On 09/10/2015 08:48, Chris Wilson wrote:
On Thu, Oct 08, 2015 at 07:31:37PM +0100, Tomas Elf wrote:
Sometimes the iterated vma objects are NULL apparently. Be aware of this while
iterating and do early exit instead of causing a NULL pointer exception.
Wrong.
-Chris
So the NULL pointer exception that I ran into multiple times during
several different test runs on the line that says "vma->pin_count" was
not because the vma pointer was NULL. Would you mind sharing your
explanation to how this might have happened? Is it because we're not
synchronizing and there's no protection against the driver deallocating
vmas while we're trying to capture them? If this all ties into the
aforementioned RCU-based solution then maybe we should go with that then.
Thanks,
Tomas
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
