Re: [patch 2.6.13-rc4] fix get_user_pages bug

[Hugh Dickins]

On Tue, 2 Aug 2005, Linus Torvalds wrote:
> 
> On the other hand, this being s390, maybe nobody cares?

You have a cruel streak.

But have I just realized a non-s390 problem with your pte_dirty
technique?  The ptep_set_wrprotect in fork's copy_one_pte.

That's specifically write-protecting the pte to force COW, but
leaving the dirty bit: so now get_user_pages will skip COW-ing it
(in all write cases, not just the peculiar ptrace force one).

We really do need to COW those in get_user_pages, don't we?
And although we could handle it by clearing dirty and doing a
set_page_dirty, it's a path we don't want to clutter further.

(Yes, there's another issue of a fork occurring while the pages
are already under get_user_pages, which is a significant issue for
InfiniBand; but I see that as a separate kind of race, which we can
reasonably disregard in this discussion - though it will need proper
attention, perhaps through Michael Tsirkin's PROT_DONTCOPY patch.)

The simple answer to Robin's pagetable update race is to say that
anyone using ptrace should be prepared for a pt race ;)

Hugh
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/