xen/netfront: harden netfront against event channel storms

The Xen netfront driver is still vulnerable for an attack via excessive
number of events sent by the backend. Fix that by using lateeoi event
channels.

For being able to detect the case of no rx responses being added while
the carrier is down a new lock is needed in order to update and test
rsp_cons and the number of seen unconsumed responses atomically.

This is part of XSA-391

Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
V2:
- don't eoi irq in case of interface set broken (Jan Beulich)
- handle carrier off + no new responses added (Jan Beulich)
V3:
- add rx_ prefix to rsp_unconsumed (Jan Beulich)
- correct xennet_set_rx_rsp_cons() spelling (Jan Beulich)

diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c
index 911f43986a8c9fdad0afa0d384f4629d32972c53..d514d96027a6ff40b3ec7ef6aac7d03de1490669 100644 (file)
--- a/drivers/net/xen-netfront.c
+++ b/drivers/net/xen-netfront.c
@@ -148,6 +148,9 @@ struct netfront_queue {
        grant_ref_t gref_rx_head;
        grant_ref_t grant_rx_ref[NET_RX_RING_SIZE];
 
+       unsigned int rx_rsp_unconsumed;
+       spinlock_t rx_cons_lock;
+
        struct page_pool *page_pool;
        struct xdp_rxq_info xdp_rxq;
 };
@@ -376,12 +379,13 @@ static int xennet_open(struct net_device *dev)
        return 0;
 }
 
-static void xennet_tx_buf_gc(struct netfront_queue *queue)
+static bool xennet_tx_buf_gc(struct netfront_queue *queue)
 {
        RING_IDX cons, prod;
        unsigned short id;
        struct sk_buff *skb;
        bool more_to_do;
+       bool work_done = false;
        const struct device *dev = &queue->info->netdev->dev;
 
        BUG_ON(!netif_carrier_ok(queue->info->netdev));
@@ -398,6 +402,8 @@ static void xennet_tx_buf_gc(struct netfront_queue *queue)
                for (cons = queue->tx.rsp_cons; cons != prod; cons++) {
                        struct xen_netif_tx_response txrsp;
 
+                       work_done = true;
+
                        RING_COPY_RESPONSE(&queue->tx, cons, &txrsp);
                        if (txrsp.status == XEN_NETIF_RSP_NULL)
                                continue;
@@ -441,11 +447,13 @@ static void xennet_tx_buf_gc(struct netfront_queue *queue)
 
        xennet_maybe_wake_tx(queue);
 
-       return;
+       return work_done;
 
  err:
        queue->info->broken = true;
        dev_alert(dev, "Disabled for further use\n");
+
+       return work_done;
 }
 
 struct xennet_gnttab_make_txreq {
@@ -834,6 +842,16 @@ static int xennet_close(struct net_device *dev)
        return 0;
 }
 
+static void xennet_set_rx_rsp_cons(struct netfront_queue *queue, RING_IDX val)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&queue->rx_cons_lock, flags);
+       queue->rx.rsp_cons = val;
+       queue->rx_rsp_unconsumed = RING_HAS_UNCONSUMED_RESPONSES(&queue->rx);
+       spin_unlock_irqrestore(&queue->rx_cons_lock, flags);
+}
+
 static void xennet_move_rx_slot(struct netfront_queue *queue, struct sk_buff *skb,
                                grant_ref_t ref)
 {
@@ -885,7 +903,7 @@ static int xennet_get_extras(struct netfront_queue *queue,
                xennet_move_rx_slot(queue, skb, ref);
        } while (extra.flags & XEN_NETIF_EXTRA_FLAG_MORE);
 
-       queue->rx.rsp_cons = cons;
+       xennet_set_rx_rsp_cons(queue, cons);
        return err;
 }
 
@@ -1039,7 +1057,7 @@ next:
        }
 
        if (unlikely(err))
-               queue->rx.rsp_cons = cons + slots;
+               xennet_set_rx_rsp_cons(queue, cons + slots);
 
        return err;
 }
@@ -1093,7 +1111,8 @@ static int xennet_fill_frags(struct netfront_queue *queue,
                        __pskb_pull_tail(skb, pull_to - skb_headlen(skb));
                }
                if (unlikely(skb_shinfo(skb)->nr_frags >= MAX_SKB_FRAGS)) {
-                       queue->rx.rsp_cons = ++cons + skb_queue_len(list);
+                       xennet_set_rx_rsp_cons(queue,
+                                              ++cons + skb_queue_len(list));
                        kfree_skb(nskb);
                        return -ENOENT;
                }
@@ -1106,7 +1125,7 @@ static int xennet_fill_frags(struct netfront_queue *queue,
                kfree_skb(nskb);
        }
 
-       queue->rx.rsp_cons = cons;
+       xennet_set_rx_rsp_cons(queue, cons);
 
        return 0;
 }
@@ -1229,7 +1248,9 @@ err:
 
                        if (unlikely(xennet_set_skb_gso(skb, gso))) {
                                __skb_queue_head(&tmpq, skb);
-                               queue->rx.rsp_cons += skb_queue_len(&tmpq);
+                               xennet_set_rx_rsp_cons(queue,
+                                                      queue->rx.rsp_cons +
+                                                      skb_queue_len(&tmpq));
                                goto err;
                        }
                }
@@ -1253,7 +1274,8 @@ err:
 
                __skb_queue_tail(&rxq, skb);
 
-               i = ++queue->rx.rsp_cons;
+               i = queue->rx.rsp_cons + 1;
+               xennet_set_rx_rsp_cons(queue, i);
                work_done++;
        }
        if (need_xdp_flush)
@@ -1417,40 +1439,79 @@ static int xennet_set_features(struct net_device *dev,
        return 0;
 }
 
-static irqreturn_t xennet_tx_interrupt(int irq, void *dev_id)
+static bool xennet_handle_tx(struct netfront_queue *queue, unsigned int *eoi)
 {
-       struct netfront_queue *queue = dev_id;
        unsigned long flags;
 
-       if (queue->info->broken)
-               return IRQ_HANDLED;
+       if (unlikely(queue->info->broken))
+               return false;
 
        spin_lock_irqsave(&queue->tx_lock, flags);
-       xennet_tx_buf_gc(queue);
+       if (xennet_tx_buf_gc(queue))
+               *eoi = 0;
        spin_unlock_irqrestore(&queue->tx_lock, flags);
 
+       return true;
+}
+
+static irqreturn_t xennet_tx_interrupt(int irq, void *dev_id)
+{
+       unsigned int eoiflag = XEN_EOI_FLAG_SPURIOUS;
+
+       if (likely(xennet_handle_tx(dev_id, &eoiflag)))
+               xen_irq_lateeoi(irq, eoiflag);
+
        return IRQ_HANDLED;
 }
 
-static irqreturn_t xennet_rx_interrupt(int irq, void *dev_id)
+static bool xennet_handle_rx(struct netfront_queue *queue, unsigned int *eoi)
 {
-       struct netfront_queue *queue = dev_id;
-       struct net_device *dev = queue->info->netdev;
+       unsigned int work_queued;
+       unsigned long flags;
 
-       if (queue->info->broken)
-               return IRQ_HANDLED;
+       if (unlikely(queue->info->broken))
+               return false;
+
+       spin_lock_irqsave(&queue->rx_cons_lock, flags);
+       work_queued = RING_HAS_UNCONSUMED_RESPONSES(&queue->rx);
+       if (work_queued > queue->rx_rsp_unconsumed) {
+               queue->rx_rsp_unconsumed = work_queued;
+               *eoi = 0;
+       } else if (unlikely(work_queued < queue->rx_rsp_unconsumed)) {
+               const struct device *dev = &queue->info->netdev->dev;
+
+               spin_unlock_irqrestore(&queue->rx_cons_lock, flags);
+               dev_alert(dev, "RX producer index going backwards\n");
+               dev_alert(dev, "Disabled for further use\n");
+               queue->info->broken = true;
+               return false;
+       }
+       spin_unlock_irqrestore(&queue->rx_cons_lock, flags);
 
-       if (likely(netif_carrier_ok(dev) &&
-                  RING_HAS_UNCONSUMED_RESPONSES(&queue->rx)))
+       if (likely(netif_carrier_ok(queue->info->netdev) && work_queued))
                napi_schedule(&queue->napi);
 
+       return true;
+}
+
+static irqreturn_t xennet_rx_interrupt(int irq, void *dev_id)
+{
+       unsigned int eoiflag = XEN_EOI_FLAG_SPURIOUS;
+
+       if (likely(xennet_handle_rx(dev_id, &eoiflag)))
+               xen_irq_lateeoi(irq, eoiflag);
+
        return IRQ_HANDLED;
 }
 
 static irqreturn_t xennet_interrupt(int irq, void *dev_id)
 {
-       xennet_tx_interrupt(irq, dev_id);
-       xennet_rx_interrupt(irq, dev_id);
+       unsigned int eoiflag = XEN_EOI_FLAG_SPURIOUS;
+
+       if (xennet_handle_tx(dev_id, &eoiflag) &&
+           xennet_handle_rx(dev_id, &eoiflag))
+               xen_irq_lateeoi(irq, eoiflag);
+
        return IRQ_HANDLED;
 }
 
@@ -1768,9 +1829,10 @@ static int setup_netfront_single(struct netfront_queue *queue)
        if (err < 0)
                goto fail;
 
-       err = bind_evtchn_to_irqhandler(queue->tx_evtchn,
-                                       xennet_interrupt,
-                                       0, queue->info->netdev->name, queue);
+       err = bind_evtchn_to_irqhandler_lateeoi(queue->tx_evtchn,
+                                               xennet_interrupt, 0,
+                                               queue->info->netdev->name,
+                                               queue);
        if (err < 0)
                goto bind_fail;
        queue->rx_evtchn = queue->tx_evtchn;
@@ -1798,18 +1860,18 @@ static int setup_netfront_split(struct netfront_queue *queue)
 
        snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name),
                 "%s-tx", queue->name);
-       err = bind_evtchn_to_irqhandler(queue->tx_evtchn,
-                                       xennet_tx_interrupt,
-                                       0, queue->tx_irq_name, queue);
+       err = bind_evtchn_to_irqhandler_lateeoi(queue->tx_evtchn,
+                                               xennet_tx_interrupt, 0,
+                                               queue->tx_irq_name, queue);
        if (err < 0)
                goto bind_tx_fail;
        queue->tx_irq = err;
 
        snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name),
                 "%s-rx", queue->name);
-       err = bind_evtchn_to_irqhandler(queue->rx_evtchn,
-                                       xennet_rx_interrupt,
-                                       0, queue->rx_irq_name, queue);
+       err = bind_evtchn_to_irqhandler_lateeoi(queue->rx_evtchn,
+                                               xennet_rx_interrupt, 0,
+                                               queue->rx_irq_name, queue);
        if (err < 0)
                goto bind_rx_fail;
        queue->rx_irq = err;
@@ -1911,6 +1973,7 @@ static int xennet_init_queue(struct netfront_queue *queue)
 
        spin_lock_init(&queue->tx_lock);
        spin_lock_init(&queue->rx_lock);
+       spin_lock_init(&queue->rx_cons_lock);
 
        timer_setup(&queue->rx_refill_timer, rx_refill_timeout, 0);