[PATCH FOR REVIEW] Potential Buffer Overflow in java_props_md.c

Alan Bateman Alan.Bateman at oracle.com
Fri Aug 3 10:40:16 UTC 2012


On 02/08/2012 03:14, David Holmes wrote:
> Andrew et al,
>
> AFAICS here:
>
>       220     encoding_variant = malloc(strlen(temp)+1);
>       221     if (encoding_variant == NULL) {
>       222         JNU_ThrowOutOfMemoryError(env, NULL);
>       223         return 0;
>       224     }
>
> we also need to do free(temp). Similarly later where we return with 
> OOM due to realloc failure, don't we also need to free what was 
> previously malloc'd?
>
> David
I see there are follow-on mails to this but just to say that this is 
System.initProperties time so if there is a malloc failure this early in 
startup then it will cause the VM initialization to fail. So for 
completeness (and perhaps native memory leak detection tools) then I 
agree but if we do have problems here then we aren't go to go very far.

-Alan.



More information about the core-libs-dev mailing list