* [PATCH 0/2] Introduce a fast boot clock and add it as trace clock
@ 2016-11-21 22:04 Joel Fernandes
2016-11-21 22:04 ` [PATCH 1/2] timekeeping: Introduce a fast boot clock derived from fast monotonic clock Joel Fernandes
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Joel Fernandes @ 2016-11-21 22:04 UTC (permalink / raw)
To: linux-kernel
Cc: Joel Fernandes, Steven Rostedt, Thomas Gleixner, John Stultz,
Ingo Molnar
Currently no trace clock can account for suspend time, using monotonic during
tracing in the suspend path means the trace times wont be advaced. Using the
boot clock with ktime_get_with_offset is not an option due to live locking
concerns in NMI context as suggested by Thomas [1].
These patches add a fast boot clock based on fast monotonic clock and adds
a trace clock based on it which solves both these issues.
Changes since RFC:
- Moved the ktime_t offsets to the end of tk_fast
- use unlikely to optimize for the monotonic case
[1] https://lkml.org/lkml/2016/11/20/75
Joel Fernandes (2):
timekeeping: Introduce a fast boot clock derived from fast monotonic
clock
trace: Add an option for boot clock as trace clock
kernel/time/timekeeping.c | 36 ++++++++++++++++++++++++++++++------
kernel/trace/trace.c | 1 +
2 files changed, 31 insertions(+), 6 deletions(-)
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Ingo Molnar <mingo@redhat.com>
--
2.8.0.rc3.226.g39d4020
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 1/2] timekeeping: Introduce a fast boot clock derived from fast monotonic clock
2016-11-21 22:04 [PATCH 0/2] Introduce a fast boot clock and add it as trace clock Joel Fernandes
@ 2016-11-21 22:04 ` Joel Fernandes
2016-11-21 22:04 ` [PATCH 2/2] trace: Add an option for boot clock as trace clock Joel Fernandes
2016-11-21 22:06 ` [PATCH 0/2] Introduce a fast boot clock and add it " Joel Fernandes
2 siblings, 0 replies; 6+ messages in thread
From: Joel Fernandes @ 2016-11-21 22:04 UTC (permalink / raw)
To: linux-kernel
Cc: Joel Fernandes, Steven Rostedt, Thomas Gleixner, John Stultz,
Ingo Molnar
Introduce a fast boot clock. It accounts for suspend time and is based on the
fast monotonic clock so that its safe to use for tracing. Next patch adds the
clock to trace clock.
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Ingo Molnar <mingo@redhat.com>
Signed-off-by: Joel Fernandes <joelaf@google.com>
---
kernel/time/timekeeping.c | 36 ++++++++++++++++++++++++++++++------
1 file changed, 30 insertions(+), 6 deletions(-)
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index 37dec7e..ca88808 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -56,6 +56,12 @@ static struct timekeeper shadow_timekeeper;
struct tk_fast {
seqcount_t seq;
struct tk_read_base base[2];
+
+ /*
+ * first dimension is based on lower seq bit,
+ * second dimension is for offset type (real, boot, tai)
+ */
+ ktime_t offsets[2][3];
};
static struct tk_fast tk_fast_mono ____cacheline_aligned;
@@ -350,14 +356,20 @@ static void update_fast_timekeeper(struct tk_read_base *tkr, struct tk_fast *tkf
/* Force readers off to base[1] */
raw_write_seqcount_latch(&tkf->seq);
- /* Update base[0] */
+ /* Update base[0] and offsets*/
memcpy(base, tkr, sizeof(*base));
+ tkf->offsets[0][TK_OFFS_REAL] = tk_core.timekeeper.offs_real;
+ tkf->offsets[0][TK_OFFS_BOOT] = tk_core.timekeeper.offs_boot;
+ tkf->offsets[0][TK_OFFS_TAI] = tk_core.timekeeper.offs_tai;
/* Force readers back to base[0] */
raw_write_seqcount_latch(&tkf->seq);
- /* Update base[1] */
+ /* Update base[1] and offsets*/
memcpy(base + 1, base, sizeof(*base));
+ tkf->offsets[1][TK_OFFS_REAL] = tk_core.timekeeper.offs_real;
+ tkf->offsets[1][TK_OFFS_BOOT] = tk_core.timekeeper.offs_boot;
+ tkf->offsets[1][TK_OFFS_TAI] = tk_core.timekeeper.offs_tai;
}
/**
@@ -392,16 +404,23 @@ static void update_fast_timekeeper(struct tk_read_base *tkr, struct tk_fast *tkf
* of the following timestamps. Callers need to be aware of that and
* deal with it.
*/
-static __always_inline u64 __ktime_get_fast_ns(struct tk_fast *tkf)
+static __always_inline u64 __ktime_get_fast_ns(struct tk_fast *tkf, int offset)
{
struct tk_read_base *tkr;
unsigned int seq;
u64 now;
+ ktime_t *off;
do {
seq = raw_read_seqcount_latch(&tkf->seq);
tkr = tkf->base + (seq & 0x01);
- now = ktime_to_ns(tkr->base);
+
+ if (unlikely((offset >= 0))) {
+ off = tkf->offsets[seq & 0x01];
+ now = ktime_to_ns(ktime_add(tkr->base, off[offset]));
+ } else {
+ now = ktime_to_ns(tkr->base);
+ }
now += timekeeping_delta_to_ns(tkr,
clocksource_delta(
@@ -415,16 +434,21 @@ static __always_inline u64 __ktime_get_fast_ns(struct tk_fast *tkf)
u64 ktime_get_mono_fast_ns(void)
{
- return __ktime_get_fast_ns(&tk_fast_mono);
+ return __ktime_get_fast_ns(&tk_fast_mono, -1);
}
EXPORT_SYMBOL_GPL(ktime_get_mono_fast_ns);
u64 ktime_get_raw_fast_ns(void)
{
- return __ktime_get_fast_ns(&tk_fast_raw);
+ return __ktime_get_fast_ns(&tk_fast_raw, -1);
}
EXPORT_SYMBOL_GPL(ktime_get_raw_fast_ns);
+u64 ktime_get_boot_fast_ns(void)
+{
+ return __ktime_get_fast_ns(&tk_fast_mono, TK_OFFS_BOOT);
+}
+
/* Suspend-time cycles value for halted fast timekeeper. */
static cycle_t cycles_at_suspend;
--
2.8.0.rc3.226.g39d4020
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 2/2] trace: Add an option for boot clock as trace clock
2016-11-21 22:04 [PATCH 0/2] Introduce a fast boot clock and add it as trace clock Joel Fernandes
2016-11-21 22:04 ` [PATCH 1/2] timekeeping: Introduce a fast boot clock derived from fast monotonic clock Joel Fernandes
@ 2016-11-21 22:04 ` Joel Fernandes
2016-11-21 22:48 ` kbuild test robot
2016-11-21 22:06 ` [PATCH 0/2] Introduce a fast boot clock and add it " Joel Fernandes
2 siblings, 1 reply; 6+ messages in thread
From: Joel Fernandes @ 2016-11-21 22:04 UTC (permalink / raw)
To: linux-kernel
Cc: Joel Fernandes, Steven Rostedt, Thomas Gleixner, John Stultz,
Ingo Molnar
Unlike monotonic clock, boot clock as a trace clock will account for
time spent in suspend useful for tracing suspend/resume. This uses
earlier introduced infrastructure for using the fast boot clock.
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Ingo Molnar <mingo@redhat.com>
Signed-off-by: Joel Fernandes <joelaf@google.com>
---
kernel/trace/trace.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index 8696ce6..f7b64db 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -1125,6 +1125,7 @@ static struct {
{ trace_clock, "perf", 1 },
{ ktime_get_mono_fast_ns, "mono", 1 },
{ ktime_get_raw_fast_ns, "mono_raw", 1 },
+ { ktime_get_boot_fast_ns, "boot", 1 },
ARCH_TRACE_CLOCKS
};
--
2.8.0.rc3.226.g39d4020
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 0/2] Introduce a fast boot clock and add it as trace clock
2016-11-21 22:04 [PATCH 0/2] Introduce a fast boot clock and add it as trace clock Joel Fernandes
2016-11-21 22:04 ` [PATCH 1/2] timekeeping: Introduce a fast boot clock derived from fast monotonic clock Joel Fernandes
2016-11-21 22:04 ` [PATCH 2/2] trace: Add an option for boot clock as trace clock Joel Fernandes
@ 2016-11-21 22:06 ` Joel Fernandes
2 siblings, 0 replies; 6+ messages in thread
From: Joel Fernandes @ 2016-11-21 22:06 UTC (permalink / raw)
To: LKML
Cc: Joel Fernandes, Steven Rostedt, Thomas Gleixner, John Stultz,
Ingo Molnar
On Mon, Nov 21, 2016 at 2:04 PM, Joel Fernandes <joelaf@google.com> wrote:
> Currently no trace clock can account for suspend time, using monotonic during
> tracing in the suspend path means the trace times wont be advaced. Using the
> boot clock with ktime_get_with_offset is not an option due to live locking
> concerns in NMI context as suggested by Thomas [1].
> These patches add a fast boot clock based on fast monotonic clock and adds
> a trace clock based on it which solves both these issues.
>
> Changes since RFC:
> - Moved the ktime_t offsets to the end of tk_fast
> - use unlikely to optimize for the monotonic case
>
Please scratch these patches, I didn't make a change I had to since
the RFC. Sorry for the noise!
> [1] https://lkml.org/lkml/2016/11/20/75
>
> Joel Fernandes (2):
> timekeeping: Introduce a fast boot clock derived from fast monotonic
> clock
> trace: Add an option for boot clock as trace clock
>
> kernel/time/timekeeping.c | 36 ++++++++++++++++++++++++++++++------
> kernel/trace/trace.c | 1 +
> 2 files changed, 31 insertions(+), 6 deletions(-)
>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: John Stultz <john.stultz@linaro.org>
> Cc: Ingo Molnar <mingo@redhat.com>
> --
> 2.8.0.rc3.226.g39d4020
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 2/2] trace: Add an option for boot clock as trace clock
2016-11-21 22:04 ` [PATCH 2/2] trace: Add an option for boot clock as trace clock Joel Fernandes
@ 2016-11-21 22:48 ` kbuild test robot
2016-11-21 22:59 ` Joel Fernandes
0 siblings, 1 reply; 6+ messages in thread
From: kbuild test robot @ 2016-11-21 22:48 UTC (permalink / raw)
To: Joel Fernandes
Cc: kbuild-all, linux-kernel, Joel Fernandes, Steven Rostedt,
Thomas Gleixner, John Stultz, Ingo Molnar
[-- Attachment #1: Type: text/plain, Size: 1358 bytes --]
Hi Joel,
[auto build test ERROR on tip/timers/core]
[also build test ERROR on v4.9-rc6 next-20161117]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Joel-Fernandes/timekeeping-Introduce-a-fast-boot-clock-derived-from-fast-monotonic-clock/20161122-063314
config: i386-randconfig-x004-201647 (attached as .config)
compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901
reproduce:
# save the attached .config to linux build tree
make ARCH=i386
All errors (new ones prefixed by >>):
>> kernel/trace/trace.c:1128:4: error: 'ktime_get_boot_fast_ns' undeclared here (not in a function)
{ ktime_get_boot_fast_ns, "boot", 1 },
^~~~~~~~~~~~~~~~~~~~~~
vim +/ktime_get_boot_fast_ns +1128 kernel/trace/trace.c
1122 { trace_clock_global, "global", 1 },
1123 { trace_clock_counter, "counter", 0 },
1124 { trace_clock_jiffies, "uptime", 0 },
1125 { trace_clock, "perf", 1 },
1126 { ktime_get_mono_fast_ns, "mono", 1 },
1127 { ktime_get_raw_fast_ns, "mono_raw", 1 },
> 1128 { ktime_get_boot_fast_ns, "boot", 1 },
1129 ARCH_TRACE_CLOCKS
1130 };
1131
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 24041 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 2/2] trace: Add an option for boot clock as trace clock
2016-11-21 22:48 ` kbuild test robot
@ 2016-11-21 22:59 ` Joel Fernandes
0 siblings, 0 replies; 6+ messages in thread
From: Joel Fernandes @ 2016-11-21 22:59 UTC (permalink / raw)
To: kbuild test robot
Cc: kbuild-all, LKML, Steven Rostedt, Thomas Gleixner, John Stultz,
Ingo Molnar
On Mon, Nov 21, 2016 at 2:48 PM, kbuild test robot <lkp@intel.com> wrote:
> Hi Joel,
>
> [auto build test ERROR on tip/timers/core]
> [also build test ERROR on v4.9-rc6 next-20161117]
> [if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
>
> url: https://github.com/0day-ci/linux/commits/Joel-Fernandes/timekeeping-Introduce-a-fast-boot-clock-derived-from-fast-monotonic-clock/20161122-063314
> config: i386-randconfig-x004-201647 (attached as .config)
> compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901
> reproduce:
> # save the attached .config to linux build tree
> make ARCH=i386
>
> All errors (new ones prefixed by >>):
>
>>> kernel/trace/trace.c:1128:4: error: 'ktime_get_boot_fast_ns' undeclared here (not in a function)
> { ktime_get_boot_fast_ns, "boot", 1 },
Looks like I didn't have enough coffee today. I did test the patches
so this is rather strange, I must have missed the header export while
respinning from another tree. I will fix this, respin and post another
version later while waiting meanwhile for any other comments (and
addressing Steven's documentation comment).
Thanks,
Joel
> ^~~~~~~~~~~~~~~~~~~~~~
>
> vim +/ktime_get_boot_fast_ns +1128 kernel/trace/trace.c
>
> 1122 { trace_clock_global, "global", 1 },
> 1123 { trace_clock_counter, "counter", 0 },
> 1124 { trace_clock_jiffies, "uptime", 0 },
> 1125 { trace_clock, "perf", 1 },
> 1126 { ktime_get_mono_fast_ns, "mono", 1 },
> 1127 { ktime_get_raw_fast_ns, "mono_raw", 1 },
>> 1128 { ktime_get_boot_fast_ns, "boot", 1 },
> 1129 ARCH_TRACE_CLOCKS
> 1130 };
> 1131
>
> ---
> 0-DAY kernel test infrastructure Open Source Technology Center
> https://lists.01.org/pipermail/kbuild-all Intel Corporation
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2016-11-21 22:59 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-11-21 22:04 [PATCH 0/2] Introduce a fast boot clock and add it as trace clock Joel Fernandes
2016-11-21 22:04 ` [PATCH 1/2] timekeeping: Introduce a fast boot clock derived from fast monotonic clock Joel Fernandes
2016-11-21 22:04 ` [PATCH 2/2] trace: Add an option for boot clock as trace clock Joel Fernandes
2016-11-21 22:48 ` kbuild test robot
2016-11-21 22:59 ` Joel Fernandes
2016-11-21 22:06 ` [PATCH 0/2] Introduce a fast boot clock and add it " Joel Fernandes
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome