On Fri, 14 Sep 2012 23:08:55 +0100, Rob Bradford <[email protected]> wrote: > From: Rob Bradford <[email protected]> > > gbm_bo_import will fail to produce a valid bo since the buffer is an SHM > buffer. This cause a crash when the NULL bo returned by gbm_bo_import is > dereferenced later.
So since gbm_bo_import() is already verifying that the handle is suitable for the use case, would it not be wiser to check it return in case it has to reject the handle for any other reason, or on the off chance that the scanout can read from any buffer? Not that I'm advocating remapping user vma like that behind the callers back! -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ wayland-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/wayland-devel
