* [PATCH 1/3] i40e: Initial support for XDP
2016-12-08 17:00 [PATCH 0/3] i40e: Support for XDP Björn Töpel
@ 2016-12-08 17:00 ` Björn Töpel
2016-12-08 17:00 ` [PATCH 2/3] i40e: Add XDP_TX support Björn Töpel
` (2 subsequent siblings)
3 siblings, 0 replies; 7+ messages in thread
From: Björn Töpel @ 2016-12-08 17:00 UTC (permalink / raw)
To: jeffrey.t.kirsher, intel-wired-lan
Cc: Björn Töpel, john.r.fastabend, magnus.karlsson, netdev
From: Björn Töpel <bjorn.topel@intel.com>
This commit adds basic XDP support for i40e derived NICs. All XDP
actions will end up in XDP_DROP.
Only the default/main VSI has support for enabling XDP.
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
drivers/net/ethernet/intel/i40e/i40e.h | 13 +++
drivers/net/ethernet/intel/i40e/i40e_ethtool.c | 3 +
drivers/net/ethernet/intel/i40e/i40e_main.c | 74 +++++++++++++
drivers/net/ethernet/intel/i40e/i40e_txrx.c | 146 ++++++++++++++++++++-----
drivers/net/ethernet/intel/i40e/i40e_txrx.h | 2 +
5 files changed, 213 insertions(+), 25 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h
index ba8d30984bee..05d805f439e6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e.h
+++ b/drivers/net/ethernet/intel/i40e/i40e.h
@@ -545,6 +545,8 @@ struct i40e_vsi {
struct i40e_ring **rx_rings;
struct i40e_ring **tx_rings;
+ struct bpf_prog *xdp_prog;
+
u32 active_filters;
u32 promisc_threshold;
@@ -904,4 +906,15 @@ i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf);
i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf);
i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf);
void i40e_print_link_message(struct i40e_vsi *vsi, bool isup);
+
+/**
+ * i40e_enabled_xdp_vsi - Check if VSI has XDP enabled
+ * @vsi: pointer to a vsi
+ *
+ * Returns true if the VSI has XDP enabled.
+ **/
+static inline bool i40e_enabled_xdp_vsi(const struct i40e_vsi *vsi)
+{
+ return vsi->xdp_prog;
+}
#endif /* _I40E_H_ */
diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index cc1465aac2ef..831bbc208fc8 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -1254,6 +1254,9 @@ static int i40e_set_ringparam(struct net_device *netdev,
if ((ring->rx_mini_pending) || (ring->rx_jumbo_pending))
return -EINVAL;
+ if (i40e_enabled_xdp_vsi(vsi))
+ return -EINVAL;
+
if (ring->tx_pending > I40E_MAX_NUM_DESCRIPTORS ||
ring->tx_pending < I40E_MIN_NUM_DESCRIPTORS ||
ring->rx_pending > I40E_MAX_NUM_DESCRIPTORS ||
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index da4cbe32eb86..db0240213f3b 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -24,6 +24,7 @@
*
******************************************************************************/
+#include <linux/bpf.h>
#include <linux/etherdevice.h>
#include <linux/of_net.h>
#include <linux/pci.h>
@@ -2431,6 +2432,13 @@ static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
struct i40e_netdev_priv *np = netdev_priv(netdev);
struct i40e_vsi *vsi = np->vsi;
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
+
+ if (max_frame > I40E_RXBUFFER_2048)
+ return -EINVAL;
+ }
+
netdev_info(netdev, "changing MTU from %d to %d\n",
netdev->mtu, new_mtu);
netdev->mtu = new_mtu;
@@ -3085,6 +3093,15 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring)
ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
writel(0, ring->tail);
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ struct bpf_prog *prog;
+
+ prog = bpf_prog_add(vsi->xdp_prog, 1);
+ if (IS_ERR(prog))
+ return PTR_ERR(prog);
+ ring->xdp_prog = prog;
+ }
+
i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
return 0;
@@ -9234,6 +9251,62 @@ static netdev_features_t i40e_features_check(struct sk_buff *skb,
return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
}
+/**
+ * i40e_xdp_setup - Add/remove an XDP program to a VSI
+ * @vsi: the VSI to add the program
+ * @prog: the XDP program
+ **/
+static int i40e_xdp_setup(struct i40e_vsi *vsi,
+ struct bpf_prog *prog)
+{
+ struct i40e_pf *pf = vsi->back;
+ struct net_device *netdev = vsi->netdev;
+ int frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
+
+ if (frame_size > I40E_RXBUFFER_2048)
+ return -EINVAL;
+
+ if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
+ return -EINVAL;
+
+ if (!i40e_enabled_xdp_vsi(vsi) && !prog)
+ return 0;
+
+ i40e_prep_for_reset(pf);
+
+ if (vsi->xdp_prog)
+ bpf_prog_put(vsi->xdp_prog);
+ vsi->xdp_prog = prog;
+
+ i40e_reset_and_rebuild(pf, true);
+ return 0;
+}
+
+/**
+ * i40e_xdp - NDO for enabled/query
+ * @dev: the netdev
+ * @xdp: XDP program
+ **/
+static int i40e_xdp(struct net_device *dev,
+ struct netdev_xdp *xdp)
+{
+ struct i40e_netdev_priv *np = netdev_priv(dev);
+ struct i40e_vsi *vsi = np->vsi;
+
+ if (vsi->type != I40E_VSI_MAIN)
+ return -EINVAL;
+
+ switch (xdp->command) {
+ case XDP_SETUP_PROG:
+ return i40e_xdp_setup(vsi, xdp->prog);
+ case XDP_QUERY_PROG:
+ xdp->prog_attached = i40e_enabled_xdp_vsi(vsi);
+ return 0;
+ default:
+ return -EINVAL;
+ }
+}
+
static const struct net_device_ops i40e_netdev_ops = {
.ndo_open = i40e_open,
.ndo_stop = i40e_close,
@@ -9270,6 +9343,7 @@ static const struct net_device_ops i40e_netdev_ops = {
.ndo_features_check = i40e_features_check,
.ndo_bridge_getlink = i40e_ndo_bridge_getlink,
.ndo_bridge_setlink = i40e_ndo_bridge_setlink,
+ .ndo_xdp = i40e_xdp,
};
/**
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index 352cf7cd2ef4..d835a51dafa6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -24,6 +24,7 @@
*
******************************************************************************/
+#include <linux/bpf.h>
#include <linux/prefetch.h>
#include <net/busy_poll.h>
#include "i40e.h"
@@ -1040,6 +1041,11 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring)
rx_ring->next_to_alloc = 0;
rx_ring->next_to_clean = 0;
rx_ring->next_to_use = 0;
+
+ if (rx_ring->xdp_prog) {
+ bpf_prog_put(rx_ring->xdp_prog);
+ rx_ring->xdp_prog = NULL;
+ }
}
/**
@@ -1600,30 +1606,104 @@ static bool i40e_add_rx_frag(struct i40e_ring *rx_ring,
}
/**
+ * i40e_run_xdp - Runs an XDP program for an Rx ring
+ * @rx_ring: Rx ring used for XDP
+ * @rx_buffer: current Rx buffer
+ * @rx_desc: current Rx descriptor
+ * @xdp_prog: the XDP program to run
+ *
+ * Returns true if the XDP program consumed the incoming frame. False
+ * means pass the frame to the good old stack.
+ **/
+static bool i40e_run_xdp(struct i40e_ring *rx_ring,
+ struct i40e_rx_buffer *rx_buffer,
+ union i40e_rx_desc *rx_desc,
+ struct bpf_prog *xdp_prog)
+{
+ u64 qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len);
+ unsigned int size = (qword & I40E_RXD_QW1_LENGTH_PBUF_MASK) >>
+ I40E_RXD_QW1_LENGTH_PBUF_SHIFT;
+ struct xdp_buff xdp;
+ u32 xdp_action;
+
+ WARN_ON(!i40e_test_staterr(rx_desc,
+ BIT(I40E_RX_DESC_STATUS_EOF_SHIFT)));
+
+ xdp.data = page_address(rx_buffer->page) + rx_buffer->page_offset;
+ xdp.data_end = xdp.data + size;
+ xdp_action = bpf_prog_run_xdp(xdp_prog, &xdp);
+
+ switch (xdp_action) {
+ case XDP_PASS:
+ return false;
+ default:
+ bpf_warn_invalid_xdp_action(xdp_action);
+ case XDP_ABORTED:
+ case XDP_TX:
+ case XDP_DROP:
+ if (likely(!i40e_page_is_reserved(rx_buffer->page))) {
+ i40e_reuse_rx_page(rx_ring, rx_buffer);
+ rx_ring->rx_stats.page_reuse_count++;
+ break;
+ }
+
+ /* we are not reusing the buffer so unmap it */
+ dma_unmap_page(rx_ring->dev, rx_buffer->dma, PAGE_SIZE,
+ DMA_FROM_DEVICE);
+ __free_pages(rx_buffer->page, 0);
+ }
+
+ /* clear contents of buffer_info */
+ rx_buffer->page = NULL;
+ return true; /* Swallowed by XDP */
+}
+
+/**
* i40e_fetch_rx_buffer - Allocate skb and populate it
* @rx_ring: rx descriptor ring to transact packets on
* @rx_desc: descriptor containing info written by hardware
+ * @skb: The allocated skb, if any
*
- * This function allocates an skb on the fly, and populates it with the page
- * data from the current receive descriptor, taking care to set up the skb
- * correctly, as well as handling calling the page recycle function if
- * necessary.
+ * Unless XDP is enabled, this function allocates an skb on the fly,
+ * and populates it with the page data from the current receive
+ * descriptor, taking care to set up the skb correctly, as well as
+ * handling calling the page recycle function if necessary.
+ *
+ * If the received frame was handled by XDP, true is
+ * returned. Otherwise, the skb is returned to the caller via the skb
+ * parameter.
*/
static inline
-struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring,
- union i40e_rx_desc *rx_desc)
+bool i40e_fetch_rx_buffer(struct i40e_ring *rx_ring,
+ union i40e_rx_desc *rx_desc,
+ struct sk_buff **skb)
{
struct i40e_rx_buffer *rx_buffer;
- struct sk_buff *skb;
struct page *page;
rx_buffer = &rx_ring->rx_bi[rx_ring->next_to_clean];
page = rx_buffer->page;
prefetchw(page);
- skb = rx_buffer->skb;
+ /* we are reusing so sync this buffer for CPU use */
+ dma_sync_single_range_for_cpu(rx_ring->dev,
+ rx_buffer->dma,
+ rx_buffer->page_offset,
+ I40E_RXBUFFER_2048,
+ DMA_FROM_DEVICE);
+
+ if (rx_ring->xdp_prog) {
+ bool xdp_consumed;
+
+ xdp_consumed = i40e_run_xdp(rx_ring, rx_buffer,
+ rx_desc, rx_ring->xdp_prog);
+ if (xdp_consumed)
+ return true;
+ }
- if (likely(!skb)) {
+ *skb = rx_buffer->skb;
+
+ if (likely(!*skb)) {
void *page_addr = page_address(page) + rx_buffer->page_offset;
/* prefetch first cache line of first page */
@@ -1633,32 +1713,25 @@ struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring,
#endif
/* allocate a skb to store the frags */
- skb = __napi_alloc_skb(&rx_ring->q_vector->napi,
- I40E_RX_HDR_SIZE,
- GFP_ATOMIC | __GFP_NOWARN);
- if (unlikely(!skb)) {
+ *skb = __napi_alloc_skb(&rx_ring->q_vector->napi,
+ I40E_RX_HDR_SIZE,
+ GFP_ATOMIC | __GFP_NOWARN);
+ if (unlikely(!*skb)) {
rx_ring->rx_stats.alloc_buff_failed++;
- return NULL;
+ return false;
}
/* we will be copying header into skb->data in
* pskb_may_pull so it is in our interest to prefetch
* it now to avoid a possible cache miss
*/
- prefetchw(skb->data);
+ prefetchw((*skb)->data);
} else {
rx_buffer->skb = NULL;
}
- /* we are reusing so sync this buffer for CPU use */
- dma_sync_single_range_for_cpu(rx_ring->dev,
- rx_buffer->dma,
- rx_buffer->page_offset,
- I40E_RXBUFFER_2048,
- DMA_FROM_DEVICE);
-
/* pull page into skb */
- if (i40e_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) {
+ if (i40e_add_rx_frag(rx_ring, rx_buffer, rx_desc, *skb)) {
/* hand second half of page back to the ring */
i40e_reuse_rx_page(rx_ring, rx_buffer);
rx_ring->rx_stats.page_reuse_count++;
@@ -1671,7 +1744,7 @@ struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring,
/* clear contents of buffer_info */
rx_buffer->page = NULL;
- return skb;
+ return false;
}
/**
@@ -1716,6 +1789,20 @@ static bool i40e_is_non_eop(struct i40e_ring *rx_ring,
}
/**
+ * i40e_update_rx_next_to_clean - Bumps the next-to-clean for an Rx ing
+ * @rx_ring: Rx ring to bump
+ **/
+static void i40e_update_rx_next_to_clean(struct i40e_ring *rx_ring)
+{
+ u32 ntc = rx_ring->next_to_clean + 1;
+
+ ntc = (ntc < rx_ring->count) ? ntc : 0;
+ rx_ring->next_to_clean = ntc;
+
+ prefetch(I40E_RX_DESC(rx_ring, ntc));
+}
+
+/**
* i40e_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
* @rx_ring: rx descriptor ring to transact packets on
* @budget: Total limit on number of packets to process
@@ -1739,6 +1826,7 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget)
u16 vlan_tag;
u8 rx_ptype;
u64 qword;
+ bool xdp_consumed;
/* return some buffers to hardware, one at a time is too slow */
if (cleaned_count >= I40E_RX_BUFFER_WRITE) {
@@ -1764,7 +1852,15 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget)
*/
dma_rmb();
- skb = i40e_fetch_rx_buffer(rx_ring, rx_desc);
+ xdp_consumed = i40e_fetch_rx_buffer(rx_ring, rx_desc, &skb);
+ if (xdp_consumed) {
+ cleaned_count++;
+
+ i40e_update_rx_next_to_clean(rx_ring);
+ total_rx_packets++;
+ continue;
+ }
+
if (!skb)
break;
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index e065321ce8ed..957d856a82c4 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -341,6 +341,8 @@ struct i40e_ring {
struct rcu_head rcu; /* to avoid race on free */
u16 next_to_alloc;
+
+ struct bpf_prog *xdp_prog;
} ____cacheline_internodealigned_in_smp;
enum i40e_latency_range {
--
2.9.3
^ permalink raw reply related [flat|nested] 7+ messages in thread* [PATCH 2/3] i40e: Add XDP_TX support
2016-12-08 17:00 [PATCH 0/3] i40e: Support for XDP Björn Töpel
2016-12-08 17:00 ` [PATCH 1/3] i40e: Initial support " Björn Töpel
@ 2016-12-08 17:00 ` Björn Töpel
2016-12-08 19:20 ` kbuild test robot
2016-12-08 17:00 ` [PATCH 3/3] i40e: Don't reset/rebuild rings on XDP program swap Björn Töpel
2016-12-08 17:52 ` [Intel-wired-lan] [PATCH 0/3] i40e: Support for XDP John Fastabend
3 siblings, 1 reply; 7+ messages in thread
From: Björn Töpel @ 2016-12-08 17:00 UTC (permalink / raw)
To: jeffrey.t.kirsher, intel-wired-lan
Cc: Björn Töpel, john.r.fastabend, magnus.karlsson, netdev
From: Björn Töpel <bjorn.topel@intel.com>
This patch adds proper XDP_TX support.
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
drivers/net/ethernet/intel/i40e/i40e.h | 5 +
drivers/net/ethernet/intel/i40e/i40e_main.c | 273 ++++++++++++++++++++-----
drivers/net/ethernet/intel/i40e/i40e_txrx.c | 304 +++++++++++++++++++++++-----
drivers/net/ethernet/intel/i40e/i40e_txrx.h | 5 +
4 files changed, 491 insertions(+), 96 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h
index 05d805f439e6..adc1f3f32729 100644
--- a/drivers/net/ethernet/intel/i40e/i40e.h
+++ b/drivers/net/ethernet/intel/i40e/i40e.h
@@ -545,6 +545,10 @@ struct i40e_vsi {
struct i40e_ring **rx_rings;
struct i40e_ring **tx_rings;
+ /* The XDP rings are Tx only, and follows the count of the
+ * regular rings, i.e. alloc_queue_pairs/num_queue_pairs
+ */
+ struct i40e_ring **xdp_rings;
struct bpf_prog *xdp_prog;
u32 active_filters;
@@ -622,6 +626,7 @@ struct i40e_q_vector {
struct i40e_ring_container rx;
struct i40e_ring_container tx;
+ struct i40e_ring_container xdp;
u8 num_ringpairs; /* total number of ring pairs in vector */
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index db0240213f3b..9310a5712ae3 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -107,6 +107,18 @@ MODULE_VERSION(DRV_VERSION);
static struct workqueue_struct *i40e_wq;
/**
+ * i40e_alloc_queue_pairs_xdp_vsi - required # of XDP queue pairs
+ * @vsi: pointer to a vsi
+ **/
+static u16 i40e_alloc_queue_pairs_xdp_vsi(const struct i40e_vsi *vsi)
+{
+ if (i40e_enabled_xdp_vsi(vsi))
+ return vsi->alloc_queue_pairs;
+
+ return 0;
+}
+
+/**
* i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
* @hw: pointer to the HW structure
* @mem: ptr to mem struct to fill out
@@ -2828,6 +2840,12 @@ static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
for (i = 0; i < vsi->num_queue_pairs && !err; i++)
err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
+ if (!i40e_enabled_xdp_vsi(vsi))
+ return err;
+
+ for (i = 0; i < vsi->num_queue_pairs && !err; i++)
+ err = i40e_setup_tx_descriptors(vsi->xdp_rings[i]);
+
return err;
}
@@ -2841,12 +2859,17 @@ static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
{
int i;
- if (!vsi->tx_rings)
- return;
+ if (vsi->tx_rings) {
+ for (i = 0; i < vsi->num_queue_pairs; i++)
+ if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
+ i40e_free_tx_resources(vsi->tx_rings[i]);
+ }
- for (i = 0; i < vsi->num_queue_pairs; i++)
- if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
- i40e_free_tx_resources(vsi->tx_rings[i]);
+ if (vsi->xdp_rings) {
+ for (i = 0; i < vsi->num_queue_pairs; i++)
+ if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
+ i40e_free_tx_resources(vsi->xdp_rings[i]);
+ }
}
/**
@@ -3121,6 +3144,12 @@ static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
err = i40e_configure_tx_ring(vsi->tx_rings[i]);
+ if (!i40e_enabled_xdp_vsi(vsi))
+ return err;
+
+ for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
+ err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
+
return err;
}
@@ -3269,7 +3298,7 @@ static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
struct i40e_hw *hw = &pf->hw;
u16 vector;
int i, q;
- u32 qp;
+ u32 qp, qp_idx = 0;
/* The interrupt indexing is offset by 1 in the PFINT_ITRn
* and PFINT_LNKLSTn registers, e.g.:
@@ -3296,16 +3325,33 @@ static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
for (q = 0; q < q_vector->num_ringpairs; q++) {
u32 val;
+ u32 nqp = qp;
+
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ nqp = vsi->base_queue +
+ vsi->xdp_rings[qp_idx]->queue_index;
+ }
val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
- (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
- (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
- (qp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
+ (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
+ (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
+ (nqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
(I40E_QUEUE_TYPE_TX
<< I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
wr32(hw, I40E_QINT_RQCTL(qp), val);
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
+ (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
+ (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
+ (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
+ (I40E_QUEUE_TYPE_TX
+ << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
+
+ wr32(hw, I40E_QINT_TQCTL(nqp), val);
+ }
+
val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
(I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
(vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
@@ -3320,6 +3366,7 @@ static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
wr32(hw, I40E_QINT_TQCTL(qp), val);
qp++;
+ qp_idx++;
}
}
@@ -3562,6 +3609,10 @@ static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
for (i = 0; i < vsi->num_queue_pairs; i++) {
wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx),
+ 0);
+ }
}
if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
@@ -3871,6 +3922,24 @@ static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
}
/**
+ * i40e_map_vector_to_xdp_ring - Assigns the XDP Tx queue to the vector
+ * @vsi: the VSI being configured
+ * @v_idx: vector index
+ * @xdp_idx: XDP Tx queue index
+ **/
+static void i40e_map_vector_to_xdp_ring(struct i40e_vsi *vsi, int v_idx,
+ int xdp_idx)
+{
+ struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
+ struct i40e_ring *xdp_ring = vsi->xdp_rings[xdp_idx];
+
+ xdp_ring->q_vector = q_vector;
+ xdp_ring->next = q_vector->xdp.ring;
+ q_vector->xdp.ring = xdp_ring;
+ q_vector->xdp.count++;
+}
+
+/**
* i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
* @vsi: the VSI being configured
*
@@ -3903,11 +3972,17 @@ static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
q_vector->rx.count = 0;
q_vector->tx.count = 0;
+ q_vector->xdp.count = 0;
q_vector->rx.ring = NULL;
q_vector->tx.ring = NULL;
+ q_vector->xdp.ring = NULL;
while (num_ringpairs--) {
i40e_map_vector_to_qp(vsi, v_start, qp_idx);
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ i40e_map_vector_to_xdp_ring(vsi, v_start,
+ qp_idx);
+ }
qp_idx++;
qp_remaining--;
}
@@ -4001,56 +4076,82 @@ static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
}
/**
- * i40e_vsi_control_tx - Start or stop a VSI's rings
+ * i40e_vsi_control_txq - Start or stop a VSI's queue
* @vsi: the VSI being configured
* @enable: start or stop the rings
+ * @pf_q: the PF queue
**/
-static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
+static int i40e_vsi_control_txq(struct i40e_vsi *vsi, bool enable, int pf_q)
{
struct i40e_pf *pf = vsi->back;
struct i40e_hw *hw = &pf->hw;
- int i, j, pf_q, ret = 0;
+ int j, ret = 0;
u32 tx_reg;
- pf_q = vsi->base_queue;
- for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
+ /* warn the TX unit of coming changes */
+ i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
+ if (!enable)
+ usleep_range(10, 20);
- /* warn the TX unit of coming changes */
- i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
- if (!enable)
- usleep_range(10, 20);
+ for (j = 0; j < 50; j++) {
+ tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
+ if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
+ ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
+ break;
+ usleep_range(1000, 2000);
+ }
+ /* Skip if the queue is already in the requested state */
+ if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
+ return 0;
- for (j = 0; j < 50; j++) {
- tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
- if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
- ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
- break;
- usleep_range(1000, 2000);
- }
- /* Skip if the queue is already in the requested state */
- if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
- continue;
+ /* turn on/off the queue */
+ if (enable) {
+ wr32(hw, I40E_QTX_HEAD(pf_q), 0);
+ tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
+ } else {
+ tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
+ }
- /* turn on/off the queue */
- if (enable) {
- wr32(hw, I40E_QTX_HEAD(pf_q), 0);
- tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
- } else {
- tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
- }
+ wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
+ /* No waiting for the Tx queue to disable */
+ if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
+ return 0;
- wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
- /* No waiting for the Tx queue to disable */
- if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
- continue;
+ /* wait for the change to finish */
+ ret = i40e_pf_txq_wait(pf, pf_q, enable);
+ if (ret) {
+ dev_info(&pf->pdev->dev,
+ "VSI seid %d Tx ring %d %sable timeout\n",
+ vsi->seid, pf_q, (enable ? "en" : "dis"));
+ return ret;
+ }
+ return 0;
+}
- /* wait for the change to finish */
- ret = i40e_pf_txq_wait(pf, pf_q, enable);
- if (ret) {
- dev_info(&pf->pdev->dev,
- "VSI seid %d Tx ring %d %sable timeout\n",
- vsi->seid, pf_q, (enable ? "en" : "dis"));
+/**
+ * i40e_vsi_control_tx - Start or stop a VSI's rings
+ * @vsi: the VSI being configured
+ * @enable: start or stop the rings
+ **/
+static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
+{
+ struct i40e_pf *pf = vsi->back;
+ struct i40e_hw *hw = &pf->hw;
+ int i, pf_q, ret = 0;
+
+ pf_q = vsi->base_queue;
+ for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
+ ret = i40e_vsi_control_txq(vsi, enable, pf_q);
+ if (ret)
break;
+ }
+
+ if (!ret && i40e_enabled_xdp_vsi(vsi)) {
+ for (i = 0; i < vsi->num_queue_pairs; i++) {
+ pf_q = vsi->base_queue + vsi->xdp_rings[i]->queue_index;
+ ret = i40e_vsi_control_txq(vsi, enable, pf_q);
+ if (ret)
+ break;
}
}
@@ -4311,6 +4412,9 @@ static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
i40e_for_each_ring(ring, q_vector->rx)
ring->q_vector = NULL;
+ i40e_for_each_ring(ring, q_vector->xdp)
+ ring->q_vector = NULL;
+
/* only VSI w/ an associated netdev is set up w/ NAPI */
if (vsi->netdev)
netif_napi_del(&q_vector->napi);
@@ -4534,6 +4638,21 @@ static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
}
}
+ if (!i40e_enabled_xdp_vsi(vsi))
+ return 0;
+
+ for (i = 0; i < vsi->num_queue_pairs; i++) {
+ pf_q = vsi->base_queue + vsi->xdp_rings[i]->queue_index;
+ /* Check and wait for the disable status of the queue */
+ ret = i40e_pf_txq_wait(pf, pf_q, false);
+ if (ret) {
+ dev_info(&pf->pdev->dev,
+ "VSI seid %d XDP Tx ring %d disable timeout\n",
+ vsi->seid, pf_q);
+ return ret;
+ }
+ }
+
return 0;
}
@@ -5474,6 +5593,8 @@ void i40e_down(struct i40e_vsi *vsi)
for (i = 0; i < vsi->num_queue_pairs; i++) {
i40e_clean_tx_ring(vsi->tx_rings[i]);
+ if (i40e_enabled_xdp_vsi(vsi))
+ i40e_clean_tx_ring(vsi->xdp_rings[i]);
i40e_clean_rx_ring(vsi->rx_rings[i]);
}
@@ -7446,6 +7567,16 @@ static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
return -ENOMEM;
vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
+ if (i40e_enabled_xdp_vsi(vsi)) {
+ size = sizeof(struct i40e_ring *) *
+ i40e_alloc_queue_pairs_xdp_vsi(vsi);
+ vsi->xdp_rings = kzalloc(size, GFP_KERNEL);
+ if (!vsi->xdp_rings) {
+ ret = -ENOMEM;
+ goto err_xdp_rings;
+ }
+ }
+
if (alloc_qvectors) {
/* allocate memory for q_vector pointers */
size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
@@ -7458,6 +7589,8 @@ static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
return ret;
err_vectors:
+ kfree(vsi->xdp_rings);
+err_xdp_rings:
kfree(vsi->tx_rings);
return ret;
}
@@ -7564,6 +7697,8 @@ static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
kfree(vsi->tx_rings);
vsi->tx_rings = NULL;
vsi->rx_rings = NULL;
+ kfree(vsi->xdp_rings);
+ vsi->xdp_rings = NULL;
}
/**
@@ -7649,6 +7784,13 @@ static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
vsi->rx_rings[i] = NULL;
}
}
+
+ if (vsi->xdp_rings && vsi->xdp_rings[0]) {
+ for (i = 0; i < vsi->alloc_queue_pairs; i++) {
+ kfree_rcu(vsi->xdp_rings[i], rcu);
+ vsi->xdp_rings[i] = NULL;
+ }
+ }
}
/**
@@ -7696,6 +7838,31 @@ static int i40e_alloc_rings(struct i40e_vsi *vsi)
vsi->rx_rings[i] = rx_ring;
}
+ if (!i40e_enabled_xdp_vsi(vsi))
+ return 0;
+
+ for (i = 0; i < vsi->alloc_queue_pairs; i++) {
+ tx_ring = kzalloc(sizeof(*tx_ring), GFP_KERNEL);
+ if (!tx_ring)
+ goto err_out;
+
+ tx_ring->queue_index = vsi->alloc_queue_pairs + i;
+ tx_ring->reg_idx = vsi->base_queue + vsi->alloc_queue_pairs + i;
+ tx_ring->ring_active = false;
+ tx_ring->vsi = vsi;
+ tx_ring->netdev = NULL;
+ tx_ring->dev = &pf->pdev->dev;
+ tx_ring->count = vsi->num_desc;
+ tx_ring->size = 0;
+ tx_ring->dcb_tc = 0;
+ if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
+ tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
+ tx_ring->tx_itr_setting = pf->tx_itr_default;
+ tx_ring->xdp_sibling = vsi->rx_rings[i];
+ vsi->xdp_rings[i] = tx_ring;
+ vsi->rx_rings[i]->xdp_sibling = tx_ring;
+ }
+
return 0;
err_out:
@@ -9921,6 +10088,7 @@ static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
struct i40e_pf *pf;
u8 enabled_tc;
int ret;
+ u16 alloc_queue_pairs;
if (!vsi)
return NULL;
@@ -9936,11 +10104,13 @@ static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
if (ret)
goto err_vsi;
- ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
+ alloc_queue_pairs = vsi->alloc_queue_pairs +
+ i40e_alloc_queue_pairs_xdp_vsi(vsi);
+ ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
if (ret < 0) {
dev_info(&pf->pdev->dev,
"failed to get tracking for %d queues for VSI %d err %d\n",
- vsi->alloc_queue_pairs, vsi->seid, ret);
+ alloc_queue_pairs, vsi->seid, ret);
goto err_vsi;
}
vsi->base_queue = ret;
@@ -9998,6 +10168,7 @@ struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
struct i40e_veb *veb = NULL;
int ret, i;
int v_idx;
+ u16 alloc_queue_pairs;
/* The requested uplink_seid must be either
* - the PF's port seid
@@ -10082,13 +10253,15 @@ struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
pf->lan_vsi = v_idx;
else if (type == I40E_VSI_SRIOV)
vsi->vf_id = param1;
+
+ alloc_queue_pairs = vsi->alloc_queue_pairs +
+ i40e_alloc_queue_pairs_xdp_vsi(vsi);
/* assign it some queues */
- ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
- vsi->idx);
+ ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
if (ret < 0) {
dev_info(&pf->pdev->dev,
"failed to get tracking for %d queues for VSI %d err=%d\n",
- vsi->alloc_queue_pairs, vsi->seid, ret);
+ alloc_queue_pairs, vsi->seid, ret);
goto err_vsi;
}
vsi->base_queue = ret;
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index d835a51dafa6..fccdec7ae102 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -520,6 +520,8 @@ static void i40e_unmap_and_free_tx_resource(struct i40e_ring *ring,
if (tx_buffer->skb) {
if (tx_buffer->tx_flags & I40E_TX_FLAGS_FD_SB)
kfree(tx_buffer->raw_buf);
+ else if (tx_buffer->tx_flags & I40E_TX_FLAGS_XDP)
+ put_page(tx_buffer->page);
else
dev_kfree_skb_any(tx_buffer->skb);
if (dma_unmap_len(tx_buffer, len))
@@ -620,6 +622,15 @@ u32 i40e_get_tx_pending(struct i40e_ring *ring, bool in_sw)
#define WB_STRIDE 4
/**
+ * i40e_page_is_reserved - check if reuse is possible
+ * @page: page struct to check
+ */
+static inline bool i40e_page_is_reserved(struct page *page)
+{
+ return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
+}
+
+/**
* i40e_clean_tx_irq - Reclaim resources after transmit completes
* @vsi: the VSI we care about
* @tx_ring: Tx ring to clean
@@ -762,6 +773,98 @@ static bool i40e_clean_tx_irq(struct i40e_vsi *vsi,
return !!budget;
}
+static bool i40e_clean_xdp_irq(struct i40e_vsi *vsi,
+ struct i40e_ring *tx_ring)
+{
+ u16 i = tx_ring->next_to_clean;
+ struct i40e_tx_buffer *tx_buf;
+ struct i40e_tx_desc *tx_head;
+ struct i40e_tx_desc *tx_desc;
+ unsigned int total_bytes = 0, total_packets = 0;
+ unsigned int budget = vsi->work_limit;
+
+ tx_buf = &tx_ring->tx_bi[i];
+ tx_desc = I40E_TX_DESC(tx_ring, i);
+ i -= tx_ring->count;
+
+ tx_head = I40E_TX_DESC(tx_ring, i40e_get_head(tx_ring));
+
+ do {
+ struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
+
+ /* if next_to_watch is not set then there is no work pending */
+ if (!eop_desc)
+ break;
+
+ /* prevent any other reads prior to eop_desc */
+ read_barrier_depends();
+
+ /* we have caught up to head, no work left to do */
+ if (tx_head == tx_desc)
+ break;
+
+ /* clear next_to_watch to prevent false hangs */
+ tx_buf->next_to_watch = NULL;
+
+ /* update the statistics for this packet */
+ total_bytes += tx_buf->bytecount;
+ total_packets += tx_buf->gso_segs;
+
+ put_page(tx_buf->page);
+
+ /* unmap skb header data */
+ dma_unmap_single(tx_ring->dev,
+ dma_unmap_addr(tx_buf, dma),
+ dma_unmap_len(tx_buf, len),
+ DMA_TO_DEVICE);
+
+ /* clear tx_buffer data */
+ tx_buf->skb = NULL;
+ dma_unmap_len_set(tx_buf, len, 0);
+
+ /* move us one more past the eop_desc for start of next pkt */
+ tx_buf++;
+ tx_desc++;
+ i++;
+ if (unlikely(!i)) {
+ i -= tx_ring->count;
+ tx_buf = tx_ring->tx_bi;
+ tx_desc = I40E_TX_DESC(tx_ring, 0);
+ }
+
+ prefetch(tx_desc);
+
+ /* update budget accounting */
+ budget--;
+ } while (likely(budget));
+
+ i += tx_ring->count;
+ tx_ring->next_to_clean = i;
+ u64_stats_update_begin(&tx_ring->syncp);
+ tx_ring->stats.bytes += total_bytes;
+ tx_ring->stats.packets += total_packets;
+ u64_stats_update_end(&tx_ring->syncp);
+ tx_ring->q_vector->tx.total_bytes += total_bytes;
+ tx_ring->q_vector->tx.total_packets += total_packets;
+
+ if (tx_ring->flags & I40E_TXR_FLAGS_WB_ON_ITR) {
+ /* check to see if there are < 4 descriptors
+ * waiting to be written back, then kick the hardware to force
+ * them to be written back in case we stay in NAPI.
+ * In this mode on X722 we do not enable Interrupt.
+ */
+ unsigned int j = i40e_get_tx_pending(tx_ring, false);
+
+ if (budget &&
+ ((j / WB_STRIDE) == 0) && (j > 0) &&
+ !test_bit(__I40E_DOWN, &vsi->state) &&
+ (I40E_DESC_UNUSED(tx_ring) != tx_ring->count))
+ tx_ring->arm_wb = true;
+ }
+
+ return !!budget;
+}
+
/**
* i40e_enable_wb_on_itr - Arm hardware to do a wb, interrupts are not enabled
* @vsi: the VSI we care about
@@ -1496,35 +1599,46 @@ static bool i40e_cleanup_headers(struct i40e_ring *rx_ring, struct sk_buff *skb)
}
/**
- * i40e_reuse_rx_page - page flip buffer and store it back on the ring
- * @rx_ring: rx descriptor ring to store buffers on
- * @old_buff: donor buffer to have page reused
+ * i40e_try_flip_rx_page - attempts to flip a page for reuse
+ * @rx_buffer: The buffer to alter
*
- * Synchronizes page for reuse by the adapter
- **/
-static void i40e_reuse_rx_page(struct i40e_ring *rx_ring,
- struct i40e_rx_buffer *old_buff)
+ * Returns true if the page was successfully flipped and can be
+ * reused.
+ */
+static bool i40e_try_flip_rx_page(struct i40e_rx_buffer *rx_buffer)
{
- struct i40e_rx_buffer *new_buff;
- u16 nta = rx_ring->next_to_alloc;
+#if (PAGE_SIZE < 8192)
+ unsigned int truesize = I40E_RXBUFFER_2048;
+#else
+ unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
+ unsigned int last_offset = PAGE_SIZE - I40E_RXBUFFER_2048;
+#endif
- new_buff = &rx_ring->rx_bi[nta];
+ /* avoid re-using remote pages */
+ if (unlikely(i40e_page_is_reserved(rx_buffer->page)))
+ return false;
- /* update, and store next to alloc */
- nta++;
- rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
+#if (PAGE_SIZE < 8192)
+ /* if we are only owner of page we can reuse it */
+ if (unlikely(page_count(rx_buffer->page) != 1))
+ return false;
- /* transfer page from old buffer to new buffer */
- *new_buff = *old_buff;
-}
+ /* flip page offset to other buffer */
+ rx_buffer->page_offset ^= truesize;
+#else
+ /* move offset up to the next cache line */
+ rx_buffer->page_offset += truesize;
-/**
- * i40e_page_is_reserved - check if reuse is possible
- * @page: page struct to check
- */
-static inline bool i40e_page_is_reserved(struct page *page)
-{
- return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
+ if (rx_buffer->page_offset > last_offset)
+ return false;
+#endif
+
+ /* Even if we own the page, we are not allowed to use atomic_set()
+ * This would break get_page_unless_zero() users.
+ */
+ get_page(rx_buffer->page);
+
+ return true;
}
/**
@@ -1555,7 +1669,6 @@ static bool i40e_add_rx_frag(struct i40e_ring *rx_ring,
unsigned int truesize = I40E_RXBUFFER_2048;
#else
unsigned int truesize = ALIGN(size, L1_CACHE_BYTES);
- unsigned int last_offset = PAGE_SIZE - I40E_RXBUFFER_2048;
#endif
/* will the data fit in the skb we allocated? if so, just
@@ -1578,34 +1691,107 @@ static bool i40e_add_rx_frag(struct i40e_ring *rx_ring,
skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
rx_buffer->page_offset, size, truesize);
- /* avoid re-using remote pages */
- if (unlikely(i40e_page_is_reserved(page)))
- return false;
+ return i40e_try_flip_rx_page(rx_buffer);
+}
-#if (PAGE_SIZE < 8192)
- /* if we are only owner of page we can reuse it */
- if (unlikely(page_count(page) != 1))
+/**
+ * i40e_xdp_xmit_tail_bump - updates the tail and sets the RS bit
+ * @xdp_ring: XDP Tx ring
+ **/
+static
+void i40e_xdp_xmit_tail_bump(struct i40e_ring *xdp_ring)
+{
+ struct i40e_tx_desc *tx_desc;
+
+ /* Set RS and bump tail */
+ tx_desc = I40E_TX_DESC(xdp_ring, xdp_ring->curr_in_use);
+ tx_desc->cmd_type_offset_bsz |=
+ cpu_to_le64(I40E_TX_DESC_CMD_RS << I40E_TXD_QW1_CMD_SHIFT);
+ /* Force memory writes to complete before letting h/w know
+ * there are new descriptors to fetch. (Only applicable for
+ * weak-ordered memory model archs, such as IA-64).
+ */
+ wmb();
+ writel(xdp_ring->curr_in_use, xdp_ring->tail);
+
+ xdp_ring->xdp_needs_tail_bump = false;
+}
+
+/**
+ * i40e_xdp_xmit - transmit a frame on the XDP Tx queue
+ * @xdp_ring: XDP Tx ring
+ * @page: current page containing the frame
+ * @page_offset: offset where the frame resides
+ * @dma: Bus address of the frame
+ * @size: size of the frame
+ *
+ * Returns true successfully sent.
+ **/
+static bool i40e_xdp_xmit(void *data, size_t size, struct page *page,
+ struct i40e_ring *xdp_ring)
+{
+ struct i40e_tx_buffer *tx_bi;
+ struct i40e_tx_desc *tx_desc;
+ u16 i = xdp_ring->next_to_use;
+ dma_addr_t dma;
+
+ if (unlikely(I40E_DESC_UNUSED(xdp_ring) < 1)) {
+ if (xdp_ring->xdp_needs_tail_bump)
+ i40e_xdp_xmit_tail_bump(xdp_ring);
+ xdp_ring->tx_stats.tx_busy++;
return false;
+ }
- /* flip page offset to other buffer */
- rx_buffer->page_offset ^= truesize;
-#else
- /* move offset up to the next cache line */
- rx_buffer->page_offset += truesize;
+ tx_bi = &xdp_ring->tx_bi[i];
+ tx_bi->bytecount = size;
+ tx_bi->gso_segs = 1;
+ tx_bi->tx_flags = I40E_TX_FLAGS_XDP;
+ tx_bi->page = page;
- if (rx_buffer->page_offset > last_offset)
+ dma = dma_map_single(xdp_ring->dev, data, size, DMA_TO_DEVICE);
+ if (dma_mapping_error(xdp_ring->dev, dma))
return false;
-#endif
- /* Even if we own the page, we are not allowed to use atomic_set()
- * This would break get_page_unless_zero() users.
- */
- get_page(rx_buffer->page);
+ /* record length, and DMA address */
+ dma_unmap_len_set(tx_bi, len, size);
+ dma_unmap_addr_set(tx_bi, dma, dma);
+ tx_desc = I40E_TX_DESC(xdp_ring, i);
+ tx_desc->buffer_addr = cpu_to_le64(dma);
+ tx_desc->cmd_type_offset_bsz = build_ctob(I40E_TX_DESC_CMD_ICRC
+ | I40E_TX_DESC_CMD_EOP,
+ 0, size, 0);
+ tx_bi->next_to_watch = tx_desc;
+ xdp_ring->curr_in_use = i++;
+ xdp_ring->next_to_use = (i < xdp_ring->count) ? i : 0;
+ xdp_ring->xdp_needs_tail_bump = true;
return true;
}
/**
+ * i40e_reuse_rx_page - page flip buffer and store it back on the ring
+ * @rx_ring: rx descriptor ring to store buffers on
+ * @old_buff: donor buffer to have page reused
+ *
+ * Synchronizes page for reuse by the adapter
+ **/
+static void i40e_reuse_rx_page(struct i40e_ring *rx_ring,
+ struct i40e_rx_buffer *old_buff)
+{
+ struct i40e_rx_buffer *new_buff;
+ u16 nta = rx_ring->next_to_alloc;
+
+ new_buff = &rx_ring->rx_bi[nta];
+
+ /* update, and store next to alloc */
+ nta++;
+ rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
+
+ /* transfer page from old buffer to new buffer */
+ *new_buff = *old_buff;
+}
+
+/**
* i40e_run_xdp - Runs an XDP program for an Rx ring
* @rx_ring: Rx ring used for XDP
* @rx_buffer: current Rx buffer
@@ -1625,6 +1811,7 @@ static bool i40e_run_xdp(struct i40e_ring *rx_ring,
I40E_RXD_QW1_LENGTH_PBUF_SHIFT;
struct xdp_buff xdp;
u32 xdp_action;
+ bool tx_ok;
WARN_ON(!i40e_test_staterr(rx_desc,
BIT(I40E_RX_DESC_STATUS_EOF_SHIFT)));
@@ -1636,21 +1823,34 @@ static bool i40e_run_xdp(struct i40e_ring *rx_ring,
switch (xdp_action) {
case XDP_PASS:
return false;
- default:
- bpf_warn_invalid_xdp_action(xdp_action);
- case XDP_ABORTED:
case XDP_TX:
+ tx_ok = i40e_xdp_xmit(xdp.data, size, rx_buffer->page,
+ rx_ring->xdp_sibling);
+ if (likely(tx_ok)) {
+ if (i40e_try_flip_rx_page(rx_buffer)) {
+ i40e_reuse_rx_page(rx_ring, rx_buffer);
+ rx_ring->rx_stats.page_reuse_count++;
+ } else {
+ dma_unmap_page(rx_ring->dev, rx_buffer->dma,
+ PAGE_SIZE, DMA_FROM_DEVICE);
+ }
+ break;
+ }
+ case XDP_ABORTED:
case XDP_DROP:
+do_drop:
if (likely(!i40e_page_is_reserved(rx_buffer->page))) {
i40e_reuse_rx_page(rx_ring, rx_buffer);
rx_ring->rx_stats.page_reuse_count++;
break;
}
-
- /* we are not reusing the buffer so unmap it */
dma_unmap_page(rx_ring->dev, rx_buffer->dma, PAGE_SIZE,
DMA_FROM_DEVICE);
__free_pages(rx_buffer->page, 0);
+ break;
+ default:
+ bpf_warn_invalid_xdp_action(xdp_action);
+ goto do_drop;
}
/* clear contents of buffer_info */
@@ -2065,6 +2265,15 @@ int i40e_napi_poll(struct napi_struct *napi, int budget)
ring->arm_wb = false;
}
+ i40e_for_each_ring(ring, q_vector->xdp) {
+ if (!i40e_clean_xdp_irq(vsi, ring)) {
+ clean_complete = false;
+ continue;
+ }
+ arm_wb |= ring->arm_wb;
+ ring->arm_wb = false;
+ }
+
/* Handle case where we are called by netpoll with a budget of 0 */
if (budget <= 0)
goto tx_only;
@@ -2077,6 +2286,9 @@ int i40e_napi_poll(struct napi_struct *napi, int budget)
i40e_for_each_ring(ring, q_vector->rx) {
int cleaned = i40e_clean_rx_irq(ring, budget_per_ring);
+ if (ring->xdp_sibling && ring->xdp_sibling->xdp_needs_tail_bump)
+ i40e_xdp_xmit_tail_bump(ring->xdp_sibling);
+
work_done += cleaned;
/* if we clean as many as budgeted, we must not be done */
if (cleaned >= budget_per_ring)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index 957d856a82c4..4d9459134e69 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -220,6 +220,7 @@ static inline unsigned int i40e_txd_use_count(unsigned int size)
#define I40E_TX_FLAGS_TSYN BIT(8)
#define I40E_TX_FLAGS_FD_SB BIT(9)
#define I40E_TX_FLAGS_UDP_TUNNEL BIT(10)
+#define I40E_TX_FLAGS_XDP BIT(11)
#define I40E_TX_FLAGS_VLAN_MASK 0xffff0000
#define I40E_TX_FLAGS_VLAN_PRIO_MASK 0xe0000000
#define I40E_TX_FLAGS_VLAN_PRIO_SHIFT 29
@@ -230,6 +231,7 @@ struct i40e_tx_buffer {
union {
struct sk_buff *skb;
void *raw_buf;
+ struct page *page;
};
unsigned int bytecount;
unsigned short gso_segs;
@@ -343,6 +345,9 @@ struct i40e_ring {
u16 next_to_alloc;
struct bpf_prog *xdp_prog;
+ struct i40e_ring *xdp_sibling; /* rx to xdp, and xdp to rx */
+ bool xdp_needs_tail_bump;
+ u16 curr_in_use;
} ____cacheline_internodealigned_in_smp;
enum i40e_latency_range {
--
2.9.3
^ permalink raw reply related [flat|nested] 7+ messages in thread