Messages in this thread | | | From | (Linus Torvalds) | Subject | Re: page_launder() bug | Date | 7 May 2001 10:59:19 -0700 |
| |
In article <Pine.A41.4.31.0105062307290.59664-100000@pandora.inf.elte.hu>, BERECZ Szabolcs <szabi@inf.elte.hu> wrote: > >there is a bug in page_launder introduced with kernel 2.4.3-ac12.
Yes.
The whole "dead_swap_page" optimization in the -ac tree is apparentrly completely bogus. It caches a value that is not valid: you cannot reliably look at whether the page has buffers etc without holding the page locked.
So calculating "dead_swap_page" without locking the page first is a sure way to cause trouble.
I can see why the bug was introduced: standard kernels _optimistically_ test whether the condition might be true before they lock the page, and decide to not even try to touch pages that look like they are probably going to be considered active.
But it is important to re-calculate the deadness after getting the lock. Before, it was just an informed guess. After the lock, it is knowledge. And you can use informed guesses for heuristics, but you must _not_ use them for any serious decisions.
Alan, please revert.
Linus - 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/
| |