From mboxrd@z Thu Jan 1 00:00:00 1970 Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.remote-ip=109.224.244.117 ARC-Seal: i=1; d=subspace.kernel.org; s=arc20261015; a=rsa-sha256; cv=none; t=1792911138; b=for2B7mOCrkkLt7I9QIDDg6RHuQ/9Ft3+9X+NBPUsxMe952Nz929RCmdZiJ0NQWi8m4E zs89teXC8sbx8VfdSuE86fxNgo5RtfLQPdYIFFZIV7nVWQzpxCjhmqGbWEDgWxy7+DqC0 KQY79GhGKLlrnE6bfkGsUs1Qp95+Yn9NfiKzO3+mmGl8vQnE2yAOTADsVN6qhpqldwLyf VsV8uleddBdYY0SrVwBcQccQoOUkfiy2b16xpur3FvOW0Zk9FTguMKDRJ1rVV1YfRePsw 5x68kog667/F9b5s2eRW65c88SVxOwLMiMLcYlWJhcewIBb0/BOC68vIhS842imt21Q== ARC-Message-Signature: i=1; d=subspace.kernel.org; s=arc20261015; a=rsa-sha256; c=relaxed/relaxed; t=1792911138; h=DMARC-Filter:DKIM-Signature:Date:To:From:Cc:Subject:Message-ID: Feedback-ID:X-Pm-Message-ID:MIME-Version:Content-Type: Content-Transfer-Encoding; bh=lwQ4f2ZvBH062ZEHXjSVyzILEqPMVfwCfFpffv1k8n0=; b=JKZhlaZMzUjxCYfpZCIy+981kylYH3xJt5Njo1J8R095Avycb11MgfR4nHhQlyElTJCg SMAsJWT6dj8v2MY0BhIsbS7Z5Eg20VtaidyVwUNU2z8fzaqOX7Oi/rq6bwleMTPwLpgMf BXOZQLKr7f0Uxvz5q7TfIj3ueOgmJ2EDS4+V2HjmeZPVxyyTpGlL6VRLA9pYZ+sxoXAIo 4T4HA+Vk5sLIlouEtyavBypPE+c5qGQ3Es7pe8XXlCgiJe1E2PR9Oj9rNXPLqFi019+mA 2qklc0i3pmRgHGCfSDLU7M6dIksqK3flX2jgG/fY9HbNpYQeGoL9Llqae/gZlUwXl+g== ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass header.from=1g4.org; spf=pass smtp.mailfrom=1g4.org; dkim=pass header.d=1g4.org header.i=@1g4.org header.a=rsa-sha256 header.s=protonmail header.b=dsFm069s; arc=none smtp.remote-ip=109.224.244.117 Received: from mail-244117.protonmail.ch (mail-244117.protonmail.ch [109.224.244.117]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature ECDSA (prime256v1) server-digest SHA256) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id D9F10606A31A; Sun, 25 Oct 2026 06:52:16 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.2 smtp.subspace.kernel.org D9F10606A31A Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=quarantine dis=none) header.from=1g4.org Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=1g4.org Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key, unprotected) header.d=1g4.org header.i=@1g4.org header.a=rsa-sha256 header.s=protonmail header.b=dsFm069s DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1g4.org; s=protonmail; t=1792911134; x=1793170334; bh=lwQ4f2ZvBH062ZEHXjSVyzILEqPMVfwCfFpffv1k8n0=; h=Date:To:From:Cc:Subject:Message-ID:Feedback-ID:From:To:Cc:Date: Subject:Reply-To:Feedback-ID:Message-ID:BIMI-Selector; b=dsFm069sFcpfabfZXMqxJf8Y2iQPD2Hw1DSKMo0/ipyfxypJyLFax3fUFZI0Wc5jz OB8ns7Z6PEE75f8UgTi1xcX6y2dq4F2NyizwKQq+AKONn/xYGdShA3xvtbg0vXbQ8a G2KBrtZD0y3qDdruiOIcy6wljnYdaRslpdPiohC+vpfYrdgrCZwGjJsndG2567u9Zi VzfpXFPqmurzPL6npUfEbw97cLxbebn5dtNpT7cL5kysXDVdz9jqdpLn0uDdNz0o3n b32KkfyDI4GjXNeUdGfvyUN2iU3AYGfuhnQ/W/QFEyMw1ovjmXPIe8P8pPqf7lwcB6 e2mFGYfNrh+pA== Date: Sun, 25 Oct 2026 06:52:06 +0000 To: anthony.l.nguyen@intel.com, przemyslaw.kitszel@intel.com From: Paul Moses Cc: andrew+netdev@lunn.ch, davem@davemloft.net, edumazet@kernel.org, kuba@kernel.org, pabeni@redhat.com, ast@kernel.org, daniel@iogearbox.net, hawk@kernel.org, john.fastabend@gmail.com, sdf@fomichev.me, maciej.fijalkowski@intel.com, vedang.patel@intel.com, jithu.joseph@intel.com, andre.guedes@intel.com, intel-wired-lan@lists.osuosl.org, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, bpf@vger.kernel.org Subject: [PATCH net] igc: track rx frame boundaries independent of skbs Message-ID: <20261025065154.2037149-1-p@1g4.org> Feedback-ID: 8253658:user:proton X-Pm-Message-ID: d7913055a968b10d30843429c75eb5326aaf17d2 X-Mailing-List: netdev@vger.kernel.org List-Id: List-Post: List-Help: List-Subscribe: List-Unsubscribe: List-Owner: Precedence: list MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable TSIP remains set on continuation descriptors, but the inline timestamp header exists only in the first buffer. Stripping it from every descriptor truncates split frames and can underflow short buffers. skb ownership cannot identify frame boundaries when XDP consumes the first buffer or AF_XDP uses zero-copy. Track rx_in_frame per ring and update it as descriptors are consumed. Strip timestamps only from first buffers, reject unsupported split frames before XDP and drain continuations. Account discarded descriptors against NAPI work without counting successful RX. Fixes: e1ed4f92a625 ("igc: Refactor Rx timestamp handling") Fixes: 26575105d6ed ("igc: Add initial XDP support") Fixes: fc9df2a0b520 ("igc: Enable RX via AF_XDP zero-copy") Cc: stable@vger.kernel.org Signed-off-by: Paul Moses --- - This supersedes "igc: only strip RX timestamp header from first buffer" following feedback from Simon Horman, Maciej Fijalkowski and Sashiko. Changes: - track frame boundaries independently of skb ownership - guard timestamp stripping and drain unsupported split frames - cover the AF_XDP zero-copy RX path --- drivers/net/ethernet/intel/igc/igc.h | 1 + drivers/net/ethernet/intel/igc/igc_main.c | 87 ++++++++++++++++------- 2 files changed, 64 insertions(+), 24 deletions(-) diff --git a/drivers/net/ethernet/intel/igc/igc.h b/drivers/net/ethernet/in= tel/igc/igc.h index 17f213cc93e4..635d49701e62 100644 --- a/drivers/net/ethernet/intel/igc/igc.h +++ b/drivers/net/ethernet/intel/igc/igc.h @@ -198,6 +198,7 @@ struct igc_ring { =09=09=09struct igc_rx_packet_stats pkt_stats; =09=09=09struct u64_stats_sync rx_syncp; =09=09=09struct sk_buff *skb; +=09=09=09bool rx_in_frame; =09=09}; =09}; =20 diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethern= et/intel/igc/igc_main.c index 1fb5f3cbe93c..9ca5558c762c 100644 --- a/drivers/net/ethernet/intel/igc/igc_main.c +++ b/drivers/net/ethernet/intel/igc/igc_main.c @@ -474,6 +474,7 @@ static void igc_clean_rx_ring(struct igc_ring *ring) =09=09igc_clean_rx_ring_page_shared(ring); =20 =09clear_ring_uses_large_buffer(ring); +=09ring->rx_in_frame =3D false; =20 =09ring->next_to_alloc =3D 0; =09ring->next_to_clean =3D 0; @@ -673,6 +674,7 @@ static void igc_configure_rx_ring(struct igc_adapter *a= dapter, =09/* reset next-to- use/clean to place SW in sync with hardware */ =09ring->next_to_clean =3D 0; =09ring->next_to_use =3D 0; +=09ring->rx_in_frame =3D false; =20 =09if (ring->xsk_pool) =09=09buf_size =3D xsk_pool_get_rx_frame_size(ring->xsk_pool); @@ -2111,10 +2113,8 @@ static bool igc_can_reuse_rx_page(struct igc_rx_buff= er *rx_buffer, * @rx_ring: Rx ring being processed * @rx_desc: Rx descriptor for current buffer * - * This function updates next to clean. If the buffer is an EOP buffer - * this function exits returning false, otherwise it will place the - * sk_buff in the next buffer to be chained and return true indicating - * that this is in fact a non-EOP buffer. + * Advance next to clean and retain frame position for the next descriptor= . + * Return true if the consumed descriptor does not end the frame. */ static bool igc_is_non_eop(struct igc_ring *rx_ring, =09=09=09 union igc_adv_rx_desc *rx_desc) @@ -2127,10 +2127,8 @@ static bool igc_is_non_eop(struct igc_ring *rx_ring, =20 =09prefetch(IGC_RX_DESC(rx_ring, ntc)); =20 -=09if (likely(igc_test_staterr(rx_desc, IGC_RXD_STAT_EOP))) -=09=09return false; - -=09return true; +=09rx_ring->rx_in_frame =3D !igc_test_staterr(rx_desc, IGC_RXD_STAT_EOP); +=09return unlikely(rx_ring->rx_in_frame); } =20 /** @@ -2601,7 +2599,7 @@ static void igc_update_rx_stats(struct igc_q_vector *= q_vector, =20 static int igc_clean_rx_irq(struct igc_q_vector *q_vector, const int budge= t) { -=09unsigned int total_bytes =3D 0, total_packets =3D 0; +=09unsigned int total_bytes =3D 0, total_packets =3D 0, work_done =3D 0; =09struct igc_adapter *adapter =3D q_vector->adapter; =09struct igc_ring *rx_ring =3D q_vector->rx.ring; =09struct sk_buff *skb =3D rx_ring->skb; @@ -2609,13 +2607,15 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_= vector, const int budget) =09int xdp_status =3D 0, rx_buffer_pgcnt; =09int xdp_res =3D 0; =20 -=09while (likely(total_packets < budget)) { +=09while (likely(work_done < budget)) { =09=09struct igc_xdp_buff ctx =3D { .rx_ts =3D NULL }; +=09=09bool first =3D !rx_ring->rx_in_frame; =09=09struct igc_rx_buffer *rx_buffer; =09=09union igc_adv_rx_desc *rx_desc; =09=09unsigned int size, truesize; =09=09int pkt_offset =3D 0; =09=09void *pktbuf; +=09=09bool tsip; =20 =09=09/* return some buffers to hardware, one at a time is too slow */ =09=09if (cleaned_count >=3D IGC_RX_BUFFER_WRITE) { @@ -2635,21 +2635,34 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_= vector, const int budget) =09=09dma_rmb(); =20 =09=09rx_buffer =3D igc_get_rx_buffer(rx_ring, size, &rx_buffer_pgcnt); +=09=09tsip =3D first && igc_test_staterr(rx_desc, IGC_RXDADV_STAT_TSIP); + +=09=09if ((!first && !skb) || +=09=09 (first && igc_xdp_is_enabled(adapter) && +=09=09 !igc_test_staterr(rx_desc, IGC_RXD_STAT_EOP)) || +=09=09 unlikely(tsip && size < IGC_TS_HDR_LEN)) { +=09=09=09if (first) +=09=09=09=09dev_core_stats_rx_dropped_inc(rx_ring->netdev); +=09=09=09xdp_res =3D IGC_XDP_CONSUMED; +=09=09=09goto xdp_done; +=09=09} + =09=09truesize =3D igc_get_rx_frame_truesize(rx_ring, size); =20 =09=09pktbuf =3D page_address(rx_buffer->page) + rx_buffer->page_offset; =20 -=09=09if (igc_test_staterr(rx_desc, IGC_RXDADV_STAT_TSIP)) { +=09=09if (tsip) { =09=09=09ctx.rx_ts =3D pktbuf; =09=09=09pkt_offset =3D IGC_TS_HDR_LEN; =09=09=09size -=3D IGC_TS_HDR_LEN; =09=09} =20 -=09=09if (igc_fpe_is_pmac_enabled(adapter) && +=09=09if (first && igc_fpe_is_pmac_enabled(adapter) && =09=09 igc_fpe_handle_mpacket(adapter, rx_desc, size, pktbuf + pkt_offs= et)) { =09=09=09/* Advance the ring next-to-clean */ =09=09=09igc_is_non_eop(rx_ring, rx_desc); =09=09=09cleaned_count++; +=09=09=09work_done++; =09=09=09continue; =09=09} =20 @@ -2664,6 +2677,12 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_v= ector, const int budget) =09=09=09xdp_res =3D igc_xdp_run_prog(adapter, &ctx.xdp); =09=09} =20 +=09=09if (xdp_res) { +=09=09=09total_bytes +=3D size; +=09=09=09total_packets++; +=09=09} + +xdp_done: =09=09if (xdp_res) { =09=09=09switch (xdp_res) { =09=09=09case IGC_XDP_CONSUMED: @@ -2676,8 +2695,7 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_ve= ctor, const int budget) =09=09=09=09break; =09=09=09} =20 -=09=09=09total_packets++; -=09=09=09total_bytes +=3D size; +=09=09=09work_done++; =09=09} else if (skb) =09=09=09igc_add_rx_frag(rx_ring, rx_buffer, skb, size); =09=09else if (ring_uses_build_skb(rx_ring)) @@ -2700,8 +2718,15 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_v= ector, const int budget) =09=09if (igc_is_non_eop(rx_ring, rx_desc)) =09=09=09continue; =20 +=09=09if (xdp_res) { +=09=09=09skb =3D NULL; +=09=09=09continue; +=09=09} + +=09=09work_done++; + =09=09/* verify the packet layout is correct */ -=09=09if (xdp_res || igc_cleanup_headers(rx_ring, rx_desc, skb)) { +=09=09if (igc_cleanup_headers(rx_ring, rx_desc, skb)) { =09=09=09skb =3D NULL; =09=09=09continue; =09=09} @@ -2717,7 +2742,6 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_ve= ctor, const int budget) =09=09/* reset skb pointer */ =09=09skb =3D NULL; =20 -=09=09/* update budget accounting */ =09=09total_packets++; =09} =20 @@ -2732,7 +2756,7 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_ve= ctor, const int budget) =09if (cleaned_count) =09=09igc_alloc_rx_buffers(rx_ring, cleaned_count); =20 -=09return total_packets; +=09return work_done; } =20 static struct sk_buff *igc_construct_skb_zc(struct igc_ring *ring, @@ -2797,10 +2821,10 @@ static struct igc_xdp_buff *xsk_buff_to_igc_ctx(str= uct xdp_buff *xdp) =20 static int igc_clean_rx_irq_zc(struct igc_q_vector *q_vector, const int bu= dget) { +=09int total_bytes =3D 0, total_packets =3D 0, work_done =3D 0; =09struct igc_adapter *adapter =3D q_vector->adapter; =09struct igc_ring *ring =3D q_vector->rx.ring; =09u16 cleaned_count =3D igc_desc_unused(ring); -=09int total_bytes =3D 0, total_packets =3D 0; =09u16 ntc =3D ring->next_to_clean; =09struct bpf_prog *prog; =09bool failure =3D false; @@ -2810,11 +2834,13 @@ static int igc_clean_rx_irq_zc(struct igc_q_vector = *q_vector, const int budget) =20 =09prog =3D READ_ONCE(adapter->xdp_prog); =20 -=09while (likely(total_packets < budget)) { +=09while (likely(work_done < budget)) { +=09=09bool first =3D !ring->rx_in_frame; =09=09union igc_adv_rx_desc *desc; =09=09struct igc_rx_buffer *bi; =09=09struct igc_xdp_buff *ctx; =09=09unsigned int size; +=09=09bool eop, tsip; =09=09int res; =20 =09=09desc =3D IGC_RX_DESC(ring, ntc); @@ -2829,11 +2855,20 @@ static int igc_clean_rx_irq_zc(struct igc_q_vector = *q_vector, const int budget) =09=09dma_rmb(); =20 =09=09bi =3D &ring->rx_buffer_info[ntc]; +=09=09tsip =3D first && igc_test_staterr(desc, IGC_RXDADV_STAT_TSIP); +=09=09eop =3D igc_test_staterr(desc, IGC_RXD_STAT_EOP); + +=09=09if (!first || !eop || unlikely(tsip && size < IGC_TS_HDR_LEN)) { +=09=09=09if (first) +=09=09=09=09dev_core_stats_rx_dropped_inc(ring->netdev); +=09=09=09res =3D IGC_XDP_CONSUMED; +=09=09=09goto xdp_done; +=09=09} =20 =09=09ctx =3D xsk_buff_to_igc_ctx(bi->xdp); =09=09ctx->rx_desc =3D desc; =20 -=09=09if (igc_test_staterr(desc, IGC_RXDADV_STAT_TSIP)) { +=09=09if (tsip) { =09=09=09ctx->rx_ts =3D bi->xdp->data; =20 =09=09=09bi->xdp->data +=3D IGC_TS_HDR_LEN; @@ -2851,6 +2886,10 @@ static int igc_clean_rx_irq_zc(struct igc_q_vector *= q_vector, const int budget) =09=09xsk_buff_dma_sync_for_cpu(bi->xdp); =20 =09=09res =3D __igc_xdp_run_prog(adapter, prog, bi->xdp); +=09=09total_bytes +=3D size; +=09=09total_packets++; + +xdp_done: =09=09switch (res) { =09=09case IGC_XDP_PASS: =09=09=09igc_dispatch_skb_zc(q_vector, desc, ctx); @@ -2865,8 +2904,8 @@ static int igc_clean_rx_irq_zc(struct igc_q_vector *q= _vector, const int budget) =09=09} =20 =09=09bi->xdp =3D NULL; -=09=09total_bytes +=3D size; -=09=09total_packets++; +=09=09ring->rx_in_frame =3D !eop; +=09=09work_done++; =09=09cleaned_count++; =09=09ntc++; =09=09if (ntc =3D=3D ring->count) @@ -2889,10 +2928,10 @@ static int igc_clean_rx_irq_zc(struct igc_q_vector = *q_vector, const int budget) =09=09=09xsk_set_rx_need_wakeup(ring->xsk_pool); =09=09else =09=09=09xsk_clear_rx_need_wakeup(ring->xsk_pool); -=09=09return total_packets; +=09=09return work_done; =09} =20 -=09return failure ? budget : total_packets; +=09return failure ? budget : work_done; } =20 static void igc_update_tx_stats(struct igc_q_vector *q_vector,