evtchn/fifo: don't spin indefinitely when setting LINK

A malicious or buggy guest can cause another domain to spin
indefinitely by repeatedly writing to an event word when the other
guest is trying to link a new event.  The cmpxchg() in
evtchn_fifo_set_link() will repeatedly fail and the loop may never
terminate.

Fixing this requires a change to the ABI which is documented in draft
H of the design.

  http://xenbits.xen.org/people/dvrabel/event-channels-H.pdf

Since a well-behaved guest only makes a limited set of state changes,
the loop can terminate early if the guest makes an invalid state
transition.

The guest may:

- clear LINKED and LINK.
- clear PENDING
- set MASKED
- clear MASKED

It is valid for the guest to mask and unmask an event at any time so
specify that it is not valid for a guest to clear MASKED if Xen is
trying to update LINK.  Indicate this to the guest with an additional
BUSY bit in the event word.  The guest must not clear MASKED if BUSY
is set and it should spin until BUSY is cleared.

The remaining valid writes (clear LINKED, clear PENDING, set MASKED,
clear MASKED by Xen) will limit the number of failures of the
cmpxchg() to at most 4.  A clear of LINKED will also terminate the
loop early. Therefore, the loop can then be limited to at most 4
iterations.

If the buggy or malicious guest does cause the loop to exit with
LINKED set and LINK unset then that buggy guest will lose events.

Reported-by: Anthony Liguori <aliguori@amazon.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>

diff --git a/xen/common/event_fifo.c b/xen/common/event_fifo.c
index 64dbfff3423d5576880f88faaf6c6d4810ab2f64..9106c55f9ae145f7f2c2ed4288d851b3ee8bcfd1 100644 (file)
--- a/xen/common/event_fifo.c
+++ b/xen/common/event_fifo.c
@@ -34,19 +34,67 @@ static inline event_word_t *evtchn_fifo_word_from_port(struct domain *d,
     return d->evtchn_fifo->event_array[p] + w;
 }
 
-static bool_t evtchn_fifo_set_link(event_word_t *word, uint32_t link)
+static int try_set_link(event_word_t *word, event_word_t *w, uint32_t link)
 {
-    event_word_t n, o, w;
+    event_word_t new, old;
 
-    w = *word;
+    if ( !(*w & (1 << EVTCHN_FIFO_LINKED)) )
+        return 0;
 
-    do {
-        if ( !(w & (1 << EVTCHN_FIFO_LINKED)) )
-            return 0;
-        o = w;
-        n = (w & ~EVTCHN_FIFO_LINK_MASK) | link;
-    } while ( (w = cmpxchg(word, o, n)) != o );
+    old = *w;
+    new = (old & ~((1 << EVTCHN_FIFO_BUSY) | EVTCHN_FIFO_LINK_MASK)) | link;
+    *w = cmpxchg(word, old, new);
+    if ( *w == old )
+        return 1;
 
+    return -EAGAIN;
+}
+
+/*
+ * Atomically set the LINK field iff it is still LINKED.
+ *
+ * The guest is only permitted to make the following changes to a
+ * LINKED event.
+ *
+ * - set MASKED
+ * - clear MASKED
+ * - clear PENDING
+ * - clear LINKED (and LINK)
+ *
+ * We block unmasking by the guest by marking the tail word as BUSY,
+ * therefore, the cmpxchg() may fail at most 4 times.
+ */
+static bool_t evtchn_fifo_set_link(const struct domain *d, event_word_t *word,
+                                   uint32_t link)
+{
+    event_word_t w;
+    unsigned int try;
+    int ret;
+
+    w = read_atomic(word);
+
+    ret = try_set_link(word, &w, link);
+    if ( ret >= 0 )
+        return ret;
+
+    /* Lock the word to prevent guest unmasking. */
+    set_bit(EVTCHN_FIFO_BUSY, word);
+
+    w = read_atomic(word);
+
+    for ( try = 0; try < 4; try++ )
+    {
+        ret = try_set_link(word, &w, link);
+        if ( ret >= 0 )
+        {
+            if ( ret == 0 )
+                clear_bit(EVTCHN_FIFO_BUSY, word);
+            return ret;
+        }
+    }
+    gdprintk(XENLOG_WARNING, "domain %d, port %d not linked\n",
+             d->domain_id, link);
+    clear_bit(EVTCHN_FIFO_BUSY, word);
     return 1;
 }
 
@@ -105,7 +153,7 @@ static void evtchn_fifo_set_pending(struct vcpu *v, struct evtchn *evtchn)
         if ( port != q->tail )
         {
             tail_word = evtchn_fifo_word_from_port(d, q->tail);
-            linked = evtchn_fifo_set_link(tail_word, port);
+            linked = evtchn_fifo_set_link(d, tail_word, port);
         }
         if ( !linked )
             write_atomic(q->head, port);
@@ -202,11 +250,12 @@ static void evtchn_fifo_print_state(struct domain *d,
 
     word = evtchn_fifo_word_from_port(d, evtchn->port);
     if ( !word )
-        printk("?   ");
+        printk("?     ");
     else if ( test_bit(EVTCHN_FIFO_LINKED, word) )
-        printk("%-4u", *word & EVTCHN_FIFO_LINK_MASK);
+        printk("%c %-4u", test_bit(EVTCHN_FIFO_BUSY, word) ? 'B' : ' ',
+               *word & EVTCHN_FIFO_LINK_MASK);
     else
-        printk("-   ");
+        printk("%c -   ", test_bit(EVTCHN_FIFO_BUSY, word) ? 'B' : ' ');
 }
 
 static const struct evtchn_port_ops evtchn_port_ops_fifo =

diff --git a/xen/include/public/event_channel.h b/xen/include/public/event_channel.h
index 4a534848e4bd58ffd06321b8c23072345fcb7846..b78ee32726dc2778f409e461c92ed5057775c7e6 100644 (file)
--- a/xen/include/public/event_channel.h
+++ b/xen/include/public/event_channel.h
@@ -343,6 +343,7 @@ typedef uint32_t event_word_t;
 #define EVTCHN_FIFO_PENDING 31
 #define EVTCHN_FIFO_MASKED  30
 #define EVTCHN_FIFO_LINKED  29
+#define EVTCHN_FIFO_BUSY    28
 
 #define EVTCHN_FIFO_LINK_BITS 17
 #define EVTCHN_FIFO_LINK_MASK ((1 << EVTCHN_FIFO_LINK_BITS) - 1)