At the moment object_get_canonical_path() crashes if the object or one
of its parents does not have a parent, for example, a KVM accelerator
object.

This adds a check for obj!=NULL in a loop to prevent the crash.
In order not to return a wrong path, this checks for currently resolved
partial path and does not add a leading slash to tell the reader that
the path is partial as the owner object is detached.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
---

I have not tested the case with obj==NULL and path!=NULL as this is
for objects which have parents which are not attached to the root
and we do not have such objects in current QEMU afaict but I kept it
just in case.

---
Changes:
v3:
* do not check for obj->parent
* return NULL or incomplete path depending on the situation
---
 qom/object.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/qom/object.c b/qom/object.c
index 0fc9720..05138ba 100644
--- a/qom/object.c
+++ b/qom/object.c
@@ -1669,7 +1669,7 @@ gchar *object_get_canonical_path(Object *obj)
     Object *root = object_get_root();
     char *newpath, *path = NULL;
 
-    while (obj != root) {
+    while (obj && obj != root) {
         char *component = object_get_canonical_path_component(obj);
 
         if (path) {
@@ -1684,7 +1684,13 @@ gchar *object_get_canonical_path(Object *obj)
         obj = obj->parent;
     }
 
-    newpath = g_strdup_printf("/%s", path ? path : "");
+    if (obj && path) {
+        newpath = g_strdup_printf("/%s", path);
+    } else if (path) {
+        newpath = g_strdup(path);
+    } else {
+        newpath = NULL;
+    }
     g_free(path);
 
     return newpath;
-- 
2.11.0

On 30/05/2018 09:17, Alexey Kardashevskiy wrote:
> At the moment object_get_canonical_path() crashes if the object or one
> of its parents does not have a parent, for example, a KVM accelerator
> object.
> 
> This adds a check for obj!=NULL in a loop to prevent the crash.
> In order not to return a wrong path, this checks for currently resolved
> partial path and does not add a leading slash to tell the reader that
> the path is partial as the owner object is detached.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> 
> I have not tested the case with obj==NULL and path!=NULL as this is
> for objects which have parents which are not attached to the root
> and we do not have such objects in current QEMU afaict but I kept it
> just in case.
> 
> ---
> Changes:
> v3:
> * do not check for obj->parent
> * return NULL or incomplete path depending on the situation
> ---
>  qom/object.c | 10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/qom/object.c b/qom/object.c
> index 0fc9720..05138ba 100644
> --- a/qom/object.c
> +++ b/qom/object.c
> @@ -1669,7 +1669,7 @@ gchar *object_get_canonical_path(Object *obj)
>      Object *root = object_get_root();
>      char *newpath, *path = NULL;
>  
> -    while (obj != root) {
> +    while (obj && obj != root) {
>          char *component = object_get_canonical_path_component(obj);
>  
>          if (path) {
> @@ -1684,7 +1684,13 @@ gchar *object_get_canonical_path(Object *obj)
>          obj = obj->parent;
>      }
>  
> -    newpath = g_strdup_printf("/%s", path ? path : "");
> +    if (obj && path) {
> +        newpath = g_strdup_printf("/%s", path);
> +    } else if (path) {
> +        newpath = g_strdup(path);
> +    } else {
> +        newpath = NULL;
> +    }
>      g_free(path);
>  
>      return newpath;

I think this is still wrong, as you wouldn't be able to get back these
objects later (e.g. by assigning the result to a link<> property).  I've
posted an alternative patch.

Paolo