monitor: Fix find_device_state() for IDs containing slashes

Recent commit 6952026120 "monitor: Tidy up find_device_state()"
assumed the function's argument is "the device's ID or QOM path" (as
documented for device_del).  It's actually either an absolute QOM
path, or a QOM path relative to /machine/peripheral/.  Such a relative
path is a device ID when it doesn't contain a slash.  When it does,
the function now always fails.  Broke iotest 200, which uses relative
path "vda/virtio-backend".

It fails because object_resolve_path_component() resolves just one
component, not a relative path.

The obvious function to resolve relative paths is
object_resolve_path().  It picks a parent automatically.  Too much
magic, we want to specify the parent.  Create new
object_resolve_path_at() for that, and use it in find_device_state().

Reported-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20211019085711.86377-1-armbru@redhat.com>
Tested-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 11 insertions, 0 deletions

diff --git a/qom/object.c b/qom/object.c
index 6be710bc40..4f0677cca9 100644
--- a/qom/object.c
+++ b/qom/object.c
@@ -2144,6 +2144,17 @@ Object *object_resolve_path(const char *path, bool *ambiguous)
     return object_resolve_path_type(path, TYPE_OBJECT, ambiguous);
 }
 
+Object *object_resolve_path_at(Object *parent, const char *path)
+{
+    g_auto(GStrv) parts = g_strsplit(path, "/", 0);
+
+    if (*path == '/') {
+        return object_resolve_abs_path(object_get_root(), parts + 1,
+                                       TYPE_OBJECT);
+    }
+    return object_resolve_abs_path(parent, parts, TYPE_OBJECT);
+}
+
 typedef struct StringProperty
 {
     char *(*get)(Object *, Error **);