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

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Wed, 2006-12-20 at 21:36 -0500, Trond Myklebust wrote:
> On Wed, 2006-12-20 at 23:15 +0100, Peter Zijlstra wrote:
> > I think this is also needed:
> 
> NAK
> 
> invalidate_inode_pages2() should _not_ be pretending that dirty pages
> are clean. This patch is incorrect both for the NFS usage and for the
> directIO usage.
> 
> In the latter case, if someone has the page mmapped, resulting in the
> page getting marked as dirty _after_ a directIO write, then it would be
> wrong to discard that data. Only dirty data from _before_ the directIO
> write should needs to be discarded (and that is achieved by unmapping,
> then cleaning the page prior to the directIO call)...
> 
> For the NFS case, the race is a bit more tricky, since you have the
> "unstable write" case which means that the page is neither marked as
> dirty, nor is entirely clean ('cos we don't know that the server has
> committed the data to permanent storage yet).

Then this patch:
http://kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.20-rc1/2.6.20-rc1-mm1/broken-out/nfs-fix-nr_file_dirty-underflow.patch

is equally wrong, right?

-
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/