diff options
Diffstat (limited to 'drivers/net/xen-netfront.c')
-rw-r--r-- | drivers/net/xen-netfront.c | 272 |
1 files changed, 169 insertions, 103 deletions
diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c index 44275908d61a..e31b98403f31 100644 --- a/drivers/net/xen-netfront.c +++ b/drivers/net/xen-netfront.c @@ -126,21 +126,17 @@ struct netfront_queue { /* * {tx,rx}_skbs store outstanding skbuffs. Free tx_skb entries - * are linked from tx_skb_freelist through skb_entry.link. - * - * NB. Freelist index entries are always going to be less than - * PAGE_OFFSET, whereas pointers to skbs will always be equal or - * greater than PAGE_OFFSET: we use this property to distinguish - * them. + * are linked from tx_skb_freelist through tx_link. */ - union skb_entry { - struct sk_buff *skb; - unsigned long link; - } tx_skbs[NET_TX_RING_SIZE]; + struct sk_buff *tx_skbs[NET_TX_RING_SIZE]; + unsigned short tx_link[NET_TX_RING_SIZE]; +#define TX_LINK_NONE 0xffff +#define TX_PENDING 0xfffe grant_ref_t gref_tx_head; grant_ref_t grant_tx_ref[NET_TX_RING_SIZE]; struct page *grant_tx_page[NET_TX_RING_SIZE]; unsigned tx_skb_freelist; + unsigned int tx_pend_queue; spinlock_t rx_lock ____cacheline_aligned_in_smp; struct xen_netif_rx_front_ring rx; @@ -173,6 +169,9 @@ struct netfront_info { bool netback_has_xdp_headroom; bool netfront_xdp_enabled; + /* Is device behaving sane? */ + bool broken; + atomic_t rx_gso_checksum_fixup; }; @@ -181,33 +180,25 @@ struct netfront_rx_info { struct xen_netif_extra_info extras[XEN_NETIF_EXTRA_TYPE_MAX - 1]; }; -static void skb_entry_set_link(union skb_entry *list, unsigned short id) -{ - list->link = id; -} - -static int skb_entry_is_link(const union skb_entry *list) -{ - BUILD_BUG_ON(sizeof(list->skb) != sizeof(list->link)); - return (unsigned long)list->skb < PAGE_OFFSET; -} - /* * Access macros for acquiring freeing slots in tx_skbs[]. */ -static void add_id_to_freelist(unsigned *head, union skb_entry *list, - unsigned short id) +static void add_id_to_list(unsigned *head, unsigned short *list, + unsigned short id) { - skb_entry_set_link(&list[id], *head); + list[id] = *head; *head = id; } -static unsigned short get_id_from_freelist(unsigned *head, - union skb_entry *list) +static unsigned short get_id_from_list(unsigned *head, unsigned short *list) { unsigned int id = *head; - *head = list[id].link; + + if (id != TX_LINK_NONE) { + *head = list[id]; + list[id] = TX_LINK_NONE; + } return id; } @@ -363,7 +354,7 @@ static int xennet_open(struct net_device *dev) unsigned int i = 0; struct netfront_queue *queue = NULL; - if (!np->queues) + if (!np->queues || np->broken) return -ENODEV; for (i = 0; i < num_queues; ++i) { @@ -391,27 +382,47 @@ static void xennet_tx_buf_gc(struct netfront_queue *queue) unsigned short id; struct sk_buff *skb; bool more_to_do; + const struct device *dev = &queue->info->netdev->dev; BUG_ON(!netif_carrier_ok(queue->info->netdev)); do { prod = queue->tx.sring->rsp_prod; + if (RING_RESPONSE_PROD_OVERFLOW(&queue->tx, prod)) { + dev_alert(dev, "Illegal number of responses %u\n", + prod - queue->tx.rsp_cons); + goto err; + } rmb(); /* Ensure we see responses up to 'rp'. */ for (cons = queue->tx.rsp_cons; cons != prod; cons++) { - struct xen_netif_tx_response *txrsp; + struct xen_netif_tx_response txrsp; - txrsp = RING_GET_RESPONSE(&queue->tx, cons); - if (txrsp->status == XEN_NETIF_RSP_NULL) + RING_COPY_RESPONSE(&queue->tx, cons, &txrsp); + if (txrsp.status == XEN_NETIF_RSP_NULL) continue; - id = txrsp->id; - skb = queue->tx_skbs[id].skb; + id = txrsp.id; + if (id >= RING_SIZE(&queue->tx)) { + dev_alert(dev, + "Response has incorrect id (%u)\n", + id); + goto err; + } + if (queue->tx_link[id] != TX_PENDING) { + dev_alert(dev, + "Response for inactive request\n"); + goto err; + } + + queue->tx_link[id] = TX_LINK_NONE; + skb = queue->tx_skbs[id]; + queue->tx_skbs[id] = NULL; if (unlikely(gnttab_query_foreign_access( queue->grant_tx_ref[id]) != 0)) { - pr_alert("%s: warning -- grant still in use by backend domain\n", - __func__); - BUG(); + dev_alert(dev, + "Grant still in use by backend domain\n"); + goto err; } gnttab_end_foreign_access_ref( queue->grant_tx_ref[id], GNTMAP_readonly); @@ -419,7 +430,7 @@ static void xennet_tx_buf_gc(struct netfront_queue *queue) &queue->gref_tx_head, queue->grant_tx_ref[id]); queue->grant_tx_ref[id] = GRANT_INVALID_REF; queue->grant_tx_page[id] = NULL; - add_id_to_freelist(&queue->tx_skb_freelist, queue->tx_skbs, id); + add_id_to_list(&queue->tx_skb_freelist, queue->tx_link, id); dev_kfree_skb_irq(skb); } @@ -429,13 +440,20 @@ static void xennet_tx_buf_gc(struct netfront_queue *queue) } while (more_to_do); xennet_maybe_wake_tx(queue); + + return; + + err: + queue->info->broken = true; + dev_alert(dev, "Disabled for further use\n"); } struct xennet_gnttab_make_txreq { struct netfront_queue *queue; struct sk_buff *skb; struct page *page; - struct xen_netif_tx_request *tx; /* Last request */ + struct xen_netif_tx_request *tx; /* Last request on ring page */ + struct xen_netif_tx_request tx_local; /* Last request local copy*/ unsigned int size; }; @@ -451,7 +469,7 @@ static void xennet_tx_setup_grant(unsigned long gfn, unsigned int offset, struct netfront_queue *queue = info->queue; struct sk_buff *skb = info->skb; - id = get_id_from_freelist(&queue->tx_skb_freelist, queue->tx_skbs); + id = get_id_from_list(&queue->tx_skb_freelist, queue->tx_link); tx = RING_GET_REQUEST(&queue->tx, queue->tx.req_prod_pvt++); ref = gnttab_claim_grant_reference(&queue->gref_tx_head); WARN_ON_ONCE(IS_ERR_VALUE((unsigned long)(int)ref)); @@ -459,34 +477,37 @@ static void xennet_tx_setup_grant(unsigned long gfn, unsigned int offset, gnttab_grant_foreign_access_ref(ref, queue->info->xbdev->otherend_id, gfn, GNTMAP_readonly); - queue->tx_skbs[id].skb = skb; + queue->tx_skbs[id] = skb; queue->grant_tx_page[id] = page; queue->grant_tx_ref[id] = ref; - tx->id = id; - tx->gref = ref; - tx->offset = offset; - tx->size = len; - tx->flags = 0; + info->tx_local.id = id; + info->tx_local.gref = ref; + info->tx_local.offset = offset; + info->tx_local.size = len; + info->tx_local.flags = 0; + + *tx = info->tx_local; + + /* + * Put the request in the pending queue, it will be set to be pending + * when the producer index is about to be raised. + */ + add_id_to_list(&queue->tx_pend_queue, queue->tx_link, id); info->tx = tx; - info->size += tx->size; + info->size += info->tx_local.size; } static struct xen_netif_tx_request *xennet_make_first_txreq( - struct netfront_queue *queue, struct sk_buff *skb, - struct page *page, unsigned int offset, unsigned int len) + struct xennet_gnttab_make_txreq *info, + unsigned int offset, unsigned int len) { - struct xennet_gnttab_make_txreq info = { - .queue = queue, - .skb = skb, - .page = page, - .size = 0, - }; + info->size = 0; - gnttab_for_one_grant(page, offset, len, xennet_tx_setup_grant, &info); + gnttab_for_one_grant(info->page, offset, len, xennet_tx_setup_grant, info); - return info.tx; + return info->tx; } static void xennet_make_one_txreq(unsigned long gfn, unsigned int offset, @@ -499,35 +520,27 @@ static void xennet_make_one_txreq(unsigned long gfn, unsigned int offset, xennet_tx_setup_grant(gfn, offset, len, data); } -static struct xen_netif_tx_request *xennet_make_txreqs( - struct netfront_queue *queue, struct xen_netif_tx_request *tx, - struct sk_buff *skb, struct page *page, +static void xennet_make_txreqs( + struct xennet_gnttab_make_txreq *info, + struct page *page, unsigned int offset, unsigned int len) { - struct xennet_gnttab_make_txreq info = { - .queue = queue, - .skb = skb, - .tx = tx, - }; - /* Skip unused frames from start of page */ page += offset >> PAGE_SHIFT; offset &= ~PAGE_MASK; while (len) { - info.page = page; - info.size = 0; + info->page = page; + info->size = 0; gnttab_foreach_grant_in_range(page, offset, len, xennet_make_one_txreq, - &info); + info); page++; offset = 0; - len -= info.size; + len -= info->size; } - - return info.tx; } /* @@ -574,19 +587,34 @@ static u16 xennet_select_queue(struct net_device *dev, struct sk_buff *skb, return queue_idx; } +static void xennet_mark_tx_pending(struct netfront_queue *queue) +{ + unsigned int i; + + while ((i = get_id_from_list(&queue->tx_pend_queue, queue->tx_link)) != + TX_LINK_NONE) + queue->tx_link[i] = TX_PENDING; +} + static int xennet_xdp_xmit_one(struct net_device *dev, struct netfront_queue *queue, struct xdp_frame *xdpf) { struct netfront_info *np = netdev_priv(dev); struct netfront_stats *tx_stats = this_cpu_ptr(np->tx_stats); + struct xennet_gnttab_make_txreq info = { + .queue = queue, + .skb = NULL, + .page = virt_to_page(xdpf->data), + }; int notify; - xennet_make_first_txreq(queue, NULL, - virt_to_page(xdpf->data), + xennet_make_first_txreq(&info, offset_in_page(xdpf->data), xdpf->len); + xennet_mark_tx_pending(queue); + RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&queue->tx, notify); if (notify) notify_remote_via_irq(queue->tx_irq); @@ -611,6 +639,8 @@ static int xennet_xdp_xmit(struct net_device *dev, int n, int nxmit = 0; int i; + if (unlikely(np->broken)) + return -ENODEV; if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) return -EINVAL; @@ -638,7 +668,7 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev { struct netfront_info *np = netdev_priv(dev); struct netfront_stats *tx_stats = this_cpu_ptr(np->tx_stats); - struct xen_netif_tx_request *tx, *first_tx; + struct xen_netif_tx_request *first_tx; unsigned int i; int notify; int slots; @@ -647,6 +677,7 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev unsigned int len; unsigned long flags; struct netfront_queue *queue = NULL; + struct xennet_gnttab_make_txreq info = { }; unsigned int num_queues = dev->real_num_tx_queues; u16 queue_index; struct sk_buff *nskb; @@ -654,6 +685,8 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev /* Drop the packet if no queues are set up */ if (num_queues < 1) goto drop; + if (unlikely(np->broken)) + goto drop; /* Determine which queue to transmit this SKB on */ queue_index = skb_get_queue_mapping(skb); queue = &np->queues[queue_index]; @@ -704,21 +737,24 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev } /* First request for the linear area. */ - first_tx = tx = xennet_make_first_txreq(queue, skb, - page, offset, len); - offset += tx->size; + info.queue = queue; + info.skb = skb; + info.page = page; + first_tx = xennet_make_first_txreq(&info, offset, len); + offset += info.tx_local.size; if (offset == PAGE_SIZE) { page++; offset = 0; } - len -= tx->size; + len -= info.tx_local.size; if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */ - tx->flags |= XEN_NETTXF_csum_blank | XEN_NETTXF_data_validated; + first_tx->flags |= XEN_NETTXF_csum_blank | + XEN_NETTXF_data_validated; else if (skb->ip_summed == CHECKSUM_UNNECESSARY) /* remote but checksummed. */ - tx->flags |= XEN_NETTXF_data_validated; + first_tx->flags |= XEN_NETTXF_data_validated; /* Optional extra info after the first request. */ if (skb_shinfo(skb)->gso_size) { @@ -727,7 +763,7 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev gso = (struct xen_netif_extra_info *) RING_GET_REQUEST(&queue->tx, queue->tx.req_prod_pvt++); - tx->flags |= XEN_NETTXF_extra_info; + first_tx->flags |= XEN_NETTXF_extra_info; gso->u.gso.size = skb_shinfo(skb)->gso_size; gso->u.gso.type = (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6) ? @@ -741,12 +777,12 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev } /* Requests for the rest of the linear area. */ - tx = xennet_make_txreqs(queue, tx, skb, page, offset, len); + xennet_make_txreqs(&info, page, offset, len); /* Requests for all the frags. */ for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) { skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; - tx = xennet_make_txreqs(queue, tx, skb, skb_frag_page(frag), + xennet_make_txreqs(&info, skb_frag_page(frag), skb_frag_off(frag), skb_frag_size(frag)); } @@ -757,6 +793,8 @@ static netdev_tx_t xennet_start_xmit(struct sk_buff *skb, struct net_device *dev /* timestamp packet in software */ skb_tx_timestamp(skb); + xennet_mark_tx_pending(queue); + RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&queue->tx, notify); if (notify) notify_remote_via_irq(queue->tx_irq); @@ -814,7 +852,7 @@ static int xennet_get_extras(struct netfront_queue *queue, RING_IDX rp) { - struct xen_netif_extra_info *extra; + struct xen_netif_extra_info extra; struct device *dev = &queue->info->netdev->dev; RING_IDX cons = queue->rx.rsp_cons; int err = 0; @@ -830,24 +868,22 @@ static int xennet_get_extras(struct netfront_queue *queue, break; } - extra = (struct xen_netif_extra_info *) - RING_GET_RESPONSE(&queue->rx, ++cons); + RING_COPY_RESPONSE(&queue->rx, ++cons, &extra); - if (unlikely(!extra->type || - extra->type >= XEN_NETIF_EXTRA_TYPE_MAX)) { + if (unlikely(!extra.type || + extra.type >= XEN_NETIF_EXTRA_TYPE_MAX)) { if (net_ratelimit()) dev_warn(dev, "Invalid extra type: %d\n", - extra->type); + extra.type); err = -EINVAL; } else { - memcpy(&extras[extra->type - 1], extra, - sizeof(*extra)); + extras[extra.type - 1] = extra; } skb = xennet_get_rx_skb(queue, cons); ref = xennet_get_rx_ref(queue, cons); xennet_move_rx_slot(queue, skb, ref); - } while (extra->flags & XEN_NETIF_EXTRA_FLAG_MORE); + } while (extra.flags & XEN_NETIF_EXTRA_FLAG_MORE); queue->rx.rsp_cons = cons; return err; @@ -905,7 +941,7 @@ static int xennet_get_responses(struct netfront_queue *queue, struct sk_buff_head *list, bool *need_xdp_flush) { - struct xen_netif_rx_response *rx = &rinfo->rx; + struct xen_netif_rx_response *rx = &rinfo->rx, rx_local; int max = XEN_NETIF_NR_SLOTS_MIN + (rx->status <= RX_COPY_THRESHOLD); RING_IDX cons = queue->rx.rsp_cons; struct sk_buff *skb = xennet_get_rx_skb(queue, cons); @@ -989,7 +1025,8 @@ next: break; } - rx = RING_GET_RESPONSE(&queue->rx, cons + slots); + RING_COPY_RESPONSE(&queue->rx, cons + slots, &rx_local); + rx = &rx_local; skb = xennet_get_rx_skb(queue, cons + slots); ref = xennet_get_rx_ref(queue, cons + slots); slots++; @@ -1044,10 +1081,11 @@ static int xennet_fill_frags(struct netfront_queue *queue, struct sk_buff *nskb; while ((nskb = __skb_dequeue(list))) { - struct xen_netif_rx_response *rx = - RING_GET_RESPONSE(&queue->rx, ++cons); + struct xen_netif_rx_response rx; skb_frag_t *nfrag = &skb_shinfo(nskb)->frags[0]; + RING_COPY_RESPONSE(&queue->rx, ++cons, &rx); + if (skb_shinfo(skb)->nr_frags == MAX_SKB_FRAGS) { unsigned int pull_to = NETFRONT_SKB_CB(skb)->pull_to; @@ -1062,7 +1100,7 @@ static int xennet_fill_frags(struct netfront_queue *queue, skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, skb_frag_page(nfrag), - rx->offset, rx->status, PAGE_SIZE); + rx.offset, rx.status, PAGE_SIZE); skb_shinfo(nskb)->nr_frags = 0; kfree_skb(nskb); @@ -1156,12 +1194,19 @@ static int xennet_poll(struct napi_struct *napi, int budget) skb_queue_head_init(&tmpq); rp = queue->rx.sring->rsp_prod; + if (RING_RESPONSE_PROD_OVERFLOW(&queue->rx, rp)) { + dev_alert(&dev->dev, "Illegal number of responses %u\n", + rp - queue->rx.rsp_cons); + queue->info->broken = true; + spin_unlock(&queue->rx_lock); + return 0; + } rmb(); /* Ensure we see queued responses up to 'rp'. */ i = queue->rx.rsp_cons; work_done = 0; while ((i != rp) && (work_done < budget)) { - memcpy(rx, RING_GET_RESPONSE(&queue->rx, i), sizeof(*rx)); + RING_COPY_RESPONSE(&queue->rx, i, rx); memset(extras, 0, sizeof(rinfo.extras)); err = xennet_get_responses(queue, &rinfo, rp, &tmpq, @@ -1286,17 +1331,18 @@ static void xennet_release_tx_bufs(struct netfront_queue *queue) for (i = 0; i < NET_TX_RING_SIZE; i++) { /* Skip over entries which are actually freelist references */ - if (skb_entry_is_link(&queue->tx_skbs[i])) + if (!queue->tx_skbs[i]) continue; - skb = queue->tx_skbs[i].skb; + skb = queue->tx_skbs[i]; + queue->tx_skbs[i] = NULL; get_page(queue->grant_tx_page[i]); gnttab_end_foreign_access(queue->grant_tx_ref[i], GNTMAP_readonly, (unsigned long)page_address(queue->grant_tx_page[i])); queue->grant_tx_page[i] = NULL; queue->grant_tx_ref[i] = GRANT_INVALID_REF; - add_id_to_freelist(&queue->tx_skb_freelist, queue->tx_skbs, i); + add_id_to_list(&queue->tx_skb_freelist, queue->tx_link, i); dev_kfree_skb_irq(skb); } } @@ -1376,6 +1422,9 @@ static irqreturn_t xennet_tx_interrupt(int irq, void *dev_id) struct netfront_queue *queue = dev_id; unsigned long flags; + if (queue->info->broken) + return IRQ_HANDLED; + spin_lock_irqsave(&queue->tx_lock, flags); xennet_tx_buf_gc(queue); spin_unlock_irqrestore(&queue->tx_lock, flags); @@ -1388,6 +1437,9 @@ static irqreturn_t xennet_rx_interrupt(int irq, void *dev_id) struct netfront_queue *queue = dev_id; struct net_device *dev = queue->info->netdev; + if (queue->info->broken) + return IRQ_HANDLED; + if (likely(netif_carrier_ok(dev) && RING_HAS_UNCONSUMED_RESPONSES(&queue->rx))) napi_schedule(&queue->napi); @@ -1409,6 +1461,10 @@ static void xennet_poll_controller(struct net_device *dev) struct netfront_info *info = netdev_priv(dev); unsigned int num_queues = dev->real_num_tx_queues; unsigned int i; + + if (info->broken) + return; + for (i = 0; i < num_queues; ++i) xennet_interrupt(0, &info->queues[i]); } @@ -1480,6 +1536,11 @@ static int xennet_xdp_set(struct net_device *dev, struct bpf_prog *prog, static int xennet_xdp(struct net_device *dev, struct netdev_bpf *xdp) { + struct netfront_info *np = netdev_priv(dev); + + if (np->broken) + return -ENODEV; + switch (xdp->command) { case XDP_SETUP_PROG: return xennet_xdp_set(dev, xdp->prog, xdp->extack); @@ -1853,13 +1914,15 @@ static int xennet_init_queue(struct netfront_queue *queue) snprintf(queue->name, sizeof(queue->name), "vif%s-q%u", devid, queue->id); - /* Initialise tx_skbs as a free chain containing every entry. */ + /* Initialise tx_skb_freelist as a free chain containing every entry. */ queue->tx_skb_freelist = 0; + queue->tx_pend_queue = TX_LINK_NONE; for (i = 0; i < NET_TX_RING_SIZE; i++) { - skb_entry_set_link(&queue->tx_skbs[i], i+1); + queue->tx_link[i] = i + 1; queue->grant_tx_ref[i] = GRANT_INVALID_REF; queue->grant_tx_page[i] = NULL; } + queue->tx_link[NET_TX_RING_SIZE - 1] = TX_LINK_NONE; /* Clear out rx_skbs */ for (i = 0; i < NET_RX_RING_SIZE; i++) { @@ -2128,6 +2191,9 @@ static int talk_to_netback(struct xenbus_device *dev, if (info->queues) xennet_destroy_queues(info); + /* For the case of a reconnect reset the "broken" indicator. */ + info->broken = false; + err = xennet_create_queues(info, &num_queues); if (err < 0) { xenbus_dev_fatal(dev, err, "creating queues"); |