[PATCH 7/7] Introduce timekeeper latch synchronization

From: Mathieu Desnoyers
Date: Sat Sep 14 2013 - 15:48:54 EST


Unlike the sequence lock, this latch synchronization scheme, proposed by
Peter Zijlstra, always keeps a readable copy of the data. Therefore,
readers will never deadlock if they nest on the writer, whether this
happens because the read-side is explicitly called within the write-side
critical section, executed in a nested interrupt (e.g. NMI), or executed
in an execution context that has lock dependency with the write-side
critical section.

The only situations in which readers have to retry is if 2 updates or
more happen concurrently with the read. Therefore, the only situation
that can trigger a reader retry involves updater progress, therefore if
a reader interrupts an update, it would interrupt progress of every
updates, and therefore the reader never has to retry.

A nice side-effect of this scheme is that the reader latency in the
case readers execute concurrently with updaters should be diminished,
because readers don't have to busy-loop when executing concurrently with
updaters, unless 2 or more updates are performed concurrently with the
read.

The overhead of this scheme is that every update must perform a copy of
struct timekeeper and then modify this copy rather than to do the update
in-place.

There should not be any significant overhead added to the read-side,
given the number of memory barriers is unchanged.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx>
Cc: John Stultz <john.stultz@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
---
include/linux/timekeeper_internal.h | 93 ++++++++-
kernel/time/ntp.c | 20 +-
kernel/time/timekeeping.c | 375 ++++++++++++++++++++---------------
3 files changed, 317 insertions(+), 171 deletions(-)

diff --git a/include/linux/timekeeper_internal.h b/include/linux/timekeeper_internal.h
index 6f0532d..af54571 100644
--- a/include/linux/timekeeper_internal.h
+++ b/include/linux/timekeeper_internal.h
@@ -151,7 +151,98 @@ struct timekeeper {
struct timekeeper_ntp ntp;
};

-extern struct timekeeper timekeeper;
+struct timekeeper_latch {
+ unsigned long head, tail;
+ struct timekeeper data[2];
+};
+
+static inline
+int timekeeper_index(struct timekeeper_latch *tl, const struct timekeeper *tk)
+{
+ return tk - &tl->data[0];
+}
+
+extern struct timekeeper_latch timekeeper_latch;
+extern raw_spinlock_t timekeeper_lock;
+
+/**
+ * timekeeper_write_begin - begin timekeeper update.
+ *
+ " @tl: struct timekeeper_latch to update.
+ * @next: pointer to next element (output parameter).
+ *
+ * The area pointed to by "next" should be considered uninitialized.
+ * The caller needs to have exclusive update access to struct timekeeper_latch.
+ */
+static inline
+void timekeeper_write_begin(struct timekeeper_latch *tl,
+ struct timekeeper **next)
+{
+ const struct timekeeper *_prev;
+ struct timekeeper *_next;
+
+ tl->head++;
+ smp_wmb(); /* Store head before storing into next entry */
+ _prev = &tl->data[tl->tail & 1];
+ _next = &tl->data[tl->head & 1];
+ *_next = *_prev; /* Copy prev content into next */
+ if (_next->clock && _next->clock->update_latch)
+ _next->clock->update_latch(_next->clock,
+ timekeeper_index(tl, _prev),
+ timekeeper_index(tl, _next));
+ *next = _next;
+}
+
+/**
+ * timekeeper_write_end - end timekeeper update.
+ *
+ " @tl: struct timekeeper_latch.
+ *
+ * The caller needs to have exclusive update access to struct timekeeper_latch.
+ */
+static inline
+void timekeeper_write_end(struct timekeeper_latch *tl)
+{
+ smp_wmb(); /* Store into next entry before storing into tail */
+ tl->tail++;
+}
+
+/**
+ * timekeeper_read_begin - begin timekeeper read.
+ *
+ " @tl: struct timekeeper_latch to read.
+ * @tail: pointer to unsigned long containing tail position (output).
+ */
+static inline
+struct timekeeper *timekeeper_read_begin(struct timekeeper_latch *tl,
+ unsigned long *tail)
+{
+ unsigned long ret;
+
+ ret = ACCESS_ONCE(tl->tail);
+ smp_rmb(); /* Load tail before loading entry */
+ *tail = ret;
+ return &tl->data[ret & 1];
+}
+
+/**
+ * timekeeper_read_retry - end timekeeper read, trigger retry if needed.
+ *
+ " @tl: struct timekeeper_latch read.
+ * @tail: tail position returned as output by timekeeper_read_begin().
+ *
+ * If timekeeper_read_retry() returns nonzero, the content of the read should
+ * be considered invalid, and the read should be performed again to
+ * reattempt reading coherent data, starting with timekeeper_read_begin().
+ */
+static inline
+int timekeeper_read_retry(struct timekeeper_latch *tl, unsigned long tail)
+{
+ smp_rmb(); /* Load entry before loading head */
+ return (ACCESS_ONCE(tl->head) - tail >= 2);
+}
+
+extern struct timekeeper *timekeeper_get_init(void);

static inline struct timespec tk_xtime(struct timekeeper *tk)
{
diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c
index 2a1b4ef..71a5d2a 100644
--- a/kernel/time/ntp.c
+++ b/kernel/time/ntp.c
@@ -395,9 +395,10 @@ static DECLARE_DELAYED_WORK(sync_cmos_work, sync_cmos_clock);

static void sync_cmos_clock(struct work_struct *work)
{
- struct timekeeper_ntp *ntp = &timekeeper.ntp;
+ struct timekeeper *tk;
+ unsigned long seq, local_tick_nsec;
struct timespec now, next;
- int fail = 1;
+ int fail = 1, ret;

/*
* If we have an externally synchronized Linux clock, then update
@@ -406,7 +407,11 @@ static void sync_cmos_clock(struct work_struct *work)
* This code is run on a timer. If the clock is set, that timer
* may not expire at the correct time. Thus, we adjust...
*/
- if (!ntp_synced(ntp)) {
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ ret = ntp_synced(&tk->ntp);
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));
+ if (!ret) {
/*
* Not synced, exit, do not restart a timer (if one is
* running, let it run out).
@@ -415,7 +420,11 @@ static void sync_cmos_clock(struct work_struct *work)
}

getnstimeofday(&now);
- if (abs(now.tv_nsec - (NSEC_PER_SEC / 2)) <= ntp->tick_nsec / 2) {
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ local_tick_nsec = tk->ntp.tick_nsec;
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));
+ if (abs(now.tv_nsec - (NSEC_PER_SEC / 2)) <= local_tick_nsec / 2) {
struct timespec adjust = now;

fail = -ENODEV;
@@ -860,8 +869,9 @@ void __hardpps(struct timekeeper_ntp *ntp, const struct timespec *phase_ts,

static int __init ntp_tick_adj_setup(char *str)
{
- struct timekeeper_ntp *ntp = &timekeeper.ntp;
+ struct timekeeper_ntp *ntp;

+ ntp = &timekeeper_get_init()->ntp;
ntp->ntp_tick_adj = simple_strtol(str, NULL, 0);
ntp->ntp_tick_adj <<= NTP_SCALE_SHIFT;

diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index 2210abb..4039691 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -28,23 +28,22 @@
#include "timekeeping_internal.h"

#define TK_CLEAR_NTP (1 << 0)
-#define TK_MIRROR (1 << 1)
-#define TK_CLOCK_WAS_SET (1 << 2)
-
-struct timekeeper timekeeper = {
- .ntp = {
- .tick_usec = TICK_USEC,
- .time_state = TIME_OK,
- .time_status = STA_UNSYNC,
- .time_constant = 2,
- .time_maxerror = NTP_PHASE_LIMIT,
- .time_esterror = NTP_PHASE_LIMIT,
+#define TK_CLOCK_WAS_SET (1 << 1)
+
+struct timekeeper_latch timekeeper_latch = {
+ .data[0] = {
+ .ntp = {
+ .tick_usec = TICK_USEC,
+ .time_state = TIME_OK,
+ .time_status = STA_UNSYNC,
+ .time_constant = 2,
+ .time_maxerror = NTP_PHASE_LIMIT,
+ .time_esterror = NTP_PHASE_LIMIT,
+ },
},
};

-static DEFINE_RAW_SPINLOCK(timekeeper_lock);
-static seqcount_t timekeeper_seq;
-static struct timekeeper shadow_timekeeper;
+DEFINE_RAW_SPINLOCK(timekeeper_lock);

/* flag for if timekeeping is suspended */
int __read_mostly timekeeping_suspended;
@@ -52,6 +51,16 @@ int __read_mostly timekeeping_suspended;
/* Flag for if there is a persistent clock on this platform */
bool __read_mostly persistent_clock_exist = false;

+/*
+ * timekeeper_get_init - get initial timekeeper structure (boot time init)
+ */
+struct timekeeper *timekeeper_get_init(void)
+{
+ struct timekeeper_latch *tl = &timekeeper_latch;
+
+ return &tl->data[tl->head & 1];
+}
+
static inline void tk_normalize_xtime(struct timekeeper *tk)
{
while (tk->xtime_nsec >= ((u64)NSEC_PER_SEC << tk->shift)) {
@@ -114,10 +123,13 @@ static void tk_setup_internals(struct timekeeper *tk, struct clocksource *clock)
cycle_t interval;
u64 tmp, ntpinterval;
struct clocksource *old_clock;
+ int tk_index = timekeeper_index(&timekeeper_latch, tk);

old_clock = tk->clock;
tk->clock = clock;
- tk->cycle_last = clock->cycle_last = clock->read(clock);
+ tk->cycle_last =
+ clock->cycle_last_latch[tk_index] =
+ clock->cycle_last = clock->read(clock);

/* Do the ns -> cycle conversion first, using original mult */
tmp = NTP_INTERVAL_LENGTH;
@@ -178,13 +190,22 @@ static inline s64 timekeeping_get_ns(struct timekeeper *tk)
cycle_t cycle_now, cycle_delta;
struct clocksource *clock;
s64 nsec;
+ int tk_index = timekeeper_index(&timekeeper_latch, tk);

/* read clocksource: */
clock = tk->clock;
- cycle_now = clock->read(clock);
+ if (clock->read_latch) {
+ cycle_now = clock->read_latch(clock, tk_index);

- /* calculate the delta since the last update_wall_time: */
- cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
+ /* calculate the delta since the last update_wall_time: */
+ cycle_delta = (cycle_now -
+ clock->cycle_last_latch[tk_index]) & clock->mask;
+ } else {
+ cycle_now = clock->read(clock);
+
+ /* calculate the delta since the last update_wall_time: */
+ cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
+ }

nsec = cycle_delta * tk->mult + tk->xtime_nsec;
nsec >>= tk->shift;
@@ -198,13 +219,22 @@ static inline s64 timekeeping_get_ns_raw(struct timekeeper *tk)
cycle_t cycle_now, cycle_delta;
struct clocksource *clock;
s64 nsec;
+ int tk_index = timekeeper_index(&timekeeper_latch, tk);

/* read clocksource: */
clock = tk->clock;
- cycle_now = clock->read(clock);
+ if (clock->read_latch) {
+ cycle_now = clock->read_latch(clock, tk_index);

- /* calculate the delta since the last update_wall_time: */
- cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
+ /* calculate the delta since the last update_wall_time: */
+ cycle_delta = (cycle_now -
+ clock->cycle_last_latch[tk_index]) & clock->mask;
+ } else {
+ cycle_now = clock->read(clock);
+
+ /* calculate the delta since the last update_wall_time: */
+ cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
+ }

/* convert delta to nanoseconds. */
nsec = clocksource_cyc2ns(cycle_delta, clock->mult, clock->shift);
@@ -225,13 +255,15 @@ static void update_pvclock_gtod(struct timekeeper *tk, bool was_set)
*/
int pvclock_gtod_register_notifier(struct notifier_block *nb)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;
int ret;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
+ timekeeper_write_begin(&timekeeper_latch, &tk);
ret = raw_notifier_chain_register(&pvclock_gtod_chain, nb);
update_pvclock_gtod(tk, true);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

return ret;
@@ -264,9 +296,6 @@ static void timekeeping_update(struct timekeeper *tk, unsigned int action)
}
update_vsyscall(tk);
update_pvclock_gtod(tk, action & TK_CLOCK_WAS_SET);
-
- if (action & TK_MIRROR)
- memcpy(&shadow_timekeeper, &timekeeper, sizeof(timekeeper));
}

/**
@@ -281,11 +310,13 @@ static void timekeeping_forward_now(struct timekeeper *tk)
cycle_t cycle_now, cycle_delta;
struct clocksource *clock;
s64 nsec;
+ int tk_index = timekeeper_index(&timekeeper_latch, tk);

clock = tk->clock;
cycle_now = clock->read(clock);
cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
- tk->cycle_last = clock->cycle_last = cycle_now;
+ tk->cycle_last = clock->cycle_last =
+ clock->cycle_last_latch[tk_index] = cycle_now;

tk->xtime_nsec += cycle_delta * tk->mult;

@@ -307,17 +338,15 @@ static void timekeeping_forward_now(struct timekeeper *tk)
*/
int __getnstimeofday(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
s64 nsecs = 0;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
-
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ts->tv_sec = tk->xtime_sec;
nsecs = timekeeping_get_ns(tk);
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

ts->tv_nsec = 0;
timespec_add_ns(ts, nsecs);
@@ -346,18 +375,17 @@ EXPORT_SYMBOL(getnstimeofday);

ktime_t ktime_get(void)
{
- struct timekeeper *tk = &timekeeper;
- unsigned int seq;
+ struct timekeeper *tk;
+ unsigned long seq;
s64 secs, nsecs;

WARN_ON(timekeeping_suspended);

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
secs = tk->xtime_sec + tk->wall_to_monotonic.tv_sec;
nsecs = timekeeping_get_ns(tk) + tk->wall_to_monotonic.tv_nsec;
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));
/*
* Use ktime_set/ktime_add_ns to create a proper ktime on
* 32-bit architectures without CONFIG_KTIME_SCALAR.
@@ -376,20 +404,19 @@ EXPORT_SYMBOL_GPL(ktime_get);
*/
void ktime_get_ts(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct timespec tomono;
s64 nsec;
- unsigned int seq;
+ unsigned long seq;

WARN_ON(timekeeping_suspended);

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ts->tv_sec = tk->xtime_sec;
nsec = timekeeping_get_ns(tk);
tomono = tk->wall_to_monotonic;
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

ts->tv_sec += tomono.tv_sec;
ts->tv_nsec = 0;
@@ -406,19 +433,17 @@ EXPORT_SYMBOL_GPL(ktime_get_ts);
*/
void timekeeping_clocktai(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
u64 nsecs;

WARN_ON(timekeeping_suspended);

do {
- seq = read_seqcount_begin(&timekeeper_seq);
-
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ts->tv_sec = tk->xtime_sec + tk->tai_offset;
nsecs = timekeeping_get_ns(tk);
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

ts->tv_nsec = 0;
timespec_add_ns(ts, nsecs);
@@ -454,14 +479,14 @@ EXPORT_SYMBOL(ktime_get_clocktai);
*/
void getnstime_raw_and_real(struct timespec *ts_raw, struct timespec *ts_real)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
s64 nsecs_raw, nsecs_real;

WARN_ON_ONCE(timekeeping_suspended);

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);

*ts_raw = tk->raw_time;
ts_real->tv_sec = tk->xtime_sec;
@@ -470,7 +495,7 @@ void getnstime_raw_and_real(struct timespec *ts_raw, struct timespec *ts_real)
nsecs_raw = timekeeping_get_ns_raw(tk);
nsecs_real = timekeeping_get_ns(tk);

- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

timespec_add_ns(ts_raw, nsecs_raw);
timespec_add_ns(ts_real, nsecs_real);
@@ -503,7 +528,7 @@ EXPORT_SYMBOL(do_gettimeofday);
*/
int do_settimeofday(const struct timespec *tv)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct timespec ts_delta, xt;
unsigned long flags;

@@ -511,7 +536,7 @@ int do_settimeofday(const struct timespec *tv)
return -EINVAL;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);

timekeeping_forward_now(tk);

@@ -523,9 +548,9 @@ int do_settimeofday(const struct timespec *tv)

tk_set_xtime(tk, tv);

- timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
+ timekeeping_update(tk, TK_CLEAR_NTP | TK_CLOCK_WAS_SET);

- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

/* signal hrtimers about time change */
@@ -543,7 +568,7 @@ EXPORT_SYMBOL(do_settimeofday);
*/
int timekeeping_inject_offset(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;
struct timespec tmp;
int ret = 0;
@@ -552,7 +577,7 @@ int timekeeping_inject_offset(struct timespec *ts)
return -EINVAL;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);

timekeeping_forward_now(tk);

@@ -567,9 +592,9 @@ int timekeeping_inject_offset(struct timespec *ts)
tk_set_wall_to_mono(tk, timespec_sub(tk->wall_to_monotonic, *ts));

error: /* even if we error out, we forwarded the time, so call update */
- timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
+ timekeeping_update(tk, TK_CLEAR_NTP | TK_CLOCK_WAS_SET);

- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

/* signal hrtimers about time change */
@@ -586,14 +611,14 @@ EXPORT_SYMBOL(timekeeping_inject_offset);
*/
s32 timekeeping_get_tai_offset(void)
{
- struct timekeeper *tk = &timekeeper;
- unsigned int seq;
+ struct timekeeper *tk;
+ unsigned long seq;
s32 ret;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ret = tk->tai_offset;
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

return ret;
}
@@ -614,13 +639,13 @@ static void __timekeeping_set_tai_offset(struct timekeeper *tk, s32 tai_offset)
*/
void timekeeping_set_tai_offset(s32 tai_offset)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);
__timekeeping_set_tai_offset(tk, tai_offset);
- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
clock_was_set();
}
@@ -632,14 +657,14 @@ void timekeeping_set_tai_offset(s32 tai_offset)
*/
static int change_clocksource(void *data)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct clocksource *new, *old;
unsigned long flags;

new = (struct clocksource *) data;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);

timekeeping_forward_now(tk);
/*
@@ -657,9 +682,9 @@ static int change_clocksource(void *data)
module_put(new->owner);
}
}
- timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
+ timekeeping_update(tk, TK_CLEAR_NTP | TK_CLOCK_WAS_SET);

- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

return 0;
@@ -674,13 +699,25 @@ static int change_clocksource(void *data)
*/
int timekeeping_notify(struct clocksource *clock)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
+ struct clocksource *current_clock;
+ unsigned long seq;

- if (tk->clock == clock)
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ current_clock = tk->clock;
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));
+ if (current_clock == clock)
return 0;
+
stop_machine(change_clocksource, clock, NULL);
tick_clock_notify();
- return tk->clock == clock ? 0 : -1;
+
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ current_clock = tk->clock;
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));
+ return current_clock == clock ? 0 : -1;
}

/**
@@ -706,16 +743,15 @@ EXPORT_SYMBOL_GPL(ktime_get_real);
*/
void getrawmonotonic(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
s64 nsecs;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
nsecs = timekeeping_get_ns_raw(tk);
*ts = tk->raw_time;
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

timespec_add_ns(ts, nsecs);
}
@@ -726,16 +762,14 @@ EXPORT_SYMBOL(getrawmonotonic);
*/
int timekeeping_valid_for_hres(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
int ret;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
-
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ret = tk->clock->flags & CLOCK_SOURCE_VALID_FOR_HRES;
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

return ret;
}
@@ -745,16 +779,14 @@ int timekeeping_valid_for_hres(void)
*/
u64 timekeeping_max_deferment(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
u64 ret;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
-
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ret = tk->clock->max_idle_ns;
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

return ret;
}
@@ -794,7 +826,7 @@ void __attribute__((weak)) read_boot_clock(struct timespec *ts)
*/
void __init timekeeping_init(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct clocksource *clock;
unsigned long flags;
struct timespec now, boot, tmp;
@@ -818,7 +850,7 @@ void __init timekeeping_init(void)
}

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);
ntp_init(&tk->ntp);

clock = clocksource_default_clock();
@@ -839,9 +871,7 @@ void __init timekeeping_init(void)
tmp.tv_nsec = 0;
tk_set_sleep_time(tk, tmp);

- memcpy(&shadow_timekeeper, &timekeeper, sizeof(timekeeper));
-
- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
}

@@ -881,7 +911,7 @@ static void __timekeeping_inject_sleeptime(struct timekeeper *tk,
*/
void timekeeping_inject_sleeptime(struct timespec *delta)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;

/*
@@ -892,15 +922,15 @@ void timekeeping_inject_sleeptime(struct timespec *delta)
return;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);

timekeeping_forward_now(tk);

__timekeeping_inject_sleeptime(tk, delta);

- timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
+ timekeeping_update(tk, TK_CLEAR_NTP | TK_CLOCK_WAS_SET);

- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

/* signal hrtimers about time change */
@@ -916,12 +946,13 @@ void timekeeping_inject_sleeptime(struct timespec *delta)
*/
static void timekeeping_resume(void)
{
- struct timekeeper *tk = &timekeeper;
- struct clocksource *clock = tk->clock;
+ struct timekeeper *tk;
+ struct clocksource *clock;
unsigned long flags;
struct timespec ts_new, ts_delta;
cycle_t cycle_now, cycle_delta;
bool suspendtime_found = false;
+ int tk_index;

read_persistent_clock(&ts_new);

@@ -929,7 +960,9 @@ static void timekeeping_resume(void)
clocksource_resume();

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);
+ clock = tk->clock;
+ tk_index = timekeeper_index(&timekeeper_latch, tk);

/*
* After system resumes, we need to calculate the suspended time and
@@ -977,11 +1010,12 @@ static void timekeeping_resume(void)
__timekeeping_inject_sleeptime(tk, &ts_delta);

/* Re-base the last cycle value */
- tk->cycle_last = clock->cycle_last = cycle_now;
+ tk->cycle_last = clock->cycle_last =
+ clock->cycle_last_latch[tk_index] = cycle_now;
tk->ntp_error = 0;
timekeeping_suspended = 0;
- timekeeping_update(tk, TK_MIRROR | TK_CLOCK_WAS_SET);
- write_seqcount_end(&timekeeper_seq);
+ timekeeping_update(tk, TK_CLOCK_WAS_SET);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

touch_softlockup_watchdog();
@@ -994,7 +1028,7 @@ static void timekeeping_resume(void)

static int timekeeping_suspend(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;
struct timespec delta, delta_delta;
static struct timespec old_delta;
@@ -1010,7 +1044,7 @@ static int timekeeping_suspend(void)
persistent_clock_exist = true;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);
timekeeping_forward_now(tk);
timekeeping_suspended = 1;

@@ -1033,7 +1067,7 @@ static int timekeeping_suspend(void)
timekeeping_suspend_time =
timespec_add(timekeeping_suspend_time, delta_delta);
}
- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

clockevents_notify(CLOCK_EVT_NOTIFY_SUSPEND, NULL);
@@ -1372,29 +1406,31 @@ static inline void old_vsyscall_fixup(struct timekeeper *tk)
static void update_wall_time(void)
{
struct clocksource *clock;
- struct timekeeper *real_tk = &timekeeper;
- struct timekeeper *tk = &shadow_timekeeper;
+ struct timekeeper *tk;
cycle_t offset;
int shift = 0, maxshift;
unsigned int action;
unsigned long flags;
+ int tk_index;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
+ timekeeper_write_begin(&timekeeper_latch, &tk);
+ tk_index = timekeeper_index(&timekeeper_latch, tk);

/* Make sure we're fully resumed: */
if (unlikely(timekeeping_suspended))
goto out;

- clock = real_tk->clock;
+ clock = tk->clock;

#ifdef CONFIG_ARCH_USES_GETTIMEOFFSET
- offset = real_tk->cycle_interval;
+ offset = tk->cycle_interval;
#else
offset = (clock->read(clock) - clock->cycle_last) & clock->mask;
#endif

/* Check if there's really nothing to do */
- if (offset < real_tk->cycle_interval)
+ if (offset < tk->cycle_interval)
goto out;

/*
@@ -1431,23 +1467,11 @@ static void update_wall_time(void)
*/
action = accumulate_nsecs_to_secs(tk);

- write_seqcount_begin(&timekeeper_seq);
/* Update clock->cycle_last with the new value */
- clock->cycle_last = tk->cycle_last;
- /*
- * Update the real timekeeper.
- *
- * We could avoid this memcpy by switching pointers, but that
- * requires changes to all other timekeeper usage sites as
- * well, i.e. move the timekeeper pointer getter into the
- * spinlocked/seqcount protected sections. And we trade this
- * memcpy under the timekeeper_seq against one before we start
- * updating.
- */
- memcpy(real_tk, tk, sizeof(*tk));
- timekeeping_update(real_tk, action);
- write_seqcount_end(&timekeeper_seq);
+ clock->cycle_last_latch[tk_index] = clock->cycle_last = tk->cycle_last;
+ timekeeping_update(tk, action);
out:
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
}

@@ -1464,13 +1488,17 @@ out:
*/
void getboottime(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
- struct timespec boottime = {
- .tv_sec = tk->wall_to_monotonic.tv_sec +
- tk->total_sleep_time.tv_sec,
- .tv_nsec = tk->wall_to_monotonic.tv_nsec +
- tk->total_sleep_time.tv_nsec
- };
+ struct timekeeper *tk;
+ struct timespec boottime;
+ unsigned long seq;
+
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ boottime.tv_sec = tk->wall_to_monotonic.tv_sec +
+ tk->total_sleep_time.tv_sec;
+ boottime.tv_nsec = tk->wall_to_monotonic.tv_nsec +
+ tk->total_sleep_time.tv_nsec;
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

set_normalized_timespec(ts, -boottime.tv_sec, -boottime.tv_nsec);
}
@@ -1487,21 +1515,20 @@ EXPORT_SYMBOL_GPL(getboottime);
*/
void get_monotonic_boottime(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct timespec tomono, sleep;
s64 nsec;
- unsigned int seq;
+ unsigned long seq;

WARN_ON(timekeeping_suspended);

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
ts->tv_sec = tk->xtime_sec;
nsec = timekeeping_get_ns(tk);
tomono = tk->wall_to_monotonic;
sleep = tk->total_sleep_time;
-
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

ts->tv_sec += tomono.tv_sec + sleep.tv_sec;
ts->tv_nsec = 0;
@@ -1532,38 +1559,57 @@ EXPORT_SYMBOL_GPL(ktime_get_boottime);
*/
void monotonic_to_bootbased(struct timespec *ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
+ struct timespec ret;
+ unsigned long seq;
+
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ ret = timespec_add(*ts, tk->total_sleep_time);
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

- *ts = timespec_add(*ts, tk->total_sleep_time);
+ *ts = ret;
}
EXPORT_SYMBOL_GPL(monotonic_to_bootbased);

unsigned long get_seconds(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
+ unsigned long seq, ret;
+
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ ret = tk->xtime_sec;
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

- return tk->xtime_sec;
+ return ret;
}
EXPORT_SYMBOL(get_seconds);

struct timespec __current_kernel_time(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
+ unsigned long seq;
+ struct timespec ret;
+
+ do {
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
+ ret = tk_xtime(tk);
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

- return tk_xtime(tk);
+ return ret;
}

struct timespec current_kernel_time(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct timespec now;
unsigned long seq;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
-
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
now = tk_xtime(tk);
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

return now;
}
@@ -1571,16 +1617,15 @@ EXPORT_SYMBOL(current_kernel_time);

struct timespec get_monotonic_coarse(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
struct timespec now, mono;
unsigned long seq;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
-
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
now = tk_xtime(tk);
mono = tk->wall_to_monotonic;
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

set_normalized_timespec(&now, now.tv_sec + mono.tv_sec,
now.tv_nsec + mono.tv_nsec);
@@ -1607,15 +1652,15 @@ void do_timer(unsigned long ticks)
void get_xtime_and_monotonic_and_sleep_offset(struct timespec *xtim,
struct timespec *wtom, struct timespec *sleep)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
*xtim = tk_xtime(tk);
*wtom = tk->wall_to_monotonic;
*sleep = tk->total_sleep_time;
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));
}

#ifdef CONFIG_HIGH_RES_TIMERS
@@ -1630,13 +1675,13 @@ void get_xtime_and_monotonic_and_sleep_offset(struct timespec *xtim,
ktime_t ktime_get_update_offsets(ktime_t *offs_real, ktime_t *offs_boot,
ktime_t *offs_tai)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
ktime_t now;
- unsigned int seq;
+ unsigned long seq;
u64 secs, nsecs;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);

secs = tk->xtime_sec;
nsecs = timekeeping_get_ns(tk);
@@ -1644,7 +1689,7 @@ ktime_t ktime_get_update_offsets(ktime_t *offs_real, ktime_t *offs_boot,
*offs_real = tk->offs_real;
*offs_boot = tk->offs_boot;
*offs_tai = tk->offs_tai;
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

now = ktime_add_ns(ktime_set(secs, 0), nsecs);
now = ktime_sub(now, *offs_real);
@@ -1657,14 +1702,14 @@ ktime_t ktime_get_update_offsets(ktime_t *offs_real, ktime_t *offs_boot,
*/
ktime_t ktime_get_monotonic_offset(void)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long seq;
struct timespec wtom;

do {
- seq = read_seqcount_begin(&timekeeper_seq);
+ tk = timekeeper_read_begin(&timekeeper_latch, &seq);
wtom = tk->wall_to_monotonic;
- } while (read_seqcount_retry(&timekeeper_seq, seq));
+ } while (timekeeper_read_retry(&timekeeper_latch, seq));

return timespec_to_ktime(wtom);
}
@@ -1675,7 +1720,7 @@ EXPORT_SYMBOL_GPL(ktime_get_monotonic_offset);
*/
int do_adjtimex(struct timex *txc)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;
struct timespec ts;
s32 orig_tai, tai;
@@ -1700,7 +1745,7 @@ int do_adjtimex(struct timex *txc)
getnstimeofday(&ts);

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);

orig_tai = tai = tk->tai_offset;
ret = __do_adjtimex(&tk->ntp, txc, &ts, &tai);
@@ -1710,7 +1755,7 @@ int do_adjtimex(struct timex *txc)
update_pvclock_gtod(tk, true);
clock_was_set_delayed();
}
- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

ntp_notify_cmos_timer();
@@ -1724,15 +1769,15 @@ int do_adjtimex(struct timex *txc)
*/
void hardpps(const struct timespec *phase_ts, const struct timespec *raw_ts)
{
- struct timekeeper *tk = &timekeeper;
+ struct timekeeper *tk;
unsigned long flags;

raw_spin_lock_irqsave(&timekeeper_lock, flags);
- write_seqcount_begin(&timekeeper_seq);
+ timekeeper_write_begin(&timekeeper_latch, &tk);

__hardpps(&tk->ntp, phase_ts, raw_ts);

- write_seqcount_end(&timekeeper_seq);
+ timekeeper_write_end(&timekeeper_latch);
raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
}
EXPORT_SYMBOL(hardpps);
--
1.7.10.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/