diff options
Diffstat (limited to 'erts/emulator/beam/erl_lock_count.c')
| -rw-r--r-- | erts/emulator/beam/erl_lock_count.c | 965 |
1 files changed, 473 insertions, 492 deletions
diff --git a/erts/emulator/beam/erl_lock_count.c b/erts/emulator/beam/erl_lock_count.c index 481e92b2cd..d2e8f47d59 100644 --- a/erts/emulator/beam/erl_lock_count.c +++ b/erts/emulator/beam/erl_lock_count.c @@ -1,7 +1,7 @@ /* * %CopyrightBegin% * - * Copyright Ericsson AB 2008-2016. All Rights Reserved. + * Copyright Ericsson AB 2008-2017. All Rights Reserved. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -18,51 +18,37 @@ * %CopyrightEnd% */ -/* - * Description: Statistics for locks. - * - * Author: Björn-Egil Dahlberg - * Date: 2008-07-03 - */ - #ifdef HAVE_CONFIG_H # include "config.h" #endif -/* Needed for VxWorks va_arg */ +#ifdef ERTS_ENABLE_LOCK_COUNT + #include "sys.h" -#ifdef ERTS_ENABLE_LOCK_COUNT +#include "global.h" #include "erl_lock_count.h" -#include "ethread.h" -#include "erl_term.h" -#include "atom.h" -#include <stdio.h> - -/* globals, dont access these without locks or blocks */ +#include "erl_thr_progress.h" -ethr_mutex lcnt_data_lock; -erts_lcnt_data_t *erts_lcnt_data; -Uint16 erts_lcnt_rt_options; -erts_lcnt_time_t timer_start; -const char *str_undefined = "undefined"; +#include "erl_node_tables.h" +#include "erl_alloc_util.h" +#include "erl_check_io.h" +#include "erl_poll.h" +#include "erl_db.h" -static ethr_tsd_key lcnt_thr_data_key; -static int lcnt_n_thr; -static erts_lcnt_thread_data_t *lcnt_thread_data[2048]; +#define LCNT_MAX_CARRIER_ENTRIES 255 -/* local functions */ +/* - Locals that are shared with the header implementation - */ -static ERTS_INLINE void lcnt_lock(void) { - ethr_mutex_lock(&lcnt_data_lock); -} +#ifdef DEBUG +int lcnt_initialization_completed__; +#endif -static ERTS_INLINE void lcnt_unlock(void) { - ethr_mutex_unlock(&lcnt_data_lock); -} +erts_lock_flags_t lcnt_category_mask__; +ethr_tsd_key lcnt_thr_data_key__; -const int log2_tab64[64] = { +const int lcnt_log2_tab64__[64] = { 63, 0, 58, 1, 59, 47, 53, 2, 60, 39, 48, 27, 54, 33, 42, 3, 61, 51, 37, 40, 49, 18, 28, 20, @@ -72,629 +58,624 @@ const int log2_tab64[64] = { 56, 45, 25, 31, 35, 16, 9, 12, 44, 24, 15, 8, 23, 7, 6, 5}; -static ERTS_INLINE int lcnt_log2(Uint64 v) { - v |= v >> 1; - v |= v >> 2; - v |= v >> 4; - v |= v >> 8; - v |= v >> 16; - v |= v >> 32; - return log2_tab64[((Uint64)((v - (v >> 1))*0x07EDD5E59A4E28C2)) >> 58]; -} - -static char* lcnt_lock_type(Uint16 flag) { - switch(flag & ERTS_LCNT_LT_ALL) { - case ERTS_LCNT_LT_SPINLOCK: return "spinlock"; - case ERTS_LCNT_LT_RWSPINLOCK: return "rw_spinlock"; - case ERTS_LCNT_LT_MUTEX: return "mutex"; - case ERTS_LCNT_LT_RWMUTEX: return "rw_mutex"; - case ERTS_LCNT_LT_PROCLOCK: return "proclock"; - default: return ""; - } -} +/* - Local variables - */ -static void lcnt_clear_stats(erts_lcnt_lock_stats_t *stats) { - ethr_atomic_set(&stats->tries, 0); - ethr_atomic_set(&stats->colls, 0); - stats->timer.s = 0; - stats->timer.ns = 0; - stats->timer_n = 0; - stats->file = (char *)str_undefined; - stats->line = 0; - sys_memzero(stats->hist.ns, sizeof(stats->hist.ns)); -} +typedef struct lcnt_static_lock_ref_ { + erts_lcnt_ref_t *reference; -static void lcnt_time(erts_lcnt_time_t *time) { - /* - * erts_sys_hrtime() is the highest resolution - * we could find, it may or may not be monotonic... - */ - ErtsMonotonicTime mtime = erts_sys_hrtime(); - time->s = (unsigned long) (mtime / 1000000000LL); - time->ns = (unsigned long) (mtime - 1000000000LL*time->s); -} + erts_lock_flags_t flags; + const char *name; + Eterm id; -static void lcnt_time_diff(erts_lcnt_time_t *d, erts_lcnt_time_t *t1, erts_lcnt_time_t *t0) { - long ds; - long dns; + struct lcnt_static_lock_ref_ *next; +} lcnt_static_lock_ref_t; - ds = t1->s - t0->s; - dns = t1->ns - t0->ns; +static ethr_atomic_t lcnt_static_lock_registry; - /* the difference should not be able to get bigger than 1 sec in ns*/ +static erts_lcnt_lock_info_list_t lcnt_current_lock_list; +static erts_lcnt_lock_info_list_t lcnt_deleted_lock_list; - if (dns < 0) { - ds -= 1; - dns += 1000000000LL; - } +static erts_lcnt_time_t lcnt_timer_start; - ASSERT(ds >= 0); +static int lcnt_preserve_info; - d->s = ds; - d->ns = dns; -} +/* local functions */ + +static void lcnt_clear_stats(erts_lcnt_lock_info_t *info) { + size_t i; + + for(i = 0; i < ERTS_LCNT_MAX_LOCK_LOCATIONS; i++) { + erts_lcnt_lock_stats_t *stats = &info->location_stats[i]; -/* difference d must be non-negative */ + sys_memzero(&stats->wait_time_histogram, sizeof(stats->wait_time_histogram)); -static void lcnt_time_add(erts_lcnt_time_t *t, erts_lcnt_time_t *d) { - t->s += d->s; - t->ns += d->ns; + stats->total_time_waited.s = 0; + stats->total_time_waited.ns = 0; - t->s += t->ns / 1000000000LL; - t->ns = t->ns % 1000000000LL; + stats->times_waited = 0; + + stats->file = NULL; + stats->line = 0; + + ethr_atomic_set(&stats->attempts, 0); + ethr_atomic_set(&stats->collisions, 0); + } + + info->location_count = 1; } -static erts_lcnt_thread_data_t *lcnt_thread_data_alloc(void) { - erts_lcnt_thread_data_t *eltd; +static lcnt_thread_data_t__ *lcnt_thread_data_alloc(void) { + lcnt_thread_data_t__ *eltd = + (lcnt_thread_data_t__*)malloc(sizeof(lcnt_thread_data_t__)); - eltd = (erts_lcnt_thread_data_t*)malloc(sizeof(erts_lcnt_thread_data_t)); - if (!eltd) { - ERTS_INTERNAL_ERROR("Lock counter failed to allocate memory!"); + if(!eltd) { + ERTS_INTERNAL_ERROR("Failed to allocate lcnt thread data."); } + eltd->timer_set = 0; eltd->lock_in_conflict = 0; - eltd->id = lcnt_n_thr++; - /* set thread data to array */ - lcnt_thread_data[eltd->id] = eltd; - return eltd; -} - -static erts_lcnt_thread_data_t *lcnt_get_thread_data(void) { - return (erts_lcnt_thread_data_t *)ethr_tsd_get(lcnt_thr_data_key); } -/* debug */ +/* - List operations - + * + * Info entries are kept in a doubly linked list where each entry is locked + * with its neighbors rather than a global lock. Deletion is rather quick, but + * insertion is still serial since the head becomes a de facto global lock. + * + * We rely on ad-hoc spinlocks to avoid "recursing" into this module. */ -#if 0 -static char* lock_opt(Uint16 flag) { - if ((flag & ERTS_LCNT_LO_WRITE) && (flag & ERTS_LCNT_LO_READ)) return "rw"; - if (flag & ERTS_LCNT_LO_READ ) return "r "; - if (flag & ERTS_LCNT_LO_WRITE) return " w"; - return "--"; -} +#define LCNT_SPINLOCK_YIELD_ITERATIONS 50 -static void print_lock_x(erts_lcnt_lock_t *lock, Uint16 flag, char *action) { - erts_aint_t w_state, r_state; - char *type; +#define LCNT_SPINLOCK_HELPER_INIT \ + Uint failed_spin_count = 0; - if (strcmp(lock->name, "run_queue") != 0) return; - type = lcnt_lock_type(lock->flag); - r_state = ethr_atomic_read(&lock->r_state); - w_state = ethr_atomic_read(&lock->w_state); +#define LCNT_SPINLOCK_HELPER_YIELD \ + do { \ + failed_spin_count++; \ + if(!(failed_spin_count % LCNT_SPINLOCK_YIELD_ITERATIONS)) { \ + erts_thr_yield(); \ + } else { \ + ERTS_SPIN_BODY; \ + } \ + } while(0) - if (lock->flag & flag) { - erts_fprintf(stderr,"%10s [%24s] [r/w state %4ld/%4ld] %2s id %T\r\n", - action, - lock->name, - r_state, - w_state, - type, - lock->id); - } +static void lcnt_unlock_list_entry(erts_lcnt_lock_info_t *info) { + ethr_atomic32_set_relb(&info->lock, 0); } -#endif - -static erts_lcnt_lock_stats_t *lcnt_get_lock_stats(erts_lcnt_lock_t *lock, char *file, unsigned int line) { - unsigned int i; - erts_lcnt_lock_stats_t *stats = NULL; - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_LOCATION) { - for (i = 0; i < lock->n_stats; i++) { - if ((lock->stats[i].file == file) && (lock->stats[i].line == line)) { - return &(lock->stats[i]); - } - } - if (lock->n_stats < ERTS_LCNT_MAX_LOCK_LOCATIONS) { - stats = &lock->stats[lock->n_stats]; - lock->n_stats++; - stats->file = file; - stats->line = line; - return stats; - } - } - return &lock->stats[0]; +static int lcnt_try_lock_list_entry(erts_lcnt_lock_info_t *info) { + return ethr_atomic32_cmpxchg_acqb(&info->lock, 1, 0) == 0; } -static void lcnt_update_stats_hist(erts_lcnt_hist_t *hist, erts_lcnt_time_t *time_wait) { - int idx; - unsigned long r; +static void lcnt_lock_list_entry(erts_lcnt_lock_info_t *info) { + LCNT_SPINLOCK_HELPER_INIT; - if (time_wait->s > 0 || time_wait->ns > ERTS_LCNT_HISTOGRAM_MAX_NS) { - idx = ERTS_LCNT_HISTOGRAM_SLOT_SIZE - 1; - } else { - r = time_wait->ns >> ERTS_LCNT_HISTOGRAM_RSHIFT; - if (r) idx = lcnt_log2(r); - else idx = 0; + while(!lcnt_try_lock_list_entry(info)) { + LCNT_SPINLOCK_HELPER_YIELD; } - hist->ns[idx]++; } -static void lcnt_update_stats(erts_lcnt_lock_stats_t *stats, int lock_in_conflict, - erts_lcnt_time_t *time_wait) { +static void lcnt_lock_list_entry_with_neighbors(erts_lcnt_lock_info_t *info) { + LCNT_SPINLOCK_HELPER_INIT; - ethr_atomic_inc(&stats->tries); + for(;;) { + if(!lcnt_try_lock_list_entry(info)) + goto retry_after_entry_failed; + if(!lcnt_try_lock_list_entry(info->next)) + goto retry_after_next_failed; + if(!lcnt_try_lock_list_entry(info->prev)) + goto retry_after_prev_failed; - if (lock_in_conflict) - ethr_atomic_inc(&stats->colls); + return; - if (time_wait) { - lcnt_time_add(&(stats->timer), time_wait); - stats->timer_n++; - lcnt_update_stats_hist(&stats->hist,time_wait); + retry_after_prev_failed: + lcnt_unlock_list_entry(info->next); + retry_after_next_failed: + lcnt_unlock_list_entry(info); + retry_after_entry_failed: + LCNT_SPINLOCK_HELPER_YIELD; } } -/* interface */ +static void lcnt_unlock_list_entry_with_neighbors(erts_lcnt_lock_info_t *info) { + lcnt_unlock_list_entry(info->prev); + lcnt_unlock_list_entry(info->next); + lcnt_unlock_list_entry(info); +} + +static void lcnt_insert_list_entry(erts_lcnt_lock_info_list_t *list, erts_lcnt_lock_info_t *info) { + erts_lcnt_lock_info_t *next, *prev; + + prev = &list->head; + + lcnt_lock_list_entry(prev); + + next = prev->next; -void erts_lcnt_init() { - erts_lcnt_thread_data_t *eltd = NULL; + lcnt_lock_list_entry(next); - /* init lock */ - if (ethr_mutex_init(&lcnt_data_lock) != 0) abort(); + info->next = next; + info->prev = prev; - /* init tsd */ - lcnt_n_thr = 0; - ethr_tsd_key_create(&lcnt_thr_data_key,"lcnt_data"); + prev->next = info; + next->prev = info; - lcnt_lock(); + lcnt_unlock_list_entry(next); + lcnt_unlock_list_entry(prev); +} + +static void lcnt_insert_list_carrier(erts_lcnt_lock_info_list_t *list, + erts_lcnt_lock_info_carrier_t *carrier) { + erts_lcnt_lock_info_t *next, *prev; + size_t i; - erts_lcnt_rt_options = ERTS_LCNT_OPT_LOCATION | ERTS_LCNT_OPT_PROCLOCK; - eltd = lcnt_thread_data_alloc(); - ethr_tsd_set(lcnt_thr_data_key, eltd); + for(i = 0; i < carrier->entry_count; i++) { + erts_lcnt_lock_info_t *info = &carrier->entries[i]; - /* init lcnt structure */ - erts_lcnt_data = (erts_lcnt_data_t*)malloc(sizeof(erts_lcnt_data_t)); - if (!erts_lcnt_data) { - ERTS_INTERNAL_ERROR("Lock counter failed to allocate memory!"); + info->prev = &carrier->entries[i - 1]; + info->next = &carrier->entries[i + 1]; } - erts_lcnt_data->current_locks = erts_lcnt_list_init(); - erts_lcnt_data->deleted_locks = erts_lcnt_list_init(); - lcnt_unlock(); + prev = &list->head; + + lcnt_lock_list_entry(prev); + + next = prev->next; + + lcnt_lock_list_entry(next); + + next->prev = &carrier->entries[carrier->entry_count - 1]; + carrier->entries[carrier->entry_count - 1].next = next; + prev->next = &carrier->entries[0]; + carrier->entries[0].prev = prev; + + lcnt_unlock_list_entry(next); + lcnt_unlock_list_entry(prev); } -void erts_lcnt_late_init() { - /* set start timer and zero statistics */ - erts_lcnt_clear_counters(); - erts_thr_install_exit_handler(erts_lcnt_thread_exit_handler); +static void lcnt_init_list(erts_lcnt_lock_info_list_t *list) { + /* Ensure that ref_count operations explode when touching the sentinels in + * DEBUG mode. */ + ethr_atomic_init(&(list->head.ref_count), -1); + ethr_atomic_init(&(list->tail.ref_count), -1); + + ethr_atomic32_init(&(list->head.lock), 0); + (list->head).next = &list->tail; + (list->head).prev = &list->tail; + + ethr_atomic32_init(&(list->tail.lock), 0); + (list->tail).next = &list->head; + (list->tail).prev = &list->head; } -/* list operations */ +/* - Carrier operations - */ -/* BEGIN ASSUMPTION: lcnt_data_lock taken */ +int lcnt_thr_progress_unmanaged_delay__(void) { + return erts_thr_progress_unmanaged_delay(); +} -erts_lcnt_lock_list_t *erts_lcnt_list_init(void) { - erts_lcnt_lock_list_t *list; +void lcnt_thr_progress_unmanaged_continue__(int handle) { + return erts_thr_progress_unmanaged_continue(handle); +} - list = (erts_lcnt_lock_list_t*)malloc(sizeof(erts_lcnt_lock_list_t)); - if (!list) { - ERTS_INTERNAL_ERROR("Lock counter failed to allocate memory!"); - } - list->head = NULL; - list->tail = NULL; - list->n = 0; - return list; +void lcnt_deallocate_carrier__(erts_lcnt_lock_info_carrier_t *carrier) { + ASSERT(ethr_atomic_read(&carrier->ref_count) == 0); + erts_free(ERTS_ALC_T_LCNT_CARRIER, (void*)carrier); } -/* only do this on the list with the deleted locks! */ -void erts_lcnt_list_clear(erts_lcnt_lock_list_t *list) { - erts_lcnt_lock_t *lock = NULL, - *next = NULL; +static void lcnt_thr_prg_cleanup_carrier(void *data) { + erts_lcnt_lock_info_carrier_t *carrier = data; + size_t entry_count, i; - lock = list->head; + /* carrier->entry_count will be replaced with garbage if it's deallocated + * on the final iteration, so we'll tuck it away to get a clean exit. */ + entry_count = carrier->entry_count; - while(lock != NULL) { - next = lock->next; - free(lock); - lock = next; - } + for(i = 0; i < entry_count; i++) { + ASSERT(ethr_atomic_read(&carrier->ref_count) >= (entry_count - i)); - list->head = NULL; - list->tail = NULL; - list->n = 0; + erts_lcnt_release_lock_info(&carrier->entries[i]); + } } -void erts_lcnt_list_insert(erts_lcnt_lock_list_t *list, erts_lcnt_lock_t *lock) { - erts_lcnt_lock_t *tail = NULL; +static void lcnt_schedule_carrier_cleanup(void *data) { + ErtsSchedulerData *esdp = erts_get_scheduler_data(); - tail = list->tail; - if (tail) { - tail->next = lock; - lock->prev = tail; + /* We can't issue cleanup jobs on anything other than normal schedulers, so + * we move to the first scheduler if required. */ + + if(!esdp || esdp->type != ERTS_SCHED_NORMAL) { + erts_schedule_misc_aux_work(1, &lcnt_schedule_carrier_cleanup, data); } else { - list->head = lock; - lock->prev = NULL; - ASSERT(!lock->next); + erts_lcnt_lock_info_carrier_t *carrier = data; + size_t carrier_size; + + carrier_size = sizeof(erts_lcnt_lock_info_carrier_t) + + sizeof(erts_lcnt_lock_info_t) * carrier->entry_count; + + erts_schedule_thr_prgr_later_cleanup_op(&lcnt_thr_prg_cleanup_carrier, + data, (ErtsThrPrgrLaterOp*)&carrier->release_entries, carrier_size); } - lock->next = NULL; - list->tail = lock; +} - list->n++; +static void lcnt_info_deallocate(erts_lcnt_lock_info_t *info) { + lcnt_release_carrier__(info->carrier); } -void erts_lcnt_list_delete(erts_lcnt_lock_list_t *list, erts_lcnt_lock_t *lock) { - if (lock->next) lock->next->prev = lock->prev; - if (lock->prev) lock->prev->next = lock->next; - if (list->head == lock) list->head = lock->next; - if (list->tail == lock) list->tail = lock->prev; +static void lcnt_info_dispose(erts_lcnt_lock_info_t *info) { + ASSERT(ethr_atomic_read(&info->ref_count) == 0); + + if(lcnt_preserve_info) { + ethr_atomic_set(&info->ref_count, 1); - lock->prev = NULL; - lock->next = NULL; - list->n--; + /* Move straight to deallocation the next time around. */ + info->dispose = &lcnt_info_deallocate; + + lcnt_insert_list_entry(&lcnt_deleted_lock_list, info); + } else { + lcnt_info_deallocate(info); + } } -/* END ASSUMPTION: lcnt_data_lock taken */ +static void lcnt_lock_info_init_helper(erts_lcnt_lock_info_t *info) { + ethr_atomic_init(&info->ref_count, 1); + ethr_atomic32_init(&info->lock, 0); -/* lock operations */ + ethr_atomic_init(&info->r_state, 0); + ethr_atomic_init(&info->w_state, 0); -/* interface to erl_threads.h */ -/* only lock on init and destroy, all others should use atomics */ -void erts_lcnt_init_lock(erts_lcnt_lock_t *lock, char *name, Uint16 flag ) { - erts_lcnt_init_lock_x(lock, name, flag, NIL); + info->dispose = &lcnt_info_dispose; + + lcnt_clear_stats(info); } -void erts_lcnt_init_lock_x(erts_lcnt_lock_t *lock, char *name, Uint16 flag, Eterm id) { - int i; - if (name == NULL) { ERTS_LCNT_CLEAR_FLAG(lock); return; } - lcnt_lock(); +erts_lcnt_lock_info_carrier_t *erts_lcnt_create_lock_info_carrier(int entry_count) { + erts_lcnt_lock_info_carrier_t *result; + size_t carrier_size, i; - lock->next = NULL; - lock->prev = NULL; - lock->flag = flag; - lock->name = name; - lock->id = id; + ASSERT(entry_count > 0 && entry_count <= LCNT_MAX_CARRIER_ENTRIES); + ASSERT(lcnt_initialization_completed__); - ethr_atomic_init(&lock->r_state, 0); - ethr_atomic_init(&lock->w_state, 0); -#ifdef DEBUG - ethr_atomic_init(&lock->flowstate, 0); -#endif + carrier_size = sizeof(erts_lcnt_lock_info_carrier_t) + + sizeof(erts_lcnt_lock_info_t) * entry_count; - lock->n_stats = 1; + result = (erts_lcnt_lock_info_carrier_t*)erts_alloc(ERTS_ALC_T_LCNT_CARRIER, carrier_size); + result->entry_count = entry_count; - for (i = 0; i < ERTS_LCNT_MAX_LOCK_LOCATIONS; i++) { - lcnt_clear_stats(&lock->stats[i]); - } + ethr_atomic_init(&result->ref_count, entry_count); - erts_lcnt_list_insert(erts_lcnt_data->current_locks, lock); - lcnt_unlock(); -} -/* init empty, instead of zero struct */ -void erts_lcnt_init_lock_empty(erts_lcnt_lock_t *lock) { - lock->next = NULL; - lock->prev = NULL; - lock->flag = 0; - lock->name = NULL; - lock->id = NIL; - ethr_atomic_init(&lock->r_state, 0); - ethr_atomic_init(&lock->w_state, 0); -#ifdef DEBUG - ethr_atomic_init(&lock->flowstate, 0); -#endif - lock->n_stats = 0; - sys_memzero(lock->stats, sizeof(lock->stats)); -} -/* destroy lock */ -void erts_lcnt_destroy_lock(erts_lcnt_lock_t *lock) { - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; - lcnt_lock(); - - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_COPYSAVE) { - erts_lcnt_lock_t *deleted_lock; - /* copy structure and insert the copy */ - deleted_lock = (erts_lcnt_lock_t*)malloc(sizeof(erts_lcnt_lock_t)); - if (!deleted_lock) { - ERTS_INTERNAL_ERROR("Lock counter failed to allocate memory!"); - } - memcpy(deleted_lock, lock, sizeof(erts_lcnt_lock_t)); - deleted_lock->next = NULL; - deleted_lock->prev = NULL; - erts_lcnt_list_insert(erts_lcnt_data->deleted_locks, deleted_lock); + for(i = 0; i < entry_count; i++) { + erts_lcnt_lock_info_t *info = &result->entries[i]; + + lcnt_lock_info_init_helper(info); + + info->carrier = result; } - /* delete original */ - erts_lcnt_list_delete(erts_lcnt_data->current_locks, lock); - ERTS_LCNT_CLEAR_FLAG(lock); - lcnt_unlock(); + return result; } -/* lock */ +void erts_lcnt_install(erts_lcnt_ref_t *ref, erts_lcnt_lock_info_carrier_t *carrier) { + ethr_sint_t swapped_carrier; -void erts_lcnt_lock_opt(erts_lcnt_lock_t *lock, Uint16 option) { - erts_aint_t r_state = 0, w_state = 0; - erts_lcnt_thread_data_t *eltd; +#ifdef DEBUG + int i; - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; + /* Verify that all locks share the same categories/static property; all + * other flags are fair game. */ + for(i = 1; i < carrier->entry_count; i++) { + const erts_lock_flags_t SIGNIFICANT_DIFF_MASK = + ERTS_LOCK_FLAGS_MASK_CATEGORY | ERTS_LOCK_FLAGS_PROPERTY_STATIC; - eltd = lcnt_get_thread_data(); - ASSERT(eltd); + erts_lcnt_lock_info_t *previous, *current; - w_state = ethr_atomic_read(&lock->w_state); + previous = &carrier->entries[i - 1]; + current = &carrier->entries[i]; - if (option & ERTS_LCNT_LO_WRITE) { - r_state = ethr_atomic_read(&lock->r_state); - ethr_atomic_inc( &lock->w_state); - } - if (option & ERTS_LCNT_LO_READ) { - ethr_atomic_inc( &lock->r_state); + ASSERT(!((previous->flags ^ current->flags) & SIGNIFICANT_DIFF_MASK)); } +#endif - /* we cannot acquire w_lock if either w or r are taken */ - /* we cannot acquire r_lock if w_lock is taken */ + swapped_carrier = ethr_atomic_cmpxchg_mb(ref, (ethr_sint_t)carrier, (ethr_sint_t)NULL); - if ((w_state > 0) || (r_state > 0)) { - eltd->lock_in_conflict = 1; - if (eltd->timer_set == 0) { - lcnt_time(&eltd->timer); - } - eltd->timer_set++; + if(swapped_carrier != (ethr_sint_t)NULL) { +#ifdef DEBUG + ASSERT(ethr_atomic_read(&carrier->ref_count) == carrier->entry_count); + ethr_atomic_set(&carrier->ref_count, 0); +#endif + + lcnt_deallocate_carrier__(carrier); } else { - eltd->lock_in_conflict = 0; + lcnt_insert_list_carrier(&lcnt_current_lock_list, carrier); } } -void erts_lcnt_lock(erts_lcnt_lock_t *lock) { - erts_aint_t w_state; - erts_lcnt_thread_data_t *eltd; +void erts_lcnt_uninstall(erts_lcnt_ref_t *ref) { + ethr_sint_t previous_carrier, swapped_carrier; - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; + previous_carrier = ethr_atomic_read(ref); + swapped_carrier = ethr_atomic_cmpxchg_mb(ref, (ethr_sint_t)NULL, previous_carrier); - w_state = ethr_atomic_read(&lock->w_state); - ethr_atomic_inc(&lock->w_state); - eltd = lcnt_get_thread_data(); + if(previous_carrier && previous_carrier == swapped_carrier) { + lcnt_schedule_carrier_cleanup((void*)previous_carrier); + } +} - ASSERT(eltd); +/* - Static lock registry - + * + * Since static locks can be trusted to never disappear, we can track them + * pretty cheaply and won't need to bother writing an "erts_lcnt_update_xx" + * variant. */ - if (w_state > 0) { - eltd->lock_in_conflict = 1; - /* only set the timer if nobody else has it - * This should only happen when proc_locks aquires several locks - * 'atomicly'. All other locks will block the thread if w_state > 0 - * i.e. locked. - */ - if (eltd->timer_set == 0) { - lcnt_time(&eltd->timer); +static void lcnt_init_static_lock_registry(void) { + ethr_atomic_init(&lcnt_static_lock_registry, (ethr_sint_t)NULL); +} + +static void lcnt_update_static_locks(void) { + lcnt_static_lock_ref_t *iterator = + (lcnt_static_lock_ref_t*)ethr_atomic_read(&lcnt_static_lock_registry); + + while(iterator != NULL) { + if(!erts_lcnt_check_enabled(iterator->flags)) { + erts_lcnt_uninstall(iterator->reference); + } else if(!erts_lcnt_check_ref_installed(iterator->reference)) { + erts_lcnt_lock_info_carrier_t *carrier = erts_lcnt_create_lock_info_carrier(1); + + erts_lcnt_init_lock_info_idx(carrier, 0, iterator->name, iterator->id, iterator->flags); + + erts_lcnt_install(iterator->reference, carrier); } - eltd->timer_set++; - } else { - eltd->lock_in_conflict = 0; + + iterator = iterator->next; } } -/* if a lock wasn't really a lock operation, bad bad process locks */ +void lcnt_register_static_lock__(erts_lcnt_ref_t *reference, const char *name, Eterm id, + erts_lock_flags_t flags) { + lcnt_static_lock_ref_t *lock = malloc(sizeof(lcnt_static_lock_ref_t)); + int retry_insertion; + + ASSERT(flags & ERTS_LOCK_FLAGS_PROPERTY_STATIC); + + lock->reference = reference; + lock->flags = flags; + lock->name = name; + lock->id = id; + + do { + ethr_sint_t swapped_head; + + lock->next = (lcnt_static_lock_ref_t*)ethr_atomic_read(&lcnt_static_lock_registry); -void erts_lcnt_lock_unaquire(erts_lcnt_lock_t *lock) { - /* should check if this thread was "waiting" */ - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; + swapped_head = ethr_atomic_cmpxchg_acqb( + &lcnt_static_lock_registry, + (ethr_sint_t)lock, + (ethr_sint_t)lock->next); - ethr_atomic_dec(&lock->w_state); + retry_insertion = (swapped_head != (ethr_sint_t)lock->next); + } while(retry_insertion); } -/* - * erts_lcnt_lock_post - * - * Used when we get a lock (i.e. directly after a lock operation) - * if the timer was set then we had to wait for the lock - * lock_post will calculate the wait time. - */ +/* - Initialization - */ + +void erts_lcnt_pre_thr_init() { + /* Ensure that the dependency hack mentioned in the header doesn't + * explode at runtime. */ + ERTS_CT_ASSERT(sizeof(LcntThrPrgrLaterOp) >= sizeof(ErtsThrPrgrLaterOp)); + ERTS_CT_ASSERT(ERTS_THR_PRGR_DHANDLE_MANAGED == + (ErtsThrPrgrDelayHandle)LCNT_THR_PRGR_DHANDLE_MANAGED); + + lcnt_init_list(&lcnt_current_lock_list); + lcnt_init_list(&lcnt_deleted_lock_list); -void erts_lcnt_lock_post(erts_lcnt_lock_t *lock) { - erts_lcnt_lock_post_x(lock, (char*)str_undefined, 0); + lcnt_init_static_lock_registry(); } -void erts_lcnt_lock_post_x(erts_lcnt_lock_t *lock, char *file, unsigned int line) { - erts_lcnt_thread_data_t *eltd; - erts_lcnt_time_t timer; - erts_lcnt_time_t time_wait; - erts_lcnt_lock_stats_t *stats; -#ifdef DEBUG - erts_aint_t flowstate; -#endif +void erts_lcnt_post_thr_init() { + /* ASSUMPTION: this is safe since it runs prior to the creation of other + * threads (Directly after ethread init). */ + + ethr_tsd_key_create(&lcnt_thr_data_key__, "lcnt_data"); - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; + erts_lcnt_thread_setup(); +} + +void erts_lcnt_late_init() { + /* Set start timer and zero all statistics */ + erts_lcnt_clear_counters(); + erts_thr_install_exit_handler(erts_lcnt_thread_exit_handler); #ifdef DEBUG - if (!(lock->flag & (ERTS_LCNT_LT_RWMUTEX | ERTS_LCNT_LT_RWSPINLOCK))) { - flowstate = ethr_atomic_read(&lock->flowstate); - ASSERT(flowstate == 0); - ethr_atomic_inc(&lock->flowstate); - } + /* It's safe to use erts_alloc and thread progress past this point. */ + lcnt_initialization_completed__ = 1; #endif +} - eltd = lcnt_get_thread_data(); - - ASSERT(eltd); +void erts_lcnt_post_startup(void) { + /* Default to capturing everything to match the behavior of the old lock + * counter build. */ + erts_lcnt_set_category_mask(ERTS_LOCK_FLAGS_MASK_CATEGORY); +} - /* if lock was in conflict, time it */ - stats = lcnt_get_lock_stats(lock, file, line); - if (eltd->timer_set) { - lcnt_time(&timer); +void erts_lcnt_thread_setup() { + lcnt_thread_data_t__ *eltd = lcnt_thread_data_alloc(); - lcnt_time_diff(&time_wait, &timer, &(eltd->timer)); - lcnt_update_stats(stats, eltd->lock_in_conflict, &time_wait); - eltd->timer_set--; - ASSERT(eltd->timer_set >= 0); - } else { - lcnt_update_stats(stats, eltd->lock_in_conflict, NULL); - } + ASSERT(eltd); + ethr_tsd_set(lcnt_thr_data_key__, eltd); } -/* unlock */ +void erts_lcnt_thread_exit_handler() { + lcnt_thread_data_t__ *eltd = lcnt_get_thread_data__(); -void erts_lcnt_unlock_opt(erts_lcnt_lock_t *lock, Uint16 option) { - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; - if (option & ERTS_LCNT_LO_WRITE) ethr_atomic_dec(&lock->w_state); - if (option & ERTS_LCNT_LO_READ ) ethr_atomic_dec(&lock->r_state); + if (eltd) { + free(eltd); + } } -void erts_lcnt_unlock(erts_lcnt_lock_t *lock) { - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; +/* - BIF interface - */ + +void erts_lcnt_retain_lock_info(erts_lcnt_lock_info_t *info) { #ifdef DEBUG - { - erts_aint_t w_state; - erts_aint_t flowstate; - - /* flowstate */ - flowstate = ethr_atomic_read(&lock->flowstate); - ASSERT(flowstate == 1); - ethr_atomic_dec(&lock->flowstate); - - /* write state */ - w_state = ethr_atomic_read(&lock->w_state); - ASSERT(w_state > 0); - } + ASSERT(ethr_atomic_inc_read_acqb(&info->ref_count) >= 2); +#else + ethr_atomic_inc_acqb(&info->ref_count); #endif - ethr_atomic_dec(&lock->w_state); } -/* trylock */ +void erts_lcnt_release_lock_info(erts_lcnt_lock_info_t *info) { + ethr_sint_t count; -void erts_lcnt_trylock_opt(erts_lcnt_lock_t *lock, int res, Uint16 option) { - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; - /* Determine lock_state via res instead of state */ - if (res != EBUSY) { - if (option & ERTS_LCNT_LO_WRITE) ethr_atomic_inc(&lock->w_state); - if (option & ERTS_LCNT_LO_READ ) ethr_atomic_inc(&lock->r_state); - lcnt_update_stats(&(lock->stats[0]), 0, NULL); + /* We need to acquire the lock before decrementing ref_count to avoid + * racing with list iteration; there's a short window between reading the + * reference to info and increasing its ref_count. */ + lcnt_lock_list_entry_with_neighbors(info); + + count = ethr_atomic_dec_read(&info->ref_count); + + ASSERT(count >= 0); + + if(count > 0) { + lcnt_unlock_list_entry_with_neighbors(info); } else { - ethr_atomic_inc(&lock->stats[0].tries); - ethr_atomic_inc(&lock->stats[0].colls); + (info->next)->prev = info->prev; + (info->prev)->next = info->next; + + lcnt_unlock_list_entry_with_neighbors(info); + + info->dispose(info); } } +erts_lock_flags_t erts_lcnt_get_category_mask() { + return lcnt_category_mask__; +} -void erts_lcnt_trylock(erts_lcnt_lock_t *lock, int res) { - /* Determine lock_state via res instead of state */ - if (erts_lcnt_rt_options & ERTS_LCNT_OPT_SUSPEND) return; - if (ERTS_LCNT_IS_LOCK_INVALID(lock)) return; - if (res != EBUSY) { -#ifdef DEBUG - { - erts_aint_t flowstate; - flowstate = ethr_atomic_read(&lock->flowstate); - ASSERT(flowstate == 0); - ethr_atomic_inc( &lock->flowstate); - } +#ifdef ERTS_ENABLE_KERNEL_POLL +/* erl_poll/erl_check_io only exports one of these variants at a time, and we + * may need to use either one depending on emulator startup flags. */ +void erts_lcnt_update_pollset_locks_nkp(int); +void erts_lcnt_update_pollset_locks_kp(int); + +void erts_lcnt_update_cio_locks_nkp(int); +void erts_lcnt_update_cio_locks_kp(int); #endif - ethr_atomic_inc(&lock->w_state); - lcnt_update_stats(&(lock->stats[0]), 0, NULL); - } else { - ethr_atomic_inc(&lock->stats[0].tries); - ethr_atomic_inc(&lock->stats[0].colls); + +void erts_lcnt_set_category_mask(erts_lock_flags_t mask) { + erts_lock_flags_t changed_categories; + + ASSERT(!(mask & ~ERTS_LOCK_FLAGS_MASK_CATEGORY)); + ASSERT(lcnt_initialization_completed__); + + changed_categories = (lcnt_category_mask__ ^ mask); + lcnt_category_mask__ = mask; + + if(changed_categories) { + lcnt_update_static_locks(); } -} -/* thread operations */ + if(changed_categories & ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION) { + erts_lcnt_update_distribution_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION); + } -void erts_lcnt_thread_setup(void) { - erts_lcnt_thread_data_t *eltd; + if(changed_categories & ERTS_LOCK_FLAGS_CATEGORY_ALLOCATOR) { + erts_lcnt_update_allocator_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_ALLOCATOR); + } - lcnt_lock(); - /* lock for thread id global update */ - eltd = lcnt_thread_data_alloc(); - lcnt_unlock(); - ASSERT(eltd); - ethr_tsd_set(lcnt_thr_data_key, eltd); -} + if(changed_categories & ERTS_LOCK_FLAGS_CATEGORY_PROCESS) { + erts_lcnt_update_process_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_PROCESS); + } -void erts_lcnt_thread_exit_handler() { - erts_lcnt_thread_data_t *eltd; + if(changed_categories & ERTS_LOCK_FLAGS_CATEGORY_IO) { +#ifdef ERTS_ENABLE_KERNEL_POLL + if(erts_use_kernel_poll) { + erts_lcnt_update_pollset_locks_kp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + erts_lcnt_update_cio_locks_kp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + } else { + erts_lcnt_update_pollset_locks_nkp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + erts_lcnt_update_cio_locks_nkp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + } +#else + erts_lcnt_update_pollset_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + erts_lcnt_update_cio_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); +#endif - eltd = ethr_tsd_get(lcnt_thr_data_key); + erts_lcnt_update_driver_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + erts_lcnt_update_port_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO); + } - if (eltd) { - free(eltd); + if(changed_categories & ERTS_LOCK_FLAGS_CATEGORY_DB) { + erts_lcnt_update_db_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_DB); } } -/* bindings for bifs */ - -Uint16 erts_lcnt_set_rt_opt(Uint16 opt) { - Uint16 prev; - prev = (erts_lcnt_rt_options & opt); - erts_lcnt_rt_options |= opt; - return prev; +void erts_lcnt_set_preserve_info(int enable) { + lcnt_preserve_info = enable; } -Uint16 erts_lcnt_clear_rt_opt(Uint16 opt) { - Uint16 prev; - prev = (erts_lcnt_rt_options & opt); - erts_lcnt_rt_options &= ~opt; - return prev; +int erts_lcnt_get_preserve_info() { + return lcnt_preserve_info; } void erts_lcnt_clear_counters(void) { - erts_lcnt_lock_t *lock; - erts_lcnt_lock_list_t *list; - erts_lcnt_lock_stats_t *stats; - int i; + erts_lcnt_lock_info_t *iterator; - lcnt_lock(); + lcnt_time__(&lcnt_timer_start); - list = erts_lcnt_data->current_locks; + iterator = NULL; + while(erts_lcnt_iterate_list(&lcnt_current_lock_list, &iterator)) { + lcnt_clear_stats(iterator); + } - for (lock = list->head; lock != NULL; lock = lock->next) { - for( i = 0; i < ERTS_LCNT_MAX_LOCK_LOCATIONS; i++) { - stats = &lock->stats[i]; - lcnt_clear_stats(stats); - } - lock->n_stats = 1; + iterator = NULL; + while(erts_lcnt_iterate_list(&lcnt_deleted_lock_list, &iterator)) { + erts_lcnt_release_lock_info(iterator); } +} + +erts_lcnt_data_t erts_lcnt_get_data(void) { + erts_lcnt_time_t timer_stop; + erts_lcnt_data_t result; - /* empty deleted locks in lock list */ - erts_lcnt_list_clear(erts_lcnt_data->deleted_locks); + lcnt_time__(&timer_stop); - lcnt_time(&timer_start); + result.timer_start = lcnt_timer_start; - lcnt_unlock(); + result.current_locks = &lcnt_current_lock_list; + result.deleted_locks = &lcnt_deleted_lock_list; + + lcnt_time_diff__(&result.duration, &timer_stop, &result.timer_start); + + return result; } -erts_lcnt_data_t *erts_lcnt_get_data(void) { - erts_lcnt_time_t timer_stop; +int erts_lcnt_iterate_list(erts_lcnt_lock_info_list_t *list, erts_lcnt_lock_info_t **iterator) { + erts_lcnt_lock_info_t *current, *next; - lcnt_lock(); + current = *iterator ? *iterator : &list->head; - lcnt_time(&timer_stop); - lcnt_time_diff(&(erts_lcnt_data->duration), &timer_stop, &timer_start); + ASSERT(current != &list->tail); - lcnt_unlock(); + lcnt_lock_list_entry(current); - return erts_lcnt_data; -} + next = current->next; + + if(next != &list->tail) { + erts_lcnt_retain_lock_info(next); + } + + lcnt_unlock_list_entry(current); + + if(current != &list->head) { + erts_lcnt_release_lock_info(current); + } + + *iterator = next; -char *erts_lcnt_lock_type(Uint16 type) { - return lcnt_lock_type(type); + return next != &list->tail; } -#endif /* ifdef ERTS_ENABLE_LOCK_COUNT */ +#endif /* #ifdef ERTS_ENABLE_LOCK_COUNT */ |