Merge branch 'xsa' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Merge xen fixes from Juergen Gross:
 "Fixes for two issues related to Xen and malicious guests:

   - Guest can force the netback driver to hog large amounts of memory

   - Denial of Service in other guests due to event storms"

* 'xsa' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/netback: don't queue unlimited number of packages
  xen/netback: fix rx queue stall detection
  xen/console: harden hvc_xen against event channel storms
  xen/netfront: harden netfront against event channel storms
  xen/blkfront: harden blkfront against event channel storms

7 files changed, 190 insertions, 65 deletions

diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c
index 8e3983e456f3..286cf1afad78 100644
--- a/drivers/block/xen-blkfront.c
+++ b/drivers/block/xen-blkfront.c
@@ -1512,9 +1512,12 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 	unsigned long flags;
 	struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)dev_id;
 	struct blkfront_info *info = rinfo->dev_info;
+	unsigned int eoiflag = XEN_EOI_FLAG_SPURIOUS;
 
-	if (unlikely(info->connected != BLKIF_STATE_CONNECTED))
+	if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) {
+		xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS);
 		return IRQ_HANDLED;
+	}
 
 	spin_lock_irqsave(&rinfo->ring_lock, flags);
  again:
@@ -1530,6 +1533,8 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 		unsigned long id;
 		unsigned int op;
 
+		eoiflag = 0;
+
 		RING_COPY_RESPONSE(&rinfo->ring, i, &bret);
 		id = bret.id;
 
@@ -1646,6 +1651,8 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 
 	spin_unlock_irqrestore(&rinfo->ring_lock, flags);
 
+	xen_irq_lateeoi(irq, eoiflag);
+
 	return IRQ_HANDLED;
 
  err:
@@ -1653,6 +1660,8 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 
 	spin_unlock_irqrestore(&rinfo->ring_lock, flags);
 
+	/* No EOI in order to avoid further interrupts. */
+
 	pr_alert("%s disabled for further use\n", info->gd->disk_name);
 	return IRQ_HANDLED;
 }
@@ -1692,8 +1701,8 @@ static int setup_blkring(struct xenbus_device *dev,
 	if (err)
 		goto fail;
 
-	err = bind_evtchn_to_irqhandler(rinfo->evtchn, blkif_interrupt, 0,
-					"blkif", rinfo);
+	err = bind_evtchn_to_irqhandler_lateeoi(rinfo->evtchn, blkif_interrupt,
+						0, "blkif", rinfo);
 	if (err <= 0) {
 		xenbus_dev_fatal(dev, err,
 				 "bind_evtchn_to_irqhandler failed");
diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h
index 4a16d6e33c09..d9dea4829c86 100644
--- a/drivers/net/xen-netback/common.h
+++ b/drivers/net/xen-netback/common.h
@@ -203,6 +203,7 @@ struct xenvif_queue { /* Per-queue data for xenvif */
 	unsigned int rx_queue_max;
 	unsigned int rx_queue_len;
 	unsigned long last_rx_time;
+	unsigned int rx_slots_needed;
 	bool stalled;
 
 	struct xenvif_copy_state rx_copy;
diff --git a/drivers/net/xen-netback/rx.c b/drivers/net/xen-netback/rx.c
index accc991d153f..dbac4c03d21a 100644
--- a/drivers/net/xen-netback/rx.c
+++ b/drivers/net/xen-netback/rx.c
@@ -33,28 +33,36 @@
 #include <xen/xen.h>
 #include <xen/events.h>
 
-static bool xenvif_rx_ring_slots_available(struct xenvif_queue *queue)
+/*
+ * Update the needed ring page slots for the first SKB queued.
+ * Note that any call sequence outside the RX thread calling this function
+ * needs to wake up the RX thread via a call of xenvif_kick_thread()
+ * afterwards in order to avoid a race with putting the thread to sleep.
+ */
+static void xenvif_update_needed_slots(struct xenvif_queue *queue,
+				       const struct sk_buff *skb)
 {
-	RING_IDX prod, cons;
-	struct sk_buff *skb;
-	int needed;
-	unsigned long flags;
-
-	spin_lock_irqsave(&queue->rx_queue.lock, flags);
+	unsigned int needed = 0;
 
-	skb = skb_peek(&queue->rx_queue);
-	if (!skb) {
-		spin_unlock_irqrestore(&queue->rx_queue.lock, flags);
-		return false;
+	if (skb) {
+		needed = DIV_ROUND_UP(skb->len, XEN_PAGE_SIZE);
+		if (skb_is_gso(skb))
+			needed++;
+		if (skb->sw_hash)
+			needed++;
 	}
 
-	needed = DIV_ROUND_UP(skb->len, XEN_PAGE_SIZE);
-	if (skb_is_gso(skb))
-		needed++;
-	if (skb->sw_hash)
-		needed++;
+	WRITE_ONCE(queue->rx_slots_needed, needed);
+}
 
-	spin_unlock_irqrestore(&queue->rx_queue.lock, flags);
+static bool xenvif_rx_ring_slots_available(struct xenvif_queue *queue)
+{
+	RING_IDX prod, cons;
+	unsigned int needed;
+
+	needed = READ_ONCE(queue->rx_slots_needed);
+	if (!needed)
+		return false;
 
 	do {
 		prod = queue->rx.sring->req_prod;
@@ -80,13 +88,19 @@ void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb)
 
 	spin_lock_irqsave(&queue->rx_queue.lock, flags);
 
-	__skb_queue_tail(&queue->rx_queue, skb);
-
-	queue->rx_queue_len += skb->len;
-	if (queue->rx_queue_len > queue->rx_queue_max) {
+	if (queue->rx_queue_len >= queue->rx_queue_max) {
 		struct net_device *dev = queue->vif->dev;
 
 		netif_tx_stop_queue(netdev_get_tx_queue(dev, queue->id));
+		kfree_skb(skb);
+		queue->vif->dev->stats.rx_dropped++;
+	} else {
+		if (skb_queue_empty(&queue->rx_queue))
+			xenvif_update_needed_slots(queue, skb);
+
+		__skb_queue_tail(&queue->rx_queue, skb);
+
+		queue->rx_queue_len += skb->len;
 	}
 
 	spin_unlock_irqrestore(&queue->rx_queue.lock, flags);
@@ -100,6 +114,8 @@ static struct sk_buff *xenvif_rx_dequeue(struct xenvif_queue *queue)
 
 	skb = __skb_dequeue(&queue->rx_queue);
 	if (skb) {
+		xenvif_update_needed_slots(queue, skb_peek(&queue->rx_queue));
+
 		queue->rx_queue_len -= skb->len;
 		if (queue->rx_queue_len < queue->rx_queue_max) {
 			struct netdev_queue *txq;
@@ -134,6 +150,7 @@ static void xenvif_rx_queue_drop_expired(struct xenvif_queue *queue)
 			break;
 		xenvif_rx_dequeue(queue);
 		kfree_skb(skb);
+		queue->vif->dev->stats.rx_dropped++;
 	}
 }
 
@@ -487,27 +504,31 @@ void xenvif_rx_action(struct xenvif_queue *queue)
 	xenvif_rx_copy_flush(queue);
 }
 
-static bool xenvif_rx_queue_stalled(struct xenvif_queue *queue)
+static RING_IDX xenvif_rx_queue_slots(const struct xenvif_queue *queue)
 {
 	RING_IDX prod, cons;
 
 	prod = queue->rx.sring->req_prod;
 	cons = queue->rx.req_cons;
 
+	return prod - cons;
+}
+
+static bool xenvif_rx_queue_stalled(const struct xenvif_queue *queue)
+{
+	unsigned int needed = READ_ONCE(queue->rx_slots_needed);
+
 	return !queue->stalled &&
-		prod - cons < 1 &&
+		xenvif_rx_queue_slots(queue) < needed &&
 		time_after(jiffies,
 			   queue->last_rx_time + queue->vif->stall_timeout);
 }
 
 static bool xenvif_rx_queue_ready(struct xenvif_queue *queue)
 {
-	RING_IDX prod, cons;
-
-	prod = queue->rx.sring->req_prod;
-	cons = queue->rx.req_cons;
+	unsigned int needed = READ_ONCE(queue->rx_slots_needed);
 
-	return queue->stalled && prod - cons >= 1;
+	return queue->stalled && xenvif_rx_queue_slots(queue) >= needed;
 }
 
 bool xenvif_have_rx_work(struct xenvif_queue *queue, bool test_kthread)
diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c
index 911f43986a8c..d514d96027a6 100644
--- 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);
 
diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index 71e0dd2c0ce5..ebaf7500f48f 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -37,6 +37,8 @@ struct xencons_info {
 	struct xenbus_device *xbdev;
 	struct xencons_interface *intf;
 	unsigned int evtchn;
+	XENCONS_RING_IDX out_cons;
+	unsigned int out_cons_same;
 	struct hvc_struct *hvc;
 	int irq;
 	int vtermno;
@@ -138,6 +140,8 @@ static int domU_read_console(uint32_t vtermno, char *buf, int len)
 	XENCONS_RING_IDX cons, prod;
 	int recv = 0;
 	struct xencons_info *xencons = vtermno_to_xencons(vtermno);
+	unsigned int eoiflag = 0;
+
 	if (xencons == NULL)
 		return -EINVAL;
 	intf = xencons->intf;
@@ -157,7 +161,27 @@ static int domU_read_console(uint32_t vtermno, char *buf, int len)
 	mb();			/* read ring before consuming */
 	intf->in_cons = cons;
 
-	notify_daemon(xencons);
+	/*
+	 * When to mark interrupt having been spurious:
+	 * - there was no new data to be read, and
+	 * - the backend did not consume some output bytes, and
+	 * - the previous round with no read data didn't see consumed bytes
+	 *   (we might have a race with an interrupt being in flight while
+	 *   updating xencons->out_cons, so account for that by allowing one
+	 *   round without any visible reason)
+	 */
+	if (intf->out_cons != xencons->out_cons) {
+		xencons->out_cons = intf->out_cons;
+		xencons->out_cons_same = 0;
+	}
+	if (recv) {
+		notify_daemon(xencons);
+	} else if (xencons->out_cons_same++ > 1) {
+		eoiflag = XEN_EOI_FLAG_SPURIOUS;
+	}
+
+	xen_irq_lateeoi(xencons->irq, eoiflag);
+
 	return recv;
 }
 
@@ -386,7 +410,7 @@ static int xencons_connect_backend(struct xenbus_device *dev,
 	if (ret)
 		return ret;
 	info->evtchn = evtchn;
-	irq = bind_evtchn_to_irq(evtchn);
+	irq = bind_interdomain_evtchn_to_irq_lateeoi(dev, evtchn);
 	if (irq < 0)
 		return irq;
 	info->irq = irq;
@@ -551,7 +575,7 @@ static int __init xen_hvc_init(void)
 			return r;
 
 		info = vtermno_to_xencons(HVC_COOKIE);
-		info->irq = bind_evtchn_to_irq(info->evtchn);
+		info->irq = bind_evtchn_to_irq_lateeoi(info->evtchn);
 	}
 	if (info->irq < 0)
 		info->irq = 0; /* NO_IRQ */
diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
index a78704ae3618..46d9295d9a6e 100644
--- a/drivers/xen/events/events_base.c
+++ b/drivers/xen/events/events_base.c
@@ -1251,6 +1251,12 @@ int bind_evtchn_to_irq(evtchn_port_t evtchn)
 }
 EXPORT_SYMBOL_GPL(bind_evtchn_to_irq);
 
+int bind_evtchn_to_irq_lateeoi(evtchn_port_t evtchn)
+{
+	return bind_evtchn_to_irq_chip(evtchn, &xen_lateeoi_chip, NULL);
+}
+EXPORT_SYMBOL_GPL(bind_evtchn_to_irq_lateeoi);
+
 static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu)
 {
 	struct evtchn_bind_ipi bind_ipi;
diff --git a/include/xen/events.h b/include/xen/events.h
index c204262d9fc2..344081e71584 100644
--- a/include/xen/events.h
+++ b/include/xen/events.h
@@ -17,6 +17,7 @@ struct xenbus_device;
 unsigned xen_evtchn_nr_channels(void);
 
 int bind_evtchn_to_irq(evtchn_port_t evtchn);
+int bind_evtchn_to_irq_lateeoi(evtchn_port_t evtchn);
 int bind_evtchn_to_irqhandler(evtchn_port_t evtchn,
 			      irq_handler_t handler,
 			      unsigned long irqflags, const char *devname,