Re: [patch] fix for OOM deadlock in swap_in (2.2.10) [Re: [test

Linus Torvalds (torvalds@transmeta.com)
Sun, 4 Jul 1999 10:49:56 -0700 (PDT)


On Sun, 4 Jul 1999, Andrea Arcangeli wrote:
>
> The first patch I sent you some time ago was buggy since I replaced the
> sigbus with a sigkill in do_page_fault, but now I force the signals only
> at the lower level (as shm and other places was just doing) and the retval
> of handle_mm_fault now _only_ tells do_page_fault if it has to fixup or
> not.

Ok. I still have your old patch, I'll just flush it so I don't confuse it
with anything else.

However, I still much prefer the 2.3.x approach (ie just returning more
than just 0/1 - a negative number means out-of-memory). In particular,
your current approach gets the ptrace() case wrong for the SIGBUS case,
and it's pretty much impossible to fix cleanly as far as I can tell.

Note that 2.3.10-pre2 also gets ptrace() wrong, but at least it's not
impossible to fix - it should just bother to check the return value it
gets from handle_mm_fault(). Right now it doesn't.

Note that ptrace() is a horrible special case, being the only thing that
accesses another process VM space (apart from vmscan which is also
horrible, in other ways). HOWEVER, it's rather bad to have a SIGBUS
problem and then when you try to debug it the debugger also gets a SIGBUS,
which is what your approach results in.

Linus

-
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/