On Fri, Feb 6, 2015 at 7:20 AM, Sasha Levin [off-list ref] wrote:

Can we modify it slightly to avoid potentially accessing invalid memory:

So I think there's a race with that.

And I'll warn you: the kernel does do speculative reads of memory that
might be invalid, not just in places like this. See the comment in
get_user_huge_page() for example, where we knowingly do speculative
reads, but hide it if DEBUG_PAGEALLOC is set.

More commonly, CONFIG_DCACHE_WORD_ACCESS is very much about doing
speculative reads. Now, that access is hidden inside an asm, so KASan
won't see it, but there might well be others.

You probably don't see them very much just because they are so rarely
a problem, and most of the time it's not to other processes stack but
to allocated structures where freeing takes long enough to basically
hide any small race..

In other words: I suspect it would be good to instead just teach KASan
about "this is a speculative read" and just suppress the warning for
those instead.

                             Linus