* s_lock.c
* Hardware-dependent implementation of spinlocks.
*
+ * When waiting for a contended spinlock we loop tightly for awhile, then
+ * delay using pg_usleep() and try again. Preferably, "awhile" should be a
+ * small multiple of the maximum time we expect a spinlock to be held. 100
+ * iterations seems about right as an initial guess. However, on a
+ * uniprocessor the loop is a waste of cycles, while in a multi-CPU scenario
+ * it's usually better to spin a bit longer than to call the kernel, so we try
+ * to adapt the spin loop count depending on whether we seem to be in a
+ * uniprocessor or multiprocessor.
*
- * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
+ * Note: you might think MIN_SPINS_PER_DELAY should be just 1, but you'd
+ * be wrong; there are platforms where that can result in a "stuck
+ * spinlock" failure. This has been seen particularly on Alphas; it seems
+ * that the first TAS after returning from kernel space will always fail
+ * on that hardware.
+ *
+ * Once we do decide to block, we use randomly increasing pg_usleep()
+ * delays. The first delay is 1 msec, then the delay randomly increases to
+ * about one second, after which we reset to 1 msec and start again. The
+ * idea here is that in the presence of heavy contention we need to
+ * increase the delay, else the spinlock holder may never get to run and
+ * release the lock. (Consider situation where spinlock holder has been
+ * nice'd down in priority by the scheduler --- it will not get scheduled
+ * until all would-be acquirers are sleeping, so if we always use a 1-msec
+ * sleep, there is a real possibility of starvation.) But we can't just
+ * clamp the delay to an upper bound, else it would take a long time to
+ * make a reasonable number of tries.
+ *
+ * We time out and declare error after NUM_DELAYS delays (thus, exactly
+ * that many tries). With the given settings, this will usually take 2 or
+ * so minutes. It seems better to fix the total number of tries (and thus
+ * the probability of unintended failure) than to fix the total time
+ * spent.
+ *
+ * Portions Copyright (c) 1996-2016, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
- * $PostgreSQL: pgsql/src/backend/storage/lmgr/s_lock.c,v 1.44 2006/05/11 21:58:22 tgl Exp $
+ * src/backend/storage/lmgr/s_lock.c
*
*-------------------------------------------------------------------------
*/
#include <unistd.h>
#include "storage/s_lock.h"
-#include "miscadmin.h"
+#include "storage/barrier.h"
+
+
+#define MIN_SPINS_PER_DELAY 10
+#define MAX_SPINS_PER_DELAY 1000
+#define NUM_DELAYS 1000
+#define MIN_DELAY_USEC 1000L
+#define MAX_DELAY_USEC 1000000L
+
+slock_t dummy_spinlock;
static int spins_per_delay = DEFAULT_SPINS_PER_DELAY;
* s_lock_stuck() - complain about a stuck spinlock
*/
static void
-s_lock_stuck(volatile slock_t *lock, const char *file, int line)
+s_lock_stuck(const char *file, int line, const char *func)
{
+ if (!func)
+ func = "(unknown)";
#if defined(S_LOCK_TEST)
fprintf(stderr,
- "\nStuck spinlock (%p) detected at %s:%d.\n",
- lock, file, line);
+ "\nStuck spinlock detected at %s, %s:%d.\n",
+ func, file, line);
exit(1);
#else
- elog(PANIC, "stuck spinlock (%p) detected at %s:%d",
- lock, file, line);
+ elog(PANIC, "stuck spinlock detected at %s, %s:%d",
+ func, file, line);
#endif
}
-
/*
* s_lock(lock) - platform-independent portion of waiting for a spinlock.
*/
+int
+s_lock(volatile slock_t *lock, const char *file, int line, const char *func)
+{
+ SpinDelayStatus delayStatus = init_spin_delay(file, line, func);
+
+ while (TAS_SPIN(lock))
+ {
+ perform_spin_delay(&delayStatus);
+ }
+
+ finish_spin_delay(&delayStatus);
+
+ return delayStatus.delays;
+}
+
+#ifdef USE_DEFAULT_S_UNLOCK
void
-s_lock(volatile slock_t *lock, const char *file, int line)
+s_unlock(volatile slock_t *lock)
{
- /*
- * We loop tightly for awhile, then delay using pg_usleep() and try again.
- * Preferably, "awhile" should be a small multiple of the maximum time we
- * expect a spinlock to be held. 100 iterations seems about right as an
- * initial guess. However, on a uniprocessor the loop is a waste of
- * cycles, while in a multi-CPU scenario it's usually better to spin a bit
- * longer than to call the kernel, so we try to adapt the spin loop count
- * depending on whether we seem to be in a uniprocessor or multiprocessor.
- *
- * Note: you might think MIN_SPINS_PER_DELAY should be just 1, but you'd
- * be wrong; there are platforms where that can result in a "stuck
- * spinlock" failure. This has been seen particularly on Alphas; it seems
- * that the first TAS after returning from kernel space will always fail
- * on that hardware.
- *
- * Once we do decide to block, we use randomly increasing pg_usleep()
- * delays. The first delay is 1 msec, then the delay randomly increases to
- * about one second, after which we reset to 1 msec and start again. The
- * idea here is that in the presence of heavy contention we need to
- * increase the delay, else the spinlock holder may never get to run and
- * release the lock. (Consider situation where spinlock holder has been
- * nice'd down in priority by the scheduler --- it will not get scheduled
- * until all would-be acquirers are sleeping, so if we always use a 1-msec
- * sleep, there is a real possibility of starvation.) But we can't just
- * clamp the delay to an upper bound, else it would take a long time to
- * make a reasonable number of tries.
- *
- * We time out and declare error after NUM_DELAYS delays (thus, exactly
- * that many tries). With the given settings, this will usually take 2 or
- * so minutes. It seems better to fix the total number of tries (and thus
- * the probability of unintended failure) than to fix the total time
- * spent.
- *
- * The pg_usleep() delays are measured in milliseconds because 1 msec is a
- * common resolution limit at the OS level for newer platforms. On older
- * platforms the resolution limit is usually 10 msec, in which case the
- * total delay before timeout will be a bit more.
- */
-#define MIN_SPINS_PER_DELAY 10
-#define MAX_SPINS_PER_DELAY 1000
-#define NUM_DELAYS 1000
-#define MIN_DELAY_MSEC 1
-#define MAX_DELAY_MSEC 1000
+#ifdef TAS_ACTIVE_WORD
+ /* HP's PA-RISC */
+ *TAS_ACTIVE_WORD(lock) = -1;
+#else
+ *lock = 0;
+#endif
+}
+#endif
- int spins = 0;
- int delays = 0;
- int cur_delay = 0;
+/*
+ * Wait while spinning on a contended spinlock.
+ */
+void
+perform_spin_delay(SpinDelayStatus *status)
+{
+ /* CPU-specific delay each time through the loop */
+ SPIN_DELAY();
- while (TAS(lock))
+ /* Block the process every spins_per_delay tries */
+ if (++(status->spins) >= spins_per_delay)
{
- /* CPU-specific delay each time through the loop */
- SPIN_DELAY();
+ if (++(status->delays) > NUM_DELAYS)
+ s_lock_stuck(status->file, status->line, status->func);
- /* Block the process every spins_per_delay tries */
- if (++spins >= spins_per_delay)
- {
- if (++delays > NUM_DELAYS)
- s_lock_stuck(lock, file, line);
+ if (status->cur_delay == 0) /* first time to delay? */
+ status->cur_delay = MIN_DELAY_USEC;
- if (cur_delay == 0) /* first time to delay? */
- cur_delay = MIN_DELAY_MSEC;
-
- pg_usleep(cur_delay * 1000L);
+ pg_usleep(status->cur_delay);
#if defined(S_LOCK_TEST)
- fprintf(stdout, "*");
- fflush(stdout);
+ fprintf(stdout, "*");
+ fflush(stdout);
#endif
- /* increase delay by a random fraction between 1X and 2X */
- cur_delay += (int) (cur_delay *
- ((double) random() / (double) MAX_RANDOM_VALUE) + 0.5);
- /* wrap back to minimum delay when max is exceeded */
- if (cur_delay > MAX_DELAY_MSEC)
- cur_delay = MIN_DELAY_MSEC;
+ /* increase delay by a random fraction between 1X and 2X */
+ status->cur_delay += (int) (status->cur_delay *
+ ((double) random() / (double) MAX_RANDOM_VALUE) + 0.5);
+ /* wrap back to minimum delay when max is exceeded */
+ if (status->cur_delay > MAX_DELAY_USEC)
+ status->cur_delay = MIN_DELAY_USEC;
- spins = 0;
- }
+ status->spins = 0;
}
+}
- /*
- * If we were able to acquire the lock without delaying, it's a good
- * indication we are in a multiprocessor. If we had to delay, it's a sign
- * (but not a sure thing) that we are in a uniprocessor. Hence, we
- * decrement spins_per_delay slowly when we had to delay, and increase it
- * rapidly when we didn't. It's expected that spins_per_delay will
- * converge to the minimum value on a uniprocessor and to the maximum
- * value on a multiprocessor.
- *
- * Note: spins_per_delay is local within our current process. We want to
- * average these observations across multiple backends, since it's
- * relatively rare for this function to even get entered, and so a single
- * backend might not live long enough to converge on a good value. That
- * is handled by the two routines below.
- */
- if (cur_delay == 0)
+/*
+ * After acquiring a spinlock, update estimates about how long to loop.
+ *
+ * If we were able to acquire the lock without delaying, it's a good
+ * indication we are in a multiprocessor. If we had to delay, it's a sign
+ * (but not a sure thing) that we are in a uniprocessor. Hence, we
+ * decrement spins_per_delay slowly when we had to delay, and increase it
+ * rapidly when we didn't. It's expected that spins_per_delay will
+ * converge to the minimum value on a uniprocessor and to the maximum
+ * value on a multiprocessor.
+ *
+ * Note: spins_per_delay is local within our current process. We want to
+ * average these observations across multiple backends, since it's
+ * relatively rare for this function to even get entered, and so a single
+ * backend might not live long enough to converge on a good value. That
+ * is handled by the two routines below.
+ */
+void
+finish_spin_delay(SpinDelayStatus *status)
+{
+ if (status->cur_delay == 0)
{
/* we never had to delay */
if (spins_per_delay < MAX_SPINS_PER_DELAY)
}
}
-
/*
* Set local copy of spins_per_delay during backend startup.
*
/*
* We use an exponential moving average with a relatively slow adaption
* rate, so that noise in any one backend's result won't affect the shared
- * value too much. As long as both inputs are within the allowed range,
+ * value too much. As long as both inputs are within the allowed range,
* the result must be too, so we need not worry about clamping the result.
*
* We deliberately truncate rather than rounding; this is so that single
);
}
#endif /* __m68k__ && !__linux__ */
-#else /* not __GNUC__ */
-
-/*
- * All non gcc
- */
-
-
-#if defined(sun3)
-static void
-tas_dummy() /* really means: extern int tas(slock_t
- * *lock); */
-{
- asm("LLA0:");
- asm(" .data");
- asm(" .text");
- asm("|#PROC# 04");
- asm(" .globl _tas");
- asm("_tas:");
- asm("|#PROLOGUE# 1");
- asm(" movel sp@(0x4),a0");
- asm(" tas a0@");
- asm(" beq LLA1");
- asm(" moveq #-128,d0");
- asm(" rts");
- asm("LLA1:");
- asm(" moveq #0,d0");
- asm(" rts");
- asm(" .data");
-}
-#endif /* sun3 */
-
-
-#if defined(__sparc__) || defined(__sparc)
-/*
- * sparc machines not using gcc
- */
-static void
-tas_dummy() /* really means: extern int tas(slock_t
- * *lock); */
-{
- asm("_tas:");
-
- /*
- * Sparc atomic test and set (sparc calls it "atomic load-store")
- */
- asm("ldstub [%r8], %r8");
- asm("retl");
- asm("nop");
-}
-#endif /* __sparc || __sparc__ */
#endif /* not __GNUC__ */
#endif /* HAVE_SPINLOCKS */