calculating proper ntp intervals when
switching back and forth between clocksources.
This patch separates the current mult value into a mult and mult_orig
pair. The mult_orig value stays constant, while the ntp clocksource
adjustments are done only to the mult value.
This allows for correct ntp interval calculation and additionally lays the
groundwork for a new notion of time, what I'm calling the monotonic-raw
time, which is introduced in a following patch.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
clocksource initialization and error accumulation. This corrects a 280ppm
drift seen on some systems using acpi_pm, and affects other clocksources as
well (likely to a lesser degree).
Signed-off-by: John Stultz <[email protected]>
Cc: Roman Zippel <[email protected]>
Cc: Steven Rostedt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
I recently noticed on one of my boxes that when synched with an NTP
server, the drift value reported for the system was ~283ppm. While in
some cases, clock hardware can be that bad, it struck me as unusual as
the system was using the acpi_pm clocksource, which is one of the more
trustworthy and accurate clocksources on x86 hardware.
I brought up another system and let it sync to the same NTP server, and
I noticed a similar 280some ppm drift.
In looking at the code, I found that the acpi_pm's constant frequency
was being computed correctly at boot-up, however once the system was up,
even without the ntp daemon running, the clocksource's frequency was
being modified by the clocksource_adjust() function.
Digging deeper, I realized that in the code that keeps track of how much
the clocksource is skewing from the ntp desired time, we were using
different lengths to establish how long an time interval was.
The clocksource was being setup with the following interval:
NTP_INTERVAL_LENGTH = NSEC_PER_SEC/NTP_INTERVAL_FREQ
While the ntp code was using the tick_length_base value:
tick_length_base ~= (tick_usec * NSEC_PER_USEC * USER_HZ)
/NTP_INTERVAL_FREQ
The subtle difference is:
(tick_usec * NSEC_PER_USEC * USER_HZ) != NSEC_PER_SEC
This difference in calculation was causing the clocksource correction
code to apply a correction factor to the clocksource so the two
intervals were the same, however this results in the actual frequency of
the clocksource to be made incorrect. I believe this difference would
affect all clocksources, although to differing degrees depending on the
clocksource resolution.
The issue was introduced when my HZ free ntp patch landed in 2.6.21-rc1,
so my apologies for the mistake, and for not noticing it until now.
The following patch, corrects the clocksource's initialization code so
it uses the same interval length as the code in ntp.c. After applying
this patch, the drift value for the same system went from ~283ppm to
only 2.635ppm.
I believe this patch to be good, however it does affect all arches and
I've only tested on x86, so some caution is advised. I do think it would
be a likely candidate for a stable 2.6.24.x release.
Any thoughts or feedback would be appreciated.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Michael Kerrisk reported that a long standing bug in the adjtimex()
system call causes glibc's adjtime(3) function to deliver the wrong
results if 'delta' is NULL.
add the ADJ_OFFSET_SS_READ API detail, which will be used by glibc
to fix this API compatibility bug.
Also see: http://bugzilla.kernel.org/show_bug.cgi?id=6761
[ [email protected]: added patch description and made it backwards compatible ]
NOTE: the new flag is defined 0xa001 so that it returns -EINVAL on
older kernels - this way glibc can use it safely. Suggested by Ulrich
Drepper.
Acked-by: Ulrich Drepper <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
To avoid lock contention, we distribute the sched_timer calls across the
cpus so they do not trigger at the same instant. However, I used NR_CPUS,
which can cause needless grouping on small smp systems depending on your
kernel config. This patch converts to using num_possible_cpus() so we
spread it as evenly as possible on every machine.
Briefly tested w/ NR_CPUS=255 and verified reduced contention.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This avoids xtime lag seen with dynticks, because while 'xtime' itself
is still not updated often, we keep a 'xtime_cache' variable around that
contains the approximate real-time that _is_ updated each time we do a
'update_wall_time()', and is thus never off by more than one tick.
IOW, this restores the original semantics for 'xtime' users, as long as
you use the proper abstraction functions (ie 'current_kernel_time()' or
'get_seconds()' depending on whether you want a timespec or just the
seconds field).
[ Updated Patch. As penance for my sins I've also yanked another #ifdef
that was added to avoid the xtime lag w/ hrtimers. ]
Signed-off-by: John Stultz <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This avoids use of the kernel-internal "xtime" variable directly outside
of the actual time-related functions. Instead, use the helper functions
that we already have available to us.
This doesn't actually change any behaviour, but this will allow us to
fix the fact that "xtime" isn't updated very often with CONFIG_NO_HZ
(because much of the realtime information is maintained as separate
offsets to 'xtime'), which has caused interfaces that use xtime directly
to get a time that is out of sync with the real-time clock by up to a
third of a second or so.
Signed-off-by: John Stultz <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
After discussing w/ Thomas over IRC, it seems the issue is the sched tick
fires on every cpu at the same time, causing extra lock contention.
This smaller change, adds an extra offset per cpu so the ticks don't line up.
This patch also drops the idle latency from 40us down to under 20us.
Signed-off-by: john stultz <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The vsyscall time() function basically returns the second portion of
xtime directly. This however means that there is about a ticks worth of
time each second where time() will return a second value less then what
gettimeofday() does.
Additionally, this window where vtime() is behind vgettimeofday() grows
when dynticks is enabled, so its probably good to get this in before
dynticks lands.
Big thanks to Sripathi for noticing this issue and creating a test case
to work with!
This patch changes the vtime() implemenation to call vgettimeofday(),
much as syscall time() implementation calls gettimeofday().
2.6.21 stable candidate too
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Convert an arch that does not currently implement sub-jiffy timekeeping to
use the generic timekeeping code.
v850 looks like it has some intent to implement sub-jiffy timekeeping, so
it may not yet be appropriate to try to convert, but I figured I'd get the
maintainer's input and submit the patch for comment.
Signed-off-by: John Stultz <[email protected]>
Cc: Miles Bader <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Currently h8/300 does not implement sub-jiffy timekeeping, so there is no
benefit to having arch specific timekeeping code.
This patch simply removes those functions and enables the generic
timekeeping code.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Yoshinori Sato <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Change mark_tsc_unstable() so it takes a string argument, which holds the
reason the TSC was marked unstable.
This is then displayed the first time mark_tsc_unstable is called.
This should help us better debug why the TSC was marked unstable on certain
systems and allow us to make sure we're not being overly paranoid when
throwing out this troublesome clocksource.
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Andi Kleen <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
In debugging a problem w/ the -rt tree, I noticed that on systems that mark
the tsc as unstable before it is registered, the TSC would still be
selected and used for a short period of time. Digging in it looks to be a
result of the mix of the clocksource list changes and my clocksource
initialization changes.
With the -rt tree, using a bad TSC, even for a short period of time can
results in a hang at boot. I was not able to reproduce this hang w/
mainline, but I'm not completely certain that someone won't trip on it.
This patch resolves the issue by initializing the jiffies clocksource
earlier so a bad TSC won't get selected just because nothing else is yet
registered.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Andi Kleen <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
I've been seeing some odd NTP behavior recently on a few boxes and
finally narrowed it down to time_offset overflowing when converted to
SHIFT_UPDATE units (which was a side effect from my HZfreeNTP patch).
This patch converts time_offset from a long to a s64 which resolves the
issue.
[[email protected]: signedness fixes]
Signed-off-by: John Stultz <[email protected]>
Cc: Roman Zippel <[email protected]>
Cc: john stultz <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
On Bob's machine clocksource is selecting PIT over the ACPI PM timer,
because he has the PIIX4 bug. That bug drops the ACPI PM timers rating
to the same as the PIT, so that's why you're getting the PIT.
Realistically, the PIT is much slower then even the triple read ACPI PM,
so the de-ranking code is probably dropping it too far.
So don't drop ACPI PM quite so low if we see the PIIX4 bug.
Signed-off-by: John Stultz <[email protected]>
Cc: Bob Tracy <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Andi Kleen <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This patch resolves the issue found here:
http://bugme.osdl.org/show_bug.cgi?id=7426
The basic summary is:
Currently we register most of i386/x86_64 clocksources at module_init
time. Then we enable clocksource selection at late_initcall time. This
causes some problems for drivers that use gettimeofday for init
calibration routines (specifically the es1968 driver in this case),
where durring module_init, the only clocksource available is the low-res
jiffies clocksource. This may cause slight calibration errors, due to
the small sampling time used.
It should be noted that drivers that require fine grained time may not
function on architectures that do not have better then jiffies
resolution timekeeping (there are a few). However, this does not
discount the reasonable need for such fine-grained timekeeping at init
time.
Thus the solution here is to register clocksources earlier (ideally when
the hardware is being initialized), and then we enable clocksource
selection at fs_initcall (before device_initcall).
This patch should probably get some testing time in -mm, since
clocksource selection is one of the most important issues for correct
timekeeping, and I've only been able to test this on a few of my own
boxes.
Signed-off-by: John Stultz <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: "David S. Miller" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
In preparation for supporting generic timekeeping, this patch cleans up
x86-64's use of vxtime.hpet_address, changing it to just hpet_address as is
also used in i386. This is necessary since the vxtime structure will be going
away.
Signed-off-by: John Stultz <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Avoid possible PIT livelock issues seen on SMP systems (and reported by
Andi), by not allowing it as a clocksource on SMP boxes.
However, since the PIT may no longer be present, we have to properly handle
the cases where SMP systems have TSC skew and fall back from the TSC.
Since the PIT isn't there, it would "fall back" to the TSC again. So this
changes the jiffies rating to 1, and the TSC-bad rating value to 0.
Thus you will get the following behavior priority on i386 systems:
tsc [if present & stable]
hpet [if present]
cyclone [if present]
acpi_pm [if present]
pit [if UP]
jiffies
Rather then the current more complicated:
tsc [if present & stable]
hpet [if present]
cyclone [if present]
acpi_pm [if present]
pit [if cpus < 4]
tsc [if present & unstable]
jiffies
Signed-off-by: John Stultz <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Remove all references to xtime in i386 and replace them w/
get/set_timeofday(). Requires some ugly and uncertain changes to APM, but
has been lightly tested to work.
Signed-off-by: John Stultz <[email protected]>
Acked-by: Ingo Molnar <[email protected]>
Acked-by: Mikael Pettersson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Apparently some systems export valid HPET addresses, but hpet_enable()
fails. Then when the HPET clocksource starts up, it only checks for a
valid HPET address, and the result is a system where time does not advance.
See http://bugme.osdl.org/show_bug.cgi?id=7062 for details.
This patch just makes sure we better check that the HPET is functional
before registering the HPET clocksource.
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
We found this issue last week w/ the -RT kernel, but it seems the same
issue is in mainline as well.
Basically it is possible for futex_unlock_pi to return without actually
freeing the lock. This is due to buggy logic in the use of
futex_handle_fault() and its attempt argument in a failure case.
Looking at futex.c the logic is as follows:
1) In futex_unlock_pi() we start w/ ret=0 and we go down to the first
futex_atomic_cmpxchg_inatomic(), where we find uval==-EFAULT. We then
jump to the pi_faulted label.
2) From pi_faulted: We increment attempt, unlock the sem and hit the
retry label.
3) From the retry label, with ret still zero, we again hit EFAULT on the
first futex_atomic_cmpxchg_inatomic(), and again goto the pi_faulted
label.
4) Again from pi_faulted: we increment attempt and enter the
conditional, where we call futex_handle_fault.
5) futex_handle_fault fails, and we goto the out_unlock_release_sem
label.
6) From out_unlock_release_sem we return, and since ret is still zero,
we return without error, while never actually unlocking the lock.
Issue #1: at the first futex_atomic_cmpxchg_inatomic() we should probably
be setting ret=-EFAULT before jumping to pi_faulted: However in our case
this doesn't really affect anything, as the glibc we're using ignores the
error value from futex_unlock_pi().
Issue #2: Look at futex_handle_fault(), its first conditional will return
-EFAULT if attempt is >= 2. However, from the "if(attempt++)
futex_handle_fault(attempt)" logic above, we'll *never* call
futex_handle_fault when attempt is less then two. So we never get a chance
to even try to fault the page in.
The following patch addresses these two issues by 1) Always setting ret to
-EFAULT if futex_handle_fault fails, and 2) Removing the = in
futex_handle_fault's (attempt >= 2) check.
I'm really not sure this is the right fix, but wanted to bring it up so
folks knew the issue is alive and well in the current -git tree. From
looking at the git logs the logic was first introduced (then later copied
to other places) in the following commit almost a year ago:
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4732efbeb997189d9f9b04708dc26bf8613ed721;hp=5b039e681b8c5f30aac9cc04385cc94be45d0823
Cc: Rusty Russell <[email protected]>
Cc: Ingo Molnar <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
Resolve problems seen w/ APM suspend.
Due to resume initialization ordering, its possible we could get a timer
interrupt before the timekeeping resume() function is called. This patch
ensures we don't do any timekeeping accounting before we're fully resumed.
(akpm: fixes the machine-freezes-on-APM-resume bug)
Signed-off-by: John Stultz <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
We're not reay to take a timer interrupt until timekeeping_init() has run.
But time_init() will start the time interrupt and if it is called with
local interrupts enabled we'll immediately take an interrupt and die.
Fix that by running timekeeping_init() prior to time_init().
We don't know _why_ local interrupts got enabled on Jesse Brandeburg's
machine. That's a separate as-yet-unsolved problem. THe patch adds a little
bit of debugging to detect that.
This whole requirement that local interrupts be held off during early boot
keeps on biting us.
Signed-off-by: John Stultz <[email protected]>
Cc: Jesse Brandeburg <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Here is the PIT fix against the TOD patches that Tim pointed out. Many
thanks to Tim for hunting this down.
Cc: Tim Mann <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
John Stultz