Re: 2.6.19 file content corruption on ext3

From: Linus Torvalds
Date: Sun Dec 17 2006 - 20:23:42 EST




On Sun, 17 Dec 2006, Andrew Morton wrote:
>
> From my quick reading, all callers of try_to_free_buffers() have already
> unmapped the page from pagetables, and given that the reported ext3 corruption
> happens on uniprocessor, non-preempt kernels, I doubt if this patch will fix
> things.

Hmm. One possible explanation: maybe the page actually _did_ get unmapped
from the page tables, but got added back?

I don't think we lock the page when faulting it in (we want it to be
uptodate, but not necessarily locked). So assuming the pageout sequence
always _does_ follow the rule that it only does try_to_free_buffers() on
pages that aren't mapped, what actually protects them from not becoming
mapped (and dirtied) during that sequence?

So we should probably do a "wait_for_page()" in do_no_page()?

Or maybe only do it for write accesses (since we don't really care about
getting mapped readably)? If so, we need to do it in the write case of
do_no_page() _and_ in the do_wp_page() case. Hmm?

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