From: Jan Beulich <jbeulich@suse.com>
Subject: evtchn: arrange for preemption in evtchn_destroy()

Especially closing of fully established interdomain channels can take
quite some time, due to the locking involved. Therefore we shouldn't
assume we can clean up still active ports all in one go. Besides adding
the necessary preemption check, also avoid pointlessly starting from
(or now really ending at) 0; 1 is the lowest numbered port which may
need closing.

Since we're now reducing ->valid_evtchns, free_xen_event_channel(),
and (at least to be on the safe side) notify_via_xen_event_channel()
need to cope with attempts to close / unbind from / send through already
closed (and no longer valid, as per port_is_valid()) ports.

This is part of XSA-344.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>

--- a/xen/common/domain.c
+++ b/xen/common/domain.c
@@ -770,12 +770,14 @@ int domain_kill(struct domain *d)
             return domain_kill(d);
         d->is_dying = DOMDYING_dying;
         argo_destroy(d);
-        evtchn_destroy(d);
         gnttab_release_mappings(d);
         vnuma_destroy(d->vnuma);
         domain_set_outstanding_pages(d, 0);
         /* fallthrough */
     case DOMDYING_dying:
+        rc = evtchn_destroy(d);
+        if ( rc )
+            break;
         rc = domain_relinquish_resources(d);
         if ( rc != 0 )
             break;
--- a/xen/common/event_channel.c
+++ b/xen/common/event_channel.c
@@ -1297,7 +1297,16 @@ int alloc_unbound_xen_event_channel(
 
 void free_xen_event_channel(struct domain *d, int port)
 {
-    BUG_ON(!port_is_valid(d, port));
+    if ( !port_is_valid(d, port) )
+    {
+        /*
+         * Make sure ->is_dying is read /after/ ->valid_evtchns, pairing
+         * with the spin_barrier() and BUG_ON() in evtchn_destroy().
+         */
+        smp_rmb();
+        BUG_ON(!d->is_dying);
+        return;
+    }
 
     evtchn_close(d, port, 0);
 }
@@ -1309,7 +1318,17 @@ void notify_via_xen_event_channel(struct
     struct domain *rd;
     unsigned long flags;
 
-    ASSERT(port_is_valid(ld, lport));
+    if ( !port_is_valid(ld, lport) )
+    {
+        /*
+         * Make sure ->is_dying is read /after/ ->valid_evtchns, pairing
+         * with the spin_barrier() and BUG_ON() in evtchn_destroy().
+         */
+        smp_rmb();
+        ASSERT(ld->is_dying);
+        return;
+    }
+
     lchn = evtchn_from_port(ld, lport);
 
     spin_lock_irqsave(&lchn->lock, flags);
@@ -1380,8 +1399,7 @@ int evtchn_init(struct domain *d, unsign
     return 0;
 }
 
-
-void evtchn_destroy(struct domain *d)
+int evtchn_destroy(struct domain *d)
 {
     unsigned int i;
 
@@ -1390,14 +1408,29 @@ void evtchn_destroy(struct domain *d)
     spin_barrier(&d->event_lock);
 
     /* Close all existing event channels. */
-    for ( i = 0; port_is_valid(d, i); i++ )
+    for ( i = d->valid_evtchns; --i; )
+    {
         evtchn_close(d, i, 0);
 
+        /*
+         * Avoid preempting when called from domain_create()'s error path,
+         * and don't check too often (choice of frequency is arbitrary).
+         */
+        if ( i && !(i & 0x3f) && d->is_dying != DOMDYING_dead &&
+             hypercall_preempt_check() )
+        {
+            write_atomic(&d->valid_evtchns, i);
+            return -ERESTART;
+        }
+    }
+
     ASSERT(!d->active_evtchns);
 
     clear_global_virq_handlers(d);
 
     evtchn_fifo_destroy(d);
+
+    return 0;
 }
 
 
--- a/xen/include/xen/sched.h
+++ b/xen/include/xen/sched.h
@@ -136,7 +136,7 @@ struct evtchn
 } __attribute__((aligned(64)));
 
 int  evtchn_init(struct domain *d, unsigned int max_port);
-void evtchn_destroy(struct domain *d); /* from domain_kill */
+int  evtchn_destroy(struct domain *d); /* from domain_kill */
 void evtchn_destroy_final(struct domain *d); /* from complete_domain_destroy */
 
 struct waitqueue_vcpu;