From 9b2bbdb227588455afcc3b03475fa9b0a35d83af Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Mon, 6 Mar 2017 18:19:39 +0200 Subject: virtio: wrap find_vqs We are going to add more parameters to find_vqs, let's wrap the call so we don't need to tweak all drivers every time. Signed-off-by: Michael S. Tsirkin --- drivers/net/caif/caif_virtio.c | 3 +-- drivers/net/virtio_net.c | 3 +-- 2 files changed, 2 insertions(+), 4 deletions(-) (limited to 'drivers/net') diff --git a/drivers/net/caif/caif_virtio.c b/drivers/net/caif/caif_virtio.c index bc0eb47eccee..6122768c8644 100644 --- a/drivers/net/caif/caif_virtio.c +++ b/drivers/net/caif/caif_virtio.c @@ -679,8 +679,7 @@ static int cfv_probe(struct virtio_device *vdev) goto err; /* Get the TX virtio ring. This is a "guest side vring". */ - err = vdev->config->find_vqs(vdev, 1, &cfv->vq_tx, &vq_cbs, &names, - NULL); + err = virtio_find_vqs(vdev, 1, &cfv->vq_tx, &vq_cbs, &names, NULL); if (err) goto err; diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index f36584616e7d..71f447ab440e 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -2079,8 +2079,7 @@ static int virtnet_find_vqs(struct virtnet_info *vi) names[txq2vq(i)] = vi->sq[i].name; } - ret = vi->vdev->config->find_vqs(vi->vdev, total_vqs, vqs, callbacks, - names, NULL); + ret = virtio_find_vqs(vi->vdev, total_vqs, vqs, callbacks, names, NULL); if (ret) goto err_find; -- cgit v1.2.3 From d45b897b11eaf9847c4b9bc99a91279041b1bbf7 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Mon, 6 Mar 2017 20:31:21 +0200 Subject: virtio_net: allow specifying context for rx With mergeable buffers we never use s/g for rx, so allow specifying context in that case. Signed-off-by: Michael S. Tsirkin --- drivers/net/virtio_net.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'drivers/net') diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index 71f447ab440e..108f923f8a69 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -2044,6 +2044,7 @@ static int virtnet_find_vqs(struct virtnet_info *vi) int ret = -ENOMEM; int i, total_vqs; const char **names; + bool *ctx; /* We expect 1 RX virtqueue followed by 1 TX virtqueue, followed by * possible N-1 RX/TX queue pairs used in multiqueue mode, followed by @@ -2062,6 +2063,13 @@ static int virtnet_find_vqs(struct virtnet_info *vi) names = kmalloc(total_vqs * sizeof(*names), GFP_KERNEL); if (!names) goto err_names; + if (vi->mergeable_rx_bufs) { + ctx = kzalloc(total_vqs * sizeof(*ctx), GFP_KERNEL); + if (!ctx) + goto err_ctx; + } else { + ctx = NULL; + } /* Parameters for control virtqueue, if any */ if (vi->has_cvq) { @@ -2077,9 +2085,12 @@ static int virtnet_find_vqs(struct virtnet_info *vi) sprintf(vi->sq[i].name, "output.%d", i); names[rxq2vq(i)] = vi->rq[i].name; names[txq2vq(i)] = vi->sq[i].name; + if (ctx) + ctx[rxq2vq(i)] = true; } - ret = virtio_find_vqs(vi->vdev, total_vqs, vqs, callbacks, names, NULL); + ret = vi->vdev->config->find_vqs(vi->vdev, total_vqs, vqs, callbacks, + names, ctx, NULL); if (ret) goto err_find; @@ -2101,6 +2112,8 @@ static int virtnet_find_vqs(struct virtnet_info *vi) return 0; err_find: + kfree(ctx); +err_ctx: kfree(names); err_names: kfree(callbacks); -- cgit v1.2.3 From 680557cf79f82623e2c4fd42733077d60a843513 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Mon, 6 Mar 2017 21:29:47 +0200 Subject: virtio_net: rework mergeable buffer handling Use the new _ctx virtio API to maintain true length for each buffer. Signed-off-by: Michael S. Tsirkin --- drivers/net/virtio_net.c | 89 +++++++++++++++++++++++------------------------- 1 file changed, 43 insertions(+), 46 deletions(-) (limited to 'drivers/net') diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index 108f923f8a69..2170a5caf68d 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -250,24 +250,6 @@ static void skb_xmit_done(struct virtqueue *vq) netif_wake_subqueue(vi->dev, vq2txq(vq)); } -static unsigned int mergeable_ctx_to_buf_truesize(unsigned long mrg_ctx) -{ - unsigned int truesize = mrg_ctx & (MERGEABLE_BUFFER_ALIGN - 1); - return (truesize + 1) * MERGEABLE_BUFFER_ALIGN; -} - -static void *mergeable_ctx_to_buf_address(unsigned long mrg_ctx) -{ - return (void *)(mrg_ctx & -MERGEABLE_BUFFER_ALIGN); - -} - -static unsigned long mergeable_buf_to_ctx(void *buf, unsigned int truesize) -{ - unsigned int size = truesize / MERGEABLE_BUFFER_ALIGN; - return (unsigned long)buf | (size - 1); -} - /* Called from bottom half context */ static struct sk_buff *page_to_skb(struct virtnet_info *vi, struct receive_queue *rq, @@ -511,15 +493,13 @@ static struct page *xdp_linearize_page(struct receive_queue *rq, while (--*num_buf) { unsigned int buflen; - unsigned long ctx; void *buf; int off; - ctx = (unsigned long)virtqueue_get_buf(rq->vq, &buflen); - if (unlikely(!ctx)) + buf = virtqueue_get_buf(rq->vq, &buflen); + if (unlikely(!buf)) goto err_buf; - buf = mergeable_ctx_to_buf_address(ctx); p = virt_to_head_page(buf); off = buf - page_address(p); @@ -548,10 +528,10 @@ err_buf: static struct sk_buff *receive_mergeable(struct net_device *dev, struct virtnet_info *vi, struct receive_queue *rq, - unsigned long ctx, + void *buf, + void *ctx, unsigned int len) { - void *buf = mergeable_ctx_to_buf_address(ctx); struct virtio_net_hdr_mrg_rxbuf *hdr = buf; u16 num_buf = virtio16_to_cpu(vi->vdev, hdr->num_buffers); struct page *page = virt_to_head_page(buf); @@ -639,7 +619,13 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, } rcu_read_unlock(); - truesize = max(len, mergeable_ctx_to_buf_truesize(ctx)); + if (unlikely(len > (unsigned long)ctx)) { + pr_debug("%s: rx error: len %u exceeds truesize 0x%lu\n", + dev->name, len, (unsigned long)ctx); + dev->stats.rx_length_errors++; + goto err_skb; + } + truesize = (unsigned long)ctx; head_skb = page_to_skb(vi, rq, page, offset, len, truesize); curr_skb = head_skb; @@ -648,7 +634,7 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, while (--num_buf) { int num_skb_frags; - ctx = (unsigned long)virtqueue_get_buf(rq->vq, &len); + buf = virtqueue_get_buf_ctx(rq->vq, &len, &ctx); if (unlikely(!ctx)) { pr_debug("%s: rx error: %d buffers out of %d missing\n", dev->name, num_buf, @@ -658,8 +644,14 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, goto err_buf; } - buf = mergeable_ctx_to_buf_address(ctx); page = virt_to_head_page(buf); + if (unlikely(len > (unsigned long)ctx)) { + pr_debug("%s: rx error: len %u exceeds truesize 0x%lu\n", + dev->name, len, (unsigned long)ctx); + dev->stats.rx_length_errors++; + goto err_skb; + } + truesize = (unsigned long)ctx; num_skb_frags = skb_shinfo(curr_skb)->nr_frags; if (unlikely(num_skb_frags == MAX_SKB_FRAGS)) { @@ -675,7 +667,6 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, head_skb->truesize += nskb->truesize; num_skb_frags = 0; } - truesize = max(len, mergeable_ctx_to_buf_truesize(ctx)); if (curr_skb != head_skb) { head_skb->data_len += len; head_skb->len += len; @@ -700,14 +691,14 @@ err_xdp: err_skb: put_page(page); while (--num_buf) { - ctx = (unsigned long)virtqueue_get_buf(rq->vq, &len); - if (unlikely(!ctx)) { + buf = virtqueue_get_buf(rq->vq, &len); + if (unlikely(!buf)) { pr_debug("%s: rx error: %d buffers missing\n", dev->name, num_buf); dev->stats.rx_length_errors++; break; } - page = virt_to_head_page(mergeable_ctx_to_buf_address(ctx)); + page = virt_to_head_page(buf); put_page(page); } err_buf: @@ -718,7 +709,7 @@ xdp_xmit: } static int receive_buf(struct virtnet_info *vi, struct receive_queue *rq, - void *buf, unsigned int len) + void *buf, unsigned int len, void **ctx) { struct net_device *dev = vi->dev; struct sk_buff *skb; @@ -729,9 +720,7 @@ static int receive_buf(struct virtnet_info *vi, struct receive_queue *rq, pr_debug("%s: short packet %i\n", dev->name, len); dev->stats.rx_length_errors++; if (vi->mergeable_rx_bufs) { - unsigned long ctx = (unsigned long)buf; - void *base = mergeable_ctx_to_buf_address(ctx); - put_page(virt_to_head_page(base)); + put_page(virt_to_head_page(buf)); } else if (vi->big_packets) { give_pages(rq, buf); } else { @@ -741,7 +730,7 @@ static int receive_buf(struct virtnet_info *vi, struct receive_queue *rq, } if (vi->mergeable_rx_bufs) - skb = receive_mergeable(dev, vi, rq, (unsigned long)buf, len); + skb = receive_mergeable(dev, vi, rq, buf, ctx, len); else if (vi->big_packets) skb = receive_big(dev, vi, rq, buf, len); else @@ -869,7 +858,7 @@ static int add_recvbuf_mergeable(struct virtnet_info *vi, struct page_frag *alloc_frag = &rq->alloc_frag; unsigned int headroom = virtnet_get_headroom(vi); char *buf; - unsigned long ctx; + void *ctx; int err; unsigned int len, hole; @@ -879,7 +868,7 @@ static int add_recvbuf_mergeable(struct virtnet_info *vi, buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; buf += headroom; /* advance address leaving hole at front of pkt */ - ctx = mergeable_buf_to_ctx(buf, len); + ctx = (void *)(unsigned long)len; get_page(alloc_frag->page); alloc_frag->offset += len + headroom; hole = alloc_frag->size - alloc_frag->offset; @@ -894,7 +883,7 @@ static int add_recvbuf_mergeable(struct virtnet_info *vi, } sg_init_one(rq->sg, buf, len); - err = virtqueue_add_inbuf(rq->vq, rq->sg, 1, (void *)ctx, gfp); + err = virtqueue_add_inbuf_ctx(rq->vq, rq->sg, 1, buf, ctx, gfp); if (err < 0) put_page(virt_to_head_page(buf)); @@ -988,10 +977,20 @@ static int virtnet_receive(struct receive_queue *rq, int budget) void *buf; struct virtnet_stats *stats = this_cpu_ptr(vi->stats); - while (received < budget && - (buf = virtqueue_get_buf(rq->vq, &len)) != NULL) { - bytes += receive_buf(vi, rq, buf, len); - received++; + if (vi->mergeable_rx_bufs) { + void *ctx; + + while (received < budget && + (buf = virtqueue_get_buf_ctx(rq->vq, &len, &ctx))) { + bytes += receive_buf(vi, rq, buf, len, ctx); + received++; + } + } else { + while (received < budget && + (buf = virtqueue_get_buf(rq->vq, &len)) != NULL) { + bytes += receive_buf(vi, rq, buf, len, NULL); + received++; + } } if (rq->vq->num_free > virtqueue_get_vring_size(rq->vq) / 2) { @@ -2014,9 +2013,7 @@ static void free_unused_bufs(struct virtnet_info *vi) while ((buf = virtqueue_detach_unused_buf(vq)) != NULL) { if (vi->mergeable_rx_bufs) { - unsigned long ctx = (unsigned long)buf; - void *base = mergeable_ctx_to_buf_address(ctx); - put_page(virt_to_head_page(base)); + put_page(virt_to_head_page(buf)); } else if (vi->big_packets) { give_pages(&vi->rq[i], buf); } else { -- cgit v1.2.3 From e377fcc8486d40867c6c217077ad0fa40977e060 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Mon, 6 Mar 2017 22:21:35 +0200 Subject: virtio_net: reduce alignment for buffers We don't need to align length to any particular value anymore. Aligning to L1 cache size probably sill makes sense to reduce false sharing. Signed-off-by: Michael S. Tsirkin --- drivers/net/virtio_net.c | 13 +------------ 1 file changed, 1 insertion(+), 12 deletions(-) (limited to 'drivers/net') diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index 2170a5caf68d..6e47e653a05d 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -53,17 +53,6 @@ module_param(gso, bool, 0444); */ DECLARE_EWMA(pkt_len, 0, 64) -/* With mergeable buffers we align buffer address and use the low bits to - * encode its true size. Buffer size is up to 1 page so we need to align to - * square root of page size to ensure we reserve enough bits to encode the true - * size. - */ -#define MERGEABLE_BUFFER_MIN_ALIGN_SHIFT ((PAGE_SHIFT + 1) / 2) - -/* Minimum alignment for mergeable packet buffers. */ -#define MERGEABLE_BUFFER_ALIGN max(L1_CACHE_BYTES, \ - 1 << MERGEABLE_BUFFER_MIN_ALIGN_SHIFT) - #define VIRTNET_DRIVER_VERSION "1.0.0" struct virtnet_stats { @@ -849,7 +838,7 @@ static unsigned int get_mergeable_buf_len(struct ewma_pkt_len *avg_pkt_len) len = hdr_len + clamp_t(unsigned int, ewma_pkt_len_read(avg_pkt_len), GOOD_PACKET_LEN, PAGE_SIZE - hdr_len); - return ALIGN(len, MERGEABLE_BUFFER_ALIGN); + return ALIGN(len, L1_CACHE_BYTES); } static int add_recvbuf_mergeable(struct virtnet_info *vi, -- cgit v1.2.3 From d85b758f72b05a774045545f24d70980e3e9aac4 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Thu, 9 Mar 2017 02:21:21 +0200 Subject: virtio_net: fix support for small rings When ring size is small (<32 entries) making buffers smaller means a full ring might not be able to hold enough buffers to fit a single large packet. Make sure a ring full of buffers is large enough to allow at least one packet of max size. Fixes: 2613af0ed18a ("virtio_net: migrate mergeable rx buffers to page frag allocators") Signed-off-by: Michael S. Tsirkin --- drivers/net/virtio_net.c | 30 ++++++++++++++++++++++++++---- 1 file changed, 26 insertions(+), 4 deletions(-) (limited to 'drivers/net') diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index 6e47e653a05d..f9381e1f365f 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -29,6 +29,7 @@ #include #include #include +#include static int napi_weight = NAPI_POLL_WEIGHT; module_param(napi_weight, int, 0444); @@ -98,6 +99,9 @@ struct receive_queue { /* RX: fragments + linear part + virtio header */ struct scatterlist sg[MAX_SKB_FRAGS + 2]; + /* Min single buffer size for mergeable buffers case. */ + unsigned int min_buf_len; + /* Name of this receive queue: input.$index */ char name[40]; }; @@ -831,13 +835,14 @@ static int add_recvbuf_big(struct virtnet_info *vi, struct receive_queue *rq, return err; } -static unsigned int get_mergeable_buf_len(struct ewma_pkt_len *avg_pkt_len) +static unsigned int get_mergeable_buf_len(struct receive_queue *rq, + struct ewma_pkt_len *avg_pkt_len) { const size_t hdr_len = sizeof(struct virtio_net_hdr_mrg_rxbuf); unsigned int len; len = hdr_len + clamp_t(unsigned int, ewma_pkt_len_read(avg_pkt_len), - GOOD_PACKET_LEN, PAGE_SIZE - hdr_len); + rq->min_buf_len - hdr_len, PAGE_SIZE - hdr_len); return ALIGN(len, L1_CACHE_BYTES); } @@ -851,7 +856,7 @@ static int add_recvbuf_mergeable(struct virtnet_info *vi, int err; unsigned int len, hole; - len = get_mergeable_buf_len(&rq->mrg_avg_pkt_len); + len = get_mergeable_buf_len(rq, &rq->mrg_avg_pkt_len); if (unlikely(!skb_page_frag_refill(len + headroom, alloc_frag, gfp))) return -ENOMEM; @@ -2023,6 +2028,21 @@ static void virtnet_del_vqs(struct virtnet_info *vi) virtnet_free_queues(vi); } +/* How large should a single buffer be so a queue full of these can fit at + * least one full packet? + * Logic below assumes the mergeable buffer header is used. + */ +static unsigned int mergeable_min_buf_len(struct virtnet_info *vi, struct virtqueue *vq) +{ + const unsigned int hdr_len = sizeof(struct virtio_net_hdr_mrg_rxbuf); + unsigned int rq_size = virtqueue_get_vring_size(vq); + unsigned int packet_len = vi->big_packets ? IP_MAX_MTU : vi->dev->max_mtu; + unsigned int buf_len = hdr_len + ETH_HLEN + VLAN_HLEN + packet_len; + unsigned int min_buf_len = DIV_ROUND_UP(buf_len, rq_size); + + return max(min_buf_len, hdr_len); +} + static int virtnet_find_vqs(struct virtnet_info *vi) { vq_callback_t **callbacks; @@ -2088,6 +2108,7 @@ static int virtnet_find_vqs(struct virtnet_info *vi) for (i = 0; i < vi->max_queue_pairs; i++) { vi->rq[i].vq = vqs[rxq2vq(i)]; + vi->rq[i].min_buf_len = mergeable_min_buf_len(vi, vi->rq[i].vq); vi->sq[i].vq = vqs[txq2vq(i)]; } @@ -2174,7 +2195,8 @@ static ssize_t mergeable_rx_buffer_size_show(struct netdev_rx_queue *queue, BUG_ON(queue_index >= vi->max_queue_pairs); avg = &vi->rq[queue_index].mrg_avg_pkt_len; - return sprintf(buf, "%u\n", get_mergeable_buf_len(avg)); + return sprintf(buf, "%u\n", + get_mergeable_buf_len(&vi->rq[queue_index], avg)); } static struct rx_queue_attribute mergeable_rx_buffer_size_attribute = -- cgit v1.2.3 From 5f24df0945b3ce3be71602c2655f689d42a24181 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Wed, 29 Mar 2017 23:12:23 +0300 Subject: virtio_net: don't reset twice on XDP on/off We already do a reset once in remove_vq_common - there appears to be no point in doing another one when we add/remove XDP. Signed-off-by: Michael S. Tsirkin --- drivers/net/virtio_net.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/net') diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index f9381e1f365f..fd5145bdfb77 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -1747,7 +1747,6 @@ static int virtnet_reset(struct virtnet_info *vi, int curr_qp, int xdp_qp) virtnet_freeze_down(dev); _remove_vq_common(vi); - dev->config->reset(dev); virtio_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE); virtio_add_status(dev, VIRTIO_CONFIG_S_DRIVER); -- cgit v1.2.3 From 56da5fd04e3d51d0d876c685b0107ca357b9fd9d Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Thu, 6 Apr 2017 12:04:47 +0300 Subject: virtio_net: tidy a couple debug statements We are printing a decimal value for truesize so we shouldn't use an "0x" prefix. Signed-off-by: Dan Carpenter Signed-off-by: Michael S. Tsirkin --- drivers/net/virtio_net.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/net') diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index fd5145bdfb77..87b5c2046341 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -613,7 +613,7 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, rcu_read_unlock(); if (unlikely(len > (unsigned long)ctx)) { - pr_debug("%s: rx error: len %u exceeds truesize 0x%lu\n", + pr_debug("%s: rx error: len %u exceeds truesize %lu\n", dev->name, len, (unsigned long)ctx); dev->stats.rx_length_errors++; goto err_skb; @@ -639,7 +639,7 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, page = virt_to_head_page(buf); if (unlikely(len > (unsigned long)ctx)) { - pr_debug("%s: rx error: len %u exceeds truesize 0x%lu\n", + pr_debug("%s: rx error: len %u exceeds truesize %lu\n", dev->name, len, (unsigned long)ctx); dev->stats.rx_length_errors++; goto err_skb; -- cgit v1.2.3