Re: pte_chain leak in rmap code (2.5.31)

[Andrew Morton <akpm@zip.com.au>]

Rik van Riel wrote:
> 
> On Mon, 12 Aug 2002, Thomas Molina wrote:
> > On Mon, 12 Aug 2002, Rik van Riel wrote:
> > > On Mon, 12 Aug 2002, Christian Ehrhardt wrote:
> > >
> > > > Note the strange use of continue and break which both achieve the same!
> > > > What was meant to happen (judging from rmap-13c) is that we break
> > > Excellent hunting!   Thank you!
> > Any chance this is the cause of the following?
> 
> Yes, quite possible.
> 

Well Adam reported it against the patched version, in which
is appears that I accidentally fixed that bug.  So we may
yet have a problem:

	for (pc = start; pc; pc = next_pc) {
		int i;

		next_pc = pc->next;
		if (next_pc)
			prefetch(next_pc);
		for (i = 0; i < NRPTE; i++) {
			pte_t *p = pc->ptes[i];

			if (!p)
				continue;
			if (victim_i == -1) 
				victim_i = i;

			switch (try_to_unmap_one(page, p)) {
			case SWAP_SUCCESS:
				/*
				 * Release a slot.  If we're releasing the
				 * first pte in the first pte_chain then
				 * pc->ptes[i] and start->ptes[victim_i] both
				 * refer to the same thing.  It works out.
				 */
				pc->ptes[i] = start->ptes[victim_i];
				start->ptes[victim_i] = NULL;
				dec_page_state(nr_reverse_maps);
				victim_i++;
				if (victim_i == NRPTE) {
					page->pte.chain = start->next;
					pte_chain_free(start);
					start = page->pte.chain;
					victim_i = 0;
				}
				break;
			case SWAP_AGAIN:
				/* Skip this pte, remembering status. */
				ret = SWAP_AGAIN;
				continue;
			case SWAP_FAIL:
				ret = SWAP_FAIL;
				goto out;
			case SWAP_ERROR:
				ret = SWAP_ERROR;
				goto out;
			}
		}
	}
out:
	return ret;
}