I was just wondering how bad the current way of writing out dirty pages is
wrt multiple page_launder() users.
We don't remove a dirty page from the inactive dirty list when writing it
out (as opposed to "direct" page->buffers ll_rw_block() IO).
When we have multiple users inside page_launder(), that means a dirty page
which is being written out (and has an additional reference gotten by the
writer) but has no page->buffers mapping yet will be moved to the
beginning of the active list and kept there until the reference is
released by the writer (since refill_inactive_scan() will not move it back
to the inactive dirty list because of the extra reference).
Remeber that we limit the amount of swap writeout's at rw_swap_page(), so
any writepage() which blocks there will have its page moved to the
_beginning_ of the active list because it has no page->buffers yet.
Linus, since you wrote that part of the code, I ask you: do you have any
reason to not remove a page being writepage()'d from the
inactive_dirty_list to avoid this kind of problems ?
(the page must be added back to the inactive_dirty_list again after the
writeout, yes).
-
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/