vl: Delay initialization of memory backends

Initialization of memory backends may take a while when
prealloc=yes is used, depending on their size. Initializing
memory backends before chardevs may delay the creation of monitor
sockets, and trigger timeouts on management software that waits
until the monitor socket is created by QEMU. See, for example,
the bug report at:
https://bugzilla.redhat.com/show_bug.cgi?id=1371211

In addition to that, allocating memory before calling
configure_accelerator() breaks the tcg_enabled() checks at
memory_region_init_*().

This patch fixes those problems by adding "memory-backend-*"
classes to the delayed-initialization list.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

diff --git a/vl.c b/vl.c
index b3c80d50770b4f08e5d4eacc85e2ce2bd8bc7212..ee557a1d3f8a4a7cb9bffb586cb4c7dc75e9262f 100644 (file)
--- a/vl.c
+++ b/vl.c
@@ -2810,6 +2810,19 @@ static bool object_create_initial(const char *type)
         return false;
     }
 
+    /* Memory allocation by backends needs to be done
+     * after configure_accelerator() (due to the tcg_enabled()
+     * checks at memory_region_init_*()).
+     *
+     * Also, allocation of large amounts of memory may delay
+     * chardev initialization for too long, and trigger timeouts
+     * on software that waits for a monitor socket to be created
+     * (e.g. libvirt).
+     */
+    if (g_str_has_prefix(type, "memory-backend-")) {
+        return false;
+    }
+
     return true;
 }