/* # of XDP queue pairs currently used by the driver */
u16 xdp_queue_pairs;
+ /* xdp_queue_pairs may be 0, when xdp is already loaded. So add this. */
+ bool xdp_enabled;
+
/* I like... big packets and I cannot lie! */
bool big_packets;
struct receive_queue *rq,
struct page *page, unsigned int offset,
unsigned int len, unsigned int truesize,
- bool hdr_valid, unsigned int metasize)
+ bool hdr_valid, unsigned int metasize,
+ unsigned int headroom)
{
struct sk_buff *skb;
struct virtio_net_hdr_mrg_rxbuf *hdr;
unsigned int copy, hdr_len, hdr_padded_len;
- char *p;
+ struct page *page_to_free = NULL;
+ int tailroom, shinfo_size;
+ char *p, *hdr_p, *buf;
p = page_address(page) + offset;
-
- /* copy small packet so we can reuse these pages for small data */
- skb = napi_alloc_skb(&rq->napi, GOOD_COPY_LEN);
- if (unlikely(!skb))
- return NULL;
-
- hdr = skb_vnet_hdr(skb);
+ hdr_p = p;
hdr_len = vi->hdr_len;
if (vi->mergeable_rx_bufs)
else
hdr_padded_len = sizeof(struct padded_vnet_hdr);
- /* hdr_valid means no XDP, so we can copy the vnet header */
- if (hdr_valid)
- memcpy(hdr, p, hdr_len);
+ /* If headroom is not 0, there is an offset between the beginning of the
+ * data and the allocated space, otherwise the data and the allocated
+ * space are aligned.
+ */
+ if (headroom) {
+ /* Buffers with headroom use PAGE_SIZE as alloc size,
+ * see add_recvbuf_mergeable() + get_mergeable_buf_len()
+ */
+ truesize = PAGE_SIZE;
+ tailroom = truesize - len - offset;
+ buf = page_address(page);
+ } else {
+ tailroom = truesize - len;
+ buf = p;
+ }
len -= hdr_len;
offset += hdr_padded_len;
p += hdr_padded_len;
+ shinfo_size = SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
+
+ /* copy small packet so we can reuse these pages */
+ if (!NET_IP_ALIGN && len > GOOD_COPY_LEN && tailroom >= shinfo_size) {
+ skb = build_skb(buf, truesize);
+ if (unlikely(!skb))
+ return NULL;
+
+ skb_reserve(skb, p - buf);
+ skb_put(skb, len);
+ goto ok;
+ }
+
+ /* copy small packet so we can reuse these pages for small data */
+ skb = napi_alloc_skb(&rq->napi, GOOD_COPY_LEN);
+ if (unlikely(!skb))
+ return NULL;
+
/* Copy all frame if it fits skb->head, otherwise
* we let virtio_net_hdr_to_skb() and GRO pull headers as needed.
*/
copy = ETH_HLEN + metasize;
skb_put_data(skb, p, copy);
- if (metasize) {
- __skb_pull(skb, metasize);
- skb_metadata_set(skb, metasize);
- }
-
len -= copy;
offset += copy;
if (len)
skb_add_rx_frag(skb, 0, page, offset, len, truesize);
else
- put_page(page);
- return skb;
+ page_to_free = page;
+ goto ok;
}
/*
if (page)
give_pages(rq, page);
+ok:
+ /* hdr_valid means no XDP, so we can copy the vnet header */
+ if (hdr_valid) {
+ hdr = skb_vnet_hdr(skb);
+ memcpy(hdr, hdr_p, hdr_len);
+ }
+ if (page_to_free)
+ put_page(page_to_free);
+
+ if (metasize) {
+ __skb_pull(skb, metasize);
+ skb_metadata_set(skb, metasize);
+ }
+
return skb;
}
return 0;
}
-static struct send_queue *virtnet_xdp_sq(struct virtnet_info *vi)
-{
- unsigned int qp;
-
- qp = vi->curr_queue_pairs - vi->xdp_queue_pairs + smp_processor_id();
- return &vi->sq[qp];
+/* when vi->curr_queue_pairs > nr_cpu_ids, the txq/sq is only used for xdp tx on
+ * the current cpu, so it does not need to be locked.
+ *
+ * Here we use marco instead of inline functions because we have to deal with
+ * three issues at the same time: 1. the choice of sq. 2. judge and execute the
+ * lock/unlock of txq 3. make sparse happy. It is difficult for two inline
+ * functions to perfectly solve these three problems at the same time.
+ */
+#define virtnet_xdp_get_sq(vi) ({ \
+ struct netdev_queue *txq; \
+ typeof(vi) v = (vi); \
+ unsigned int qp; \
+ \
+ if (v->curr_queue_pairs > nr_cpu_ids) { \
+ qp = v->curr_queue_pairs - v->xdp_queue_pairs; \
+ qp += smp_processor_id(); \
+ txq = netdev_get_tx_queue(v->dev, qp); \
+ __netif_tx_acquire(txq); \
+ } else { \
+ qp = smp_processor_id() % v->curr_queue_pairs; \
+ txq = netdev_get_tx_queue(v->dev, qp); \
+ __netif_tx_lock(txq, raw_smp_processor_id()); \
+ } \
+ v->sq + qp; \
+})
+
+#define virtnet_xdp_put_sq(vi, q) { \
+ struct netdev_queue *txq; \
+ typeof(vi) v = (vi); \
+ \
+ txq = netdev_get_tx_queue(v->dev, (q) - v->sq); \
+ if (v->curr_queue_pairs > nr_cpu_ids) \
+ __netif_tx_release(txq); \
+ else \
+ __netif_tx_unlock(txq); \
}
static int virtnet_xdp_xmit(struct net_device *dev,
unsigned int len;
int packets = 0;
int bytes = 0;
- int drops = 0;
+ int nxmit = 0;
int kicks = 0;
- int ret, err;
void *ptr;
+ int ret;
int i;
/* Only allow ndo_xdp_xmit if XDP is loaded on dev, as this
if (!xdp_prog)
return -ENXIO;
- sq = virtnet_xdp_sq(vi);
+ sq = virtnet_xdp_get_sq(vi);
if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) {
ret = -EINVAL;
- drops = n;
goto out;
}
for (i = 0; i < n; i++) {
struct xdp_frame *xdpf = frames[i];
- err = __virtnet_xdp_xmit_one(vi, sq, xdpf);
- if (err) {
- xdp_return_frame_rx_napi(xdpf);
- drops++;
- }
+ if (__virtnet_xdp_xmit_one(vi, sq, xdpf))
+ break;
+ nxmit++;
}
- ret = n - drops;
+ ret = nxmit;
if (flags & XDP_XMIT_FLUSH) {
if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq))
sq->stats.bytes += bytes;
sq->stats.packets += packets;
sq->stats.xdp_tx += n;
- sq->stats.xdp_tx_drops += drops;
+ sq->stats.xdp_tx_drops += n - nxmit;
sq->stats.kicks += kicks;
u64_stats_update_end(&sq->stats.syncp);
+ virtnet_xdp_put_sq(vi, sq);
return ret;
}
static unsigned int virtnet_get_headroom(struct virtnet_info *vi)
{
- return vi->xdp_queue_pairs ? VIRTIO_XDP_HEADROOM : 0;
+ return vi->xdp_enabled ? VIRTIO_XDP_HEADROOM : 0;
}
/* We copy the packet for XDP in the following cases:
if (unlikely(!xdpf))
goto err_xdp;
err = virtnet_xdp_xmit(dev, 1, &xdpf, 0);
- if (unlikely(err < 0)) {
+ if (unlikely(!err)) {
+ xdp_return_frame_rx_napi(xdpf);
+ } else if (unlikely(err < 0)) {
trace_xdp_exception(vi->dev, xdp_prog, act);
goto err_xdp;
}
{
struct page *page = buf;
struct sk_buff *skb =
- page_to_skb(vi, rq, page, 0, len, PAGE_SIZE, true, 0);
+ page_to_skb(vi, rq, page, 0, len, PAGE_SIZE, true, 0, 0);
stats->bytes += len - vi->hdr_len;
if (unlikely(!skb))
put_page(page);
head_skb = page_to_skb(vi, rq, xdp_page, offset,
len, PAGE_SIZE, false,
- metasize);
+ metasize, headroom);
return head_skb;
}
break;
if (unlikely(!xdpf))
goto err_xdp;
err = virtnet_xdp_xmit(dev, 1, &xdpf, 0);
- if (unlikely(err < 0)) {
+ if (unlikely(!err)) {
+ xdp_return_frame_rx_napi(xdpf);
+ } else if (unlikely(err < 0)) {
trace_xdp_exception(vi->dev, xdp_prog, act);
if (unlikely(xdp_page != page))
put_page(xdp_page);
}
head_skb = page_to_skb(vi, rq, page, offset, len, truesize, !xdp_prog,
- metasize);
+ metasize, headroom);
curr_skb = head_skb;
if (unlikely(!curr_skb))
xdp_do_flush();
if (xdp_xmit & VIRTIO_XDP_TX) {
- sq = virtnet_xdp_sq(vi);
+ sq = virtnet_xdp_get_sq(vi);
if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq)) {
u64_stats_update_begin(&sq->stats.syncp);
sq->stats.kicks++;
u64_stats_update_end(&sq->stats.syncp);
}
+ virtnet_xdp_put_sq(vi, sq);
}
return received;
}
virtqueue_set_affinity(vi->rq[i].vq, mask);
virtqueue_set_affinity(vi->sq[i].vq, mask);
- __netif_set_xps_queue(vi->dev, cpumask_bits(mask), i, false);
+ __netif_set_xps_queue(vi->dev, cpumask_bits(mask), i, XPS_CPUS);
cpumask_clear(mask);
}
static void virtnet_get_strings(struct net_device *dev, u32 stringset, u8 *data)
{
struct virtnet_info *vi = netdev_priv(dev);
- char *p = (char *)data;
unsigned int i, j;
+ u8 *p = data;
switch (stringset) {
case ETH_SS_STATS:
for (i = 0; i < vi->curr_queue_pairs; i++) {
- for (j = 0; j < VIRTNET_RQ_STATS_LEN; j++) {
- snprintf(p, ETH_GSTRING_LEN, "rx_queue_%u_%s",
- i, virtnet_rq_stats_desc[j].desc);
- p += ETH_GSTRING_LEN;
- }
+ for (j = 0; j < VIRTNET_RQ_STATS_LEN; j++)
+ ethtool_sprintf(&p, "rx_queue_%u_%s", i,
+ virtnet_rq_stats_desc[j].desc);
}
for (i = 0; i < vi->curr_queue_pairs; i++) {
- for (j = 0; j < VIRTNET_SQ_STATS_LEN; j++) {
- snprintf(p, ETH_GSTRING_LEN, "tx_queue_%u_%s",
- i, virtnet_sq_stats_desc[j].desc);
- p += ETH_GSTRING_LEN;
- }
+ for (j = 0; j < VIRTNET_SQ_STATS_LEN; j++)
+ ethtool_sprintf(&p, "tx_queue_%u_%s", i,
+ virtnet_sq_stats_desc[j].desc);
}
break;
}
/* XDP requires extra queues for XDP_TX */
if (curr_qp + xdp_qp > vi->max_queue_pairs) {
- NL_SET_ERR_MSG_MOD(extack, "Too few free TX rings available");
- netdev_warn(dev, "request %i queues but max is %i\n",
+ netdev_warn(dev, "XDP request %i queues but max is %i. XDP_TX and XDP_REDIRECT will operate in a slower locked tx mode.\n",
curr_qp + xdp_qp, vi->max_queue_pairs);
- return -ENOMEM;
+ xdp_qp = 0;
}
old_prog = rtnl_dereference(vi->rq[0].xdp_prog);
vi->xdp_queue_pairs = xdp_qp;
if (prog) {
+ vi->xdp_enabled = true;
for (i = 0; i < vi->max_queue_pairs; i++) {
rcu_assign_pointer(vi->rq[i].xdp_prog, prog);
if (i == 0 && !old_prog)
virtnet_clear_guest_offloads(vi);
}
+ } else {
+ vi->xdp_enabled = false;
}
for (i = 0; i < vi->max_queue_pairs; i++) {
int err;
if ((dev->features ^ features) & NETIF_F_LRO) {
- if (vi->xdp_queue_pairs)
+ if (vi->xdp_enabled)
return -EBUSY;
if (features & NETIF_F_LRO)
return -ENOMEM;
/* Set up network device as normal. */
- dev->priv_flags |= IFF_UNICAST_FLT | IFF_LIVE_ADDR_CHANGE;
+ dev->priv_flags |= IFF_UNICAST_FLT | IFF_LIVE_ADDR_CHANGE |
+ IFF_TX_SKB_NO_LINEAR;
dev->netdev_ops = &virtnet_netdev;
dev->features = NETIF_F_HIGHDMA;
projects / linux-block.git / blobdiff