project-base/server-skynet-source-3rd-jemalloc
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Introduce lockedint module.

Browse Source 
This pulls out the various abstractions where some stats counter is sometimes an
atomic, sometimes a plain variable, sometimes always protected by a lock,
sometimes protected by reads but not writes, etc.  With this change, these cases
are treated consistently, and access patterns tagged.

In the process, we fix a few missed-update bugs (where one caller assumes
"protected-by-a-lock" semantics and another does not).
This commit is contained in:
David Goldblatt 2020-03-08 20:43:41 -07:00 committed by David Goldblatt
parent acd0bf6a26
commit 356aaa7dc6
7 changed files with 341 additions and 290 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
include/jemalloc/internal/arena_inlines_b.h
View File

@ -148,14 +148,14 @@ arena_decay_extent(tsdn_t *tsdn,arena_t *arena, ehooks_t *ehooks,
extent_dalloc_wrapper(tsdn, arena, ehooks, edata); extent_dalloc_wrapper(tsdn, arena, ehooks, edata);
if (config_stats) { if (config_stats) {
/* Update stats accordingly. */ /* Update stats accordingly. */
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_stats_add_u64(tsdn, &arena->stats, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->decay_dirty.stats->nmadvise, 1); &arena->decay_dirty.stats->nmadvise, 1);
arena_stats_add_u64(tsdn, &arena->stats, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->decay_dirty.stats->purged, extent_size >> LG_PAGE); &arena->decay_dirty.stats->purged, extent_size >> LG_PAGE);
arena_stats_sub_zu(tsdn, &arena->stats, &arena->stats.mapped, locked_dec_zu(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
extent_size); &arena->stats.mapped, extent_size);
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
} }
} }

179
include/jemalloc/internal/arena_stats.h
View File

@ -2,6 +2,7 @@
#define JEMALLOC_INTERNAL_ARENA_STATS_H #define JEMALLOC_INTERNAL_ARENA_STATS_H
#include "jemalloc/internal/atomic.h" #include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/lockedint.h"
#include "jemalloc/internal/mutex.h" #include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/mutex_prof.h" #include "jemalloc/internal/mutex_prof.h"
#include "jemalloc/internal/pa.h" #include "jemalloc/internal/pa.h"
@ -9,40 +10,28 @@
JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
/*
* In those architectures that support 64-bit atomics, we use atomic updates for
* our 64-bit values. Otherwise, we use a plain uint64_t and synchronize
* externally.
*/
#ifdef JEMALLOC_ATOMIC_U64
typedef atomic_u64_t arena_stats_u64_t;
#else
/* Must hold the arena stats mutex while reading atomically. */
typedef uint64_t arena_stats_u64_t;
#endif
typedef struct arena_stats_large_s arena_stats_large_t; typedef struct arena_stats_large_s arena_stats_large_t;
struct arena_stats_large_s { struct arena_stats_large_s {
/* /*
* Total number of allocation/deallocation requests served directly by * Total number of allocation/deallocation requests served directly by
* the arena. * the arena.
*/ */
arena_stats_u64_t nmalloc; locked_u64_t nmalloc;
arena_stats_u64_t ndalloc; locked_u64_t ndalloc;
/* /*
* Number of allocation requests that correspond to this size class. * Number of allocation requests that correspond to this size class.
* This includes requests served by tcache, though tcache only * This includes requests served by tcache, though tcache only
* periodically merges into this counter. * periodically merges into this counter.
*/ */
arena_stats_u64_t nrequests; /* Partially derived. */ locked_u64_t nrequests; /* Partially derived. */
/* /*
* Number of tcache fills / flushes for large (similarly, periodically * Number of tcache fills / flushes for large (similarly, periodically
* merged). Note that there is no large tcache batch-fill currently * merged). Note that there is no large tcache batch-fill currently
* (i.e. only fill 1 at a time); however flush may be batched. * (i.e. only fill 1 at a time); however flush may be batched.
*/ */
arena_stats_u64_t nfills; /* Partially derived. */ locked_u64_t nfills; /* Partially derived. */
arena_stats_u64_t nflushes; /* Partially derived. */ locked_u64_t nflushes; /* Partially derived. */
/* Current number of allocations of this size class. */ /* Current number of allocations of this size class. */
size_t curlextents; /* Derived. */ size_t curlextents; /* Derived. */
@ -51,11 +40,11 @@ struct arena_stats_large_s {
typedef struct arena_stats_decay_s arena_stats_decay_t; typedef struct arena_stats_decay_s arena_stats_decay_t;
struct arena_stats_decay_s { struct arena_stats_decay_s {
/* Total number of purge sweeps. */ /* Total number of purge sweeps. */
arena_stats_u64_t npurge; locked_u64_t npurge;
/* Total number of madvise calls made. */ /* Total number of madvise calls made. */
arena_stats_u64_t nmadvise; locked_u64_t nmadvise;
/* Total number of pages purged. */ /* Total number of pages purged. */
arena_stats_u64_t purged; locked_u64_t purged;
}; };
typedef struct arena_stats_extents_s arena_stats_extents_t; typedef struct arena_stats_extents_s arena_stats_extents_t;
@ -81,19 +70,19 @@ struct arena_stats_extents_s {
*/ */
typedef struct arena_stats_s arena_stats_t; typedef struct arena_stats_s arena_stats_t;
struct arena_stats_s { struct arena_stats_s {
#ifndef JEMALLOC_ATOMIC_U64 LOCKEDINT_MTX_DECLARE(mtx)
malloc_mutex_t mtx;
#endif
/* Number of bytes currently mapped, excluding retained memory. */ /*
atomic_zu_t mapped; /* Partially derived. */ * Number of bytes currently mapped, excluding retained memory.
*/
locked_zu_t mapped; /* Partially derived. */
/* /*
* Number of unused virtual memory bytes currently retained. Retained * Number of unused virtual memory bytes currently retained. Retained
* bytes are technically mapped (though always decommitted or purged), * bytes are technically mapped (though always decommitted or purged),
* but they are excluded from the mapped statistic (above). * but they are excluded from the mapped statistic (above).
*/ */
atomic_zu_t retained; /* Derived. */ locked_zu_t retained; /* Derived. */
/* Number of edata_t structs allocated by base, but not being used. */ /* Number of edata_t structs allocated by base, but not being used. */
atomic_zu_t edata_avail; atomic_zu_t edata_avail;
@ -107,11 +96,11 @@ struct arena_stats_s {
atomic_zu_t metadata_thp; atomic_zu_t metadata_thp;
atomic_zu_t allocated_large; /* Derived. */ atomic_zu_t allocated_large; /* Derived. */
arena_stats_u64_t nmalloc_large; /* Derived. */ locked_u64_t nmalloc_large; /* Derived. */
arena_stats_u64_t ndalloc_large; /* Derived. */ locked_u64_t ndalloc_large; /* Derived. */
arena_stats_u64_t nfills_large; /* Derived. */ locked_u64_t nfills_large; /* Derived. */
arena_stats_u64_t nflushes_large; /* Derived. */ locked_u64_t nflushes_large; /* Derived. */
arena_stats_u64_t nrequests_large; /* Derived. */ locked_u64_t nrequests_large; /* Derived. */
/* /*
* The stats logically owned by the pa_shard in the same arena. This * The stats logically owned by the pa_shard in the same arena. This
@ -139,138 +128,32 @@ arena_stats_init(tsdn_t *tsdn, arena_stats_t *arena_stats) {
assert(((char *)arena_stats)[i] == 0); assert(((char *)arena_stats)[i] == 0);
} }
} }
#ifndef JEMALLOC_ATOMIC_U64 if (LOCKEDINT_MTX_INIT(LOCKEDINT_MTX(arena_stats->mtx), "arena_stats",
if (malloc_mutex_init(&arena_stats->mtx, "arena_stats",
WITNESS_RANK_ARENA_STATS, malloc_mutex_rank_exclusive)) { WITNESS_RANK_ARENA_STATS, malloc_mutex_rank_exclusive)) {
return true; return true;
} }
#endif
/* Memory is zeroed, so there is no need to clear stats. */ /* Memory is zeroed, so there is no need to clear stats. */
return false; return false;
} }
static inline void
arena_stats_lock(tsdn_t *tsdn, arena_stats_t *arena_stats) {
#ifndef JEMALLOC_ATOMIC_U64
malloc_mutex_lock(tsdn, &arena_stats->mtx);
#endif
}
static inline void
arena_stats_unlock(tsdn_t *tsdn, arena_stats_t *arena_stats) {
#ifndef JEMALLOC_ATOMIC_U64
malloc_mutex_unlock(tsdn, &arena_stats->mtx);
#endif
}
static inline uint64_t
arena_stats_read_u64(tsdn_t *tsdn, arena_stats_t *arena_stats,
arena_stats_u64_t *p) {
#ifdef JEMALLOC_ATOMIC_U64
return atomic_load_u64(p, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
return *p;
#endif
}
static inline void
arena_stats_add_u64(tsdn_t *tsdn, arena_stats_t *arena_stats,
arena_stats_u64_t *p, uint64_t x) {
#ifdef JEMALLOC_ATOMIC_U64
atomic_fetch_add_u64(p, x, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
*p += x;
#endif
}
static inline void
arena_stats_sub_u64(tsdn_t *tsdn, arena_stats_t *arena_stats,
arena_stats_u64_t *p, uint64_t x) {
#ifdef JEMALLOC_ATOMIC_U64
uint64_t r = atomic_fetch_sub_u64(p, x, ATOMIC_RELAXED);
assert(r - x <= r);
#else
malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
*p -= x;
assert(*p + x >= *p);
#endif
}
/*
* Non-atomically sets *dst += src. *dst needs external synchronization.
* This lets us avoid the cost of a fetch_add when its unnecessary (note that
* the types here are atomic).
*/
static inline void
arena_stats_accum_u64(arena_stats_u64_t *dst, uint64_t src) {
#ifdef JEMALLOC_ATOMIC_U64
uint64_t cur_dst = atomic_load_u64(dst, ATOMIC_RELAXED);
atomic_store_u64(dst, src + cur_dst, ATOMIC_RELAXED);
#else
*dst += src;
#endif
}
static inline size_t
arena_stats_read_zu(tsdn_t *tsdn, arena_stats_t *arena_stats,
atomic_zu_t *p) {
#ifdef JEMALLOC_ATOMIC_U64
return atomic_load_zu(p, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
return atomic_load_zu(p, ATOMIC_RELAXED);
#endif
}
static inline void
arena_stats_add_zu(tsdn_t *tsdn, arena_stats_t *arena_stats,
atomic_zu_t *p, size_t x) {
#ifdef JEMALLOC_ATOMIC_U64
atomic_fetch_add_zu(p, x, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
size_t cur = atomic_load_zu(p, ATOMIC_RELAXED);
atomic_store_zu(p, cur + x, ATOMIC_RELAXED);
#endif
}
static inline void
arena_stats_sub_zu(tsdn_t *tsdn, arena_stats_t *arena_stats,
atomic_zu_t *p, size_t x) {
#ifdef JEMALLOC_ATOMIC_U64
size_t r = atomic_fetch_sub_zu(p, x, ATOMIC_RELAXED);
assert(r - x <= r);
#else
malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
size_t cur = atomic_load_zu(p, ATOMIC_RELAXED);
atomic_store_zu(p, cur - x, ATOMIC_RELAXED);
#endif
}
/* Like the _u64 variant, needs an externally synchronized *dst. */
static inline void
arena_stats_accum_zu(atomic_zu_t *dst, size_t src) {
size_t cur_dst = atomic_load_zu(dst, ATOMIC_RELAXED);
atomic_store_zu(dst, src + cur_dst, ATOMIC_RELAXED);
}
static inline void static inline void
arena_stats_large_flush_nrequests_add(tsdn_t *tsdn, arena_stats_t *arena_stats, arena_stats_large_flush_nrequests_add(tsdn_t *tsdn, arena_stats_t *arena_stats,
szind_t szind, uint64_t nrequests) { szind_t szind, uint64_t nrequests) {
arena_stats_lock(tsdn, arena_stats); LOCKEDINT_MTX_LOCK(tsdn, arena_stats->mtx);
arena_stats_large_t *lstats = &arena_stats->lstats[szind - SC_NBINS]; arena_stats_large_t *lstats = &arena_stats->lstats[szind - SC_NBINS];
arena_stats_add_u64(tsdn, arena_stats, &lstats->nrequests, nrequests); locked_inc_u64(tsdn, LOCKEDINT_MTX(arena_stats->mtx),
arena_stats_add_u64(tsdn, arena_stats, &lstats->nflushes, 1); &lstats->nrequests, nrequests);
arena_stats_unlock(tsdn, arena_stats); locked_inc_u64(tsdn, LOCKEDINT_MTX(arena_stats->mtx),
&lstats->nflushes, 1);
LOCKEDINT_MTX_UNLOCK(tsdn, arena_stats->mtx);
} }
static inline void static inline void
arena_stats_mapped_add(tsdn_t *tsdn, arena_stats_t *arena_stats, size_t size) { arena_stats_mapped_add(tsdn_t *tsdn, arena_stats_t *arena_stats, size_t size) {
arena_stats_lock(tsdn, arena_stats); LOCKEDINT_MTX_LOCK(tsdn, arena_stats->mtx);
arena_stats_add_zu(tsdn, arena_stats, &arena_stats->mapped, size); locked_inc_zu(tsdn, LOCKEDINT_MTX(arena_stats->mtx),
arena_stats_unlock(tsdn, arena_stats); &arena_stats->mapped, size);
LOCKEDINT_MTX_UNLOCK(tsdn, arena_stats->mtx);
} }
#endif /* JEMALLOC_INTERNAL_ARENA_STATS_H */ #endif /* JEMALLOC_INTERNAL_ARENA_STATS_H */

26
include/jemalloc/internal/atomic.h
View File

@ -51,6 +51,20 @@
#define ATOMIC_ACQ_REL atomic_memory_order_acq_rel #define ATOMIC_ACQ_REL atomic_memory_order_acq_rel
#define ATOMIC_SEQ_CST atomic_memory_order_seq_cst #define ATOMIC_SEQ_CST atomic_memory_order_seq_cst
/*
* Another convenience -- simple atomic helper functions.
*/
#define JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(type, short_type, \
lg_size) \
JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, lg_size) \
ATOMIC_INLINE void \
atomic_load_add_store_##short_type(atomic_##short_type##_t *a, \
type inc) { \
type oldval = atomic_load_##short_type(a, ATOMIC_RELAXED); \
type newval = oldval + inc; \
atomic_store_##short_type(a, newval, ATOMIC_RELAXED); \
}
/* /*
* Not all platforms have 64-bit atomics. If we do, this #define exposes that * Not all platforms have 64-bit atomics. If we do, this #define exposes that
* fact. * fact.
@ -67,18 +81,18 @@ JEMALLOC_GENERATE_ATOMICS(void *, p, LG_SIZEOF_PTR)
*/ */
JEMALLOC_GENERATE_ATOMICS(bool, b, 0) JEMALLOC_GENERATE_ATOMICS(bool, b, 0)
JEMALLOC_GENERATE_INT_ATOMICS(unsigned, u, LG_SIZEOF_INT) JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(unsigned, u, LG_SIZEOF_INT)
JEMALLOC_GENERATE_INT_ATOMICS(size_t, zu, LG_SIZEOF_PTR) JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(size_t, zu, LG_SIZEOF_PTR)
JEMALLOC_GENERATE_INT_ATOMICS(ssize_t, zd, LG_SIZEOF_PTR) JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(ssize_t, zd, LG_SIZEOF_PTR)
JEMALLOC_GENERATE_INT_ATOMICS(uint8_t, u8, 0) JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(uint8_t, u8, 0)
JEMALLOC_GENERATE_INT_ATOMICS(uint32_t, u32, 2) JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(uint32_t, u32, 2)
#ifdef JEMALLOC_ATOMIC_U64 #ifdef JEMALLOC_ATOMIC_U64
JEMALLOC_GENERATE_INT_ATOMICS(uint64_t, u64, 3) JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(uint64_t, u64, 3)
#endif #endif
#undef ATOMIC_INLINE #undef ATOMIC_INLINE

151
include/jemalloc/internal/lockedint.h Normal file
View File

@ -0,0 +1,151 @@
#ifndef JEMALLOC_INTERNAL_LOCKEDINT_H
#define JEMALLOC_INTERNAL_LOCKEDINT_H
/*
* In those architectures that support 64-bit atomics, we use atomic updates for
* our 64-bit values. Otherwise, we use a plain uint64_t and synchronize
* externally.
*/
typedef struct locked_u64_s locked_u64_t;
#ifdef JEMALLOC_ATOMIC_U64
struct locked_u64_s {
atomic_u64_t val;
};
#else
/* Must hold the associated mutex. */
struct locked_u64_s {
uint64_t val;
};
#endif
typedef struct locked_zu_s locked_zu_t;
struct locked_zu_s {
atomic_zu_t val;
};
#ifndef JEMALLOC_ATOMIC_U64
# define LOCKEDINT_MTX_DECLARE(name) malloc_mutex_t name;
# define LOCKEDINT_MTX_INIT(ptr, name, rank, rank_mode) \
malloc_mutex_init(ptr, name, rank, rank_mode)
# define LOCKEDINT_MTX(mtx) (&(mtx))
# define LOCKEDINT_MTX_LOCK(tsdn, mu) malloc_mutex_lock(tsdn, &(mu))
# define LOCKEDINT_MTX_UNLOCK(tsdn, mu) malloc_mutex_unlock(tsdn, &(mu))
#else
# define LOCKEDINT_MTX_DECLARE(name)
# define LOCKEDINT_MTX(ptr) NULL
# define LOCKEDINT_MTX_INIT(ptr, name, rank, rank_mode) false
# define LOCKEDINT_MTX_LOCK(tsdn, mu) do {} while (0)
# define LOCKEDINT_MTX_UNLOCK(tsdn, mu) do {} while (0)
#endif
static inline uint64_t
locked_read_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p) {
#ifdef JEMALLOC_ATOMIC_U64
return atomic_load_u64(&p->val, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, mtx);
return p->val;
#endif
}
static inline void
locked_inc_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
uint64_t x) {
#ifdef JEMALLOC_ATOMIC_U64
atomic_fetch_add_u64(&p->val, x, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, mtx);
p->val += x;
#endif
}
static inline void
locked_dec_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
uint64_t x) {
#ifdef JEMALLOC_ATOMIC_U64
uint64_t r = atomic_fetch_sub_u64(&p->val, x, ATOMIC_RELAXED);
assert(r - x <= r);
#else
malloc_mutex_assert_owner(tsdn, mtx);
p->val -= x;
assert(p->val + x >= p->val);
#endif
}
/*
* Non-atomically sets *dst += src. *dst needs external synchronization.
* This lets us avoid the cost of a fetch_add when its unnecessary (note that
* the types here are atomic).
*/
static inline void
locked_inc_u64_unsynchronized(locked_u64_t *dst, uint64_t src) {
#ifdef JEMALLOC_ATOMIC_U64
uint64_t cur_dst = atomic_load_u64(&dst->val, ATOMIC_RELAXED);
atomic_store_u64(&dst->val, src + cur_dst, ATOMIC_RELAXED);
#else
dst->val += src;
#endif
}
static inline uint64_t
locked_read_u64_unsynchronized(locked_u64_t *p) {
#ifdef JEMALLOC_ATOMIC_U64
return atomic_load_u64(&p->val, ATOMIC_RELAXED);
#else
return p->val;
#endif
}
static inline size_t
locked_read_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p) {
#ifdef JEMALLOC_ATOMIC_U64
return atomic_load_zu(&p->val, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, mtx);
return atomic_load_zu(&p->val, ATOMIC_RELAXED);
#endif
}
static inline void
locked_inc_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p,
size_t x) {
#ifdef JEMALLOC_ATOMIC_U64
atomic_fetch_add_zu(&p->val, x, ATOMIC_RELAXED);
#else
malloc_mutex_assert_owner(tsdn, mtx);
size_t cur = atomic_load_zu(&p->val, ATOMIC_RELAXED);
atomic_store_zu(&p->val, cur + x, ATOMIC_RELAXED);
#endif
}
static inline void
locked_dec_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p,
size_t x) {
#ifdef JEMALLOC_ATOMIC_U64
size_t r = atomic_fetch_sub_zu(&p->val, x, ATOMIC_RELAXED);
assert(r - x <= r);
#else
malloc_mutex_assert_owner(tsdn, mtx);
size_t cur = atomic_load_zu(&p->val, ATOMIC_RELAXED);
atomic_store_zu(&p->val, cur - x, ATOMIC_RELAXED);
#endif
}
/* Like the _u64 variant, needs an externally synchronized *dst. */
static inline void
locked_inc_zu_unsynchronized(locked_zu_t *dst, size_t src) {
size_t cur_dst = atomic_load_zu(&dst->val, ATOMIC_RELAXED);
atomic_store_zu(&dst->val, src + cur_dst, ATOMIC_RELAXED);
}
/*
* Unlike the _u64 variant, this is safe to call unconditionally.
*/
static inline size_t
locked_read_atomic_zu(locked_zu_t *p) {
return atomic_load_zu(&p->val, ATOMIC_RELAXED);
}
#endif /* JEMALLOC_INTERNAL_LOCKEDINT_H */

132
src/arena.c
View File

@ -93,80 +93,89 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
base_stats_get(tsdn, arena->base, &base_allocated, &base_resident, base_stats_get(tsdn, arena->base, &base_allocated, &base_resident,
&base_mapped, &metadata_thp); &base_mapped, &metadata_thp);
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_stats_accum_zu(&astats->mapped, base_mapped locked_inc_zu_unsynchronized(&astats->mapped, base_mapped
+ arena_stats_read_zu(tsdn, &arena->stats, &arena->stats.mapped)); + locked_read_zu(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
arena_stats_accum_zu(&astats->retained, &arena->stats.mapped));
locked_inc_zu_unsynchronized(&astats->retained,
ecache_npages_get(&arena->pa_shard.ecache_retained) << LG_PAGE); ecache_npages_get(&arena->pa_shard.ecache_retained) << LG_PAGE);
atomic_store_zu(&astats->edata_avail, atomic_store_zu(&astats->edata_avail,
atomic_load_zu(&arena->pa_shard.edata_cache.count, ATOMIC_RELAXED), atomic_load_zu(&arena->pa_shard.edata_cache.count, ATOMIC_RELAXED),
ATOMIC_RELAXED); ATOMIC_RELAXED);
arena_stats_accum_u64(&astats->decay_dirty.npurge, locked_inc_u64_unsynchronized(&astats->decay_dirty.npurge,
arena_stats_read_u64(tsdn, &arena->stats, locked_read_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.decay_dirty.npurge)); &arena->stats.decay_dirty.npurge));
arena_stats_accum_u64(&astats->decay_dirty.nmadvise, locked_inc_u64_unsynchronized(&astats->decay_dirty.nmadvise,
arena_stats_read_u64(tsdn, &arena->stats, locked_read_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.decay_dirty.nmadvise)); &arena->stats.decay_dirty.nmadvise));
arena_stats_accum_u64(&astats->decay_dirty.purged, locked_inc_u64_unsynchronized(&astats->decay_dirty.purged,
arena_stats_read_u64(tsdn, &arena->stats, locked_read_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.decay_dirty.purged)); &arena->stats.decay_dirty.purged));
arena_stats_accum_u64(&astats->decay_muzzy.npurge, locked_inc_u64_unsynchronized(&astats->decay_muzzy.npurge,
arena_stats_read_u64(tsdn, &arena->stats, locked_read_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.decay_muzzy.npurge)); &arena->stats.decay_muzzy.npurge));
arena_stats_accum_u64(&astats->decay_muzzy.nmadvise, locked_inc_u64_unsynchronized(&astats->decay_muzzy.nmadvise,
arena_stats_read_u64(tsdn, &arena->stats, locked_read_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.decay_muzzy.nmadvise)); &arena->stats.decay_muzzy.nmadvise));
arena_stats_accum_u64(&astats->decay_muzzy.purged, locked_inc_u64_unsynchronized(&astats->decay_muzzy.purged,
arena_stats_read_u64(tsdn, &arena->stats, locked_read_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.decay_muzzy.purged)); &arena->stats.decay_muzzy.purged));
arena_stats_accum_zu(&astats->base, base_allocated); atomic_load_add_store_zu(&astats->base, base_allocated);
arena_stats_accum_zu(&astats->internal, arena_internal_get(arena)); atomic_load_add_store_zu(&astats->internal, arena_internal_get(arena));
arena_stats_accum_zu(&astats->metadata_thp, metadata_thp); atomic_load_add_store_zu(&astats->metadata_thp, metadata_thp);
arena_stats_accum_zu(&astats->resident, base_resident + atomic_load_add_store_zu(&astats->resident, base_resident +
(((atomic_load_zu(&arena->nactive, ATOMIC_RELAXED) + (((atomic_load_zu(&arena->nactive, ATOMIC_RELAXED) +
ecache_npages_get(&arena->pa_shard.ecache_dirty) + ecache_npages_get(&arena->pa_shard.ecache_dirty) +
ecache_npages_get(&arena->pa_shard.ecache_muzzy)) << LG_PAGE))); ecache_npages_get(&arena->pa_shard.ecache_muzzy)) << LG_PAGE)));
arena_stats_accum_zu(&astats->pa_shard_stats.abandoned_vm, atomic_load_add_store_zu(&astats->pa_shard_stats.abandoned_vm,
atomic_load_zu(&arena->stats.pa_shard_stats.abandoned_vm, atomic_load_zu(&arena->stats.pa_shard_stats.abandoned_vm,
ATOMIC_RELAXED)); ATOMIC_RELAXED));
for (szind_t i = 0; i < SC_NSIZES - SC_NBINS; i++) { for (szind_t i = 0; i < SC_NSIZES - SC_NBINS; i++) {
uint64_t nmalloc = arena_stats_read_u64(tsdn, &arena->stats, uint64_t nmalloc = locked_read_u64(tsdn,
LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.lstats[i].nmalloc); &arena->stats.lstats[i].nmalloc);
arena_stats_accum_u64(&lstats[i].nmalloc, nmalloc); locked_inc_u64_unsynchronized(&lstats[i].nmalloc, nmalloc);
arena_stats_accum_u64(&astats->nmalloc_large, nmalloc); locked_inc_u64_unsynchronized(&astats->nmalloc_large,
nmalloc);
uint64_t ndalloc = arena_stats_read_u64(tsdn, &arena->stats, uint64_t ndalloc = locked_read_u64(tsdn,
LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.lstats[i].ndalloc); &arena->stats.lstats[i].ndalloc);
arena_stats_accum_u64(&lstats[i].ndalloc, ndalloc); locked_inc_u64_unsynchronized(&lstats[i].ndalloc, ndalloc);
arena_stats_accum_u64(&astats->ndalloc_large, ndalloc); locked_inc_u64_unsynchronized(&astats->ndalloc_large,
ndalloc);
uint64_t nrequests = arena_stats_read_u64(tsdn, &arena->stats, uint64_t nrequests = locked_read_u64(tsdn,
LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.lstats[i].nrequests); &arena->stats.lstats[i].nrequests);
arena_stats_accum_u64(&lstats[i].nrequests, locked_inc_u64_unsynchronized(&lstats[i].nrequests,
nmalloc + nrequests); nmalloc + nrequests);
arena_stats_accum_u64(&astats->nrequests_large, locked_inc_u64_unsynchronized(&astats->nrequests_large,
nmalloc + nrequests); nmalloc + nrequests);
/* nfill == nmalloc for large currently. */ /* nfill == nmalloc for large currently. */
arena_stats_accum_u64(&lstats[i].nfills, nmalloc); locked_inc_u64_unsynchronized(&lstats[i].nfills, nmalloc);
arena_stats_accum_u64(&astats->nfills_large, nmalloc); locked_inc_u64_unsynchronized(&astats->nfills_large,
nmalloc);
uint64_t nflush = arena_stats_read_u64(tsdn, &arena->stats, uint64_t nflush = locked_read_u64(tsdn,
LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.lstats[i].nflushes); &arena->stats.lstats[i].nflushes);
arena_stats_accum_u64(&lstats[i].nflushes, nflush); locked_inc_u64_unsynchronized(&lstats[i].nflushes, nflush);
arena_stats_accum_u64(&astats->nflushes_large, nflush); locked_inc_u64_unsynchronized(&astats->nflushes_large,
nflush);
assert(nmalloc >= ndalloc); assert(nmalloc >= ndalloc);
assert(nmalloc - ndalloc <= SIZE_T_MAX); assert(nmalloc - ndalloc <= SIZE_T_MAX);
size_t curlextents = (size_t)(nmalloc - ndalloc); size_t curlextents = (size_t)(nmalloc - ndalloc);
lstats[i].curlextents += curlextents; lstats[i].curlextents += curlextents;
arena_stats_accum_zu(&astats->allocated_large, atomic_load_add_store_zu(&astats->allocated_large,
curlextents * sz_index2size(SC_NBINS + i)); curlextents * sz_index2size(SC_NBINS + i));
} }
@ -195,7 +204,7 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
ATOMIC_RELAXED); ATOMIC_RELAXED);
} }
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
/* tcache_bytes counts currently cached bytes. */ /* tcache_bytes counts currently cached bytes. */
atomic_store_zu(&astats->tcache_bytes, 0, ATOMIC_RELAXED); atomic_store_zu(&astats->tcache_bytes, 0, ATOMIC_RELAXED);
@ -204,13 +213,13 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
ql_foreach(descriptor, &arena->cache_bin_array_descriptor_ql, link) { ql_foreach(descriptor, &arena->cache_bin_array_descriptor_ql, link) {
for (szind_t i = 0; i < SC_NBINS; i++) { for (szind_t i = 0; i < SC_NBINS; i++) {
cache_bin_t *tbin = &descriptor->bins_small[i]; cache_bin_t *tbin = &descriptor->bins_small[i];
arena_stats_accum_zu(&astats->tcache_bytes, atomic_load_add_store_zu(&astats->tcache_bytes,
cache_bin_ncached_get(tbin, &tcache_bin_info[i]) cache_bin_ncached_get(tbin,
* sz_index2size(i)); &tcache_bin_info[i]) * sz_index2size(i));
} }
for (szind_t i = 0; i < nhbins - SC_NBINS; i++) { for (szind_t i = 0; i < nhbins - SC_NBINS; i++) {
cache_bin_t *tbin = &descriptor->bins_large[i]; cache_bin_t *tbin = &descriptor->bins_large[i];
arena_stats_accum_zu(&astats->tcache_bytes, atomic_load_add_store_zu(&astats->tcache_bytes,
cache_bin_ncached_get(tbin, cache_bin_ncached_get(tbin,
&tcache_bin_info[i + SC_NBINS]) &tcache_bin_info[i + SC_NBINS])
* sz_index2size(i + SC_NBINS)); * sz_index2size(i + SC_NBINS));
@ -397,7 +406,7 @@ arena_large_malloc_stats_update(tsdn_t *tsdn, arena_t *arena, size_t usize) {
index = sz_size2index(usize); index = sz_size2index(usize);
hindex = (index >= SC_NBINS) ? index - SC_NBINS : 0; hindex = (index >= SC_NBINS) ? index - SC_NBINS : 0;
arena_stats_add_u64(tsdn, &arena->stats, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.lstats[hindex].nmalloc, 1); &arena->stats.lstats[hindex].nmalloc, 1);
} }
@ -413,7 +422,7 @@ arena_large_dalloc_stats_update(tsdn_t *tsdn, arena_t *arena, size_t usize) {
index = sz_size2index(usize); index = sz_size2index(usize);
hindex = (index >= SC_NBINS) ? index - SC_NBINS : 0; hindex = (index >= SC_NBINS) ? index - SC_NBINS : 0;
arena_stats_add_u64(tsdn, &arena->stats, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.lstats[hindex].ndalloc, 1); &arena->stats.lstats[hindex].ndalloc, 1);
} }
@ -466,13 +475,14 @@ arena_extent_alloc_large(tsdn_t *tsdn, arena_t *arena, size_t usize,
if (edata != NULL) { if (edata != NULL) {
if (config_stats) { if (config_stats) {
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_large_malloc_stats_update(tsdn, arena, usize); arena_large_malloc_stats_update(tsdn, arena, usize);
if (mapped_add != 0) { if (mapped_add != 0) {
arena_stats_add_zu(tsdn, &arena->stats, locked_inc_zu(tsdn,
LOCKEDINT_MTX(arena->stats.mtx),
&arena->stats.mapped, mapped_add); &arena->stats.mapped, mapped_add);
} }
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
} }
arena_nactive_add(arena, esize >> LG_PAGE); arena_nactive_add(arena, esize >> LG_PAGE);
} }
@ -487,10 +497,10 @@ arena_extent_alloc_large(tsdn_t *tsdn, arena_t *arena, size_t usize,
void void
arena_extent_dalloc_large_prep(tsdn_t *tsdn, arena_t *arena, edata_t *edata) { arena_extent_dalloc_large_prep(tsdn_t *tsdn, arena_t *arena, edata_t *edata) {
if (config_stats) { if (config_stats) {
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_large_dalloc_stats_update(tsdn, arena, arena_large_dalloc_stats_update(tsdn, arena,
edata_usize_get(edata)); edata_usize_get(edata));
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
} }
arena_nactive_sub(arena, edata_size_get(edata) >> LG_PAGE); arena_nactive_sub(arena, edata_size_get(edata) >> LG_PAGE);
} }
@ -502,9 +512,9 @@ arena_extent_ralloc_large_shrink(tsdn_t *tsdn, arena_t *arena, edata_t *edata,
size_t udiff = oldusize - usize; size_t udiff = oldusize - usize;
if (config_stats) { if (config_stats) {
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_large_ralloc_stats_update(tsdn, arena, oldusize, usize); arena_large_ralloc_stats_update(tsdn, arena, oldusize, usize);
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
} }
arena_nactive_sub(arena, udiff >> LG_PAGE); arena_nactive_sub(arena, udiff >> LG_PAGE);
} }
@ -516,9 +526,9 @@ arena_extent_ralloc_large_expand(tsdn_t *tsdn, arena_t *arena, edata_t *edata,
size_t udiff = usize - oldusize; size_t udiff = usize - oldusize;
if (config_stats) { if (config_stats) {
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_large_ralloc_stats_update(tsdn, arena, oldusize, usize); arena_large_ralloc_stats_update(tsdn, arena, oldusize, usize);
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
} }
arena_nactive_add(arena, udiff >> LG_PAGE); arena_nactive_add(arena, udiff >> LG_PAGE);
} }
@ -894,16 +904,16 @@ arena_decay_stashed(tsdn_t *tsdn, arena_t *arena, ehooks_t *ehooks,
} }
if (config_stats) { if (config_stats) {
arena_stats_lock(tsdn, &arena->stats); LOCKEDINT_MTX_LOCK(tsdn, arena->stats.mtx);
arena_stats_add_u64(tsdn, &arena->stats, &decay->stats->npurge, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
1); &decay->stats->npurge, 1);
arena_stats_add_u64(tsdn, &arena->stats, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
&decay->stats->nmadvise, nmadvise); &decay->stats->nmadvise, nmadvise);
arena_stats_add_u64(tsdn, &arena->stats, &decay->stats->purged, locked_inc_u64(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
npurged); &decay->stats->purged, npurged);
arena_stats_sub_zu(tsdn, &arena->stats, &arena->stats.mapped, locked_dec_zu(tsdn, LOCKEDINT_MTX(arena->stats.mtx),
nunmapped << LG_PAGE); &arena->stats.mapped, nunmapped << LG_PAGE);
arena_stats_unlock(tsdn, &arena->stats); LOCKEDINT_MTX_UNLOCK(tsdn, arena->stats.mtx);
} }
return npurged; return npurged;

128
src/ctl.c
View File

@ -676,28 +676,19 @@ static const ctl_named_node_t super_root_node[] = {
* synchronized by the ctl mutex. * synchronized by the ctl mutex.
*/ */
static void static void
ctl_accum_arena_stats_u64(arena_stats_u64_t *dst, arena_stats_u64_t *src) { ctl_accum_locked_u64(locked_u64_t *dst, locked_u64_t *src) {
#ifdef JEMALLOC_ATOMIC_U64 locked_inc_u64_unsynchronized(dst,
uint64_t cur_dst = atomic_load_u64(dst, ATOMIC_RELAXED); locked_read_u64_unsynchronized(src));
uint64_t cur_src = atomic_load_u64(src, ATOMIC_RELAXED);
atomic_store_u64(dst, cur_dst + cur_src, ATOMIC_RELAXED);
#else
*dst += *src;
#endif
}
/* Likewise: with ctl mutex synchronization, reading is simple. */
static uint64_t
ctl_arena_stats_read_u64(arena_stats_u64_t *p) {
#ifdef JEMALLOC_ATOMIC_U64
return atomic_load_u64(p, ATOMIC_RELAXED);
#else
return *p;
#endif
} }
static void static void
accum_atomic_zu(atomic_zu_t *dst, atomic_zu_t *src) { ctl_accum_locked_zu(locked_zu_t *dst, locked_zu_t *src) {
locked_inc_zu_unsynchronized(dst,
locked_read_atomic_zu(src));
}
static void
ctl_accum_atomic_zu(atomic_zu_t *dst, atomic_zu_t *src) {
size_t cur_dst = atomic_load_zu(dst, ATOMIC_RELAXED); size_t cur_dst = atomic_load_zu(dst, ATOMIC_RELAXED);
size_t cur_src = atomic_load_zu(src, ATOMIC_RELAXED); size_t cur_src = atomic_load_zu(src, ATOMIC_RELAXED);
atomic_store_zu(dst, cur_dst + cur_src, ATOMIC_RELAXED); atomic_store_zu(dst, cur_dst + cur_src, ATOMIC_RELAXED);
@ -870,26 +861,26 @@ ctl_arena_stats_sdmerge(ctl_arena_t *ctl_sdarena, ctl_arena_t *ctl_arena,
ctl_arena_stats_t *astats = ctl_arena->astats; ctl_arena_stats_t *astats = ctl_arena->astats;
if (!destroyed) { if (!destroyed) {
accum_atomic_zu(&sdstats->astats.mapped, ctl_accum_locked_zu(&sdstats->astats.mapped,
&astats->astats.mapped); &astats->astats.mapped);
accum_atomic_zu(&sdstats->astats.retained, ctl_accum_locked_zu(&sdstats->astats.retained,
&astats->astats.retained); &astats->astats.retained);
accum_atomic_zu(&sdstats->astats.edata_avail, ctl_accum_atomic_zu(&sdstats->astats.edata_avail,
&astats->astats.edata_avail); &astats->astats.edata_avail);
} }
ctl_accum_arena_stats_u64(&sdstats->astats.decay_dirty.npurge, ctl_accum_locked_u64(&sdstats->astats.decay_dirty.npurge,
&astats->astats.decay_dirty.npurge); &astats->astats.decay_dirty.npurge);
ctl_accum_arena_stats_u64(&sdstats->astats.decay_dirty.nmadvise, ctl_accum_locked_u64(&sdstats->astats.decay_dirty.nmadvise,
&astats->astats.decay_dirty.nmadvise); &astats->astats.decay_dirty.nmadvise);
ctl_accum_arena_stats_u64(&sdstats->astats.decay_dirty.purged, ctl_accum_locked_u64(&sdstats->astats.decay_dirty.purged,
&astats->astats.decay_dirty.purged); &astats->astats.decay_dirty.purged);
ctl_accum_arena_stats_u64(&sdstats->astats.decay_muzzy.npurge, ctl_accum_locked_u64(&sdstats->astats.decay_muzzy.npurge,
&astats->astats.decay_muzzy.npurge); &astats->astats.decay_muzzy.npurge);
ctl_accum_arena_stats_u64(&sdstats->astats.decay_muzzy.nmadvise, ctl_accum_locked_u64(&sdstats->astats.decay_muzzy.nmadvise,
&astats->astats.decay_muzzy.nmadvise); &astats->astats.decay_muzzy.nmadvise);
ctl_accum_arena_stats_u64(&sdstats->astats.decay_muzzy.purged, ctl_accum_locked_u64(&sdstats->astats.decay_muzzy.purged,
&astats->astats.decay_muzzy.purged); &astats->astats.decay_muzzy.purged);
#define OP(mtx) malloc_mutex_prof_merge( \ #define OP(mtx) malloc_mutex_prof_merge( \
@ -900,13 +891,13 @@ ctl_arena_stats_sdmerge(ctl_arena_t *ctl_sdarena, ctl_arena_t *ctl_arena,
MUTEX_PROF_ARENA_MUTEXES MUTEX_PROF_ARENA_MUTEXES
#undef OP #undef OP
if (!destroyed) { if (!destroyed) {
accum_atomic_zu(&sdstats->astats.base, ctl_accum_atomic_zu(&sdstats->astats.base,
&astats->astats.base); &astats->astats.base);
accum_atomic_zu(&sdstats->astats.internal, ctl_accum_atomic_zu(&sdstats->astats.internal,
&astats->astats.internal); &astats->astats.internal);
accum_atomic_zu(&sdstats->astats.resident, ctl_accum_atomic_zu(&sdstats->astats.resident,
&astats->astats.resident); &astats->astats.resident);
accum_atomic_zu(&sdstats->astats.metadata_thp, ctl_accum_atomic_zu(&sdstats->astats.metadata_thp,
&astats->astats.metadata_thp); &astats->astats.metadata_thp);
} else { } else {
assert(atomic_load_zu( assert(atomic_load_zu(
@ -925,24 +916,25 @@ MUTEX_PROF_ARENA_MUTEXES
sdstats->nflushes_small += astats->nflushes_small; sdstats->nflushes_small += astats->nflushes_small;
if (!destroyed) { if (!destroyed) {
accum_atomic_zu(&sdstats->astats.allocated_large, ctl_accum_atomic_zu(&sdstats->astats.allocated_large,
&astats->astats.allocated_large); &astats->astats.allocated_large);
} else { } else {
assert(atomic_load_zu(&astats->astats.allocated_large, assert(atomic_load_zu(&astats->astats.allocated_large,
ATOMIC_RELAXED) == 0); ATOMIC_RELAXED) == 0);
} }
ctl_accum_arena_stats_u64(&sdstats->astats.nmalloc_large, ctl_accum_locked_u64(&sdstats->astats.nmalloc_large,
&astats->astats.nmalloc_large); &astats->astats.nmalloc_large);
ctl_accum_arena_stats_u64(&sdstats->astats.ndalloc_large, ctl_accum_locked_u64(&sdstats->astats.ndalloc_large,
&astats->astats.ndalloc_large); &astats->astats.ndalloc_large);
ctl_accum_arena_stats_u64(&sdstats->astats.nrequests_large, ctl_accum_locked_u64(&sdstats->astats.nrequests_large,
&astats->astats.nrequests_large); &astats->astats.nrequests_large);
ctl_accum_arena_stats_u64(&sdstats->astats.nflushes_large, ctl_accum_locked_u64(&sdstats->astats.nflushes_large,
&astats->astats.nflushes_large); &astats->astats.nflushes_large);
accum_atomic_zu(&sdstats->astats.pa_shard_stats.abandoned_vm, ctl_accum_atomic_zu(
&sdstats->astats.pa_shard_stats.abandoned_vm,
&astats->astats.pa_shard_stats.abandoned_vm); &astats->astats.pa_shard_stats.abandoned_vm);
accum_atomic_zu(&sdstats->astats.tcache_bytes, ctl_accum_atomic_zu(&sdstats->astats.tcache_bytes,
&astats->astats.tcache_bytes); &astats->astats.tcache_bytes);
if (ctl_arena->arena_ind == 0) { if (ctl_arena->arena_ind == 0) {
@ -978,11 +970,11 @@ MUTEX_PROF_ARENA_MUTEXES
/* Merge stats for large allocations. */ /* Merge stats for large allocations. */
for (i = 0; i < SC_NSIZES - SC_NBINS; i++) { for (i = 0; i < SC_NSIZES - SC_NBINS; i++) {
ctl_accum_arena_stats_u64(&sdstats->lstats[i].nmalloc, ctl_accum_locked_u64(&sdstats->lstats[i].nmalloc,
&astats->lstats[i].nmalloc); &astats->lstats[i].nmalloc);
ctl_accum_arena_stats_u64(&sdstats->lstats[i].ndalloc, ctl_accum_locked_u64(&sdstats->lstats[i].ndalloc,
&astats->lstats[i].ndalloc); &astats->lstats[i].ndalloc);
ctl_accum_arena_stats_u64(&sdstats->lstats[i].nrequests, ctl_accum_locked_u64(&sdstats->lstats[i].nrequests,
&astats->lstats[i].nrequests); &astats->lstats[i].nrequests);
if (!destroyed) { if (!destroyed) {
sdstats->lstats[i].curlextents += sdstats->lstats[i].curlextents +=
@ -994,17 +986,17 @@ MUTEX_PROF_ARENA_MUTEXES
/* Merge extents stats. */ /* Merge extents stats. */
for (i = 0; i < SC_NPSIZES; i++) { for (i = 0; i < SC_NPSIZES; i++) {
accum_atomic_zu(&sdstats->estats[i].ndirty, ctl_accum_atomic_zu(&sdstats->estats[i].ndirty,
&astats->estats[i].ndirty); &astats->estats[i].ndirty);
accum_atomic_zu(&sdstats->estats[i].nmuzzy, ctl_accum_atomic_zu(&sdstats->estats[i].nmuzzy,
&astats->estats[i].nmuzzy); &astats->estats[i].nmuzzy);
accum_atomic_zu(&sdstats->estats[i].nretained, ctl_accum_atomic_zu(&sdstats->estats[i].nretained,
&astats->estats[i].nretained); &astats->estats[i].nretained);
accum_atomic_zu(&sdstats->estats[i].dirty_bytes, ctl_accum_atomic_zu(&sdstats->estats[i].dirty_bytes,
&astats->estats[i].dirty_bytes); &astats->estats[i].dirty_bytes);
accum_atomic_zu(&sdstats->estats[i].muzzy_bytes, ctl_accum_atomic_zu(&sdstats->estats[i].muzzy_bytes,
&astats->estats[i].muzzy_bytes); &astats->estats[i].muzzy_bytes);
accum_atomic_zu(&sdstats->estats[i].retained_bytes, ctl_accum_atomic_zu(&sdstats->estats[i].retained_bytes,
&astats->estats[i].retained_bytes); &astats->estats[i].retained_bytes);
} }
} }
@ -1104,10 +1096,10 @@ ctl_refresh(tsdn_t *tsdn) {
&ctl_sarena->astats->astats.metadata_thp, ATOMIC_RELAXED); &ctl_sarena->astats->astats.metadata_thp, ATOMIC_RELAXED);
ctl_stats->resident = atomic_load_zu( ctl_stats->resident = atomic_load_zu(
&ctl_sarena->astats->astats.resident, ATOMIC_RELAXED); &ctl_sarena->astats->astats.resident, ATOMIC_RELAXED);
ctl_stats->mapped = atomic_load_zu( ctl_stats->mapped = locked_read_atomic_zu(
&ctl_sarena->astats->astats.mapped, ATOMIC_RELAXED); &ctl_sarena->astats->astats.mapped);
ctl_stats->retained = atomic_load_zu( ctl_stats->retained = locked_read_atomic_zu(
&ctl_sarena->astats->astats.retained, ATOMIC_RELAXED); &ctl_sarena->astats->astats.retained);
ctl_background_thread_stats_read(tsdn); ctl_background_thread_stats_read(tsdn);
@ -2916,10 +2908,10 @@ CTL_RO_GEN(stats_arenas_i_pactive, arenas_i(mib[2])->pactive, size_t)
CTL_RO_GEN(stats_arenas_i_pdirty, arenas_i(mib[2])->pdirty, size_t) CTL_RO_GEN(stats_arenas_i_pdirty, arenas_i(mib[2])->pdirty, size_t)
CTL_RO_GEN(stats_arenas_i_pmuzzy, arenas_i(mib[2])->pmuzzy, size_t) CTL_RO_GEN(stats_arenas_i_pmuzzy, arenas_i(mib[2])->pmuzzy, size_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_mapped, CTL_RO_CGEN(config_stats, stats_arenas_i_mapped,
atomic_load_zu(&arenas_i(mib[2])->astats->astats.mapped, ATOMIC_RELAXED), locked_read_atomic_zu(&arenas_i(mib[2])->astats->astats.mapped),
size_t) size_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_retained, CTL_RO_CGEN(config_stats, stats_arenas_i_retained,
atomic_load_zu(&arenas_i(mib[2])->astats->astats.retained, ATOMIC_RELAXED), locked_read_atomic_zu(&arenas_i(mib[2])->astats->astats.retained),
size_t) size_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_extent_avail, CTL_RO_CGEN(config_stats, stats_arenas_i_extent_avail,
atomic_load_zu(&arenas_i(mib[2])->astats->astats.edata_avail, atomic_load_zu(&arenas_i(mib[2])->astats->astats.edata_avail,
@ -2927,23 +2919,23 @@ CTL_RO_CGEN(config_stats, stats_arenas_i_extent_avail,
size_t) size_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_npurge, CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_npurge,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.decay_dirty.npurge), uint64_t) &arenas_i(mib[2])->astats->astats.decay_dirty.npurge), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_nmadvise, CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_nmadvise,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.decay_dirty.nmadvise), uint64_t) &arenas_i(mib[2])->astats->astats.decay_dirty.nmadvise), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_purged, CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_purged,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.decay_dirty.purged), uint64_t) &arenas_i(mib[2])->astats->astats.decay_dirty.purged), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_muzzy_npurge, CTL_RO_CGEN(config_stats, stats_arenas_i_muzzy_npurge,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.decay_muzzy.npurge), uint64_t) &arenas_i(mib[2])->astats->astats.decay_muzzy.npurge), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_muzzy_nmadvise, CTL_RO_CGEN(config_stats, stats_arenas_i_muzzy_nmadvise,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.decay_muzzy.nmadvise), uint64_t) &arenas_i(mib[2])->astats->astats.decay_muzzy.nmadvise), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_muzzy_purged, CTL_RO_CGEN(config_stats, stats_arenas_i_muzzy_purged,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.decay_muzzy.purged), uint64_t) &arenas_i(mib[2])->astats->astats.decay_muzzy.purged), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_base, CTL_RO_CGEN(config_stats, stats_arenas_i_base,
@ -2982,23 +2974,23 @@ CTL_RO_CGEN(config_stats, stats_arenas_i_large_allocated,
atomic_load_zu(&arenas_i(mib[2])->astats->astats.allocated_large, atomic_load_zu(&arenas_i(mib[2])->astats->astats.allocated_large,
ATOMIC_RELAXED), size_t) ATOMIC_RELAXED), size_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_large_nmalloc, CTL_RO_CGEN(config_stats, stats_arenas_i_large_nmalloc,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.nmalloc_large), uint64_t) &arenas_i(mib[2])->astats->astats.nmalloc_large), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_large_ndalloc, CTL_RO_CGEN(config_stats, stats_arenas_i_large_ndalloc,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.ndalloc_large), uint64_t) &arenas_i(mib[2])->astats->astats.ndalloc_large), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_large_nrequests, CTL_RO_CGEN(config_stats, stats_arenas_i_large_nrequests,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.nrequests_large), uint64_t) &arenas_i(mib[2])->astats->astats.nrequests_large), uint64_t)
/* /*
* Note: "nmalloc_large" here instead of "nfills" in the read. This is * Note: "nmalloc_large" here instead of "nfills" in the read. This is
* intentional (large has no batch fill). * intentional (large has no batch fill).
*/ */
CTL_RO_CGEN(config_stats, stats_arenas_i_large_nfills, CTL_RO_CGEN(config_stats, stats_arenas_i_large_nfills,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.nmalloc_large), uint64_t) &arenas_i(mib[2])->astats->astats.nmalloc_large), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_large_nflushes, CTL_RO_CGEN(config_stats, stats_arenas_i_large_nflushes,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->astats.nflushes_large), uint64_t) &arenas_i(mib[2])->astats->astats.nflushes_large), uint64_t)
/* Lock profiling related APIs below. */ /* Lock profiling related APIs below. */
@ -3124,13 +3116,13 @@ stats_arenas_i_bins_j_index(tsdn_t *tsdn, const size_t *mib,
} }
CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_nmalloc, CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_nmalloc,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->lstats[mib[4]].nmalloc), uint64_t) &arenas_i(mib[2])->astats->lstats[mib[4]].nmalloc), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_ndalloc, CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_ndalloc,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->lstats[mib[4]].ndalloc), uint64_t) &arenas_i(mib[2])->astats->lstats[mib[4]].ndalloc), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_nrequests, CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_nrequests,
ctl_arena_stats_read_u64( locked_read_u64_unsynchronized(
&arenas_i(mib[2])->astats->lstats[mib[4]].nrequests), uint64_t) &arenas_i(mib[2])->astats->lstats[mib[4]].nrequests), uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_curlextents, CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_curlextents,
arenas_i(mib[2])->astats->lstats[mib[4]].curlextents, size_t) arenas_i(mib[2])->astats->lstats[mib[4]].curlextents, size_t)

3
src/extent.c
View File

@ -199,7 +199,8 @@ extents_abandon_vm(tsdn_t *tsdn, arena_t *arena, ehooks_t *ehooks,
ecache_t *ecache, edata_t *edata, bool growing_retained) { ecache_t *ecache, edata_t *edata, bool growing_retained) {
size_t sz = edata_size_get(edata); size_t sz = edata_size_get(edata);
if (config_stats) { if (config_stats) {
arena_stats_accum_zu(&arena->pa_shard.stats->abandoned_vm, sz); atomic_fetch_add_zu(&arena->pa_shard.stats->abandoned_vm, sz,
ATOMIC_RELAXED);
} }
/* /*
* Leak extent after making sure its pages have already been purged, so * Leak extent after making sure its pages have already been purged, so