On Fri 30-09-16 01:59:05, Christoph Hellwig wrote:
> On Tue, Sep 27, 2016 at 06:43:29PM +0200, Jan Kara wrote:
> > Hello,
> >
> > these patches fix races when invalidating hole pages in DAX mappings. See
> > changelogs for details. The series is based on my patches to write-protect
> > DAX PTEs because we really need to closely track dirtiness (and cleanness!)
> > of radix tree entries in DAX mappings in order to avoid discarding valid
> > dirty bits leading to missed cache flushes on fsync(2).
>
> Except for the VM magic I don't feel qualified to review this looks fine
> to me. But don't we need to switch ext4 away from going through the DIO
> path first before the limited invalidation is effective there?
> Otherwise we'll still get the full range invalidation from
> generic_file_direct_write?
After patch 5/6 the additional invalidation we get from
generic_file_direct_write() for ext4 is racy but not doing any harm since
we later do a reliable invalidation directly from dax_io() which patch 6/6
takes care to update. So it will be inefficient but it will work. Once ext4
is converted we get rid of this redundancy.
BTW I have locally patches that avoid this inefficiency for ext4 but I
figured there's no point in posting them since we want to convert ext4 to
iomap interface sooner rather than later.
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
