diff options
Diffstat (limited to 'net/core/skbuff.c')
-rw-r--r-- | net/core/skbuff.c | 237 |
1 files changed, 170 insertions, 67 deletions
diff --git a/net/core/skbuff.c b/net/core/skbuff.c index a31ff4d83ecc..eb7d33b41e71 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -79,15 +79,44 @@ #include <linux/capability.h> #include <linux/user_namespace.h> #include <linux/indirect_call_wrapper.h> +#include <linux/textsearch.h> #include "dev.h" #include "sock_destructor.h" -struct kmem_cache *skbuff_head_cache __ro_after_init; +struct kmem_cache *skbuff_cache __ro_after_init; static struct kmem_cache *skbuff_fclone_cache __ro_after_init; #ifdef CONFIG_SKB_EXTENSIONS static struct kmem_cache *skbuff_ext_cache __ro_after_init; #endif + +/* skb_small_head_cache and related code is only supported + * for CONFIG_SLAB and CONFIG_SLUB. + * As soon as SLOB is removed from the kernel, we can clean up this. + */ +#if !defined(CONFIG_SLOB) +# define HAVE_SKB_SMALL_HEAD_CACHE 1 +#endif + +#ifdef HAVE_SKB_SMALL_HEAD_CACHE +static struct kmem_cache *skb_small_head_cache __ro_after_init; + +#define SKB_SMALL_HEAD_SIZE SKB_HEAD_ALIGN(MAX_TCP_HEADER) + +/* We want SKB_SMALL_HEAD_CACHE_SIZE to not be a power of two. + * This should ensure that SKB_SMALL_HEAD_HEADROOM is a unique + * size, and we can differentiate heads from skb_small_head_cache + * vs system slabs by looking at their size (skb_end_offset()). + */ +#define SKB_SMALL_HEAD_CACHE_SIZE \ + (is_power_of_2(SKB_SMALL_HEAD_SIZE) ? \ + (SKB_SMALL_HEAD_SIZE + L1_CACHE_BYTES) : \ + SKB_SMALL_HEAD_SIZE) + +#define SKB_SMALL_HEAD_HEADROOM \ + SKB_WITH_OVERHEAD(SKB_SMALL_HEAD_CACHE_SIZE) +#endif /* HAVE_SKB_SMALL_HEAD_CACHE */ + int sysctl_max_skb_frags __read_mostly = MAX_SKB_FRAGS; EXPORT_SYMBOL(sysctl_max_skb_frags); @@ -256,7 +285,7 @@ static struct sk_buff *napi_skb_cache_get(void) struct sk_buff *skb; if (unlikely(!nc->skb_count)) { - nc->skb_count = kmem_cache_alloc_bulk(skbuff_head_cache, + nc->skb_count = kmem_cache_alloc_bulk(skbuff_cache, GFP_ATOMIC, NAPI_SKB_CACHE_BULK, nc->skb_cache); @@ -265,7 +294,7 @@ static struct sk_buff *napi_skb_cache_get(void) } skb = nc->skb_cache[--nc->skb_count]; - kasan_unpoison_object_data(skbuff_head_cache, skb); + kasan_unpoison_object_data(skbuff_cache, skb); return skb; } @@ -323,7 +352,7 @@ struct sk_buff *slab_build_skb(void *data) struct sk_buff *skb; unsigned int size; - skb = kmem_cache_alloc(skbuff_head_cache, GFP_ATOMIC); + skb = kmem_cache_alloc(skbuff_cache, GFP_ATOMIC); if (unlikely(!skb)) return NULL; @@ -374,7 +403,7 @@ struct sk_buff *__build_skb(void *data, unsigned int frag_size) { struct sk_buff *skb; - skb = kmem_cache_alloc(skbuff_head_cache, GFP_ATOMIC); + skb = kmem_cache_alloc(skbuff_cache, GFP_ATOMIC); if (unlikely(!skb)) return NULL; @@ -386,8 +415,6 @@ struct sk_buff *__build_skb(void *data, unsigned int frag_size) /* build_skb() is wrapper over __build_skb(), that specifically * takes care of skb->head and skb->pfmemalloc - * This means that if @frag_size is not zero, then @data must be backed - * by a page fragment, not kmalloc() or vmalloc() */ struct sk_buff *build_skb(void *data, unsigned int frag_size) { @@ -406,7 +433,7 @@ EXPORT_SYMBOL(build_skb); * build_skb_around - build a network buffer around provided skb * @skb: sk_buff provide by caller, must be memset cleared * @data: data buffer provided by caller - * @frag_size: size of data, or 0 if head was kmalloced + * @frag_size: size of data */ struct sk_buff *build_skb_around(struct sk_buff *skb, void *data, unsigned int frag_size) @@ -428,7 +455,7 @@ EXPORT_SYMBOL(build_skb_around); /** * __napi_build_skb - build a network buffer * @data: data buffer provided by caller - * @frag_size: size of data, or 0 if head was kmalloced + * @frag_size: size of data * * Version of __build_skb() that uses NAPI percpu caches to obtain * skbuff_head instead of inplace allocation. @@ -452,7 +479,7 @@ static struct sk_buff *__napi_build_skb(void *data, unsigned int frag_size) /** * napi_build_skb - build a network buffer * @data: data buffer provided by caller - * @frag_size: size of data, or 0 if head was kmalloced + * @frag_size: size of data * * Version of __napi_build_skb() that takes care of skb->head_frag * and skb->pfmemalloc when the data is a page or page fragment. @@ -479,17 +506,37 @@ EXPORT_SYMBOL(napi_build_skb); * may be used. Otherwise, the packet data may be discarded until enough * memory is free */ -static void *kmalloc_reserve(size_t size, gfp_t flags, int node, +static void *kmalloc_reserve(unsigned int *size, gfp_t flags, int node, bool *pfmemalloc) { - void *obj; bool ret_pfmemalloc = false; + unsigned int obj_size; + void *obj; + + obj_size = SKB_HEAD_ALIGN(*size); +#ifdef HAVE_SKB_SMALL_HEAD_CACHE + if (obj_size <= SKB_SMALL_HEAD_CACHE_SIZE && + !(flags & KMALLOC_NOT_NORMAL_BITS)) { + /* skb_small_head_cache has non power of two size, + * likely forcing SLUB to use order-3 pages. + * We deliberately attempt a NOMEMALLOC allocation only. + */ + obj = kmem_cache_alloc_node(skb_small_head_cache, + flags | __GFP_NOMEMALLOC | __GFP_NOWARN, + node); + if (obj) { + *size = SKB_SMALL_HEAD_CACHE_SIZE; + goto out; + } + } +#endif + *size = obj_size = kmalloc_size_roundup(obj_size); /* * Try a regular allocation, when that fails and we're not entitled * to the reserves, fail. */ - obj = kmalloc_node_track_caller(size, + obj = kmalloc_node_track_caller(obj_size, flags | __GFP_NOMEMALLOC | __GFP_NOWARN, node); if (obj || !(gfp_pfmemalloc_allowed(flags))) @@ -497,7 +544,7 @@ static void *kmalloc_reserve(size_t size, gfp_t flags, int node, /* Try again but now we are using pfmemalloc reserves */ ret_pfmemalloc = true; - obj = kmalloc_node_track_caller(size, flags, node); + obj = kmalloc_node_track_caller(obj_size, flags, node); out: if (pfmemalloc) @@ -534,12 +581,11 @@ struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask, { struct kmem_cache *cache; struct sk_buff *skb; - unsigned int osize; bool pfmemalloc; u8 *data; cache = (flags & SKB_ALLOC_FCLONE) - ? skbuff_fclone_cache : skbuff_head_cache; + ? skbuff_fclone_cache : skbuff_cache; if (sk_memalloc_socks() && (flags & SKB_ALLOC_RX)) gfp_mask |= __GFP_MEMALLOC; @@ -559,18 +605,14 @@ struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask, * aligned memory blocks, unless SLUB/SLAB debug is enabled. * Both skb->head and skb_shared_info are cache line aligned. */ - size = SKB_DATA_ALIGN(size); - size += SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); - osize = kmalloc_size_roundup(size); - data = kmalloc_reserve(osize, gfp_mask, node, &pfmemalloc); + data = kmalloc_reserve(&size, gfp_mask, node, &pfmemalloc); if (unlikely(!data)) goto nodata; /* kmalloc_size_roundup() might give us more room than requested. * Put skb_shared_info exactly at the end of allocated zone, * to allow max possible filling before reallocation. */ - size = SKB_WITH_OVERHEAD(osize); - prefetchw(data + size); + prefetchw(data + SKB_WITH_OVERHEAD(size)); /* * Only clear those fields we need to clear, not those that we will @@ -578,7 +620,7 @@ struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask, * the tail pointer in struct sk_buff! */ memset(skb, 0, offsetof(struct sk_buff, tail)); - __build_skb_around(skb, data, osize); + __build_skb_around(skb, data, size); skb->pfmemalloc = pfmemalloc; if (flags & SKB_ALLOC_FCLONE) { @@ -633,8 +675,7 @@ struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int len, goto skb_success; } - len += SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); - len = SKB_DATA_ALIGN(len); + len = SKB_HEAD_ALIGN(len); if (sk_memalloc_socks()) gfp_mask |= __GFP_MEMALLOC; @@ -733,8 +774,7 @@ struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, unsigned int len, data = page_frag_alloc_1k(&nc->page_small, gfp_mask); pfmemalloc = NAPI_SMALL_PAGE_PFMEMALLOC(nc->page_small); } else { - len += SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); - len = SKB_DATA_ALIGN(len); + len = SKB_HEAD_ALIGN(len); data = page_frag_alloc(&nc->page, len, gfp_mask); pfmemalloc = nc->page.pfmemalloc; @@ -810,6 +850,16 @@ static bool skb_pp_recycle(struct sk_buff *skb, void *data) return page_pool_return_skb_page(virt_to_page(data)); } +static void skb_kfree_head(void *head, unsigned int end_offset) +{ +#ifdef HAVE_SKB_SMALL_HEAD_CACHE + if (end_offset == SKB_SMALL_HEAD_HEADROOM) + kmem_cache_free(skb_small_head_cache, head); + else +#endif + kfree(head); +} + static void skb_free_head(struct sk_buff *skb) { unsigned char *head = skb->head; @@ -819,7 +869,7 @@ static void skb_free_head(struct sk_buff *skb) return; skb_free_frag(head); } else { - kfree(head); + skb_kfree_head(head, skb_end_offset(skb)); } } @@ -871,7 +921,7 @@ static void kfree_skbmem(struct sk_buff *skb) switch (skb->fclone) { case SKB_FCLONE_UNAVAILABLE: - kmem_cache_free(skbuff_head_cache, skb); + kmem_cache_free(skbuff_cache, skb); return; case SKB_FCLONE_ORIG: @@ -932,6 +982,21 @@ void __kfree_skb(struct sk_buff *skb) } EXPORT_SYMBOL(__kfree_skb); +static __always_inline +bool __kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason) +{ + if (unlikely(!skb_unref(skb))) + return false; + + DEBUG_NET_WARN_ON_ONCE(reason <= 0 || reason >= SKB_DROP_REASON_MAX); + + if (reason == SKB_CONSUMED) + trace_consume_skb(skb, __builtin_return_address(0)); + else + trace_kfree_skb(skb, __builtin_return_address(0), reason); + return true; +} + /** * kfree_skb_reason - free an sk_buff with special reason * @skb: buffer to free @@ -944,28 +1009,58 @@ EXPORT_SYMBOL(__kfree_skb); void __fix_address kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason) { - if (unlikely(!skb_unref(skb))) + if (__kfree_skb_reason(skb, reason)) + __kfree_skb(skb); +} +EXPORT_SYMBOL(kfree_skb_reason); + +#define KFREE_SKB_BULK_SIZE 16 + +struct skb_free_array { + unsigned int skb_count; + void *skb_array[KFREE_SKB_BULK_SIZE]; +}; + +static void kfree_skb_add_bulk(struct sk_buff *skb, + struct skb_free_array *sa, + enum skb_drop_reason reason) +{ + /* if SKB is a clone, don't handle this case */ + if (unlikely(skb->fclone != SKB_FCLONE_UNAVAILABLE)) { + __kfree_skb(skb); return; + } - DEBUG_NET_WARN_ON_ONCE(reason <= 0 || reason >= SKB_DROP_REASON_MAX); + skb_release_all(skb, reason); + sa->skb_array[sa->skb_count++] = skb; - if (reason == SKB_CONSUMED) - trace_consume_skb(skb); - else - trace_kfree_skb(skb, __builtin_return_address(0), reason); - __kfree_skb(skb); + if (unlikely(sa->skb_count == KFREE_SKB_BULK_SIZE)) { + kmem_cache_free_bulk(skbuff_cache, KFREE_SKB_BULK_SIZE, + sa->skb_array); + sa->skb_count = 0; + } } -EXPORT_SYMBOL(kfree_skb_reason); -void kfree_skb_list_reason(struct sk_buff *segs, - enum skb_drop_reason reason) +void __fix_address +kfree_skb_list_reason(struct sk_buff *segs, enum skb_drop_reason reason) { + struct skb_free_array sa; + + sa.skb_count = 0; + while (segs) { struct sk_buff *next = segs->next; - kfree_skb_reason(segs, reason); + if (__kfree_skb_reason(segs, reason)) { + skb_poison_list(segs); + kfree_skb_add_bulk(segs, &sa, reason); + } + segs = next; } + + if (sa.skb_count) + kmem_cache_free_bulk(skbuff_cache, sa.skb_count, sa.skb_array); } EXPORT_SYMBOL(kfree_skb_list_reason); @@ -1094,7 +1189,7 @@ void consume_skb(struct sk_buff *skb) if (!skb_unref(skb)) return; - trace_consume_skb(skb); + trace_consume_skb(skb, __builtin_return_address(0)); __kfree_skb(skb); } EXPORT_SYMBOL(consume_skb); @@ -1109,7 +1204,7 @@ EXPORT_SYMBOL(consume_skb); */ void __consume_stateless_skb(struct sk_buff *skb) { - trace_consume_skb(skb); + trace_consume_skb(skb, __builtin_return_address(0)); skb_release_data(skb, SKB_CONSUMED); kfree_skbmem(skb); } @@ -1119,15 +1214,15 @@ static void napi_skb_cache_put(struct sk_buff *skb) struct napi_alloc_cache *nc = this_cpu_ptr(&napi_alloc_cache); u32 i; - kasan_poison_object_data(skbuff_head_cache, skb); + kasan_poison_object_data(skbuff_cache, skb); nc->skb_cache[nc->skb_count++] = skb; if (unlikely(nc->skb_count == NAPI_SKB_CACHE_SIZE)) { for (i = NAPI_SKB_CACHE_HALF; i < NAPI_SKB_CACHE_SIZE; i++) - kasan_unpoison_object_data(skbuff_head_cache, + kasan_unpoison_object_data(skbuff_cache, nc->skb_cache[i]); - kmem_cache_free_bulk(skbuff_head_cache, NAPI_SKB_CACHE_HALF, + kmem_cache_free_bulk(skbuff_cache, NAPI_SKB_CACHE_HALF, nc->skb_cache + NAPI_SKB_CACHE_HALF); nc->skb_count = NAPI_SKB_CACHE_HALF; } @@ -1165,7 +1260,7 @@ void napi_consume_skb(struct sk_buff *skb, int budget) return; /* if reaching here SKB is ready to free */ - trace_consume_skb(skb); + trace_consume_skb(skb, __builtin_return_address(0)); /* if SKB is a clone, don't handle this case */ if (skb->fclone != SKB_FCLONE_UNAVAILABLE) { @@ -1311,14 +1406,18 @@ EXPORT_SYMBOL_GPL(skb_morph); int mm_account_pinned_pages(struct mmpin *mmp, size_t size) { - unsigned long max_pg, num_pg, new_pg, old_pg; + unsigned long max_pg, num_pg, new_pg, old_pg, rlim; struct user_struct *user; if (capable(CAP_IPC_LOCK) || !size) return 0; + rlim = rlimit(RLIMIT_MEMLOCK); + if (rlim == RLIM_INFINITY) + return 0; + num_pg = (size >> PAGE_SHIFT) + 2; /* worst case */ - max_pg = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; + max_pg = rlim >> PAGE_SHIFT; user = mmp->user ? : current_user(); old_pg = atomic_long_read(&user->locked_vm); @@ -1711,7 +1810,7 @@ struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t gfp_mask) if (skb_pfmemalloc(skb)) gfp_mask |= __GFP_MEMALLOC; - n = kmem_cache_alloc(skbuff_head_cache, gfp_mask); + n = kmem_cache_alloc(skbuff_cache, gfp_mask); if (!n) return NULL; @@ -1893,10 +1992,7 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, if (skb_pfmemalloc(skb)) gfp_mask |= __GFP_MEMALLOC; - size = SKB_DATA_ALIGN(size); - size += SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); - size = kmalloc_size_roundup(size); - data = kmalloc_reserve(size, gfp_mask, NUMA_NO_NODE, NULL); + data = kmalloc_reserve(&size, gfp_mask, NUMA_NO_NODE, NULL); if (!data) goto nodata; size = SKB_WITH_OVERHEAD(size); @@ -1959,7 +2055,7 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, return 0; nofrags: - kfree(data); + skb_kfree_head(data, size); nodata: return -ENOMEM; } @@ -4584,7 +4680,7 @@ static void skb_extensions_init(void) {} void __init skb_init(void) { - skbuff_head_cache = kmem_cache_create_usercopy("skbuff_head_cache", + skbuff_cache = kmem_cache_create_usercopy("skbuff_head_cache", sizeof(struct sk_buff), 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, @@ -4596,6 +4692,19 @@ void __init skb_init(void) 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); +#ifdef HAVE_SKB_SMALL_HEAD_CACHE + /* usercopy should only access first SKB_SMALL_HEAD_HEADROOM bytes. + * struct skb_shared_info is located at the end of skb->head, + * and should not be copied to/from user. + */ + skb_small_head_cache = kmem_cache_create_usercopy("skbuff_small_head", + SKB_SMALL_HEAD_CACHE_SIZE, + 0, + SLAB_HWCACHE_ALIGN | SLAB_PANIC, + 0, + SKB_SMALL_HEAD_HEADROOM, + NULL); +#endif skb_extensions_init(); } @@ -5450,7 +5559,7 @@ void kfree_skb_partial(struct sk_buff *skb, bool head_stolen) { if (head_stolen) { skb_release_head_state(skb); - kmem_cache_free(skbuff_head_cache, skb); + kmem_cache_free(skbuff_cache, skb); } else { __kfree_skb(skb); } @@ -6244,10 +6353,7 @@ static int pskb_carve_inside_header(struct sk_buff *skb, const u32 off, if (skb_pfmemalloc(skb)) gfp_mask |= __GFP_MEMALLOC; - size = SKB_DATA_ALIGN(size); - size += SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); - size = kmalloc_size_roundup(size); - data = kmalloc_reserve(size, gfp_mask, NUMA_NO_NODE, NULL); + data = kmalloc_reserve(&size, gfp_mask, NUMA_NO_NODE, NULL); if (!data) return -ENOMEM; size = SKB_WITH_OVERHEAD(size); @@ -6263,7 +6369,7 @@ static int pskb_carve_inside_header(struct sk_buff *skb, const u32 off, if (skb_cloned(skb)) { /* drop the old head gracefully */ if (skb_orphan_frags(skb, gfp_mask)) { - kfree(data); + skb_kfree_head(data, size); return -ENOMEM; } for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) @@ -6363,10 +6469,7 @@ static int pskb_carve_inside_nonlinear(struct sk_buff *skb, const u32 off, if (skb_pfmemalloc(skb)) gfp_mask |= __GFP_MEMALLOC; - size = SKB_DATA_ALIGN(size); - size += SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); - size = kmalloc_size_roundup(size); - data = kmalloc_reserve(size, gfp_mask, NUMA_NO_NODE, NULL); + data = kmalloc_reserve(&size, gfp_mask, NUMA_NO_NODE, NULL); if (!data) return -ENOMEM; size = SKB_WITH_OVERHEAD(size); @@ -6374,7 +6477,7 @@ static int pskb_carve_inside_nonlinear(struct sk_buff *skb, const u32 off, memcpy((struct skb_shared_info *)(data + size), skb_shinfo(skb), offsetof(struct skb_shared_info, frags[0])); if (skb_orphan_frags(skb, gfp_mask)) { - kfree(data); + skb_kfree_head(data, size); return -ENOMEM; } shinfo = (struct skb_shared_info *)(data + size); @@ -6410,7 +6513,7 @@ static int pskb_carve_inside_nonlinear(struct sk_buff *skb, const u32 off, /* skb_frag_unref() is not needed here as shinfo->nr_frags = 0. */ if (skb_has_frag_list(skb)) kfree_skb_list(skb_shinfo(skb)->frag_list); - kfree(data); + skb_kfree_head(data, size); return -ENOMEM; } skb_release_data(skb, SKB_CONSUMED); |