專題:Linux內存管理專題
關鍵詞:slab/slub/slob、slab描述符、kmalloc、本地/共享對象緩沖池、slabs_partial/slabs_full/slabs_free、avail/limit/batchcount。
從Linux內存管理框架圖可以知道:slab/slub/slob都是基於伙伴系統。
伙伴系統是以page為單位進行操作的。但是很多場景並不需要如此大的內存分配,slab就是用在這種場景的。
本章節主要內容:從slab相關數據結構講起,對slab有一個靜態的認識;然后介紹slab從創建描述符->分配緩存->釋放緩存->銷毀描述符介紹整個slab生命周;最后介紹基於slab分配器的kmalloc的運行原理。
slab分配器最終還是由伙伴系統來分配出實際的物理頁面,只不過slab分配器在這些連續的物理頁面上實現了自己的算法,以此來對小內存塊進行管理。
slab分配器相關重要函數有:
struct kmem_cache *kmem_cache_create(const char *, size_t, size_t,---------創建slab描述符kmem_cache,此時並沒有真正分配內存 unsigned long, void (*)(void *)); void *kmem_cache_alloc(struct kmem_cache *, gfp_t flags);------------------分配slab緩存對象 void kmem_cache_free(struct kmem_cache *, void *);-------------------------釋放slab緩存對象 void kmem_cache_destroy(struct kmem_cache *);-----------------------------銷毀slab描述符
每個slab由多少個頁面組成?
每個slab由一個或多個連續頁面組成,最低一個,物理連續。
slab需要的物理內存在什么時候分配?
首先kmem_cache_create是並不分配頁面,等到kmem_cache_alloc時才有可能分配頁面。首先從本地緩沖池和共享緩沖池、三大鏈表都沒有空閑對象時,才會去分配2^gfporder個頁面,然后掛入到slabs_free中。
slab描述符中空閑對象過多,是否要回收?
有兩種方式回收空閑對象:
(1)使用kmem_cache_free釋放對象,當本地和共享對象緩沖池中空閑對象數目ac->avail大於ac->limit時,系統會主動脂肪batchcount個對象。當所有空閑數目大於系統空閑對象數目極限值,並且slab沒有活躍對象時,可以銷毀此slab,回收內存。
(2)系統注冊了delayed_work,定時掃描slab描述符,回收一部分空閑對象,在cache_reap中實現。
slab的cache colour着色區作用?
使不同slab上同一個相對位置slab對象的起始地址在高速緩存中相互錯開,有利於改善高速緩存的行能。
另一個利用cache場景是Per-CPU類型本地對象緩沖池。兩個優點:讓一個對象盡可能地運行在同一個CPU上;訪問Per-CPU類型本地對象緩沖池不需要獲取額外自選鎖。
1. slab相關數據結構
slab對象的描述符struct kmem_cache:
struct kmem_cache { struct array_cache __percpu *cpu_cache; /* 1) Cache tunables. Protected by slab_mutex */ unsigned int batchcount;-----------------------------------表示當前CPU本地緩沖池array_cache為空時,從共享緩沖池或者slabs_partial/slabs_free列表中獲取對象的數目。 unsigned int limit;----------------------------------------表示當本地對象緩沖池空閑對象數目大於limit時就會主動釋放batchcount個對象,便於內核回收和銷毀slab。 unsigned int shared; unsigned int size;-----------------------------------------align過后的對象長度 struct reciprocal_value reciprocal_buffer_size; /* 2) touched by every alloc & free from the backend */ unsigned int flags; /* constant flags */------------分配掩碼 unsigned int num; /* # of objs per slab */----------slab中有多少個對象 /* 3) cache_grow/shrink */ /* order of pgs per slab (2^n) */ unsigned int gfporder;------------------------------------此slab占用z^gfporder個頁面 /* force GFP flags, e.g. GFP_DMA */ gfp_t allocflags; size_t colour; /* cache colouring range */----一個slab有幾個不同的cache line unsigned int colour_off; /* colour offset */----------一個cache order的長度,和L1 Cache Line長度相同
struct kmem_cache *freelist_cache; unsigned int freelist_size; /* constructor func */ void (*ctor)(void *obj); /* 4) cache creation/removal */ const char *name;----------------------------------------slab描述符的名稱 struct list_head list; int refcount;--------------------------------------------被引用的次數,供slab描述符銷毀參考 int object_size;-----------------------------------------對象的實際大小 int align;-----------------------------------------------對齊的大小 /* 5) statistics */ #ifdef CONFIG_DEBUG_SLAB unsigned long num_active; unsigned long num_allocations; unsigned long high_mark; unsigned long grown; unsigned long reaped; unsigned long errors; unsigned long max_freeable; unsigned long node_allocs; unsigned long node_frees; unsigned long node_overflow; atomic_t allochit; atomic_t allocmiss; atomic_t freehit; atomic_t freemiss; /* * If debugging is enabled, then the allocator can add additional * fields and/or padding to every object. size contains the total * object size including these internal fields, the following two * variables contain the offset to the user object and its size. */ int obj_offset; #endif /* CONFIG_DEBUG_SLAB */ #ifdef CONFIG_MEMCG_KMEM struct memcg_cache_params memcg_params; #endif struct kmem_cache_node *node[MAX_NUMNODES];-------slab對應的節點的struct kmem_cache_node數據結構 }
本地CPU緩沖池struct array_cache:
struct array_cache { unsigned int avail;-------------對象緩沖池中可用的對象數目 unsigned int limit; unsigned int batchcount; unsigned int touched;----------從緩沖池移除一個對象時,touched置1;收縮緩存時,touched置0。 void *entry[];-----------------保存對象的實體 };
內存節點的slab列表:
/* * The slab lists for all objects. */ struct kmem_cache_node { spinlock_t list_lock; #ifdef CONFIG_SLAB struct list_head slabs_partial; /* partial list first, better asm code */----slab鏈表中部分對象空閑 struct list_head slabs_full;----------------------------------------------------slab鏈表中沒有對象空閑 struct list_head slabs_free;----------------------------------------------------slab鏈表中所有對象空閑 unsigned long free_objects;-----------------------------------------------------三個鏈表中空閑對象數目 unsigned int free_limit;--------------------------------------------------------slab中可容許的空閑對象數目最大閾值。 unsigned int colour_next; /* Per-node cache coloring */ struct array_cache *shared; /* shared per node */----------------------------多核CPU公用的共享對象緩沖池 struct alien_cache **alien; /* on other nodes */ unsigned long next_reap; /* updated without locking */ int free_touched; /* updated without locking */ #endif #ifdef CONFIG_SLUB unsigned long nr_partial; struct list_head partial; #ifdef CONFIG_SLUB_DEBUG atomic_long_t nr_slabs; atomic_long_t total_objects; struct list_head full; #endif #endif };
SLAB Flags
/* * Flags to pass to kmem_cache_create(). * The ones marked DEBUG are only valid if CONFIG_SLAB_DEBUG is set. */ #define SLAB_DEBUG_FREE 0x00000100UL /* DEBUG: Perform (expensive) checks on free */ #define SLAB_RED_ZONE 0x00000400UL /* DEBUG: Red zone objs in a cache */ #define SLAB_POISON 0x00000800UL /* DEBUG: Poison objects */ #define SLAB_HWCACHE_ALIGN 0x00002000UL /* Align objs on cache lines */ #define SLAB_CACHE_DMA 0x00004000UL /* Use GFP_DMA memory */ #define SLAB_STORE_USER 0x00010000UL /* DEBUG: Store the last owner for bug hunting */ #define SLAB_PANIC 0x00040000UL /* Panic if kmem_cache_create() fails */
2. 創建slab描述符
kmem_cache_create的最主要功能就是填充struct kmem_cache,主要參數有:
name:slab描述符的名稱
size:緩存對象的大小
align:對齊的大小
flags:分配掩碼
ctor:對象的構造函數
kmem_cache_create函數調用核心流程是:
kmem_cache_create-----------------------------進行合法性檢查,以及是否有現成slab描述符可用 do_kmem_cache_create----------------------將主要參數配置到slab描述符,然后將得到的描述符加入slab_caches全局鏈表中。 __kmem_cache_create-------------------是創建slab描述符的核心進行對齊操作,計算需要頁面,對象數目,對slab着色等等操作。 calculate_slab_order--------------計算slab對象需要的大小,以及一個slab描述符需要多少page setup_cpu_cache-------------------繼續配置slab描述符
struct kmem_cache * kmem_cache_create(const char *name, size_t size, size_t align, unsigned long flags, void (*ctor)(void *)) { ... s = __kmem_cache_alias(name, size, align, flags, ctor);----------------檢查是否有現成的slab描述符可用,有即跳轉到out_unlock。 if (s) goto out_unlock; cache_name = kstrdup_const(name, GFP_KERNEL); if (!cache_name) { err = -ENOMEM; goto out_unlock; } s = do_kmem_cache_create(cache_name, size, size,----------------------調用do_kmem_cache_create創建slab描述符 calculate_alignment(flags, align, size), flags, ctor, NULL, NULL); ... return s; }
static struct kmem_cache * do_kmem_cache_create(const char *name, size_t object_size, size_t size, size_t align, unsigned long flags, void (*ctor)(void *), struct mem_cgroup *memcg, struct kmem_cache *root_cache) { struct kmem_cache *s; int err; err = -ENOMEM; s = kmem_cache_zalloc(kmem_cache, GFP_KERNEL);-----------------------分配一個struct kmem_cache結構體 if (!s) goto out; s->name = name; s->object_size = object_size; s->size = size; s->align = align; s->ctor = ctor;-----------------------------------------------------將參數填入struct kmem_cache結構體 ... err = __kmem_cache_create(s, flags);------------------------------- ... s->refcount = 1; list_add(&s->list, &slab_caches);----------------------------------將創建的slab描述符加入到全局變量slab_caches中 ... }
__kmem_cache_create是創建slab描述符的核心:
int __kmem_cache_create (struct kmem_cache *cachep, unsigned long flags) { size_t left_over, freelist_size; size_t ralign = BYTES_PER_WORD; gfp_t gfp; int err; size_t size = cachep->size; ... /* * Check that size is in terms of words. This is needed to avoid * unaligned accesses for some archs when redzoning is used, and makes * sure any on-slab bufctl's are also correctly aligned. */ if (size & (BYTES_PER_WORD - 1)) { size += (BYTES_PER_WORD - 1); size &= ~(BYTES_PER_WORD - 1);----------------------4字節對齊 } if (flags & SLAB_RED_ZONE) { ralign = REDZONE_ALIGN; /* If redzoning, ensure that the second redzone is suitably * aligned, by adjusting the object size accordingly. */ size += REDZONE_ALIGN - 1; size &= ~(REDZONE_ALIGN - 1); } /* 3) caller mandated alignment */ if (ralign < cachep->align) { ralign = cachep->align; } /* disable debug if necessary */ if (ralign > __alignof__(unsigned long long)) flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER); /* * 4) Store it. */ cachep->align = ralign;------------------------------對齊大小設置到struct kmem_cache if (slab_is_available())-----------------------------slab_state>=UP時,可以使用GFP_KERNEL分配,否則只能使用GFP_NOWAIT gfp = GFP_KERNEL; else gfp = GFP_NOWAIT; ... /* * Determine if the slab management is 'on' or 'off' slab. * (bootstrapping cannot cope with offslab caches so don't do * it too early on. Always use on-slab management when * SLAB_NOLEAKTRACE to avoid recursive calls into kmemleak) */ if ((size >= (PAGE_SIZE >> 5)) && !slab_early_init && !(flags & SLAB_NOLEAKTRACE)) /* * Size is large, assume best to place the slab management obj * off-slab (should allow better packing of objs). */ flags |= CFLGS_OFF_SLAB; size = ALIGN(size, cachep->align);------------------按照cachep->align對size進行對齊 /* * We should restrict the number of objects in a slab to implement * byte sized index. Refer comment on SLAB_OBJ_MIN_SIZE definition. */ if (FREELIST_BYTE_INDEX && size < SLAB_OBJ_MIN_SIZE) size = ALIGN(SLAB_OBJ_MIN_SIZE, cachep->align); left_over = calculate_slab_order(cachep, size, cachep->align, flags); if (!cachep->num) return -E2BIG; freelist_size = calculate_freelist_size(cachep->num, cachep->align); /* * If the slab has been placed off-slab, and we have enough space then * move it on-slab. This is at the expense of any extra colouring. */ if (flags & CFLGS_OFF_SLAB && left_over >= freelist_size) { flags &= ~CFLGS_OFF_SLAB; left_over -= freelist_size; } if (flags & CFLGS_OFF_SLAB) { /* really off slab. No need for manual alignment */ freelist_size = calculate_freelist_size(cachep->num, 0); #ifdef CONFIG_PAGE_POISONING /* If we're going to use the generic kernel_map_pages() * poisoning, then it's going to smash the contents of * the redzone and userword anyhow, so switch them off. */ if (size % PAGE_SIZE == 0 && flags & SLAB_POISON) flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER); #endif } cachep->colour_off = cache_line_size();----------------------------------------L1 Cache line大小,由CONFIG_ARM_L1_CACHE_SHIFT配置,此處為64B。 /* Offset must be a multiple of the alignment. */ if (cachep->colour_off < cachep->align) cachep->colour_off = cachep->align; cachep->colour = left_over / cachep->colour_off; cachep->freelist_size = freelist_size; cachep->flags = flags; cachep->allocflags = __GFP_COMP; if (CONFIG_ZONE_DMA_FLAG && (flags & SLAB_CACHE_DMA)) cachep->allocflags |= GFP_DMA; cachep->size = size; cachep->reciprocal_buffer_size = reciprocal_value(size); if (flags & CFLGS_OFF_SLAB) { cachep->freelist_cache = kmalloc_slab(freelist_size, 0u); /* * This is a possibility for one of the kmalloc_{dma,}_caches. * But since we go off slab only for object size greater than * PAGE_SIZE/8, and kmalloc_{dma,}_caches get created * in ascending order,this should not happen at all. * But leave a BUG_ON for some lucky dude. */ BUG_ON(ZERO_OR_NULL_PTR(cachep->freelist_cache)); } err = setup_cpu_cache(cachep, gfp);-------------------------------------根據slab_state狀態進行不同處理,計算limit/batchcount,分配本地對象緩沖池,共享對象緩沖池 if (err) { __kmem_cache_shutdown(cachep); return err; } return 0; }
slab_state用於表示slab分配器的狀態:
/* * State of the slab allocator. * * This is used to describe the states of the allocator during bootup. * Allocators use this to gradually bootstrap themselves. Most allocators * have the problem that the structures used for managing slab caches are * allocated from slab caches themselves. */ enum slab_state { DOWN, /* No slab functionality yet */ PARTIAL, /* SLUB: kmem_cache_node available */ PARTIAL_NODE, /* SLAB: kmalloc size for node struct available */ UP, /* Slab caches usable but not all extras yet */ FULL /* Everything is working */------------------------完全初始化 };
calculate_slab_order計算slab的大小,返回值是page order。同時也計算此slab中可以容納多少個同樣大小的對象。
static size_t calculate_slab_order(struct kmem_cache *cachep, size_t size, size_t align, unsigned long flags) { unsigned long offslab_limit; size_t left_over = 0; int gfporder; for (gfporder = 0; gfporder <= KMALLOC_MAX_ORDER; gfporder++) {------從gfporder=0開始,直到KMALLOC_MAX_ORDER=10,即從4KB到4MB大小。 unsigned int num; size_t remainder; cache_estimate(gfporder, size, align, flags, &remainder, &num); if (!num)---------------------------------------------------------不等於0則表示gfporder已經滿足條件,最低分配到一個size大小的對象。等於0則繼續下一次for循環。 continue; /* Can't handle number of objects more than SLAB_OBJ_MAX_NUM */ if (num > SLAB_OBJ_MAX_NUM)--------------------------------------slab中對象最大數目,SLAB_OBJ_MAX_NUM為255,所以所有的slab對象不超過255 break; if (flags & CFLGS_OFF_SLAB) { size_t freelist_size_per_obj = sizeof(freelist_idx_t); /* * Max number of objs-per-slab for caches which * use off-slab slabs. Needed to avoid a possible * looping condition in cache_grow(). */ if (IS_ENABLED(CONFIG_DEBUG_SLAB_LEAK)) freelist_size_per_obj += sizeof(char); offslab_limit = size; offslab_limit /= freelist_size_per_obj; if (num > offslab_limit) break; } /* Found something acceptable - save it away */ cachep->num = num; cachep->gfporder = gfporder; left_over = remainder;-------------------------------------------確定對象個數和需要的頁面數 ...
if (left_over * 8 <= (PAGE_SIZE << gfporder))-------------------滿足着色條件,退出for循環。 break;
}
return left_over;
}
cache_eastimate根據當前大小2^gfporder來計算可以容納多少個對象,以及剩下多少空間用於着色。
static void cache_estimate(unsigned long gfporder, size_t buffer_size, size_t align, int flags, size_t *left_over, unsigned int *num) { int nr_objs; size_t mgmt_size; size_t slab_size = PAGE_SIZE << gfporder; ... if (flags & CFLGS_OFF_SLAB) { mgmt_size = 0; nr_objs = slab_size / buffer_size; } else { nr_objs = calculate_nr_objs(slab_size, buffer_size,--------------可以容納對象數 sizeof(freelist_idx_t), align); mgmt_size = calculate_freelist_size(nr_objs, align); } *num = nr_objs; *left_over = slab_size - nr_objs*buffer_size - mgmt_size;------------除去對象大小和管理slab額外開銷外,剩余部分 }
3. 分配slab對象
kmem_cache_alloc是slab分配緩存對象的核心函數,在slab分配緩存過程中是全程關閉本地中斷的。
kmem_cache_alloc-->slab_alloc-->__do_cache_alloc是關中斷的。
static __always_inline void * slab_alloc(struct kmem_cache *cachep, gfp_t flags, unsigned long caller) { ... local_irq_save(save_flags); objp = __do_cache_alloc(cachep, flags);-------------------------全程關本地中斷 local_irq_restore(save_flags); ... }
由於沒有定義NUMA,所以__do_cache_alloc就僅通過___cache_alloc來分配緩存。
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags) { void *objp; struct array_cache *ac; bool force_refill = false; check_irq_off(); ac = cpu_cache_get(cachep);----------------------------------------獲取本地對象緩沖池 if (likely(ac->avail)) {-------------------------------------------本地對象緩沖池是否有空閑對象 ac->touched = 1; objp = ac_get_obj(cachep, ac, flags, false);-------------------從本地對象緩沖池中分配一個對象 /* * Allow for the possibility all avail objects are not allowed * by the current flags */ if (objp) { STATS_INC_ALLOCHIT(cachep); goto out;-------------------------------------------------如果成功獲得objp,那么直接返回指針。 } force_refill = true; } STATS_INC_ALLOCMISS(cachep); objp = cache_alloc_refill(cachep, flags, force_refill);------------是slab分配緩存的核心 /* * the 'ac' may be updated by cache_alloc_refill(), * and kmemleak_erase() requires its correct value. */ ac = cpu_cache_get(cachep); out: /* * To avoid a false negative, if an object that is in one of the * per-CPU caches is leaked, we need to make sure kmemleak doesn't * treat the array pointers as a reference to the object. */ if (objp) kmemleak_erase(&ac->entry[ac->avail]); return objp; }
cache_alloc_refill是slab分配緩存的核心:
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags, bool force_refill) { int batchcount; struct kmem_cache_node *n; struct array_cache *ac; int node; check_irq_off(); node = numa_mem_id(); if (unlikely(force_refill)) goto force_grow; retry: ac = cpu_cache_get(cachep);-----------------------------------------獲取本地對象緩沖池ac batchcount = ac->batchcount; if (!ac->touched && batchcount > BATCHREFILL_LIMIT) { /* * If there was little recent activity on this cache, then * perform only a partial refill. Otherwise we could generate * refill bouncing. */ batchcount = BATCHREFILL_LIMIT; } n = get_node(cachep, node);-----------------------------------------找到對應的slab節點 BUG_ON(ac->avail > 0 || !n); spin_lock(&n->list_lock); /* See if we can refill from the shared array */ if (n->shared && transfer_objects(ac, n->shared, batchcount)) {----判斷共享對象緩沖池(n->shared)是否有空想對象。tansfer_objects嘗試遷移batchcount個空閑對象到ac中。 n->shared->touched = 1; goto alloc_done; } while (batchcount > 0) {---------------------------------嘗試從slabs_partial/slabs_free中分配對象 struct list_head *entry; struct page *page; /* Get slab alloc is to come from. */ entry = n->slabs_partial.next; if (entry == &n->slabs_partial) { n->free_touched = 1; entry = n->slabs_free.next; if (entry == &n->slabs_free) goto must_grow;-----------------------------如果slabs_partial/slabs_free都為空,則跳到must_grow分配對象。 } page = list_entry(entry, struct page, lru); check_spinlock_acquired(cachep); /* * The slab was either on partial or free list so * there must be at least one object available for * allocation. */ BUG_ON(page->active >= cachep->num); while (page->active < cachep->num && batchcount--) { STATS_INC_ALLOCED(cachep); STATS_INC_ACTIVE(cachep); STATS_SET_HIGH(cachep); ac_put_obj(cachep, ac, slab_get_obj(cachep, page, node));---------------------ac_put_obj將slab_get_obj獲取到的對象遷移到ac中。 } /* move slabp to correct slabp list: */ list_del(&page->lru); if (page->active == cachep->num)------------------------將獲取到的slab掛到合適的鏈表。 list_add(&page->lru, &n->slabs_full); else list_add(&page->lru, &n->slabs_partial); } must_grow: n->free_objects -= ac->avail; alloc_done: spin_unlock(&n->list_lock); if (unlikely(!ac->avail)) {--------------------------------ac->avail為0表示從共享對象緩沖池、slabs_free/slabs_partial都失敗了。 int x; force_grow: x = cache_grow(cachep, flags | GFP_THISNODE, node, NULL);---在cachep中創建一個slab,並掛到slabs_free鏈表中。 /* cache_grow can reenable interrupts, then ac could change. */ ac = cpu_cache_get(cachep); node = numa_mem_id(); /* no objects in sight? abort */ if (!x && (ac->avail == 0 || force_refill)) return NULL; if (!ac->avail) /* objects refilled by interrupt? */ goto retry; } ac->touched = 1; return ac_get_obj(cachep, ac, flags, force_refill); }
4. 釋放slab對象
slab釋放對象通過kmem_cache_free進行,在釋放過程中也是全程關中斷的。
一個slab描述符中可能有多個對象,因此釋放對象需要兩個參數才能確定釋放內容。
void kmem_cache_free(struct kmem_cache *cachep, void *objp) { unsigned long flags; cachep = cache_from_obj(cachep, objp);-----------------------------通過對象找到slab描述符 if (!cachep) return; local_irq_save(flags); debug_check_no_locks_freed(objp, cachep->object_size); if (!(cachep->flags & SLAB_DEBUG_OBJECTS)) debug_check_no_obj_freed(objp, cachep->object_size); __cache_free(cachep, objp, _RET_IP_);-------------------------------關本地中斷 local_irq_restore(flags); trace_kmem_cache_free(_RET_IP_, objp); }
cache_from_obj通過要釋放對象虛擬地址,找到所在頁面,繼而找到對應的struct kmem_cache結構體。
然后將轉換得到的slab描述符和入參描述符對比,即可判斷兩者是否有效。
static inline struct kmem_cache *cache_from_obj(struct kmem_cache *s, void *x) { ... page = virt_to_head_page(x);-----------由virt_to_page找到對應的page,再找到first_page。 cachep = page->slab_cache;-------------first_page中有指向slab描述符的slab_cache if (slab_equal_or_root(cachep, s))-----判斷兩者是否吻合 return cachep; ... return s; }
__cache_free是釋放slab對象的核心:
首先通過slab描述符找到本地對象緩沖池;
然后判斷ac->avail和ac->limit大小,如果avail超過limit,則需要cache_flusharray去回收空閑對象;
最后ac_put_obj將對象釋放到本地對象緩沖池ac中,釋放過程結束。
static inline void __cache_free(struct kmem_cache *cachep, void *objp, unsigned long caller) { struct array_cache *ac = cpu_cache_get(cachep);----------------找到本地對象緩沖池 check_irq_off(); kmemleak_free_recursive(objp, cachep->flags); objp = cache_free_debugcheck(cachep, objp, caller); kmemcheck_slab_free(cachep, objp, cachep->object_size); /* * Skip calling cache_free_alien() when the platform is not numa. * This will avoid cache misses that happen while accessing slabp (which * is per page memory reference) to get nodeid. Instead use a global * variable to skip the call, which is mostly likely to be present in * the cache. */ if (nr_online_nodes > 1 && cache_free_alien(cachep, objp)) return; if (ac->avail < ac->limit) { STATS_INC_FREEHIT(cachep); } else { STATS_INC_FREEMISS(cachep); cache_flusharray(cachep, ac);---------------------------------嘗試回收空閑對象 } ac_put_obj(cachep, ac, objp);-------------------------------------將對象釋放到本地對象緩沖池ac中 }
5. kmalloc分配函數
kmalloc函數基於slab機制,分配的內存大小也是對齊到2^order個字節。
分配的時候是從kmalloc-xxx的slab描述符種分配一個對象。
這些kmalloc-xxx的slab描述符是由create_kmalloc_caches在系統初始換的時候創建的。
PS:下面代碼根據slub進行分析。
5.1 kmalloc slab描述符創建
create_kmalloc_caches的調用路徑是start_kernel-->mm_init-->kmem_cache_init-->create_kmalloc_caches。
再初始化之前,弄明白這三個參數KMALLOC_SHIFT_LOW, KMALLOC_SHIFT_HIGH, KMALLOC_SHIFT_MAX很重要。
#define CONFIG_ARM_L1_CACHE_SHIFT 6----------------------------------------6,對應64B ================================================= #define L1_CACHE_SHIFT CONFIG_ARM_L1_CACHE_SHIFT #define L1_CACHE_BYTES (1 << L1_CACHE_SHIFT)------------------------即為64B /* * Memory returned by kmalloc() may be used for DMA, so we must make * sure that all such allocations are cache aligned. Otherwise, * unrelated code may cause parts of the buffer to be read into the * cache before the transfer is done, causing old data to be seen by * the CPU. */ #define ARCH_DMA_MINALIGN L1_CACHE_BYTES-------------------------------和L1 Cache對齊,即64B對齊 ================================================= #define ARCH_KMALLOC_MINALIGN ARCH_DMA_MINALIGN---------------------------即為64B #define KMALLOC_MIN_SIZE ARCH_DMA_MINALIGN--------------------------------即為64B #define KMALLOC_SHIFT_LOW ilog2(ARCH_DMA_MINALIGN)------------------------位移量為6,對應64B =================================================/* * SLUB directly allocates requests fitting in to an order-1 page * (PAGE_SIZE*2). Larger requests are passed to the page allocator. */ #define KMALLOC_SHIFT_HIGH (PAGE_SHIFT + 1)---------------------------位移量為13,對應8KB大小 #define KMALLOC_SHIFT_MAX (MAX_ORDER + PAGE_SHIFT)--------------------位移量為23,對應8MB大小
所以:
KMALLOC_MIN_SIZE=64 KMALLOC_SHIFT_LOW=6 KMALLOC_SHIFT_HIGH=13 KMALLOC_SHIFT_MAX=23
對於kmalloc尺寸小於192B從哪個slab描述符中分配緩存,進行了特殊的映射。
/* * Conversion table for small slabs sizes / 8 to the index in the * kmalloc array. This is necessary for slabs < 192 since we have non power * of two cache sizes there. The size of larger slabs can be determined using * fls. */ static s8 size_index[24] = { 3, /* 8 */ 4, /* 16 */ 5, /* 24 */ 5, /* 32 */ 6, /* 40 */ 6, /* 48 */ 6, /* 56 */ 6, /* 64 */ 1, /* 72 */ 1, /* 80 */ 1, /* 88 */ 1, /* 96 */ 7, /* 104 */ 7, /* 112 */ 7, /* 120 */ 7, /* 128 */ 2, /* 136 */ 2, /* 144 */ 2, /* 152 */ 2, /* 160 */ 2, /* 168 */ 2, /* 176 */ 2, /* 184 */ 2 /* 192 */ };
size_index的數值對應kmalloc_caches的下標,kmalloc_caches的內容由create_kmalloc_caches創建。
/* * Create the kmalloc array. Some of the regular kmalloc arrays * may already have been created because they were needed to * enable allocations for slab creation. */ void __init create_kmalloc_caches(unsigned long flags) { int i; /* * Patch up the size_index table if we have strange large alignment * requirements for the kmalloc array. This is only the case for * MIPS it seems. The standard arches will not generate any code here. * * Largest permitted alignment is 256 bytes due to the way we * handle the index determination for the smaller caches. * * Make sure that nothing crazy happens if someone starts tinkering * around with ARCH_KMALLOC_MINALIGN */ BUILD_BUG_ON(KMALLOC_MIN_SIZE > 256 || (KMALLOC_MIN_SIZE & (KMALLOC_MIN_SIZE - 1))); for (i = 8; i < KMALLOC_MIN_SIZE; i += 8) { int elem = size_index_elem(i); if (elem >= ARRAY_SIZE(size_index)) break; size_index[elem] = KMALLOC_SHIFT_LOW;---------------------------- } if (KMALLOC_MIN_SIZE >= 64) { /* * The 96 byte size cache is not used if the alignment * is 64 byte. */ for (i = 64 + 8; i <= 96; i += 8) size_index[size_index_elem(i)] = 7; } if (KMALLOC_MIN_SIZE >= 128) { /* * The 192 byte sized cache is not used if the alignment * is 128 byte. Redirect kmalloc to use the 256 byte cache * instead. */ for (i = 128 + 8; i <= 192; i += 8) size_index[size_index_elem(i)] = 8; } for (i = KMALLOC_SHIFT_LOW; i <= KMALLOC_SHIFT_HIGH; i++) {------------------------從order=8開始到order=13,這里創建kmalloc-64/kmalloc-128/kmalloc-256/kmalloc-512/kmalloc-1024/kmalloc-2048/kmalloc-4096/kmalloc-8192 if (!kmalloc_caches[i]) { kmalloc_caches[i] = create_kmalloc_cache(NULL, 1 << i, flags); } /* * Caches that are not of the two-to-the-power-of size. * These have to be created immediately after the * earlier power of two caches */ if (KMALLOC_MIN_SIZE <= 32 && !kmalloc_caches[1] && i == 6)----------------KMALLOC_MIN_SIZE為64,跳過 kmalloc_caches[1] = create_kmalloc_cache(NULL, 96, flags); if (KMALLOC_MIN_SIZE <= 64 && !kmalloc_caches[2] && i == 7)----------------創建kmalloc-192 kmalloc_caches[2] = create_kmalloc_cache(NULL, 192, flags); } /* Kmalloc array is now usable */ slab_state = UP; for (i = 0; i <= KMALLOC_SHIFT_HIGH; i++) { struct kmem_cache *s = kmalloc_caches[i]; char *n; if (s) { n = kasprintf(GFP_NOWAIT, "kmalloc-%d", kmalloc_size(i));--------------修改slab描述符名稱 BUG_ON(!n); s->name = n; } } ... }
其中size_index經過重映射之后變成了如下。
所以8B/16B/24B/32B/40B/48B/56B/64B都使用kmalloc-64;
72B/80B/88B/96B/104B/112B/120B/128B都是用kmalloc-128;
136B/144B/152B/160B/168B/176B/184B/192B都使用kmalloc-192。
size_index[0]=6 /*8*/
size_index[1]=6 /*16*/
size_index[2]=6 /*24*/
size_index[3]=6 /*32*/
size_index[4]=6 /*40*/
size_index[5]=6 /*48*/
size_index[6]=6 /*56*/
size_index[7]=6 /*64*/
size_index[8]=7 /*72*/
size_index[9]=7 /*80*/
size_index[10]=7 /*88*/
size_index[11]=7 /*96*/
size_index[12]=7 /*104*/
size_index[13]=7 /*112*/
size_index[14]=7 /*120*/
size_index[15]=7 /*128*/
size_index[16]=2 /*136*/
size_index[17]=2 /*144*/
size_index[18]=2 /*152*/
size_index[19]=2 /*160*/
size_index[20]=2 /*168*/
size_index[21]=2 /*176*/
size_index[22]=2 /*184*/
size_index[23]=2 /*192*/
看看/proc/slabinfo中的最終結果如何?
kmalloc-8192 12 12 8192 4 8 : tunables 0 0 0 : slabdata 3 3 0 kmalloc-4096 61 88 4096 8 8 : tunables 0 0 0 : slabdata 11 11 0 kmalloc-2048 48 48 2048 16 8 : tunables 0 0 0 : slabdata 3 3 0 kmalloc-1024 96 96 1024 16 4 : tunables 0 0 0 : slabdata 6 6 0 kmalloc-512 384 384 512 16 2 : tunables 0 0 0 : slabdata 24 24 0 kmalloc-256 208 208 256 16 1 : tunables 0 0 0 : slabdata 13 13 0 kmalloc-192 441 441 192 21 1 : tunables 0 0 0 : slabdata 21 21 0 kmalloc-128 1280 1280 128 32 1 : tunables 0 0 0 : slabdata 40 40 0 kmalloc-64 4416 4416 64 64 1 : tunables 0 0 0 : slabdata 69 69 0
5.2 kmalloc
kmalloc是按字節分配內存的接口,針對不同大小采取了不同的操作。
KMALLOC_MAX_CACHE_SIZE是一個分界線,大於8KB的內存分配需要kmalloc_large進行處理。
另外對於小於等於192B,通過size_index映射到不同kmalloc-xxx slab描述符。
大於192B小於KMALLOC_MAX_CACHE_SIZE,通過fls找到對應的kmalloc_caches索引號。
static __always_inline void *kmalloc(size_t size, gfp_t flags) { if (__builtin_constant_p(size)) { if (size > KMALLOC_MAX_CACHE_SIZE)---------------------------------大於8KB內存使用kmalloc_large來分配 return kmalloc_large(size, flags); #ifndef CONFIG_SLOB if (!(flags & GFP_DMA)) { int index = kmalloc_index(size);-------------------------------找到slab描述符 if (!index) return ZERO_SIZE_PTR; return kmem_cache_alloc_trace(kmalloc_caches[index],-----------調用slab_alloc分配緩存 flags, size); } #endif } return __kmalloc(size, flags);-----------------------------------------另一種情況分支 } 不同分配器分支,這里取slub: void *__kmalloc(size_t size, gfp_t flags) { struct kmem_cache *s; void *ret; if (unlikely(size > KMALLOC_MAX_CACHE_SIZE))--------------------------再一次檢查8KB這個大小,kmalloc_large分配8KB+緩存 return kmalloc_large(size, flags); s = kmalloc_slab(size, flags);----------------------------------------從預分配slab描述符中找到struct kmem_cache。 if (unlikely(ZERO_OR_NULL_PTR(s))) return s; ret = slab_alloc(s, flags, _RET_IP_);---------------------------------調用slab_alloc分配緩存 trace_kmalloc(_RET_IP_, ret, size, s->size, flags); kasan_kmalloc(s, ret, size); return ret; } /* * Find the kmem_cache structure that serves a given size of * allocation */ struct kmem_cache *kmalloc_slab(size_t size, gfp_t flags) { int index; if (unlikely(size > KMALLOC_MAX_SIZE)) { WARN_ON_ONCE(!(flags & __GFP_NOWARN)); return NULL; } if (size <= 192) { if (!size) return ZERO_SIZE_PTR; index = size_index[size_index_elem(size)];------------------小於等於192B大小通過size_index得出slab描述符索引 } else index = fls(size - 1);--------------------------------------fls根據大小計算most-significant位索引,范圍從192B~8KB。 #ifdef CONFIG_ZONE_DMA if (unlikely((flags & GFP_DMA))) return kmalloc_dma_caches[index]; #endif return kmalloc_caches[index]; }
為了提高分配緩存的速度,降低函數調用路徑。關鍵函數進行了__always_inline修飾。
kmem_cache_alloc slab_alloc--> slab_alloc_node--> static __always_inline void *slab_alloc(struct kmem_cache *s, gfp_t gfpflags, unsigned long addr) { return slab_alloc_node(s, gfpflags, NUMA_NO_NODE, addr); } static __always_inline void *slab_alloc_node(struct kmem_cache *s, gfp_t gfpflags, int node, unsigned long addr)
6. 創建slab描述符實驗
7. slab分配器相關調試接口
7.1 解讀/proc/slabinfo
/proc/slabinfo是slab分配器的統計信息,打開CONFIG_DEBUG_SLAB可以獲取更多信息。
slabinfo - version: 2.1 (statistics) # name <active_objs> <num_objs> <objsize> <objperslab> <pagesperslab> : tunables <limit> <batchcount> <sharedfactor> : slabdata <active_slabs> <num_slabs> <sharedavail> : globalstat <listallocs> <maxobjs> <grown> <reaped> <error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow> : cpustat <allochit> <allocmiss> <freehit> <freemiss>... kmalloc-4194304 0 0 4194304 1 1024 : tunables 1 1 0 : slabdata 0 0 0 : globalstat 0 0 0 0 0 0 0 0 0 : cpustat 0 0 0 0 kmalloc-2097152 0 0 2097152 1 512 : tunables 1 1 0 : slabdata 0 0 0 : globalstat 0 0 0 0 0 0 0 0 0 : cpustat 0 0 0 0 kmalloc-1048576 0 0 1048576 1 256 : tunables 1 1 0 : slabdata 0 0 0 : globalstat 0 0 0 0 0 0 0 0 0 : cpustat 0 0 0 0 kmalloc-524288 0 0 524288 1 128 : tunables 1 1 0 : slabdata 0 0 0 : globalstat 0 0 0 0 0 0 0 0 0 : cpustat 0 0 0 0 kmalloc-262144 1 1 262144 1 64 : tunables 1 1 0 : slabdata 1 1 0 : globalstat 1 1 1 0 0 0 0 0 0 : cpustat 0 1 0 0 kmalloc-131072 0 0 131072 1 32 : tunables 8 4 0 : slabdata 0 0 0 : globalstat 0 0 0 0 0 0 0 0 0 : cpustat 0 0 0 0 kmalloc-65536 1 1 65536 1 16 : tunables 8 4 0 : slabdata 1 1 0 : globalstat 2 2 2 1 0 0 0 0 0 : cpustat 0 2 1 0 kmalloc-32768 1 1 32768 1 8 : tunables 8 4 0 : slabdata 1 1 0 : globalstat 2 2 2 1 0 0 0 0 0 : cpustat 0 2 1 0 kmalloc-16384 3 3 16384 1 4 : tunables 8 4 0 : slabdata 3 3 0 : globalstat 4 3 3 0 0 0 0 0 0 : cpustat 0 4 1 0 kmalloc-8192 7 7 8192 1 2 : tunables 8 4 0 : slabdata 7 7 0 : globalstat 9 8 8 1 0 0 0 0 0 : cpustat 0 9 2 0 kmalloc-4096 29 78 4096 1 1 : tunables 24 12 8 : slabdata 29 78 0 : globalstat 105 105 105 10 0 20 0 0 0 : cpustat 878 139 974 34 kmalloc-2048 18 18 4096 1 1 : tunables 24 12 8 : slabdata 18 18 0 : globalstat 19 18 19 1 0 0 0 0 0 : cpustat 4 19 5 0 kmalloc-1024 135 135 4096 1 1 : tunables 24 12 8 : slabdata 135 135 0 : globalstat 135 135 135 0 0 0 0 0 0 : cpustat 42 135 43 0 kmalloc-512 425 425 4096 1 1 : tunables 24 12 8 : slabdata 425 425 0 : globalstat 425 425 425 0 0 0 0 0 0 : cpustat 137 425 137 0 kmalloc-256 112 112 4096 1 1 : tunables 24 12 8 : slabdata 112 112 0 : globalstat 121 119 120 8 0 0 0 0 0 : cpustat 608 121 619 0 kmalloc-192 248 248 4096 1 1 : tunables 24 12 8 : slabdata 248 248 0 : globalstat 248 248 248 0 0 0 0 0 0 : cpustat 11 248 11 0 kmalloc-128 977 977 4096 1 1 : tunables 24 12 8 : slabdata 977 977 0 : globalstat 981 977 981 4 0 0 0 0 0 : cpustat 437 981 443 0 kmalloc-64 26746 26838 64 63 1 : tunables 32 16 8 : slabdata 426 426 48 : globalstat 26845 26838 426 0 0 0 0 0 0 : cpustat 27112 2168 2589 23 kmem_cache 142 142 4096 1 2 : tunables 24 12 8 : slabdata 142 142 0 : globalstat 142 142 142 0 0 0 0 0 0 : cpustat 0 142 0 0
8 kmem相關Tracepoint
kmem跟蹤事件主要跟蹤內核slab和page的分配和釋放行為,主要可以分為5大部分。
這些events的詳細解釋參考:Documentation/trace/events-kmem.txt。
8.1 Slab allocation of small objects of unknown type (kmalloc)
那些函數調用?Trace什么樣子?有什么用途?
kfree---------------------------kfree
kmalloc-------------------------kmalloc/__kmalloc等類kmalloc函數
kmalloc_node--------------------kmalloc_node/__kmalloc_node等類kmalloc_node函數
kmalloc_node和kmalloc的區別是多了個node參數,對NUMA系統來說需要node進行區分。在非NUMA系統,意義不大。
相關Log如下,從中可以看出調用者call_site,分配內存地址ptr,請求分配大小bytes_req,實際分配大小bytes_alloc,分配掩碼gfp_flags。
bytes_alloc>=bytes_req,並且進行了2^order對齊;但是call_site是個地址,可讀性較差。
# tracer: nop # # entries-in-buffer/entries-written: 14/14 #P:4 # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / delay # TASK-PID CPU# |||| TIMESTAMP FUNCTION # | | | |||| | | sh-647 [001] .... 843.814154: kmalloc: call_site=c012fbec ptr=ee042600 bytes_req=200 bytes_alloc=256 gfp_flags=GFP_KERNEL|GFP_ZERO sh-647 [001] .... 843.815146: kmalloc: call_site=c0175e4c ptr=eeab2580 bytes_req=104 bytes_alloc=128 gfp_flags=GFP_KERNEL sh-647 [001] .... 843.815185: kmalloc: call_site=c0174818 ptr=ee042a00 bytes_req=224 bytes_alloc=256 gfp_flags=GFP_KERNEL sh-647 [001] .... 843.816017: kfree: call_site=c0176744 ptr= (null) sh-647 [001] .... 843.816029: kfree: call_site=c017674c ptr=ee042a00 sh-647 [001] .... 843.816129: kfree: call_site=c0175eb4 ptr=eeab2580 sh-647 [001] .... 843.816143: kfree: call_site=c012ebdc ptr=ee042600 sh-647 [001] .... 843.816149: kfree: call_site=c01300e0 ptr= (null) sh-647 [001] .... 843.816776: kmalloc: call_site=c0184928 ptr=ee9994c0 bytes_req=12 bytes_alloc=64 gfp_flags=GFP_KERNEL sh-647 [001] .... 843.816868: kfree: call_site=c014ff80 ptr=ee9994c0 ...
對call_site進行一下簡單的改造,使其可以直接打印字符串:
diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h old mode 100644 new mode 100755 index 4ad10ba..5c404bb --- a/include/trace/events/kmem.h +++ b/include/trace/events/kmem.h @@ -34,7 +34,7 @@ DECLARE_EVENT_CLASS(kmem_alloc, __entry->gfp_flags = gfp_flags; ), - TP_printk("call_site=%lx ptr=%p bytes_req=%zu bytes_alloc=%zu gfp_flags=%s", + TP_printk("call_site=%pf ptr=%p bytes_req=%zu bytes_alloc=%zu gfp_flags=%s", __entry->call_site, __entry->ptr, __entry->bytes_req, @@ -87,7 +87,7 @@ DECLARE_EVENT_CLASS(kmem_alloc_node, __entry->node = node; ), - TP_printk("call_site=%lx ptr=%p bytes_req=%zu bytes_alloc=%zu gfp_flags=%s node=%d", + TP_printk("call_site=%pf ptr=%p bytes_req=%zu bytes_alloc=%zu gfp_flags=%s node=%d", __entry->call_site, __entry->ptr, __entry->bytes_req, @@ -130,7 +130,7 @@ DECLARE_EVENT_CLASS(kmem_free, __entry->ptr = ptr; ), - TP_printk("call_site=%lx ptr=%p", __entry->call_site, __entry->ptr) + TP_printk("call_site=%pf ptr=%p", __entry->call_site, __entry->ptr) ); DEFINE_EVENT(kmem_free, kfree,
修改后的結果如下,ptr是kmalloc和kfree的聯系樞紐,兩者必須成對,不然就可能存在內存泄露。
同時可以看到同一個ptr的kmalloc和kfree的call_site,對此內存的申請釋放路徑就有個大概的了解。
# tracer: nop # # entries-in-buffer/entries-written: 15/15 #P:4 # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / delay # TASK-PID CPU# |||| TIMESTAMP FUNCTION # | | | |||| | | sh-640 [000] .... 97.451247: kmalloc: call_site=tracepoint_probe_register ptr=ee3ef400 bytes_req=24 bytes_alloc=64 gfp_flags=GFP_KERNEL sh-646 [001] .... 102.511304: kmalloc: call_site=do_execveat_common ptr=ee0dd400 bytes_req=200 bytes_alloc=256 gfp_flags=GFP_KERNEL|GFP_ZERO sh-646 [001] .... 102.513041: kmalloc: call_site=load_elf_binary ptr=eead9880 bytes_req=104 bytes_alloc=128 gfp_flags=GFP_KERNEL sh-646 [001] .... 102.513149: kmalloc: call_site=load_elf_phdrs ptr=ee0dd000 bytes_req=224 bytes_alloc=256 gfp_flags=GFP_KERNEL sh-646 [001] .... 102.513831: kfree: call_site=load_elf_binary ptr= (null) sh-646 [001] .... 102.513878: kfree: call_site=load_elf_binary ptr=ee0dd000 sh-646 [001] .... 102.513981: kfree: call_site=load_elf_binary ptr=eead9880 sh-646 [001] .... 102.513996: kfree: call_site=free_bprm ptr=ee0dd400 sh-646 [001] .... 102.514002: kfree: call_site=do_execveat_common ptr= (null) sh-646 [001] .... 102.514629: kmalloc: call_site=proc_self_follow_link ptr=ed4aaf80 bytes_req=12 bytes_alloc=64 gfp_flags=GFP_KERNEL sh-646 [001] .... 102.514721: kfree: call_site=kfree_put_link ptr=ed4aaf80 ...
所以基於kmalloc/kmalloc_node/kfree這幾個events,可以判斷一個進程分配了多少內存;在運行過程中是否存在內存泄露,即kmalloc沒有對應的kfree。
8.2 Slab allocation of small objects of known type
kmem_cache_alloc/kmem_cache_alloc_node/kmem_cache_free基本上和類kmalloc函數一一對應,兩者的使用和表達的含義基本一致。只是對應的分配函數不一樣。
kmem_cache_alloc------------------------kmem_cache_alloc
kmem_cache_alloc_node-------------------kmem_cache_alloc_node
kmem_cache_free-------------------------kmem_cache_free
kmem_cache_alloc類事件的用途和kmalloc類基本差不多,可以通過call_site找到調用者;可以通過kmem_cache_alloc和kmem_cache_free是否成對出現而判斷內存泄露問題。
實例如下:
# tracer: nop # # entries-in-buffer/entries-written: 80/80 #P:4 # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / delay # TASK-PID CPU# |||| TIMESTAMP FUNCTION # | | | |||| | | sh-640 [000] .... 598.446568: kmem_cache_alloc: call_site=getname_flags ptr=ed4af000 bytes_req=4096 bytes_alloc=4096 gfp_flags=GFP_KERNEL sh-640 [000] .... 598.446611: kmem_cache_alloc: call_site=get_empty_filp ptr=eeb17ac0 bytes_req=192 bytes_alloc=192 gfp_flags=GFP_KERNEL|GFP_ZERO sh-640 [000] .... 598.446673: kmem_cache_alloc: call_site=__d_alloc ptr=ee445660 bytes_req=136 bytes_alloc=136 gfp_flags=GFP_KERNEL sh-640 [000] .... 598.446751: kmem_cache_free: call_site=putname ptr=ed4af000 sh-640 [000] .... 598.446808: kmem_cache_alloc: call_site=SyS_getcwd ptr=ed4af000 bytes_req=4096 bytes_alloc=4096 gfp_flags=GFP_KERNEL sh-640 [000] .... 598.446839: kmem_cache_free: call_site=SyS_getcwd ptr=ed4af000 sh-640 [000] .... 601.702831: kmem_cache_alloc: call_site=getname_flags ptr=ed4af000 bytes_req=4096 bytes_alloc=4096 gfp_flags=GFP_KERNEL sh-640 [000] .... 601.702884: kmem_cache_alloc: call_site=get_empty_filp ptr=eeb17ac0 bytes_req=192 bytes_alloc=192 gfp_flags=GFP_KERNEL|GFP_ZERO sh-640 [000] .... 601.703028: kmem_cache_free: call_site=putname ptr=ed4af000 sh-640 [000] .... 601.703560: kmem_cache_alloc: call_site=copy_process.part.12 ptr=ee9af080 bytes_req=1280 bytes_alloc=1280 gfp_flags=GFP_KERNEL ...
8.3 Page allocation
mm_page_alloc
mm_page_alloc_zone_locked
mm_page_free
mm_page_free_batched
8.4 Per-CPU Allocator Activity
mm_page_alloc_zone_locked
mm_page_pcpu_drain
8.5 External Fragmentation
mm_page_alloc_extfrag