diff options
Diffstat (limited to 'dep/jemalloc/src/arena.c')
| -rw-r--r-- | dep/jemalloc/src/arena.c | 2928 |
1 files changed, 1359 insertions, 1569 deletions
diff --git a/dep/jemalloc/src/arena.c b/dep/jemalloc/src/arena.c index 7f939b3cd77..05a787f89d9 100644 --- a/dep/jemalloc/src/arena.c +++ b/dep/jemalloc/src/arena.c @@ -4,176 +4,71 @@ /******************************************************************************/ /* Data. */ -size_t opt_lg_qspace_max = LG_QSPACE_MAX_DEFAULT; -size_t opt_lg_cspace_max = LG_CSPACE_MAX_DEFAULT; ssize_t opt_lg_dirty_mult = LG_DIRTY_MULT_DEFAULT; -uint8_t const *small_size2bin; - -/* Various bin-related settings. */ -unsigned nqbins; -unsigned ncbins; -unsigned nsbins; -unsigned nbins; -size_t qspace_max; -size_t cspace_min; -size_t cspace_max; -size_t sspace_min; -size_t sspace_max; - -size_t lg_mspace; -size_t mspace_mask; +arena_bin_info_t arena_bin_info[NBINS]; -/* - * const_small_size2bin is a static constant lookup table that in the common - * case can be used as-is for small_size2bin. For dynamically linked programs, - * this avoids a page of memory overhead per process. - */ -#define S2B_1(i) i, -#define S2B_2(i) S2B_1(i) S2B_1(i) -#define S2B_4(i) S2B_2(i) S2B_2(i) -#define S2B_8(i) S2B_4(i) S2B_4(i) +JEMALLOC_ALIGNED(CACHELINE) +const uint8_t small_size2bin[] = { +#define S2B_8(i) i, #define S2B_16(i) S2B_8(i) S2B_8(i) #define S2B_32(i) S2B_16(i) S2B_16(i) #define S2B_64(i) S2B_32(i) S2B_32(i) #define S2B_128(i) S2B_64(i) S2B_64(i) #define S2B_256(i) S2B_128(i) S2B_128(i) -/* - * The number of elements in const_small_size2bin is dependent on page size - * and on the definition for SUBPAGE. If SUBPAGE changes, the '- 255' must also - * change, along with the addition/removal of static lookup table element - * definitions. - */ -static const uint8_t const_small_size2bin[STATIC_PAGE_SIZE - 255] = { - S2B_1(0xffU) /* 0 */ -#if (LG_QUANTUM == 4) -/* 16-byte quantum **********************/ -# ifdef JEMALLOC_TINY -# if (LG_TINY_MIN == 2) - S2B_4(0) /* 4 */ - S2B_4(1) /* 8 */ - S2B_8(2) /* 16 */ -# define S2B_QMIN 2 -# elif (LG_TINY_MIN == 3) - S2B_8(0) /* 8 */ - S2B_8(1) /* 16 */ -# define S2B_QMIN 1 -# else -# error "Unsupported LG_TINY_MIN" -# endif -# else - S2B_16(0) /* 16 */ -# define S2B_QMIN 0 -# endif - S2B_16(S2B_QMIN + 1) /* 32 */ - S2B_16(S2B_QMIN + 2) /* 48 */ - S2B_16(S2B_QMIN + 3) /* 64 */ - S2B_16(S2B_QMIN + 4) /* 80 */ - S2B_16(S2B_QMIN + 5) /* 96 */ - S2B_16(S2B_QMIN + 6) /* 112 */ - S2B_16(S2B_QMIN + 7) /* 128 */ -# define S2B_CMIN (S2B_QMIN + 8) -#else -/* 8-byte quantum ***********************/ -# ifdef JEMALLOC_TINY -# if (LG_TINY_MIN == 2) - S2B_4(0) /* 4 */ - S2B_4(1) /* 8 */ -# define S2B_QMIN 1 -# else -# error "Unsupported LG_TINY_MIN" -# endif -# else - S2B_8(0) /* 8 */ -# define S2B_QMIN 0 -# endif - S2B_8(S2B_QMIN + 1) /* 16 */ - S2B_8(S2B_QMIN + 2) /* 24 */ - S2B_8(S2B_QMIN + 3) /* 32 */ - S2B_8(S2B_QMIN + 4) /* 40 */ - S2B_8(S2B_QMIN + 5) /* 48 */ - S2B_8(S2B_QMIN + 6) /* 56 */ - S2B_8(S2B_QMIN + 7) /* 64 */ - S2B_8(S2B_QMIN + 8) /* 72 */ - S2B_8(S2B_QMIN + 9) /* 80 */ - S2B_8(S2B_QMIN + 10) /* 88 */ - S2B_8(S2B_QMIN + 11) /* 96 */ - S2B_8(S2B_QMIN + 12) /* 104 */ - S2B_8(S2B_QMIN + 13) /* 112 */ - S2B_8(S2B_QMIN + 14) /* 120 */ - S2B_8(S2B_QMIN + 15) /* 128 */ -# define S2B_CMIN (S2B_QMIN + 16) -#endif -/****************************************/ - S2B_64(S2B_CMIN + 0) /* 192 */ - S2B_64(S2B_CMIN + 1) /* 256 */ - S2B_64(S2B_CMIN + 2) /* 320 */ - S2B_64(S2B_CMIN + 3) /* 384 */ - S2B_64(S2B_CMIN + 4) /* 448 */ - S2B_64(S2B_CMIN + 5) /* 512 */ -# define S2B_SMIN (S2B_CMIN + 6) - S2B_256(S2B_SMIN + 0) /* 768 */ - S2B_256(S2B_SMIN + 1) /* 1024 */ - S2B_256(S2B_SMIN + 2) /* 1280 */ - S2B_256(S2B_SMIN + 3) /* 1536 */ - S2B_256(S2B_SMIN + 4) /* 1792 */ - S2B_256(S2B_SMIN + 5) /* 2048 */ - S2B_256(S2B_SMIN + 6) /* 2304 */ - S2B_256(S2B_SMIN + 7) /* 2560 */ - S2B_256(S2B_SMIN + 8) /* 2816 */ - S2B_256(S2B_SMIN + 9) /* 3072 */ - S2B_256(S2B_SMIN + 10) /* 3328 */ - S2B_256(S2B_SMIN + 11) /* 3584 */ - S2B_256(S2B_SMIN + 12) /* 3840 */ -#if (STATIC_PAGE_SHIFT == 13) - S2B_256(S2B_SMIN + 13) /* 4096 */ - S2B_256(S2B_SMIN + 14) /* 4352 */ - S2B_256(S2B_SMIN + 15) /* 4608 */ - S2B_256(S2B_SMIN + 16) /* 4864 */ - S2B_256(S2B_SMIN + 17) /* 5120 */ - S2B_256(S2B_SMIN + 18) /* 5376 */ - S2B_256(S2B_SMIN + 19) /* 5632 */ - S2B_256(S2B_SMIN + 20) /* 5888 */ - S2B_256(S2B_SMIN + 21) /* 6144 */ - S2B_256(S2B_SMIN + 22) /* 6400 */ - S2B_256(S2B_SMIN + 23) /* 6656 */ - S2B_256(S2B_SMIN + 24) /* 6912 */ - S2B_256(S2B_SMIN + 25) /* 7168 */ - S2B_256(S2B_SMIN + 26) /* 7424 */ - S2B_256(S2B_SMIN + 27) /* 7680 */ - S2B_256(S2B_SMIN + 28) /* 7936 */ -#endif -}; -#undef S2B_1 -#undef S2B_2 -#undef S2B_4 +#define S2B_512(i) S2B_256(i) S2B_256(i) +#define S2B_1024(i) S2B_512(i) S2B_512(i) +#define S2B_2048(i) S2B_1024(i) S2B_1024(i) +#define S2B_4096(i) S2B_2048(i) S2B_2048(i) +#define S2B_8192(i) S2B_4096(i) S2B_4096(i) +#define SIZE_CLASS(bin, delta, size) \ + S2B_##delta(bin) + SIZE_CLASSES #undef S2B_8 #undef S2B_16 #undef S2B_32 #undef S2B_64 #undef S2B_128 #undef S2B_256 -#undef S2B_QMIN -#undef S2B_CMIN -#undef S2B_SMIN +#undef S2B_512 +#undef S2B_1024 +#undef S2B_2048 +#undef S2B_4096 +#undef S2B_8192 +#undef SIZE_CLASS +}; /******************************************************************************/ /* Function prototypes for non-inline static functions. */ +static void arena_avail_insert(arena_t *arena, arena_chunk_t *chunk, + size_t pageind, size_t npages, bool maybe_adjac_pred, + bool maybe_adjac_succ); +static void arena_avail_remove(arena_t *arena, arena_chunk_t *chunk, + size_t pageind, size_t npages, bool maybe_adjac_pred, + bool maybe_adjac_succ); static void arena_run_split(arena_t *arena, arena_run_t *run, size_t size, - bool large, bool zero); + bool large, size_t binind, bool zero); static arena_chunk_t *arena_chunk_alloc(arena_t *arena); static void arena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk); +static arena_run_t *arena_run_alloc_helper(arena_t *arena, size_t size, + bool large, size_t binind, bool zero); static arena_run_t *arena_run_alloc(arena_t *arena, size_t size, bool large, - bool zero); + size_t binind, bool zero); +static arena_chunk_t *chunks_dirty_iter_cb(arena_chunk_tree_t *tree, + arena_chunk_t *chunk, void *arg); static void arena_purge(arena_t *arena, bool all); -static void arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty); +static void arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty, + bool cleaned); static void arena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize); static void arena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize, bool dirty); +static arena_run_t *arena_bin_runs_first(arena_bin_t *bin); +static void arena_bin_runs_insert(arena_bin_t *bin, arena_run_t *run); +static void arena_bin_runs_remove(arena_bin_t *bin, arena_run_t *run); +static arena_run_t *arena_bin_nonfull_run_tryget(arena_bin_t *bin); static arena_run_t *arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin); static void *arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin); -static size_t arena_bin_run_size_calc(arena_bin_t *bin, size_t min_run_size); static void arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run, arena_bin_t *bin); static void arena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk, @@ -186,11 +81,9 @@ static bool arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, size_t oldsize, size_t size, size_t extra, bool zero); static bool arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, bool zero); -static bool small_size2bin_init(void); -#ifdef JEMALLOC_DEBUG -static void small_size2bin_validate(void); -#endif -static bool small_size2bin_init_hard(void); +static size_t bin_info_run_size_calc(arena_bin_info_t *bin_info, + size_t min_run_size); +static void bin_info_init(void); /******************************************************************************/ @@ -207,8 +100,8 @@ arena_run_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) } /* Generate red-black tree functions. */ -rb_gen(static JEMALLOC_ATTR(unused), arena_run_tree_, arena_run_tree_t, - arena_chunk_map_t, u.rb_link, arena_run_comp) +rb_gen(static UNUSED, arena_run_tree_, arena_run_tree_t, arena_chunk_map_t, + u.rb_link, arena_run_comp) static inline int arena_avail_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) @@ -217,9 +110,6 @@ arena_avail_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) size_t a_size = a->bits & ~PAGE_MASK; size_t b_size = b->bits & ~PAGE_MASK; - assert((a->bits & CHUNK_MAP_KEY) == CHUNK_MAP_KEY || (a->bits & - CHUNK_MAP_DIRTY) == (b->bits & CHUNK_MAP_DIRTY)); - ret = (a_size > b_size) - (a_size < b_size); if (ret == 0) { uintptr_t a_mapelm, b_mapelm; @@ -242,129 +132,311 @@ arena_avail_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) } /* Generate red-black tree functions. */ -rb_gen(static JEMALLOC_ATTR(unused), arena_avail_tree_, arena_avail_tree_t, - arena_chunk_map_t, u.rb_link, arena_avail_comp) +rb_gen(static UNUSED, arena_avail_tree_, arena_avail_tree_t, arena_chunk_map_t, + u.rb_link, arena_avail_comp) + +static inline int +arena_chunk_dirty_comp(arena_chunk_t *a, arena_chunk_t *b) +{ + + assert(a != NULL); + assert(b != NULL); + + /* + * Short-circuit for self comparison. The following comparison code + * would come to the same result, but at the cost of executing the slow + * path. + */ + if (a == b) + return (0); + + /* + * Order such that chunks with higher fragmentation are "less than" + * those with lower fragmentation -- purging order is from "least" to + * "greatest". Fragmentation is measured as: + * + * mean current avail run size + * -------------------------------- + * mean defragmented avail run size + * + * navail + * ----------- + * nruns_avail nruns_avail-nruns_adjac + * = ========================= = ----------------------- + * navail nruns_avail + * ----------------------- + * nruns_avail-nruns_adjac + * + * The following code multiplies away the denominator prior to + * comparison, in order to avoid division. + * + */ + { + size_t a_val = (a->nruns_avail - a->nruns_adjac) * + b->nruns_avail; + size_t b_val = (b->nruns_avail - b->nruns_adjac) * + a->nruns_avail; + + if (a_val < b_val) + return (1); + if (a_val > b_val) + return (-1); + } + /* + * Break ties by chunk address. For fragmented chunks, report lower + * addresses as "lower", so that fragmentation reduction happens first + * at lower addresses. However, use the opposite ordering for + * unfragmented chunks, in order to increase the chances of + * re-allocating dirty runs. + */ + { + uintptr_t a_chunk = (uintptr_t)a; + uintptr_t b_chunk = (uintptr_t)b; + int ret = ((a_chunk > b_chunk) - (a_chunk < b_chunk)); + if (a->nruns_adjac == 0) { + assert(b->nruns_adjac == 0); + ret = -ret; + } + return (ret); + } +} + +/* Generate red-black tree functions. */ +rb_gen(static UNUSED, arena_chunk_dirty_, arena_chunk_tree_t, arena_chunk_t, + dirty_link, arena_chunk_dirty_comp) + +static inline bool +arena_avail_adjac_pred(arena_chunk_t *chunk, size_t pageind) +{ + bool ret; + + if (pageind-1 < map_bias) + ret = false; + else { + ret = (arena_mapbits_allocated_get(chunk, pageind-1) == 0); + assert(ret == false || arena_mapbits_dirty_get(chunk, + pageind-1) != arena_mapbits_dirty_get(chunk, pageind)); + } + return (ret); +} + +static inline bool +arena_avail_adjac_succ(arena_chunk_t *chunk, size_t pageind, size_t npages) +{ + bool ret; + + if (pageind+npages == chunk_npages) + ret = false; + else { + assert(pageind+npages < chunk_npages); + ret = (arena_mapbits_allocated_get(chunk, pageind+npages) == 0); + assert(ret == false || arena_mapbits_dirty_get(chunk, pageind) + != arena_mapbits_dirty_get(chunk, pageind+npages)); + } + return (ret); +} + +static inline bool +arena_avail_adjac(arena_chunk_t *chunk, size_t pageind, size_t npages) +{ + + return (arena_avail_adjac_pred(chunk, pageind) || + arena_avail_adjac_succ(chunk, pageind, npages)); +} + +static void +arena_avail_insert(arena_t *arena, arena_chunk_t *chunk, size_t pageind, + size_t npages, bool maybe_adjac_pred, bool maybe_adjac_succ) +{ + + assert(npages == (arena_mapbits_unallocated_size_get(chunk, pageind) >> + LG_PAGE)); + + /* + * chunks_dirty is keyed by nruns_{avail,adjac}, so the chunk must be + * removed and reinserted even if the run to be inserted is clean. + */ + if (chunk->ndirty != 0) + arena_chunk_dirty_remove(&arena->chunks_dirty, chunk); + + if (maybe_adjac_pred && arena_avail_adjac_pred(chunk, pageind)) + chunk->nruns_adjac++; + if (maybe_adjac_succ && arena_avail_adjac_succ(chunk, pageind, npages)) + chunk->nruns_adjac++; + chunk->nruns_avail++; + assert(chunk->nruns_avail > chunk->nruns_adjac); + + if (arena_mapbits_dirty_get(chunk, pageind) != 0) { + arena->ndirty += npages; + chunk->ndirty += npages; + } + if (chunk->ndirty != 0) + arena_chunk_dirty_insert(&arena->chunks_dirty, chunk); + + arena_avail_tree_insert(&arena->runs_avail, arena_mapp_get(chunk, + pageind)); +} + +static void +arena_avail_remove(arena_t *arena, arena_chunk_t *chunk, size_t pageind, + size_t npages, bool maybe_adjac_pred, bool maybe_adjac_succ) +{ + + assert(npages == (arena_mapbits_unallocated_size_get(chunk, pageind) >> + LG_PAGE)); + + /* + * chunks_dirty is keyed by nruns_{avail,adjac}, so the chunk must be + * removed and reinserted even if the run to be removed is clean. + */ + if (chunk->ndirty != 0) + arena_chunk_dirty_remove(&arena->chunks_dirty, chunk); + + if (maybe_adjac_pred && arena_avail_adjac_pred(chunk, pageind)) + chunk->nruns_adjac--; + if (maybe_adjac_succ && arena_avail_adjac_succ(chunk, pageind, npages)) + chunk->nruns_adjac--; + chunk->nruns_avail--; + assert(chunk->nruns_avail > chunk->nruns_adjac || (chunk->nruns_avail + == 0 && chunk->nruns_adjac == 0)); + + if (arena_mapbits_dirty_get(chunk, pageind) != 0) { + arena->ndirty -= npages; + chunk->ndirty -= npages; + } + if (chunk->ndirty != 0) + arena_chunk_dirty_insert(&arena->chunks_dirty, chunk); + + arena_avail_tree_remove(&arena->runs_avail, arena_mapp_get(chunk, + pageind)); +} static inline void * -arena_run_reg_alloc(arena_run_t *run, arena_bin_t *bin) +arena_run_reg_alloc(arena_run_t *run, arena_bin_info_t *bin_info) { void *ret; + unsigned regind; + bitmap_t *bitmap = (bitmap_t *)((uintptr_t)run + + (uintptr_t)bin_info->bitmap_offset); - assert(run->magic == ARENA_RUN_MAGIC); assert(run->nfree > 0); + assert(bitmap_full(bitmap, &bin_info->bitmap_info) == false); + regind = bitmap_sfu(bitmap, &bin_info->bitmap_info); + ret = (void *)((uintptr_t)run + (uintptr_t)bin_info->reg0_offset + + (uintptr_t)(bin_info->reg_interval * regind)); run->nfree--; - ret = run->avail; - if (ret != NULL) { - /* Double free can cause assertion failure.*/ - assert(ret != NULL); - /* Write-after free can cause assertion failure. */ - assert((uintptr_t)ret >= (uintptr_t)run + - (uintptr_t)bin->reg0_offset); - assert((uintptr_t)ret < (uintptr_t)run->next); - assert(((uintptr_t)ret - ((uintptr_t)run + - (uintptr_t)bin->reg0_offset)) % (uintptr_t)bin->reg_size == - 0); - run->avail = *(void **)ret; - return (ret); - } - ret = run->next; - run->next = (void *)((uintptr_t)ret + (uintptr_t)bin->reg_size); - assert(ret != NULL); + if (regind == run->nextind) + run->nextind++; + assert(regind < run->nextind); return (ret); } static inline void arena_run_reg_dalloc(arena_run_t *run, void *ptr) { - - assert(run->nfree < run->bin->nregs); + arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); + size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + size_t mapbits = arena_mapbits_get(chunk, pageind); + size_t binind = arena_ptr_small_binind_get(ptr, mapbits); + arena_bin_info_t *bin_info = &arena_bin_info[binind]; + unsigned regind = arena_run_regind(run, bin_info, ptr); + bitmap_t *bitmap = (bitmap_t *)((uintptr_t)run + + (uintptr_t)bin_info->bitmap_offset); + + assert(run->nfree < bin_info->nregs); /* Freeing an interior pointer can cause assertion failure. */ assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)run->bin->reg0_offset)) % (uintptr_t)run->bin->reg_size - == 0); - /* - * Freeing a pointer lower than region zero can cause assertion - * failure. - */ + (uintptr_t)bin_info->reg0_offset)) % + (uintptr_t)bin_info->reg_interval == 0); assert((uintptr_t)ptr >= (uintptr_t)run + - (uintptr_t)run->bin->reg0_offset); - /* - * Freeing a pointer past in the run's frontier can cause assertion - * failure. - */ - assert((uintptr_t)ptr < (uintptr_t)run->next); + (uintptr_t)bin_info->reg0_offset); + /* Freeing an unallocated pointer can cause assertion failure. */ + assert(bitmap_get(bitmap, &bin_info->bitmap_info, regind)); - *(void **)ptr = run->avail; - run->avail = ptr; + bitmap_unset(bitmap, &bin_info->bitmap_info, regind); run->nfree++; } -#ifdef JEMALLOC_DEBUG static inline void -arena_chunk_validate_zeroed(arena_chunk_t *chunk, size_t run_ind) +arena_run_zero(arena_chunk_t *chunk, size_t run_ind, size_t npages) +{ + + VALGRIND_MAKE_MEM_UNDEFINED((void *)((uintptr_t)chunk + (run_ind << + LG_PAGE)), (npages << LG_PAGE)); + memset((void *)((uintptr_t)chunk + (run_ind << LG_PAGE)), 0, + (npages << LG_PAGE)); +} + +static inline void +arena_run_page_validate_zeroed(arena_chunk_t *chunk, size_t run_ind) { size_t i; - size_t *p = (size_t *)((uintptr_t)chunk + (run_ind << PAGE_SHIFT)); + UNUSED size_t *p = (size_t *)((uintptr_t)chunk + (run_ind << LG_PAGE)); - for (i = 0; i < PAGE_SIZE / sizeof(size_t); i++) + VALGRIND_MAKE_MEM_DEFINED((void *)((uintptr_t)chunk + (run_ind << + LG_PAGE)), PAGE); + for (i = 0; i < PAGE / sizeof(size_t); i++) assert(p[i] == 0); } -#endif static void arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, - bool zero) + size_t binind, bool zero) { arena_chunk_t *chunk; - size_t old_ndirty, run_ind, total_pages, need_pages, rem_pages, i; + size_t run_ind, total_pages, need_pages, rem_pages, i; size_t flag_dirty; - arena_avail_tree_t *runs_avail; + + assert((large && binind == BININD_INVALID) || (large == false && binind + != BININD_INVALID)); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - old_ndirty = chunk->ndirty; - run_ind = (unsigned)(((uintptr_t)run - (uintptr_t)chunk) - >> PAGE_SHIFT); - flag_dirty = chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY; - runs_avail = (flag_dirty != 0) ? &arena->runs_avail_dirty : - &arena->runs_avail_clean; - total_pages = (chunk->map[run_ind-map_bias].bits & ~PAGE_MASK) >> - PAGE_SHIFT; - assert((chunk->map[run_ind+total_pages-1-map_bias].bits & - CHUNK_MAP_DIRTY) == flag_dirty); - need_pages = (size >> PAGE_SHIFT); + run_ind = (unsigned)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE); + flag_dirty = arena_mapbits_dirty_get(chunk, run_ind); + total_pages = arena_mapbits_unallocated_size_get(chunk, run_ind) >> + LG_PAGE; + assert(arena_mapbits_dirty_get(chunk, run_ind+total_pages-1) == + flag_dirty); + need_pages = (size >> LG_PAGE); assert(need_pages > 0); assert(need_pages <= total_pages); rem_pages = total_pages - need_pages; - arena_avail_tree_remove(runs_avail, &chunk->map[run_ind-map_bias]); + arena_avail_remove(arena, chunk, run_ind, total_pages, true, true); + if (config_stats) { + /* + * Update stats_cactive if nactive is crossing a chunk + * multiple. + */ + size_t cactive_diff = CHUNK_CEILING((arena->nactive + + need_pages) << LG_PAGE) - CHUNK_CEILING(arena->nactive << + LG_PAGE); + if (cactive_diff != 0) + stats_cactive_add(cactive_diff); + } arena->nactive += need_pages; /* Keep track of trailing unused pages for later use. */ if (rem_pages > 0) { if (flag_dirty != 0) { - chunk->map[run_ind+need_pages-map_bias].bits = - (rem_pages << PAGE_SHIFT) | CHUNK_MAP_DIRTY; - chunk->map[run_ind+total_pages-1-map_bias].bits = - (rem_pages << PAGE_SHIFT) | CHUNK_MAP_DIRTY; + arena_mapbits_unallocated_set(chunk, run_ind+need_pages, + (rem_pages << LG_PAGE), CHUNK_MAP_DIRTY); + arena_mapbits_unallocated_set(chunk, + run_ind+total_pages-1, (rem_pages << LG_PAGE), + CHUNK_MAP_DIRTY); } else { - chunk->map[run_ind+need_pages-map_bias].bits = - (rem_pages << PAGE_SHIFT) | - (chunk->map[run_ind+need_pages-map_bias].bits & - CHUNK_MAP_UNZEROED); - chunk->map[run_ind+total_pages-1-map_bias].bits = - (rem_pages << PAGE_SHIFT) | - (chunk->map[run_ind+total_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED); + arena_mapbits_unallocated_set(chunk, run_ind+need_pages, + (rem_pages << LG_PAGE), + arena_mapbits_unzeroed_get(chunk, + run_ind+need_pages)); + arena_mapbits_unallocated_set(chunk, + run_ind+total_pages-1, (rem_pages << LG_PAGE), + arena_mapbits_unzeroed_get(chunk, + run_ind+total_pages-1)); } - arena_avail_tree_insert(runs_avail, - &chunk->map[run_ind+need_pages-map_bias]); - } - - /* Update dirty page accounting. */ - if (flag_dirty != 0) { - chunk->ndirty -= need_pages; - arena->ndirty -= need_pages; + arena_avail_insert(arena, chunk, run_ind+need_pages, rem_pages, + false, true); } /* @@ -379,28 +451,21 @@ arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, * zeroed (i.e. never before touched). */ for (i = 0; i < need_pages; i++) { - if ((chunk->map[run_ind+i-map_bias].bits - & CHUNK_MAP_UNZEROED) != 0) { - memset((void *)((uintptr_t) - chunk + ((run_ind+i) << - PAGE_SHIFT)), 0, - PAGE_SIZE); - } -#ifdef JEMALLOC_DEBUG - else { - arena_chunk_validate_zeroed( + if (arena_mapbits_unzeroed_get(chunk, + run_ind+i) != 0) { + arena_run_zero(chunk, run_ind+i, + 1); + } else if (config_debug) { + arena_run_page_validate_zeroed( chunk, run_ind+i); } -#endif } } else { /* * The run is dirty, so all pages must be * zeroed. */ - memset((void *)((uintptr_t)chunk + (run_ind << - PAGE_SHIFT)), 0, (need_pages << - PAGE_SHIFT)); + arena_run_zero(chunk, run_ind, need_pages); } } @@ -408,10 +473,9 @@ arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, * Set the last element first, in case the run only contains one * page (i.e. both statements set the same element). */ - chunk->map[run_ind+need_pages-1-map_bias].bits = - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED | flag_dirty; - chunk->map[run_ind-map_bias].bits = size | flag_dirty | - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + arena_mapbits_large_set(chunk, run_ind+need_pages-1, 0, + flag_dirty); + arena_mapbits_large_set(chunk, run_ind, size, flag_dirty); } else { assert(zero == false); /* @@ -419,44 +483,34 @@ arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, * small run, so that arena_dalloc_bin_run() has the ability to * conditionally trim clean pages. */ - chunk->map[run_ind-map_bias].bits = - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_UNZEROED) | - CHUNK_MAP_ALLOCATED | flag_dirty; -#ifdef JEMALLOC_DEBUG + arena_mapbits_small_set(chunk, run_ind, 0, binind, flag_dirty); /* * The first page will always be dirtied during small run * initialization, so a validation failure here would not * actually cause an observable failure. */ - if (flag_dirty == 0 && - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_UNZEROED) - == 0) - arena_chunk_validate_zeroed(chunk, run_ind); -#endif + if (config_debug && flag_dirty == 0 && + arena_mapbits_unzeroed_get(chunk, run_ind) == 0) + arena_run_page_validate_zeroed(chunk, run_ind); for (i = 1; i < need_pages - 1; i++) { - chunk->map[run_ind+i-map_bias].bits = (i << PAGE_SHIFT) - | (chunk->map[run_ind+i-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_ALLOCATED; -#ifdef JEMALLOC_DEBUG - if (flag_dirty == 0 && - (chunk->map[run_ind+i-map_bias].bits & - CHUNK_MAP_UNZEROED) == 0) - arena_chunk_validate_zeroed(chunk, run_ind+i); -#endif + arena_mapbits_small_set(chunk, run_ind+i, i, binind, 0); + if (config_debug && flag_dirty == 0 && + arena_mapbits_unzeroed_get(chunk, run_ind+i) == 0) { + arena_run_page_validate_zeroed(chunk, + run_ind+i); + } } - chunk->map[run_ind+need_pages-1-map_bias].bits = ((need_pages - - 1) << PAGE_SHIFT) | - (chunk->map[run_ind+need_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_ALLOCATED | flag_dirty; -#ifdef JEMALLOC_DEBUG - if (flag_dirty == 0 && - (chunk->map[run_ind+need_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) == 0) { - arena_chunk_validate_zeroed(chunk, + arena_mapbits_small_set(chunk, run_ind+need_pages-1, + need_pages-1, binind, flag_dirty); + if (config_debug && flag_dirty == 0 && + arena_mapbits_unzeroed_get(chunk, run_ind+need_pages-1) == + 0) { + arena_run_page_validate_zeroed(chunk, run_ind+need_pages-1); } -#endif } + VALGRIND_MAKE_MEM_UNDEFINED((void *)((uintptr_t)chunk + (run_ind << + LG_PAGE)), (need_pages << LG_PAGE)); } static arena_chunk_t * @@ -466,40 +520,32 @@ arena_chunk_alloc(arena_t *arena) size_t i; if (arena->spare != NULL) { - arena_avail_tree_t *runs_avail; - chunk = arena->spare; arena->spare = NULL; - /* Insert the run into the appropriate runs_avail_* tree. */ - if ((chunk->map[0].bits & CHUNK_MAP_DIRTY) == 0) - runs_avail = &arena->runs_avail_clean; - else - runs_avail = &arena->runs_avail_dirty; - assert((chunk->map[0].bits & ~PAGE_MASK) == arena_maxclass); - assert((chunk->map[chunk_npages-1-map_bias].bits & ~PAGE_MASK) - == arena_maxclass); - assert((chunk->map[0].bits & CHUNK_MAP_DIRTY) == - (chunk->map[chunk_npages-1-map_bias].bits & - CHUNK_MAP_DIRTY)); - arena_avail_tree_insert(runs_avail, &chunk->map[0]); + assert(arena_mapbits_allocated_get(chunk, map_bias) == 0); + assert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0); + assert(arena_mapbits_unallocated_size_get(chunk, map_bias) == + arena_maxclass); + assert(arena_mapbits_unallocated_size_get(chunk, + chunk_npages-1) == arena_maxclass); + assert(arena_mapbits_dirty_get(chunk, map_bias) == + arena_mapbits_dirty_get(chunk, chunk_npages-1)); } else { bool zero; size_t unzeroed; zero = false; malloc_mutex_unlock(&arena->lock); - chunk = (arena_chunk_t *)chunk_alloc(chunksize, false, &zero); + chunk = (arena_chunk_t *)chunk_alloc(chunksize, chunksize, + false, &zero, arena->dss_prec); malloc_mutex_lock(&arena->lock); if (chunk == NULL) return (NULL); -#ifdef JEMALLOC_STATS - arena->stats.mapped += chunksize; -#endif + if (config_stats) + arena->stats.mapped += chunksize; chunk->arena = arena; - ql_elm_new(chunk, link_dirty); - chunk->dirtied = false; /* * Claim that no pages are in use, since the header is merely @@ -507,85 +553,87 @@ arena_chunk_alloc(arena_t *arena) */ chunk->ndirty = 0; + chunk->nruns_avail = 0; + chunk->nruns_adjac = 0; + /* * Initialize the map to contain one maximal free untouched run. * Mark the pages as zeroed iff chunk_alloc() returned a zeroed * chunk. */ unzeroed = zero ? 0 : CHUNK_MAP_UNZEROED; - chunk->map[0].bits = arena_maxclass | unzeroed; + arena_mapbits_unallocated_set(chunk, map_bias, arena_maxclass, + unzeroed); /* * There is no need to initialize the internal page map entries * unless the chunk is not zeroed. */ if (zero == false) { for (i = map_bias+1; i < chunk_npages-1; i++) - chunk->map[i-map_bias].bits = unzeroed; - } -#ifdef JEMALLOC_DEBUG - else { - for (i = map_bias+1; i < chunk_npages-1; i++) - assert(chunk->map[i-map_bias].bits == unzeroed); + arena_mapbits_unzeroed_set(chunk, i, unzeroed); + } else if (config_debug) { + VALGRIND_MAKE_MEM_DEFINED( + (void *)arena_mapp_get(chunk, map_bias+1), + (void *)((uintptr_t) + arena_mapp_get(chunk, chunk_npages-1) + - (uintptr_t)arena_mapp_get(chunk, map_bias+1))); + for (i = map_bias+1; i < chunk_npages-1; i++) { + assert(arena_mapbits_unzeroed_get(chunk, i) == + unzeroed); + } } -#endif - chunk->map[chunk_npages-1-map_bias].bits = arena_maxclass | - unzeroed; - - /* Insert the run into the runs_avail_clean tree. */ - arena_avail_tree_insert(&arena->runs_avail_clean, - &chunk->map[0]); + arena_mapbits_unallocated_set(chunk, chunk_npages-1, + arena_maxclass, unzeroed); } + /* Insert the run into the runs_avail tree. */ + arena_avail_insert(arena, chunk, map_bias, chunk_npages-map_bias, + false, false); + return (chunk); } static void arena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk) { - arena_avail_tree_t *runs_avail; + assert(arena_mapbits_allocated_get(chunk, map_bias) == 0); + assert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0); + assert(arena_mapbits_unallocated_size_get(chunk, map_bias) == + arena_maxclass); + assert(arena_mapbits_unallocated_size_get(chunk, chunk_npages-1) == + arena_maxclass); + assert(arena_mapbits_dirty_get(chunk, map_bias) == + arena_mapbits_dirty_get(chunk, chunk_npages-1)); /* - * Remove run from the appropriate runs_avail_* tree, so that the arena - * does not use it. + * Remove run from the runs_avail tree, so that the arena does not use + * it. */ - if ((chunk->map[0].bits & CHUNK_MAP_DIRTY) == 0) - runs_avail = &arena->runs_avail_clean; - else - runs_avail = &arena->runs_avail_dirty; - arena_avail_tree_remove(runs_avail, &chunk->map[0]); + arena_avail_remove(arena, chunk, map_bias, chunk_npages-map_bias, + false, false); if (arena->spare != NULL) { arena_chunk_t *spare = arena->spare; arena->spare = chunk; - if (spare->dirtied) { - ql_remove(&chunk->arena->chunks_dirty, spare, - link_dirty); - arena->ndirty -= spare->ndirty; - } malloc_mutex_unlock(&arena->lock); - chunk_dealloc((void *)spare, chunksize); + chunk_dealloc((void *)spare, chunksize, true); malloc_mutex_lock(&arena->lock); -#ifdef JEMALLOC_STATS - arena->stats.mapped -= chunksize; -#endif + if (config_stats) + arena->stats.mapped -= chunksize; } else arena->spare = chunk; } static arena_run_t * -arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) +arena_run_alloc_helper(arena_t *arena, size_t size, bool large, size_t binind, + bool zero) { - arena_chunk_t *chunk; arena_run_t *run; arena_chunk_map_t *mapelm, key; - assert(size <= arena_maxclass); - assert((size & PAGE_MASK) == 0); - - /* Search the arena's chunks for the lowest best fit. */ key.bits = size | CHUNK_MAP_KEY; - mapelm = arena_avail_tree_nsearch(&arena->runs_avail_dirty, &key); + mapelm = arena_avail_tree_nsearch(&arena->runs_avail, &key); if (mapelm != NULL) { arena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm); size_t pageind = (((uintptr_t)mapelm - @@ -593,31 +641,38 @@ arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) + map_bias; run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); + LG_PAGE)); + arena_run_split(arena, run, size, large, binind, zero); return (run); } - mapelm = arena_avail_tree_nsearch(&arena->runs_avail_clean, &key); - if (mapelm != NULL) { - arena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm); - size_t pageind = (((uintptr_t)mapelm - - (uintptr_t)run_chunk->map) / sizeof(arena_chunk_map_t)) - + map_bias; - run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); + return (NULL); +} + +static arena_run_t * +arena_run_alloc(arena_t *arena, size_t size, bool large, size_t binind, + bool zero) +{ + arena_chunk_t *chunk; + arena_run_t *run; + + assert(size <= arena_maxclass); + assert((size & PAGE_MASK) == 0); + assert((large && binind == BININD_INVALID) || (large == false && binind + != BININD_INVALID)); + + /* Search the arena's chunks for the lowest best fit. */ + run = arena_run_alloc_helper(arena, size, large, binind, zero); + if (run != NULL) return (run); - } /* * No usable runs. Create a new chunk from which to allocate the run. */ chunk = arena_chunk_alloc(arena); if (chunk != NULL) { - run = (arena_run_t *)((uintptr_t)chunk + (map_bias << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); + run = (arena_run_t *)((uintptr_t)chunk + (map_bias << LG_PAGE)); + arena_run_split(arena, run, size, large, binind, zero); return (run); } @@ -626,78 +681,46 @@ arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) * sufficient memory available while this one dropped arena->lock in * arena_chunk_alloc(), so search one more time. */ - mapelm = arena_avail_tree_nsearch(&arena->runs_avail_dirty, &key); - if (mapelm != NULL) { - arena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm); - size_t pageind = (((uintptr_t)mapelm - - (uintptr_t)run_chunk->map) / sizeof(arena_chunk_map_t)) - + map_bias; - - run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); - return (run); - } - mapelm = arena_avail_tree_nsearch(&arena->runs_avail_clean, &key); - if (mapelm != NULL) { - arena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm); - size_t pageind = (((uintptr_t)mapelm - - (uintptr_t)run_chunk->map) / sizeof(arena_chunk_map_t)) - + map_bias; - - run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); - return (run); - } - - return (NULL); + return (arena_run_alloc_helper(arena, size, large, binind, zero)); } static inline void arena_maybe_purge(arena_t *arena) { + size_t npurgeable, threshold; + + /* Don't purge if the option is disabled. */ + if (opt_lg_dirty_mult < 0) + return; + /* Don't purge if all dirty pages are already being purged. */ + if (arena->ndirty <= arena->npurgatory) + return; + npurgeable = arena->ndirty - arena->npurgatory; + threshold = (arena->nactive >> opt_lg_dirty_mult); + /* + * Don't purge unless the number of purgeable pages exceeds the + * threshold. + */ + if (npurgeable <= threshold) + return; - /* Enforce opt_lg_dirty_mult. */ - if (opt_lg_dirty_mult >= 0 && arena->ndirty > arena->npurgatory && - (arena->ndirty - arena->npurgatory) > chunk_npages && - (arena->nactive >> opt_lg_dirty_mult) < (arena->ndirty - - arena->npurgatory)) - arena_purge(arena, false); + arena_purge(arena, false); } -static inline void -arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) +static inline size_t +arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk, bool all) { + size_t npurged; ql_head(arena_chunk_map_t) mapelms; arena_chunk_map_t *mapelm; - size_t pageind, flag_unzeroed; -#ifdef JEMALLOC_DEBUG - size_t ndirty; -#endif -#ifdef JEMALLOC_STATS + size_t pageind, npages; size_t nmadvise; -#endif ql_new(&mapelms); - flag_unzeroed = -#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED - /* - * madvise(..., MADV_DONTNEED) results in zero-filled pages for anonymous - * mappings, but not for file-backed mappings. - */ -# ifdef JEMALLOC_SWAP - swap_enabled ? CHUNK_MAP_UNZEROED : -# endif - 0; -#else - CHUNK_MAP_UNZEROED; -#endif - /* * If chunk is the spare, temporarily re-allocate it, 1) so that its - * run is reinserted into runs_avail_dirty, and 2) so that it cannot be + * run is reinserted into runs_avail, and 2) so that it cannot be * completely discarded by another thread while arena->lock is dropped * by this thread. Note that the arena_run_dalloc() call will * implicitly deallocate the chunk, so no explicit action is required @@ -711,124 +734,135 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) * run. */ if (chunk == arena->spare) { - assert((chunk->map[0].bits & CHUNK_MAP_DIRTY) != 0); + assert(arena_mapbits_dirty_get(chunk, map_bias) != 0); + assert(arena_mapbits_dirty_get(chunk, chunk_npages-1) != 0); + arena_chunk_alloc(arena); } - /* Temporarily allocate all free dirty runs within chunk. */ - for (pageind = map_bias; pageind < chunk_npages;) { - mapelm = &chunk->map[pageind-map_bias]; - if ((mapelm->bits & CHUNK_MAP_ALLOCATED) == 0) { - size_t npages; + if (config_stats) + arena->stats.purged += chunk->ndirty; - npages = mapelm->bits >> PAGE_SHIFT; - assert(pageind + npages <= chunk_npages); - if (mapelm->bits & CHUNK_MAP_DIRTY) { - size_t i; + /* + * Operate on all dirty runs if there is no clean/dirty run + * fragmentation. + */ + if (chunk->nruns_adjac == 0) + all = true; - arena_avail_tree_remove( - &arena->runs_avail_dirty, mapelm); + /* + * Temporarily allocate free dirty runs within chunk. If all is false, + * only operate on dirty runs that are fragments; otherwise operate on + * all dirty runs. + */ + for (pageind = map_bias; pageind < chunk_npages; pageind += npages) { + mapelm = arena_mapp_get(chunk, pageind); + if (arena_mapbits_allocated_get(chunk, pageind) == 0) { + size_t run_size = + arena_mapbits_unallocated_size_get(chunk, pageind); - mapelm->bits = (npages << PAGE_SHIFT) | - flag_unzeroed | CHUNK_MAP_LARGE | - CHUNK_MAP_ALLOCATED; - /* - * Update internal elements in the page map, so - * that CHUNK_MAP_UNZEROED is properly set. - */ - for (i = 1; i < npages - 1; i++) { - chunk->map[pageind+i-map_bias].bits = - flag_unzeroed; - } - if (npages > 1) { - chunk->map[ - pageind+npages-1-map_bias].bits = - flag_unzeroed | CHUNK_MAP_LARGE | - CHUNK_MAP_ALLOCATED; - } + npages = run_size >> LG_PAGE; + assert(pageind + npages <= chunk_npages); + assert(arena_mapbits_dirty_get(chunk, pageind) == + arena_mapbits_dirty_get(chunk, pageind+npages-1)); - arena->nactive += npages; + if (arena_mapbits_dirty_get(chunk, pageind) != 0 && + (all || arena_avail_adjac(chunk, pageind, + npages))) { + arena_run_t *run = (arena_run_t *)((uintptr_t) + chunk + (uintptr_t)(pageind << LG_PAGE)); + + arena_run_split(arena, run, run_size, true, + BININD_INVALID, false); /* Append to list for later processing. */ ql_elm_new(mapelm, u.ql_link); ql_tail_insert(&mapelms, mapelm, u.ql_link); } - - pageind += npages; } else { - /* Skip allocated run. */ - if (mapelm->bits & CHUNK_MAP_LARGE) - pageind += mapelm->bits >> PAGE_SHIFT; - else { + /* Skip run. */ + if (arena_mapbits_large_get(chunk, pageind) != 0) { + npages = arena_mapbits_large_size_get(chunk, + pageind) >> LG_PAGE; + } else { + size_t binind; + arena_bin_info_t *bin_info; arena_run_t *run = (arena_run_t *)((uintptr_t) - chunk + (uintptr_t)(pageind << PAGE_SHIFT)); + chunk + (uintptr_t)(pageind << LG_PAGE)); - assert((mapelm->bits >> PAGE_SHIFT) == 0); - assert(run->magic == ARENA_RUN_MAGIC); - pageind += run->bin->run_size >> PAGE_SHIFT; + assert(arena_mapbits_small_runind_get(chunk, + pageind) == 0); + binind = arena_bin_index(arena, run->bin); + bin_info = &arena_bin_info[binind]; + npages = bin_info->run_size >> LG_PAGE; } } } assert(pageind == chunk_npages); - -#ifdef JEMALLOC_DEBUG - ndirty = chunk->ndirty; -#endif -#ifdef JEMALLOC_STATS - arena->stats.purged += chunk->ndirty; -#endif - arena->ndirty -= chunk->ndirty; - chunk->ndirty = 0; - ql_remove(&arena->chunks_dirty, chunk, link_dirty); - chunk->dirtied = false; + assert(chunk->ndirty == 0 || all == false); + assert(chunk->nruns_adjac == 0); malloc_mutex_unlock(&arena->lock); -#ifdef JEMALLOC_STATS - nmadvise = 0; -#endif + if (config_stats) + nmadvise = 0; + npurged = 0; ql_foreach(mapelm, &mapelms, u.ql_link) { - size_t pageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) / - sizeof(arena_chunk_map_t)) + map_bias; - size_t npages = mapelm->bits >> PAGE_SHIFT; + bool unzeroed; + size_t flag_unzeroed, i; + pageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) / + sizeof(arena_chunk_map_t)) + map_bias; + npages = arena_mapbits_large_size_get(chunk, pageind) >> + LG_PAGE; assert(pageind + npages <= chunk_npages); -#ifdef JEMALLOC_DEBUG - assert(ndirty >= npages); - ndirty -= npages; -#endif - -#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED - madvise((void *)((uintptr_t)chunk + (pageind << PAGE_SHIFT)), - (npages << PAGE_SHIFT), MADV_DONTNEED); -#elif defined(JEMALLOC_PURGE_MADVISE_FREE) - madvise((void *)((uintptr_t)chunk + (pageind << PAGE_SHIFT)), - (npages << PAGE_SHIFT), MADV_FREE); -#else -# error "No method defined for purging unused dirty pages." -#endif - -#ifdef JEMALLOC_STATS - nmadvise++; -#endif - } -#ifdef JEMALLOC_DEBUG - assert(ndirty == 0); -#endif + unzeroed = pages_purge((void *)((uintptr_t)chunk + (pageind << + LG_PAGE)), (npages << LG_PAGE)); + flag_unzeroed = unzeroed ? CHUNK_MAP_UNZEROED : 0; + /* + * Set the unzeroed flag for all pages, now that pages_purge() + * has returned whether the pages were zeroed as a side effect + * of purging. This chunk map modification is safe even though + * the arena mutex isn't currently owned by this thread, + * because the run is marked as allocated, thus protecting it + * from being modified by any other thread. As long as these + * writes don't perturb the first and last elements' + * CHUNK_MAP_ALLOCATED bits, behavior is well defined. + */ + for (i = 0; i < npages; i++) { + arena_mapbits_unzeroed_set(chunk, pageind+i, + flag_unzeroed); + } + npurged += npages; + if (config_stats) + nmadvise++; + } malloc_mutex_lock(&arena->lock); -#ifdef JEMALLOC_STATS - arena->stats.nmadvise += nmadvise; -#endif + if (config_stats) + arena->stats.nmadvise += nmadvise; /* Deallocate runs. */ for (mapelm = ql_first(&mapelms); mapelm != NULL; mapelm = ql_first(&mapelms)) { - size_t pageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) / - sizeof(arena_chunk_map_t)) + map_bias; - arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)(pageind << PAGE_SHIFT)); + arena_run_t *run; + pageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) / + sizeof(arena_chunk_map_t)) + map_bias; + run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)(pageind << + LG_PAGE)); ql_remove(&mapelms, mapelm, u.ql_link); - arena_run_dalloc(arena, run, false); + arena_run_dalloc(arena, run, false, true); } + + return (npurged); +} + +static arena_chunk_t * +chunks_dirty_iter_cb(arena_chunk_tree_t *tree, arena_chunk_t *chunk, void *arg) +{ + size_t *ndirty = (size_t *)arg; + + assert(chunk->ndirty != 0); + *ndirty += chunk->ndirty; + return (NULL); } static void @@ -836,36 +870,43 @@ arena_purge(arena_t *arena, bool all) { arena_chunk_t *chunk; size_t npurgatory; -#ifdef JEMALLOC_DEBUG - size_t ndirty = 0; + if (config_debug) { + size_t ndirty = 0; - ql_foreach(chunk, &arena->chunks_dirty, link_dirty) { - assert(chunk->dirtied); - ndirty += chunk->ndirty; + arena_chunk_dirty_iter(&arena->chunks_dirty, NULL, + chunks_dirty_iter_cb, (void *)&ndirty); + assert(ndirty == arena->ndirty); } - assert(ndirty == arena->ndirty); -#endif - assert(arena->ndirty > arena->npurgatory); - assert(arena->ndirty > chunk_npages || all); - assert((arena->nactive >> opt_lg_dirty_mult) < arena->ndirty || all); + assert(arena->ndirty > arena->npurgatory || all); + assert((arena->nactive >> opt_lg_dirty_mult) < (arena->ndirty - + arena->npurgatory) || all); -#ifdef JEMALLOC_STATS - arena->stats.npurge++; -#endif + if (config_stats) + arena->stats.npurge++; /* * Compute the minimum number of pages that this thread should try to * purge, and add the result to arena->npurgatory. This will keep * multiple threads from racing to reduce ndirty below the threshold. */ - npurgatory = arena->ndirty - arena->npurgatory; - if (all == false) - npurgatory -= arena->nactive >> opt_lg_dirty_mult; + { + size_t npurgeable = arena->ndirty - arena->npurgatory; + + if (all == false) { + size_t threshold = (arena->nactive >> + opt_lg_dirty_mult); + + npurgatory = npurgeable - threshold; + } else + npurgatory = npurgeable; + } arena->npurgatory += npurgatory; while (npurgatory > 0) { + size_t npurgeable, npurged, nunpurged; + /* Get next chunk with dirty pages. */ - chunk = ql_first(&arena->chunks_dirty); + chunk = arena_chunk_dirty_first(&arena->chunks_dirty); if (chunk == NULL) { /* * This thread was unable to purge as many pages as @@ -876,23 +917,15 @@ arena_purge(arena_t *arena, bool all) arena->npurgatory -= npurgatory; return; } - while (chunk->ndirty == 0) { - ql_remove(&arena->chunks_dirty, chunk, link_dirty); - chunk->dirtied = false; - chunk = ql_first(&arena->chunks_dirty); - if (chunk == NULL) { - /* Same logic as for above. */ - arena->npurgatory -= npurgatory; - return; - } - } + npurgeable = chunk->ndirty; + assert(npurgeable != 0); - if (chunk->ndirty > npurgatory) { + if (npurgeable > npurgatory && chunk->nruns_adjac == 0) { /* - * This thread will, at a minimum, purge all the dirty - * pages in chunk, so set npurgatory to reflect this - * thread's commitment to purge the pages. This tends - * to reduce the chances of the following scenario: + * This thread will purge all the dirty pages in chunk, + * so set npurgatory to reflect this thread's intent to + * purge the pages. This tends to reduce the chances + * of the following scenario: * * 1) This thread sets arena->npurgatory such that * (arena->ndirty - arena->npurgatory) is at the @@ -906,13 +939,20 @@ arena_purge(arena_t *arena, bool all) * because all of the purging work being done really * needs to happen. */ - arena->npurgatory += chunk->ndirty - npurgatory; - npurgatory = chunk->ndirty; + arena->npurgatory += npurgeable - npurgatory; + npurgatory = npurgeable; } - arena->npurgatory -= chunk->ndirty; - npurgatory -= chunk->ndirty; - arena_chunk_purge(arena, chunk); + /* + * Keep track of how many pages are purgeable, versus how many + * actually get purged, and adjust counters accordingly. + */ + arena->npurgatory -= npurgeable; + npurgatory -= npurgeable; + npurged = arena_chunk_purge(arena, chunk, all); + nunpurged = npurgeable - npurged; + arena->npurgatory += nunpurged; + npurgatory += nunpurged; } } @@ -926,96 +966,95 @@ arena_purge_all(arena_t *arena) } static void -arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty) +arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty, bool cleaned) { arena_chunk_t *chunk; size_t size, run_ind, run_pages, flag_dirty; - arena_avail_tree_t *runs_avail; chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) - >> PAGE_SHIFT); + run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE); assert(run_ind >= map_bias); assert(run_ind < chunk_npages); - if ((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_LARGE) != 0) { - size = chunk->map[run_ind-map_bias].bits & ~PAGE_MASK; - assert(size == PAGE_SIZE || - (chunk->map[run_ind+(size>>PAGE_SHIFT)-1-map_bias].bits & - ~PAGE_MASK) == 0); - assert((chunk->map[run_ind+(size>>PAGE_SHIFT)-1-map_bias].bits & - CHUNK_MAP_LARGE) != 0); - assert((chunk->map[run_ind+(size>>PAGE_SHIFT)-1-map_bias].bits & - CHUNK_MAP_ALLOCATED) != 0); - } else - size = run->bin->run_size; - run_pages = (size >> PAGE_SHIFT); + if (arena_mapbits_large_get(chunk, run_ind) != 0) { + size = arena_mapbits_large_size_get(chunk, run_ind); + assert(size == PAGE || + arena_mapbits_large_size_get(chunk, + run_ind+(size>>LG_PAGE)-1) == 0); + } else { + size_t binind = arena_bin_index(arena, run->bin); + arena_bin_info_t *bin_info = &arena_bin_info[binind]; + size = bin_info->run_size; + } + run_pages = (size >> LG_PAGE); + if (config_stats) { + /* + * Update stats_cactive if nactive is crossing a chunk + * multiple. + */ + size_t cactive_diff = CHUNK_CEILING(arena->nactive << LG_PAGE) - + CHUNK_CEILING((arena->nactive - run_pages) << LG_PAGE); + if (cactive_diff != 0) + stats_cactive_sub(cactive_diff); + } arena->nactive -= run_pages; /* * The run is dirty if the caller claims to have dirtied it, as well as - * if it was already dirty before being allocated. + * if it was already dirty before being allocated and the caller + * doesn't claim to have cleaned it. */ - if ((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) != 0) + assert(arena_mapbits_dirty_get(chunk, run_ind) == + arena_mapbits_dirty_get(chunk, run_ind+run_pages-1)); + if (cleaned == false && arena_mapbits_dirty_get(chunk, run_ind) != 0) dirty = true; flag_dirty = dirty ? CHUNK_MAP_DIRTY : 0; - runs_avail = dirty ? &arena->runs_avail_dirty : - &arena->runs_avail_clean; /* Mark pages as unallocated in the chunk map. */ if (dirty) { - chunk->map[run_ind-map_bias].bits = size | CHUNK_MAP_DIRTY; - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - CHUNK_MAP_DIRTY; - - chunk->ndirty += run_pages; - arena->ndirty += run_pages; + arena_mapbits_unallocated_set(chunk, run_ind, size, + CHUNK_MAP_DIRTY); + arena_mapbits_unallocated_set(chunk, run_ind+run_pages-1, size, + CHUNK_MAP_DIRTY); } else { - chunk->map[run_ind-map_bias].bits = size | - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_UNZEROED); - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - (chunk->map[run_ind+run_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED); + arena_mapbits_unallocated_set(chunk, run_ind, size, + arena_mapbits_unzeroed_get(chunk, run_ind)); + arena_mapbits_unallocated_set(chunk, run_ind+run_pages-1, size, + arena_mapbits_unzeroed_get(chunk, run_ind+run_pages-1)); } /* Try to coalesce forward. */ if (run_ind + run_pages < chunk_npages && - (chunk->map[run_ind+run_pages-map_bias].bits & CHUNK_MAP_ALLOCATED) - == 0 && (chunk->map[run_ind+run_pages-map_bias].bits & - CHUNK_MAP_DIRTY) == flag_dirty) { - size_t nrun_size = chunk->map[run_ind+run_pages-map_bias].bits & - ~PAGE_MASK; - size_t nrun_pages = nrun_size >> PAGE_SHIFT; + arena_mapbits_allocated_get(chunk, run_ind+run_pages) == 0 && + arena_mapbits_dirty_get(chunk, run_ind+run_pages) == flag_dirty) { + size_t nrun_size = arena_mapbits_unallocated_size_get(chunk, + run_ind+run_pages); + size_t nrun_pages = nrun_size >> LG_PAGE; /* * Remove successor from runs_avail; the coalesced run is * inserted later. */ - assert((chunk->map[run_ind+run_pages+nrun_pages-1-map_bias].bits - & ~PAGE_MASK) == nrun_size); - assert((chunk->map[run_ind+run_pages+nrun_pages-1-map_bias].bits - & CHUNK_MAP_ALLOCATED) == 0); - assert((chunk->map[run_ind+run_pages+nrun_pages-1-map_bias].bits - & CHUNK_MAP_DIRTY) == flag_dirty); - arena_avail_tree_remove(runs_avail, - &chunk->map[run_ind+run_pages-map_bias]); + assert(arena_mapbits_unallocated_size_get(chunk, + run_ind+run_pages+nrun_pages-1) == nrun_size); + assert(arena_mapbits_dirty_get(chunk, + run_ind+run_pages+nrun_pages-1) == flag_dirty); + arena_avail_remove(arena, chunk, run_ind+run_pages, nrun_pages, + false, true); size += nrun_size; run_pages += nrun_pages; - chunk->map[run_ind-map_bias].bits = size | - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_FLAGS_MASK); - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - (chunk->map[run_ind+run_pages-1-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); + arena_mapbits_unallocated_size_set(chunk, run_ind, size); + arena_mapbits_unallocated_size_set(chunk, run_ind+run_pages-1, + size); } /* Try to coalesce backward. */ - if (run_ind > map_bias && (chunk->map[run_ind-1-map_bias].bits & - CHUNK_MAP_ALLOCATED) == 0 && (chunk->map[run_ind-1-map_bias].bits & - CHUNK_MAP_DIRTY) == flag_dirty) { - size_t prun_size = chunk->map[run_ind-1-map_bias].bits & - ~PAGE_MASK; - size_t prun_pages = prun_size >> PAGE_SHIFT; + if (run_ind > map_bias && arena_mapbits_allocated_get(chunk, run_ind-1) + == 0 && arena_mapbits_dirty_get(chunk, run_ind-1) == flag_dirty) { + size_t prun_size = arena_mapbits_unallocated_size_get(chunk, + run_ind-1); + size_t prun_pages = prun_size >> LG_PAGE; run_ind -= prun_pages; @@ -1023,52 +1062,33 @@ arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty) * Remove predecessor from runs_avail; the coalesced run is * inserted later. */ - assert((chunk->map[run_ind-map_bias].bits & ~PAGE_MASK) - == prun_size); - assert((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_ALLOCATED) - == 0); - assert((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) - == flag_dirty); - arena_avail_tree_remove(runs_avail, - &chunk->map[run_ind-map_bias]); + assert(arena_mapbits_unallocated_size_get(chunk, run_ind) == + prun_size); + assert(arena_mapbits_dirty_get(chunk, run_ind) == flag_dirty); + arena_avail_remove(arena, chunk, run_ind, prun_pages, true, + false); size += prun_size; run_pages += prun_pages; - chunk->map[run_ind-map_bias].bits = size | - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_FLAGS_MASK); - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - (chunk->map[run_ind+run_pages-1-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); + arena_mapbits_unallocated_size_set(chunk, run_ind, size); + arena_mapbits_unallocated_size_set(chunk, run_ind+run_pages-1, + size); } /* Insert into runs_avail, now that coalescing is complete. */ - assert((chunk->map[run_ind-map_bias].bits & ~PAGE_MASK) == - (chunk->map[run_ind+run_pages-1-map_bias].bits & ~PAGE_MASK)); - assert((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) == - (chunk->map[run_ind+run_pages-1-map_bias].bits & CHUNK_MAP_DIRTY)); - arena_avail_tree_insert(runs_avail, &chunk->map[run_ind-map_bias]); - - if (dirty) { - /* - * Insert into chunks_dirty before potentially calling - * arena_chunk_dealloc(), so that chunks_dirty and - * arena->ndirty are consistent. - */ - if (chunk->dirtied == false) { - ql_tail_insert(&arena->chunks_dirty, chunk, link_dirty); - chunk->dirtied = true; - } - } - - /* - * Deallocate chunk if it is now completely unused. The bit - * manipulation checks whether the first run is unallocated and extends - * to the end of the chunk. - */ - if ((chunk->map[0].bits & (~PAGE_MASK | CHUNK_MAP_ALLOCATED)) == - arena_maxclass) + assert(arena_mapbits_unallocated_size_get(chunk, run_ind) == + arena_mapbits_unallocated_size_get(chunk, run_ind+run_pages-1)); + assert(arena_mapbits_dirty_get(chunk, run_ind) == + arena_mapbits_dirty_get(chunk, run_ind+run_pages-1)); + arena_avail_insert(arena, chunk, run_ind, run_pages, true, true); + + /* Deallocate chunk if it is now completely unused. */ + if (size == arena_maxclass) { + assert(run_ind == map_bias); + assert(run_pages == (arena_maxclass >> LG_PAGE)); arena_chunk_dealloc(arena, chunk); + } /* * It is okay to do dirty page processing here even if the chunk was @@ -1085,9 +1105,9 @@ static void arena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize) { - size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> PAGE_SHIFT; - size_t head_npages = (oldsize - newsize) >> PAGE_SHIFT; - size_t flag_dirty = chunk->map[pageind-map_bias].bits & CHUNK_MAP_DIRTY; + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + size_t head_npages = (oldsize - newsize) >> LG_PAGE; + size_t flag_dirty = arena_mapbits_dirty_get(chunk, pageind); assert(oldsize > newsize); @@ -1096,44 +1116,30 @@ arena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, * leading run as separately allocated. Set the last element of each * run first, in case of single-page runs. */ - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_ALLOCATED) != 0); - chunk->map[pageind+head_npages-1-map_bias].bits = flag_dirty | - (chunk->map[pageind+head_npages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind-map_bias].bits = (oldsize - newsize) - | flag_dirty | (chunk->map[pageind-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - -#ifdef JEMALLOC_DEBUG - { - size_t tail_npages = newsize >> PAGE_SHIFT; - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & ~PAGE_MASK) == 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & CHUNK_MAP_DIRTY) == flag_dirty); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & CHUNK_MAP_ALLOCATED) != 0); - } -#endif - chunk->map[pageind+head_npages-map_bias].bits = newsize | flag_dirty | - (chunk->map[pageind+head_npages-map_bias].bits & - CHUNK_MAP_FLAGS_MASK) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - - arena_run_dalloc(arena, run, false); + assert(arena_mapbits_large_size_get(chunk, pageind) == oldsize); + arena_mapbits_large_set(chunk, pageind+head_npages-1, 0, flag_dirty); + arena_mapbits_large_set(chunk, pageind, oldsize-newsize, flag_dirty); + + if (config_debug) { + UNUSED size_t tail_npages = newsize >> LG_PAGE; + assert(arena_mapbits_large_size_get(chunk, + pageind+head_npages+tail_npages-1) == 0); + assert(arena_mapbits_dirty_get(chunk, + pageind+head_npages+tail_npages-1) == flag_dirty); + } + arena_mapbits_large_set(chunk, pageind+head_npages, newsize, + flag_dirty); + + arena_run_dalloc(arena, run, false, false); } static void arena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize, bool dirty) { - size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> PAGE_SHIFT; - size_t head_npages = newsize >> PAGE_SHIFT; - size_t tail_npages = (oldsize - newsize) >> PAGE_SHIFT; - size_t flag_dirty = chunk->map[pageind-map_bias].bits & - CHUNK_MAP_DIRTY; + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + size_t head_npages = newsize >> LG_PAGE; + size_t flag_dirty = arena_mapbits_dirty_get(chunk, pageind); assert(oldsize > newsize); @@ -1142,87 +1148,120 @@ arena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, * trailing run as separately allocated. Set the last element of each * run first, in case of single-page runs. */ - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_ALLOCATED) != 0); - chunk->map[pageind+head_npages-1-map_bias].bits = flag_dirty | - (chunk->map[pageind+head_npages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind-map_bias].bits = newsize | flag_dirty | - (chunk->map[pageind-map_bias].bits & CHUNK_MAP_UNZEROED) | - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - ~PAGE_MASK) == 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - CHUNK_MAP_ALLOCATED) != 0); - chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits = - flag_dirty | - (chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind+head_npages-map_bias].bits = (oldsize - newsize) | - flag_dirty | (chunk->map[pageind+head_npages-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + assert(arena_mapbits_large_size_get(chunk, pageind) == oldsize); + arena_mapbits_large_set(chunk, pageind+head_npages-1, 0, flag_dirty); + arena_mapbits_large_set(chunk, pageind, newsize, flag_dirty); + + if (config_debug) { + UNUSED size_t tail_npages = (oldsize - newsize) >> LG_PAGE; + assert(arena_mapbits_large_size_get(chunk, + pageind+head_npages+tail_npages-1) == 0); + assert(arena_mapbits_dirty_get(chunk, + pageind+head_npages+tail_npages-1) == flag_dirty); + } + arena_mapbits_large_set(chunk, pageind+head_npages, oldsize-newsize, + flag_dirty); arena_run_dalloc(arena, (arena_run_t *)((uintptr_t)run + newsize), - dirty); + dirty, false); } static arena_run_t * -arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) +arena_bin_runs_first(arena_bin_t *bin) { - arena_chunk_map_t *mapelm; - arena_run_t *run; - - /* Look for a usable run. */ - mapelm = arena_run_tree_first(&bin->runs); + arena_chunk_map_t *mapelm = arena_run_tree_first(&bin->runs); if (mapelm != NULL) { arena_chunk_t *chunk; size_t pageind; - - /* run is guaranteed to have available space. */ - arena_run_tree_remove(&bin->runs, mapelm); + arena_run_t *run; chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(mapelm); pageind = ((((uintptr_t)mapelm - (uintptr_t)chunk->map) / sizeof(arena_chunk_map_t))) + map_bias; run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) - << PAGE_SHIFT)); -#ifdef JEMALLOC_STATS - bin->stats.reruns++; -#endif + arena_mapbits_small_runind_get(chunk, pageind)) << + LG_PAGE)); return (run); } + + return (NULL); +} + +static void +arena_bin_runs_insert(arena_bin_t *bin, arena_run_t *run) +{ + arena_chunk_t *chunk = CHUNK_ADDR2BASE(run); + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + arena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind); + + assert(arena_run_tree_search(&bin->runs, mapelm) == NULL); + + arena_run_tree_insert(&bin->runs, mapelm); +} + +static void +arena_bin_runs_remove(arena_bin_t *bin, arena_run_t *run) +{ + arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + arena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind); + + assert(arena_run_tree_search(&bin->runs, mapelm) != NULL); + + arena_run_tree_remove(&bin->runs, mapelm); +} + +static arena_run_t * +arena_bin_nonfull_run_tryget(arena_bin_t *bin) +{ + arena_run_t *run = arena_bin_runs_first(bin); + if (run != NULL) { + arena_bin_runs_remove(bin, run); + if (config_stats) + bin->stats.reruns++; + } + return (run); +} + +static arena_run_t * +arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) +{ + arena_run_t *run; + size_t binind; + arena_bin_info_t *bin_info; + + /* Look for a usable run. */ + run = arena_bin_nonfull_run_tryget(bin); + if (run != NULL) + return (run); /* No existing runs have any space available. */ + binind = arena_bin_index(arena, bin); + bin_info = &arena_bin_info[binind]; + /* Allocate a new run. */ malloc_mutex_unlock(&bin->lock); /******************************/ malloc_mutex_lock(&arena->lock); - run = arena_run_alloc(arena, bin->run_size, false, false); + run = arena_run_alloc(arena, bin_info->run_size, false, binind, false); if (run != NULL) { + bitmap_t *bitmap = (bitmap_t *)((uintptr_t)run + + (uintptr_t)bin_info->bitmap_offset); + /* Initialize run internals. */ run->bin = bin; - run->avail = NULL; - run->next = (void *)((uintptr_t)run + - (uintptr_t)bin->reg0_offset); - run->nfree = bin->nregs; -#ifdef JEMALLOC_DEBUG - run->magic = ARENA_RUN_MAGIC; -#endif + run->nextind = 0; + run->nfree = bin_info->nregs; + bitmap_init(bitmap, &bin_info->bitmap_info); } malloc_mutex_unlock(&arena->lock); /********************************/ malloc_mutex_lock(&bin->lock); if (run != NULL) { -#ifdef JEMALLOC_STATS - bin->stats.nruns++; - bin->stats.curruns++; - if (bin->stats.curruns > bin->stats.highruns) - bin->stats.highruns = bin->stats.curruns; -#endif + if (config_stats) { + bin->stats.nruns++; + bin->stats.curruns++; + } return (run); } @@ -1231,25 +1270,9 @@ arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) * sufficient memory available while this one dropped bin->lock above, * so search one more time. */ - mapelm = arena_run_tree_first(&bin->runs); - if (mapelm != NULL) { - arena_chunk_t *chunk; - size_t pageind; - - /* run is guaranteed to have available space. */ - arena_run_tree_remove(&bin->runs, mapelm); - - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(mapelm); - pageind = ((((uintptr_t)mapelm - (uintptr_t)chunk->map) / - sizeof(arena_chunk_map_t))) + map_bias; - run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) - << PAGE_SHIFT)); -#ifdef JEMALLOC_STATS - bin->stats.reruns++; -#endif + run = arena_bin_nonfull_run_tryget(bin); + if (run != NULL) return (run); - } return (NULL); } @@ -1259,8 +1282,12 @@ static void * arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin) { void *ret; + size_t binind; + arena_bin_info_t *bin_info; arena_run_t *run; + binind = arena_bin_index(arena, bin); + bin_info = &arena_bin_info[binind]; bin->runcur = NULL; run = arena_bin_nonfull_run_get(arena, bin); if (bin->runcur != NULL && bin->runcur->nfree > 0) { @@ -1268,22 +1295,21 @@ arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin) * Another thread updated runcur while this one ran without the * bin lock in arena_bin_nonfull_run_get(). */ - assert(bin->runcur->magic == ARENA_RUN_MAGIC); assert(bin->runcur->nfree > 0); - ret = arena_run_reg_alloc(bin->runcur, bin); + ret = arena_run_reg_alloc(bin->runcur, bin_info); if (run != NULL) { arena_chunk_t *chunk; /* * arena_run_alloc() may have allocated run, or it may - * have pulled it from the bin's run tree. Therefore + * have pulled run from the bin's run tree. Therefore * it is unsafe to make any assumptions about how run * has previously been used, and arena_bin_lower_run() * must be called, as if a region were just deallocated * from the run. */ chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - if (run->nfree == bin->nregs) + if (run->nfree == bin_info->nregs) arena_dalloc_bin_run(arena, chunk, run, bin); else arena_bin_lower_run(arena, chunk, run, bin); @@ -1296,34 +1322,14 @@ arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin) bin->runcur = run; - assert(bin->runcur->magic == ARENA_RUN_MAGIC); assert(bin->runcur->nfree > 0); - return (arena_run_reg_alloc(bin->runcur, bin)); -} - -#ifdef JEMALLOC_PROF -void -arena_prof_accum(arena_t *arena, uint64_t accumbytes) -{ - - if (prof_interval != 0) { - arena->prof_accumbytes += accumbytes; - if (arena->prof_accumbytes >= prof_interval) { - prof_idump(); - arena->prof_accumbytes -= prof_interval; - } - } + return (arena_run_reg_alloc(bin->runcur, bin_info)); } -#endif -#ifdef JEMALLOC_TCACHE void -arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind -# ifdef JEMALLOC_PROF - , uint64_t prof_accumbytes -# endif - ) +arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind, + uint64_t prof_accumbytes) { unsigned i, nfill; arena_bin_t *bin; @@ -1332,143 +1338,83 @@ arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind assert(tbin->ncached == 0); -#ifdef JEMALLOC_PROF - malloc_mutex_lock(&arena->lock); - arena_prof_accum(arena, prof_accumbytes); - malloc_mutex_unlock(&arena->lock); -#endif + if (config_prof && arena_prof_accum(arena, prof_accumbytes)) + prof_idump(); bin = &arena->bins[binind]; malloc_mutex_lock(&bin->lock); - for (i = 0, nfill = (tbin->ncached_max >> 1); i < nfill; i++) { + for (i = 0, nfill = (tcache_bin_info[binind].ncached_max >> + tbin->lg_fill_div); i < nfill; i++) { if ((run = bin->runcur) != NULL && run->nfree > 0) - ptr = arena_run_reg_alloc(run, bin); + ptr = arena_run_reg_alloc(run, &arena_bin_info[binind]); else ptr = arena_bin_malloc_hard(arena, bin); if (ptr == NULL) break; - *(void **)ptr = tbin->avail; - tbin->avail = ptr; - } -#ifdef JEMALLOC_STATS - bin->stats.allocated += (i - tbin->ncached) * bin->reg_size; - bin->stats.nmalloc += i; - bin->stats.nrequests += tbin->tstats.nrequests; - bin->stats.nfills++; - tbin->tstats.nrequests = 0; -#endif + if (config_fill && opt_junk) { + arena_alloc_junk_small(ptr, &arena_bin_info[binind], + true); + } + /* Insert such that low regions get used first. */ + tbin->avail[nfill - 1 - i] = ptr; + } + if (config_stats) { + bin->stats.allocated += i * arena_bin_info[binind].reg_size; + bin->stats.nmalloc += i; + bin->stats.nrequests += tbin->tstats.nrequests; + bin->stats.nfills++; + tbin->tstats.nrequests = 0; + } malloc_mutex_unlock(&bin->lock); tbin->ncached = i; - if (tbin->ncached > tbin->high_water) - tbin->high_water = tbin->ncached; } -#endif -/* - * Calculate bin->run_size such that it meets the following constraints: - * - * *) bin->run_size >= min_run_size - * *) bin->run_size <= arena_maxclass - * *) run header overhead <= RUN_MAX_OVRHD (or header overhead relaxed). - * *) run header size < PAGE_SIZE - * - * bin->nregs and bin->reg0_offset are also calculated here, since these - * settings are all interdependent. - */ -static size_t -arena_bin_run_size_calc(arena_bin_t *bin, size_t min_run_size) +void +arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info, bool zero) { - size_t try_run_size, good_run_size; - uint32_t try_nregs, good_nregs; - uint32_t try_hdr_size, good_hdr_size; -#ifdef JEMALLOC_PROF - uint32_t try_ctx0_offset, good_ctx0_offset; -#endif - uint32_t try_reg0_offset, good_reg0_offset; - - assert(min_run_size >= PAGE_SIZE); - assert(min_run_size <= arena_maxclass); - - /* - * Calculate known-valid settings before entering the run_size - * expansion loop, so that the first part of the loop always copies - * valid settings. - * - * The do..while loop iteratively reduces the number of regions until - * the run header and the regions no longer overlap. A closed formula - * would be quite messy, since there is an interdependency between the - * header's mask length and the number of regions. - */ - try_run_size = min_run_size; - try_nregs = ((try_run_size - sizeof(arena_run_t)) / bin->reg_size) - + 1; /* Counter-act try_nregs-- in loop. */ - do { - try_nregs--; - try_hdr_size = sizeof(arena_run_t); -#ifdef JEMALLOC_PROF - if (opt_prof && prof_promote == false) { - /* Pad to a quantum boundary. */ - try_hdr_size = QUANTUM_CEILING(try_hdr_size); - try_ctx0_offset = try_hdr_size; - /* Add space for one (prof_ctx_t *) per region. */ - try_hdr_size += try_nregs * sizeof(prof_ctx_t *); - } else - try_ctx0_offset = 0; -#endif - try_reg0_offset = try_run_size - (try_nregs * bin->reg_size); - } while (try_hdr_size > try_reg0_offset); - /* run_size expansion loop. */ - do { - /* - * Copy valid settings before trying more aggressive settings. - */ - good_run_size = try_run_size; - good_nregs = try_nregs; - good_hdr_size = try_hdr_size; -#ifdef JEMALLOC_PROF - good_ctx0_offset = try_ctx0_offset; -#endif - good_reg0_offset = try_reg0_offset; - - /* Try more aggressive settings. */ - try_run_size += PAGE_SIZE; - try_nregs = ((try_run_size - sizeof(arena_run_t)) / - bin->reg_size) + 1; /* Counter-act try_nregs-- in loop. */ - do { - try_nregs--; - try_hdr_size = sizeof(arena_run_t); -#ifdef JEMALLOC_PROF - if (opt_prof && prof_promote == false) { - /* Pad to a quantum boundary. */ - try_hdr_size = QUANTUM_CEILING(try_hdr_size); - try_ctx0_offset = try_hdr_size; - /* - * Add space for one (prof_ctx_t *) per region. - */ - try_hdr_size += try_nregs * - sizeof(prof_ctx_t *); - } -#endif - try_reg0_offset = try_run_size - (try_nregs * - bin->reg_size); - } while (try_hdr_size > try_reg0_offset); - } while (try_run_size <= arena_maxclass - && try_run_size <= arena_maxclass - && RUN_MAX_OVRHD * (bin->reg_size << 3) > RUN_MAX_OVRHD_RELAX - && (try_reg0_offset << RUN_BFP) > RUN_MAX_OVRHD * try_run_size - && try_hdr_size < PAGE_SIZE); - - assert(good_hdr_size <= good_reg0_offset); + if (zero) { + size_t redzone_size = bin_info->redzone_size; + memset((void *)((uintptr_t)ptr - redzone_size), 0xa5, + redzone_size); + memset((void *)((uintptr_t)ptr + bin_info->reg_size), 0xa5, + redzone_size); + } else { + memset((void *)((uintptr_t)ptr - bin_info->redzone_size), 0xa5, + bin_info->reg_interval); + } +} - /* Copy final settings. */ - bin->run_size = good_run_size; - bin->nregs = good_nregs; -#ifdef JEMALLOC_PROF - bin->ctx0_offset = good_ctx0_offset; -#endif - bin->reg0_offset = good_reg0_offset; +void +arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info) +{ + size_t size = bin_info->reg_size; + size_t redzone_size = bin_info->redzone_size; + size_t i; + bool error = false; + + for (i = 1; i <= redzone_size; i++) { + unsigned byte; + if ((byte = *(uint8_t *)((uintptr_t)ptr - i)) != 0xa5) { + error = true; + malloc_printf("<jemalloc>: Corrupt redzone " + "%zu byte%s before %p (size %zu), byte=%#x\n", i, + (i == 1) ? "" : "s", ptr, size, byte); + } + } + for (i = 0; i < redzone_size; i++) { + unsigned byte; + if ((byte = *(uint8_t *)((uintptr_t)ptr + size + i)) != 0xa5) { + error = true; + malloc_printf("<jemalloc>: Corrupt redzone " + "%zu byte%s after end of %p (size %zu), byte=%#x\n", + i, (i == 1) ? "" : "s", ptr, size, byte); + } + } + if (opt_abort && error) + abort(); - return (good_run_size); + memset((void *)((uintptr_t)ptr - redzone_size), 0x5a, + bin_info->reg_interval); } void * @@ -1479,14 +1425,14 @@ arena_malloc_small(arena_t *arena, size_t size, bool zero) arena_run_t *run; size_t binind; - binind = small_size2bin[size]; - assert(binind < nbins); + binind = SMALL_SIZE2BIN(size); + assert(binind < NBINS); bin = &arena->bins[binind]; - size = bin->reg_size; + size = arena_bin_info[binind].reg_size; malloc_mutex_lock(&bin->lock); if ((run = bin->runcur) != NULL && run->nfree > 0) - ret = arena_run_reg_alloc(run, bin); + ret = arena_run_reg_alloc(run, &arena_bin_info[binind]); else ret = arena_bin_malloc_hard(arena, bin); @@ -1495,29 +1441,32 @@ arena_malloc_small(arena_t *arena, size_t size, bool zero) return (NULL); } -#ifdef JEMALLOC_STATS - bin->stats.allocated += size; - bin->stats.nmalloc++; - bin->stats.nrequests++; -#endif - malloc_mutex_unlock(&bin->lock); -#ifdef JEMALLOC_PROF - if (isthreaded == false) { - malloc_mutex_lock(&arena->lock); - arena_prof_accum(arena, size); - malloc_mutex_unlock(&arena->lock); + if (config_stats) { + bin->stats.allocated += size; + bin->stats.nmalloc++; + bin->stats.nrequests++; } -#endif + malloc_mutex_unlock(&bin->lock); + if (config_prof && isthreaded == false && arena_prof_accum(arena, size)) + prof_idump(); if (zero == false) { -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ret, 0xa5, size); - else if (opt_zero) - memset(ret, 0, size); -#endif - } else + if (config_fill) { + if (opt_junk) { + arena_alloc_junk_small(ret, + &arena_bin_info[binind], false); + } else if (opt_zero) + memset(ret, 0, size); + } + } else { + if (config_fill && opt_junk) { + arena_alloc_junk_small(ret, &arena_bin_info[binind], + true); + } + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); memset(ret, 0, size); + } + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); return (ret); } @@ -1526,242 +1475,119 @@ void * arena_malloc_large(arena_t *arena, size_t size, bool zero) { void *ret; + UNUSED bool idump; /* Large allocation. */ size = PAGE_CEILING(size); malloc_mutex_lock(&arena->lock); - ret = (void *)arena_run_alloc(arena, size, true, zero); + ret = (void *)arena_run_alloc(arena, size, true, BININD_INVALID, zero); if (ret == NULL) { malloc_mutex_unlock(&arena->lock); return (NULL); } -#ifdef JEMALLOC_STATS - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns; - } -#endif -#ifdef JEMALLOC_PROF - arena_prof_accum(arena, size); -#endif + if (config_stats) { + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; + } + if (config_prof) + idump = arena_prof_accum_locked(arena, size); malloc_mutex_unlock(&arena->lock); + if (config_prof && idump) + prof_idump(); if (zero == false) { -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ret, 0xa5, size); - else if (opt_zero) - memset(ret, 0, size); -#endif + if (config_fill) { + if (opt_junk) + memset(ret, 0xa5, size); + else if (opt_zero) + memset(ret, 0, size); + } } return (ret); } -void * -arena_malloc(size_t size, bool zero) -{ - - assert(size != 0); - assert(QUANTUM_CEILING(size) <= arena_maxclass); - - if (size <= small_maxclass) { -#ifdef JEMALLOC_TCACHE - tcache_t *tcache; - - if ((tcache = tcache_get()) != NULL) - return (tcache_alloc_small(tcache, size, zero)); - else - -#endif - return (arena_malloc_small(choose_arena(), size, zero)); - } else { -#ifdef JEMALLOC_TCACHE - if (size <= tcache_maxclass) { - tcache_t *tcache; - - if ((tcache = tcache_get()) != NULL) - return (tcache_alloc_large(tcache, size, zero)); - else { - return (arena_malloc_large(choose_arena(), - size, zero)); - } - } else -#endif - return (arena_malloc_large(choose_arena(), size, zero)); - } -} - /* Only handles large allocations that require more than page alignment. */ void * -arena_palloc(arena_t *arena, size_t size, size_t alloc_size, size_t alignment, - bool zero) +arena_palloc(arena_t *arena, size_t size, size_t alignment, bool zero) { void *ret; - size_t offset; + size_t alloc_size, leadsize, trailsize; + arena_run_t *run; arena_chunk_t *chunk; assert((size & PAGE_MASK) == 0); alignment = PAGE_CEILING(alignment); + alloc_size = size + alignment - PAGE; malloc_mutex_lock(&arena->lock); - ret = (void *)arena_run_alloc(arena, alloc_size, true, zero); - if (ret == NULL) { + run = arena_run_alloc(arena, alloc_size, true, BININD_INVALID, zero); + if (run == NULL) { malloc_mutex_unlock(&arena->lock); return (NULL); } + chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ret); - - offset = (uintptr_t)ret & (alignment - 1); - assert((offset & PAGE_MASK) == 0); - assert(offset < alloc_size); - if (offset == 0) - arena_run_trim_tail(arena, chunk, ret, alloc_size, size, false); - else { - size_t leadsize, trailsize; - - leadsize = alignment - offset; - if (leadsize > 0) { - arena_run_trim_head(arena, chunk, ret, alloc_size, - alloc_size - leadsize); - ret = (void *)((uintptr_t)ret + leadsize); - } - - trailsize = alloc_size - leadsize - size; - if (trailsize != 0) { - /* Trim trailing space. */ - assert(trailsize < alloc_size); - arena_run_trim_tail(arena, chunk, ret, size + trailsize, - size, false); - } + leadsize = ALIGNMENT_CEILING((uintptr_t)run, alignment) - + (uintptr_t)run; + assert(alloc_size >= leadsize + size); + trailsize = alloc_size - leadsize - size; + ret = (void *)((uintptr_t)run + leadsize); + if (leadsize != 0) { + arena_run_trim_head(arena, chunk, run, alloc_size, alloc_size - + leadsize); + } + if (trailsize != 0) { + arena_run_trim_tail(arena, chunk, ret, size + trailsize, size, + false); } -#ifdef JEMALLOC_STATS - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns; + if (config_stats) { + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; } -#endif malloc_mutex_unlock(&arena->lock); -#ifdef JEMALLOC_FILL - if (zero == false) { + if (config_fill && zero == false) { if (opt_junk) memset(ret, 0xa5, size); else if (opt_zero) memset(ret, 0, size); } -#endif return (ret); } -/* Return the size of the allocation pointed to by ptr. */ -size_t -arena_salloc(const void *ptr) -{ - size_t ret; - arena_chunk_t *chunk; - size_t pageind, mapbits; - - assert(ptr != NULL); - assert(CHUNK_ADDR2BASE(ptr) != ptr); - - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapbits = chunk->map[pageind-map_bias].bits; - assert((mapbits & CHUNK_MAP_ALLOCATED) != 0); - if ((mapbits & CHUNK_MAP_LARGE) == 0) { - arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapbits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - assert(run->magic == ARENA_RUN_MAGIC); - assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)run->bin->reg0_offset)) % run->bin->reg_size == - 0); - ret = run->bin->reg_size; - } else { - assert(((uintptr_t)ptr & PAGE_MASK) == 0); - ret = mapbits & ~PAGE_MASK; - assert(ret != 0); - } - - return (ret); -} - -#ifdef JEMALLOC_PROF void arena_prof_promoted(const void *ptr, size_t size) { arena_chunk_t *chunk; size_t pageind, binind; + cassert(config_prof); assert(ptr != NULL); assert(CHUNK_ADDR2BASE(ptr) != ptr); - assert(isalloc(ptr) == PAGE_SIZE); + assert(isalloc(ptr, false) == PAGE); + assert(isalloc(ptr, true) == PAGE); + assert(size <= SMALL_MAXCLASS); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - binind = small_size2bin[size]; - assert(binind < nbins); - chunk->map[pageind-map_bias].bits = (chunk->map[pageind-map_bias].bits & - ~CHUNK_MAP_CLASS_MASK) | ((binind+1) << CHUNK_MAP_CLASS_SHIFT); -} + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + binind = SMALL_SIZE2BIN(size); + assert(binind < NBINS); + arena_mapbits_large_binind_set(chunk, pageind, binind); -size_t -arena_salloc_demote(const void *ptr) -{ - size_t ret; - arena_chunk_t *chunk; - size_t pageind, mapbits; - - assert(ptr != NULL); - assert(CHUNK_ADDR2BASE(ptr) != ptr); - - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapbits = chunk->map[pageind-map_bias].bits; - assert((mapbits & CHUNK_MAP_ALLOCATED) != 0); - if ((mapbits & CHUNK_MAP_LARGE) == 0) { - arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapbits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - assert(run->magic == ARENA_RUN_MAGIC); - assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)run->bin->reg0_offset)) % run->bin->reg_size == - 0); - ret = run->bin->reg_size; - } else { - assert(((uintptr_t)ptr & PAGE_MASK) == 0); - ret = mapbits & ~PAGE_MASK; - if (prof_promote && ret == PAGE_SIZE && (mapbits & - CHUNK_MAP_CLASS_MASK) != 0) { - size_t binind = ((mapbits & CHUNK_MAP_CLASS_MASK) >> - CHUNK_MAP_CLASS_SHIFT) - 1; - assert(binind < nbins); - ret = chunk->arena->bins[binind].reg_size; - } - assert(ret != 0); - } - - return (ret); + assert(isalloc(ptr, false) == PAGE); + assert(isalloc(ptr, true) == size); } -#endif static void arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run, @@ -1771,17 +1597,18 @@ arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run, /* Dissociate run from bin. */ if (run == bin->runcur) bin->runcur = NULL; - else if (bin->nregs != 1) { - size_t run_pageind = (((uintptr_t)run - (uintptr_t)chunk)) >> - PAGE_SHIFT; - arena_chunk_map_t *run_mapelm = - &chunk->map[run_pageind-map_bias]; - /* - * This block's conditional is necessary because if the run - * only contains one region, then it never gets inserted into - * the non-full runs tree. - */ - arena_run_tree_remove(&bin->runs, run_mapelm); + else { + size_t binind = arena_bin_index(chunk->arena, bin); + arena_bin_info_t *bin_info = &arena_bin_info[binind]; + + if (bin_info->nregs != 1) { + /* + * This block's conditional is necessary because if the + * run only contains one region, then it never gets + * inserted into the non-full runs tree. + */ + arena_bin_runs_remove(bin, run); + } } } @@ -1789,18 +1616,26 @@ static void arena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, arena_bin_t *bin) { + size_t binind; + arena_bin_info_t *bin_info; size_t npages, run_ind, past; assert(run != bin->runcur); - assert(arena_run_tree_search(&bin->runs, &chunk->map[ - (((uintptr_t)run-(uintptr_t)chunk)>>PAGE_SHIFT)-map_bias]) == NULL); + assert(arena_run_tree_search(&bin->runs, + arena_mapp_get(chunk, ((uintptr_t)run-(uintptr_t)chunk)>>LG_PAGE)) + == NULL); + + binind = arena_bin_index(chunk->arena, run->bin); + bin_info = &arena_bin_info[binind]; malloc_mutex_unlock(&bin->lock); /******************************/ - npages = bin->run_size >> PAGE_SHIFT; - run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> PAGE_SHIFT); - past = (size_t)((PAGE_CEILING((uintptr_t)run->next) - (uintptr_t)chunk) - >> PAGE_SHIFT); + npages = bin_info->run_size >> LG_PAGE; + run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE); + past = (size_t)(PAGE_CEILING((uintptr_t)run + + (uintptr_t)bin_info->reg0_offset + (uintptr_t)(run->nextind * + bin_info->reg_interval - bin_info->redzone_size) - + (uintptr_t)chunk) >> LG_PAGE); malloc_mutex_lock(&arena->lock); /* @@ -1808,32 +1643,24 @@ arena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, * trim the clean pages before deallocating the dirty portion of the * run. */ - if ((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) == 0 && past - - run_ind < npages) { - /* - * Trim clean pages. Convert to large run beforehand. Set the - * last map element first, in case this is a one-page run. - */ - chunk->map[run_ind+npages-1-map_bias].bits = CHUNK_MAP_LARGE | - (chunk->map[run_ind+npages-1-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); - chunk->map[run_ind-map_bias].bits = bin->run_size | - CHUNK_MAP_LARGE | (chunk->map[run_ind-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); - arena_run_trim_tail(arena, chunk, run, (npages << PAGE_SHIFT), - ((past - run_ind) << PAGE_SHIFT), false); + assert(arena_mapbits_dirty_get(chunk, run_ind) == + arena_mapbits_dirty_get(chunk, run_ind+npages-1)); + if (arena_mapbits_dirty_get(chunk, run_ind) == 0 && past - run_ind < + npages) { + /* Trim clean pages. Convert to large run beforehand. */ + assert(npages > 0); + arena_mapbits_large_set(chunk, run_ind, bin_info->run_size, 0); + arena_mapbits_large_set(chunk, run_ind+npages-1, 0, 0); + arena_run_trim_tail(arena, chunk, run, (npages << LG_PAGE), + ((past - run_ind) << LG_PAGE), false); /* npages = past - run_ind; */ } -#ifdef JEMALLOC_DEBUG - run->magic = 0; -#endif - arena_run_dalloc(arena, run, true); + arena_run_dalloc(arena, run, true, false); malloc_mutex_unlock(&arena->lock); /****************************/ malloc_mutex_lock(&bin->lock); -#ifdef JEMALLOC_STATS - bin->stats.curruns--; -#endif + if (config_stats) + bin->stats.curruns--; } static void @@ -1842,157 +1669,114 @@ arena_bin_lower_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, { /* - * Make sure that bin->runcur always refers to the lowest non-full run, - * if one exists. + * Make sure that if bin->runcur is non-NULL, it refers to the lowest + * non-full run. It is okay to NULL runcur out rather than proactively + * keeping it pointing at the lowest non-full run. */ - if (bin->runcur == NULL) - bin->runcur = run; - else if ((uintptr_t)run < (uintptr_t)bin->runcur) { + if ((uintptr_t)run < (uintptr_t)bin->runcur) { /* Switch runcur. */ - if (bin->runcur->nfree > 0) { - arena_chunk_t *runcur_chunk = - CHUNK_ADDR2BASE(bin->runcur); - size_t runcur_pageind = (((uintptr_t)bin->runcur - - (uintptr_t)runcur_chunk)) >> PAGE_SHIFT; - arena_chunk_map_t *runcur_mapelm = - &runcur_chunk->map[runcur_pageind-map_bias]; - - /* Insert runcur. */ - arena_run_tree_insert(&bin->runs, runcur_mapelm); - } + if (bin->runcur->nfree > 0) + arena_bin_runs_insert(bin, bin->runcur); bin->runcur = run; - } else { - size_t run_pageind = (((uintptr_t)run - - (uintptr_t)chunk)) >> PAGE_SHIFT; - arena_chunk_map_t *run_mapelm = - &chunk->map[run_pageind-map_bias]; - - assert(arena_run_tree_search(&bin->runs, run_mapelm) == NULL); - arena_run_tree_insert(&bin->runs, run_mapelm); - } + if (config_stats) + bin->stats.reruns++; + } else + arena_bin_runs_insert(bin, run); } void -arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, +arena_dalloc_bin_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr, arena_chunk_map_t *mapelm) { size_t pageind; arena_run_t *run; arena_bin_t *bin; -#if (defined(JEMALLOC_FILL) || defined(JEMALLOC_STATS)) - size_t size; -#endif + arena_bin_info_t *bin_info; + size_t size, binind; - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) << PAGE_SHIFT)); - assert(run->magic == ARENA_RUN_MAGIC); + arena_mapbits_small_runind_get(chunk, pageind)) << LG_PAGE)); bin = run->bin; -#if (defined(JEMALLOC_FILL) || defined(JEMALLOC_STATS)) - size = bin->reg_size; -#endif + binind = arena_ptr_small_binind_get(ptr, mapelm->bits); + bin_info = &arena_bin_info[binind]; + if (config_fill || config_stats) + size = bin_info->reg_size; -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ptr, 0x5a, size); -#endif + if (config_fill && opt_junk) + arena_dalloc_junk_small(ptr, bin_info); arena_run_reg_dalloc(run, ptr); - if (run->nfree == bin->nregs) { + if (run->nfree == bin_info->nregs) { arena_dissociate_bin_run(chunk, run, bin); arena_dalloc_bin_run(arena, chunk, run, bin); } else if (run->nfree == 1 && run != bin->runcur) arena_bin_lower_run(arena, chunk, run, bin); -#ifdef JEMALLOC_STATS - bin->stats.allocated -= size; - bin->stats.ndalloc++; -#endif + if (config_stats) { + bin->stats.allocated -= size; + bin->stats.ndalloc++; + } } -#ifdef JEMALLOC_STATS void -arena_stats_merge(arena_t *arena, size_t *nactive, size_t *ndirty, - arena_stats_t *astats, malloc_bin_stats_t *bstats, - malloc_large_stats_t *lstats) +arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, + size_t pageind, arena_chunk_map_t *mapelm) { - unsigned i; + arena_run_t *run; + arena_bin_t *bin; - malloc_mutex_lock(&arena->lock); - *nactive += arena->nactive; - *ndirty += arena->ndirty; + run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - + arena_mapbits_small_runind_get(chunk, pageind)) << LG_PAGE)); + bin = run->bin; + malloc_mutex_lock(&bin->lock); + arena_dalloc_bin_locked(arena, chunk, ptr, mapelm); + malloc_mutex_unlock(&bin->lock); +} - astats->mapped += arena->stats.mapped; - astats->npurge += arena->stats.npurge; - astats->nmadvise += arena->stats.nmadvise; - astats->purged += arena->stats.purged; - astats->allocated_large += arena->stats.allocated_large; - astats->nmalloc_large += arena->stats.nmalloc_large; - astats->ndalloc_large += arena->stats.ndalloc_large; - astats->nrequests_large += arena->stats.nrequests_large; +void +arena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr, + size_t pageind) +{ + arena_chunk_map_t *mapelm; - for (i = 0; i < nlclasses; i++) { - lstats[i].nmalloc += arena->stats.lstats[i].nmalloc; - lstats[i].ndalloc += arena->stats.lstats[i].ndalloc; - lstats[i].nrequests += arena->stats.lstats[i].nrequests; - lstats[i].highruns += arena->stats.lstats[i].highruns; - lstats[i].curruns += arena->stats.lstats[i].curruns; + if (config_debug) { + /* arena_ptr_small_binind_get() does extra sanity checking. */ + assert(arena_ptr_small_binind_get(ptr, arena_mapbits_get(chunk, + pageind)) != BININD_INVALID); } - malloc_mutex_unlock(&arena->lock); + mapelm = arena_mapp_get(chunk, pageind); + arena_dalloc_bin(arena, chunk, ptr, pageind, mapelm); +} - for (i = 0; i < nbins; i++) { - arena_bin_t *bin = &arena->bins[i]; +void +arena_dalloc_large_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr) +{ - malloc_mutex_lock(&bin->lock); - bstats[i].allocated += bin->stats.allocated; - bstats[i].nmalloc += bin->stats.nmalloc; - bstats[i].ndalloc += bin->stats.ndalloc; - bstats[i].nrequests += bin->stats.nrequests; -#ifdef JEMALLOC_TCACHE - bstats[i].nfills += bin->stats.nfills; - bstats[i].nflushes += bin->stats.nflushes; -#endif - bstats[i].nruns += bin->stats.nruns; - bstats[i].reruns += bin->stats.reruns; - bstats[i].highruns += bin->stats.highruns; - bstats[i].curruns += bin->stats.curruns; - malloc_mutex_unlock(&bin->lock); + if (config_fill || config_stats) { + size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + size_t size = arena_mapbits_large_size_get(chunk, pageind); + + if (config_fill && config_stats && opt_junk) + memset(ptr, 0x5a, size); + if (config_stats) { + arena->stats.ndalloc_large++; + arena->stats.allocated_large -= size; + arena->stats.lstats[(size >> LG_PAGE) - 1].ndalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns--; + } } + + arena_run_dalloc(arena, (arena_run_t *)ptr, true, false); } -#endif void arena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr) { - /* Large allocation. */ -#ifdef JEMALLOC_FILL -# ifndef JEMALLOC_STATS - if (opt_junk) -# endif -#endif - { -#if (defined(JEMALLOC_FILL) || defined(JEMALLOC_STATS)) - size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> - PAGE_SHIFT; - size_t size = chunk->map[pageind-map_bias].bits & ~PAGE_MASK; -#endif - -#ifdef JEMALLOC_FILL -# ifdef JEMALLOC_STATS - if (opt_junk) -# endif - memset(ptr, 0x5a, size); -#endif -#ifdef JEMALLOC_STATS - arena->stats.ndalloc_large++; - arena->stats.allocated_large -= size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].ndalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns--; -#endif - } - - arena_run_dalloc(arena, (arena_run_t *)ptr, true); + malloc_mutex_lock(&arena->lock); + arena_dalloc_large_locked(arena, chunk, ptr); + malloc_mutex_unlock(&arena->lock); } static void @@ -2009,24 +1793,19 @@ arena_ralloc_large_shrink(arena_t *arena, arena_chunk_t *chunk, void *ptr, malloc_mutex_lock(&arena->lock); arena_run_trim_tail(arena, chunk, (arena_run_t *)ptr, oldsize, size, true); -#ifdef JEMALLOC_STATS - arena->stats.ndalloc_large++; - arena->stats.allocated_large -= oldsize; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].ndalloc++; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].curruns--; - - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns; - } -#endif + if (config_stats) { + arena->stats.ndalloc_large++; + arena->stats.allocated_large -= oldsize; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].ndalloc++; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].curruns--; + + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; + } malloc_mutex_unlock(&arena->lock); } @@ -2034,20 +1813,19 @@ static bool arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, size_t oldsize, size_t size, size_t extra, bool zero) { - size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - size_t npages = oldsize >> PAGE_SHIFT; + size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + size_t npages = oldsize >> LG_PAGE; size_t followsize; - assert(oldsize == (chunk->map[pageind-map_bias].bits & ~PAGE_MASK)); + assert(oldsize == arena_mapbits_large_size_get(chunk, pageind)); /* Try to extend the run. */ assert(size + extra > oldsize); malloc_mutex_lock(&arena->lock); if (pageind + npages < chunk_npages && - (chunk->map[pageind+npages-map_bias].bits - & CHUNK_MAP_ALLOCATED) == 0 && (followsize = - chunk->map[pageind+npages-map_bias].bits & ~PAGE_MASK) >= size - - oldsize) { + arena_mapbits_allocated_get(chunk, pageind+npages) == 0 && + (followsize = arena_mapbits_unallocated_size_get(chunk, + pageind+npages)) >= size - oldsize) { /* * The next run is available and sufficiently large. Split the * following run, then merge the first part with the existing @@ -2057,10 +1835,11 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, size_t splitsize = (oldsize + followsize <= size + extra) ? followsize : size + extra - oldsize; arena_run_split(arena, (arena_run_t *)((uintptr_t)chunk + - ((pageind+npages) << PAGE_SHIFT)), splitsize, true, zero); + ((pageind+npages) << LG_PAGE)), splitsize, true, + BININD_INVALID, zero); size = oldsize + splitsize; - npages = size >> PAGE_SHIFT; + npages = size >> LG_PAGE; /* * Mark the extended run as dirty if either portion of the run @@ -2070,34 +1849,24 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, * arena_run_dalloc() with the dirty argument set to false * (which is when dirty flag consistency would really matter). */ - flag_dirty = (chunk->map[pageind-map_bias].bits & - CHUNK_MAP_DIRTY) | - (chunk->map[pageind+npages-1-map_bias].bits & - CHUNK_MAP_DIRTY); - chunk->map[pageind-map_bias].bits = size | flag_dirty - | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind+npages-1-map_bias].bits = flag_dirty | - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - -#ifdef JEMALLOC_STATS - arena->stats.ndalloc_large++; - arena->stats.allocated_large -= oldsize; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].ndalloc++; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].curruns--; - - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - - 1].curruns; + flag_dirty = arena_mapbits_dirty_get(chunk, pageind) | + arena_mapbits_dirty_get(chunk, pageind+npages-1); + arena_mapbits_large_set(chunk, pageind, size, flag_dirty); + arena_mapbits_large_set(chunk, pageind+npages-1, 0, flag_dirty); + + if (config_stats) { + arena->stats.ndalloc_large++; + arena->stats.allocated_large -= oldsize; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].ndalloc++; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].curruns--; + + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; } -#endif malloc_mutex_unlock(&arena->lock); return (false); } @@ -2119,12 +1888,10 @@ arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, psize = PAGE_CEILING(size + extra); if (psize == oldsize) { /* Same size class. */ -#ifdef JEMALLOC_FILL - if (opt_junk && size < oldsize) { + if (config_fill && opt_junk && size < oldsize) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif return (false); } else { arena_chunk_t *chunk; @@ -2132,16 +1899,13 @@ arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); arena = chunk->arena; - assert(arena->magic == ARENA_MAGIC); if (psize < oldsize) { -#ifdef JEMALLOC_FILL /* Fill before shrinking in order avoid a race. */ - if (opt_junk) { + if (config_fill && opt_junk) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif arena_ralloc_large_shrink(arena, chunk, ptr, oldsize, psize); return (false); @@ -2149,12 +1913,11 @@ arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, bool ret = arena_ralloc_large_grow(arena, chunk, ptr, oldsize, PAGE_CEILING(size), psize - PAGE_CEILING(size), zero); -#ifdef JEMALLOC_FILL - if (ret == false && zero == false && opt_zero) { + if (config_fill && ret == false && zero == false && + opt_zero) { memset((void *)((uintptr_t)ptr + oldsize), 0, size - oldsize); } -#endif return (ret); } } @@ -2169,24 +1932,22 @@ arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra, * Avoid moving the allocation if the size class can be left the same. */ if (oldsize <= arena_maxclass) { - if (oldsize <= small_maxclass) { - assert(choose_arena()->bins[small_size2bin[ - oldsize]].reg_size == oldsize); - if ((size + extra <= small_maxclass && - small_size2bin[size + extra] == - small_size2bin[oldsize]) || (size <= oldsize && + if (oldsize <= SMALL_MAXCLASS) { + assert(arena_bin_info[SMALL_SIZE2BIN(oldsize)].reg_size + == oldsize); + if ((size + extra <= SMALL_MAXCLASS && + SMALL_SIZE2BIN(size + extra) == + SMALL_SIZE2BIN(oldsize)) || (size <= oldsize && size + extra >= oldsize)) { -#ifdef JEMALLOC_FILL - if (opt_junk && size < oldsize) { + if (config_fill && opt_junk && size < oldsize) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif return (ptr); } } else { assert(size <= arena_maxclass); - if (size + extra > small_maxclass) { + if (size + extra > SMALL_MAXCLASS) { if (arena_ralloc_large(ptr, oldsize, size, extra, zero) == false) return (ptr); @@ -2199,8 +1960,9 @@ arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra, } void * -arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, - size_t alignment, bool zero) +arena_ralloc(arena_t *arena, void *ptr, size_t oldsize, size_t size, + size_t extra, size_t alignment, bool zero, bool try_tcache_alloc, + bool try_tcache_dalloc) { void *ret; size_t copysize; @@ -2210,25 +1972,31 @@ arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, if (ret != NULL) return (ret); - /* * size and oldsize are different enough that we need to move the * object. In that case, fall back to allocating new space and * copying. */ - if (alignment != 0) - ret = ipalloc(size + extra, alignment, zero); - else - ret = arena_malloc(size + extra, zero); + if (alignment != 0) { + size_t usize = sa2u(size + extra, alignment); + if (usize == 0) + return (NULL); + ret = ipallocx(usize, alignment, zero, try_tcache_alloc, arena); + } else + ret = arena_malloc(arena, size + extra, zero, try_tcache_alloc); if (ret == NULL) { if (extra == 0) return (NULL); /* Try again, this time without extra. */ - if (alignment != 0) - ret = ipalloc(size, alignment, zero); - else - ret = arena_malloc(size, zero); + if (alignment != 0) { + size_t usize = sa2u(size, alignment); + if (usize == 0) + return (NULL); + ret = ipallocx(usize, alignment, zero, try_tcache_alloc, + arena); + } else + ret = arena_malloc(arena, size, zero, try_tcache_alloc); if (ret == NULL) return (NULL); @@ -2241,313 +2009,305 @@ arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, * expectation that the extra bytes will be reliably preserved. */ copysize = (size < oldsize) ? size : oldsize; + VALGRIND_MAKE_MEM_UNDEFINED(ret, copysize); memcpy(ret, ptr, copysize); - idalloc(ptr); + iqallocx(ptr, try_tcache_dalloc); + return (ret); +} + +dss_prec_t +arena_dss_prec_get(arena_t *arena) +{ + dss_prec_t ret; + + malloc_mutex_lock(&arena->lock); + ret = arena->dss_prec; + malloc_mutex_unlock(&arena->lock); return (ret); } +void +arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec) +{ + + malloc_mutex_lock(&arena->lock); + arena->dss_prec = dss_prec; + malloc_mutex_unlock(&arena->lock); +} + +void +arena_stats_merge(arena_t *arena, const char **dss, size_t *nactive, + size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats, + malloc_large_stats_t *lstats) +{ + unsigned i; + + malloc_mutex_lock(&arena->lock); + *dss = dss_prec_names[arena->dss_prec]; + *nactive += arena->nactive; + *ndirty += arena->ndirty; + + astats->mapped += arena->stats.mapped; + astats->npurge += arena->stats.npurge; + astats->nmadvise += arena->stats.nmadvise; + astats->purged += arena->stats.purged; + astats->allocated_large += arena->stats.allocated_large; + astats->nmalloc_large += arena->stats.nmalloc_large; + astats->ndalloc_large += arena->stats.ndalloc_large; + astats->nrequests_large += arena->stats.nrequests_large; + + for (i = 0; i < nlclasses; i++) { + lstats[i].nmalloc += arena->stats.lstats[i].nmalloc; + lstats[i].ndalloc += arena->stats.lstats[i].ndalloc; + lstats[i].nrequests += arena->stats.lstats[i].nrequests; + lstats[i].curruns += arena->stats.lstats[i].curruns; + } + malloc_mutex_unlock(&arena->lock); + + for (i = 0; i < NBINS; i++) { + arena_bin_t *bin = &arena->bins[i]; + + malloc_mutex_lock(&bin->lock); + bstats[i].allocated += bin->stats.allocated; + bstats[i].nmalloc += bin->stats.nmalloc; + bstats[i].ndalloc += bin->stats.ndalloc; + bstats[i].nrequests += bin->stats.nrequests; + if (config_tcache) { + bstats[i].nfills += bin->stats.nfills; + bstats[i].nflushes += bin->stats.nflushes; + } + bstats[i].nruns += bin->stats.nruns; + bstats[i].reruns += bin->stats.reruns; + bstats[i].curruns += bin->stats.curruns; + malloc_mutex_unlock(&bin->lock); + } +} + bool arena_new(arena_t *arena, unsigned ind) { unsigned i; arena_bin_t *bin; - size_t prev_run_size; arena->ind = ind; + arena->nthreads = 0; if (malloc_mutex_init(&arena->lock)) return (true); -#ifdef JEMALLOC_STATS - memset(&arena->stats, 0, sizeof(arena_stats_t)); - arena->stats.lstats = (malloc_large_stats_t *)base_alloc(nlclasses * - sizeof(malloc_large_stats_t)); - if (arena->stats.lstats == NULL) - return (true); - memset(arena->stats.lstats, 0, nlclasses * - sizeof(malloc_large_stats_t)); -# ifdef JEMALLOC_TCACHE - ql_new(&arena->tcache_ql); -# endif -#endif + if (config_stats) { + memset(&arena->stats, 0, sizeof(arena_stats_t)); + arena->stats.lstats = + (malloc_large_stats_t *)base_alloc(nlclasses * + sizeof(malloc_large_stats_t)); + if (arena->stats.lstats == NULL) + return (true); + memset(arena->stats.lstats, 0, nlclasses * + sizeof(malloc_large_stats_t)); + if (config_tcache) + ql_new(&arena->tcache_ql); + } -#ifdef JEMALLOC_PROF - arena->prof_accumbytes = 0; -#endif + if (config_prof) + arena->prof_accumbytes = 0; + + arena->dss_prec = chunk_dss_prec_get(); /* Initialize chunks. */ - ql_new(&arena->chunks_dirty); + arena_chunk_dirty_new(&arena->chunks_dirty); arena->spare = NULL; arena->nactive = 0; arena->ndirty = 0; arena->npurgatory = 0; - arena_avail_tree_new(&arena->runs_avail_clean); - arena_avail_tree_new(&arena->runs_avail_dirty); + arena_avail_tree_new(&arena->runs_avail); /* Initialize bins. */ - prev_run_size = PAGE_SIZE; - - i = 0; -#ifdef JEMALLOC_TINY - /* (2^n)-spaced tiny bins. */ - for (; i < ntbins; i++) { + for (i = 0; i < NBINS; i++) { bin = &arena->bins[i]; if (malloc_mutex_init(&bin->lock)) return (true); bin->runcur = NULL; arena_run_tree_new(&bin->runs); - - bin->reg_size = (1U << (LG_TINY_MIN + i)); - - prev_run_size = arena_bin_run_size_calc(bin, prev_run_size); - -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif + if (config_stats) + memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); } -#endif - /* Quantum-spaced bins. */ - for (; i < ntbins + nqbins; i++) { - bin = &arena->bins[i]; - if (malloc_mutex_init(&bin->lock)) - return (true); - bin->runcur = NULL; - arena_run_tree_new(&bin->runs); + return (false); +} - bin->reg_size = (i - ntbins + 1) << LG_QUANTUM; +/* + * Calculate bin_info->run_size such that it meets the following constraints: + * + * *) bin_info->run_size >= min_run_size + * *) bin_info->run_size <= arena_maxclass + * *) run header overhead <= RUN_MAX_OVRHD (or header overhead relaxed). + * *) bin_info->nregs <= RUN_MAXREGS + * + * bin_info->nregs, bin_info->bitmap_offset, and bin_info->reg0_offset are also + * calculated here, since these settings are all interdependent. + */ +static size_t +bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) +{ + size_t pad_size; + size_t try_run_size, good_run_size; + uint32_t try_nregs, good_nregs; + uint32_t try_hdr_size, good_hdr_size; + uint32_t try_bitmap_offset, good_bitmap_offset; + uint32_t try_ctx0_offset, good_ctx0_offset; + uint32_t try_redzone0_offset, good_redzone0_offset; - prev_run_size = arena_bin_run_size_calc(bin, prev_run_size); + assert(min_run_size >= PAGE); + assert(min_run_size <= arena_maxclass); -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif + /* + * Determine redzone size based on minimum alignment and minimum + * redzone size. Add padding to the end of the run if it is needed to + * align the regions. The padding allows each redzone to be half the + * minimum alignment; without the padding, each redzone would have to + * be twice as large in order to maintain alignment. + */ + if (config_fill && opt_redzone) { + size_t align_min = ZU(1) << (ffs(bin_info->reg_size) - 1); + if (align_min <= REDZONE_MINSIZE) { + bin_info->redzone_size = REDZONE_MINSIZE; + pad_size = 0; + } else { + bin_info->redzone_size = align_min >> 1; + pad_size = bin_info->redzone_size; + } + } else { + bin_info->redzone_size = 0; + pad_size = 0; } + bin_info->reg_interval = bin_info->reg_size + + (bin_info->redzone_size << 1); - /* Cacheline-spaced bins. */ - for (; i < ntbins + nqbins + ncbins; i++) { - bin = &arena->bins[i]; - if (malloc_mutex_init(&bin->lock)) - return (true); - bin->runcur = NULL; - arena_run_tree_new(&bin->runs); - - bin->reg_size = cspace_min + ((i - (ntbins + nqbins)) << - LG_CACHELINE); - - prev_run_size = arena_bin_run_size_calc(bin, prev_run_size); - -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif + /* + * Calculate known-valid settings before entering the run_size + * expansion loop, so that the first part of the loop always copies + * valid settings. + * + * The do..while loop iteratively reduces the number of regions until + * the run header and the regions no longer overlap. A closed formula + * would be quite messy, since there is an interdependency between the + * header's mask length and the number of regions. + */ + try_run_size = min_run_size; + try_nregs = ((try_run_size - sizeof(arena_run_t)) / + bin_info->reg_interval) + + 1; /* Counter-act try_nregs-- in loop. */ + if (try_nregs > RUN_MAXREGS) { + try_nregs = RUN_MAXREGS + + 1; /* Counter-act try_nregs-- in loop. */ } + do { + try_nregs--; + try_hdr_size = sizeof(arena_run_t); + /* Pad to a long boundary. */ + try_hdr_size = LONG_CEILING(try_hdr_size); + try_bitmap_offset = try_hdr_size; + /* Add space for bitmap. */ + try_hdr_size += bitmap_size(try_nregs); + if (config_prof && opt_prof && prof_promote == false) { + /* Pad to a quantum boundary. */ + try_hdr_size = QUANTUM_CEILING(try_hdr_size); + try_ctx0_offset = try_hdr_size; + /* Add space for one (prof_ctx_t *) per region. */ + try_hdr_size += try_nregs * sizeof(prof_ctx_t *); + } else + try_ctx0_offset = 0; + try_redzone0_offset = try_run_size - (try_nregs * + bin_info->reg_interval) - pad_size; + } while (try_hdr_size > try_redzone0_offset); - /* Subpage-spaced bins. */ - for (; i < nbins; i++) { - bin = &arena->bins[i]; - if (malloc_mutex_init(&bin->lock)) - return (true); - bin->runcur = NULL; - arena_run_tree_new(&bin->runs); + /* run_size expansion loop. */ + do { + /* + * Copy valid settings before trying more aggressive settings. + */ + good_run_size = try_run_size; + good_nregs = try_nregs; + good_hdr_size = try_hdr_size; + good_bitmap_offset = try_bitmap_offset; + good_ctx0_offset = try_ctx0_offset; + good_redzone0_offset = try_redzone0_offset; - bin->reg_size = sspace_min + ((i - (ntbins + nqbins + ncbins)) - << LG_SUBPAGE); + /* Try more aggressive settings. */ + try_run_size += PAGE; + try_nregs = ((try_run_size - sizeof(arena_run_t) - pad_size) / + bin_info->reg_interval) + + 1; /* Counter-act try_nregs-- in loop. */ + if (try_nregs > RUN_MAXREGS) { + try_nregs = RUN_MAXREGS + + 1; /* Counter-act try_nregs-- in loop. */ + } + do { + try_nregs--; + try_hdr_size = sizeof(arena_run_t); + /* Pad to a long boundary. */ + try_hdr_size = LONG_CEILING(try_hdr_size); + try_bitmap_offset = try_hdr_size; + /* Add space for bitmap. */ + try_hdr_size += bitmap_size(try_nregs); + if (config_prof && opt_prof && prof_promote == false) { + /* Pad to a quantum boundary. */ + try_hdr_size = QUANTUM_CEILING(try_hdr_size); + try_ctx0_offset = try_hdr_size; + /* + * Add space for one (prof_ctx_t *) per region. + */ + try_hdr_size += try_nregs * + sizeof(prof_ctx_t *); + } + try_redzone0_offset = try_run_size - (try_nregs * + bin_info->reg_interval) - pad_size; + } while (try_hdr_size > try_redzone0_offset); + } while (try_run_size <= arena_maxclass + && try_run_size <= arena_maxclass + && RUN_MAX_OVRHD * (bin_info->reg_interval << 3) > + RUN_MAX_OVRHD_RELAX + && (try_redzone0_offset << RUN_BFP) > RUN_MAX_OVRHD * try_run_size + && try_nregs < RUN_MAXREGS); - prev_run_size = arena_bin_run_size_calc(bin, prev_run_size); + assert(good_hdr_size <= good_redzone0_offset); -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif - } + /* Copy final settings. */ + bin_info->run_size = good_run_size; + bin_info->nregs = good_nregs; + bin_info->bitmap_offset = good_bitmap_offset; + bin_info->ctx0_offset = good_ctx0_offset; + bin_info->reg0_offset = good_redzone0_offset + bin_info->redzone_size; -#ifdef JEMALLOC_DEBUG - arena->magic = ARENA_MAGIC; -#endif + assert(bin_info->reg0_offset - bin_info->redzone_size + (bin_info->nregs + * bin_info->reg_interval) + pad_size == bin_info->run_size); - return (false); + return (good_run_size); } -#ifdef JEMALLOC_DEBUG static void -small_size2bin_validate(void) +bin_info_init(void) { - size_t i, size, binind; - - assert(small_size2bin[0] == 0xffU); - i = 1; -# ifdef JEMALLOC_TINY - /* Tiny. */ - for (; i < (1U << LG_TINY_MIN); i++) { - size = pow2_ceil(1U << LG_TINY_MIN); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - assert(small_size2bin[i] == binind); - } - for (; i < qspace_min; i++) { - size = pow2_ceil(i); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - assert(small_size2bin[i] == binind); - } -# endif - /* Quantum-spaced. */ - for (; i <= qspace_max; i++) { - size = QUANTUM_CEILING(i); - binind = ntbins + (size >> LG_QUANTUM) - 1; - assert(small_size2bin[i] == binind); - } - /* Cacheline-spaced. */ - for (; i <= cspace_max; i++) { - size = CACHELINE_CEILING(i); - binind = ntbins + nqbins + ((size - cspace_min) >> - LG_CACHELINE); - assert(small_size2bin[i] == binind); - } - /* Sub-page. */ - for (; i <= sspace_max; i++) { - size = SUBPAGE_CEILING(i); - binind = ntbins + nqbins + ncbins + ((size - sspace_min) - >> LG_SUBPAGE); - assert(small_size2bin[i] == binind); - } + arena_bin_info_t *bin_info; + size_t prev_run_size = PAGE; + +#define SIZE_CLASS(bin, delta, size) \ + bin_info = &arena_bin_info[bin]; \ + bin_info->reg_size = size; \ + prev_run_size = bin_info_run_size_calc(bin_info, prev_run_size);\ + bitmap_info_init(&bin_info->bitmap_info, bin_info->nregs); + SIZE_CLASSES +#undef SIZE_CLASS } -#endif -static bool -small_size2bin_init(void) -{ - - if (opt_lg_qspace_max != LG_QSPACE_MAX_DEFAULT - || opt_lg_cspace_max != LG_CSPACE_MAX_DEFAULT - || sizeof(const_small_size2bin) != small_maxclass + 1) - return (small_size2bin_init_hard()); - - small_size2bin = const_small_size2bin; -#ifdef JEMALLOC_DEBUG - assert(sizeof(const_small_size2bin) == small_maxclass + 1); - small_size2bin_validate(); -#endif - return (false); -} - -static bool -small_size2bin_init_hard(void) -{ - size_t i, size, binind; - uint8_t *custom_small_size2bin; - - assert(opt_lg_qspace_max != LG_QSPACE_MAX_DEFAULT - || opt_lg_cspace_max != LG_CSPACE_MAX_DEFAULT - || sizeof(const_small_size2bin) != small_maxclass + 1); - - custom_small_size2bin = (uint8_t *)base_alloc(small_maxclass + 1); - if (custom_small_size2bin == NULL) - return (true); - - custom_small_size2bin[0] = 0xffU; - i = 1; -#ifdef JEMALLOC_TINY - /* Tiny. */ - for (; i < (1U << LG_TINY_MIN); i++) { - size = pow2_ceil(1U << LG_TINY_MIN); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - custom_small_size2bin[i] = binind; - } - for (; i < qspace_min; i++) { - size = pow2_ceil(i); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - custom_small_size2bin[i] = binind; - } -#endif - /* Quantum-spaced. */ - for (; i <= qspace_max; i++) { - size = QUANTUM_CEILING(i); - binind = ntbins + (size >> LG_QUANTUM) - 1; - custom_small_size2bin[i] = binind; - } - /* Cacheline-spaced. */ - for (; i <= cspace_max; i++) { - size = CACHELINE_CEILING(i); - binind = ntbins + nqbins + ((size - cspace_min) >> - LG_CACHELINE); - custom_small_size2bin[i] = binind; - } - /* Sub-page. */ - for (; i <= sspace_max; i++) { - size = SUBPAGE_CEILING(i); - binind = ntbins + nqbins + ncbins + ((size - sspace_min) >> - LG_SUBPAGE); - custom_small_size2bin[i] = binind; - } - - small_size2bin = custom_small_size2bin; -#ifdef JEMALLOC_DEBUG - small_size2bin_validate(); -#endif - return (false); -} - -bool +void arena_boot(void) { size_t header_size; unsigned i; - /* Set variables according to the value of opt_lg_[qc]space_max. */ - qspace_max = (1U << opt_lg_qspace_max); - cspace_min = CACHELINE_CEILING(qspace_max); - if (cspace_min == qspace_max) - cspace_min += CACHELINE; - cspace_max = (1U << opt_lg_cspace_max); - sspace_min = SUBPAGE_CEILING(cspace_max); - if (sspace_min == cspace_max) - sspace_min += SUBPAGE; - assert(sspace_min < PAGE_SIZE); - sspace_max = PAGE_SIZE - SUBPAGE; - -#ifdef JEMALLOC_TINY - assert(LG_QUANTUM >= LG_TINY_MIN); -#endif - assert(ntbins <= LG_QUANTUM); - nqbins = qspace_max >> LG_QUANTUM; - ncbins = ((cspace_max - cspace_min) >> LG_CACHELINE) + 1; - nsbins = ((sspace_max - sspace_min) >> LG_SUBPAGE) + 1; - nbins = ntbins + nqbins + ncbins + nsbins; - - /* - * The small_size2bin lookup table uses uint8_t to encode each bin - * index, so we cannot support more than 256 small size classes. This - * limit is difficult to exceed (not even possible with 16B quantum and - * 4KiB pages), and such configurations are impractical, but - * nonetheless we need to protect against this case in order to avoid - * undefined behavior. - * - * Further constrain nbins to 255 if prof_promote is true, since all - * small size classes, plus a "not small" size class must be stored in - * 8 bits of arena_chunk_map_t's bits field. - */ -#ifdef JEMALLOC_PROF - if (opt_prof && prof_promote) { - if (nbins > 255) { - char line_buf[UMAX2S_BUFSIZE]; - malloc_write("<jemalloc>: Too many small size classes ("); - malloc_write(u2s(nbins, 10, line_buf)); - malloc_write(" > max 255)\n"); - abort(); - } - } else -#endif - if (nbins > 256) { - char line_buf[UMAX2S_BUFSIZE]; - malloc_write("<jemalloc>: Too many small size classes ("); - malloc_write(u2s(nbins, 10, line_buf)); - malloc_write(" > max 256)\n"); - abort(); - } - - if (small_size2bin_init()) - return (true); - /* * Compute the header size such that it is large enough to contain the * page map. The page map is biased to omit entries for the header @@ -2562,14 +2322,44 @@ arena_boot(void) */ map_bias = 0; for (i = 0; i < 3; i++) { - header_size = offsetof(arena_chunk_t, map) - + (sizeof(arena_chunk_map_t) * (chunk_npages-map_bias)); - map_bias = (header_size >> PAGE_SHIFT) + ((header_size & - PAGE_MASK) != 0); + header_size = offsetof(arena_chunk_t, map) + + (sizeof(arena_chunk_map_t) * (chunk_npages-map_bias)); + map_bias = (header_size >> LG_PAGE) + ((header_size & PAGE_MASK) + != 0); } assert(map_bias > 0); - arena_maxclass = chunksize - (map_bias << PAGE_SHIFT); + arena_maxclass = chunksize - (map_bias << LG_PAGE); - return (false); + bin_info_init(); +} + +void +arena_prefork(arena_t *arena) +{ + unsigned i; + + malloc_mutex_prefork(&arena->lock); + for (i = 0; i < NBINS; i++) + malloc_mutex_prefork(&arena->bins[i].lock); +} + +void +arena_postfork_parent(arena_t *arena) +{ + unsigned i; + + for (i = 0; i < NBINS; i++) + malloc_mutex_postfork_parent(&arena->bins[i].lock); + malloc_mutex_postfork_parent(&arena->lock); +} + +void +arena_postfork_child(arena_t *arena) +{ + unsigned i; + + for (i = 0; i < NBINS; i++) + malloc_mutex_postfork_child(&arena->bins[i].lock); + malloc_mutex_postfork_child(&arena->lock); } |