From 269009893146c495f41e9572dd9319e787c2eba9 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:52 +0100 Subject: [PATCH 01/11] xsk: recycle buffer in case Rx queue was full Add missing xsk_buff_free() call when __xsk_rcv_zc() failed to produce descriptor to XSK Rx queue. Fixes: 24ea50127ecf ("xsk: support mbuf on ZC RX") Acked-by: Magnus Karlsson Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-2-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- net/xdp/xsk.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c index 9f13aa3353e3..1eadfac03cc4 100644 --- a/net/xdp/xsk.c +++ b/net/xdp/xsk.c @@ -167,8 +167,10 @@ static int xsk_rcv_zc(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len) contd = XDP_PKT_CONTD; err = __xsk_rcv_zc(xs, xskb, len, contd); - if (err || likely(!frags)) - goto out; + if (err) + goto err; + if (likely(!frags)) + return 0; xskb_list = &xskb->pool->xskb_list; list_for_each_entry_safe(pos, tmp, xskb_list, xskb_list_node) { @@ -177,11 +179,13 @@ static int xsk_rcv_zc(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len) len = pos->xdp.data_end - pos->xdp.data; err = __xsk_rcv_zc(xs, pos, len, contd); if (err) - return err; + goto err; list_del(&pos->xskb_list_node); } -out: + return 0; +err: + xsk_buff_free(xdp); return err; } From f7f6aa8e24383fbb11ac55942e66da9660110f80 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:53 +0100 Subject: [PATCH 02/11] xsk: make xsk_buff_pool responsible for clearing xdp_buff::flags XDP multi-buffer support introduced XDP_FLAGS_HAS_FRAGS flag that is used by drivers to notify data path whether xdp_buff contains fragments or not. Data path looks up mentioned flag on first buffer that occupies the linear part of xdp_buff, so drivers only modify it there. This is sufficient for SKB and XDP_DRV modes as usually xdp_buff is allocated on stack or it resides within struct representing driver's queue and fragments are carried via skb_frag_t structs. IOW, we are dealing with only one xdp_buff. ZC mode though relies on list of xdp_buff structs that is carried via xsk_buff_pool::xskb_list, so ZC data path has to make sure that fragments do *not* have XDP_FLAGS_HAS_FRAGS set. Otherwise, xsk_buff_free() could misbehave if it would be executed against xdp_buff that carries a frag with XDP_FLAGS_HAS_FRAGS flag set. Such scenario can take place when within supplied XDP program bpf_xdp_adjust_tail() is used with negative offset that would in turn release the tail fragment from multi-buffer frame. Calling xsk_buff_free() on tail fragment with XDP_FLAGS_HAS_FRAGS would result in releasing all the nodes from xskb_list that were produced by driver before XDP program execution, which is not what is intended - only tail fragment should be deleted from xskb_list and then it should be put onto xsk_buff_pool::free_list. Such multi-buffer frame will never make it up to user space, so from AF_XDP application POV there would be no traffic running, however due to free_list getting constantly new nodes, driver will be able to feed HW Rx queue with recycled buffers. Bottom line is that instead of traffic being redirected to user space, it would be continuously dropped. To fix this, let us clear the mentioned flag on xsk_buff_pool side during xdp_buff initialization, which is what should have been done right from the start of XSK multi-buffer support. Fixes: 1bbc04de607b ("ice: xsk: add RX multi-buffer support") Fixes: 1c9ba9c14658 ("i40e: xsk: add RX multi-buffer support") Fixes: 24ea50127ecf ("xsk: support mbuf on ZC RX") Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-3-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/i40e/i40e_xsk.c | 1 - drivers/net/ethernet/intel/ice/ice_xsk.c | 1 - include/net/xdp_sock_drv.h | 1 + net/xdp/xsk_buff_pool.c | 1 + 4 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c index af7d5fa6cdc1..82aca0d16a3e 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c +++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c @@ -498,7 +498,6 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget) xdp_res = i40e_run_xdp_zc(rx_ring, first, xdp_prog); i40e_handle_xdp_result_zc(rx_ring, first, rx_desc, &rx_packets, &rx_bytes, xdp_res, &failure); - first->flags = 0; next_to_clean = next_to_process; if (failure) break; diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c index 5d1ae8e4058a..d9073a618ad6 100644 --- a/drivers/net/ethernet/intel/ice/ice_xsk.c +++ b/drivers/net/ethernet/intel/ice/ice_xsk.c @@ -895,7 +895,6 @@ int ice_clean_rx_irq_zc(struct ice_rx_ring *rx_ring, int budget) if (!first) { first = xdp; - xdp_buff_clear_frags_flag(first); } else if (ice_add_xsk_frag(rx_ring, first, xdp, size)) { break; } diff --git a/include/net/xdp_sock_drv.h b/include/net/xdp_sock_drv.h index 526c1e7f505e..9819e2af0378 100644 --- a/include/net/xdp_sock_drv.h +++ b/include/net/xdp_sock_drv.h @@ -164,6 +164,7 @@ static inline void xsk_buff_set_size(struct xdp_buff *xdp, u32 size) xdp->data = xdp->data_hard_start + XDP_PACKET_HEADROOM; xdp->data_meta = xdp->data; xdp->data_end = xdp->data + size; + xdp->flags = 0; } static inline dma_addr_t xsk_buff_raw_get_dma(struct xsk_buff_pool *pool, diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c index 28711cc44ced..ce60ecd48a4d 100644 --- a/net/xdp/xsk_buff_pool.c +++ b/net/xdp/xsk_buff_pool.c @@ -555,6 +555,7 @@ struct xdp_buff *xp_alloc(struct xsk_buff_pool *pool) xskb->xdp.data = xskb->xdp.data_hard_start + XDP_PACKET_HEADROOM; xskb->xdp.data_meta = xskb->xdp.data; + xskb->xdp.flags = 0; if (pool->dma_need_sync) { dma_sync_single_range_for_device(pool->dev, xskb->dma, 0, From c5114710c8ce86b8317e9b448f4fd15c711c2a82 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:54 +0100 Subject: [PATCH 03/11] xsk: fix usage of multi-buffer BPF helpers for ZC XDP Currently when packet is shrunk via bpf_xdp_adjust_tail() and memory type is set to MEM_TYPE_XSK_BUFF_POOL, null ptr dereference happens: [1136314.192256] BUG: kernel NULL pointer dereference, address: 0000000000000034 [1136314.203943] #PF: supervisor read access in kernel mode [1136314.213768] #PF: error_code(0x0000) - not-present page [1136314.223550] PGD 0 P4D 0 [1136314.230684] Oops: 0000 [#1] PREEMPT SMP NOPTI [1136314.239621] CPU: 8 PID: 54203 Comm: xdpsock Not tainted 6.6.0+ #257 [1136314.250469] Hardware name: Intel Corporation S2600WFT/S2600WFT, BIOS SE5C620.86B.02.01.0008.031920191559 03/19/2019 [1136314.265615] RIP: 0010:__xdp_return+0x6c/0x210 [1136314.274653] Code: ad 00 48 8b 47 08 49 89 f8 a8 01 0f 85 9b 01 00 00 0f 1f 44 00 00 f0 41 ff 48 34 75 32 4c 89 c7 e9 79 cd 80 ff 83 fe 03 75 17 41 34 01 0f 85 02 01 00 00 48 89 cf e9 22 cc 1e 00 e9 3d d2 86 [1136314.302907] RSP: 0018:ffffc900089f8db0 EFLAGS: 00010246 [1136314.312967] RAX: ffffc9003168aed0 RBX: ffff8881c3300000 RCX: 0000000000000000 [1136314.324953] RDX: 0000000000000000 RSI: 0000000000000003 RDI: ffffc9003168c000 [1136314.336929] RBP: 0000000000000ae0 R08: 0000000000000002 R09: 0000000000010000 [1136314.348844] R10: ffffc9000e495000 R11: 0000000000000040 R12: 0000000000000001 [1136314.360706] R13: 0000000000000524 R14: ffffc9003168aec0 R15: 0000000000000001 [1136314.373298] FS: 00007f8df8bbcb80(0000) GS:ffff8897e0e00000(0000) knlGS:0000000000000000 [1136314.386105] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [1136314.396532] CR2: 0000000000000034 CR3: 00000001aa912002 CR4: 00000000007706f0 [1136314.408377] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [1136314.420173] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [1136314.431890] PKRU: 55555554 [1136314.439143] Call Trace: [1136314.446058] [1136314.452465] ? __die+0x20/0x70 [1136314.459881] ? page_fault_oops+0x15b/0x440 [1136314.468305] ? exc_page_fault+0x6a/0x150 [1136314.476491] ? asm_exc_page_fault+0x22/0x30 [1136314.484927] ? __xdp_return+0x6c/0x210 [1136314.492863] bpf_xdp_adjust_tail+0x155/0x1d0 [1136314.501269] bpf_prog_ccc47ae29d3b6570_xdp_sock_prog+0x15/0x60 [1136314.511263] ice_clean_rx_irq_zc+0x206/0xc60 [ice] [1136314.520222] ? ice_xmit_zc+0x6e/0x150 [ice] [1136314.528506] ice_napi_poll+0x467/0x670 [ice] [1136314.536858] ? ttwu_do_activate.constprop.0+0x8f/0x1a0 [1136314.546010] __napi_poll+0x29/0x1b0 [1136314.553462] net_rx_action+0x133/0x270 [1136314.561619] __do_softirq+0xbe/0x28e [1136314.569303] do_softirq+0x3f/0x60 This comes from __xdp_return() call with xdp_buff argument passed as NULL which is supposed to be consumed by xsk_buff_free() call. To address this properly, in ZC case, a node that represents the frag being removed has to be pulled out of xskb_list. Introduce appropriate xsk helpers to do such node operation and use them accordingly within bpf_xdp_adjust_tail(). Fixes: 24ea50127ecf ("xsk: support mbuf on ZC RX") Acked-by: Magnus Karlsson # For the xsk header part Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-4-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- include/net/xdp_sock_drv.h | 26 +++++++++++++++++++++++ net/core/filter.c | 42 ++++++++++++++++++++++++++++++++------ 2 files changed, 62 insertions(+), 6 deletions(-) diff --git a/include/net/xdp_sock_drv.h b/include/net/xdp_sock_drv.h index 9819e2af0378..c9aec9ab6191 100644 --- a/include/net/xdp_sock_drv.h +++ b/include/net/xdp_sock_drv.h @@ -159,6 +159,23 @@ static inline struct xdp_buff *xsk_buff_get_frag(struct xdp_buff *first) return ret; } +static inline void xsk_buff_del_tail(struct xdp_buff *tail) +{ + struct xdp_buff_xsk *xskb = container_of(tail, struct xdp_buff_xsk, xdp); + + list_del(&xskb->xskb_list_node); +} + +static inline struct xdp_buff *xsk_buff_get_tail(struct xdp_buff *first) +{ + struct xdp_buff_xsk *xskb = container_of(first, struct xdp_buff_xsk, xdp); + struct xdp_buff_xsk *frag; + + frag = list_last_entry(&xskb->pool->xskb_list, struct xdp_buff_xsk, + xskb_list_node); + return &frag->xdp; +} + static inline void xsk_buff_set_size(struct xdp_buff *xdp, u32 size) { xdp->data = xdp->data_hard_start + XDP_PACKET_HEADROOM; @@ -351,6 +368,15 @@ static inline struct xdp_buff *xsk_buff_get_frag(struct xdp_buff *first) return NULL; } +static inline void xsk_buff_del_tail(struct xdp_buff *tail) +{ +} + +static inline struct xdp_buff *xsk_buff_get_tail(struct xdp_buff *first) +{ + return NULL; +} + static inline void xsk_buff_set_size(struct xdp_buff *xdp, u32 size) { } diff --git a/net/core/filter.c b/net/core/filter.c index 24061f29c9dd..36fb5ae8af69 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -83,6 +83,7 @@ #include #include #include +#include #include "dev.h" @@ -4096,6 +4097,40 @@ static int bpf_xdp_frags_increase_tail(struct xdp_buff *xdp, int offset) return 0; } +static void bpf_xdp_shrink_data_zc(struct xdp_buff *xdp, int shrink, + struct xdp_mem_info *mem_info, bool release) +{ + struct xdp_buff *zc_frag = xsk_buff_get_tail(xdp); + + if (release) { + xsk_buff_del_tail(zc_frag); + __xdp_return(NULL, mem_info, false, zc_frag); + } else { + zc_frag->data_end -= shrink; + } +} + +static bool bpf_xdp_shrink_data(struct xdp_buff *xdp, skb_frag_t *frag, + int shrink) +{ + struct xdp_mem_info *mem_info = &xdp->rxq->mem; + bool release = skb_frag_size(frag) == shrink; + + if (mem_info->type == MEM_TYPE_XSK_BUFF_POOL) { + bpf_xdp_shrink_data_zc(xdp, shrink, mem_info, release); + goto out; + } + + if (release) { + struct page *page = skb_frag_page(frag); + + __xdp_return(page_address(page), mem_info, false, NULL); + } + +out: + return release; +} + static int bpf_xdp_frags_shrink_tail(struct xdp_buff *xdp, int offset) { struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); @@ -4110,12 +4145,7 @@ static int bpf_xdp_frags_shrink_tail(struct xdp_buff *xdp, int offset) len_free += shrink; offset -= shrink; - - if (skb_frag_size(frag) == shrink) { - struct page *page = skb_frag_page(frag); - - __xdp_return(page_address(page), &xdp->rxq->mem, - false, NULL); + if (bpf_xdp_shrink_data(xdp, frag, shrink)) { n_frags_free++; } else { skb_frag_size_sub(frag, shrink); From ad2047cf5d9313200e308612aed516548873d124 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:55 +0100 Subject: [PATCH 04/11] ice: work on pre-XDP prog frag count Fix an OOM panic in XDP_DRV mode when a XDP program shrinks a multi-buffer packet by 4k bytes and then redirects it to an AF_XDP socket. Since support for handling multi-buffer frames was added to XDP, usage of bpf_xdp_adjust_tail() helper within XDP program can free the page that given fragment occupies and in turn decrease the fragment count within skb_shared_info that is embedded in xdp_buff struct. In current ice driver codebase, it can become problematic when page recycling logic decides not to reuse the page. In such case, __page_frag_cache_drain() is used with ice_rx_buf::pagecnt_bias that was not adjusted after refcount of page was changed by XDP prog which in turn does not drain the refcount to 0 and page is never freed. To address this, let us store the count of frags before the XDP program was executed on Rx ring struct. This will be used to compare with current frag count from skb_shared_info embedded in xdp_buff. A smaller value in the latter indicates that XDP prog freed frag(s). Then, for given delta decrement pagecnt_bias for XDP_DROP verdict. While at it, let us also handle the EOP frag within ice_set_rx_bufs_act() to make our life easier, so all of the adjustments needed to be applied against freed frags are performed in the single place. Fixes: 2fba7dc5157b ("ice: Add support for XDP multi-buffer on Rx side") Acked-by: Magnus Karlsson Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-5-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/ice/ice_txrx.c | 14 ++++++--- drivers/net/ethernet/intel/ice/ice_txrx.h | 1 + drivers/net/ethernet/intel/ice/ice_txrx_lib.h | 31 +++++++++++++------ 3 files changed, 32 insertions(+), 14 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c index 74d13cc5a3a7..0c9b4aa8a049 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx.c +++ b/drivers/net/ethernet/intel/ice/ice_txrx.c @@ -603,9 +603,7 @@ out_failure: ret = ICE_XDP_CONSUMED; } exit: - rx_buf->act = ret; - if (unlikely(xdp_buff_has_frags(xdp))) - ice_set_rx_bufs_act(xdp, rx_ring, ret); + ice_set_rx_bufs_act(xdp, rx_ring, ret); } /** @@ -893,14 +891,17 @@ ice_add_xdp_frag(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp, } if (unlikely(sinfo->nr_frags == MAX_SKB_FRAGS)) { - if (unlikely(xdp_buff_has_frags(xdp))) - ice_set_rx_bufs_act(xdp, rx_ring, ICE_XDP_CONSUMED); + ice_set_rx_bufs_act(xdp, rx_ring, ICE_XDP_CONSUMED); return -ENOMEM; } __skb_fill_page_desc_noacc(sinfo, sinfo->nr_frags++, rx_buf->page, rx_buf->page_offset, size); sinfo->xdp_frags_size += size; + /* remember frag count before XDP prog execution; bpf_xdp_adjust_tail() + * can pop off frags but driver has to handle it on its own + */ + rx_ring->nr_frags = sinfo->nr_frags; if (page_is_pfmemalloc(rx_buf->page)) xdp_buff_set_frag_pfmemalloc(xdp); @@ -1251,6 +1252,7 @@ int ice_clean_rx_irq(struct ice_rx_ring *rx_ring, int budget) xdp->data = NULL; rx_ring->first_desc = ntc; + rx_ring->nr_frags = 0; continue; construct_skb: if (likely(ice_ring_uses_build_skb(rx_ring))) @@ -1266,10 +1268,12 @@ construct_skb: ICE_XDP_CONSUMED); xdp->data = NULL; rx_ring->first_desc = ntc; + rx_ring->nr_frags = 0; break; } xdp->data = NULL; rx_ring->first_desc = ntc; + rx_ring->nr_frags = 0; stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_RXE_S); if (unlikely(ice_test_staterr(rx_desc->wb.status_error0, diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.h b/drivers/net/ethernet/intel/ice/ice_txrx.h index b3379ff73674..af955b0e5dc5 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx.h +++ b/drivers/net/ethernet/intel/ice/ice_txrx.h @@ -358,6 +358,7 @@ struct ice_rx_ring { struct ice_tx_ring *xdp_ring; struct ice_rx_ring *next; /* pointer to next ring in q_vector */ struct xsk_buff_pool *xsk_pool; + u32 nr_frags; dma_addr_t dma; /* physical address of ring */ u16 rx_buf_len; u8 dcb_tc; /* Traffic class of ring */ diff --git a/drivers/net/ethernet/intel/ice/ice_txrx_lib.h b/drivers/net/ethernet/intel/ice/ice_txrx_lib.h index 762047508619..afcead4baef4 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx_lib.h +++ b/drivers/net/ethernet/intel/ice/ice_txrx_lib.h @@ -12,26 +12,39 @@ * act: action to store onto Rx buffers related to XDP buffer parts * * Set action that should be taken before putting Rx buffer from first frag - * to one before last. Last one is handled by caller of this function as it - * is the EOP frag that is currently being processed. This function is - * supposed to be called only when XDP buffer contains frags. + * to the last. */ static inline void ice_set_rx_bufs_act(struct xdp_buff *xdp, const struct ice_rx_ring *rx_ring, const unsigned int act) { - const struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); - u32 first = rx_ring->first_desc; - u32 nr_frags = sinfo->nr_frags; + u32 sinfo_frags = xdp_get_shared_info_from_buff(xdp)->nr_frags; + u32 nr_frags = rx_ring->nr_frags + 1; + u32 idx = rx_ring->first_desc; u32 cnt = rx_ring->count; struct ice_rx_buf *buf; for (int i = 0; i < nr_frags; i++) { - buf = &rx_ring->rx_buf[first]; + buf = &rx_ring->rx_buf[idx]; buf->act = act; - if (++first == cnt) - first = 0; + if (++idx == cnt) + idx = 0; + } + + /* adjust pagecnt_bias on frags freed by XDP prog */ + if (sinfo_frags < rx_ring->nr_frags && act == ICE_XDP_CONSUMED) { + u32 delta = rx_ring->nr_frags - sinfo_frags; + + while (delta) { + if (idx == 0) + idx = cnt - 1; + else + idx--; + buf = &rx_ring->rx_buf[idx]; + buf->pagecnt_bias--; + delta--; + } } } From 83014323c642b8faa2d64a5f303b41c019322478 Mon Sep 17 00:00:00 2001 From: Tirthendu Sarkar Date: Wed, 24 Jan 2024 20:15:56 +0100 Subject: [PATCH 05/11] i40e: handle multi-buffer packets that are shrunk by xdp prog XDP programs can shrink packets by calling the bpf_xdp_adjust_tail() helper function. For multi-buffer packets this may lead to reduction of frag count stored in skb_shared_info area of the xdp_buff struct. This results in issues with the current handling of XDP_PASS and XDP_DROP cases. For XDP_PASS, currently skb is being built using frag count of xdp_buffer before it was processed by XDP prog and thus will result in an inconsistent skb when frag count gets reduced by XDP prog. To fix this, get correct frag count while building the skb instead of using pre-obtained frag count. For XDP_DROP, current page recycling logic will not reuse the page but instead will adjust the pagecnt_bias so that the page can be freed. This again results in inconsistent behavior as the page refcnt has already been changed by the helper while freeing the frag(s) as part of shrinking the packet. To fix this, only adjust pagecnt_bias for buffers that are stillpart of the packet post-xdp prog run. Fixes: e213ced19bef ("i40e: add support for XDP multi-buffer Rx") Reported-by: Maciej Fijalkowski Signed-off-by: Tirthendu Sarkar Link: https://lore.kernel.org/r/20240124191602.566724-6-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/i40e/i40e_txrx.c | 40 ++++++++++++--------- 1 file changed, 23 insertions(+), 17 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index 971ba3322038..1f0a0f13a334 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -2087,7 +2087,8 @@ static void i40e_put_rx_buffer(struct i40e_ring *rx_ring, static void i40e_process_rx_buffs(struct i40e_ring *rx_ring, int xdp_res, struct xdp_buff *xdp) { - u32 next = rx_ring->next_to_clean; + u32 nr_frags = xdp_get_shared_info_from_buff(xdp)->nr_frags; + u32 next = rx_ring->next_to_clean, i = 0; struct i40e_rx_buffer *rx_buffer; xdp->flags = 0; @@ -2100,10 +2101,10 @@ static void i40e_process_rx_buffs(struct i40e_ring *rx_ring, int xdp_res, if (!rx_buffer->page) continue; - if (xdp_res == I40E_XDP_CONSUMED) - rx_buffer->pagecnt_bias++; - else + if (xdp_res != I40E_XDP_CONSUMED) i40e_rx_buffer_flip(rx_buffer, xdp->frame_sz); + else if (i++ <= nr_frags) + rx_buffer->pagecnt_bias++; /* EOP buffer will be put in i40e_clean_rx_irq() */ if (next == rx_ring->next_to_process) @@ -2117,20 +2118,20 @@ static void i40e_process_rx_buffs(struct i40e_ring *rx_ring, int xdp_res, * i40e_construct_skb - Allocate skb and populate it * @rx_ring: rx descriptor ring to transact packets on * @xdp: xdp_buff pointing to the data - * @nr_frags: number of buffers for the packet * * This function allocates an skb. It then populates it with the page * data from the current receive descriptor, taking care to set up the * skb correctly. */ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, - struct xdp_buff *xdp, - u32 nr_frags) + struct xdp_buff *xdp) { unsigned int size = xdp->data_end - xdp->data; struct i40e_rx_buffer *rx_buffer; + struct skb_shared_info *sinfo; unsigned int headlen; struct sk_buff *skb; + u32 nr_frags = 0; /* prefetch first cache line of first page */ net_prefetch(xdp->data); @@ -2168,6 +2169,10 @@ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, sizeof(long))); + if (unlikely(xdp_buff_has_frags(xdp))) { + sinfo = xdp_get_shared_info_from_buff(xdp); + nr_frags = sinfo->nr_frags; + } rx_buffer = i40e_rx_bi(rx_ring, rx_ring->next_to_clean); /* update all of the pointers */ size -= headlen; @@ -2187,9 +2192,8 @@ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, } if (unlikely(xdp_buff_has_frags(xdp))) { - struct skb_shared_info *sinfo, *skinfo = skb_shinfo(skb); + struct skb_shared_info *skinfo = skb_shinfo(skb); - sinfo = xdp_get_shared_info_from_buff(xdp); memcpy(&skinfo->frags[skinfo->nr_frags], &sinfo->frags[0], sizeof(skb_frag_t) * nr_frags); @@ -2212,17 +2216,17 @@ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, * i40e_build_skb - Build skb around an existing buffer * @rx_ring: Rx descriptor ring to transact packets on * @xdp: xdp_buff pointing to the data - * @nr_frags: number of buffers for the packet * * This function builds an skb around an existing Rx buffer, taking care * to set up the skb correctly and avoid any memcpy overhead. */ static struct sk_buff *i40e_build_skb(struct i40e_ring *rx_ring, - struct xdp_buff *xdp, - u32 nr_frags) + struct xdp_buff *xdp) { unsigned int metasize = xdp->data - xdp->data_meta; + struct skb_shared_info *sinfo; struct sk_buff *skb; + u32 nr_frags; /* Prefetch first cache line of first page. If xdp->data_meta * is unused, this points exactly as xdp->data, otherwise we @@ -2231,6 +2235,11 @@ static struct sk_buff *i40e_build_skb(struct i40e_ring *rx_ring, */ net_prefetch(xdp->data_meta); + if (unlikely(xdp_buff_has_frags(xdp))) { + sinfo = xdp_get_shared_info_from_buff(xdp); + nr_frags = sinfo->nr_frags; + } + /* build an skb around the page buffer */ skb = napi_build_skb(xdp->data_hard_start, xdp->frame_sz); if (unlikely(!skb)) @@ -2243,9 +2252,6 @@ static struct sk_buff *i40e_build_skb(struct i40e_ring *rx_ring, skb_metadata_set(skb, metasize); if (unlikely(xdp_buff_has_frags(xdp))) { - struct skb_shared_info *sinfo; - - sinfo = xdp_get_shared_info_from_buff(xdp); xdp_update_skb_shared_info(skb, nr_frags, sinfo->xdp_frags_size, nr_frags * xdp->frame_sz, @@ -2589,9 +2595,9 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget, total_rx_bytes += size; } else { if (ring_uses_build_skb(rx_ring)) - skb = i40e_build_skb(rx_ring, xdp, nfrags); + skb = i40e_build_skb(rx_ring, xdp); else - skb = i40e_construct_skb(rx_ring, xdp, nfrags); + skb = i40e_construct_skb(rx_ring, xdp); /* drop if we failed to retrieve a buffer */ if (!skb) { From 2ee788c06493d02ee85855414cca39825e768aaf Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:57 +0100 Subject: [PATCH 06/11] ice: remove redundant xdp_rxq_info registration xdp_rxq_info struct can be registered by drivers via two functions - xdp_rxq_info_reg() and __xdp_rxq_info_reg(). The latter one allows drivers that support XDP multi-buffer to set up xdp_rxq_info::frag_size which in turn will make it possible to grow the packet via bpf_xdp_adjust_tail() BPF helper. Currently, ice registers xdp_rxq_info in two spots: 1) ice_setup_rx_ring() // via xdp_rxq_info_reg(), BUG 2) ice_vsi_cfg_rxq() // via __xdp_rxq_info_reg(), OK Cited commit under fixes tag took care of setting up frag_size and updated registration scheme in 2) but it did not help as 1) is called before 2) and as shown above it uses old registration function. This means that 2) sees that xdp_rxq_info is already registered and never calls __xdp_rxq_info_reg() which leaves us with xdp_rxq_info::frag_size being set to 0. To fix this misbehavior, simply remove xdp_rxq_info_reg() call from ice_setup_rx_ring(). Fixes: 2fba7dc5157b ("ice: Add support for XDP multi-buffer on Rx side") Acked-by: Magnus Karlsson Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-7-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/ice/ice_txrx.c | 5 ----- 1 file changed, 5 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c index 0c9b4aa8a049..97d41d6ebf1f 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx.c +++ b/drivers/net/ethernet/intel/ice/ice_txrx.c @@ -513,11 +513,6 @@ int ice_setup_rx_ring(struct ice_rx_ring *rx_ring) if (ice_is_xdp_ena_vsi(rx_ring->vsi)) WRITE_ONCE(rx_ring->xdp_prog, rx_ring->vsi->xdp_prog); - if (rx_ring->vsi->type == ICE_VSI_PF && - !xdp_rxq_info_is_reg(&rx_ring->xdp_rxq)) - if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev, - rx_ring->q_index, rx_ring->q_vector->napi.napi_id)) - goto err; return 0; err: From 290779905d09d5fdf6caa4f58ddefc3f4db0c0a9 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:58 +0100 Subject: [PATCH 07/11] intel: xsk: initialize skb_frag_t::bv_offset in ZC drivers Ice and i40e ZC drivers currently set offset of a frag within skb_shared_info to 0, which is incorrect. xdp_buffs that come from xsk_buff_pool always have 256 bytes of a headroom, so they need to be taken into account to retrieve xdp_buff::data via skb_frag_address(). Otherwise, bpf_xdp_frags_increase_tail() would be starting its job from xdp_buff::data_hard_start which would result in overwriting existing payload. Fixes: 1c9ba9c14658 ("i40e: xsk: add RX multi-buffer support") Fixes: 1bbc04de607b ("ice: xsk: add RX multi-buffer support") Acked-by: Magnus Karlsson Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-8-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/i40e/i40e_xsk.c | 3 ++- drivers/net/ethernet/intel/ice/ice_xsk.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c index 82aca0d16a3e..11500003af0d 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c +++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c @@ -414,7 +414,8 @@ i40e_add_xsk_frag(struct i40e_ring *rx_ring, struct xdp_buff *first, } __skb_fill_page_desc_noacc(sinfo, sinfo->nr_frags++, - virt_to_page(xdp->data_hard_start), 0, size); + virt_to_page(xdp->data_hard_start), + XDP_PACKET_HEADROOM, size); sinfo->xdp_frags_size += size; xsk_buff_add_frag(xdp); diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c index d9073a618ad6..8b81a1677045 100644 --- a/drivers/net/ethernet/intel/ice/ice_xsk.c +++ b/drivers/net/ethernet/intel/ice/ice_xsk.c @@ -825,7 +825,8 @@ ice_add_xsk_frag(struct ice_rx_ring *rx_ring, struct xdp_buff *first, } __skb_fill_page_desc_noacc(sinfo, sinfo->nr_frags++, - virt_to_page(xdp->data_hard_start), 0, size); + virt_to_page(xdp->data_hard_start), + XDP_PACKET_HEADROOM, size); sinfo->xdp_frags_size += size; xsk_buff_add_frag(xdp); From 3de38c87174225487fc93befeea7d380db80aef6 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:15:59 +0100 Subject: [PATCH 08/11] ice: update xdp_rxq_info::frag_size for ZC enabled Rx queue Now that ice driver correctly sets up frag_size in xdp_rxq_info, let us make it work for ZC multi-buffer as well. ice_rx_ring::rx_buf_len for ZC is being set via xsk_pool_get_rx_frame_size() and this needs to be propagated up to xdp_rxq_info. Use a bigger hammer and instead of unregistering only xdp_rxq_info's memory model, unregister it altogether and register it again and have xdp_rxq_info with correct frag_size value. Fixes: 1bbc04de607b ("ice: xsk: add RX multi-buffer support") Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-9-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/ice/ice_base.c | 37 ++++++++++++++--------- 1 file changed, 23 insertions(+), 14 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c index 533b923cae2d..7ac847718882 100644 --- a/drivers/net/ethernet/intel/ice/ice_base.c +++ b/drivers/net/ethernet/intel/ice/ice_base.c @@ -547,19 +547,27 @@ int ice_vsi_cfg_rxq(struct ice_rx_ring *ring) ring->rx_buf_len = ring->vsi->rx_buf_len; if (ring->vsi->type == ICE_VSI_PF) { - if (!xdp_rxq_info_is_reg(&ring->xdp_rxq)) - /* coverity[check_return] */ - __xdp_rxq_info_reg(&ring->xdp_rxq, ring->netdev, - ring->q_index, - ring->q_vector->napi.napi_id, - ring->vsi->rx_buf_len); + if (!xdp_rxq_info_is_reg(&ring->xdp_rxq)) { + err = __xdp_rxq_info_reg(&ring->xdp_rxq, ring->netdev, + ring->q_index, + ring->q_vector->napi.napi_id, + ring->rx_buf_len); + if (err) + return err; + } ring->xsk_pool = ice_xsk_pool(ring); if (ring->xsk_pool) { - xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq); + xdp_rxq_info_unreg(&ring->xdp_rxq); ring->rx_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool); + err = __xdp_rxq_info_reg(&ring->xdp_rxq, ring->netdev, + ring->q_index, + ring->q_vector->napi.napi_id, + ring->rx_buf_len); + if (err) + return err; err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, MEM_TYPE_XSK_BUFF_POOL, NULL); @@ -571,13 +579,14 @@ int ice_vsi_cfg_rxq(struct ice_rx_ring *ring) dev_info(dev, "Registered XDP mem model MEM_TYPE_XSK_BUFF_POOL on Rx ring %d\n", ring->q_index); } else { - if (!xdp_rxq_info_is_reg(&ring->xdp_rxq)) - /* coverity[check_return] */ - __xdp_rxq_info_reg(&ring->xdp_rxq, - ring->netdev, - ring->q_index, - ring->q_vector->napi.napi_id, - ring->vsi->rx_buf_len); + if (!xdp_rxq_info_is_reg(&ring->xdp_rxq)) { + err = __xdp_rxq_info_reg(&ring->xdp_rxq, ring->netdev, + ring->q_index, + ring->q_vector->napi.napi_id, + ring->rx_buf_len); + if (err) + return err; + } err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, MEM_TYPE_PAGE_SHARED, From fbadd83a612c3b7aad2987893faca6bd24aaebb3 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:16:00 +0100 Subject: [PATCH 09/11] xdp: reflect tail increase for MEM_TYPE_XSK_BUFF_POOL XSK ZC Rx path calculates the size of data that will be posted to XSK Rx queue via subtracting xdp_buff::data_end from xdp_buff::data. In bpf_xdp_frags_increase_tail(), when underlying memory type of xdp_rxq_info is MEM_TYPE_XSK_BUFF_POOL, add offset to data_end in tail fragment, so that later on user space will be able to take into account the amount of bytes added by XDP program. Fixes: 24ea50127ecf ("xsk: support mbuf on ZC RX") Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-10-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- net/core/filter.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/net/core/filter.c b/net/core/filter.c index 36fb5ae8af69..ef3e78b6a39c 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -4093,6 +4093,8 @@ static int bpf_xdp_frags_increase_tail(struct xdp_buff *xdp, int offset) memset(skb_frag_address(frag) + skb_frag_size(frag), 0, offset); skb_frag_size_add(frag, offset); sinfo->xdp_frags_size += offset; + if (rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL) + xsk_buff_get_tail(xdp)->data_end += offset; return 0; } From a045d2f2d03d23e7db6772dd83e0ba2705dfad93 Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:16:01 +0100 Subject: [PATCH 10/11] i40e: set xdp_rxq_info::frag_size i40e support XDP multi-buffer so it is supposed to use __xdp_rxq_info_reg() instead of xdp_rxq_info_reg() and set the frag_size. It can not be simply converted at existing callsite because rx_buf_len could be un-initialized, so let us register xdp_rxq_info within i40e_configure_rx_ring(), which happen to be called with already initialized rx_buf_len value. Commit 5180ff1364bc ("i40e: use int for i40e_status") converted 'err' to int, so two variables to deal with return codes are not needed within i40e_configure_rx_ring(). Remove 'ret' and use 'err' to handle status from xdp_rxq_info registration. Fixes: e213ced19bef ("i40e: add support for XDP multi-buffer Rx") Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-11-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/i40e/i40e_main.c | 40 ++++++++++++--------- drivers/net/ethernet/intel/i40e/i40e_txrx.c | 9 ----- 2 files changed, 24 insertions(+), 25 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index ae8f9f135725..d3b00d8ed39a 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -3588,40 +3588,48 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring) struct i40e_hmc_obj_rxq rx_ctx; int err = 0; bool ok; - int ret; bitmap_zero(ring->state, __I40E_RING_STATE_NBITS); /* clear the context structure first */ memset(&rx_ctx, 0, sizeof(rx_ctx)); - if (ring->vsi->type == I40E_VSI_MAIN) - xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq); + ring->rx_buf_len = vsi->rx_buf_len; + + /* XDP RX-queue info only needed for RX rings exposed to XDP */ + if (ring->vsi->type != I40E_VSI_MAIN) + goto skip; + + if (!xdp_rxq_info_is_reg(&ring->xdp_rxq)) { + err = __xdp_rxq_info_reg(&ring->xdp_rxq, ring->netdev, + ring->queue_index, + ring->q_vector->napi.napi_id, + ring->rx_buf_len); + if (err) + return err; + } ring->xsk_pool = i40e_xsk_pool(ring); if (ring->xsk_pool) { - ring->rx_buf_len = - xsk_pool_get_rx_frame_size(ring->xsk_pool); - ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, + ring->rx_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool); + err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, MEM_TYPE_XSK_BUFF_POOL, NULL); - if (ret) - return ret; + if (err) + return err; dev_info(&vsi->back->pdev->dev, "Registered XDP mem model MEM_TYPE_XSK_BUFF_POOL on Rx ring %d\n", ring->queue_index); } else { - ring->rx_buf_len = vsi->rx_buf_len; - if (ring->vsi->type == I40E_VSI_MAIN) { - ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, - MEM_TYPE_PAGE_SHARED, - NULL); - if (ret) - return ret; - } + err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, + MEM_TYPE_PAGE_SHARED, + NULL); + if (err) + return err; } +skip: xdp_init_buff(&ring->xdp, i40e_rx_pg_size(ring) / 2, &ring->xdp_rxq); rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len, diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index 1f0a0f13a334..0d7177083708 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -1548,7 +1548,6 @@ void i40e_free_rx_resources(struct i40e_ring *rx_ring) int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring) { struct device *dev = rx_ring->dev; - int err; u64_stats_init(&rx_ring->syncp); @@ -1569,14 +1568,6 @@ int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring) rx_ring->next_to_process = 0; rx_ring->next_to_use = 0; - /* XDP RX-queue info only needed for RX rings exposed to XDP */ - if (rx_ring->vsi->type == I40E_VSI_MAIN) { - err = xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev, - rx_ring->queue_index, rx_ring->q_vector->napi.napi_id); - if (err < 0) - return err; - } - rx_ring->xdp_prog = rx_ring->vsi->xdp_prog; rx_ring->rx_bi = From 0cbb08707c932b3f004bc1a8ec6200ef572c1f5f Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 24 Jan 2024 20:16:02 +0100 Subject: [PATCH 11/11] i40e: update xdp_rxq_info::frag_size for ZC enabled Rx queue Now that i40e driver correctly sets up frag_size in xdp_rxq_info, let us make it work for ZC multi-buffer as well. i40e_ring::rx_buf_len for ZC is being set via xsk_pool_get_rx_frame_size() and this needs to be propagated up to xdp_rxq_info. Fixes: 1c9ba9c14658 ("i40e: xsk: add RX multi-buffer support") Acked-by: Magnus Karlsson Signed-off-by: Maciej Fijalkowski Link: https://lore.kernel.org/r/20240124191602.566724-12-maciej.fijalkowski@intel.com Signed-off-by: Alexei Starovoitov --- drivers/net/ethernet/intel/i40e/i40e_main.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index d3b00d8ed39a..6e7fd473abfd 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -3611,7 +3611,14 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring) ring->xsk_pool = i40e_xsk_pool(ring); if (ring->xsk_pool) { + xdp_rxq_info_unreg(&ring->xdp_rxq); ring->rx_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool); + err = __xdp_rxq_info_reg(&ring->xdp_rxq, ring->netdev, + ring->queue_index, + ring->q_vector->napi.napi_id, + ring->rx_buf_len); + if (err) + return err; err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, MEM_TYPE_XSK_BUFF_POOL, NULL);