Skip to contentSkip to conversation

vmxnet3: Add XDP support.

RFC PATCH Superseded William Tu 2 participants 3 messages Jan 8, 2023 – Jan 10, 2023

Subsystems (3)

v11 of 22 Superseded Latest is v22 →

lore
Share & tools
lore

William Tu

author
#0 lore raw

RFC PATCH v11 1/1 vmxnet3: Add XDP support.

The patch adds native-mode XDP support: XDP DROP, PASS, TX, and REDIRECT.

Background:
The vmxnet3 rx consists of three rings: ring0, ring1, and dataring. For r0 and r1, buffers at r0 are allocated using alloc_skb APIs and dma mapped to the ring's descriptor. If LRO is enabled and packet size larger than 3K, VMXNET3_MAX_SKB_BUF_SIZE, then r1 is used to mapped the rest of the buffer larger than VMXNET3_MAX_SKB_BUF_SIZE. Each buffer in r1 is allocated using alloc_page. So for LRO packets, the payload will be in one buffer from r0 and multiple from r1, for non-LRO packets, only one descriptor in r0 is used for packet size less than 3k.

When receiving a packet, the first descriptor will have the sop (start of packet) bit set, and the last descriptor will have the eop (end of packet) bit set. Non-LRO packets will have only one descriptor with both sop and eop set.

Other than r0 and r1, vmxnet3 dataring is specifically designed for handling packets with small size, usually 128 bytes, defined in VMXNET3_DEF_RXDATA_DESC_SIZE, by simply copying the packet from the backend driver in ESXi to the ring's memory region at front-end vmxnet3 driver, in order to avoid memory mapping/unmapping overhead. In summary, packet size:

    A. < 128B: use dataring
    B. 128B - 3K: use ring0 (VMXNET3_RX_BUF_SKB)
    C. > 3K: use ring0 and ring1 (VMXNET3_RX_BUF_SKB + VMXNET3_RX_BUF_PAGE)

As a result, the patch adds XDP support for packets using dataring and r0 (case A and B), not the large packet size when LRO is enabled.

XDP Implementation:
When user loads and XDP prog, vmxnet3 driver checks configurations, such as mtu, lro, and re-allocate the rx buffer size for reserving the extra headroom, XDP_PACKET_HEADROOM, for XDP frame. The XDP prog will then be associated with every rx queue of the device. Note that when using dataring for small packet size, vmxnet3 (front-end driver) doesn't control the buffer allocation, as a result we allocate a new page and copy packet from the dataring to XDP frame.

The receive side of XDP is implemented for case A and B, by invoking the bpf program at vmxnet3_rq_rx_complete and handle its returned action. The vmxnet3_process_xdp(), vmxnet3_process_xdp_small() function handles the ring0 and dataring case separately, and decides the next journey of the packet afterward.

For TX, vmxnet3 has split header design. Outgoing packets are parsed first and protocol headers (L2/L3/L4) are copied to the backend. The rest of the payload are dma mapped. Since XDP_TX does not parse the packet protocol, the entire XDP frame is dma mapped for transmission and transmitted in a batch. Later on, the frame is freed and recycled back to the memory pool.

Performance:
Tested using two VMs inside one ESXi vSphere 7.0 machine, using single core on each vmxnet3 device, sender using DPDK testpmd tx-mode attached to vmxnet3 device, sending 64B or 512B UDP packet.

VM1 txgen:

$ dpdk-testpmd -l 0-3 -n 1 -- -i --nb-cores=3 \
--forward-mode=txonly --eth-peer=0,<mac addr of vm2>
option: add "--txonly-multi-flow"
option: use --txpkts=512 or 64 byte

VM2 running XDP:

$ ./samples/bpf/xdp_rxq_info -d ens160 -a <options> --skb-mode
$ ./samples/bpf/xdp_rxq_info -d ens160 -a <options>
options: XDP_DROP, XDP_PASS, XDP_TX

To test REDIRECT to cpu 0, use

$ ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e drop

Single core performance comparison with skb-mode.
64B: skb-mode -> native-mode
XDP_DROP: 1.6Mpps -> 2.4Mpps
XDP_PASS: 338Kpps -> 367Kpps
XDP_TX: 1.1Mpps -> 2.3Mpps
REDIRECT-drop: 1.3Mpps -> 2.3Mpps

512B: skb-mode -> native-mode
XDP_DROP: 863Kpps -> 1.3Mpps
XDP_PASS: 275Kpps -> 376Kpps
XDP_TX: 554Kpps -> 1.2Mpps
REDIRECT-drop: 659Kpps -> 1.2Mpps

Limitations:
a. LRO will be disabled when users load XDP program b. MTU will be checked and limit to
VMXNET3_MAX_SKB_BUF_SIZE(3K) - XDP_PACKET_HEADROOM(256) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))

---
v10 -> v11:
work on feedbacks from Alexander Duyck
internal feedback from Guolin and Ronak
- fix the issue of xdp_return_frame_bulk, move to up level of vmxnet3_unmap_tx_buf and some refactoring
- refactor and simplify vmxnet3_tq_cleanup
- disable XDP when LRO is enabled, suggested by Ronak
- diff of v10..v11
https://github.com/williamtu/net-next/compare/e46f2cf5c18f..162e8849903b

v9 -> v10:
- Mark as RFC as we're waiting for internal review Feedback from Alexander Duyck
- fix dma mapping leak of ndo_xdp_xmit case
- remove unused MAP_INVALID and adjist bitfield

v8 -> v9:
new
- add rxdataring support (need extra copy but still fast)
- update performance number (much better than v8!)
https://youtu.be/4lm1CSCi78Q

- work on feedbacks from Alexander Duyck and Alexander Lobakin Alexander Lobakin
- use xdp_return_frame_bulk and see some performance improvement
- use xdp_do_flush not xdp_do_flush_map
- fix several alignment issues, formatting issues, minor code restructure, remove 2 dead functions, unrelated add/del of new lines, add braces when logical ops nearby, endianness conversion
- remove several oneliner goto label
- anonymous union of skb and xdpf
- remove xdp_enabled and use xdp prog directly to check
- use bitsfields macro --> I decide to do it later as there are many unrelated places needed to change.

Alexander Duyck
- use bitfield for tbi map type
- anonymous union of skb and xdpf
- remove use of modulus ops, cpu % tq_number

others
- fix issue reported from kernel test robot using sparse

v7 -> v8:
- work on feedbacks from Gerhard Engleder and Alexander
- change memory model to use page pool API, rewrite some of the XDP processing code
- attach bpf xdp prog to adapter, not per rx queue
- I reference some of the XDP implementation from drivers/net/ethernet/mediatek and drivers/net/ethernet/stmicro/stmmac/
- drop support for rxdataring for this version
- redo performance evaluation and demo here
https://youtu.be/T7_0yrCXCe0
- check using /sys/kernel/debug/kmemleak

I tested the patch using below script:
while [ true ]; do
timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_DROP --skb-mode timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_DROP timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_PASS --skb-mode timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_PASS timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_TX --skb-mode timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_TX timeout 20 ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e drop timeout 20 ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e pass done

---
 drivers/net/Kconfig                   |   1 +
 drivers/net/vmxnet3/Makefile          |   2 +-
 drivers/net/vmxnet3/vmxnet3_drv.c     | 239 ++++++++++++---
 drivers/net/vmxnet3/vmxnet3_ethtool.c |  14 +
 drivers/net/vmxnet3/vmxnet3_int.h     |  44 ++-
 drivers/net/vmxnet3/vmxnet3_xdp.c     | 425 ++++++++++++++++++++++++++
 drivers/net/vmxnet3/vmxnet3_xdp.h     |  41 +++
 7 files changed, 717 insertions(+), 49 deletions(-)
 create mode 100644 drivers/net/vmxnet3/vmxnet3_xdp.c
 create mode 100644 drivers/net/vmxnet3/vmxnet3_xdp.h
drivers/net/Kconfig +1 -0
index 9e63b8c43f3e..a4419d661bdd 100644
@@ -571,6 +571,7 @@ config VMXNET3
 	tristate "VMware VMXNET3 ethernet driver"
 	depends on PCI && INET
 	depends on PAGE_SIZE_LESS_THAN_64KB
+	select PAGE_POOL
 	help
 	  This driver supports VMware's vmxnet3 virtual ethernet NIC.
 	  To compile this driver as a module, choose M here: the
drivers/net/vmxnet3/Makefile +1 -1
index a666a88ac1ff..f82870c10205 100644
@@ -32,4 +32,4 @@
 
 obj-$(CONFIG_VMXNET3) += vmxnet3.o
 
-vmxnet3-objs := vmxnet3_drv.o vmxnet3_ethtool.o
+vmxnet3-objs := vmxnet3_drv.o vmxnet3_ethtool.o vmxnet3_xdp.o
drivers/net/vmxnet3/vmxnet3_drv.c +201 -38

442 lines not on this page · raw message

drivers/net/vmxnet3/vmxnet3_ethtool.c +14 -0
index 18cf7c723201..6f542236b26e 100644
@@ -76,6 +76,10 @@ vmxnet3_tq_driver_stats[] = {
 					 copy_skb_header) },
 	{ "  giant hdr",	offsetof(struct vmxnet3_tq_driver_stats,
 					 oversized_hdr) },
+	{ "  xdp xmit",		offsetof(struct vmxnet3_tq_driver_stats,
+					 xdp_xmit) },
+	{ "  xdp xmit err",	offsetof(struct vmxnet3_tq_driver_stats,
+					 xdp_xmit_err) },
 };
 
 /* per rq stats maintained by the device */
@@ -106,6 +110,16 @@ vmxnet3_rq_driver_stats[] = {
 					 drop_fcs) },
 	{ "  rx buf alloc fail", offsetof(struct vmxnet3_rq_driver_stats,
 					  rx_buf_alloc_failure) },
+	{ "     xdp packets", offsetof(struct vmxnet3_rq_driver_stats,
+				       xdp_packets) },
+	{ "     xdp tx", offsetof(struct vmxnet3_rq_driver_stats,
+				  xdp_tx) },
+	{ "     xdp redirects", offsetof(struct vmxnet3_rq_driver_stats,
+					 xdp_redirects) },
+	{ "     xdp drops", offsetof(struct vmxnet3_rq_driver_stats,
+				     xdp_drops) },
+	{ "     xdp aborted", offsetof(struct vmxnet3_rq_driver_stats,
+				       xdp_aborted) },
 };
 
 /* global stats maintained by the driver */
drivers/net/vmxnet3/vmxnet3_int.h +34 -10

108 lines not on this page · raw message

drivers/net/vmxnet3/vmxnet3_xdp.c new file +425 -0

425 lines not on this page · raw message

drivers/net/vmxnet3/vmxnet3_xdp.h new file +41 -0

41 lines not on this page · raw message

Signature · 1 line
2.25.1
#1 lore raw
Quoted 205 lines · William Tu

The patch adds native-mode XDP support: XDP DROP, PASS, TX, and REDIRECT.

Background:
The vmxnet3 rx consists of three rings: ring0, ring1, and dataring. For r0 and r1, buffers at r0 are allocated using alloc_skb APIs and dma mapped to the ring's descriptor. If LRO is enabled and packet size larger than 3K, VMXNET3_MAX_SKB_BUF_SIZE, then r1 is used to mapped the rest of the buffer larger than VMXNET3_MAX_SKB_BUF_SIZE. Each buffer in r1 is allocated using alloc_page. So for LRO packets, the payload will be in one buffer from r0 and multiple from r1, for non-LRO packets, only one descriptor in r0 is used for packet size less than 3k.

When receiving a packet, the first descriptor will have the sop (start of packet) bit set, and the last descriptor will have the eop (end of packet) bit set. Non-LRO packets will have only one descriptor with both sop and eop set.

Other than r0 and r1, vmxnet3 dataring is specifically designed for handling packets with small size, usually 128 bytes, defined in VMXNET3_DEF_RXDATA_DESC_SIZE, by simply copying the packet from the backend driver in ESXi to the ring's memory region at front-end vmxnet3 driver, in order to avoid memory mapping/unmapping overhead. In summary, packet size:

    A. < 128B: use dataring
    B. 128B - 3K: use ring0 (VMXNET3_RX_BUF_SKB)
    C. > 3K: use ring0 and ring1 (VMXNET3_RX_BUF_SKB + VMXNET3_RX_BUF_PAGE)

As a result, the patch adds XDP support for packets using dataring and r0 (case A and B), not the large packet size when LRO is enabled.

XDP Implementation:
When user loads and XDP prog, vmxnet3 driver checks configurations, such as mtu, lro, and re-allocate the rx buffer size for reserving the extra headroom, XDP_PACKET_HEADROOM, for XDP frame. The XDP prog will then be associated with every rx queue of the device. Note that when using dataring for small packet size, vmxnet3 (front-end driver) doesn't control the buffer allocation, as a result we allocate a new page and copy packet from the dataring to XDP frame.

The receive side of XDP is implemented for case A and B, by invoking the bpf program at vmxnet3_rq_rx_complete and handle its returned action. The vmxnet3_process_xdp(), vmxnet3_process_xdp_small() function handles the ring0 and dataring case separately, and decides the next journey of the packet afterward.

For TX, vmxnet3 has split header design. Outgoing packets are parsed first and protocol headers (L2/L3/L4) are copied to the backend. The rest of the payload are dma mapped. Since XDP_TX does not parse the packet protocol, the entire XDP frame is dma mapped for transmission and transmitted in a batch. Later on, the frame is freed and recycled back to the memory pool.

Performance:
Tested using two VMs inside one ESXi vSphere 7.0 machine, using single core on each vmxnet3 device, sender using DPDK testpmd tx-mode attached to vmxnet3 device, sending 64B or 512B UDP packet.

VM1 txgen:

$ dpdk-testpmd -l 0-3 -n 1 -- -i --nb-cores=3 \
--forward-mode=txonly --eth-peer=0,<mac addr of vm2>
option: add "--txonly-multi-flow"
option: use --txpkts=512 or 64 byte

VM2 running XDP:

$ ./samples/bpf/xdp_rxq_info -d ens160 -a <options> --skb-mode
$ ./samples/bpf/xdp_rxq_info -d ens160 -a <options>
options: XDP_DROP, XDP_PASS, XDP_TX

To test REDIRECT to cpu 0, use

$ ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e drop

Single core performance comparison with skb-mode.
64B: skb-mode -> native-mode
XDP_DROP: 1.6Mpps -> 2.4Mpps
XDP_PASS: 338Kpps -> 367Kpps
XDP_TX: 1.1Mpps -> 2.3Mpps
REDIRECT-drop: 1.3Mpps -> 2.3Mpps

512B: skb-mode -> native-mode
XDP_DROP: 863Kpps -> 1.3Mpps
XDP_PASS: 275Kpps -> 376Kpps
XDP_TX: 554Kpps -> 1.2Mpps
REDIRECT-drop: 659Kpps -> 1.2Mpps

Limitations:
a. LRO will be disabled when users load XDP program b. MTU will be checked and limit to
VMXNET3_MAX_SKB_BUF_SIZE(3K) - XDP_PACKET_HEADROOM(256) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))

---
v10 -> v11:
work on feedbacks from Alexander Duyck
internal feedback from Guolin and Ronak
- fix the issue of xdp_return_frame_bulk, move to up level of vmxnet3_unmap_tx_buf and some refactoring
- refactor and simplify vmxnet3_tq_cleanup
- disable XDP when LRO is enabled, suggested by Ronak
- diff of v10..v11
https://github.com/williamtu/net-next/compare/e46f2cf5c18f..162e8849903b

v9 -> v10:
- Mark as RFC as we're waiting for internal review Feedback from Alexander Duyck
- fix dma mapping leak of ndo_xdp_xmit case
- remove unused MAP_INVALID and adjist bitfield

v8 -> v9:
new
- add rxdataring support (need extra copy but still fast)
- update performance number (much better than v8!)
https://youtu.be/4lm1CSCi78Q

- work on feedbacks from Alexander Duyck and Alexander Lobakin Alexander Lobakin
- use xdp_return_frame_bulk and see some performance improvement
- use xdp_do_flush not xdp_do_flush_map
- fix several alignment issues, formatting issues, minor code restructure, remove 2 dead functions, unrelated add/del of new lines, add braces when logical ops nearby, endianness conversion
- remove several oneliner goto label
- anonymous union of skb and xdpf
- remove xdp_enabled and use xdp prog directly to check
- use bitsfields macro --> I decide to do it later as there are many unrelated places needed to change.

Alexander Duyck
- use bitfield for tbi map type
- anonymous union of skb and xdpf
- remove use of modulus ops, cpu % tq_number

others
- fix issue reported from kernel test robot using sparse

v7 -> v8:
- work on feedbacks from Gerhard Engleder and Alexander
- change memory model to use page pool API, rewrite some of the XDP processing code
- attach bpf xdp prog to adapter, not per rx queue
- I reference some of the XDP implementation from drivers/net/ethernet/mediatek and drivers/net/ethernet/stmicro/stmmac/
- drop support for rxdataring for this version
- redo performance evaluation and demo here
https://youtu.be/T7_0yrCXCe0
- check using /sys/kernel/debug/kmemleak

I tested the patch using below script:
while [ true ]; do
timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_DROP --skb-mode timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_DROP timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_PASS --skb-mode timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_PASS timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_TX --skb-mode timeout 20 ./samples/bpf/xdp_rxq_info -d ens160 -a XDP_TX timeout 20 ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e drop timeout 20 ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e pass done

---
 drivers/net/Kconfig                   |   1 +
 drivers/net/vmxnet3/Makefile          |   2 +-
 drivers/net/vmxnet3/vmxnet3_drv.c     | 239 ++++++++++++---
 drivers/net/vmxnet3/vmxnet3_ethtool.c |  14 +
 drivers/net/vmxnet3/vmxnet3_int.h     |  44 ++-
 drivers/net/vmxnet3/vmxnet3_xdp.c     | 425 ++++++++++++++++++++++++++
 drivers/net/vmxnet3/vmxnet3_xdp.h     |  41 +++
 7 files changed, 717 insertions(+), 49 deletions(-)
 create mode 100644 drivers/net/vmxnet3/vmxnet3_xdp.c
 create mode 100644 drivers/net/vmxnet3/vmxnet3_xdp.h
drivers/net/Kconfig +1 -0
index 9e63b8c43f3e..a4419d661bdd 100644
@@ -571,6 +571,7 @@ config VMXNET3
 	tristate "VMware VMXNET3 ethernet driver"
 	depends on PCI && INET
 	depends on PAGE_SIZE_LESS_THAN_64KB
+	select PAGE_POOL
 	help
 	  This driver supports VMware's vmxnet3 virtual ethernet NIC.
 	  To compile this driver as a module, choose M here: the
drivers/net/vmxnet3/Makefile +1 -1
index a666a88ac1ff..f82870c10205 100644
@@ -32,4 +32,4 @@
 
 obj-$(CONFIG_VMXNET3) += vmxnet3.o
 
-vmxnet3-objs := vmxnet3_drv.o vmxnet3_ethtool.o
+vmxnet3-objs := vmxnet3_drv.o vmxnet3_ethtool.o vmxnet3_xdp.o
drivers/net/vmxnet3/vmxnet3_drv.c +1 -0
index d3e7b27eb933..6e40e40c9329 100644
@@ -28,6 +28,7 @@
 #include <net/ip6_checksum.h>
 
 #include "vmxnet3_int.h"
+#include "vmxnet3_xdp.h"
 
 char vmxnet3_driver_name[] = "vmxnet3";
 #define VMXNET3_DRIVER_DESC "VMware vmxnet3 virtual NIC driver"

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 328

Quoted 9 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -3
@@ -322,44 +323,58 @@ static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size)
 #endif /* __BIG_ENDIAN_BITFIELD  */
 
 
-static void
-vmxnet3_unmap_tx_buf(struct vmxnet3_tx_buf_info *tbi,
-		     struct pci_dev *pdev)
+static u32
+vmxnet3_unmap_tx_buf(struct vmxnet3_tx_buf_info *tbi, struct pci_dev *pdev,
+		     struct xdp_frame_bulk *bq)

Is the bq value being used anywhere in here? We probably need to either drop it.

Quoted 6 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -1
 {
-	if (tbi->map_type == VMXNET3_MAP_SINGLE)
+	u32 map_type = tbi->map_type;
+
+	if (map_type & VMXNET3_MAP_SINGLE)
 		dma_unmap_single(&pdev->dev, tbi->dma_addr, tbi->len,

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 340

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +4 -3
 				 DMA_TO_DEVICE);
-	else if (tbi->map_type == VMXNET3_MAP_PAGE)
+	else if (map_type & VMXNET3_MAP_PAGE)
 		dma_unmap_page(&pdev->dev, tbi->dma_addr, tbi->len,
 			       DMA_TO_DEVICE);
-	else
-		BUG_ON(tbi->map_type != VMXNET3_MAP_NONE);
+	else if (map_type & ~(VMXNET3_MAP_SINGLE | VMXNET3_MAP_PAGE |
+			      VMXNET3_MAP_XDP))
+		BUG_ON(map_type != VMXNET3_MAP_NONE);

What you might just do here is drop the "else if" in favor of an "else". Your bug on would basically just need to check for:

	else
		BUG_ON((map_type & ~VMXNET_MAP_XDP));

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 347

Quoted 6 lines
drivers/net/vmxnet3/vmxnet3_drv.c +2 -0
 	tbi->map_type = VMXNET3_MAP_NONE; /* to help debugging */
+
+	return map_type;
 }
 
 

In reality we only need to worry about the map_type of the eop buffer when we are dealing with any frame. You might actually pull the map_type out in vmxnet3_unmap_pkt below and just pass that to the end.

Quoted 12 lines
drivers/net/vmxnet3/vmxnet3_drv.c +4 -2
 static int
 vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq,
-		  struct pci_dev *pdev,	struct vmxnet3_adapter *adapter)
+		  struct pci_dev *pdev,	struct vmxnet3_adapter *adapter,
+		  struct xdp_frame_bulk *bq)
 {
-	struct sk_buff *skb;
+	struct vmxnet3_tx_buf_info *tbi;
 	int entries = 0;
+	u32 map_type;
 
 	/* no out of order completion */

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 363

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +1 -4
 	BUG_ON(tq->buf_info[eop_idx].sop_idx != tq->tx_ring.next2comp);
 	BUG_ON(VMXNET3_TXDESC_GET_EOP(&(tq->tx_ring.base[eop_idx].txd)) != 1);
 
-	skb = tq->buf_info[eop_idx].skb;
-	BUG_ON(skb == NULL);
-	tq->buf_info[eop_idx].skb = NULL;
-
+	tbi = &tq->buf_info[eop_idx];
 	VMXNET3_INC_RING_IDX_ONLY(eop_idx, tq->tx_ring.size);
 

We may want to record the map_type here for the tbi that contains the skb or XDP frame. Then we don't need to record the map_type in the loop below and can save ourselves a few cycles.

Also we may want to preserve the BUG_ON check, but here it would be:

	BUG_ON(tbi->skb == NULL);
Quoted 7 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -3
 	while (tq->tx_ring.next2comp != eop_idx) {
-		vmxnet3_unmap_tx_buf(tq->buf_info + tq->tx_ring.next2comp,
-				     pdev);
-
+		map_type = vmxnet3_unmap_tx_buf(tq->buf_info +
+						tq->tx_ring.next2comp, pdev,
+						bq);

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 377

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +10 -0
+		/* xdpf and skb are in an anonymous union, if set we need to
+		 * free a buffer.
+		 */
+		if (tbi->skb) {
+			if (map_type & VMXNET3_MAP_XDP)
+				xdp_return_frame_bulk(tbi->xdpf, bq);
+			else
+				dev_kfree_skb_any(tbi->skb);
+			tbi->skb = NULL;
+		}

So this code could probably be moved to the end of the function where the dev_kfree_skb_any originally was. Then you can drop the "if(tbi-

Quoted 1 line

skb)" and instead just have the map_type dealt with there. This should

save you from having to read and store map_type multiple times.

drivers/net/vmxnet3/vmxnet3_drv.c vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq, · line 389

Quoted 11 lines
drivers/net/vmxnet3/vmxnet3_drv.c +0 -1
 		/* update next2comp w/o tx_lock. Since we are marking more,
 		 * instead of less, tx ring entries avail, the worst case is
 		 * that the tx routine incorrectly re-queues a pkt due to
@@ -369,7 +384,6 @@ vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq,
 		entries++;
 	}
 
-	dev_kfree_skb_any(skb);
 	return entries;
 }
 

No point moving this into the loop when it can always be processed at the end. Basically we just needc to pull the if statement you added above down to here.

Quoted 30 lines
drivers/net/vmxnet3/vmxnet3_drv.c +5 -1
@@ -379,8 +393,10 @@ vmxnet3_tq_tx_complete(struct vmxnet3_tx_queue *tq,
 			struct vmxnet3_adapter *adapter)
 {
 	int completed = 0;
+	struct xdp_frame_bulk bq;
 	union Vmxnet3_GenericDesc *gdesc;
 
+	xdp_frame_bulk_init(&bq);
 	gdesc = tq->comp_ring.base + tq->comp_ring.next2proc;
 	while (VMXNET3_TCD_GET_GEN(&gdesc->tcd) == tq->comp_ring.gen) {
 		/* Prevent any &gdesc->tcd field from being (speculatively)
@@ -390,11 +406,12 @@ vmxnet3_tq_tx_complete(struct vmxnet3_tx_queue *tq,
 
 		completed += vmxnet3_unmap_pkt(VMXNET3_TCD_GET_TXIDX(
 					       &gdesc->tcd), tq, adapter->pdev,
-					       adapter);
+					       adapter, &bq);
 
 		vmxnet3_comp_ring_adv_next2proc(&tq->comp_ring);
 		gdesc = tq->comp_ring.base + tq->comp_ring.next2proc;
 	}
+	xdp_flush_frame_bulk(&bq);
 
 	if (completed) {
 		spin_lock(&tq->tx_lock);
@@ -414,26 +431,33 @@ static void
 vmxnet3_tq_cleanup(struct vmxnet3_tx_queue *tq,
 		   struct vmxnet3_adapter *adapter)
 {
+	struct xdp_frame_bulk bq;

drivers/net/vmxnet3/vmxnet3_drv.c static void · line 444

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -0
+	u32 map_type;
 	int i;
 
+	xdp_frame_bulk_init(&bq);
+
 	while (tq->tx_ring.next2comp != tq->tx_ring.next2fill) {
 		struct vmxnet3_tx_buf_info *tbi;
 
 		tbi = tq->buf_info + tq->tx_ring.next2comp;
 

Rather than have vmxnet3_unmap_tx_buf return the map_type you might just read it yourself here.

Quoted 26 lines
drivers/net/vmxnet3/vmxnet3_drv.c +10 -7
-		vmxnet3_unmap_tx_buf(tbi, adapter->pdev);
+		map_type = vmxnet3_unmap_tx_buf(tbi, adapter->pdev, &bq);
 		if (tbi->skb) {
-			dev_kfree_skb_any(tbi->skb);
+			if (map_type & VMXNET3_MAP_XDP)
+				xdp_return_frame_bulk(tbi->xdpf, &bq);
+			else
+				dev_kfree_skb_any(tbi->skb);
 			tbi->skb = NULL;
 		}
 		vmxnet3_cmd_ring_adv_next2comp(&tq->tx_ring);
 	}
 
-	/* sanity check, verify all buffers are indeed unmapped and freed */
-	for (i = 0; i < tq->tx_ring.size; i++) {
-		BUG_ON(tq->buf_info[i].skb != NULL ||
-		       tq->buf_info[i].map_type != VMXNET3_MAP_NONE);
-	}
+	xdp_flush_frame_bulk(&bq);
+
+	/* sanity check, verify all buffers are indeed unmapped */
+	for (i = 0; i < tq->tx_ring.size; i++)
+		BUG_ON(tq->buf_info[i].map_type != VMXNET3_MAP_NONE);
 
 	tq->tx_ring.gen = VMXNET3_INIT_GEN;
 	tq->tx_ring.next2fill = tq->tx_ring.next2comp = 0;
author
#2 lore raw

Hi Alexander,

Thanks for your review!

Quoted 4 lines · Alexander H Duyck
Quoted 2 lines · William Tu

The patch adds native-mode XDP support: XDP DROP, PASS, TX, and REDIRECT.

<...>

Quoted 6 lines
Quoted 6 lines

char vmxnet3_driver_name[] = "vmxnet3";
#define VMXNET3_DRIVER_DESC "VMware vmxnet3 virtual NIC driver"

drivers/net/vmxnet3/vmxnet3_drv.c +0 -0
@@ -322,44 +323,58 @@ static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size)
 #endif /* __BIG_ENDIAN_BITFIELD  */
 
 
Quoted 9 lines

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 328

Quoted 6 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -3
-static void
-vmxnet3_unmap_tx_buf(struct vmxnet3_tx_buf_info *tbi,
-                  struct pci_dev *pdev)
+static u32
+vmxnet3_unmap_tx_buf(struct vmxnet3_tx_buf_info *tbi, struct pci_dev *pdev,
+                  struct xdp_frame_bulk *bq)

Is the bq value being used anywhere in here? We probably need to either drop it.

yes, my mistake.

Quoted 18 lines
Quoted 7 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -1
 {
-     if (tbi->map_type == VMXNET3_MAP_SINGLE)
+     u32 map_type = tbi->map_type;
+
+     if (map_type & VMXNET3_MAP_SINGLE)
              dma_unmap_single(&pdev->dev, tbi->dma_addr, tbi->len,

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 340

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +4 -3
                               DMA_TO_DEVICE);
-     else if (tbi->map_type == VMXNET3_MAP_PAGE)
+     else if (map_type & VMXNET3_MAP_PAGE)
              dma_unmap_page(&pdev->dev, tbi->dma_addr, tbi->len,
                             DMA_TO_DEVICE);
-     else
-             BUG_ON(tbi->map_type != VMXNET3_MAP_NONE);
+     else if (map_type & ~(VMXNET3_MAP_SINGLE | VMXNET3_MAP_PAGE |
+                           VMXNET3_MAP_XDP))
+             BUG_ON(map_type != VMXNET3_MAP_NONE);
Quoted 4 lines

What you might just do here is drop the "else if" in favor of an "else". Your bug on would basically just need to check for:

        else
                BUG_ON((map_type & ~VMXNET_MAP_XDP));

Will do it, thanks!

Quoted 35 lines

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 347

Quoted 7 lines
drivers/net/vmxnet3/vmxnet3_drv.c +2 -0
      tbi->map_type = VMXNET3_MAP_NONE; /* to help debugging */
+
+     return map_type;
 }
 
 

In reality we only need to worry about the map_type of the eop buffer when we are dealing with any frame. You might actually pull the map_type out in vmxnet3_unmap_pkt below and just pass that to the end.

Quoted 12 lines
drivers/net/vmxnet3/vmxnet3_drv.c +4 -2
 static int
 vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq,
-               struct pci_dev *pdev, struct vmxnet3_adapter *adapter)
+               struct pci_dev *pdev, struct vmxnet3_adapter *adapter,
+               struct xdp_frame_bulk *bq)
 {
-     struct sk_buff *skb;
+     struct vmxnet3_tx_buf_info *tbi;
      int entries = 0;
+     u32 map_type;
 
      /* no out of order completion */

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 363

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +1 -4
      BUG_ON(tq->buf_info[eop_idx].sop_idx != tq->tx_ring.next2comp);
      BUG_ON(VMXNET3_TXDESC_GET_EOP(&(tq->tx_ring.base[eop_idx].txd)) != 1);
 
-     skb = tq->buf_info[eop_idx].skb;
-     BUG_ON(skb == NULL);
-     tq->buf_info[eop_idx].skb = NULL;
-
+     tbi = &tq->buf_info[eop_idx];
      VMXNET3_INC_RING_IDX_ONLY(eop_idx, tq->tx_ring.size);
 
Quoted 6 lines

We may want to record the map_type here for the tbi that contains the skb or XDP frame. Then we don't need to record the map_type in the loop below and can save ourselves a few cycles.

Also we may want to preserve the BUG_ON check, but here it would be:

        BUG_ON(tbi->skb == NULL);

Got it, thanks!

Quoted 31 lines
Quoted 8 lines
drivers/net/vmxnet3/vmxnet3_drv.c +3 -3
      while (tq->tx_ring.next2comp != eop_idx) {
-             vmxnet3_unmap_tx_buf(tq->buf_info + tq->tx_ring.next2comp,
-                                  pdev);
-
+             map_type = vmxnet3_unmap_tx_buf(tq->buf_info +
+                                             tq->tx_ring.next2comp, pdev,
+                                             bq);

drivers/net/vmxnet3/vmxnet3_drv.c static u32 get_bitfield32(const __le32 *bitfield, u32 pos, u32 size) · line 377

Quoted 10 lines
drivers/net/vmxnet3/vmxnet3_drv.c +10 -0
+             /* xdpf and skb are in an anonymous union, if set we need to
+              * free a buffer.
+              */
+             if (tbi->skb) {
+                     if (map_type & VMXNET3_MAP_XDP)
+                             xdp_return_frame_bulk(tbi->xdpf, bq);
+                     else
+                             dev_kfree_skb_any(tbi->skb);
+                     tbi->skb = NULL;
+             }

So this code could probably be moved to the end of the function where the dev_kfree_skb_any originally was. Then you can drop the "if(tbi-

Quoted 1 line

skb)" and instead just have the map_type dealt with there. This should

save you from having to read and store map_type multiple times.

Quoted 7 lines

/* update next2comp w/o tx_lock. Since we are marking more,
* instead of less, tx ring entries avail, the worst case is
* that the tx routine incorrectly re-queues a pkt due to

drivers/net/vmxnet3/vmxnet3_drv.c +0 -0
@@ -369,7 +384,6 @@ vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq,
              entries++;
      }
 
Quoted 8 lines

drivers/net/vmxnet3/vmxnet3_drv.c vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq, · line 389

Quoted 4 lines
drivers/net/vmxnet3/vmxnet3_drv.c +0 -1
-     dev_kfree_skb_any(skb);
      return entries;
 }
 

No point moving this into the loop when it can always be processed at the end. Basically we just needc to pull the if statement you added above down to here.

right, will move out of the loop.

Quoted 39 lines
Quoted 39 lines
drivers/net/vmxnet3/vmxnet3_drv.c +8 -1
@@ -379,8 +393,10 @@ vmxnet3_tq_tx_complete(struct vmxnet3_tx_queue *tq,
                      struct vmxnet3_adapter *adapter)
 {
      int completed = 0;
+     struct xdp_frame_bulk bq;
      union Vmxnet3_GenericDesc *gdesc;
 
+     xdp_frame_bulk_init(&bq);
      gdesc = tq->comp_ring.base + tq->comp_ring.next2proc;
      while (VMXNET3_TCD_GET_GEN(&gdesc->tcd) == tq->comp_ring.gen) {
              /* Prevent any &gdesc->tcd field from being (speculatively)
@@ -390,11 +406,12 @@ vmxnet3_tq_tx_complete(struct vmxnet3_tx_queue *tq,
 
              completed += vmxnet3_unmap_pkt(VMXNET3_TCD_GET_TXIDX(
                                             &gdesc->tcd), tq, adapter->pdev,
-                                            adapter);
+                                            adapter, &bq);
 
              vmxnet3_comp_ring_adv_next2proc(&tq->comp_ring);
              gdesc = tq->comp_ring.base + tq->comp_ring.next2proc;
      }
+     xdp_flush_frame_bulk(&bq);
 
      if (completed) {
              spin_lock(&tq->tx_lock);
@@ -414,26 +431,33 @@ static void
 vmxnet3_tq_cleanup(struct vmxnet3_tx_queue *tq,
                 struct vmxnet3_adapter *adapter)
 {
+     struct xdp_frame_bulk bq;
+     u32 map_type;
      int i;
 
+     xdp_frame_bulk_init(&bq);
+
      while (tq->tx_ring.next2comp != tq->tx_ring.next2fill) {
              struct vmxnet3_tx_buf_info *tbi;
 
Quoted 5 lines

drivers/net/vmxnet3/vmxnet3_drv.c static void · line 444

Quoted 2 lines
drivers/net/vmxnet3/vmxnet3_drv.c +0 -0
              tbi = tq->buf_info + tq->tx_ring.next2comp;
 

Rather than have vmxnet3_unmap_tx_buf return the map_type you might just read it yourself here.

Will do it, thanks
William