From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) (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 579FF606A31C for ; Tue, 17 Nov 2026 07:58:31 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key, unprotected) header.d=google.com header.i=@google.com header.a=rsa-sha256 header.s=20260707 header.b=ZqJ/kHrV Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 67CC8402C9; Tue, 17 Nov 2026 08:58:22 +0100 (CET) Received: from mail-qk1-f200.google.com (mail-qk1-f200.google.com [209.85.222.200]) by mails.dpdk.org (Postfix) with ESMTPS id EE6164025A for ; Sun, 15 Nov 2026 15:00:18 +0100 (CET) Received: by mail-qk1-f200.google.com with SMTP id af79cd13be357-946038e22f5so226804985a.0 for ; Sun, 15 Nov 2026 06:00:18 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20260707; t=1794751218; x=1795356018; darn=dpdk.org; h=content-type:cc:to:from:subject:message-id:mime-version:date:from :to:cc:subject:date:message-id:reply-to:content-type; bh=+tsdmPr9zdxUlnyJZbTXwJxc6lS/D4sb3Fho4B0iaFw=; b=ZqJ/kHrV1pe9SZinGm3WmJi8VogHOanlrJwHbj/JEC58ueq5Z7O21b1fRDfbklb00c JOyYCJx3y3Yd57B80Z8qxIdjK1XNTN/qUB5FdNNCzjoWoLfx/Prw1MRTTpXQKPLKXQK9 Q7wpAhNqa9/+zTBGNtSvPGInfD8ZMZDYQxfUkllLar0JM+iCc6iLP1sh4x8vcUybcDbg IZHavy0RvZts9RP3w6aOrPEUWcNvmUT74gmahcYu+phtGAYEX2yPtX6Ll3EhL0ZKQ+vT T8az84jr3JNIfdLQDYA1C9KVjdG76WxMcocusIGJQ7KUvVfz/QPAFDABpAk4bF7pM25/ Da9w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20260707; t=1794751218; x=1795356018; h=content-type:cc:to:from:subject:message-id:mime-version:date :x-gm-message-state:from:to:cc:subject:date:message-id:reply-to :content-type; bh=+tsdmPr9zdxUlnyJZbTXwJxc6lS/D4sb3Fho4B0iaFw=; b=WQ7rQ7uni9aj5WIKzGmpUjYFMNfvrf0iH91xkRphz8CI7fzMEDFXnOoXSMpkK0dDol zur31ag/MXGMIZ5fApt5W+Ja7KuCFjEeFDSBOQ3d5l13KQuA/DaPU3V1oVTZtY9Yl5nv U3WlH5XbZEUi0sXpYLfL5VjVEKNBXd3o/Z3SpteO/rD3YfQ2Zypd+drVZLxSsI0q9HZW b8Qo5GTWJIsdJzX7xj4Un/AMjs2i0RujNvYG8Lw6s9oHze1Jdol1LlsgVb3VXfKjGDzJ y1sLvGCEazQxQSU8CN2/IROUKeNszVzjfR8H2g4mVMaDlEbyOTT2NgzIzw1/ReIlu6ic H5MQ== X-Gm-Message-State: APuU4Z+G7nWIrUw8tLEIZWQosQk7uwUWMxJisfXMiTBymt4B/JoZUlfs kaax8rS05Uy/lRQsBNIt/SrDwOvV0i/MLrsebnaD1EuFKLhdhr5cdRX2njpHc1W8ZdsDiitN4J8 bVCH06uarwPQMgJFXXRle5ofaI2YFPq4Ww69htrLKYQZtyOP0/ghyXdDxr9FT7n1L4Tgz7Nxhqo GZX8aMHt8N91yIquSP2hvZjA479czU X-Received: from qkao9.prod.google.com ([2002:a05:620a:a809:b0:93f:c1b:9052]) (user=rnwolfe job=prod-delivery.src-stubby-dispatcher) by 2002:a05:620a:a114:b0:949:5e26:e927 with SMTP id af79cd13be357-9495e26ee0bmr766960085a.73.1794751217240; Sun, 15 Nov 2026 06:00:17 -0800 (PST) Date: Sun, 15 Nov 2026 09:00:16 -0500 X-Archived-By: Mime-Version: 1.0 X-Mailer: git-send-email 2.56.0.679.gc527c65306-goog Message-ID: <20261115140016.3830109-1-rnwolfe@google.com> Subject: [PATCH] net/gve: fix GQI QPL jumbo Tx FIFO reclaim and doorbell From: Ryan Wolfe To: dev@dpdk.org, Junfeng Guo , Joshua Washington Cc: stable@dpdk.org, Harshitha Ramamurthy , Rushil Gupta Content-Type: text/plain; charset="UTF-8" X-Mailman-Approved-At: Tue, 17 Nov 2026 08:58:20 +0100 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Archive: List-Post: List-Help: List-Subscribe: , X-Spam-Status: No, score=-26.3 required=3.0 tests=BAYES_00,DKIMWL_WL_MED, DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_CR_TRAILER,INCLUDES_PATCH,MAILING_LIST_MULTI,SPF_HELO_NONE, SPF_PASS,USER_AGENT_GIT,USER_IN_DEF_DKIM_WL autolearn=ham autolearn_force=no version=4.0.1 X-Spam-Checker-Version: SpamAssassin 4.0.1 (2024-03-26) on aws-us-west-2-korg-listserv-alma10-1 In GQI Queue Page List (GQI_QPL) mode each Tx queue copies outgoing frames into a fixed QPL bounce FIFO (512 pages = 2 MiB on GCE, ~236 8,832-byte cacheline-aligned jumbo frames) even when nb_tx_desc is configured to 2048. Under sustained jumbo bursts three bugs cause unnecessary packet drops and mbuf tracking hazards: 1. Entry cleanup in gve_tx_burst_qpl() only ran when nb_free < free_thresh (32) or fifo_avail == 0. With 2048 descriptors, 236 jumbo frames leave nb_free = 1811 and non-zero remainder bytes in fifo_avail, so entry cleanup never ran and every QPL wrap stalled mid-burst. Reclaim when fifo_avail drops below half of fifo_size, and do not count normal FIFO/ring-full backpressure as hardware stats.errors. 2. When is_fifo_avail() failed mid-burst after queuing packets, the driver polled gve_tx_clean() once before writing qtx_tail, so the device had not yet started draining the newly queued frames, and a single GVE_TX_MAX_FREE_SZ batch after many small frames could stop short of freeing a jumbo slot. Flush qtx_tail first and loop gve_tx_clean() while next_to_clean advances. 3. sw_ring[sw_id] and the first TSO header slice were allocated/written before verifying total FIFO space, gve_free_bulk_mbuf() did not clear the first freed slot, and gve_tx_burst_ra() skipped mid-burst ring cleanup and wrote pkt_len instead of data_len into the first segment descriptor of multi-segment non-TSO packets. Fixes: a46583cf43c8 ("net/gve: support Rx/Tx") Cc: stable@dpdk.org Signed-off-by: Ryan Wolfe --- drivers/net/gve/gve_tx.c | 68 +++++++++++++++++++++++++++------------- 1 file changed, 47 insertions(+), 21 deletions(-) diff --git a/drivers/net/gve/gve_tx.c b/drivers/net/gve/gve_tx.c index 279669165f..2cf091f5be 100644 --- a/drivers/net/gve/gve_tx.c +++ b/drivers/net/gve/gve_tx.c @@ -19,6 +19,7 @@ gve_free_bulk_mbuf(struct rte_mbuf **txep, int num) for (s = 0; s < num; s++) { if (txep[s] != NULL) { m = rte_pktmbuf_prefree_seg(txep[s]); + txep[s] = NULL; if (m != NULL) break; } @@ -200,7 +201,7 @@ gve_tx_fill_seg_desc(volatile union gve_tx_desc *desc, uint64_t ol_flags, } static inline bool -is_fifo_avail(struct gve_tx_queue *txq, uint16_t len) +is_fifo_avail(struct gve_tx_queue *txq, uint32_t len) { if (txq->fifo_avail < len) return false; @@ -210,6 +211,26 @@ is_fifo_avail(struct gve_tx_queue *txq, uint16_t len) return false; return true; } + +static inline bool +gve_tx_reclaim_fifo_qpl(struct gve_tx_queue *txq, uint32_t tx_tail, + uint32_t need) +{ + if (txq->tx_tail != tx_tail) { + rte_write32(rte_cpu_to_be_32(tx_tail), txq->qtx_tail); + txq->tx_tail = tx_tail; + } + + do { + uint32_t prev_ntc = txq->next_to_clean; + + gve_tx_clean(txq); + if (is_fifo_avail(txq, need)) + return true; + if (txq->next_to_clean == prev_ntc) + return false; + } while (true); +} static inline uint64_t gve_tx_alloc_from_fifo(struct gve_tx_queue *txq, uint16_t tx_id, uint16_t len) { @@ -266,7 +287,8 @@ gve_tx_burst_qpl(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) txr = txq->tx_desc_ring; - if (txq->nb_free < txq->free_thresh || txq->fifo_avail == 0) + if (txq->nb_free < txq->free_thresh || + txq->fifo_avail < (txq->fifo_size >> 1)) gve_tx_clean(txq); if (txq->sw_nb_free < txq->free_thresh) @@ -304,22 +326,23 @@ gve_tx_burst_qpl(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) (uint32_t)(tx_offload.l2_len + tx_offload.l3_len + tx_offload.l4_len) : tx_pkt->pkt_len; - if (!is_fifo_avail(txq, hlen)) { - gve_tx_clean(txq); - if (!is_fifo_avail(txq, hlen)) + if (ol_flags & RTE_MBUF_F_TX_TCP_SEG) { + uint32_t data_len, need; + + if (unlikely(tx_pkt->pkt_len <= hlen)) + goto end_of_tx; + data_len = tx_pkt->pkt_len - hlen; + need = RTE_ALIGN(hlen, RTE_CACHE_LINE_SIZE) + + RTE_ALIGN(data_len, RTE_CACHE_LINE_SIZE); + if (!is_fifo_avail(txq, need) && + !gve_tx_reclaim_fifo_qpl(txq, tx_tail, need)) goto end_of_tx; + } else if (!is_fifo_avail(txq, hlen) && + !gve_tx_reclaim_fifo_qpl(txq, tx_tail, hlen)) { + goto end_of_tx; } fifo_addr = gve_tx_alloc_from_fifo(txq, tx_id, hlen); - /* For TSO, check if there's enough fifo space for data first */ - if (ol_flags & RTE_MBUF_F_TX_TCP_SEG) { - if (!is_fifo_avail(txq, tx_pkt->pkt_len - hlen)) { - gve_tx_clean(txq); - if (!is_fifo_avail(txq, tx_pkt->pkt_len - hlen)) - goto end_of_tx; - } - } - qpl_write_addr = (void *)(size_t)(fifo_addr + txq->fifo_base); mbuf_header_addr = rte_pktmbuf_read(tx_pkt, 0, hlen, qpl_write_addr); @@ -370,13 +393,14 @@ gve_tx_burst_qpl(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) end_of_tx: if (nb_tx) { - rte_write32(rte_cpu_to_be_32(tx_tail), txq->qtx_tail); - txq->tx_tail = tx_tail; + if (txq->tx_tail != tx_tail) { + rte_write32(rte_cpu_to_be_32(tx_tail), txq->qtx_tail); + txq->tx_tail = tx_tail; + } txq->sw_tail = sw_id; txq->stats.packets += nb_tx; txq->stats.bytes += bytes; - txq->stats.errors += nb_pkts - nb_tx; } return nb_tx; @@ -411,8 +435,11 @@ gve_tx_burst_ra(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) if (ol_flags & RTE_MBUF_F_TX_TCP_SEG) nb_used++; - if (txq->nb_free < nb_used) - goto end_of_tx; + if (txq->nb_free < nb_used) { + gve_tx_clean(txq); + if (txq->nb_free < nb_used) + goto end_of_tx; + } tx_offload.l2_len = tx_pkt->l2_len; tx_offload.l3_len = tx_pkt->l3_len; @@ -424,7 +451,7 @@ gve_tx_burst_ra(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) hlen = ol_flags & RTE_MBUF_F_TX_TCP_SEG ? (uint32_t)(tx_offload.l2_len + tx_offload.l3_len + tx_offload.l4_len) : - tx_pkt->pkt_len; + tx_pkt->data_len; /* * if tso, the driver needs to fill 2 descs for 1 mbuf * so only put this mbuf into the 1st tx entry in sw ring @@ -465,7 +492,6 @@ gve_tx_burst_ra(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) txq->stats.packets += nb_tx; txq->stats.bytes += bytes; - txq->stats.errors += nb_pkts - nb_tx; } return nb_tx; -- 2.56.0.679.gc527c65306-goog