From: Paul Durrant <pdurrant@amazon.com>

Currently a single watch on /local/domain/X/backend is registered by each
QEMU process running in service domain X (where X is usually 0). The purpose
of this watch is to ensure that QEMU is notified when the Xen toolstack
creates a new device backend area.
Such a backend area is specific to a single frontend area created for a
specific guest domain and, since each QEMU process is also created to service
a specfic guest domain, it is unnecessary and inefficient to notify all QEMU
processes.
Only the QEMU process associated with the same guest domain need
receive the notification. This patch re-factors the watch registration code
such that notifications are targetted appropriately.

Reported-by: Jerome Leseinne <jerome.leseinne@gmail.com>
Signed-off-by: Paul Durrant <pdurrant@amazon.com>
---
Cc: Edwin Torok <edvin.torok@citrix.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Anthony Perard <anthony.perard@citrix.com>
---
 hw/xen/xen-backend.c         | 10 ++++++++++
 hw/xen/xen-bus.c             | 38 ++++++++++++++++++++++++++++--------
 include/hw/xen/xen-backend.h |  1 +
 include/hw/xen/xen-bus.h     |  3 ++-
 4 files changed, 43 insertions(+), 9 deletions(-)

diff --git a/hw/xen/xen-backend.c b/hw/xen/xen-backend.c
index 10199fb58d..f2711fe4a7 100644
--- a/hw/xen/xen-backend.c
+++ b/hw/xen/xen-backend.c
@@ -41,6 +41,11 @@ static void xen_backend_table_add(XenBackendImpl *impl)
     g_hash_table_insert(xen_backend_table_get(), (void *)impl->type, impl);
 }
 
+static void **xen_backend_table_keys(unsigned int *count)
+{
+    return g_hash_table_get_keys_as_array(xen_backend_table_get(), count);
+}
+
 static const XenBackendImpl *xen_backend_table_lookup(const char *type)
 {
     return g_hash_table_lookup(xen_backend_table_get(), type);
@@ -70,6 +75,11 @@ void xen_backend_register(const XenBackendInfo *info)
     xen_backend_table_add(impl);
 }
 
+const char **xen_backend_get_types(unsigned int *count)
+{
+    return (const char **)xen_backend_table_keys(count);
+}
+
 static QLIST_HEAD(, XenBackendInstance) backend_list;
 
 static void xen_backend_list_add(XenBackendInstance *backend)
diff --git a/hw/xen/xen-bus.c b/hw/xen/xen-bus.c
index 9ce1c9540b..c83da93bf3 100644
--- a/hw/xen/xen-bus.c
+++ b/hw/xen/xen-bus.c
@@ -430,7 +430,13 @@ static void xen_bus_unrealize(BusState *bus)
     trace_xen_bus_unrealize();
 
     if (xenbus->backend_watch) {
-        xen_bus_remove_watch(xenbus, xenbus->backend_watch, NULL);
+        unsigned int i;
+
+        for (i = 0; i < xenbus->backend_types; i++) {
+            xen_bus_remove_watch(xenbus, xenbus->backend_watch[i], NULL);
+        }
+
+        g_free(xenbus->backend_watch);
         xenbus->backend_watch = NULL;
     }
 
@@ -446,8 +452,11 @@ static void xen_bus_unrealize(BusState *bus)
 
 static void xen_bus_realize(BusState *bus, Error **errp)
 {
+    char *key = g_strdup_printf("%u", xen_domid);
     XenBus *xenbus = XEN_BUS(bus);
     unsigned int domid;
+    const char **type;
+    unsigned int i;
     Error *local_err = NULL;
 
     trace_xen_bus_realize();
@@ -469,19 +478,32 @@ static void xen_bus_realize(BusState *bus, Error **errp)
 
     module_call_init(MODULE_INIT_XEN_BACKEND);
 
-    xenbus->backend_watch =
-        xen_bus_add_watch(xenbus, "", /* domain root node */
-                          "backend", xen_bus_backend_changed, &local_err);
-    if (local_err) {
-        /* This need not be treated as a hard error so don't propagate */
-        error_reportf_err(local_err,
-                          "failed to set up enumeration watch: ");
+    type = xen_backend_get_types(&xenbus->backend_types);
+    xenbus->backend_watch = g_new(XenWatch *, xenbus->backend_types);
+
+    for (i = 0; i < xenbus->backend_types; i++) {
+        char *node = g_strdup_printf("backend/%s", type[i]);
+
+        xenbus->backend_watch[i] =
+            xen_bus_add_watch(xenbus, node, key, xen_bus_backend_changed,
+                              &local_err);
+        if (local_err) {
+            /* This need not be treated as a hard error so don't propagate */
+            error_reportf_err(local_err,
+                              "failed to set up '%s' enumeration watch: ",
+                              type[i]);
+        }
+
+        g_free(node);
     }
 
+    g_free(type);
+    g_free(key);
     return;
 
 fail:
     xen_bus_unrealize(bus);
+    g_free(key);
 }
 
 static void xen_bus_unplug_request(HotplugHandler *hotplug,
diff --git a/include/hw/xen/xen-backend.h b/include/hw/xen/xen-backend.h
index 010d712638..aac2fd454d 100644
--- a/include/hw/xen/xen-backend.h
+++ b/include/hw/xen/xen-backend.h
@@ -31,6 +31,7 @@ void xen_backend_set_device(XenBackendInstance *backend,
 XenDevice *xen_backend_get_device(XenBackendInstance *backend);
 
 void xen_backend_register(const XenBackendInfo *info);
+const char **xen_backend_get_types(unsigned int *nr);
 
 void xen_backend_device_create(XenBus *xenbus, const char *type,
                                const char *name, QDict *opts, Error **errp);
diff --git a/include/hw/xen/xen-bus.h b/include/hw/xen/xen-bus.h
index 3df696136f..6bdbf3ff82 100644
--- a/include/hw/xen/xen-bus.h
+++ b/include/hw/xen/xen-bus.h
@@ -66,7 +66,8 @@ struct XenBus {
     domid_t backend_id;
     struct xs_handle *xsh;
     XenWatchList *watch_list;
-    XenWatch *backend_watch;
+    unsigned int backend_types;
+    XenWatch **backend_watch;
     QLIST_HEAD(, XenDevice) inactive_devices;
 };
 
-- 
2.20.1

On Wed, Sep 23, 2020 at 04:57:31PM +0100, Paul Durrant wrote:
> From: Paul Durrant <pdurrant@amazon.com>
> 
> Currently a single watch on /local/domain/X/backend is registered by each
> QEMU process running in service domain X (where X is usually 0). The purpose
> of this watch is to ensure that QEMU is notified when the Xen toolstack
> creates a new device backend area.
> Such a backend area is specific to a single frontend area created for a
> specific guest domain and, since each QEMU process is also created to service
> a specfic guest domain, it is unnecessary and inefficient to notify all QEMU
> processes.
> Only the QEMU process associated with the same guest domain need
> receive the notification. This patch re-factors the watch registration code
> such that notifications are targetted appropriately.
> 
> Reported-by: Jerome Leseinne <jerome.leseinne@gmail.com>
> Signed-off-by: Paul Durrant <pdurrant@amazon.com>
> 
> diff --git a/hw/xen/xen-backend.c b/hw/xen/xen-backend.c
> index 10199fb58d..f2711fe4a7 100644
> --- a/hw/xen/xen-backend.c
> +++ b/hw/xen/xen-backend.c
> @@ -41,6 +41,11 @@ static void xen_backend_table_add(XenBackendImpl *impl)
>      g_hash_table_insert(xen_backend_table_get(), (void *)impl->type, impl);
>  }
>  
> +static void **xen_backend_table_keys(unsigned int *count)
> +{
> +    return g_hash_table_get_keys_as_array(xen_backend_table_get(), count);

That could be cast to (const gchar **) as the GLib doc suggest, or (const
char **) since gchar and char are the same.
https://developer.gnome.org/glib/stable/glib-Hash-Tables.html#g-hash-table-get-keys-as-array

> +}
> +
>  static const XenBackendImpl *xen_backend_table_lookup(const char *type)
>  {
>      return g_hash_table_lookup(xen_backend_table_get(), type);
> diff --git a/hw/xen/xen-bus.c b/hw/xen/xen-bus.c
> index 9ce1c9540b..c83da93bf3 100644
> --- a/hw/xen/xen-bus.c
> +++ b/hw/xen/xen-bus.c
> @@ -430,7 +430,13 @@ static void xen_bus_unrealize(BusState *bus)
>      trace_xen_bus_unrealize();
>  
>      if (xenbus->backend_watch) {
> -        xen_bus_remove_watch(xenbus, xenbus->backend_watch, NULL);
> +        unsigned int i;
> +
> +        for (i = 0; i < xenbus->backend_types; i++) {
> +            xen_bus_remove_watch(xenbus, xenbus->backend_watch[i], NULL);

We should check if backend_watch[i] is NULL.

> +        }
> +
> +        g_free(xenbus->backend_watch);
>          xenbus->backend_watch = NULL;
>      }
>  

The rest of the patch looks fine. Next improvement is to only look at
only one backend type in xen_bus_backend_changed() since there is now a
watch per backend type :-), but that would be for another day.

Cheers,

-- 
Anthony PERARD

> -----Original Message-----
> From: Anthony PERARD <anthony.perard@citrix.com>
> Sent: 30 September 2020 12:43
> To: Paul Durrant <paul@xen.org>
> Cc: xen-devel@lists.xenproject.org; qemu-devel@nongnu.org; Paul Durrant <pdurrant@amazon.com>; Jerome
> Leseinne <jerome.leseinne@gmail.com>; Edwin Torok <edvin.torok@citrix.com>; Stefano Stabellini
> <sstabellini@kernel.org>
> Subject: Re: [PATCH] xen-bus: reduce scope of backend watch
> 
> On Wed, Sep 23, 2020 at 04:57:31PM +0100, Paul Durrant wrote:
> > From: Paul Durrant <pdurrant@amazon.com>
> >
> > Currently a single watch on /local/domain/X/backend is registered by each
> > QEMU process running in service domain X (where X is usually 0). The purpose
> > of this watch is to ensure that QEMU is notified when the Xen toolstack
> > creates a new device backend area.
> > Such a backend area is specific to a single frontend area created for a
> > specific guest domain and, since each QEMU process is also created to service
> > a specfic guest domain, it is unnecessary and inefficient to notify all QEMU
> > processes.
> > Only the QEMU process associated with the same guest domain need
> > receive the notification. This patch re-factors the watch registration code
> > such that notifications are targetted appropriately.
> >
> > Reported-by: Jerome Leseinne <jerome.leseinne@gmail.com>
> > Signed-off-by: Paul Durrant <pdurrant@amazon.com>
> >
> > diff --git a/hw/xen/xen-backend.c b/hw/xen/xen-backend.c
> > index 10199fb58d..f2711fe4a7 100644
> > --- a/hw/xen/xen-backend.c
> > +++ b/hw/xen/xen-backend.c
> > @@ -41,6 +41,11 @@ static void xen_backend_table_add(XenBackendImpl *impl)
> >      g_hash_table_insert(xen_backend_table_get(), (void *)impl->type, impl);
> >  }
> >
> > +static void **xen_backend_table_keys(unsigned int *count)
> > +{
> > +    return g_hash_table_get_keys_as_array(xen_backend_table_get(), count);
> 
> That could be cast to (const gchar **) as the GLib doc suggest, or (const
> char **) since gchar and char are the same.
> https://developer.gnome.org/glib/stable/glib-Hash-Tables.html#g-hash-table-get-keys-as-array
> 

Ok, I'll re-arrange the const-ing to cast at the inner level.

> > +}
> > +
> >  static const XenBackendImpl *xen_backend_table_lookup(const char *type)
> >  {
> >      return g_hash_table_lookup(xen_backend_table_get(), type);
> > diff --git a/hw/xen/xen-bus.c b/hw/xen/xen-bus.c
> > index 9ce1c9540b..c83da93bf3 100644
> > --- a/hw/xen/xen-bus.c
> > +++ b/hw/xen/xen-bus.c
> > @@ -430,7 +430,13 @@ static void xen_bus_unrealize(BusState *bus)
> >      trace_xen_bus_unrealize();
> >
> >      if (xenbus->backend_watch) {
> > -        xen_bus_remove_watch(xenbus, xenbus->backend_watch, NULL);
> > +        unsigned int i;
> > +
> > +        for (i = 0; i < xenbus->backend_types; i++) {
> > +            xen_bus_remove_watch(xenbus, xenbus->backend_watch[i], NULL);
> 
> We should check if backend_watch[i] is NULL.
> 

Yes, I'll add a check.

> > +        }
> > +
> > +        g_free(xenbus->backend_watch);
> >          xenbus->backend_watch = NULL;
> >      }
> >
> 
> The rest of the patch looks fine. Next improvement is to only look at
> only one backend type in xen_bus_backend_changed() since there is now a
> watch per backend type :-), but that would be for another day.
> 

Might not be too tricky. I'll see if I can come up with a follow-up patch.

  Paul

> Cheers,
> 
> --
> Anthony PERARD