netdev - [PATCH net] vmxnet3: Fix tx queue race condition with XDP

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <20250124090211.110328-1-sankararaman.jayaraman@broadcom.com>
Date: Fri, 24 Jan 2025 14:32:11 +0530
From: Sankararaman Jayaraman <sankararaman.jayaraman@...adcom.com>
To: netdev@...r.kernel.org
Cc: sankararaman.jayaraman@...adcom.com,
	ronak.doshi@...adcom.com,
	bcm-kernel-feedback-list@...adcom.com,
	andrew+netdev@...n.ch,
	davem@...emloft.net,
	u9012063@...il.com,
	kuba@...nel.org,
	edumazet@...gle.com,
	pabeni@...hat.com,
	ast@...nel.org,
	alexandr.lobakin@...el.com,
	alexanderduyck@...com,
	bpf@...r.kernel.org,
	daniel@...earbox.net,
	hawk@...nel.org,
	john.fastabend@...il.com
Subject: [PATCH net] vmxnet3: Fix tx queue race condition with XDP

If XDP traffic runs on a CPU which is greater than or equal to
the number of the Tx queues of the NIC, then vmxnet3_xdp_get_tq()
always picks up queue 0 for transmission as it uses reciprocal scale
instead of simple modulo operation.

vmxnet3_xdp_xmit() and vmxnet3_xdp_xmit_frame() use the above
returned queue without any locking which can lead to race conditions
when multiple XDP xmits run in parallel on different CPU's.

This patch uses a simple module scheme when the current CPU equals or
exceeds the number of Tx queues on the NIC. It also adds locking in
vmxnet3_xdp_xmit() and vmxnet3_xdp_xmit_frame() functions.

Fixes: 54f00cce1178 ("vmxnet3: Add XDP support.")
Signed-off-by: Sankararaman Jayaraman <sankararaman.jayaraman@...adcom.com>
Signed-off-by: Ronak Doshi <ronak.doshi@...adcom.com>
---
 drivers/net/vmxnet3/vmxnet3_xdp.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/drivers/net/vmxnet3/vmxnet3_xdp.c b/drivers/net/vmxnet3/vmxnet3_xdp.c
index 1341374a4588..5f177e77cfcb 100644
--- a/drivers/net/vmxnet3/vmxnet3_xdp.c
+++ b/drivers/net/vmxnet3/vmxnet3_xdp.c
@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 /*
  * Linux driver for VMware's vmxnet3 ethernet NIC.
- * Copyright (C) 2008-2023, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2025, VMware, Inc. All Rights Reserved.
  * Maintained by: pv-drivers@...are.com
  *
  */
@@ -28,7 +28,7 @@ vmxnet3_xdp_get_tq(struct vmxnet3_adapter *adapter)
 	if (likely(cpu < tq_number))
 		tq = &adapter->tx_queue[cpu];
 	else
-		tq = &adapter->tx_queue[reciprocal_scale(cpu, tq_number)];
+		tq = &adapter->tx_queue[cpu % tq_number];
 
 	return tq;
 }
@@ -123,7 +123,9 @@ vmxnet3_xdp_xmit_frame(struct vmxnet3_adapter *adapter,
 	struct page *page;
 	u32 buf_size;
 	u32 dw2;
+	unsigned long irq_flags;
 
+	spin_lock_irqsave(&tq->tx_lock, irq_flags);
 	dw2 = (tq->tx_ring.gen ^ 0x1) << VMXNET3_TXD_GEN_SHIFT;
 	dw2 |= xdpf->len;
 	ctx.sop_txd = tq->tx_ring.base + tq->tx_ring.next2fill;
@@ -134,6 +136,7 @@ vmxnet3_xdp_xmit_frame(struct vmxnet3_adapter *adapter,
 
 	if (vmxnet3_cmd_ring_desc_avail(&tq->tx_ring) == 0) {
 		tq->stats.tx_ring_full++;
+		spin_unlock_irqrestore(&tq->tx_lock, irq_flags);
 		return -ENOSPC;
 	}
 
@@ -142,8 +145,10 @@ vmxnet3_xdp_xmit_frame(struct vmxnet3_adapter *adapter,
 		tbi->dma_addr = dma_map_single(&adapter->pdev->dev,
 					       xdpf->data, buf_size,
 					       DMA_TO_DEVICE);
-		if (dma_mapping_error(&adapter->pdev->dev, tbi->dma_addr))
+		if (dma_mapping_error(&adapter->pdev->dev, tbi->dma_addr)) {
+			spin_unlock_irqrestore(&tq->tx_lock, irq_flags);
 			return -EFAULT;
+		}
 		tbi->map_type |= VMXNET3_MAP_SINGLE;
 	} else { /* XDP buffer from page pool */
 		page = virt_to_page(xdpf->data);
@@ -182,6 +187,7 @@ vmxnet3_xdp_xmit_frame(struct vmxnet3_adapter *adapter,
 	dma_wmb();
 	gdesc->dword[2] = cpu_to_le32(le32_to_cpu(gdesc->dword[2]) ^
 						  VMXNET3_TXD_GEN);
+	spin_unlock_irqrestore(&tq->tx_lock, irq_flags);
 
 	/* No need to handle the case when tx_num_deferred doesn't reach
 	 * threshold. Backend driver at hypervisor side will poll and reset
@@ -226,6 +232,7 @@ vmxnet3_xdp_xmit(struct net_device *dev,
 	struct vmxnet3_adapter *adapter = netdev_priv(dev);
 	struct vmxnet3_tx_queue *tq;
 	int i;
+	struct netdev_queue *nq;
 
 	if (unlikely(test_bit(VMXNET3_STATE_BIT_QUIESCED, &adapter->state)))
 		return -ENETDOWN;
@@ -236,6 +243,9 @@ vmxnet3_xdp_xmit(struct net_device *dev,
 	if (tq->stopped)
 		return -ENETDOWN;
 
+	nq = netdev_get_tx_queue(adapter->netdev, tq->qid);
+
+	__netif_tx_lock(nq, smp_processor_id());
 	for (i = 0; i < n; i++) {
 		if (vmxnet3_xdp_xmit_frame(adapter, frames[i], tq, true)) {
 			tq->stats.xdp_xmit_err++;
@@ -243,6 +253,7 @@ vmxnet3_xdp_xmit(struct net_device *dev,
 		}
 	}
 	tq->stats.xdp_xmit += i;
+	__netif_tx_unlock(nq);
 
 	return i;
 }
-- 
2.25.1