Re: [PATCH 4/5] xen: events, exposes evtchn_alloc_unbound_domain

[Jean Guyader <jean.guyader@citrix.com>]

[-- Attachment #1: Type: text/plain, Size: 2331 bytes --]

On 09/08 11:40, Jean Guyader wrote:
> On 9 August 2012 11:35, Tim Deegan <tim@xen.org> wrote:
> > At 11:23 +0100 on 09 Aug (1344511426), Ian Campbell wrote:
> >> On Thu, 2012-08-09 at 11:06 +0100, Tim Deegan wrote:
> >> > At 20:50 +0100 on 03 Aug (1344027053), Jean Guyader wrote:
> >> > >
> >> > > Exposes evtchn_alloc_unbound_domain to the rest of
> >> > > Xen so we can create allocated unbound evtchn within Xen.
> >> > >
> >> > > Signed-off-by: Jean Guyader <jean.guyader@citrix.com>
> >> >
> >> > > @@ -161,18 +163,18 @@ static long evtchn_alloc_unbound(evtchn_alloc_unbound_t *alloc)
> >> > >  {
> >> > >      struct evtchn *chn;
> >> > >      struct domain *d;
> >> > > -    int            port;
> >> > > +    evtchn_port_t  port;
> >> > >      domid_t        dom = alloc->dom;
> >> > > -    long           rc;
> >> > > +    int            rc;
> >> >
> >> > The function returns long; if you're tidying this up to be an int, might
> >> > as well change the return type too.
> >>
> >> I'm not sure if this is relevant but Jan just sent a patch to "make all
> >> (native) hypercalls consistently have "long" return type". I
> >> think/suspect this rc here turns into the result of the hypercall?
> >>
> >> Jan's patch was motivated by something to do with sign extension when a
> >> hypercall's int return is written to the long in the multicall arg
> >> struct which causes strangeness. Perhaps not totally relevant to
> >> evtchn_alloc which is unlikely to be in a MC.
> >
> > Yes, this eventually ends up in a hypercall handler, but s/long/int/
> > here doesn't cause problems because
> >  - rc is only ever set to an 'int' value here so we can't lose data
> >    from the type being too narrow; and
> >  - Those int values get cast up to long (either in here or in the
> >    caller) directly, which will sign-extend the.
> >
> > It really doesn't matter whether this function returns an int or a long,
> > but it's a bit untidy to change it half-way.
> >
> 
> The main reason why I changed it only base ERROR_EXIT_DOM expects an int based
> on the format string. I guess I could cast the long in int for the
> call to ERROR_EXIT_DOM
> but that doesn't really look nice either.
> 

Hi,

Here is a new version that should address the comments from Tim and Jan.

Signed-off-by: Jean Guyader <jean.guyader@citrix.com>

Jean

[-- Attachment #2: evtchn_alloc_unbound_domain.patch --]
[-- Type: text/x-diff, Size: 2881 bytes --]

commit c43dbcee9c4e9d65520f9a562b39e8e6455efc36
Author: Jean Guyader <jean.guyader@citrix.com>
Date:   Thu Aug 2 16:19:23 2012 +0100

    xen: events, exposes evtchn_alloc_unbound_domain
    
    Exposes evtchn_alloc_unbound_domain to the rest of
    Xen so we can create allocated unbound evtchn within Xen.

diff --git a/xen/common/event_channel.c b/xen/common/event_channel.c
index 53777f8..880395e 100644
--- a/xen/common/event_channel.c
+++ b/xen/common/event_channel.c
@@ -159,9 +159,8 @@ static int get_free_port(struct domain *d)
 
 static long evtchn_alloc_unbound(evtchn_alloc_unbound_t *alloc)
 {
-    struct evtchn *chn;
     struct domain *d;
-    int            port;
+    evtchn_port_t  port;
     domid_t        dom = alloc->dom;
     long           rc;
 
@@ -169,26 +168,47 @@ static long evtchn_alloc_unbound(evtchn_alloc_unbound_t *alloc)
     if ( rc )
         return rc;
 
+    rc = evtchn_alloc_unbound_domain(d, &port,
+            alloc->remote_dom == DOMID_SELF ? current->domain->domain_id
+                                            : alloc->remote_dom);
+    if ( rc )
+        ERROR_EXIT_DOM((int)rc, d);
+
+    alloc->port = port;
+
+ out:
+    rcu_unlock_domain(d);
+
+    return rc;
+}
+
+int evtchn_alloc_unbound_domain(struct domain *d, evtchn_port_t *port,
+                                domid_t remote_domid)
+{
+    struct evtchn *chn;
+    int           rc;
+    int           free_port;
+
     spin_lock(&d->event_lock);
 
-    if ( (port = get_free_port(d)) < 0 )
-        ERROR_EXIT_DOM(port, d);
-    chn = evtchn_from_port(d, port);
+    rc = free_port = get_free_port(d);
+    if ( free_port < 0 )
+        goto out;
 
-    rc = xsm_evtchn_unbound(d, chn, alloc->remote_dom);
+    chn = evtchn_from_port(d, free_port);
+    rc = xsm_evtchn_unbound(d, chn, remote_domid);
     if ( rc )
         goto out;
 
     chn->state = ECS_UNBOUND;
-    if ( (chn->u.unbound.remote_domid = alloc->remote_dom) == DOMID_SELF )
-        chn->u.unbound.remote_domid = current->domain->domain_id;
+    chn->u.unbound.remote_domid = remote_domid;
 
-    alloc->port = port;
+    *port = free_port;
+    /* Everything is fine, returns 0 */
+    rc = 0;
 
  out:
     spin_unlock(&d->event_lock);
-    rcu_unlock_domain(d);
-
     return rc;
 }
 
diff --git a/xen/include/xen/event.h b/xen/include/xen/event.h
index 71c3e92..1a0c832 100644
--- a/xen/include/xen/event.h
+++ b/xen/include/xen/event.h
@@ -69,6 +69,9 @@ int guest_enabled_event(struct vcpu *v, uint32_t virq);
 /* Notify remote end of a Xen-attached event channel.*/
 void notify_via_xen_event_channel(struct domain *ld, int lport);
 
+int evtchn_alloc_unbound_domain(struct domain *d, evtchn_port_t *port,
+                                domid_t remote_domid);
+
 /* Internal event channel object accessors */
 #define bucket_from_port(d,p) \
     ((d)->evtchn[(p)/EVTCHNS_PER_BUCKET])

[-- Attachment #3: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel