yes, we may have. But no such spec or requirement that I am aware of.
Returning EFAULT is being nice to the stupid user, but SIGSEGV'ing the
offender (e.g. if she overflows the stack) is not breaking any standards.
> At one time there was an optimized path for the double-copy: do the
> verify_area early, then bypass it when doing the real copy later. I guess
> this path no longer exists?
It is still done in several places. For example, have a look at
generic_file_read() in mm/filemap.c. It uses verify_area() at the top and
uses __copy_to_user() in the actual actor (file_read_actor).
Regards,
Tigran.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/