Added some relevant lists to CC (please don't forget about this).
On Tue 24-12-13 17:44:35, Verma, Vishal L wrote:
> I would like to attend LSF/MM 2014 - I am especially interested in
> discussions around persistent memory. I have been studying EXT4 and
> researching on what we'd have to do to best enable it for
> byte-addressable persistent memory.
Do you have some concrete suggestions or observations regarding this?
Because discussing abstract ideas without actually trying something out
beforehand doesn't usually result in a useful discussion...
> I am relatively new in the Linux kernel world, and my previous
> experience has been with device driver work - I wrote a SCSI SG_IO
> translation layer for the NVMe driver. I also contributed to PMFS -
> https://github.com/linux-pmfs/pmfs which was our go-to route for
> enabling PM in the kernel before we decided EXT4 would be a better
> choice.
I would be interested what were your reasons for the decision. Can you
elaborate a bit (that's just my personal curiosity)?
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
On Mon, 2013-12-30 at 22:04 +0100, Jan Kara wrote:
> Added some relevant lists to CC (please don't forget about this).
Thanks, I wasn't sure I had needed to, for an attend request, but it
makes sense now.
>
> On Tue 24-12-13 17:44:35, Verma, Vishal L wrote:
> > I would like to attend LSF/MM 2014 - I am especially interested in
> > discussions around persistent memory. I have been studying EXT4 and
> > researching on what we'd have to do to best enable it for
> > byte-addressable persistent memory.
> Do you have some concrete suggestions or observations regarding this?
> Because discussing abstract ideas without actually trying something out
> beforehand doesn't usually result in a useful discussion...
I've started looking into the EXT4 layout to see if we'd need any
changes there, especially for optimizing huge page mappings. This is
really a work in progress, and I don't yet have any observations. I was
hoping to be involved in the discussions Matthew proposed here:
http://thread.gmane.org/gmane.linux.file-systems/80965
>
> > I am relatively new in the Linux kernel world, and my previous
> > experience has been with device driver work - I wrote a SCSI SG_IO
> > translation layer for the NVMe driver. I also contributed to PMFS -
> > https://github.com/linux-pmfs/pmfs which was our go-to route for
> > enabling PM in the kernel before we decided EXT4 would be a better
> > choice.
> I would be interested what were your reasons for the decision. Can you
> elaborate a bit (that's just my personal curiosity)?
Certainly. The crux of it was that we'd have to spend a lot more time
getting things related to file system 'correctness' fixed, such as
locking in the journaling code, that other file systems already have
figured out. EXT4 is already 'trusted', and something completely new
would take a long time for adoption.
-Vishal