From 80d2eefcb4c84aa9018b2a997ab3a4c567bc821a Mon Sep 17 00:00:00 2001 From: Sebastian Andrzej Siewior Date: Mon, 25 Mar 2024 08:40:30 +0100 Subject: net: Use backlog-NAPI to clean up the defer_list. The defer_list is a per-CPU list which is used to free skbs outside of the socket lock and on the CPU on which they have been allocated. The list is processed during NAPI callbacks so ideally the list is cleaned up. Should the amount of skbs on the list exceed a certain water mark then the softirq is triggered remotely on the target CPU by invoking a remote function call. The raise of the softirqs via a remote function call leads to waking the ksoftirqd on PREEMPT_RT which is undesired. The backlog-NAPI threads already provide the infrastructure which can be utilized to perform the cleanup of the defer_list. The NAPI state is updated with the input_pkt_queue.lock acquired. It order not to break the state, it is needed to also wake the backlog-NAPI thread with the lock held. This requires to acquire the use the lock in rps_lock_irq*() if the backlog-NAPI threads are used even with RPS disabled. Move the logic of remotely starting softirqs to clean up the defer_list into kick_defer_list_purge(). Make sure a lock is held in rps_lock_irq*() if backlog-NAPI threads are used. Schedule backlog-NAPI for defer_list cleanup if backlog-NAPI is available. Acked-by: Jakub Kicinski Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Paolo Abeni --- net/core/skbuff.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/net/core/skbuff.c b/net/core/skbuff.c index b99127712e67..17617c29be2d 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -7039,8 +7039,8 @@ nodefer: __kfree_skb(skb); /* Make sure to trigger NET_RX_SOFTIRQ on the remote CPU * if we are unlucky enough (this seems very unlikely). */ - if (unlikely(kick) && !cmpxchg(&sd->defer_ipi_scheduled, 0, 1)) - smp_call_function_single_async(cpu, &sd->defer_csd); + if (unlikely(kick)) + kick_defer_list_purge(sd, cpu); } static void skb_splice_csum_page(struct sk_buff *skb, struct page *page, -- cgit v1.2.3-59-g8ed1b From 6e9b01909a811555ff3326cf80a5847169c57806 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Tue, 26 Mar 2024 21:02:12 -0700 Subject: net: remove gfp_mask from napi_alloc_skb() __napi_alloc_skb() is napi_alloc_skb() with the added flexibility of choosing gfp_mask. This is a NAPI function, so GFP_ATOMIC is implied. The only practical choice the caller has is whether to set __GFP_NOWARN. But that's a false choice, too, allocation failures in atomic context will happen, and printing warnings in logs, effectively for a packet drop, is both too much and very likely non-actionable. This leads me to a conclusion that most uses of napi_alloc_skb() are simply misguided, and should use __GFP_NOWARN in the first place. We also have a "standard" way of reporting allocation failures via the queue stat API (qstats::rx-alloc-fail). The direct motivation for this patch is that one of the drivers used at Meta calls napi_alloc_skb() (so prior to this patch without __GFP_NOWARN), and the resulting OOM warning is the top networking warning in our fleet. Reviewed-by: Alexander Lobakin Reviewed-by: Simon Horman Reviewed-by: Eric Dumazet Link: https://lore.kernel.org/r/20240327040213.3153864-1-kuba@kernel.org Signed-off-by: Jakub Kicinski --- Documentation/mm/page_frags.rst | 2 +- Documentation/translations/zh_CN/mm/page_frags.rst | 2 +- drivers/net/ethernet/intel/i40e/i40e_txrx.c | 4 +--- drivers/net/ethernet/intel/i40e/i40e_xsk.c | 3 +-- drivers/net/ethernet/intel/iavf/iavf_txrx.c | 4 +--- drivers/net/ethernet/intel/ice/ice_txrx.c | 3 +-- drivers/net/ethernet/intel/ice/ice_xsk.c | 3 +-- drivers/net/ethernet/intel/idpf/idpf_txrx.c | 5 ++--- drivers/net/ethernet/intel/igc/igc_main.c | 3 +-- drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 3 +-- drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 5 ++--- include/linux/skbuff.h | 8 +------- net/core/skbuff.c | 9 ++++----- 13 files changed, 18 insertions(+), 36 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/Documentation/mm/page_frags.rst b/Documentation/mm/page_frags.rst index a81617e688a8..503ca6cdb804 100644 --- a/Documentation/mm/page_frags.rst +++ b/Documentation/mm/page_frags.rst @@ -25,7 +25,7 @@ to be disabled when executing the fragment allocation. The network stack uses two separate caches per CPU to handle fragment allocation. The netdev_alloc_cache is used by callers making use of the netdev_alloc_frag and __netdev_alloc_skb calls. The napi_alloc_cache is -used by callers of the __napi_alloc_frag and __napi_alloc_skb calls. The +used by callers of the __napi_alloc_frag and napi_alloc_skb calls. The main difference between these two calls is the context in which they may be called. The "netdev" prefixed functions are usable in any context as these functions will disable interrupts, while the "napi" prefixed functions are diff --git a/Documentation/translations/zh_CN/mm/page_frags.rst b/Documentation/translations/zh_CN/mm/page_frags.rst index 20bd3fafdc8c..a5b22486a913 100644 --- a/Documentation/translations/zh_CN/mm/page_frags.rst +++ b/Documentation/translations/zh_CN/mm/page_frags.rst @@ -25,7 +25,7 @@ sk_buff->head使用，或者用于skb_shared_info的 “frags” 部分。 网络堆栈在每个CPU使用两个独立的缓存来处理碎片分配。netdev_alloc_cache被使用 netdev_alloc_frag和__netdev_alloc_skb调用的调用者使用。napi_alloc_cache -被调用__napi_alloc_frag和__napi_alloc_skb的调用者使用。这两个调用的主要区别是 +被调用__napi_alloc_frag和napi_alloc_skb的调用者使用。这两个调用的主要区别是 它们可能被调用的环境。“netdev” 前缀的函数可以在任何上下文中使用，因为这些函数 将禁用中断，而 ”napi“ 前缀的函数只可以在softirq上下文中使用。 diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index 0d7177083708..ac2fcc5ac595 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -2144,9 +2144,7 @@ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, */ /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, - I40E_RX_HDR_SIZE, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&rx_ring->q_vector->napi, I40E_RX_HDR_SIZE); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c index 11500003af0d..a85b425794df 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c +++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c @@ -301,8 +301,7 @@ static struct sk_buff *i40e_construct_skb_zc(struct i40e_ring *rx_ring, net_prefetch(xdp->data_meta); /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, totalsize, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&rx_ring->q_vector->napi, totalsize); if (unlikely(!skb)) goto out; diff --git a/drivers/net/ethernet/intel/iavf/iavf_txrx.c b/drivers/net/ethernet/intel/iavf/iavf_txrx.c index b71484c87a84..32bb604a1382 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_txrx.c +++ b/drivers/net/ethernet/intel/iavf/iavf_txrx.c @@ -1334,9 +1334,7 @@ static struct sk_buff *iavf_construct_skb(struct iavf_ring *rx_ring, net_prefetch(va); /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, - IAVF_RX_HDR_SIZE, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&rx_ring->q_vector->napi, IAVF_RX_HDR_SIZE); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c index 97d41d6ebf1f..8bb743f78fcb 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx.c +++ b/drivers/net/ethernet/intel/ice/ice_txrx.c @@ -1051,8 +1051,7 @@ ice_construct_skb(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp) } /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, ICE_RX_HDR_SIZE, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&rx_ring->q_vector->napi, ICE_RX_HDR_SIZE); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c index 1857220d27fe..aa81d1162b81 100644 --- a/drivers/net/ethernet/intel/ice/ice_xsk.c +++ b/drivers/net/ethernet/intel/ice/ice_xsk.c @@ -555,8 +555,7 @@ ice_construct_skb_zc(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp) } net_prefetch(xdp->data_meta); - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, totalsize, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&rx_ring->q_vector->napi, totalsize); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c index 6dd7a66bb897..f940f650cd78 100644 --- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c +++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c @@ -3005,8 +3005,7 @@ struct sk_buff *idpf_rx_construct_skb(struct idpf_queue *rxq, /* prefetch first cache line of first page */ net_prefetch(va); /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rxq->q_vector->napi, IDPF_RX_HDR_SIZE, - GFP_ATOMIC); + skb = napi_alloc_skb(&rxq->q_vector->napi, IDPF_RX_HDR_SIZE); if (unlikely(!skb)) { idpf_rx_put_page(rx_buf); @@ -3060,7 +3059,7 @@ static struct sk_buff *idpf_rx_hdr_construct_skb(struct idpf_queue *rxq, struct sk_buff *skb; /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rxq->q_vector->napi, size, GFP_ATOMIC); + skb = napi_alloc_skb(&rxq->q_vector->napi, size); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c index 35ad40a803cb..0cd923c8ac9b 100644 --- a/drivers/net/ethernet/intel/igc/igc_main.c +++ b/drivers/net/ethernet/intel/igc/igc_main.c @@ -2712,8 +2712,7 @@ static struct sk_buff *igc_construct_skb_zc(struct igc_ring *ring, net_prefetch(xdp->data_meta); - skb = __napi_alloc_skb(&ring->q_vector->napi, totalsize, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&ring->q_vector->napi, totalsize); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c index d34d715c59eb..397cb773fabb 100644 --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c @@ -220,8 +220,7 @@ static struct sk_buff *ixgbe_construct_skb_zc(struct ixgbe_ring *rx_ring, net_prefetch(xdp->data_meta); /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, totalsize, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&rx_ring->q_vector->napi, totalsize); if (unlikely(!skb)) return NULL; diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c index 24cd80490d19..bcdde68a099a 100644 --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c @@ -5109,9 +5109,8 @@ static struct sk_buff *stmmac_construct_skb_zc(struct stmmac_channel *ch, unsigned int datasize = xdp->data_end - xdp->data; struct sk_buff *skb; - skb = __napi_alloc_skb(&ch->rxtx_napi, - xdp->data_end - xdp->data_hard_start, - GFP_ATOMIC | __GFP_NOWARN); + skb = napi_alloc_skb(&ch->rxtx_napi, + xdp->data_end - xdp->data_hard_start); if (unlikely(!skb)) return NULL; diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index 517e546a120a..b7f1ecdaec38 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -3350,13 +3350,7 @@ static inline void *napi_alloc_frag_align(unsigned int fragsz, return __napi_alloc_frag_align(fragsz, -align); } -struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, - unsigned int length, gfp_t gfp_mask); -static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, - unsigned int length) -{ - return __napi_alloc_skb(napi, length, GFP_ATOMIC); -} +struct sk_buff *napi_alloc_skb(struct napi_struct *napi, unsigned int length); void napi_consume_skb(struct sk_buff *skb, int budget); void napi_skb_free_stolen_head(struct sk_buff *skb); diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 17617c29be2d..a1be84be5d35 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -775,10 +775,9 @@ skb_fail: EXPORT_SYMBOL(__netdev_alloc_skb); /** - * __napi_alloc_skb - allocate skbuff for rx in a specific NAPI instance + * napi_alloc_skb - allocate skbuff for rx in a specific NAPI instance * @napi: napi instance this buffer was allocated for * @len: length to allocate - * @gfp_mask: get_free_pages mask, passed to alloc_skb and alloc_pages * * Allocate a new sk_buff for use in NAPI receive. This buffer will * attempt to allocate the head from a special reserved region used @@ -787,9 +786,9 @@ EXPORT_SYMBOL(__netdev_alloc_skb); * * %NULL is returned if there is no free memory. */ -struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, unsigned int len, - gfp_t gfp_mask) +struct sk_buff *napi_alloc_skb(struct napi_struct *napi, unsigned int len) { + gfp_t gfp_mask = GFP_ATOMIC | __GFP_NOWARN; struct napi_alloc_cache *nc; struct sk_buff *skb; bool pfmemalloc; @@ -860,7 +859,7 @@ skb_success: skb_fail: return skb; } -EXPORT_SYMBOL(__napi_alloc_skb); +EXPORT_SYMBOL(napi_alloc_skb); void skb_add_rx_frag_netmem(struct sk_buff *skb, int i, netmem_ref netmem, int off, int size, unsigned int truesize) -- cgit v1.2.3-59-g8ed1b From 4a96a4e807c390a9d91b450ebe04eeb2e0ecc076 Mon Sep 17 00:00:00 2001 From: Alexander Lobakin Date: Fri, 29 Mar 2024 17:55:06 +0100 Subject: page_pool: check for PP direct cache locality later Since we have pool->p.napi (Jakub) and pool->cpuid (Lorenzo) to check whether it's safe to use direct recycling, we can use both globally for each page instead of relying solely on @allow_direct argument. Let's assume that @allow_direct means "I'm sure it's local, don't waste time rechecking this" and when it's false, try the mentioned params to still recycle the page directly. If neither is true, we'll lose some CPU cycles, but then it surely won't be hotpath. On the other hand, paths where it's possible to use direct cache, but not possible to safely set @allow_direct, will benefit from this move. The whole propagation of @napi_safe through a dozen of skb freeing functions can now go away, which saves us some stack space. Signed-off-by: Alexander Lobakin Link: https://lore.kernel.org/r/20240329165507.3240110-2-aleksander.lobakin@intel.com Signed-off-by: Jakub Kicinski --- include/linux/skbuff.h | 12 ++++----- net/core/page_pool.c | 31 +++++++++++++++++++--- net/core/skbuff.c | 70 ++++++++++++++++---------------------------------- net/ipv4/esp4.c | 2 +- net/ipv6/esp6.c | 2 +- 5 files changed, 58 insertions(+), 59 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index b7f1ecdaec38..03ea36a82cdd 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -3510,25 +3510,25 @@ int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb, unsigned int headroom); int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb, struct bpf_prog *prog); -bool napi_pp_put_page(struct page *page, bool napi_safe); +bool napi_pp_put_page(struct page *page); static inline void -skb_page_unref(const struct sk_buff *skb, struct page *page, bool napi_safe) +skb_page_unref(const struct sk_buff *skb, struct page *page) { #ifdef CONFIG_PAGE_POOL - if (skb->pp_recycle && napi_pp_put_page(page, napi_safe)) + if (skb->pp_recycle && napi_pp_put_page(page)) return; #endif put_page(page); } static inline void -napi_frag_unref(skb_frag_t *frag, bool recycle, bool napi_safe) +napi_frag_unref(skb_frag_t *frag, bool recycle) { struct page *page = skb_frag_page(frag); #ifdef CONFIG_PAGE_POOL - if (recycle && napi_pp_put_page(page, napi_safe)) + if (recycle && napi_pp_put_page(page)) return; #endif put_page(page); @@ -3544,7 +3544,7 @@ napi_frag_unref(skb_frag_t *frag, bool recycle, bool napi_safe) */ static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) { - napi_frag_unref(frag, recycle, false); + napi_frag_unref(frag, recycle); } /** diff --git a/net/core/page_pool.c b/net/core/page_pool.c index dd364d738c00..9d56257e444b 100644 --- a/net/core/page_pool.c +++ b/net/core/page_pool.c @@ -690,8 +690,7 @@ __page_pool_put_page(struct page_pool *pool, struct page *page, page_pool_dma_sync_for_device(pool, page, dma_sync_size); - if (allow_direct && in_softirq() && - page_pool_recycle_in_cache(page, pool)) + if (allow_direct && page_pool_recycle_in_cache(page, pool)) return NULL; /* Page found as candidate for recycling */ @@ -716,9 +715,35 @@ __page_pool_put_page(struct page_pool *pool, struct page *page, return NULL; } +static bool page_pool_napi_local(const struct page_pool *pool) +{ + const struct napi_struct *napi; + u32 cpuid; + + if (unlikely(!in_softirq())) + return false; + + /* Allow direct recycle if we have reasons to believe that we are + * in the same context as the consumer would run, so there's + * no possible race. + * __page_pool_put_page() makes sure we're not in hardirq context + * and interrupts are enabled prior to accessing the cache. + */ + cpuid = smp_processor_id(); + if (READ_ONCE(pool->cpuid) == cpuid) + return true; + + napi = READ_ONCE(pool->p.napi); + + return napi && READ_ONCE(napi->list_owner) == cpuid; +} + void page_pool_put_unrefed_page(struct page_pool *pool, struct page *page, unsigned int dma_sync_size, bool allow_direct) { + if (!allow_direct) + allow_direct = page_pool_napi_local(pool); + page = __page_pool_put_page(pool, page, dma_sync_size, allow_direct); if (page && !page_pool_recycle_in_ring(pool, page)) { /* Cache full, fallback to free pages */ @@ -969,7 +994,7 @@ void page_pool_use_xdp_mem(struct page_pool *pool, void (*disconnect)(void *), static void page_pool_disable_direct_recycling(struct page_pool *pool) { /* Disable direct recycling based on pool->cpuid. - * Paired with READ_ONCE() in napi_pp_put_page(). + * Paired with READ_ONCE() in page_pool_napi_local(). */ WRITE_ONCE(pool->cpuid, -1); diff --git a/net/core/skbuff.c b/net/core/skbuff.c index a1be84be5d35..2a5ce6667bbb 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -1004,11 +1004,8 @@ int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb, EXPORT_SYMBOL(skb_cow_data_for_xdp); #if IS_ENABLED(CONFIG_PAGE_POOL) -bool napi_pp_put_page(struct page *page, bool napi_safe) +bool napi_pp_put_page(struct page *page) { - bool allow_direct = false; - struct page_pool *pp; - page = compound_head(page); /* page->pp_magic is OR'ed with PP_SIGNATURE after the allocation @@ -1021,39 +1018,18 @@ bool napi_pp_put_page(struct page *page, bool napi_safe) if (unlikely(!is_pp_page(page))) return false; - pp = page->pp; - - /* Allow direct recycle if we have reasons to believe that we are - * in the same context as the consumer would run, so there's - * no possible race. - * __page_pool_put_page() makes sure we're not in hardirq context - * and interrupts are enabled prior to accessing the cache. - */ - if (napi_safe || in_softirq()) { - const struct napi_struct *napi = READ_ONCE(pp->p.napi); - unsigned int cpuid = smp_processor_id(); - - allow_direct = napi && READ_ONCE(napi->list_owner) == cpuid; - allow_direct |= READ_ONCE(pp->cpuid) == cpuid; - } - - /* Driver set this to memory recycling info. Reset it on recycle. - * This will *not* work for NIC using a split-page memory model. - * The page will be returned to the pool here regardless of the - * 'flipped' fragment being in use or not. - */ - page_pool_put_full_page(pp, page, allow_direct); + page_pool_put_full_page(page->pp, page, false); return true; } EXPORT_SYMBOL(napi_pp_put_page); #endif -static bool skb_pp_recycle(struct sk_buff *skb, void *data, bool napi_safe) +static bool skb_pp_recycle(struct sk_buff *skb, void *data) { if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle) return false; - return napi_pp_put_page(virt_to_page(data), napi_safe); + return napi_pp_put_page(virt_to_page(data)); } /** @@ -1095,12 +1071,12 @@ static void skb_kfree_head(void *head, unsigned int end_offset) kfree(head); } -static void skb_free_head(struct sk_buff *skb, bool napi_safe) +static void skb_free_head(struct sk_buff *skb) { unsigned char *head = skb->head; if (skb->head_frag) { - if (skb_pp_recycle(skb, head, napi_safe)) + if (skb_pp_recycle(skb, head)) return; skb_free_frag(head); } else { @@ -1108,8 +1084,7 @@ static void skb_free_head(struct sk_buff *skb, bool napi_safe) } } -static void skb_release_data(struct sk_buff *skb, enum skb_drop_reason reason, - bool napi_safe) +static void skb_release_data(struct sk_buff *skb, enum skb_drop_reason reason) { struct skb_shared_info *shinfo = skb_shinfo(skb); int i; @@ -1126,13 +1101,13 @@ static void skb_release_data(struct sk_buff *skb, enum skb_drop_reason reason, } for (i = 0; i < shinfo->nr_frags; i++) - napi_frag_unref(&shinfo->frags[i], skb->pp_recycle, napi_safe); + napi_frag_unref(&shinfo->frags[i], skb->pp_recycle); free_head: if (shinfo->frag_list) kfree_skb_list_reason(shinfo->frag_list, reason); - skb_free_head(skb, napi_safe); + skb_free_head(skb); exit: /* When we clone an SKB we copy the reycling bit. The pp_recycle * bit is only set on the head though, so in order to avoid races @@ -1193,12 +1168,11 @@ void skb_release_head_state(struct sk_buff *skb) } /* Free everything but the sk_buff shell. */ -static void skb_release_all(struct sk_buff *skb, enum skb_drop_reason reason, - bool napi_safe) +static void skb_release_all(struct sk_buff *skb, enum skb_drop_reason reason) { skb_release_head_state(skb); if (likely(skb->head)) - skb_release_data(skb, reason, napi_safe); + skb_release_data(skb, reason); } /** @@ -1212,7 +1186,7 @@ static void skb_release_all(struct sk_buff *skb, enum skb_drop_reason reason, void __kfree_skb(struct sk_buff *skb) { - skb_release_all(skb, SKB_DROP_REASON_NOT_SPECIFIED, false); + skb_release_all(skb, SKB_DROP_REASON_NOT_SPECIFIED); kfree_skbmem(skb); } EXPORT_SYMBOL(__kfree_skb); @@ -1269,7 +1243,7 @@ static void kfree_skb_add_bulk(struct sk_buff *skb, return; } - skb_release_all(skb, reason, false); + skb_release_all(skb, reason); sa->skb_array[sa->skb_count++] = skb; if (unlikely(sa->skb_count == KFREE_SKB_BULK_SIZE)) { @@ -1443,7 +1417,7 @@ EXPORT_SYMBOL(consume_skb); void __consume_stateless_skb(struct sk_buff *skb) { trace_consume_skb(skb, __builtin_return_address(0)); - skb_release_data(skb, SKB_CONSUMED, false); + skb_release_data(skb, SKB_CONSUMED); kfree_skbmem(skb); } @@ -1470,7 +1444,7 @@ static void napi_skb_cache_put(struct sk_buff *skb) void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason) { - skb_release_all(skb, reason, true); + skb_release_all(skb, reason); napi_skb_cache_put(skb); } @@ -1508,7 +1482,7 @@ void napi_consume_skb(struct sk_buff *skb, int budget) return; } - skb_release_all(skb, SKB_CONSUMED, !!budget); + skb_release_all(skb, SKB_CONSUMED); napi_skb_cache_put(skb); } EXPORT_SYMBOL(napi_consume_skb); @@ -1639,7 +1613,7 @@ EXPORT_SYMBOL_GPL(alloc_skb_for_msg); */ struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src) { - skb_release_all(dst, SKB_CONSUMED, false); + skb_release_all(dst, SKB_CONSUMED); return __skb_clone(dst, src); } EXPORT_SYMBOL_GPL(skb_morph); @@ -2271,9 +2245,9 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, if (skb_has_frag_list(skb)) skb_clone_fraglist(skb); - skb_release_data(skb, SKB_CONSUMED, false); + skb_release_data(skb, SKB_CONSUMED); } else { - skb_free_head(skb, false); + skb_free_head(skb); } off = (data + nhead) - skb->head; @@ -6574,12 +6548,12 @@ static int pskb_carve_inside_header(struct sk_buff *skb, const u32 off, skb_frag_ref(skb, i); if (skb_has_frag_list(skb)) skb_clone_fraglist(skb); - skb_release_data(skb, SKB_CONSUMED, false); + skb_release_data(skb, SKB_CONSUMED); } else { /* we can reuse existing recount- all we did was * relocate values */ - skb_free_head(skb, false); + skb_free_head(skb); } skb->head = data; @@ -6714,7 +6688,7 @@ static int pskb_carve_inside_nonlinear(struct sk_buff *skb, const u32 off, skb_kfree_head(data, size); return -ENOMEM; } - skb_release_data(skb, SKB_CONSUMED, false); + skb_release_data(skb, SKB_CONSUMED); skb->head = data; skb->head_frag = 0; diff --git a/net/ipv4/esp4.c b/net/ipv4/esp4.c index d33d12421814..3d647c9a7a21 100644 --- a/net/ipv4/esp4.c +++ b/net/ipv4/esp4.c @@ -114,7 +114,7 @@ static void esp_ssg_unref(struct xfrm_state *x, void *tmp, struct sk_buff *skb) */ if (req->src != req->dst) for (sg = sg_next(req->src); sg; sg = sg_next(sg)) - skb_page_unref(skb, sg_page(sg), false); + skb_page_unref(skb, sg_page(sg)); } #ifdef CONFIG_INET_ESPINTCP diff --git a/net/ipv6/esp6.c b/net/ipv6/esp6.c index 7371886d4f9f..fe8d53f5a5ee 100644 --- a/net/ipv6/esp6.c +++ b/net/ipv6/esp6.c @@ -131,7 +131,7 @@ static void esp_ssg_unref(struct xfrm_state *x, void *tmp, struct sk_buff *skb) */ if (req->src != req->dst) for (sg = sg_next(req->src); sg; sg = sg_next(sg)) - skb_page_unref(skb, sg_page(sg), false); + skb_page_unref(skb, sg_page(sg)); } #ifdef CONFIG_INET6_ESPINTCP -- cgit v1.2.3-59-g8ed1b From 4308811ba90118ae1b71a95fee79ab7dada6400c Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 7 Apr 2024 08:06:06 +0000 Subject: net: display more skb fields in skb_dump() Print these additional fields in skb_dump() to ease debugging. - mac_len - csum_start (in v2, at Willem suggestion) - csum_offset (in v2, at Willem suggestion) - priority - mark - alloc_cpu - vlan_all - encapsulation - inner_protocol - inner_mac_header - inner_network_header - inner_transport_header Signed-off-by: Eric Dumazet Reviewed-by: Willem de Bruijn Signed-off-by: David S. Miller --- net/core/skbuff.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 2a5ce6667bbb..21cd01641f4c 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -1304,22 +1304,28 @@ void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt) has_trans = skb_transport_header_was_set(skb); printk("%sskb len=%u headroom=%u headlen=%u tailroom=%u\n" - "mac=(%d,%d) net=(%d,%d) trans=%d\n" + "mac=(%d,%d) mac_len=%u net=(%d,%d) trans=%d\n" "shinfo(txflags=%u nr_frags=%u gso(size=%hu type=%u segs=%hu))\n" - "csum(0x%x ip_summed=%u complete_sw=%u valid=%u level=%u)\n" - "hash(0x%x sw=%u l4=%u) proto=0x%04x pkttype=%u iif=%d\n", + "csum(0x%x start=%u offset=%u ip_summed=%u complete_sw=%u valid=%u level=%u)\n" + "hash(0x%x sw=%u l4=%u) proto=0x%04x pkttype=%u iif=%d\n" + "priority=0x%x mark=0x%x alloc_cpu=%u vlan_all=0x%x\n" + "encapsulation=%d inner(proto=0x%04x, mac=%u, net=%u, trans=%u)\n", level, skb->len, headroom, skb_headlen(skb), tailroom, has_mac ? skb->mac_header : -1, has_mac ? skb_mac_header_len(skb) : -1, + skb->mac_len, skb->network_header, has_trans ? skb_network_header_len(skb) : -1, has_trans ? skb->transport_header : -1, sh->tx_flags, sh->nr_frags, sh->gso_size, sh->gso_type, sh->gso_segs, - skb->csum, skb->ip_summed, skb->csum_complete_sw, - skb->csum_valid, skb->csum_level, + skb->csum, skb->csum_start, skb->csum_offset, skb->ip_summed, + skb->csum_complete_sw, skb->csum_valid, skb->csum_level, skb->hash, skb->sw_hash, skb->l4_hash, - ntohs(skb->protocol), skb->pkt_type, skb->skb_iif); + ntohs(skb->protocol), skb->pkt_type, skb->skb_iif, + skb->priority, skb->mark, skb->alloc_cpu, skb->vlan_all, + skb->encapsulation, skb->inner_protocol, skb->inner_mac_header, + skb->inner_network_header, skb->inner_transport_header); if (dev) printk("%sdev name=%s feat=%pNF\n", -- cgit v1.2.3-59-g8ed1b From f58f3c9563409e618e591d0d540316286cb0665f Mon Sep 17 00:00:00 2001 From: Mina Almasry Date: Mon, 8 Apr 2024 08:29:58 -0700 Subject: net: remove napi_frag_unref With the changes in the last patches, napi_frag_unref() is now reduandant. Remove it and use skb_page_unref directly. Signed-off-by: Mina Almasry Reviewed-by: Dragos Tatulea Reviewed-by: Eric Dumazet Reviewed-by: Jacob Keller Link: https://lore.kernel.org/r/20240408153000.2152844-4-almasrymina@google.com Signed-off-by: Jakub Kicinski --- include/linux/skbuff.h | 8 +------- net/core/skbuff.c | 2 +- 2 files changed, 2 insertions(+), 8 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index c0ff85bb087a..7135a3e94afd 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -3531,12 +3531,6 @@ skb_page_unref(struct page *page, bool recycle) put_page(page); } -static inline void -napi_frag_unref(skb_frag_t *frag, bool recycle) -{ - skb_page_unref(skb_frag_page(frag), recycle); -} - /** * __skb_frag_unref - release a reference on a paged fragment. * @frag: the paged fragment @@ -3547,7 +3541,7 @@ napi_frag_unref(skb_frag_t *frag, bool recycle) */ static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) { - napi_frag_unref(frag, recycle); + skb_page_unref(skb_frag_page(frag), recycle); } /** diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 21cd01641f4c..888874ef8566 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -1101,7 +1101,7 @@ static void skb_release_data(struct sk_buff *skb, enum skb_drop_reason reason) } for (i = 0; i < shinfo->nr_frags; i++) - napi_frag_unref(&shinfo->frags[i], skb->pp_recycle); + __skb_frag_unref(&shinfo->frags[i], skb->pp_recycle); free_head: if (shinfo->frag_list) -- cgit v1.2.3-59-g8ed1b From 7cb31c46b9cc37a5e564667fe46daf9a35dbafdd Mon Sep 17 00:00:00 2001 From: Pavel Begunkov Date: Wed, 10 Apr 2024 02:28:09 +0100 Subject: net: cache for same cpu skb_attempt_defer_free Optimise skb_attempt_defer_free() when run by the same CPU the skb was allocated on. Instead of __kfree_skb() -> kmem_cache_free() we can disable softirqs and put the buffer into cpu local caches. CPU bound TCP ping pong style benchmarking (i.e. netbench) showed a 1% throughput increase (392.2 -> 396.4 Krps). Cross checking with profiles, the total CPU share of skb_attempt_defer_free() dropped by 0.6%. Note, I'd expect the win doubled with rx only benchmarks, as the optimisation is for the receive path, but the test spends >55% of CPU doing writes. Signed-off-by: Pavel Begunkov Reviewed-by: Eric Dumazet Link: https://lore.kernel.org/r/a887463fb219d973ec5ad275e31194812571f1f5.1712711977.git.asml.silence@gmail.com Signed-off-by: Jakub Kicinski --- net/core/skbuff.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'net/core/skbuff.c') diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 888874ef8566..18612f255ef5 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -6974,6 +6974,19 @@ free_now: EXPORT_SYMBOL(__skb_ext_put); #endif /* CONFIG_SKB_EXTENSIONS */ +static void kfree_skb_napi_cache(struct sk_buff *skb) +{ + /* if SKB is a clone, don't handle this case */ + if (skb->fclone != SKB_FCLONE_UNAVAILABLE) { + __kfree_skb(skb); + return; + } + + local_bh_disable(); + __napi_kfree_skb(skb, SKB_DROP_REASON_NOT_SPECIFIED); + local_bh_enable(); +} + /** * skb_attempt_defer_free - queue skb for remote freeing * @skb: buffer @@ -6992,7 +7005,7 @@ void skb_attempt_defer_free(struct sk_buff *skb) if (WARN_ON_ONCE(cpu >= nr_cpu_ids) || !cpu_online(cpu) || cpu == raw_smp_processor_id()) { -nodefer: __kfree_skb(skb); +nodefer: kfree_skb_napi_cache(skb); return; } -- cgit v1.2.3-59-g8ed1b From d8415a165c43905ccf9ad7a4c7eaf6006b91a00b Mon Sep 17 00:00:00 2001 From: Pavel Begunkov Date: Wed, 10 Apr 2024 02:28:10 +0100 Subject: net: use SKB_CONSUMED in skb_attempt_defer_free() skb_attempt_defer_free() is used to free already processed skbs, so pass SKB_CONSUMED as the reason in kfree_skb_napi_cache(). Suggested-by: Jason Xing Suggested-by: Eric Dumazet Signed-off-by: Pavel Begunkov Reviewed-by: Eric Dumazet Link: https://lore.kernel.org/r/bcf5dbdda79688b074ab7ae2238535840a6d3fc2.1712711977.git.asml.silence@gmail.com Signed-off-by: Jakub Kicinski --- net/core/skbuff.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/core/skbuff.c') diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 18612f255ef5..ab970ded8a7b 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -6983,7 +6983,7 @@ static void kfree_skb_napi_cache(struct sk_buff *skb) } local_bh_disable(); - __napi_kfree_skb(skb, SKB_DROP_REASON_NOT_SPECIFIED); + __napi_kfree_skb(skb, SKB_CONSUMED); local_bh_enable(); } -- cgit v1.2.3-59-g8ed1b From f6d827b180bda01f8805bf5e85307419b0d6f890 Mon Sep 17 00:00:00 2001 From: Mina Almasry Date: Wed, 10 Apr 2024 12:05:01 -0700 Subject: net: move skb ref helpers to new header Add a new header, linux/skbuff_ref.h, which contains all the skb_*_ref() helpers. Many of the consumers of skbuff.h do not actually use any of the skb ref helpers, and we can speed up compilation a bit by minimizing this header file. Additionally in the later patch in the series we add page_pool support to skb_frag_ref(), which requires some page_pool dependencies. We can now add these dependencies to skbuff_ref.h instead of a very ubiquitous skbuff.h Signed-off-by: Mina Almasry Link: https://lore.kernel.org/r/20240410190505.1225848-2-almasrymina@google.com Signed-off-by: Jakub Kicinski --- .../chelsio/inline_crypto/ch_ktls/chcr_ktls.c | 1 + drivers/net/ethernet/marvell/sky2.c | 1 + drivers/net/ethernet/mellanox/mlx4/en_rx.c | 1 + drivers/net/ethernet/sun/cassini.c | 1 + drivers/net/veth.c | 1 + drivers/net/xen-netback/netback.c | 1 + include/linux/skbuff.h | 63 ------------------ include/linux/skbuff_ref.h | 75 ++++++++++++++++++++++ net/core/gro.c | 1 + net/core/skbuff.c | 1 + net/ipv4/esp4.c | 1 + net/ipv4/tcp_output.c | 1 + net/ipv6/esp6.c | 1 + net/tls/tls_device.c | 1 + net/tls/tls_device_fallback.c | 1 + net/tls/tls_strp.c | 1 + 16 files changed, 89 insertions(+), 63 deletions(-) create mode 100644 include/linux/skbuff_ref.h (limited to 'net/core/skbuff.c') diff --git a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c index 6482728794dd..e8e460a92e0e 100644 --- a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c +++ b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c @@ -10,6 +10,7 @@ #include #include #include +#include #include "chcr_ktls.h" static LIST_HEAD(uld_ctx_list); diff --git a/drivers/net/ethernet/marvell/sky2.c b/drivers/net/ethernet/marvell/sky2.c index 07720841a8d7..f3f7f4cc27b3 100644 --- a/drivers/net/ethernet/marvell/sky2.c +++ b/drivers/net/ethernet/marvell/sky2.c @@ -34,6 +34,7 @@ #include #include #include +#include #include diff --git a/drivers/net/ethernet/mellanox/mlx4/en_rx.c b/drivers/net/ethernet/mellanox/mlx4/en_rx.c index eac49657bd07..8328df8645d5 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_rx.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_rx.c @@ -42,6 +42,7 @@ #include #include #include +#include #include #if IS_ENABLED(CONFIG_IPV6) diff --git a/drivers/net/ethernet/sun/cassini.c b/drivers/net/ethernet/sun/cassini.c index bfb903506367..8f1f43dbb76d 100644 --- a/drivers/net/ethernet/sun/cassini.c +++ b/drivers/net/ethernet/sun/cassini.c @@ -73,6 +73,7 @@ #include #include #include +#include #include #include #include diff --git a/drivers/net/veth.c b/drivers/net/veth.c index bcdfbf61eb66..426e68a95067 100644 --- a/drivers/net/veth.c +++ b/drivers/net/veth.c @@ -26,6 +26,7 @@ #include #include #include +#include #include #define DRV_NAME "veth" diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index ef76850d9bcd..48254fc07d64 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -38,6 +38,7 @@ #include #include #include +#include #include diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index 7135a3e94afd..4072a7ee3859 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -3492,73 +3492,10 @@ static inline struct page *skb_frag_page(const skb_frag_t *frag) return netmem_to_page(frag->netmem); } -/** - * __skb_frag_ref - take an addition reference on a paged fragment. - * @frag: the paged fragment - * - * Takes an additional reference on the paged fragment @frag. - */ -static inline void __skb_frag_ref(skb_frag_t *frag) -{ - get_page(skb_frag_page(frag)); -} - -/** - * skb_frag_ref - take an addition reference on a paged fragment of an skb. - * @skb: the buffer - * @f: the fragment offset. - * - * Takes an additional reference on the @f'th paged fragment of @skb. - */ -static inline void skb_frag_ref(struct sk_buff *skb, int f) -{ - __skb_frag_ref(&skb_shinfo(skb)->frags[f]); -} - int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb, unsigned int headroom); int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb, struct bpf_prog *prog); -bool napi_pp_put_page(struct page *page); - -static inline void -skb_page_unref(struct page *page, bool recycle) -{ -#ifdef CONFIG_PAGE_POOL - if (recycle && napi_pp_put_page(page)) - return; -#endif - put_page(page); -} - -/** - * __skb_frag_unref - release a reference on a paged fragment. - * @frag: the paged fragment - * @recycle: recycle the page if allocated via page_pool - * - * Releases a reference on the paged fragment @frag - * or recycles the page via the page_pool API. - */ -static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) -{ - skb_page_unref(skb_frag_page(frag), recycle); -} - -/** - * skb_frag_unref - release a reference on a paged fragment of an skb. - * @skb: the buffer - * @f: the fragment offset - * - * Releases a reference on the @f'th paged fragment of @skb. - */ -static inline void skb_frag_unref(struct sk_buff *skb, int f) -{ - struct skb_shared_info *shinfo = skb_shinfo(skb); - - if (!skb_zcopy_managed(skb)) - __skb_frag_unref(&shinfo->frags[f], skb->pp_recycle); -} - /** * skb_frag_address - gets the address of the data contained in a paged fragment * @frag: the paged fragment buffer diff --git a/include/linux/skbuff_ref.h b/include/linux/skbuff_ref.h new file mode 100644 index 000000000000..11f0a4063403 --- /dev/null +++ b/include/linux/skbuff_ref.h @@ -0,0 +1,75 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * Skb ref helpers. + * + */ + +#ifndef _LINUX_SKBUFF_REF_H +#define _LINUX_SKBUFF_REF_H + +#include + +/** + * __skb_frag_ref - take an addition reference on a paged fragment. + * @frag: the paged fragment + * + * Takes an additional reference on the paged fragment @frag. + */ +static inline void __skb_frag_ref(skb_frag_t *frag) +{ + get_page(skb_frag_page(frag)); +} + +/** + * skb_frag_ref - take an addition reference on a paged fragment of an skb. + * @skb: the buffer + * @f: the fragment offset. + * + * Takes an additional reference on the @f'th paged fragment of @skb. + */ +static inline void skb_frag_ref(struct sk_buff *skb, int f) +{ + __skb_frag_ref(&skb_shinfo(skb)->frags[f]); +} + +bool napi_pp_put_page(struct page *page); + +static inline void +skb_page_unref(struct page *page, bool recycle) +{ +#ifdef CONFIG_PAGE_POOL + if (recycle && napi_pp_put_page(page)) + return; +#endif + put_page(page); +} + +/** + * __skb_frag_unref - release a reference on a paged fragment. + * @frag: the paged fragment + * @recycle: recycle the page if allocated via page_pool + * + * Releases a reference on the paged fragment @frag + * or recycles the page via the page_pool API. + */ +static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle) +{ + skb_page_unref(skb_frag_page(frag), recycle); +} + +/** + * skb_frag_unref - release a reference on a paged fragment of an skb. + * @skb: the buffer + * @f: the fragment offset + * + * Releases a reference on the @f'th paged fragment of @skb. + */ +static inline void skb_frag_unref(struct sk_buff *skb, int f) +{ + struct skb_shared_info *shinfo = skb_shinfo(skb); + + if (!skb_zcopy_managed(skb)) + __skb_frag_unref(&shinfo->frags[f], skb->pp_recycle); +} + +#endif /* _LINUX_SKBUFF_REF_H */ diff --git a/net/core/gro.c b/net/core/gro.c index 83f35d99a682..2459ab697f7f 100644 --- a/net/core/gro.c +++ b/net/core/gro.c @@ -3,6 +3,7 @@ #include #include #include +#include #define MAX_GRO_SKBS 8 diff --git a/net/core/skbuff.c b/net/core/skbuff.c index ab970ded8a7b..2554a6f5f386 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -51,6 +51,7 @@ #endif #include #include +#include #include #include #include diff --git a/net/ipv4/esp4.c b/net/ipv4/esp4.c index 40330253f076..dff04580318f 100644 --- a/net/ipv4/esp4.c +++ b/net/ipv4/esp4.c @@ -20,6 +20,7 @@ #include #include #include +#include #include diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 9282fafc0e61..61119d42b0fd 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -44,6 +44,7 @@ #include #include #include +#include #include diff --git a/net/ipv6/esp6.c b/net/ipv6/esp6.c index fb431d0a3475..6bc0a84c8d05 100644 --- a/net/ipv6/esp6.c +++ b/net/ipv6/esp6.c @@ -36,6 +36,7 @@ #include #include #include +#include #include diff --git a/net/tls/tls_device.c b/net/tls/tls_device.c index bf8ed36b1ad6..ab6e694f7bc2 100644 --- a/net/tls/tls_device.c +++ b/net/tls/tls_device.c @@ -37,6 +37,7 @@ #include #include #include +#include #include "tls.h" #include "trace.h" diff --git a/net/tls/tls_device_fallback.c b/net/tls/tls_device_fallback.c index 4e7228f275fa..f9e3d3d90dcf 100644 --- a/net/tls/tls_device_fallback.c +++ b/net/tls/tls_device_fallback.c @@ -33,6 +33,7 @@ #include #include #include +#include #include "tls.h" diff --git a/net/tls/tls_strp.c b/net/tls/tls_strp.c index ca1e0e198ceb..58c4b06f4f0c 100644 --- a/net/tls/tls_strp.c +++ b/net/tls/tls_strp.c @@ -2,6 +2,7 @@ /* Copyright (c) 2016 Tom Herbert */ #include +#include #include #include #include -- cgit v1.2.3-59-g8ed1b From a580ea994fd37f4105028f5a85c38ff6508a2b25 Mon Sep 17 00:00:00 2001 From: Mina Almasry Date: Wed, 10 Apr 2024 12:05:02 -0700 Subject: net: mirror skb frag ref/unref helpers Refactor some of the skb frag ref/unref helpers for improved clarity. Implement napi_pp_get_page() to be the mirror counterpart of napi_pp_put_page(). Implement skb_page_ref() to be the mirror of skb_page_unref(). Improve __skb_frag_ref() to become a mirror counterpart of __skb_frag_unref(). Previously unref could handle pp & non-pp pages, while the ref could only handle non-pp pages. Now both the ref & unref helpers can correctly handle both pp & non-pp pages. Now that __skb_frag_ref() can handle both pp & non-pp pages, remove skb_pp_frag_ref(), and use __skb_frag_ref() instead. This lets us remove pp specific handling from skb_try_coalesce. Additionally, since __skb_frag_ref() can now handle both pp & non-pp pages, a latent issue in skb_shift() should now be fixed. Previously this function would do a non-pp ref & pp unref on potential pp frags (fragfrom). After this patch, skb_shift() should correctly do a pp ref/unref on pp frags. Signed-off-by: Mina Almasry Reviewed-by: Dragos Tatulea Reviewed-by: Jacob Keller Link: https://lore.kernel.org/r/20240410190505.1225848-3-almasrymina@google.com Signed-off-by: Jakub Kicinski --- .../chelsio/inline_crypto/ch_ktls/chcr_ktls.c | 2 +- drivers/net/ethernet/sun/cassini.c | 4 +- drivers/net/veth.c | 2 +- include/linux/skbuff_ref.h | 39 ++++++++++++++++-- net/core/skbuff.c | 46 ++-------------------- net/tls/tls_device_fallback.c | 2 +- 6 files changed, 44 insertions(+), 51 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c index e8e460a92e0e..3832c2e8ea5a 100644 --- a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c +++ b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c @@ -1659,7 +1659,7 @@ static void chcr_ktls_copy_record_in_skb(struct sk_buff *nskb, for (i = 0; i < record->num_frags; i++) { skb_shinfo(nskb)->frags[i] = record->frags[i]; /* increase the frag ref count */ - __skb_frag_ref(&skb_shinfo(nskb)->frags[i]); + __skb_frag_ref(&skb_shinfo(nskb)->frags[i], nskb->pp_recycle); } skb_shinfo(nskb)->nr_frags = record->num_frags; diff --git a/drivers/net/ethernet/sun/cassini.c b/drivers/net/ethernet/sun/cassini.c index 8f1f43dbb76d..f058e154a3bc 100644 --- a/drivers/net/ethernet/sun/cassini.c +++ b/drivers/net/ethernet/sun/cassini.c @@ -2000,7 +2000,7 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, skb->len += hlen - swivel; skb_frag_fill_page_desc(frag, page->buffer, off, hlen - swivel); - __skb_frag_ref(frag); + __skb_frag_ref(frag, skb->pp_recycle); /* any more data? */ if ((words[0] & RX_COMP1_SPLIT_PKT) && ((dlen -= hlen) > 0)) { @@ -2024,7 +2024,7 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, frag++; skb_frag_fill_page_desc(frag, page->buffer, 0, hlen); - __skb_frag_ref(frag); + __skb_frag_ref(frag, skb->pp_recycle); RX_USED_ADD(page, hlen + cp->crc_size); } diff --git a/drivers/net/veth.c b/drivers/net/veth.c index 426e68a95067..0b0293629329 100644 --- a/drivers/net/veth.c +++ b/drivers/net/veth.c @@ -717,7 +717,7 @@ static void veth_xdp_get(struct xdp_buff *xdp) return; for (i = 0; i < sinfo->nr_frags; i++) - __skb_frag_ref(&sinfo->frags[i]); + __skb_frag_ref(&sinfo->frags[i], false); } static int veth_convert_skb_to_xdp_buff(struct veth_rq *rq, diff --git a/include/linux/skbuff_ref.h b/include/linux/skbuff_ref.h index 11f0a4063403..4dcdbe9fbc5f 100644 --- a/include/linux/skbuff_ref.h +++ b/include/linux/skbuff_ref.h @@ -8,16 +8,47 @@ #define _LINUX_SKBUFF_REF_H #include +#include + +#ifdef CONFIG_PAGE_POOL +static inline bool is_pp_page(struct page *page) +{ + return (page->pp_magic & ~0x3UL) == PP_SIGNATURE; +} + +static inline bool napi_pp_get_page(struct page *page) +{ + page = compound_head(page); + + if (!is_pp_page(page)) + return false; + + page_pool_ref_page(page); + return true; +} +#endif + +static inline void skb_page_ref(struct page *page, bool recycle) +{ +#ifdef CONFIG_PAGE_POOL + if (recycle && napi_pp_get_page(page)) + return; +#endif + get_page(page); +} /** * __skb_frag_ref - take an addition reference on a paged fragment. * @frag: the paged fragment + * @recycle: skb->pp_recycle param of the parent skb. False if no parent skb. * - * Takes an additional reference on the paged fragment @frag. + * Takes an additional reference on the paged fragment @frag. Obtains the + * correct reference count depending on whether skb->pp_recycle is set and + * whether the frag is a page pool frag. */ -static inline void __skb_frag_ref(skb_frag_t *frag) +static inline void __skb_frag_ref(skb_frag_t *frag, bool recycle) { - get_page(skb_frag_page(frag)); + skb_page_ref(skb_frag_page(frag), recycle); } /** @@ -29,7 +60,7 @@ static inline void __skb_frag_ref(skb_frag_t *frag) */ static inline void skb_frag_ref(struct sk_buff *skb, int f) { - __skb_frag_ref(&skb_shinfo(skb)->frags[f]); + __skb_frag_ref(&skb_shinfo(skb)->frags[f], skb->pp_recycle); } bool napi_pp_put_page(struct page *page); diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 2554a6f5f386..ea052fa710d8 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -907,11 +907,6 @@ static void skb_clone_fraglist(struct sk_buff *skb) skb_get(list); } -static bool is_pp_page(struct page *page) -{ - return (page->pp_magic & ~0x3UL) == PP_SIGNATURE; -} - int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb, unsigned int headroom) { @@ -1033,37 +1028,6 @@ static bool skb_pp_recycle(struct sk_buff *skb, void *data) return napi_pp_put_page(virt_to_page(data)); } -/** - * skb_pp_frag_ref() - Increase fragment references of a page pool aware skb - * @skb: page pool aware skb - * - * Increase the fragment reference count (pp_ref_count) of a skb. This is - * intended to gain fragment references only for page pool aware skbs, - * i.e. when skb->pp_recycle is true, and not for fragments in a - * non-pp-recycling skb. It has a fallback to increase references on normal - * pages, as page pool aware skbs may also have normal page fragments. - */ -static int skb_pp_frag_ref(struct sk_buff *skb) -{ - struct skb_shared_info *shinfo; - struct page *head_page; - int i; - - if (!skb->pp_recycle) - return -EINVAL; - - shinfo = skb_shinfo(skb); - - for (i = 0; i < shinfo->nr_frags; i++) { - head_page = compound_head(skb_frag_page(&shinfo->frags[i])); - if (likely(is_pp_page(head_page))) - page_pool_ref_page(head_page); - else - page_ref_inc(head_page); - } - return 0; -} - static void skb_kfree_head(void *head, unsigned int end_offset) { if (end_offset == SKB_SMALL_HEAD_HEADROOM) @@ -4176,7 +4140,7 @@ int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen) to++; } else { - __skb_frag_ref(fragfrom); + __skb_frag_ref(fragfrom, skb->pp_recycle); skb_frag_page_copy(fragto, fragfrom); skb_frag_off_copy(fragto, fragfrom); skb_frag_size_set(fragto, todo); @@ -4826,7 +4790,7 @@ normal: } *nskb_frag = (i < 0) ? skb_head_frag_to_page_desc(frag_skb) : *frag; - __skb_frag_ref(nskb_frag); + __skb_frag_ref(nskb_frag, nskb->pp_recycle); size = skb_frag_size(nskb_frag); if (pos < offset) { @@ -5957,10 +5921,8 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, /* if the skb is not cloned this does nothing * since we set nr_frags to 0. */ - if (skb_pp_frag_ref(from)) { - for (i = 0; i < from_shinfo->nr_frags; i++) - __skb_frag_ref(&from_shinfo->frags[i]); - } + for (i = 0; i < from_shinfo->nr_frags; i++) + __skb_frag_ref(&from_shinfo->frags[i], from->pp_recycle); to->truesize += delta; to->len += len; diff --git a/net/tls/tls_device_fallback.c b/net/tls/tls_device_fallback.c index f9e3d3d90dcf..9237dded4467 100644 --- a/net/tls/tls_device_fallback.c +++ b/net/tls/tls_device_fallback.c @@ -278,7 +278,7 @@ static int fill_sg_in(struct scatterlist *sg_in, for (i = 0; remaining > 0; i++) { skb_frag_t *frag = &record->frags[i]; - __skb_frag_ref(frag); + __skb_frag_ref(frag, false); sg_set_page(sg_in + i, skb_frag_page(frag), skb_frag_size(frag), skb_frag_off(frag)); -- cgit v1.2.3-59-g8ed1b From 4d0470b9ad73e965f5a1e52f1deb0edbb6d03c89 Mon Sep 17 00:00:00 2001 From: Jason Xing Date: Fri, 12 Apr 2024 11:07:18 +0800 Subject: net: save some cycles when doing skb_attempt_defer_free() Normally, we don't face these two exceptions very often meanwhile we have some chance to meet the condition where the current cpu id is the same as skb->alloc_cpu. One simple test that can help us see the frequency of this statement 'cpu == raw_smp_processor_id()': 1. running iperf -s and iperf -c [ip] -P [MAX CPU] 2. using BPF to capture skb_attempt_defer_free() I can see around 4% chance that happens to satisfy the statement. So moving this statement at the beginning can save some cycles in most cases. Signed-off-by: Jason Xing Reviewed-by: Alexander Lobakin Signed-off-by: David S. Miller --- net/core/skbuff.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/net/core/skbuff.c b/net/core/skbuff.c index ea052fa710d8..37c858dc11a6 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -6965,9 +6965,9 @@ void skb_attempt_defer_free(struct sk_buff *skb) unsigned int defer_max; bool kick; - if (WARN_ON_ONCE(cpu >= nr_cpu_ids) || - !cpu_online(cpu) || - cpu == raw_smp_processor_id()) { + if (cpu == raw_smp_processor_id() || + WARN_ON_ONCE(cpu >= nr_cpu_ids) || + !cpu_online(cpu)) { nodefer: kfree_skb_napi_cache(skb); return; } -- cgit v1.2.3-59-g8ed1b From 7ab4f16f9e2440e797eae88812f800458e5879d2 Mon Sep 17 00:00:00 2001 From: Pavel Begunkov Date: Fri, 19 Apr 2024 12:08:39 +0100 Subject: net: extend ubuf_info callback to ops structure We'll need to associate additional callbacks with ubuf_info, introduce a structure holding ubuf_info callbacks. Apart from a more smarter io_uring notification management introduced in next patches, it can be used to generalise msg_zerocopy_put_abort() and also store ->sg_from_iter, which is currently passed in struct msghdr. Reviewed-by: Jens Axboe Reviewed-by: David Ahern Signed-off-by: Pavel Begunkov Reviewed-by: Willem de Bruijn Link: https://lore.kernel.org/all/a62015541de49c0e2a8a0377a1d5d0a5aeb07016.1713369317.git.asml.silence@gmail.com Signed-off-by: Jakub Kicinski --- drivers/net/tap.c | 2 +- drivers/net/tun.c | 2 +- drivers/net/xen-netback/common.h | 5 ++--- drivers/net/xen-netback/interface.c | 2 +- drivers/net/xen-netback/netback.c | 11 ++++++++--- drivers/vhost/net.c | 8 ++++++-- include/linux/skbuff.h | 19 +++++++++++-------- io_uring/notif.c | 18 +++++++++++++----- net/core/skbuff.c | 16 ++++++++++------ 9 files changed, 53 insertions(+), 30 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/drivers/net/tap.c b/drivers/net/tap.c index 9f0495e8df4d..bfdd3875fe86 100644 --- a/drivers/net/tap.c +++ b/drivers/net/tap.c @@ -754,7 +754,7 @@ static ssize_t tap_get_user(struct tap_queue *q, void *msg_control, skb_zcopy_init(skb, msg_control); } else if (msg_control) { struct ubuf_info *uarg = msg_control; - uarg->callback(NULL, uarg, false); + uarg->ops->complete(NULL, uarg, false); } dev_queue_xmit(skb); diff --git a/drivers/net/tun.c b/drivers/net/tun.c index 0b3f21cba552..b7401d990680 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -1906,7 +1906,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, skb_zcopy_init(skb, msg_control); } else if (msg_control) { struct ubuf_info *uarg = msg_control; - uarg->callback(NULL, uarg, false); + uarg->ops->complete(NULL, uarg, false); } skb_reset_network_header(skb); diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h index 1fcbd83f7ff2..17421da139f2 100644 --- a/drivers/net/xen-netback/common.h +++ b/drivers/net/xen-netback/common.h @@ -390,9 +390,8 @@ bool xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb); void xenvif_carrier_on(struct xenvif *vif); -/* Callback from stack when TX packet can be released */ -void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf, - bool zerocopy_success); +/* Callbacks from stack when TX packet can be released */ +extern const struct ubuf_info_ops xenvif_ubuf_ops; static inline pending_ring_idx_t nr_pending_reqs(struct xenvif_queue *queue) { diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c index 7cff90aa8d24..65db5f14465f 100644 --- a/drivers/net/xen-netback/interface.c +++ b/drivers/net/xen-netback/interface.c @@ -593,7 +593,7 @@ int xenvif_init_queue(struct xenvif_queue *queue) for (i = 0; i < MAX_PENDING_REQS; i++) { queue->pending_tx_info[i].callback_struct = (struct ubuf_info_msgzc) - { { .callback = xenvif_zerocopy_callback }, + { { .ops = &xenvif_ubuf_ops }, { { .ctx = NULL, .desc = i } } }; queue->grant_tx_handle[i] = NETBACK_INVALID_HANDLE; diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index ef76850d9bcd..bab7e43ea05b 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -1156,7 +1156,7 @@ static int xenvif_handle_frag_list(struct xenvif_queue *queue, struct sk_buff *s uarg = skb_shinfo(skb)->destructor_arg; /* increase inflight counter to offset decrement in callback */ atomic_inc(&queue->inflight_packets); - uarg->callback(NULL, uarg, true); + uarg->ops->complete(NULL, uarg, true); skb_shinfo(skb)->destructor_arg = NULL; /* Fill the skb with the new (local) frags. */ @@ -1278,8 +1278,9 @@ static int xenvif_tx_submit(struct xenvif_queue *queue) return work_done; } -void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf_base, - bool zerocopy_success) +static void xenvif_zerocopy_callback(struct sk_buff *skb, + struct ubuf_info *ubuf_base, + bool zerocopy_success) { unsigned long flags; pending_ring_idx_t index; @@ -1312,6 +1313,10 @@ void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf_base, xenvif_skb_zerocopy_complete(queue); } +const struct ubuf_info_ops xenvif_ubuf_ops = { + .complete = xenvif_zerocopy_callback, +}; + static inline void xenvif_tx_dealloc_action(struct xenvif_queue *queue) { struct gnttab_unmap_grant_ref *gop; diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index c64ded183f8d..f16279351db5 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -380,7 +380,7 @@ static void vhost_zerocopy_signal_used(struct vhost_net *net, } } -static void vhost_zerocopy_callback(struct sk_buff *skb, +static void vhost_zerocopy_complete(struct sk_buff *skb, struct ubuf_info *ubuf_base, bool success) { struct ubuf_info_msgzc *ubuf = uarg_to_msgzc(ubuf_base); @@ -408,6 +408,10 @@ static void vhost_zerocopy_callback(struct sk_buff *skb, rcu_read_unlock_bh(); } +static const struct ubuf_info_ops vhost_ubuf_ops = { + .complete = vhost_zerocopy_complete, +}; + static inline unsigned long busy_clock(void) { return local_clock() >> 10; @@ -879,7 +883,7 @@ static void handle_tx_zerocopy(struct vhost_net *net, struct socket *sock) vq->heads[nvq->upend_idx].len = VHOST_DMA_IN_PROGRESS; ubuf->ctx = nvq->ubufs; ubuf->desc = nvq->upend_idx; - ubuf->ubuf.callback = vhost_zerocopy_callback; + ubuf->ubuf.ops = &vhost_ubuf_ops; ubuf->ubuf.flags = SKBFL_ZEROCOPY_FRAG; refcount_set(&ubuf->ubuf.refcnt, 1); msg.msg_control = &ctl; diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index 9d24aec064e8..a110e97e074a 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -527,6 +527,11 @@ enum { #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \ SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS) +struct ubuf_info_ops { + void (*complete)(struct sk_buff *, struct ubuf_info *, + bool zerocopy_success); +}; + /* * The callback notifies userspace to release buffers when skb DMA is done in * lower device, the skb last reference should be 0 when calling this. @@ -536,8 +541,7 @@ enum { * The desc field is used to track userspace buffer index. */ struct ubuf_info { - void (*callback)(struct sk_buff *, struct ubuf_info *, - bool zerocopy_success); + const struct ubuf_info_ops *ops; refcount_t refcnt; u8 flags; }; @@ -1662,14 +1666,13 @@ static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) } #endif +extern const struct ubuf_info_ops msg_zerocopy_ubuf_ops; + struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, struct ubuf_info *uarg); void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); -void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, - bool success); - int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk, struct sk_buff *skb, struct iov_iter *from, size_t length); @@ -1757,13 +1760,13 @@ static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) static inline void net_zcopy_put(struct ubuf_info *uarg) { if (uarg) - uarg->callback(NULL, uarg, true); + uarg->ops->complete(NULL, uarg, true); } static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) { if (uarg) { - if (uarg->callback == msg_zerocopy_callback) + if (uarg->ops == &msg_zerocopy_ubuf_ops) msg_zerocopy_put_abort(uarg, have_uref); else if (have_uref) net_zcopy_put(uarg); @@ -1777,7 +1780,7 @@ static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) if (uarg) { if (!skb_zcopy_is_nouarg(skb)) - uarg->callback(skb, uarg, zerocopy_success); + uarg->ops->complete(skb, uarg, zerocopy_success); skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY; } diff --git a/io_uring/notif.c b/io_uring/notif.c index d3e703c37aba..204f0fc7d966 100644 --- a/io_uring/notif.c +++ b/io_uring/notif.c @@ -24,7 +24,7 @@ static void io_notif_complete_tw_ext(struct io_kiocb *notif, struct io_tw_state io_req_task_complete(notif, ts); } -static void io_tx_ubuf_callback(struct sk_buff *skb, struct ubuf_info *uarg, +static void io_tx_ubuf_complete(struct sk_buff *skb, struct ubuf_info *uarg, bool success) { struct io_notif_data *nd = container_of(uarg, struct io_notif_data, uarg); @@ -45,19 +45,27 @@ static void io_tx_ubuf_callback_ext(struct sk_buff *skb, struct ubuf_info *uarg, else if (!success && !nd->zc_copied) WRITE_ONCE(nd->zc_copied, true); } - io_tx_ubuf_callback(skb, uarg, success); + io_tx_ubuf_complete(skb, uarg, success); } +static const struct ubuf_info_ops io_ubuf_ops = { + .complete = io_tx_ubuf_complete, +}; + +static const struct ubuf_info_ops io_ubuf_ops_ext = { + .complete = io_tx_ubuf_callback_ext, +}; + void io_notif_set_extended(struct io_kiocb *notif) { struct io_notif_data *nd = io_notif_to_data(notif); - if (nd->uarg.callback != io_tx_ubuf_callback_ext) { + if (nd->uarg.ops != &io_ubuf_ops_ext) { nd->account_pages = 0; nd->zc_report = false; nd->zc_used = false; nd->zc_copied = false; - nd->uarg.callback = io_tx_ubuf_callback_ext; + nd->uarg.ops = &io_ubuf_ops_ext; notif->io_task_work.func = io_notif_complete_tw_ext; } } @@ -80,7 +88,7 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx) nd = io_notif_to_data(notif); nd->uarg.flags = IO_NOTIF_UBUF_FLAGS; - nd->uarg.callback = io_tx_ubuf_callback; + nd->uarg.ops = &io_ubuf_ops; refcount_set(&nd->uarg.refcnt, 1); return notif; } diff --git a/net/core/skbuff.c b/net/core/skbuff.c index b99127712e67..9d36750459f3 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -1708,7 +1708,7 @@ static struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size) return NULL; } - uarg->ubuf.callback = msg_zerocopy_callback; + uarg->ubuf.ops = &msg_zerocopy_ubuf_ops; uarg->id = ((u32)atomic_inc_return(&sk->sk_zckey)) - 1; uarg->len = 1; uarg->bytelen = size; @@ -1734,7 +1734,7 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, u32 bytelen, next; /* there might be non MSG_ZEROCOPY users */ - if (uarg->callback != msg_zerocopy_callback) + if (uarg->ops != &msg_zerocopy_ubuf_ops) return NULL; /* realloc only when socket is locked (TCP, UDP cork), @@ -1845,8 +1845,8 @@ release: sock_put(sk); } -void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, - bool success) +static void msg_zerocopy_complete(struct sk_buff *skb, struct ubuf_info *uarg, + bool success) { struct ubuf_info_msgzc *uarg_zc = uarg_to_msgzc(uarg); @@ -1855,7 +1855,6 @@ void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg, if (refcount_dec_and_test(&uarg->refcnt)) __msg_zerocopy_callback(uarg_zc); } -EXPORT_SYMBOL_GPL(msg_zerocopy_callback); void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref) { @@ -1865,10 +1864,15 @@ void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref) uarg_to_msgzc(uarg)->len--; if (have_uref) - msg_zerocopy_callback(NULL, uarg, true); + msg_zerocopy_complete(NULL, uarg, true); } EXPORT_SYMBOL_GPL(msg_zerocopy_put_abort); +const struct ubuf_info_ops msg_zerocopy_ubuf_ops = { + .complete = msg_zerocopy_complete, +}; +EXPORT_SYMBOL_GPL(msg_zerocopy_ubuf_ops); + int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, struct msghdr *msg, int len, struct ubuf_info *uarg) -- cgit v1.2.3-59-g8ed1b From 65bada80dec1f2108a751644773b2120bd789934 Mon Sep 17 00:00:00 2001 From: Pavel Begunkov Date: Fri, 19 Apr 2024 12:08:40 +0100 Subject: net: add callback for setting a ubuf_info to skb At the moment an skb can only have one ubuf_info associated with it, which might be a performance problem for zerocopy sends in cases like TCP via io_uring. Add a callback for assigning ubuf_info to skb, this way we will implement smarter assignment later like linking ubuf_info together. Note, it's an optional callback, which should be compatible with skb_zcopy_set(), that's because the net stack might potentially decide to clone an skb and take another reference to ubuf_info whenever it wishes. Also, a correct implementation should always be able to bind to an skb without prior ubuf_info, otherwise we could end up in a situation when the send would not be able to progress. Reviewed-by: Jens Axboe Reviewed-by: David Ahern Signed-off-by: Pavel Begunkov Reviewed-by: Willem de Bruijn Link: https://lore.kernel.org/all/b7918aadffeb787c84c9e72e34c729dc04f3a45d.1713369317.git.asml.silence@gmail.com Signed-off-by: Jakub Kicinski --- include/linux/skbuff.h | 2 ++ net/core/skbuff.c | 20 ++++++++++++++------ 2 files changed, 16 insertions(+), 6 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index a110e97e074a..ced69f37977f 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -530,6 +530,8 @@ enum { struct ubuf_info_ops { void (*complete)(struct sk_buff *, struct ubuf_info *, bool zerocopy_success); + /* has to be compatible with skb_zcopy_set() */ + int (*link_skb)(struct sk_buff *skb, struct ubuf_info *uarg); }; /* diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 9d36750459f3..4a0c10685cd2 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -1880,11 +1880,18 @@ int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, struct ubuf_info *orig_uarg = skb_zcopy(skb); int err, orig_len = skb->len; - /* An skb can only point to one uarg. This edge case happens when - * TCP appends to an skb, but zerocopy_realloc triggered a new alloc. - */ - if (orig_uarg && uarg != orig_uarg) - return -EEXIST; + if (uarg->ops->link_skb) { + err = uarg->ops->link_skb(skb, uarg); + if (err) + return err; + } else { + /* An skb can only point to one uarg. This edge case happens + * when TCP appends to an skb, but zerocopy_realloc triggered + * a new alloc. + */ + if (orig_uarg && uarg != orig_uarg) + return -EEXIST; + } err = __zerocopy_sg_from_iter(msg, sk, skb, &msg->msg_iter, len); if (err == -EFAULT || (err == -EMSGSIZE && skb->len == orig_len)) { @@ -1898,7 +1905,8 @@ int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, return err; } - skb_zcopy_set(skb, uarg, NULL); + if (!uarg->ops->link_skb) + skb_zcopy_set(skb, uarg, NULL); return skb->len - orig_len; } EXPORT_SYMBOL_GPL(skb_zerocopy_iter_stream); -- cgit v1.2.3-59-g8ed1b From a86a0661b86f310c0b73a30c829648864f0b2619 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 29 Apr 2024 13:40:21 +0000 Subject: net: move sysctl_max_skb_frags to net_hotdata sysctl_max_skb_frags is used in TCP and MPTCP fast paths, move it to net_hodata for better cache locality. Signed-off-by: Eric Dumazet Reviewed-by: David Ahern Link: https://lore.kernel.org/r/20240429134025.1233626-2-edumazet@google.com Signed-off-by: Jakub Kicinski --- include/linux/skbuff.h | 2 -- include/net/hotdata.h | 1 + net/core/hotdata.c | 1 + net/core/skbuff.c | 5 +---- net/core/sysctl_net_core.c | 2 +- net/ipv4/tcp.c | 3 ++- net/mptcp/protocol.c | 3 ++- 7 files changed, 8 insertions(+), 9 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index adf75d69770c..36b133f04d30 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -353,8 +353,6 @@ struct sk_buff; #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS -extern int sysctl_max_skb_frags; - /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to * segment using its current segmentation instead. */ diff --git a/include/net/hotdata.h b/include/net/hotdata.h index 003667a1efd6..a6cff6590426 100644 --- a/include/net/hotdata.h +++ b/include/net/hotdata.h @@ -38,6 +38,7 @@ struct net_hotdata { int max_backlog; int dev_tx_weight; int dev_rx_weight; + int sysctl_max_skb_frags; }; #define inet_ehash_secret net_hotdata.tcp_protocol.secret diff --git a/net/core/hotdata.c b/net/core/hotdata.c index c8a7a451c18a..f17cbb4807b9 100644 --- a/net/core/hotdata.c +++ b/net/core/hotdata.c @@ -18,5 +18,6 @@ struct net_hotdata net_hotdata __cacheline_aligned = { .max_backlog = 1000, .dev_tx_weight = 64, .dev_rx_weight = 64, + .sysctl_max_skb_frags = MAX_SKB_FRAGS, }; EXPORT_SYMBOL(net_hotdata); diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 0c8b82750000..65779b8f0b12 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -109,9 +109,6 @@ static struct kmem_cache *skbuff_ext_cache __ro_after_init; #define SKB_SMALL_HEAD_HEADROOM \ SKB_WITH_OVERHEAD(SKB_SMALL_HEAD_CACHE_SIZE) -int sysctl_max_skb_frags __read_mostly = MAX_SKB_FRAGS; -EXPORT_SYMBOL(sysctl_max_skb_frags); - /* kcm_write_msgs() relies on casting paged frags to bio_vec to use * iov_iter_bvec(). These static asserts ensure the cast is valid is long as the * netmem is a page. @@ -7040,7 +7037,7 @@ static void skb_splice_csum_page(struct sk_buff *skb, struct page *page, ssize_t skb_splice_from_iter(struct sk_buff *skb, struct iov_iter *iter, ssize_t maxsize, gfp_t gfp) { - size_t frag_limit = READ_ONCE(sysctl_max_skb_frags); + size_t frag_limit = READ_ONCE(net_hotdata.sysctl_max_skb_frags); struct page *pages[8], **ppages = pages; ssize_t spliced = 0, ret = 0; unsigned int i; diff --git a/net/core/sysctl_net_core.c b/net/core/sysctl_net_core.c index 903ab4a51c17..e75375d54b9e 100644 --- a/net/core/sysctl_net_core.c +++ b/net/core/sysctl_net_core.c @@ -595,7 +595,7 @@ static struct ctl_table net_core_table[] = { }, { .procname = "max_skb_frags", - .data = &sysctl_max_skb_frags, + .data = &net_hotdata.sysctl_max_skb_frags, .maxlen = sizeof(int), .mode = 0644, .proc_handler = proc_dointvec_minmax, diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 4ec0f4feee00..388f6e115bf1 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -280,6 +280,7 @@ #include #include #include +#include #include /* Track pending CMSGs. */ @@ -1188,7 +1189,7 @@ new_segment: if (!skb_can_coalesce(skb, i, pfrag->page, pfrag->offset)) { - if (i >= READ_ONCE(sysctl_max_skb_frags)) { + if (i >= READ_ONCE(net_hotdata.sysctl_max_skb_frags)) { tcp_mark_push(tp, skb); goto new_segment; } diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 4b13ca362efa..aff17597e6a7 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -20,6 +20,7 @@ #include #endif #include +#include #include #include #include "protocol.h" @@ -1272,7 +1273,7 @@ static int mptcp_sendmsg_frag(struct sock *sk, struct sock *ssk, i = skb_shinfo(skb)->nr_frags; can_coalesce = skb_can_coalesce(skb, i, dfrag->page, offset); - if (!can_coalesce && i >= READ_ONCE(sysctl_max_skb_frags)) { + if (!can_coalesce && i >= READ_ONCE(net_hotdata.sysctl_max_skb_frags)) { tcp_mark_push(tcp_sk(ssk), skb); goto alloc_skb; } -- cgit v1.2.3-59-g8ed1b From d480dc76d9f8a72671a6f9a7d987d2389c2e2ee4 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 29 Apr 2024 13:40:22 +0000 Subject: net: move sysctl_skb_defer_max to net_hotdata sysctl_skb_defer_max is used in TCP fast path, move it to net_hodata. Signed-off-by: Eric Dumazet Reviewed-by: David Ahern Link: https://lore.kernel.org/r/20240429134025.1233626-3-edumazet@google.com Signed-off-by: Jakub Kicinski --- include/net/hotdata.h | 1 + net/core/dev.c | 1 - net/core/dev.h | 1 - net/core/hotdata.c | 1 + net/core/skbuff.c | 2 +- net/core/sysctl_net_core.c | 2 +- 6 files changed, 4 insertions(+), 4 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/include/net/hotdata.h b/include/net/hotdata.h index a6cff6590426..290499f72e18 100644 --- a/include/net/hotdata.h +++ b/include/net/hotdata.h @@ -39,6 +39,7 @@ struct net_hotdata { int dev_tx_weight; int dev_rx_weight; int sysctl_max_skb_frags; + int sysctl_skb_defer_max; }; #define inet_ehash_secret net_hotdata.tcp_protocol.secret diff --git a/net/core/dev.c b/net/core/dev.c index c9e59eff8ec8..cd7ba50eac15 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -4450,7 +4450,6 @@ EXPORT_SYMBOL(__dev_direct_xmit); *************************************************************************/ static DEFINE_PER_CPU(struct task_struct *, backlog_napi); -unsigned int sysctl_skb_defer_max __read_mostly = 64; int weight_p __read_mostly = 64; /* old backlog weight */ int dev_weight_rx_bias __read_mostly = 1; /* bias for backlog weight */ int dev_weight_tx_bias __read_mostly = 1; /* bias for output_queue quota */ diff --git a/net/core/dev.h b/net/core/dev.h index 8572d2c8dc4a..b7b518bc2be5 100644 --- a/net/core/dev.h +++ b/net/core/dev.h @@ -36,7 +36,6 @@ int dev_addr_init(struct net_device *dev); void dev_addr_check(struct net_device *dev); /* sysctls not referred to from outside net/core/ */ -extern unsigned int sysctl_skb_defer_max; extern int netdev_unregister_timeout_secs; extern int weight_p; extern int dev_weight_rx_bias; diff --git a/net/core/hotdata.c b/net/core/hotdata.c index f17cbb4807b9..a359ff160d54 100644 --- a/net/core/hotdata.c +++ b/net/core/hotdata.c @@ -19,5 +19,6 @@ struct net_hotdata net_hotdata __cacheline_aligned = { .dev_tx_weight = 64, .dev_rx_weight = 64, .sysctl_max_skb_frags = MAX_SKB_FRAGS, + .sysctl_skb_defer_max = 64, }; EXPORT_SYMBOL(net_hotdata); diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 65779b8f0b12..5f382e94b4d1 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -6985,7 +6985,7 @@ nodefer: kfree_skb_napi_cache(skb); DEBUG_NET_WARN_ON_ONCE(skb->destructor); sd = &per_cpu(softnet_data, cpu); - defer_max = READ_ONCE(sysctl_skb_defer_max); + defer_max = READ_ONCE(net_hotdata.sysctl_skb_defer_max); if (READ_ONCE(sd->defer_count) >= defer_max) goto nodefer; diff --git a/net/core/sysctl_net_core.c b/net/core/sysctl_net_core.c index e75375d54b9e..118c78615543 100644 --- a/net/core/sysctl_net_core.c +++ b/net/core/sysctl_net_core.c @@ -654,7 +654,7 @@ static struct ctl_table net_core_table[] = { }, { .procname = "skb_defer_max", - .data = &sysctl_skb_defer_max, + .data = &net_hotdata.sysctl_skb_defer_max, .maxlen = sizeof(unsigned int), .mode = 0644, .proc_handler = proc_dointvec_minmax, -- cgit v1.2.3-59-g8ed1b From d091e579b864fa790dd6a0cd537a22c383126681 Mon Sep 17 00:00:00 2001 From: Felix Fietkau Date: Sat, 27 Apr 2024 20:24:19 +0200 Subject: net: core: reject skb_copy(_expand) for fraglist GSO skbs SKB_GSO_FRAGLIST skbs must not be linearized, otherwise they become invalid. Return NULL if such an skb is passed to skb_copy or skb_copy_expand, in order to prevent a crash on a potential later call to skb_gso_segment. Fixes: 3a1296a38d0c ("net: Support GRO/GSO fraglist chaining.") Signed-off-by: Felix Fietkau Signed-off-by: David S. Miller --- net/core/skbuff.c | 27 +++++++++++++++++++-------- 1 file changed, 19 insertions(+), 8 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/net/core/skbuff.c b/net/core/skbuff.c index b99127712e67..4096e679f61c 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -2123,11 +2123,17 @@ static inline int skb_alloc_rx_flag(const struct sk_buff *skb) struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t gfp_mask) { - int headerlen = skb_headroom(skb); - unsigned int size = skb_end_offset(skb) + skb->data_len; - struct sk_buff *n = __alloc_skb(size, gfp_mask, - skb_alloc_rx_flag(skb), NUMA_NO_NODE); + struct sk_buff *n; + unsigned int size; + int headerlen; + + if (WARN_ON_ONCE(skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST)) + return NULL; + headerlen = skb_headroom(skb); + size = skb_end_offset(skb) + skb->data_len; + n = __alloc_skb(size, gfp_mask, + skb_alloc_rx_flag(skb), NUMA_NO_NODE); if (!n) return NULL; @@ -2455,12 +2461,17 @@ struct sk_buff *skb_copy_expand(const struct sk_buff *skb, /* * Allocate the copy buffer */ - struct sk_buff *n = __alloc_skb(newheadroom + skb->len + newtailroom, - gfp_mask, skb_alloc_rx_flag(skb), - NUMA_NO_NODE); - int oldheadroom = skb_headroom(skb); int head_copy_len, head_copy_off; + struct sk_buff *n; + int oldheadroom; + + if (WARN_ON_ONCE(skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST)) + return NULL; + oldheadroom = skb_headroom(skb); + n = __alloc_skb(newheadroom + skb->len + newtailroom, + gfp_mask, skb_alloc_rx_flag(skb), + NUMA_NO_NODE); if (!n) return NULL; -- cgit v1.2.3-59-g8ed1b From 173e7622ccb3f46834bd4176ed363f435e142942 Mon Sep 17 00:00:00 2001 From: Mina Almasry Date: Thu, 2 May 2024 10:54:22 -0700 Subject: Revert "net: mirror skb frag ref/unref helpers" This reverts commit a580ea994fd37f4105028f5a85c38ff6508a2b25. This revert is to resolve Dragos's report of page_pool leak here: https://lore.kernel.org/lkml/20240424165646.1625690-2-dtatulea@nvidia.com/ The reverted patch interacts very badly with commit 2cc3aeb5eccc ("skbuff: Fix a potential race while recycling page_pool packets"). The reverted commit hopes that the pp_recycle + is_pp_page variables do not change between the skb_frag_ref and skb_frag_unref operation. If such a change occurs, the skb_frag_ref/unref will not operate on the same reference type. In the case of Dragos's report, the grabbed ref was a pp ref, but the unref was a page ref, because the pp_recycle setting on the skb was changed. Attempting to fix this issue on the fly is risky. Lets revert and I hope to reland this with better understanding and testing to ensure we don't regress some edge case while streamlining skb reffing. Fixes: a580ea994fd3 ("net: mirror skb frag ref/unref helpers") Reported-by: Dragos Tatulea Signed-off-by: Mina Almasry Link: https://lore.kernel.org/r/20240502175423.2456544-1-almasrymina@google.com Signed-off-by: Jakub Kicinski --- .../chelsio/inline_crypto/ch_ktls/chcr_ktls.c | 2 +- drivers/net/ethernet/sun/cassini.c | 4 +- drivers/net/veth.c | 2 +- include/linux/skbuff_ref.h | 39 ++---------------- net/core/skbuff.c | 46 ++++++++++++++++++++-- net/tls/tls_device_fallback.c | 2 +- 6 files changed, 51 insertions(+), 44 deletions(-) (limited to 'net/core/skbuff.c') diff --git a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c index 3832c2e8ea5a..e8e460a92e0e 100644 --- a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c +++ b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c @@ -1659,7 +1659,7 @@ static void chcr_ktls_copy_record_in_skb(struct sk_buff *nskb, for (i = 0; i < record->num_frags; i++) { skb_shinfo(nskb)->frags[i] = record->frags[i]; /* increase the frag ref count */ - __skb_frag_ref(&skb_shinfo(nskb)->frags[i], nskb->pp_recycle); + __skb_frag_ref(&skb_shinfo(nskb)->frags[i]); } skb_shinfo(nskb)->nr_frags = record->num_frags; diff --git a/drivers/net/ethernet/sun/cassini.c b/drivers/net/ethernet/sun/cassini.c index f058e154a3bc..8f1f43dbb76d 100644 --- a/drivers/net/ethernet/sun/cassini.c +++ b/drivers/net/ethernet/sun/cassini.c @@ -2000,7 +2000,7 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, skb->len += hlen - swivel; skb_frag_fill_page_desc(frag, page->buffer, off, hlen - swivel); - __skb_frag_ref(frag, skb->pp_recycle); + __skb_frag_ref(frag); /* any more data? */ if ((words[0] & RX_COMP1_SPLIT_PKT) && ((dlen -= hlen) > 0)) { @@ -2024,7 +2024,7 @@ static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc, frag++; skb_frag_fill_page_desc(frag, page->buffer, 0, hlen); - __skb_frag_ref(frag, skb->pp_recycle); + __skb_frag_ref(frag); RX_USED_ADD(page, hlen + cp->crc_size); } diff --git a/drivers/net/veth.c b/drivers/net/veth.c index 0b0293629329..426e68a95067 100644 --- a/drivers/net/veth.c +++ b/drivers/net/veth.c @@ -717,7 +717,7 @@ static void veth_xdp_get(struct xdp_buff *xdp) return; for (i = 0; i < sinfo->nr_frags; i++) - __skb_frag_ref(&sinfo->frags[i], false); + __skb_frag_ref(&sinfo->frags[i]); } static int veth_convert_skb_to_xdp_buff(struct veth_rq *rq, diff --git a/include/linux/skbuff_ref.h b/include/linux/skbuff_ref.h index 4dcdbe9fbc5f..11f0a4063403 100644 --- a/include/linux/skbuff_ref.h +++ b/include/linux/skbuff_ref.h @@ -8,47 +8,16 @@ #define _LINUX_SKBUFF_REF_H #include -#include - -#ifdef CONFIG_PAGE_POOL -static inline bool is_pp_page(struct page *page) -{ - return (page->pp_magic & ~0x3UL) == PP_SIGNATURE; -} - -static inline bool napi_pp_get_page(struct page *page) -{ - page = compound_head(page); - - if (!is_pp_page(page)) - return false; - - page_pool_ref_page(page); - return true; -} -#endif - -static inline void skb_page_ref(struct page *page, bool recycle) -{ -#ifdef CONFIG_PAGE_POOL - if (recycle && napi_pp_get_page(page)) - return; -#endif - get_page(page); -} /** * __skb_frag_ref - take an addition reference on a paged fragment. * @frag: the paged fragment - * @recycle: skb->pp_recycle param of the parent skb. False if no parent skb. * - * Takes an additional reference on the paged fragment @frag. Obtains the - * correct reference count depending on whether skb->pp_recycle is set and - * whether the frag is a page pool frag. + * Takes an additional reference on the paged fragment @frag. */ -static inline void __skb_frag_ref(skb_frag_t *frag, bool recycle) +static inline void __skb_frag_ref(skb_frag_t *frag) { - skb_page_ref(skb_frag_page(frag), recycle); + get_page(skb_frag_page(frag)); } /** @@ -60,7 +29,7 @@ static inline void __skb_frag_ref(skb_frag_t *frag, bool recycle) */ static inline void skb_frag_ref(struct sk_buff *skb, int f) { - __skb_frag_ref(&skb_shinfo(skb)->frags[f], skb->pp_recycle); + __skb_frag_ref(&skb_shinfo(skb)->frags[f]); } bool napi_pp_put_page(struct page *page); diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 28cd640a6ea9..466999a7515e 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -904,6 +904,11 @@ static void skb_clone_fraglist(struct sk_buff *skb) skb_get(list); } +static bool is_pp_page(struct page *page) +{ + return (page->pp_magic & ~0x3UL) == PP_SIGNATURE; +} + int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb, unsigned int headroom) { @@ -1025,6 +1030,37 @@ static bool skb_pp_recycle(struct sk_buff *skb, void *data) return napi_pp_put_page(virt_to_page(data)); } +/** + * skb_pp_frag_ref() - Increase fragment references of a page pool aware skb + * @skb: page pool aware skb + * + * Increase the fragment reference count (pp_ref_count) of a skb. This is + * intended to gain fragment references only for page pool aware skbs, + * i.e. when skb->pp_recycle is true, and not for fragments in a + * non-pp-recycling skb. It has a fallback to increase references on normal + * pages, as page pool aware skbs may also have normal page fragments. + */ +static int skb_pp_frag_ref(struct sk_buff *skb) +{ + struct skb_shared_info *shinfo; + struct page *head_page; + int i; + + if (!skb->pp_recycle) + return -EINVAL; + + shinfo = skb_shinfo(skb); + + for (i = 0; i < shinfo->nr_frags; i++) { + head_page = compound_head(skb_frag_page(&shinfo->frags[i])); + if (likely(is_pp_page(head_page))) + page_pool_ref_page(head_page); + else + page_ref_inc(head_page); + } + return 0; +} + static void skb_kfree_head(void *head, unsigned int end_offset) { if (end_offset == SKB_SMALL_HEAD_HEADROOM) @@ -4160,7 +4196,7 @@ int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen) to++; } else { - __skb_frag_ref(fragfrom, skb->pp_recycle); + __skb_frag_ref(fragfrom); skb_frag_page_copy(fragto, fragfrom); skb_frag_off_copy(fragto, fragfrom); skb_frag_size_set(fragto, todo); @@ -4810,7 +4846,7 @@ normal: } *nskb_frag = (i < 0) ? skb_head_frag_to_page_desc(frag_skb) : *frag; - __skb_frag_ref(nskb_frag, nskb->pp_recycle); + __skb_frag_ref(nskb_frag); size = skb_frag_size(nskb_frag); if (pos < offset) { @@ -5941,8 +5977,10 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, /* if the skb is not cloned this does nothing * since we set nr_frags to 0. */ - for (i = 0; i < from_shinfo->nr_frags; i++) - __skb_frag_ref(&from_shinfo->frags[i], from->pp_recycle); + if (skb_pp_frag_ref(from)) { + for (i = 0; i < from_shinfo->nr_frags; i++) + __skb_frag_ref(&from_shinfo->frags[i]); + } to->truesize += delta; to->len += len; diff --git a/net/tls/tls_device_fallback.c b/net/tls/tls_device_fallback.c index 9237dded4467..f9e3d3d90dcf 100644 --- a/net/tls/tls_device_fallback.c +++ b/net/tls/tls_device_fallback.c @@ -278,7 +278,7 @@ static int fill_sg_in(struct scatterlist *sg_in, for (i = 0; remaining > 0; i++) { skb_frag_t *frag = &record->frags[i]; - __skb_frag_ref(frag, false); + __skb_frag_ref(frag); sg_set_page(sg_in + i, skb_frag_page(frag), skb_frag_size(frag), skb_frag_off(frag)); -- cgit v1.2.3-59-g8ed1b