Re: [PATCH] mm: fix page_mkclean_one (was: 2.6.19 file contentcorruption on ext3)

[Linus Torvalds]

On Wed, 20 Dec 2006, Peter Zijlstra wrote:
>
> I think this is also needed:

Yeah, that looks about right. Although I think it should go above the 
"try_to_release_page()", because right now we do that "ttrp()" with the 
dirty bit set, and we should let the low-level filesystem just know that 
it's simply not interesting any more (and, indeed, "try_to_free_buffers()" 
too, for that matter).

Anyway, I think that's a detail. I'd rather know whether this all actually 
makes any difference what-so-ever to the corruption behaviour of Andrei 
&co. 

Maybe the UP ARM case is some strange dcache alias issue with PIO IDE, and 
the only reason that started showing up at the same time is the different 
IO loads. Who knows.

[ Although I think you may have been on the right track with that dcache 
  flushing stuff in "page_mkclean()".. It might not have been quite 
  all there, but I think we should go back and look very closely at 
  page_mkclean() regardless of any other issues! ]

So far, my whole "cancel_dirty_page/clean_page_dirty_for_io" patch has 
really been just a "try to make the code _look_ sane. I don't think we 
have a single report that the patch actually makes any difference yet.

		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/