lkml.org 
[lkml]   [2000]   [Oct]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: 2.4.0-test10-pre3:Oops in mm/filemap.c:filemap_write_pa


On 20 Oct 2000, Trond Myklebust wrote:
>
> Under NFS the problem is that pages can (and *should*) be invalidated
> despite there being pending write backs. The server can trigger the
> need for a cache invalidation at any time.
> The existence of file locks that aren't page aligned, as well as
> partial page writeback ensure that we cannot make the equivalence
> page has pending write == page is correct.

Note that this is not what my patch really says at all.

My patch says "ok, we should invalidate this page, because somebody thinks
it may be bad. HOWEVER, we cannot reasonably do that, because we have
users of the page (which may be completely unrelated to pending writes),
and invalidating this page is _certain_ to cause corruption".

Basically think of the case of somebody having a shared mmap over NFS. He
has dirty data in his pages, and he expects those to be written out to the
server. There is NO QUESTION about this fact.

Imagine somebody (possibly even the same person) then releasing or getting
a file lock, possibly in some other part of the file.

Without that added test, all those dirty local pages just got completely
thrown away.

Now, you have a choice: you can KNOW that you're doing something horribly
and utterly wrong (throwing away data), or you can SUSPECT that you might
cause non-coherency between the client and the server. You cannot get
both.

I'm saying that we're much much better off guaranteeing local consistency
over knowingly breaking local consistency over a uncertain global
consistency issue. Especially as NFS has never guaranteed global
consistency in the first place, and was not designed to do that.

Linus

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

\
 
 \ /
  Last update: 2005-03-22 12:41    [W:0.301 / U:0.832 seconds]
©2003-2014 Jasper Spaans. Advertise on this site