Andi Kleen wrote:
>
> On Wed, Mar 29, 2000 at 03:55:28AM -0700, Jeff V. Merkey wrote:
> >
> >
> > Andi Kleen wrote:
> > >
> > > > This is implemented the same way it is in the FAT and NCPFS file
> > > > systems. How do they write the data back when changed?
> > >
> > > I don't know. Unless I'm missing something (which is quite possible)
> > > they don't sync.
> > > Have you tested shared mappings to be functional ?
> > >
> >
> > No, I have not. This could possibly mean that the other FS's are
>
> Shame on me, I missed that:
>
> int nwfs_mmap(struct file * file, struct vm_area_struct * vma)
> {
> struct inode *inode = file->f_dentry->d_inode;
>
> if (vma->vm_flags & VM_SHARED) // only PAGE_COW or read-only supported now
> return -EINVAL;
>
> I guess supporting shared mappings would be useful though for a
> general purpose filesystem, because more and more programs use them.
>
> To fix it I think it is needed to implement unmap and sync
> vma operations that write the data to the file.
>
Be my guest. I'll add your name and code to the NWFS release if you
want to put it in. Pavel can own the mirrorng, and rip it out of the FS
proper and put it at the raid level (and I'll change the FS to use it
this way -- Darren Major's preference anyway) You may own all the mmap,
bmap, smap, etc-map issues (I have not implemented several of them but
need to). I have to spend too much time working on NT file systems
(which are so ugly compared to the simplicity and elegance of Linux
VFS).
Jeff
> That still has some problems over fully page cache integrated filesystems
> like ext2 (user has to call msync to see updates in read/write and it is
> slower), but would at least make them work.
>
> -Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 21:00:24 EST