// SPDX-License-Identifier: GPL-2.0-only/* Copyright (c) 2022 Meta Platforms, Inc. and affiliates. */#include <linux/mm.h>#include <linux/llist.h>#include <linux/bpf.h>#include <linux/irq_work.h>#include <linux/bpf_mem_alloc.h>#include <linux/memcontrol.h>#include <asm/local.h> /* Any context (including NMI) BPF specific memory allocator. * * Tracing BPF programs can attach to kprobe and fentry. Hence they * run in unknown context where calling plain kmalloc() might not be safe. * * Front-end kmalloc() with per-cpu per-bucket cache of free elements. * Refill this cache asynchronously from irq_work. * * CPU_0 buckets * 16 32 64 96 128 196 256 512 1024 2048 4096 * ... * CPU_N buckets * 16 32 64 96 128 196 256 512 1024 2048 4096 * * The buckets are prefilled at the start. * BPF programs always run with migration disabled. * It's safe to allocate from cache of the current cpu with irqs disabled. * Free-ing is always done into bucket of the current cpu as well. * irq_work trims extra free elements from buckets with kfree * and refills them with kmalloc, so global kmalloc logic takes care * of freeing objects allocated by one cpu and freed on another. * * Every allocated objected is padded with extra 8 bytes that contains * struct llist_node. */#define LLIST_NODE_SZ sizeof(struct llist_node) #define BPF_MEM_ALLOC_SIZE_MAX 4096 /* similar to kmalloc, but sizeof == 8 bucket is gone */static u8 size_index[24] __ro_after_init = { 3, /* 8 */ 3, /* 16 */ 4, /* 24 */ 4, /* 32 */ 5, /* 40 */ 5, /* 48 */ 5, /* 56 */ 5, /* 64 */ 1, /* 72 */ 1, /* 80 */ 1, /* 88 */ 1, /* 96 */ 6, /* 104 */ 6, /* 112 */ 6, /* 120 */ 6, /* 128 */ 2, /* 136 */ 2, /* 144 */ 2, /* 152 */ 2, /* 160 */ 2, /* 168 */ 2, /* 176 */ 2, /* 184 */ 2 /* 192 */}; static int bpf_mem_cache_idx(size_t size){ if (!size || size > BPF_MEM_ALLOC_SIZE_MAX) return -1; if (size <= 192) return size_index[(size - 1) / 8] - 1; return fls(size - 1) - 2;} #define NUM_CACHES 11 struct bpf_mem_cache { /* per-cpu list of free objects of size 'unit_size'. * All accesses are done with interrupts disabled and 'active' counter * protection with __llist_add() and __llist_del_first(). */ struct llist_head free_llist; local_t active; /* Operations on the free_list from unit_alloc/unit_free/bpf_mem_refill * are sequenced by per-cpu 'active' counter. But unit_free() cannot * fail. When 'active' is busy the unit_free() will add an object to * free_llist_extra. */ struct llist_head free_llist_extra; struct irq_work refill_work; struct obj_cgroup *objcg; int unit_size; /* count of objects in free_llist */ int free_cnt; int low_watermark, high_watermark, batch; int percpu_size; bool draining; struct bpf_mem_cache *tgt; /* list of objects to be freed after RCU GP */ struct llist_head free_by_rcu; struct llist_node *free_by_rcu_tail; struct llist_head waiting_for_gp; struct llist_node *waiting_for_gp_tail; struct rcu_head rcu; atomic_t call_rcu_in_progress; struct llist_head free_llist_extra_rcu; /* list of objects to be freed after RCU tasks trace GP */ struct llist_head free_by_rcu_ttrace; struct llist_head waiting_for_gp_ttrace; struct rcu_head rcu_ttrace; atomic_t call_rcu_ttrace_in_progress;}; struct bpf_mem_caches { struct bpf_mem_cache cache[NUM_CACHES];}; static const u16 sizes[NUM_CACHES] = {96, 192, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096}; static struct llist_node notrace *__llist_del_first(struct llist_head *head){ struct llist_node *entry, *next; entry = head->first; if (!entry) return NULL; next = entry->next; head->first = next; return entry;} static void *__alloc(struct bpf_mem_cache *c, int node, gfp_t flags){ if (c->percpu_size) { void __percpu **obj = kmalloc_node(c->percpu_size, flags, node); void __percpu *pptr = __alloc_percpu_gfp(c->unit_size, 8, flags); if (!obj || !pptr) { free_percpu(pptr); kfree(obj); return NULL; } obj[1] = pptr; return obj; } return kmalloc_node(c->unit_size, flags | __GFP_ZERO, node);} static struct mem_cgroup *get_memcg(const struct bpf_mem_cache *c){#ifdef CONFIG_MEMCG if (c->objcg) return get_mem_cgroup_from_objcg(c->objcg); return root_mem_cgroup;#else return NULL;#endif} static void inc_active(struct bpf_mem_cache *c, unsigned long *flags){ if (IS_ENABLED(CONFIG_PREEMPT_RT)) /* In RT irq_work runs in per-cpu kthread, so disable * interrupts to avoid preemption and interrupts and * reduce the chance of bpf prog executing on this cpu * when active counter is busy. */ local_irq_save(*flags); /* alloc_bulk runs from irq_work which will not preempt a bpf * program that does unit_alloc/unit_free since IRQs are * disabled there. There is no race to increment 'active' * counter. It protects free_llist from corruption in case NMI * bpf prog preempted this loop. */ WARN_ON_ONCE(local_inc_return(&c->active) != 1);} static void dec_active(struct bpf_mem_cache *c, unsigned long *flags){ local_dec(&c->active); if (IS_ENABLED(CONFIG_PREEMPT_RT)) local_irq_restore(*flags);} static void add_obj_to_free_list(struct bpf_mem_cache *c, void *obj){ unsigned long flags; inc_active(c, &flags); __llist_add(obj, &c->free_llist); c->free_cnt++; dec_active(c, &flags);} /* Mostly runs from irq_work except __init phase. */static void alloc_bulk(struct bpf_mem_cache *c, int cnt, int node, bool atomic){ struct mem_cgroup *memcg = NULL, *old_memcg; gfp_t gfp; void *obj; int i; gfp = __GFP_NOWARN | __GFP_ACCOUNT; gfp |= atomic ? GFP_NOWAIT : GFP_KERNEL; for (i = 0; i < cnt; i++) { /* * For every 'c' llist_del_first(&c->free_by_rcu_ttrace); is * done only by one CPU == current CPU. Other CPUs might * llist_add() and llist_del_all() in parallel. */ obj = llist_del_first(&c->free_by_rcu_ttrace); if (!obj) break; add_obj_to_free_list(c, obj); } if (i >= cnt) return; for (; i < cnt; i++) { obj = llist_del_first(&c->waiting_for_gp_ttrace); if (!obj) break; add_obj_to_free_list(c, obj); } if (i >= cnt) return; memcg = get_memcg(c); old_memcg = set_active_memcg(memcg); for (; i < cnt; i++) { /* Allocate, but don't deplete atomic reserves that typical * GFP_ATOMIC would do. irq_work runs on this cpu and kmalloc * will allocate from the current numa node which is what we * want here. */ obj = __alloc(c, node, gfp); if (!obj) break; add_obj_to_free_list(c, obj); } set_active_memcg(old_memcg); mem_cgroup_put(memcg);} static void free_one(void *obj, bool percpu){ if (percpu) { free_percpu(((void __percpu **)obj)[1]); kfree(obj); return; } kfree(obj);} static int free_all(struct llist_node *llnode, bool percpu){ struct llist_node *pos, *t; int cnt = 0; llist_for_each_safe(pos, t, llnode) { free_one(pos, percpu); cnt++; } return cnt;} static void __free_rcu(struct rcu_head *head){ struct bpf_mem_cache *c = container_of(head, struct bpf_mem_cache, rcu_ttrace); free_all(llist_del_all(&c->waiting_for_gp_ttrace), !!c->percpu_size); atomic_set(&c->call_rcu_ttrace_in_progress, 0);} static void __free_rcu_tasks_trace(struct rcu_head *head){ /* If RCU Tasks Trace grace period implies RCU grace period, * there is no need to invoke call_rcu(). */ if (rcu_trace_implies_rcu_gp()) __free_rcu(head); else call_rcu(head, __free_rcu);} static void enque_to_free(struct bpf_mem_cache *c, void *obj){ struct llist_node *llnode = obj; /* bpf_mem_cache is a per-cpu object. Freeing happens in irq_work. * Nothing races to add to free_by_rcu_ttrace list. */ llist_add(llnode, &c->free_by_rcu_ttrace);} static void do_call_rcu_ttrace(struct bpf_mem_cache *c){ struct llist_node *llnode, *t; if (atomic_xchg(&c->call_rcu_ttrace_in_progress, 1)) { if (unlikely(READ_ONCE(c->draining))) { llnode = llist_del_all(&c->free_by_rcu_ttrace); free_all(llnode, !!c->percpu_size); } return; } WARN_ON_ONCE(!llist_empty(&c->waiting_for_gp_ttrace)); llist_for_each_safe(llnode, t, llist_del_all(&c->free_by_rcu_ttrace)) llist_add(llnode, &c->waiting_for_gp_ttrace); if (unlikely(READ_ONCE(c->draining))) { __free_rcu(&c->rcu_ttrace); return; } /* Use call_rcu_tasks_trace() to wait for sleepable progs to finish. * If RCU Tasks Trace grace period implies RCU grace period, free * these elements directly, else use call_rcu() to wait for normal * progs to finish and finally do free_one() on each element. */ call_rcu_tasks_trace(&c->rcu_ttrace, __free_rcu_tasks_trace);} static void free_bulk(struct bpf_mem_cache *c){ struct bpf_mem_cache *tgt = c->tgt; struct llist_node *llnode, *t; unsigned long flags; int cnt; WARN_ON_ONCE(tgt->unit_size != c->unit_size); WARN_ON_ONCE(tgt->percpu_size != c->percpu_size); do { inc_active(c, &flags); llnode = __llist_del_first(&c->free_llist); if (llnode) cnt = --c->free_cnt; else cnt = 0; dec_active(c, &flags); if (llnode) enque_to_free(tgt, llnode); } while (cnt > (c->high_watermark + c->low_watermark) / 2); /* and drain free_llist_extra */ llist_for_each_safe(llnode, t, llist_del_all(&c->free_llist_extra)) enque_to_free(tgt, llnode); do_call_rcu_ttrace(tgt);} static void __free_by_rcu(struct rcu_head *head){ struct bpf_mem_cache *c = container_of(head, struct bpf_mem_cache, rcu); struct bpf_mem_cache *tgt = c->tgt; struct llist_node *llnode; WARN_ON_ONCE(tgt->unit_size != c->unit_size); WARN_ON_ONCE(tgt->percpu_size != c->percpu_size); llnode = llist_del_all(&c->waiting_for_gp); if (!llnode) goto out; llist_add_batch(llnode, c->waiting_for_gp_tail, &tgt->free_by_rcu_ttrace); /* Objects went through regular RCU GP. Send them to RCU tasks trace */ do_call_rcu_ttrace(tgt);out: atomic_set(&c->call_rcu_in_progress, 0);} static void check_free_by_rcu(struct bpf_mem_cache *c){ struct llist_node *llnode, *t; unsigned long flags; /* drain free_llist_extra_rcu */ if (unlikely(!llist_empty(&c->free_llist_extra_rcu))) { inc_active(c, &flags); llist_for_each_safe(llnode, t, llist_del_all(&c->free_llist_extra_rcu)) if (__llist_add(llnode, &c->free_by_rcu)) c->free_by_rcu_tail = llnode; dec_active(c, &flags); } if (llist_empty(&c->free_by_rcu)) return; if (atomic_xchg(&c->call_rcu_in_progress, 1)) { /* * Instead of kmalloc-ing new rcu_head and triggering 10k * call_rcu() to hit rcutree.qhimark and force RCU to notice * the overload just ask RCU to hurry up. There could be many * objects in free_by_rcu list. * This hint reduces memory consumption for an artificial * benchmark from 2 Gbyte to 150 Mbyte. */ rcu_request_urgent_qs_task(current); return; } WARN_ON_ONCE(!llist_empty(&c->waiting_for_gp)); inc_active(c, &flags); WRITE_ONCE(c->waiting_for_gp.first, __llist_del_all(&c->free_by_rcu)); c->waiting_for_gp_tail = c->free_by_rcu_tail; dec_active(c, &flags); if (unlikely(READ_ONCE(c->draining))) { free_all(llist_del_all(&c->waiting_for_gp), !!c->percpu_size); atomic_set(&c->call_rcu_in_progress, 0); } else { call_rcu_hurry(&c->rcu, __free_by_rcu); }} static void bpf_mem_refill(struct irq_work *work){ struct bpf_mem_cache *c = container_of(work, struct bpf_mem_cache, refill_work); int cnt; /* Racy access to free_cnt. It doesn't need to be 100% accurate */ cnt = c->free_cnt; if (cnt < c->low_watermark) /* irq_work runs on this cpu and kmalloc will allocate * from the current numa node which is what we want here. */ alloc_bulk(c, c->batch, NUMA_NO_NODE, true); else if (cnt > c->high_watermark) free_bulk(c); check_free_by_rcu(c);} static void notrace irq_work_raise(struct bpf_mem_cache *c){ irq_work_queue(&c->refill_work);} /* For typical bpf map case that uses bpf_mem_cache_alloc and single bucket * the freelist cache will be elem_size * 64 (or less) on each cpu. * * For bpf programs that don't have statically known allocation sizes and * assuming (low_mark + high_mark) / 2 as an average number of elements per * bucket and all buckets are used the total amount of memory in freelists * on each cpu will be: * 64*16 + 64*32 + 64*64 + 64*96 + 64*128 + 64*196 + 64*256 + 32*512 + 16*1024 + 8*2048 + 4*4096 * == ~ 116 Kbyte using below heuristic. * Initialized, but unused bpf allocator (not bpf map specific one) will * consume ~ 11 Kbyte per cpu. * Typical case will be between 11K and 116K closer to 11K. * bpf progs can and should share bpf_mem_cache when possible. * * Percpu allocation is typically rare. To avoid potential unnecessary large * memory consumption, set low_mark = 1 and high_mark = 3, resulting in c->batch = 1. */static void init_refill_work(struct bpf_mem_cache *c){ init_irq_work(&c->refill_work, bpf_mem_refill); if (c->percpu_size) { c->low_watermark = 1; c->high_watermark = 3; } else if (c->unit_size <= 256) { c->low_watermark = 32; c->high_watermark = 96; } else { /* When page_size == 4k, order-0 cache will have low_mark == 2 * and high_mark == 6 with batch alloc of 3 individual pages at * a time. * 8k allocs and above low == 1, high == 3, batch == 1. */ c->low_watermark = max(32 * 256 / c->unit_size, 1); c->high_watermark = max(96 * 256 / c->unit_size, 3); } c->batch = max((c->high_watermark - c->low_watermark) / 4 * 3, 1);} static void prefill_mem_cache(struct bpf_mem_cache *c, int cpu){ int cnt = 1; /* To avoid consuming memory, for non-percpu allocation, assume that * 1st run of bpf prog won't be doing more than 4 map_update_elem from * irq disabled region if unit size is less than or equal to 256. * For all other cases, let us just do one allocation. */ if (!c->percpu_size && c->unit_size <= 256)Lines 1–500 of 1019. Load more
// SPDX-License-Identifier: GPL-2.0-only /* Copyright (c) 2022 Meta Platforms, Inc. and affiliates. */ #include <linux/mm.h> #include <linux/llist.h> #include <linux/bpf.h> #include <linux/irq_work.h> #include <linux/bpf_mem_alloc.h> #include <linux/memcontrol.h> #include <asm/local.h> /* Any context (including NMI) BPF specific memory allocator. * * Tracing BPF programs can attach to kprobe and fentry. Hence they * run in unknown context where calling plain kmalloc() might not be safe. * * Front-end kmalloc() with per-cpu per-bucket cache of free elements. * Refill this cache asynchronously from irq_work. * * CPU_0 buckets * 16 32 64 96 128 196 256 512 1024 2048 4096 * ... * CPU_N buckets * 16 32 64 96 128 196 256 512 1024 2048 4096 * * The buckets are prefilled at the start. * BPF programs always run with migration disabled. * It's safe to allocate from cache of the current cpu with irqs disabled. * Free-ing is always done into bucket of the current cpu as well. * irq_work trims extra free elements from buckets with kfree * and refills them with kmalloc, so global kmalloc logic takes care * of freeing objects allocated by one cpu and freed on another. * * Every allocated objected is padded with extra 8 bytes that contains * struct llist_node. */ #define LLIST_NODE_SZ sizeof(struct llist_node) #define BPF_MEM_ALLOC_SIZE_MAX 4096 /* similar to kmalloc, but sizeof == 8 bucket is gone */ static u8 size_index[24] __ro_after_init = { 3, /* 8 */ 3, /* 16 */ 4, /* 24 */ 4, /* 32 */ 5, /* 40 */ 5, /* 48 */ 5, /* 56 */ 5, /* 64 */ 1, /* 72 */ 1, /* 80 */ 1, /* 88 */ 1, /* 96 */ 6, /* 104 */ 6, /* 112 */ 6, /* 120 */ 6, /* 128 */ 2, /* 136 */ 2, /* 144 */ 2, /* 152 */ 2, /* 160 */ 2, /* 168 */ 2, /* 176 */ 2, /* 184 */ 2 /* 192 */ }; static int bpf_mem_cache_idx(size_t size) { if (!size || size > BPF_MEM_ALLOC_SIZE_MAX) return -1; if (size <= 192) return size_index[(size - 1) / 8] - 1; return fls(size - 1) - 2; } #define NUM_CACHES 11 struct bpf_mem_cache { /* per-cpu list of free objects of size 'unit_size'. * All accesses are done with interrupts disabled and 'active' counter * protection with __llist_add() and __llist_del_first(). */ struct llist_head free_llist; local_t active; /* Operations on the free_list from unit_alloc/unit_free/bpf_mem_refill * are sequenced by per-cpu 'active' counter. But unit_free() cannot * fail. When 'active' is busy the unit_free() will add an object to * free_llist_extra. */ struct llist_head free_llist_extra; struct irq_work refill_work; struct obj_cgroup *objcg; int unit_size; /* count of objects in free_llist */ int free_cnt; int low_watermark, high_watermark, batch; int percpu_size; bool draining; struct bpf_mem_cache *tgt; /* list of objects to be freed after RCU GP */ struct llist_head free_by_rcu; struct llist_node *free_by_rcu_tail; struct llist_head waiting_for_gp; struct llist_node *waiting_for_gp_tail; struct rcu_head rcu; atomic_t call_rcu_in_progress; struct llist_head free_llist_extra_rcu; /* list of objects to be freed after RCU tasks trace GP */ struct llist_head free_by_rcu_ttrace; struct llist_head waiting_for_gp_ttrace; struct rcu_head rcu_ttrace; atomic_t call_rcu_ttrace_in_progress; }; struct bpf_mem_caches { struct bpf_mem_cache cache[NUM_CACHES]; }; static const u16 sizes[NUM_CACHES] = {96, 192, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096}; static struct llist_node notrace *__llist_del_first(struct llist_head *head) { struct llist_node *entry, *next; entry = head->first; if (!entry) return NULL; next = entry->next; head->first = next; return entry; } static void *__alloc(struct bpf_mem_cache *c, int node, gfp_t flags) { if (c->percpu_size) { void __percpu **obj = kmalloc_node(c->percpu_size, flags, node); void __percpu *pptr = __alloc_percpu_gfp(c->unit_size, 8, flags); if (!obj || !pptr) { free_percpu(pptr); kfree(obj); return NULL; } obj[1] = pptr; return obj; } return kmalloc_node(c->unit_size, flags | __GFP_ZERO, node); } static struct mem_cgroup *get_memcg(const struct bpf_mem_cache *c) { #ifdef CONFIG_MEMCG if (c->objcg) return get_mem_cgroup_from_objcg(c->objcg); return root_mem_cgroup; #else return NULL; #endif } static void inc_active(struct bpf_mem_cache *c, unsigned long *flags) { if (IS_ENABLED(CONFIG_PREEMPT_RT)) /* In RT irq_work runs in per-cpu kthread, so disable * interrupts to avoid preemption and interrupts and * reduce the chance of bpf prog executing on this cpu * when active counter is busy. */ local_irq_save(*flags); /* alloc_bulk runs from irq_work which will not preempt a bpf * program that does unit_alloc/unit_free since IRQs are * disabled there. There is no race to increment 'active' * counter. It protects free_llist from corruption in case NMI * bpf prog preempted this loop. */ WARN_ON_ONCE(local_inc_return(&c->active) != 1); } static void dec_active(struct bpf_mem_cache *c, unsigned long *flags) { local_dec(&c->active); if (IS_ENABLED(CONFIG_PREEMPT_RT)) local_irq_restore(*flags); } static void add_obj_to_free_list(struct bpf_mem_cache *c, void *obj) { unsigned long flags; inc_active(c, &flags); __llist_add(obj, &c->free_llist); c->free_cnt++; dec_active(c, &flags); } /* Mostly runs from irq_work except __init phase. */ static void alloc_bulk(struct bpf_mem_cache *c, int cnt, int node, bool atomic) { struct mem_cgroup *memcg = NULL, *old_memcg; gfp_t gfp; void *obj; int i; gfp = __GFP_NOWARN | __GFP_ACCOUNT; gfp |= atomic ? GFP_NOWAIT : GFP_KERNEL; for (i = 0; i < cnt; i++) { /* * For every 'c' llist_del_first(&c->free_by_rcu_ttrace); is * done only by one CPU == current CPU. Other CPUs might * llist_add() and llist_del_all() in parallel. */ obj = llist_del_first(&c->free_by_rcu_ttrace); if (!obj) break; add_obj_to_free_list(c, obj); } if (i >= cnt) return; for (; i < cnt; i++) { obj = llist_del_first(&c->waiting_for_gp_ttrace); if (!obj) break; add_obj_to_free_list(c, obj); } if (i >= cnt) return; memcg = get_memcg(c); old_memcg = set_active_memcg(memcg); for (; i < cnt; i++) { /* Allocate, but don't deplete atomic reserves that typical * GFP_ATOMIC would do. irq_work runs on this cpu and kmalloc * will allocate from the current numa node which is what we * want here. */ obj = __alloc(c, node, gfp); if (!obj) break; add_obj_to_free_list(c, obj); } set_active_memcg(old_memcg); mem_cgroup_put(memcg); } static void free_one(void *obj, bool percpu) { if (percpu) { free_percpu(((void __percpu **)obj)[1]); kfree(obj); return; } kfree(obj); } static int free_all(struct llist_node *llnode, bool percpu) { struct llist_node *pos, *t; int cnt = 0; llist_for_each_safe(pos, t, llnode) { free_one(pos, percpu); cnt++; } return cnt; } static void __free_rcu(struct rcu_head *head) { struct bpf_mem_cache *c = container_of(head, struct bpf_mem_cache, rcu_ttrace); free_all(llist_del_all(&c->waiting_for_gp_ttrace), !!c->percpu_size); atomic_set(&c->call_rcu_ttrace_in_progress, 0); } static void __free_rcu_tasks_trace(struct rcu_head *head) { /* If RCU Tasks Trace grace period implies RCU grace period, * there is no need to invoke call_rcu(). */ if (rcu_trace_implies_rcu_gp()) __free_rcu(head); else call_rcu(head, __free_rcu); } static void enque_to_free(struct bpf_mem_cache *c, void *obj) { struct llist_node *llnode = obj; /* bpf_mem_cache is a per-cpu object. Freeing happens in irq_work. * Nothing races to add to free_by_rcu_ttrace list. */ llist_add(llnode, &c->free_by_rcu_ttrace); } static void do_call_rcu_ttrace(struct bpf_mem_cache *c) { struct llist_node *llnode, *t; if (atomic_xchg(&c->call_rcu_ttrace_in_progress, 1)) { if (unlikely(READ_ONCE(c->draining))) { llnode = llist_del_all(&c->free_by_rcu_ttrace); free_all(llnode, !!c->percpu_size); } return; } WARN_ON_ONCE(!llist_empty(&c->waiting_for_gp_ttrace)); llist_for_each_safe(llnode, t, llist_del_all(&c->free_by_rcu_ttrace)) llist_add(llnode, &c->waiting_for_gp_ttrace); if (unlikely(READ_ONCE(c->draining))) { __free_rcu(&c->rcu_ttrace); return; } /* Use call_rcu_tasks_trace() to wait for sleepable progs to finish. * If RCU Tasks Trace grace period implies RCU grace period, free * these elements directly, else use call_rcu() to wait for normal * progs to finish and finally do free_one() on each element. */ call_rcu_tasks_trace(&c->rcu_ttrace, __free_rcu_tasks_trace); } static void free_bulk(struct bpf_mem_cache *c) { struct bpf_mem_cache *tgt = c->tgt; struct llist_node *llnode, *t; unsigned long flags; int cnt; WARN_ON_ONCE(tgt->unit_size != c->unit_size); WARN_ON_ONCE(tgt->percpu_size != c->percpu_size); do { inc_active(c, &flags); llnode = __llist_del_first(&c->free_llist); if (llnode) cnt = --c->free_cnt; else cnt = 0; dec_active(c, &flags); if (llnode) enque_to_free(tgt, llnode); } while (cnt > (c->high_watermark + c->low_watermark) / 2); /* and drain free_llist_extra */ llist_for_each_safe(llnode, t, llist_del_all(&c->free_llist_extra)) enque_to_free(tgt, llnode); do_call_rcu_ttrace(tgt); } static void __free_by_rcu(struct rcu_head *head) { struct bpf_mem_cache *c = container_of(head, struct bpf_mem_cache, rcu); struct bpf_mem_cache *tgt = c->tgt; struct llist_node *llnode; WARN_ON_ONCE(tgt->unit_size != c->unit_size); WARN_ON_ONCE(tgt->percpu_size != c->percpu_size); llnode = llist_del_all(&c->waiting_for_gp); if (!llnode) goto out; llist_add_batch(llnode, c->waiting_for_gp_tail, &tgt->free_by_rcu_ttrace); /* Objects went through regular RCU GP. Send them to RCU tasks trace */ do_call_rcu_ttrace(tgt); out: atomic_set(&c->call_rcu_in_progress, 0); } static void check_free_by_rcu(struct bpf_mem_cache *c) { struct llist_node *llnode, *t; unsigned long flags; /* drain free_llist_extra_rcu */ if (unlikely(!llist_empty(&c->free_llist_extra_rcu))) { inc_active(c, &flags); llist_for_each_safe(llnode, t, llist_del_all(&c->free_llist_extra_rcu)) if (__llist_add(llnode, &c->free_by_rcu)) c->free_by_rcu_tail = llnode; dec_active(c, &flags); } if (llist_empty(&c->free_by_rcu)) return; if (atomic_xchg(&c->call_rcu_in_progress, 1)) { /* * Instead of kmalloc-ing new rcu_head and triggering 10k * call_rcu() to hit rcutree.qhimark and force RCU to notice * the overload just ask RCU to hurry up. There could be many * objects in free_by_rcu list. * This hint reduces memory consumption for an artificial * benchmark from 2 Gbyte to 150 Mbyte. */ rcu_request_urgent_qs_task(current); return; } WARN_ON_ONCE(!llist_empty(&c->waiting_for_gp)); inc_active(c, &flags); WRITE_ONCE(c->waiting_for_gp.first, __llist_del_all(&c->free_by_rcu)); c->waiting_for_gp_tail = c->free_by_rcu_tail; dec_active(c, &flags); if (unlikely(READ_ONCE(c->draining))) { free_all(llist_del_all(&c->waiting_for_gp), !!c->percpu_size); atomic_set(&c->call_rcu_in_progress, 0); } else { call_rcu_hurry(&c->rcu, __free_by_rcu); } } static void bpf_mem_refill(struct irq_work *work) { struct bpf_mem_cache *c = container_of(work, struct bpf_mem_cache, refill_work); int cnt; /* Racy access to free_cnt. It doesn't need to be 100% accurate */ cnt = c->free_cnt; if (cnt < c->low_watermark) /* irq_work runs on this cpu and kmalloc will allocate * from the current numa node which is what we want here. */ alloc_bulk(c, c->batch, NUMA_NO_NODE, true); else if (cnt > c->high_watermark) free_bulk(c); check_free_by_rcu(c); } static void notrace irq_work_raise(struct bpf_mem_cache *c) { irq_work_queue(&c->refill_work); } /* For typical bpf map case that uses bpf_mem_cache_alloc and single bucket * the freelist cache will be elem_size * 64 (or less) on each cpu. * * For bpf programs that don't have statically known allocation sizes and * assuming (low_mark + high_mark) / 2 as an average number of elements per * bucket and all buckets are used the total amount of memory in freelists * on each cpu will be: * 64*16 + 64*32 + 64*64 + 64*96 + 64*128 + 64*196 + 64*256 + 32*512 + 16*1024 + 8*2048 + 4*4096 * == ~ 116 Kbyte using below heuristic. * Initialized, but unused bpf allocator (not bpf map specific one) will * consume ~ 11 Kbyte per cpu. * Typical case will be between 11K and 116K closer to 11K. * bpf progs can and should share bpf_mem_cache when possible. * * Percpu allocation is typically rare. To avoid potential unnecessary large * memory consumption, set low_mark = 1 and high_mark = 3, resulting in c->batch = 1. */ static void init_refill_work(struct bpf_mem_cache *c) { init_irq_work(&c->refill_work, bpf_mem_refill); if (c->percpu_size) { c->low_watermark = 1; c->high_watermark = 3; } else if (c->unit_size <= 256) { c->low_watermark = 32; c->high_watermark = 96; } else { /* When page_size == 4k, order-0 cache will have low_mark == 2 * and high_mark == 6 with batch alloc of 3 individual pages at * a time. * 8k allocs and above low == 1, high == 3, batch == 1. */ c->low_watermark = max(32 * 256 / c->unit_size, 1); c->high_watermark = max(96 * 256 / c->unit_size, 3); } c->batch = max((c->high_watermark - c->low_watermark) / 4 * 3, 1); } static void prefill_mem_cache(struct bpf_mem_cache *c, int cpu) { int cnt = 1; /* To avoid consuming memory, for non-percpu allocation, assume that * 1st run of bpf prog won't be doing more than 4 map_update_elem from * irq disabled region if unit size is less than or equal to 256. * For all other cases, let us just do one allocation. */ if (!c->percpu_size && c->unit_size <= 256)