Re: Race in new page migration code?

[Hugh Dickins <hugh@veritas.com>]

On Sun, 15 Jan 2006, Christoph Lameter wrote:
> On Sun, 15 Jan 2006, Hugh Dickins wrote:
> > 
> > Good.  And whether it's your or Nick's patch that goes in, please also
> > remove that PageReserved test which you recently put in check_pte_range.
> 
> Zero pages are still marked reserved AFAIK. Why not check for it?

Indeed they are, at present and quite likely into posterity.  But
they're not a common case here, and migrate_page_add now handles them
silently, so why bother to complicate it with an unnecessary check?

You added the test because the WARN_ON fired, you're now doing the
right thing and removing the WARN_ON because it was inevitably racy,
so it would make sense to remove the PageReserved test too.

If you look through the rest of mm/, you'll find 2.6.15 removed all
the PageReserved tests, except at the low level in page_alloc.c:
so it's retrograde for you to be adding one back here.  Testing
PageLRU would be more relevant, if you insist on such a test.

Or have you found the zero page mapcount distorting get_stats stats?
If that's an issue, then better add a commented test for it there.

Hmm, that battery of unusual tests at the start of migrate_page_add
is odd: the tests don't quite match the comment, and it isn't clear
what reasoning lies behind the comment anyway.

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