Skip to content

Commit 6d6b39f

Browse files
Shenwei WangPaolo Abeni
Shenwei Wang
authored and
Paolo Abeni
committed
net: fec: add initial XDP support
This patch adds the initial XDP support to Freescale driver. It supports XDP_PASS, XDP_DROP and XDP_REDIRECT actions. Upcoming patches will add support for XDP_TX and Zero Copy features. As the patch is rather large, the part of codes to collect the statistics is separated and will prepare a dedicated patch for that part. I just tested with the application of xdpsock. -- Native here means running command of "xdpsock -i eth0" -- SKB-Mode means running command of "xdpsock -S -i eth0" The following are the testing result relating to XDP mode: root@imx8qxpc0mek:~/bpf# ./xdpsock -i eth0 sock0@eth0:0 rxdrop xdp-drv pps pkts 1.00 rx 371347 2717794 tx 0 0 root@imx8qxpc0mek:~/bpf# ./xdpsock -S -i eth0 sock0@eth0:0 rxdrop xdp-skb pps pkts 1.00 rx 202229 404528 tx 0 0 root@imx8qxpc0mek:~/bpf# ./xdp2 eth0 proto 0: 496708 pkt/s proto 0: 505469 pkt/s proto 0: 505283 pkt/s proto 0: 505443 pkt/s proto 0: 505465 pkt/s root@imx8qxpc0mek:~/bpf# ./xdp2 -S eth0 proto 0: 0 pkt/s proto 17: 118778 pkt/s proto 17: 118989 pkt/s proto 0: 1 pkt/s proto 17: 118987 pkt/s proto 0: 0 pkt/s proto 17: 118943 pkt/s proto 17: 118976 pkt/s proto 0: 1 pkt/s proto 17: 119006 pkt/s proto 0: 0 pkt/s proto 17: 119071 pkt/s proto 17: 119092 pkt/s Signed-off-by: Shenwei Wang <[email protected]> Reported-by: kernel test robot <[email protected]> Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Paolo Abeni <[email protected]>
1 parent 598d298 commit 6d6b39f

File tree

2 files changed

+226
-2
lines changed

2 files changed

+226
-2
lines changed

drivers/net/ethernet/freescale/fec.h

+3-1
Original file line numberDiff line numberDiff line change
@@ -348,7 +348,6 @@ struct bufdesc_ex {
348348
*/
349349

350350
#define FEC_ENET_XDP_HEADROOM (XDP_PACKET_HEADROOM)
351-
352351
#define FEC_ENET_RX_PAGES 256
353352
#define FEC_ENET_RX_FRSIZE (PAGE_SIZE - FEC_ENET_XDP_HEADROOM \
354353
- SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
@@ -663,6 +662,9 @@ struct fec_enet_private {
663662

664663
struct imx_sc_ipc *ipc_handle;
665664

665+
/* XDP BPF Program */
666+
struct bpf_prog *xdp_prog;
667+
666668
u64 ethtool_stats[];
667669
};
668670

drivers/net/ethernet/freescale/fec_main.c

+223-1
Original file line numberDiff line numberDiff line change
@@ -89,6 +89,11 @@ static const u16 fec_enet_vlan_pri_to_queue[8] = {0, 0, 1, 1, 1, 2, 2, 2};
8989
#define FEC_ENET_OPD_V 0xFFF0
9090
#define FEC_MDIO_PM_TIMEOUT 100 /* ms */
9191

92+
#define FEC_ENET_XDP_PASS 0
93+
#define FEC_ENET_XDP_CONSUMED BIT(0)
94+
#define FEC_ENET_XDP_TX BIT(1)
95+
#define FEC_ENET_XDP_REDIR BIT(2)
96+
9297
struct fec_devinfo {
9398
u32 quirks;
9499
};
@@ -418,13 +423,14 @@ static int
418423
fec_enet_create_page_pool(struct fec_enet_private *fep,
419424
struct fec_enet_priv_rx_q *rxq, int size)
420425
{
426+
struct bpf_prog *xdp_prog = READ_ONCE(fep->xdp_prog);
421427
struct page_pool_params pp_params = {
422428
.order = 0,
423429
.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV,
424430
.pool_size = size,
425431
.nid = dev_to_node(&fep->pdev->dev),
426432
.dev = &fep->pdev->dev,
427-
.dma_dir = DMA_FROM_DEVICE,
433+
.dma_dir = xdp_prog ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE,
428434
.offset = FEC_ENET_XDP_HEADROOM,
429435
.max_len = FEC_ENET_RX_FRSIZE,
430436
};
@@ -1499,6 +1505,59 @@ static void fec_enet_update_cbd(struct fec_enet_priv_rx_q *rxq,
14991505
bdp->cbd_bufaddr = cpu_to_fec32(phys_addr);
15001506
}
15011507

1508+
static u32
1509+
fec_enet_run_xdp(struct fec_enet_private *fep, struct bpf_prog *prog,
1510+
struct xdp_buff *xdp, struct fec_enet_priv_rx_q *rxq, int index)
1511+
{
1512+
unsigned int sync, len = xdp->data_end - xdp->data;
1513+
u32 ret = FEC_ENET_XDP_PASS;
1514+
struct page *page;
1515+
int err;
1516+
u32 act;
1517+
1518+
act = bpf_prog_run_xdp(prog, xdp);
1519+
1520+
/* Due xdp_adjust_tail: DMA sync for_device cover max len CPU touch */
1521+
sync = xdp->data_end - xdp->data_hard_start - FEC_ENET_XDP_HEADROOM;
1522+
sync = max(sync, len);
1523+
1524+
switch (act) {
1525+
case XDP_PASS:
1526+
ret = FEC_ENET_XDP_PASS;
1527+
break;
1528+
1529+
case XDP_REDIRECT:
1530+
err = xdp_do_redirect(fep->netdev, xdp, prog);
1531+
if (!err) {
1532+
ret = FEC_ENET_XDP_REDIR;
1533+
} else {
1534+
ret = FEC_ENET_XDP_CONSUMED;
1535+
page = virt_to_head_page(xdp->data);
1536+
page_pool_put_page(rxq->page_pool, page, sync, true);
1537+
}
1538+
break;
1539+
1540+
default:
1541+
bpf_warn_invalid_xdp_action(fep->netdev, prog, act);
1542+
fallthrough;
1543+
1544+
case XDP_TX:
1545+
bpf_warn_invalid_xdp_action(fep->netdev, prog, act);
1546+
fallthrough;
1547+
1548+
case XDP_ABORTED:
1549+
fallthrough; /* handle aborts by dropping packet */
1550+
1551+
case XDP_DROP:
1552+
ret = FEC_ENET_XDP_CONSUMED;
1553+
page = virt_to_head_page(xdp->data);
1554+
page_pool_put_page(rxq->page_pool, page, sync, true);
1555+
break;
1556+
}
1557+
1558+
return ret;
1559+
}
1560+
15021561
/* During a receive, the bd_rx.cur points to the current incoming buffer.
15031562
* When we update through the ring, if the next incoming buffer has
15041563
* not been given to the system, we just set the empty indicator,
@@ -1520,6 +1579,9 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
15201579
u16 vlan_tag;
15211580
int index = 0;
15221581
bool need_swap = fep->quirks & FEC_QUIRK_SWAP_FRAME;
1582+
struct bpf_prog *xdp_prog = READ_ONCE(fep->xdp_prog);
1583+
u32 ret, xdp_result = FEC_ENET_XDP_PASS;
1584+
struct xdp_buff xdp;
15231585
struct page *page;
15241586

15251587
#ifdef CONFIG_M532x
@@ -1531,6 +1593,7 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
15311593
* These get messed up if we get called due to a busy condition.
15321594
*/
15331595
bdp = rxq->bd.cur;
1596+
xdp_init_buff(&xdp, PAGE_SIZE, &rxq->xdp_rxq);
15341597

15351598
while (!((status = fec16_to_cpu(bdp->cbd_sc)) & BD_ENET_RX_EMPTY)) {
15361599

@@ -1580,6 +1643,17 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
15801643
prefetch(page_address(page));
15811644
fec_enet_update_cbd(rxq, bdp, index);
15821645

1646+
if (xdp_prog) {
1647+
xdp_buff_clear_frags_flag(&xdp);
1648+
xdp_prepare_buff(&xdp, page_address(page),
1649+
FEC_ENET_XDP_HEADROOM, pkt_len, false);
1650+
1651+
ret = fec_enet_run_xdp(fep, xdp_prog, &xdp, rxq, index);
1652+
xdp_result |= ret;
1653+
if (ret != FEC_ENET_XDP_PASS)
1654+
goto rx_processing_done;
1655+
}
1656+
15831657
/* The packet length includes FCS, but we don't want to
15841658
* include that when passing upstream as it messes up
15851659
* bridging applications.
@@ -1675,6 +1749,10 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
16751749
writel(0, rxq->bd.reg_desc_active);
16761750
}
16771751
rxq->bd.cur = bdp;
1752+
1753+
if (xdp_result & FEC_ENET_XDP_REDIR)
1754+
xdp_do_flush_map();
1755+
16781756
return pkt_received;
16791757
}
16801758

@@ -3518,6 +3596,148 @@ static u16 fec_enet_select_queue(struct net_device *ndev, struct sk_buff *skb,
35183596
return fec_enet_vlan_pri_to_queue[vlan_tag >> 13];
35193597
}
35203598

3599+
static int fec_enet_bpf(struct net_device *dev, struct netdev_bpf *bpf)
3600+
{
3601+
struct fec_enet_private *fep = netdev_priv(dev);
3602+
bool is_run = netif_running(dev);
3603+
struct bpf_prog *old_prog;
3604+
3605+
switch (bpf->command) {
3606+
case XDP_SETUP_PROG:
3607+
if (is_run) {
3608+
napi_disable(&fep->napi);
3609+
netif_tx_disable(dev);
3610+
}
3611+
3612+
old_prog = xchg(&fep->xdp_prog, bpf->prog);
3613+
fec_restart(dev);
3614+
3615+
if (is_run) {
3616+
napi_enable(&fep->napi);
3617+
netif_tx_start_all_queues(dev);
3618+
}
3619+
3620+
if (old_prog)
3621+
bpf_prog_put(old_prog);
3622+
3623+
return 0;
3624+
3625+
case XDP_SETUP_XSK_POOL:
3626+
return -EOPNOTSUPP;
3627+
3628+
default:
3629+
return -EOPNOTSUPP;
3630+
}
3631+
}
3632+
3633+
static int
3634+
fec_enet_xdp_get_tx_queue(struct fec_enet_private *fep, int cpu)
3635+
{
3636+
int index = cpu;
3637+
3638+
if (unlikely(index < 0))
3639+
index = 0;
3640+
3641+
while (index >= fep->num_tx_queues)
3642+
index -= fep->num_tx_queues;
3643+
3644+
return index;
3645+
}
3646+
3647+
static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
3648+
struct fec_enet_priv_tx_q *txq,
3649+
struct xdp_frame *frame)
3650+
{
3651+
unsigned int index, status, estatus;
3652+
struct bufdesc *bdp, *last_bdp;
3653+
dma_addr_t dma_addr;
3654+
int entries_free;
3655+
3656+
entries_free = fec_enet_get_free_txdesc_num(txq);
3657+
if (entries_free < MAX_SKB_FRAGS + 1) {
3658+
netdev_err(fep->netdev, "NOT enough BD for SG!\n");
3659+
return NETDEV_TX_OK;
3660+
}
3661+
3662+
/* Fill in a Tx ring entry */
3663+
bdp = txq->bd.cur;
3664+
last_bdp = bdp;
3665+
status = fec16_to_cpu(bdp->cbd_sc);
3666+
status &= ~BD_ENET_TX_STATS;
3667+
3668+
index = fec_enet_get_bd_index(bdp, &txq->bd);
3669+
3670+
dma_addr = dma_map_single(&fep->pdev->dev, frame->data,
3671+
frame->len, DMA_TO_DEVICE);
3672+
if (dma_mapping_error(&fep->pdev->dev, dma_addr))
3673+
return FEC_ENET_XDP_CONSUMED;
3674+
3675+
status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
3676+
if (fep->bufdesc_ex)
3677+
estatus = BD_ENET_TX_INT;
3678+
3679+
bdp->cbd_bufaddr = cpu_to_fec32(dma_addr);
3680+
bdp->cbd_datlen = cpu_to_fec16(frame->len);
3681+
3682+
if (fep->bufdesc_ex) {
3683+
struct bufdesc_ex *ebdp = (struct bufdesc_ex *)bdp;
3684+
3685+
if (fep->quirks & FEC_QUIRK_HAS_AVB)
3686+
estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
3687+
3688+
ebdp->cbd_bdu = 0;
3689+
ebdp->cbd_esc = cpu_to_fec32(estatus);
3690+
}
3691+
3692+
index = fec_enet_get_bd_index(last_bdp, &txq->bd);
3693+
txq->tx_skbuff[index] = NULL;
3694+
3695+
/* Send it on its way. Tell FEC it's ready, interrupt when done,
3696+
* it's the last BD of the frame, and to put the CRC on the end.
3697+
*/
3698+
status |= (BD_ENET_TX_READY | BD_ENET_TX_TC);
3699+
bdp->cbd_sc = cpu_to_fec16(status);
3700+
3701+
/* If this was the last BD in the ring, start at the beginning again. */
3702+
bdp = fec_enet_get_nextdesc(last_bdp, &txq->bd);
3703+
3704+
txq->bd.cur = bdp;
3705+
3706+
return 0;
3707+
}
3708+
3709+
static int fec_enet_xdp_xmit(struct net_device *dev,
3710+
int num_frames,
3711+
struct xdp_frame **frames,
3712+
u32 flags)
3713+
{
3714+
struct fec_enet_private *fep = netdev_priv(dev);
3715+
struct fec_enet_priv_tx_q *txq;
3716+
int cpu = smp_processor_id();
3717+
struct netdev_queue *nq;
3718+
unsigned int queue;
3719+
int i;
3720+
3721+
queue = fec_enet_xdp_get_tx_queue(fep, cpu);
3722+
txq = fep->tx_queue[queue];
3723+
nq = netdev_get_tx_queue(fep->netdev, queue);
3724+
3725+
__netif_tx_lock(nq, cpu);
3726+
3727+
for (i = 0; i < num_frames; i++)
3728+
fec_enet_txq_xmit_frame(fep, txq, frames[i]);
3729+
3730+
/* Make sure the update to bdp and tx_skbuff are performed. */
3731+
wmb();
3732+
3733+
/* Trigger transmission start */
3734+
writel(0, txq->bd.reg_desc_active);
3735+
3736+
__netif_tx_unlock(nq);
3737+
3738+
return num_frames;
3739+
}
3740+
35213741
static const struct net_device_ops fec_netdev_ops = {
35223742
.ndo_open = fec_enet_open,
35233743
.ndo_stop = fec_enet_close,
@@ -3532,6 +3752,8 @@ static const struct net_device_ops fec_netdev_ops = {
35323752
.ndo_poll_controller = fec_poll_controller,
35333753
#endif
35343754
.ndo_set_features = fec_set_features,
3755+
.ndo_bpf = fec_enet_bpf,
3756+
.ndo_xdp_xmit = fec_enet_xdp_xmit,
35353757
};
35363758

35373759
static const unsigned short offset_des_active_rxq[] = {

0 commit comments

Comments
 (0)