Skip to contentSkip to conversation

vmxnet3: upgrade to version 7

PATCH New Ronak Doshi 2 participants 10 messages Jun 8, 2022 – Jun 9, 2022

Subsystems (3)
lore
Share & tools
lore
#7 lore raw

PATCH v3 7/8 vmxnet3: use ext1 field to indicate encapsulated packet

In reply to Ronak Doshi

Till vmxnet3 version 6, om field of transmit descriptor was used to indicate encapsulated offload packet and msscof was used to indirectly indicate TSO/CSO. From version 7 and later, ext1 field will be used to indicate whether packet is encapsulated or not and om fields will continue to indicate if the packet is TSO or CSO.

---
 drivers/net/vmxnet3/vmxnet3_defs.h | 14 ++++++++------
 drivers/net/vmxnet3/vmxnet3_drv.c  | 18 +++++++++++++++---
 2 files changed, 23 insertions(+), 9 deletions(-)
drivers/net/vmxnet3/vmxnet3_defs.h +8 -6
index cb9dc72f2b3d..41d6767283a6 100644
@@ -148,17 +148,17 @@ struct Vmxnet3_TxDesc {
 
 #ifdef __BIG_ENDIAN_BITFIELD
 	u32 msscof:14;  /* MSS, checksum offset, flags */
-	u32 ext1:1;
+	u32 ext1:1;     /* set to 1 to indicate inner csum/tso, vmxnet3 v7 */
 	u32 dtype:1;    /* descriptor type */
-	u32 oco:1;
+	u32 oco:1;      /* Outer csum offload */
 	u32 gen:1;      /* generation bit */
 	u32 len:14;
 #else
 	u32 len:14;
 	u32 gen:1;      /* generation bit */
-	u32 oco:1;
+	u32 oco:1;      /* Outer csum offload */
 	u32 dtype:1;    /* descriptor type */
-	u32 ext1:1;
+	u32 ext1:1;     /* set to 1 to indicate inner csum/tso, vmxnet3 v7 */
 	u32 msscof:14;  /* MSS, checksum offset, flags */
 #endif  /* __BIG_ENDIAN_BITFIELD */
 
@@ -262,11 +262,13 @@ struct Vmxnet3_RxCompDesc {
 	u32		rqID:10;      /* rx queue/ring ID */
 	u32		sop:1;        /* Start of Packet */
 	u32		eop:1;        /* End of Packet */
-	u32		ext1:2;
+	u32		ext1:2;       /* bit 0: indicating v4/v6/.. is for inner header */
+				      /* bit 1: indicating rssType is based on inner header */
 	u32		rxdIdx:12;    /* Index of the RxDesc */
 #else
 	u32		rxdIdx:12;    /* Index of the RxDesc */
-	u32		ext1:2;
+	u32		ext1:2;       /* bit 0: indicating v4/v6/.. is for inner header */
+				      /* bit 1: indicating rssType is based on inner header */
 	u32		eop:1;        /* End of Packet */
 	u32		sop:1;        /* Start of Packet */
 	u32		rqID:10;      /* rx queue/ring ID */
drivers/net/vmxnet3/vmxnet3_drv.c +15 -3
index 2ee3a2e39f10..5c42b4a008e8 100644
@@ -1161,7 +1161,12 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
 	if (ctx.mss) {
 		if (VMXNET3_VERSION_GE_4(adapter) && skb->encapsulation) {
 			gdesc->txd.hlen = ctx.l4_offset + ctx.l4_hdr_size;
-			gdesc->txd.om = VMXNET3_OM_ENCAP;
+			if (VMXNET3_VERSION_GE_7(adapter)) {
+				gdesc->txd.om = VMXNET3_OM_TSO;
+				gdesc->txd.ext1 = 1;
+			} else {
+				gdesc->txd.om = VMXNET3_OM_ENCAP;
+			}
 			gdesc->txd.msscof = ctx.mss;
 
 			if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)
@@ -1178,8 +1183,15 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
 			    skb->encapsulation) {
 				gdesc->txd.hlen = ctx.l4_offset +
 						  ctx.l4_hdr_size;
-				gdesc->txd.om = VMXNET3_OM_ENCAP;
-				gdesc->txd.msscof = 0;		/* Reserved */
+				if (VMXNET3_VERSION_GE_7(adapter)) {
+					gdesc->txd.om = VMXNET3_OM_CSUM;
+					gdesc->txd.msscof = ctx.l4_offset +
+							    skb->csum_offset;
+					gdesc->txd.ext1 = 1;
+				} else {
+					gdesc->txd.om = VMXNET3_OM_ENCAP;
+					gdesc->txd.msscof = 0;		/* Reserved */
+				}
 			} else {
 				gdesc->txd.hlen = ctx.l4_offset;
 				gdesc->txd.om = VMXNET3_OM_CSUM;
Signature · 1 line
2.11.0
#1 lore raw

PATCH v3 1/8 vmxnet3: prepare for version 7 changes

In reply to Ronak Doshi

vmxnet3 is currently at version 6 and this patch initiates the preparation to accommodate changes for upto version 7. Introduced utility macros for vmxnet3 version 7 comparison and update Copyright information.

---
 drivers/net/vmxnet3/Makefile          | 2 +-
 drivers/net/vmxnet3/upt1_defs.h       | 2 +-
 drivers/net/vmxnet3/vmxnet3_defs.h    | 2 +-
 drivers/net/vmxnet3/vmxnet3_drv.c     | 2 +-
 drivers/net/vmxnet3/vmxnet3_ethtool.c | 2 +-
 drivers/net/vmxnet3/vmxnet3_int.h     | 5 ++++-
 6 files changed, 9 insertions(+), 6 deletions(-)
drivers/net/vmxnet3/Makefile +1 -1
index 7a38925f4165..a666a88ac1ff 100644
@@ -2,7 +2,7 @@
 #
 # Linux driver for VMware's vmxnet3 ethernet NIC.
 #
-# Copyright (C) 2007-2021, VMware, Inc. All Rights Reserved.
+# Copyright (C) 2007-2022, VMware, Inc. All Rights Reserved.
 #
 # This program is free software; you can redistribute it and/or modify it
 # under the terms of the GNU General Public License as published by the
drivers/net/vmxnet3/upt1_defs.h +1 -1
index f9f3a23d1698..41c0660a0c54 100644
@@ -1,7 +1,7 @@
 /*
  * Linux driver for VMware's vmxnet3 ethernet NIC.
  *
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License as published by the
drivers/net/vmxnet3/vmxnet3_defs.h +1 -1
index 74d4e8bc4abc..9f91ebb10137 100644
@@ -1,7 +1,7 @@
 /*
  * Linux driver for VMware's vmxnet3 ethernet NIC.
  *
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License as published by the
drivers/net/vmxnet3/vmxnet3_drv.c +1 -1
index 93e8d119d45f..6fc6a2a26161 100644
@@ -1,7 +1,7 @@
 /*
  * Linux driver for VMware's vmxnet3 ethernet NIC.
  *
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License as published by the
drivers/net/vmxnet3/vmxnet3_ethtool.c +1 -1
index 3172d46c0335..e41e76757c5b 100644
@@ -1,7 +1,7 @@
 /*
  * Linux driver for VMware's vmxnet3 ethernet NIC.
  *
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License as published by the
drivers/net/vmxnet3/vmxnet3_int.h +4 -1
index 7027ff483fa5..5251c3439d6a 100644
@@ -1,7 +1,7 @@
 /*
  * Linux driver for VMware's vmxnet3 ethernet NIC.
  *
- * Copyright (C) 2008-2021, VMware, Inc. All Rights Reserved.
+ * Copyright (C) 2008-2022, VMware, Inc. All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License as published by the
@@ -81,6 +81,7 @@
 	#define VMXNET3_RSS
 #endif
 
+#define VMXNET3_REV_7		6	/* Vmxnet3 Rev. 7 */
 #define VMXNET3_REV_6		5	/* Vmxnet3 Rev. 6 */
 #define VMXNET3_REV_5		4	/* Vmxnet3 Rev. 5 */
 #define VMXNET3_REV_4		3	/* Vmxnet3 Rev. 4 */
@@ -431,6 +432,8 @@ struct vmxnet3_adapter {
 	(adapter->version >= VMXNET3_REV_5 + 1)
 #define VMXNET3_VERSION_GE_6(adapter) \
 	(adapter->version >= VMXNET3_REV_6 + 1)
+#define VMXNET3_VERSION_GE_7(adapter) \
+	(adapter->version >= VMXNET3_REV_7 + 1)
 
 /* must be a multiple of VMXNET3_RING_SIZE_ALIGN */
 #define VMXNET3_DEF_TX_RING_SIZE    512
Signature · 1 line
2.11.0
#0 lore raw

v3 cover vmxnet3: upgrade to version 7

vmxnet3 emulation has recently added several new features including support for uniform passthrough(UPT). To make UPT work vmxnet3 has to be enhanced as per the new specification. This patch series extends the vmxnet3 driver to leverage these new features.

Compatibility is maintained using existing vmxnet3 versioning mechanism as follows:
- new features added to vmxnet3 emulation are associated with new vmxnet3 version viz. vmxnet3 version 7.
- emulation advertises all the versions it supports to the driver.
- during initialization, vmxnet3 driver picks the highest version number supported by both the emulation and the driver and configures emulation to run at that version.

In particular, following changes are introduced:

Patch 1:
This patch introduces utility macros for vmxnet3 version 7 comparison and updates Copyright information.

Patch 2:
This patch adds new capability registers to fine control enablement of individual features based on emulation and passthrough.

Patch 3:
This patch adds support for large passthrough BAR register.

Patch 4:
This patch adds support for out of order rx completion processing.

Patch 5:
This patch introduces new command to set ring buffer sizes to pass this information to the hardware.

Patch 6:
For better performance, hardware has a requirement to limit number of TSO descriptors. This patch adds that support.

Patch 7:
With vmxnet3 version 7, new descriptor fields are used to indicate encapsulation offload.

Patch 8:
With all vmxnet3 version 7 changes incorporated in the vmxnet3 driver, with this patch, the driver can configure emulation to run at vmxnet3 version 7.

Changes in v2->v3:
- use correct byte ordering for ringBufSize

Changes in v2:
- use local rss_fields variable for the rss capability checks in patch 2

Ronak Doshi (8):
vmxnet3: prepare for version 7 changes
vmxnet3: add support for capability registers
vmxnet3: add support for large passthrough BAR register
vmxnet3: add support for out of order rx completion
vmxnet3: add command to set ring buffer sizes
vmxnet3: limit number of TXDs used for TSO packet
vmxnet3: use ext1 field to indicate encapsulated packet
vmxnet3: update to version 7

 drivers/net/vmxnet3/Makefile          |   2 +-
 drivers/net/vmxnet3/upt1_defs.h       |   2 +-
 drivers/net/vmxnet3/vmxnet3_defs.h    |  80 ++++++++--
 drivers/net/vmxnet3/vmxnet3_drv.c     | 291 ++++++++++++++++++++++++++++++----
 drivers/net/vmxnet3/vmxnet3_ethtool.c | 116 ++++++++++++--
 drivers/net/vmxnet3/vmxnet3_int.h     |  24 ++-
 6 files changed, 457 insertions(+), 58 deletions(-)
Signature · 1 line
2.11.0
#3 lore raw

PATCH v3 3/8 vmxnet3: add support for large passthrough BAR register

For vmxnet3 to work in UPT mode, the BAR sizes have been increased. The PT page has been extended to 2 pages and also includes OOB pages as a part of PT BAR. This patch enhances vmxnet3 to use appropriate BAR offsets based on the capability registered. To use new offsets, VMXNET3_CAP_LARGE_BAR needs to be set by the device. If it is not set then the device will use legacy PT page layout.

---
 drivers/net/vmxnet3/vmxnet3_defs.h    | 14 ++++++++++++--
 drivers/net/vmxnet3/vmxnet3_drv.c     | 25 ++++++++++++++++++++-----
 drivers/net/vmxnet3/vmxnet3_ethtool.c |  6 +++---
 drivers/net/vmxnet3/vmxnet3_int.h     |  3 +++
 4 files changed, 38 insertions(+), 10 deletions(-)
drivers/net/vmxnet3/vmxnet3_defs.h +12 -2
index 0157155ff677..8d626611ab2d 100644
@@ -57,8 +57,18 @@ enum {
 	VMXNET3_REG_RXPROD2	= 0xA00	 /* Rx Producer Index for ring 2 */
 };
 
-#define VMXNET3_PT_REG_SIZE     4096	/* BAR 0 */
-#define VMXNET3_VD_REG_SIZE     4096	/* BAR 1 */
+/* For Large PT BAR, the following offset to DB register */
+enum {
+	VMXNET3_REG_LB_TXPROD   = 0x1000, /* Tx Producer Index */
+	VMXNET3_REG_LB_RXPROD   = 0x1400, /* Rx Producer Index for ring 1 */
+	VMXNET3_REG_LB_RXPROD2  = 0x1800, /* Rx Producer Index for ring 2 */
+};
+
+#define VMXNET3_PT_REG_SIZE         4096		/* BAR 0 */
+#define VMXNET3_LARGE_PT_REG_SIZE   8192		/* large PT pages */
+#define VMXNET3_VD_REG_SIZE         4096		/* BAR 1 */
+#define VMXNET3_LARGE_BAR0_REG_SIZE (4096 * 4096)	/* LARGE BAR 0 */
+#define VMXNET3_OOB_REG_SIZE        (4094 * 4096)	/* OOB pages */
 
 #define VMXNET3_REG_ALIGN       8	/* All registers are 8-byte aligned. */
 #define VMXNET3_REG_ALIGN_MASK  0x7
drivers/net/vmxnet3/vmxnet3_drv.c +20 -5
index edc4f23d4965..93f237db463d 100644
@@ -1207,7 +1207,7 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
 	if (tx_num_deferred >= le32_to_cpu(tq->shared->txThreshold)) {
 		tq->shared->txNumDeferred = 0;
 		VMXNET3_WRITE_BAR0_REG(adapter,
-				       VMXNET3_REG_TXPROD + tq->qid * 8,
+				       adapter->tx_prod_offset + tq->qid * 8,
 				       tq->tx_ring.next2fill);
 	}
 
@@ -1359,8 +1359,8 @@ static int
 vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
 		       struct vmxnet3_adapter *adapter, int quota)
 {
-	static const u32 rxprod_reg[2] = {
-		VMXNET3_REG_RXPROD, VMXNET3_REG_RXPROD2
+	u32 rxprod_reg[2] = {
+		adapter->rx_prod_offset, adapter->rx_prod2_offset
 	};
 	u32 num_pkts = 0;
 	bool skip_page_frags = false;
@@ -2783,9 +2783,9 @@ vmxnet3_activate_dev(struct vmxnet3_adapter *adapter)
 
 	for (i = 0; i < adapter->num_rx_queues; i++) {
 		VMXNET3_WRITE_BAR0_REG(adapter,
-				VMXNET3_REG_RXPROD + i * VMXNET3_REG_ALIGN,
+				adapter->rx_prod_offset + i * VMXNET3_REG_ALIGN,
 				adapter->rx_queue[i].rx_ring[0].next2fill);
-		VMXNET3_WRITE_BAR0_REG(adapter, (VMXNET3_REG_RXPROD2 +
+		VMXNET3_WRITE_BAR0_REG(adapter, (adapter->rx_prod2_offset +
 				(i * VMXNET3_REG_ALIGN)),
 				adapter->rx_queue[i].rx_ring[1].next2fill);
 	}
@@ -3608,6 +3608,10 @@ vmxnet3_probe_device(struct pci_dev *pdev,
 	if (VMXNET3_VERSION_GE_7(adapter)) {
 		adapter->devcap_supported[0] = VMXNET3_READ_BAR1_REG(adapter, VMXNET3_REG_DCR);
 		adapter->ptcap_supported[0] = VMXNET3_READ_BAR1_REG(adapter, VMXNET3_REG_PTCR);
+		if (adapter->devcap_supported[0] & (1UL << VMXNET3_CAP_LARGE_BAR)) {
+			adapter->dev_caps[0] = adapter->devcap_supported[0] &
+							(1UL << VMXNET3_CAP_LARGE_BAR);
+		}
 		if (adapter->dev_caps[0])
 			VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_DCR, adapter->dev_caps[0]);
 
@@ -3617,6 +3621,17 @@ vmxnet3_probe_device(struct pci_dev *pdev,
 		spin_unlock_irqrestore(&adapter->cmd_lock, flags);
 	}
 
+	if (VMXNET3_VERSION_GE_7(adapter) &&
+	    adapter->dev_caps[0] & (1UL << VMXNET3_CAP_LARGE_BAR)) {
+		adapter->tx_prod_offset = VMXNET3_REG_LB_TXPROD;
+		adapter->rx_prod_offset = VMXNET3_REG_LB_RXPROD;
+		adapter->rx_prod2_offset = VMXNET3_REG_LB_RXPROD2;
+	} else {
+		adapter->tx_prod_offset = VMXNET3_REG_TXPROD;
+		adapter->rx_prod_offset = VMXNET3_REG_RXPROD;
+		adapter->rx_prod2_offset = VMXNET3_REG_RXPROD2;
+	}
+
 	if (VMXNET3_VERSION_GE_6(adapter)) {
 		spin_lock_irqsave(&adapter->cmd_lock, flags);
 		VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_CMD,
drivers/net/vmxnet3/vmxnet3_ethtool.c +3 -3
index 458f2da1ebab..397b268f7bc5 100644
@@ -520,7 +520,7 @@ vmxnet3_get_regs(struct net_device *netdev, struct ethtool_regs *regs, void *p)
 	for (i = 0; i < adapter->num_tx_queues; i++) {
 		struct vmxnet3_tx_queue *tq = &adapter->tx_queue[i];
 
-		buf[j++] = VMXNET3_READ_BAR0_REG(adapter, VMXNET3_REG_TXPROD +
+		buf[j++] = VMXNET3_READ_BAR0_REG(adapter, adapter->tx_prod_offset +
 						 i * VMXNET3_REG_ALIGN);
 
 		buf[j++] = VMXNET3_GET_ADDR_LO(tq->tx_ring.basePA);
@@ -548,9 +548,9 @@ vmxnet3_get_regs(struct net_device *netdev, struct ethtool_regs *regs, void *p)
 	for (i = 0; i < adapter->num_rx_queues; i++) {
 		struct vmxnet3_rx_queue *rq = &adapter->rx_queue[i];
 
-		buf[j++] =  VMXNET3_READ_BAR0_REG(adapter, VMXNET3_REG_RXPROD +
+		buf[j++] =  VMXNET3_READ_BAR0_REG(adapter, adapter->rx_prod_offset +
 						  i * VMXNET3_REG_ALIGN);
-		buf[j++] =  VMXNET3_READ_BAR0_REG(adapter, VMXNET3_REG_RXPROD2 +
+		buf[j++] =  VMXNET3_READ_BAR0_REG(adapter, adapter->rx_prod2_offset +
 						  i * VMXNET3_REG_ALIGN);
 
 		buf[j++] = VMXNET3_GET_ADDR_LO(rq->rx_ring[0].basePA);
drivers/net/vmxnet3/vmxnet3_int.h +3 -0
index a7c8f80702c2..a4f832f0ad5b 100644
@@ -406,6 +406,9 @@ struct vmxnet3_adapter {
 	u32    devcap_supported[8];
 	u32    ptcap_supported[8];
 	u32    dev_caps[8];
+	u16    tx_prod_offset;
+	u16    rx_prod_offset;
+	u16    rx_prod2_offset;
 };
 
 #define VMXNET3_WRITE_BAR0_REG(adapter, reg, val)  \
Signature · 1 line
2.11.0
#4 lore raw

PATCH v3 4/8 vmxnet3: add support for out of order rx completion

In reply to Ronak Doshi

Currently, vmxnet3 processes rx completions in-order i.e. no out of order completion descriptor is expected. With UPT, if hardware supports LRO, then hardware can report out of order rx completions. This patch enhances vmxnet3 to add this support. This supports gets effective only when the corresponding feature bit is set.

Also, minor enhancements are done for performance.

---
 drivers/net/vmxnet3/vmxnet3_drv.c | 70 ++++++++++++++++++++++++++++++++-------
 drivers/net/vmxnet3/vmxnet3_int.h |  5 +++
 2 files changed, 63 insertions(+), 12 deletions(-)
drivers/net/vmxnet3/vmxnet3_drv.c +58 -12
index 93f237db463d..94ca3bc1d540 100644
@@ -585,6 +585,7 @@ vmxnet3_rq_alloc_rx_buf(struct vmxnet3_rx_queue *rq, u32 ring_idx,
 
 		rbi = rbi_base + ring->next2fill;
 		gd = ring->base + ring->next2fill;
+		rbi->comp_state = VMXNET3_RXD_COMP_PENDING;
 
 		if (rbi->buf_type == VMXNET3_RX_BUF_SKB) {
 			if (rbi->skb == NULL) {
@@ -644,8 +645,10 @@ vmxnet3_rq_alloc_rx_buf(struct vmxnet3_rx_queue *rq, u32 ring_idx,
 
 		/* Fill the last buffer but dont mark it ready, or else the
 		 * device will think that the queue is full */
-		if (num_allocated == num_to_alloc)
+		if (num_allocated == num_to_alloc) {
+			rbi->comp_state = VMXNET3_RXD_COMP_DONE;
 			break;
+		}
 
 		gd->dword[2] |= cpu_to_le32(ring->gen << VMXNET3_RXD_GEN_SHIFT);
 		num_allocated++;
@@ -1367,6 +1370,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
 	struct Vmxnet3_RxCompDesc *rcd;
 	struct vmxnet3_rx_ctx *ctx = &rq->rx_ctx;
 	u16 segCnt = 0, mss = 0;
+	int comp_offset, fill_offset;
 #ifdef __BIG_ENDIAN_BITFIELD
 	struct Vmxnet3_RxDesc rxCmdDesc;
 	struct Vmxnet3_RxCompDesc rxComp;
@@ -1639,9 +1643,15 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
 
 rcd_done:
 		/* device may have skipped some rx descs */
-		ring->next2comp = idx;
-		num_to_alloc = vmxnet3_cmd_ring_desc_avail(ring);
 		ring = rq->rx_ring + ring_idx;
+		rbi->comp_state = VMXNET3_RXD_COMP_DONE;
+
+		comp_offset = vmxnet3_cmd_ring_desc_avail(ring);
+		fill_offset = (idx > ring->next2fill ? 0 : ring->size) +
+			      idx - ring->next2fill - 1;
+		if (!ring->isOutOfOrder || fill_offset >= comp_offset)
+			ring->next2comp = idx;
+		num_to_alloc = vmxnet3_cmd_ring_desc_avail(ring);
 
 		/* Ensure that the writes to rxd->gen bits will be observed
 		 * after all other writes to rxd objects.
@@ -1649,18 +1659,38 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
 		dma_wmb();
 
 		while (num_to_alloc) {
-			vmxnet3_getRxDesc(rxd, &ring->base[ring->next2fill].rxd,
-					  &rxCmdDesc);
-			BUG_ON(!rxd->addr);
-
-			/* Recv desc is ready to be used by the device */
-			rxd->gen = ring->gen;
-			vmxnet3_cmd_ring_adv_next2fill(ring);
-			num_to_alloc--;
+			rbi = rq->buf_info[ring_idx] + ring->next2fill;
+			if (!(adapter->dev_caps[0] & (1UL << VMXNET3_CAP_OOORX_COMP)))
+				goto refill_buf;
+			if (ring_idx == 0) {
+				/* ring0 Type1 buffers can get skipped; re-fill them */
+				if (rbi->buf_type != VMXNET3_RX_BUF_SKB)
+					goto refill_buf;
+			}
+			if (rbi->comp_state == VMXNET3_RXD_COMP_DONE) {
+refill_buf:
+				vmxnet3_getRxDesc(rxd, &ring->base[ring->next2fill].rxd,
+						  &rxCmdDesc);
+				WARN_ON(!rxd->addr);
+
+				/* Recv desc is ready to be used by the device */
+				rxd->gen = ring->gen;
+				vmxnet3_cmd_ring_adv_next2fill(ring);
+				rbi->comp_state = VMXNET3_RXD_COMP_PENDING;
+				num_to_alloc--;
+			} else {
+				/* rx completion hasn't occurred */
+				ring->isOutOfOrder = 1;
+				break;
+			}
+		}
+
+		if (num_to_alloc == 0) {
+			ring->isOutOfOrder = 0;
 		}
 
 		/* if needed, update the register */
-		if (unlikely(rq->shared->updateRxProd)) {
+		if (unlikely(rq->shared->updateRxProd) && (ring->next2fill & 0xf) == 0) {
 			VMXNET3_WRITE_BAR0_REG(adapter,
 					       rxprod_reg[ring_idx] + rq->qid * 8,
 					       ring->next2fill);
@@ -1824,6 +1854,7 @@ vmxnet3_rq_init(struct vmxnet3_rx_queue *rq,
 		memset(rq->rx_ring[i].base, 0, rq->rx_ring[i].size *
 		       sizeof(struct Vmxnet3_RxDesc));
 		rq->rx_ring[i].gen = VMXNET3_INIT_GEN;
+		rq->rx_ring[i].isOutOfOrder = 0;
 	}
 	if (vmxnet3_rq_alloc_rx_buf(rq, 0, rq->rx_ring[0].size - 1,
 				    adapter) == 0) {
@@ -2014,8 +2045,17 @@ vmxnet3_poll_rx_only(struct napi_struct *napi, int budget)
 	rxd_done = vmxnet3_rq_rx_complete(rq, adapter, budget);
 
 	if (rxd_done < budget) {
+		struct Vmxnet3_RxCompDesc *rcd;
+#ifdef __BIG_ENDIAN_BITFIELD
+		struct Vmxnet3_RxCompDesc rxComp;
+#endif
 		napi_complete_done(napi, rxd_done);
 		vmxnet3_enable_intr(adapter, rq->comp_ring.intr_idx);
+		/* after unmasking the interrupt, check if any descriptors were completed */
+		vmxnet3_getRxComp(rcd, &rq->comp_ring.base[rq->comp_ring.next2proc].rcd,
+				  &rxComp);
+		if (rcd->gen == rq->comp_ring.gen && napi_reschedule(napi))
+			vmxnet3_disable_intr(adapter, rq->comp_ring.intr_idx);
 	}
 	return rxd_done;
 }
@@ -3612,6 +3652,12 @@ vmxnet3_probe_device(struct pci_dev *pdev,
 			adapter->dev_caps[0] = adapter->devcap_supported[0] &
 							(1UL << VMXNET3_CAP_LARGE_BAR);
 		}
+		if (!(adapter->ptcap_supported[0] & (1UL << VMXNET3_DCR_ERROR)) &&
+		    adapter->ptcap_supported[0] & (1UL << VMXNET3_CAP_OOORX_COMP) &&
+		    adapter->devcap_supported[0] & (1UL << VMXNET3_CAP_OOORX_COMP)) {
+			adapter->dev_caps[0] |= adapter->devcap_supported[0] &
+						(1UL << VMXNET3_CAP_OOORX_COMP);
+		}
 		if (adapter->dev_caps[0])
 			VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_DCR, adapter->dev_caps[0]);
 
drivers/net/vmxnet3/vmxnet3_int.h +5 -0
index a4f832f0ad5b..5b495ef253e8 100644
@@ -136,6 +136,7 @@ struct vmxnet3_cmd_ring {
 	u32		next2fill;
 	u32		next2comp;
 	u8		gen;
+	u8              isOutOfOrder;
 	dma_addr_t	basePA;
 };
 
@@ -260,9 +261,13 @@ enum vmxnet3_rx_buf_type {
 	VMXNET3_RX_BUF_PAGE = 2
 };
 
+#define VMXNET3_RXD_COMP_PENDING        0
+#define VMXNET3_RXD_COMP_DONE           1
+
 struct vmxnet3_rx_buf_info {
 	enum vmxnet3_rx_buf_type buf_type;
 	u16     len;
+	u8      comp_state;
 	union {
 		struct sk_buff *skb;
 		struct page    *page;
Signature · 1 line
2.11.0
#8 lore raw

PATCH v3 8/8 vmxnet3: update to version 7

In reply to Ronak Doshi

With all vmxnet3 version 7 changes incorporated in the vmxnet3 driver, the driver can configure emulation to run at vmxnet3 version 7, provided the emulation advertises support for version 7.

---
 drivers/net/vmxnet3/vmxnet3_drv.c | 7 ++++++-
 drivers/net/vmxnet3/vmxnet3_int.h | 4 ++--
 2 files changed, 8 insertions(+), 3 deletions(-)
drivers/net/vmxnet3/vmxnet3_drv.c +6 -1
index 5c42b4a008e8..1565e1808a19 100644
@@ -3659,7 +3659,12 @@ vmxnet3_probe_device(struct pci_dev *pdev,
 		goto err_alloc_pci;
 
 	ver = VMXNET3_READ_BAR1_REG(adapter, VMXNET3_REG_VRRS);
-	if (ver & (1 << VMXNET3_REV_6)) {
+	if (ver & (1 << VMXNET3_REV_7)) {
+		VMXNET3_WRITE_BAR1_REG(adapter,
+				       VMXNET3_REG_VRRS,
+				       1 << VMXNET3_REV_7);
+		adapter->version = VMXNET3_REV_7 + 1;
+	} else if (ver & (1 << VMXNET3_REV_6)) {
 		VMXNET3_WRITE_BAR1_REG(adapter,
 				       VMXNET3_REG_VRRS,
 				       1 << VMXNET3_REV_6);
drivers/net/vmxnet3/vmxnet3_int.h +2 -2
index cb87731f5f1c..3367db23aa13 100644
@@ -69,12 +69,12 @@
 /*
  * Version numbers
  */
-#define VMXNET3_DRIVER_VERSION_STRING   "1.6.0.0-k"
+#define VMXNET3_DRIVER_VERSION_STRING   "1.7.0.0-k"
 
 /* Each byte of this 32-bit integer encodes a version number in
  * VMXNET3_DRIVER_VERSION_STRING.
  */
-#define VMXNET3_DRIVER_VERSION_NUM      0x01060000
+#define VMXNET3_DRIVER_VERSION_NUM      0x01070000
 
 #if defined(CONFIG_PCI_MSI)
 	/* RSS only makes sense if MSI-X is supported. */
Signature · 1 line
2.11.0
#5 lore raw

PATCH v3 5/8 vmxnet3: add command to set ring buffer sizes

In reply to Ronak Doshi

This patch adds a new command to set ring buffer sizes. This is required to pass the buffer size information to passthrough devices. For performance reasons, with version7 and later, ring1 will contain only mtu size buffers (bound to 3K). Packets > 3K will use both ring1 and ring2.

Also, ring sizes are round down to power of 2 and ring2 default size is increased to 512.

---
 drivers/net/vmxnet3/vmxnet3_defs.h    | 11 +++++++
 drivers/net/vmxnet3/vmxnet3_drv.c     | 57 +++++++++++++++++++++++++++--------
 drivers/net/vmxnet3/vmxnet3_ethtool.c |  7 +++++
 drivers/net/vmxnet3/vmxnet3_int.h     |  3 +-
 4 files changed, 65 insertions(+), 13 deletions(-)
drivers/net/vmxnet3/vmxnet3_defs.h +11 -0
index 8d626611ab2d..415e4c9993ef 100644
@@ -99,6 +99,9 @@ enum {
 	VMXNET3_CMD_SET_COALESCE,
 	VMXNET3_CMD_REGISTER_MEMREGS,
 	VMXNET3_CMD_SET_RSS_FIELDS,
+	VMXNET3_CMD_RESERVED4,
+	VMXNET3_CMD_RESERVED5,
+	VMXNET3_CMD_SET_RING_BUFFER_SIZE,
 
 	VMXNET3_CMD_FIRST_GET = 0xF00D0000,
 	VMXNET3_CMD_GET_QUEUE_STATUS = VMXNET3_CMD_FIRST_GET,
@@ -743,6 +746,13 @@ enum Vmxnet3_RSSField {
 	VMXNET3_RSS_FIELDS_ESPIP6 = 0x0020,
 };
 
+struct Vmxnet3_RingBufferSize {
+	__le16             ring1BufSizeType0;
+	__le16             ring1BufSizeType1;
+	__le16             ring2BufSizeType1;
+	__le16             pad;
+};
+
 /* If the command data <= 16 bytes, use the shared memory directly.
  * otherwise, use variable length configuration descriptor.
  */
@@ -750,6 +760,7 @@ union Vmxnet3_CmdInfo {
 	struct Vmxnet3_VariableLenConfDesc	varConf;
 	struct Vmxnet3_SetPolling		setPolling;
 	enum   Vmxnet3_RSSField                 setRssFields;
+	struct Vmxnet3_RingBufferSize           ringBufSize;
 	__le64					data[2];
 };
 
drivers/net/vmxnet3/vmxnet3_drv.c +45 -12

82 lines not on this page · raw message

drivers/net/vmxnet3/vmxnet3_ethtool.c +7 -0
index 397b268f7bc5..ce3993282c0f 100644
@@ -718,6 +718,13 @@ vmxnet3_set_ringparam(struct net_device *netdev,
 	new_rx_ring2_size = min_t(u32, new_rx_ring2_size,
 				  VMXNET3_RX_RING2_MAX_SIZE);
 
+	/* For v7 and later, keep ring size power of 2 for UPT */
+	if (VMXNET3_VERSION_GE_7(adapter)) {
+		new_tx_ring_size = rounddown_pow_of_two(new_tx_ring_size);
+		new_rx_ring_size = rounddown_pow_of_two(new_rx_ring_size);
+		new_rx_ring2_size = rounddown_pow_of_two(new_rx_ring2_size);
+	}
+
 	/* rx data ring buffer size has to be a multiple of
 	 * VMXNET3_RXDATA_DESC_SIZE_ALIGN
 	 */
drivers/net/vmxnet3/vmxnet3_int.h +2 -1
index 5b495ef253e8..cb87731f5f1c 100644
@@ -408,6 +408,7 @@ struct vmxnet3_adapter {
 	dma_addr_t pm_conf_pa;
 	dma_addr_t rss_conf_pa;
 	bool   queuesExtEnabled;
+	struct Vmxnet3_RingBufferSize     ringBufSize;
 	u32    devcap_supported[8];
 	u32    ptcap_supported[8];
 	u32    dev_caps[8];
@@ -449,7 +450,7 @@ struct vmxnet3_adapter {
 /* must be a multiple of VMXNET3_RING_SIZE_ALIGN */
 #define VMXNET3_DEF_TX_RING_SIZE    512
 #define VMXNET3_DEF_RX_RING_SIZE    1024
-#define VMXNET3_DEF_RX_RING2_SIZE   256
+#define VMXNET3_DEF_RX_RING2_SIZE   512
 
 #define VMXNET3_DEF_RXDATA_DESC_SIZE 128
 
Signature · 1 line
2.11.0
#6 lore raw

PATCH v3 6/8 vmxnet3: limit number of TXDs used for TSO packet

In reply to Ronak Doshi

Currently, vmxnet3 does not have a limit on number of descriptors used for a TSO packet. However, with UPT, for hardware performance reasons, this patch limits the number of transmit descriptors to 24 for a TSO packet.

---
 drivers/net/vmxnet3/vmxnet3_defs.h |  2 ++
 drivers/net/vmxnet3/vmxnet3_drv.c  | 17 +++++++++++++++++
 2 files changed, 19 insertions(+)
drivers/net/vmxnet3/vmxnet3_defs.h +2 -0
index 415e4c9993ef..cb9dc72f2b3d 100644
@@ -400,6 +400,8 @@ union Vmxnet3_GenericDesc {
 
 /* max # of tx descs for a non-tso pkt */
 #define VMXNET3_MAX_TXD_PER_PKT 16
+/* max # of tx descs for a tso pkt */
+#define VMXNET3_MAX_TSO_TXD_PER_PKT 24
 
 /* Max size of a single rx buffer */
 #define VMXNET3_MAX_RX_BUF_SIZE  ((1 << 14) - 1)
drivers/net/vmxnet3/vmxnet3_drv.c +17 -0
index c2f99c01a130..2ee3a2e39f10 100644
@@ -1061,6 +1061,23 @@ vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
 			}
 			tq->stats.copy_skb_header++;
 		}
+		if (unlikely(count > VMXNET3_MAX_TSO_TXD_PER_PKT)) {
+			/* tso pkts must not use more than
+			 * VMXNET3_MAX_TSO_TXD_PER_PKT entries
+			 */
+			if (skb_linearize(skb) != 0) {
+				tq->stats.drop_too_many_frags++;
+				goto drop_pkt;
+			}
+			tq->stats.linearized++;
+
+			/* recalculate the # of descriptors to use */
+			count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + 1;
+			if (unlikely(count > VMXNET3_MAX_TSO_TXD_PER_PKT)) {
+				tq->stats.drop_too_many_frags++;
+				goto drop_pkt;
+			}
+		}
 		if (skb->encapsulation) {
 			vmxnet3_prepare_inner_tso(skb, &ctx);
 		} else {
Signature · 1 line
2.11.0
#2 lore raw

PATCH v3 2/8 vmxnet3: add support for capability registers

In reply to Ronak Doshi

This patch enhances vmxnet3 to suuport capability registers which allows it to enable features selectively. The DCR register tracks the capabilities vmxnet3 device supports. The PTCR register states the capabilities that the passthrough device supports.

With the help of these registers, vmxnet3 can enable only those features which the passthrough device supoprts. This allows smooth trasition to Uniform-Passthrough (UPT) mode if the virtual nic requests it. If PTCR register returns nothing or error it means UPT is not being requested and vnic will continue in emulation mode.

---
 drivers/net/vmxnet3/vmxnet3_defs.h    |  37 ++++++++++++-
 drivers/net/vmxnet3/vmxnet3_drv.c     |  97 ++++++++++++++++++++++++++++++++
 drivers/net/vmxnet3/vmxnet3_ethtool.c | 101 ++++++++++++++++++++++++++++++++--
 drivers/net/vmxnet3/vmxnet3_int.h     |   4 ++
 4 files changed, 233 insertions(+), 6 deletions(-)
drivers/net/vmxnet3/vmxnet3_defs.h +36 -1
index 9f91ebb10137..0157155ff677 100644
@@ -40,7 +40,13 @@ enum {
 	VMXNET3_REG_MACL	= 0x28,	/* MAC Address Low */
 	VMXNET3_REG_MACH	= 0x30,	/* MAC Address High */
 	VMXNET3_REG_ICR		= 0x38,	/* Interrupt Cause Register */
-	VMXNET3_REG_ECR		= 0x40	/* Event Cause Register */
+	VMXNET3_REG_ECR		= 0x40, /* Event Cause Register */
+	VMXNET3_REG_DCR         = 0x48, /* Device capability register,
+					 * from 0x48 to 0x80
+					 */
+	VMXNET3_REG_PTCR        = 0x88, /* Passthru capbility register
+					 * from 0x88 to 0xb0
+					 */
 };
 
 /* BAR 0 */
@@ -101,6 +107,9 @@ enum {
 	VMXNET3_CMD_GET_RESERVED2,
 	VMXNET3_CMD_GET_RESERVED3,
 	VMXNET3_CMD_GET_MAX_QUEUES_CONF,
+	VMXNET3_CMD_GET_RESERVED4,
+	VMXNET3_CMD_GET_MAX_CAPABILITIES,
+	VMXNET3_CMD_GET_DCR0_REG,
 };
 
 /*
@@ -801,4 +810,30 @@ struct Vmxnet3_DriverShared {
 #define VMXNET3_LINK_UP         (10000 << 16 | 1)    /* 10 Gbps, up */
 #define VMXNET3_LINK_DOWN       0
 
+#define VMXNET3_DCR_ERROR                          31   /* error when bit 31 of DCR is set */
+#define VMXNET3_CAP_UDP_RSS                        0    /* bit 0 of DCR 0 */
+#define VMXNET3_CAP_ESP_RSS_IPV4                   1    /* bit 1 of DCR 0 */
+#define VMXNET3_CAP_GENEVE_CHECKSUM_OFFLOAD        2    /* bit 2 of DCR 0 */
+#define VMXNET3_CAP_GENEVE_TSO                     3    /* bit 3 of DCR 0 */
+#define VMXNET3_CAP_VXLAN_CHECKSUM_OFFLOAD         4    /* bit 4 of DCR 0 */
+#define VMXNET3_CAP_VXLAN_TSO                      5    /* bit 5 of DCR 0 */
+#define VMXNET3_CAP_GENEVE_OUTER_CHECKSUM_OFFLOAD  6    /* bit 6 of DCR 0 */
+#define VMXNET3_CAP_VXLAN_OUTER_CHECKSUM_OFFLOAD   7    /* bit 7 of DCR 0 */
+#define VMXNET3_CAP_PKT_STEERING_IPV4              8    /* bit 8 of DCR 0 */
+#define VMXNET3_CAP_VERSION_4_MAX                  VMXNET3_CAP_PKT_STEERING_IPV4
+#define VMXNET3_CAP_ESP_RSS_IPV6                   9    /* bit 9 of DCR 0 */
+#define VMXNET3_CAP_VERSION_5_MAX                  VMXNET3_CAP_ESP_RSS_IPV6
+#define VMXNET3_CAP_ESP_OVER_UDP_RSS               10   /* bit 10 of DCR 0 */
+#define VMXNET3_CAP_INNER_RSS                      11   /* bit 11 of DCR 0 */
+#define VMXNET3_CAP_INNER_ESP_RSS                  12   /* bit 12 of DCR 0 */
+#define VMXNET3_CAP_CRC32_HASH_FUNC                13   /* bit 13 of DCR 0 */
+#define VMXNET3_CAP_VERSION_6_MAX                  VMXNET3_CAP_CRC32_HASH_FUNC
+#define VMXNET3_CAP_OAM_FILTER                     14   /* bit 14 of DCR 0 */
+#define VMXNET3_CAP_ESP_QS                         15   /* bit 15 of DCR 0 */
+#define VMXNET3_CAP_LARGE_BAR                      16   /* bit 16 of DCR 0 */
+#define VMXNET3_CAP_OOORX_COMP                     17   /* bit 17 of DCR 0 */
+#define VMXNET3_CAP_VERSION_7_MAX                  18
+/* when new capability is introduced, update VMXNET3_CAP_MAX */
+#define VMXNET3_CAP_MAX                            VMXNET3_CAP_VERSION_7_MAX
+
 #endif /* _VMXNET3_DEFS_H_ */
drivers/net/vmxnet3/vmxnet3_drv.c +97 -0

121 lines not on this page · raw message

drivers/net/vmxnet3/vmxnet3_ethtool.c +96 -5

131 lines not on this page · raw message

drivers/net/vmxnet3/vmxnet3_int.h +4 -0
index 5251c3439d6a..a7c8f80702c2 100644
@@ -403,6 +403,9 @@ struct vmxnet3_adapter {
 	dma_addr_t pm_conf_pa;
 	dma_addr_t rss_conf_pa;
 	bool   queuesExtEnabled;
+	u32    devcap_supported[8];
+	u32    ptcap_supported[8];
+	u32    dev_caps[8];
 };
 
 #define VMXNET3_WRITE_BAR0_REG(adapter, reg, val)  \
@@ -497,6 +500,7 @@ void vmxnet3_set_ethtool_ops(struct net_device *netdev);
 
 void vmxnet3_get_stats64(struct net_device *dev,
 			 struct rtnl_link_stats64 *stats);
+bool vmxnet3_check_ptcapability(u32 cap_supported, u32 cap);
 
 extern char vmxnet3_driver_name[];
 #endif
Signature · 1 line
2.11.0
#9 lore raw
In reply to Ronak Doshi
This series was applied to netdev/net-next.git (master) by Paolo Abeni <[email protected]>:

Hello:

This series was applied to netdev/net-next.git (master) by Paolo Abeni <[email protected]>:

Quoted 14 lines · On Tue, 7 Jun 2022 20:23:45 -0700 you wrote:

vmxnet3 emulation has recently added several new features including support for uniform passthrough(UPT). To make UPT work vmxnet3 has to be enhanced as per the new specification. This patch series extends the vmxnet3 driver to leverage these new features.

Compatibility is maintained using existing vmxnet3 versioning mechanism as follows:
- new features added to vmxnet3 emulation are associated with new vmxnet3 version viz. vmxnet3 version 7.
- emulation advertises all the versions it supports to the driver.
- during initialization, vmxnet3 driver picks the highest version number supported by both the emulation and the driver and configures emulation to run at that version.

Quoted 1 line

[...]

Here is the summary with links:
- [v3,net-next,1/8] vmxnet3: prepare for version 7 changes

    https://git.kernel.org/netdev/net-next/c/55f0395fcace
  - [v3,net-next,2/8] vmxnet3: add support for capability registers
    https://git.kernel.org/netdev/net-next/c/6f91f4ba046e
  - [v3,net-next,3/8] vmxnet3: add support for large passthrough BAR register
    https://git.kernel.org/netdev/net-next/c/543fb6740541
  - [v3,net-next,4/8] vmxnet3: add support for out of order rx completion
    https://git.kernel.org/netdev/net-next/c/2c5a5748105a
  - [v3,net-next,5/8] vmxnet3: add command to set ring buffer sizes
    (no matching commit)
  - [v3,net-next,6/8] vmxnet3: limit number of TXDs used for TSO packet
    https://git.kernel.org/netdev/net-next/c/d2857b99a74b
  - [v3,net-next,7/8] vmxnet3: use ext1 field to indicate encapsulated packet
    https://git.kernel.org/netdev/net-next/c/60cafa0395c2
  - [v3,net-next,8/8] vmxnet3: update to version 7
    https://git.kernel.org/netdev/net-next/c/acc38e041bd3

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html