On Wed, 2002-01-02 at 10:53, James A Goodwin wrote:
> I have two questions about I/O using the Linux XFS DMAPI implementation:
>
> 1. Some filesystems store a little bit of the file data in the i-node.
> This results in the first X bytes staying resident even after a
> dm_punch_hole command tries to free them. Is this the case with XFS, and
> if so, how many bytes are stored this way?
There is never any file data in the xfs inode - the original data
structures allowed for it, but it was never implemented. Things like
mmap etc probably got too nasty.
>
> 2. I just noticed that the DM_WRITE_SYNC flag is not supported for calls to
> dm_write_invis(). This is a fairly critical piece of missing
> functionality. Is there any chance this flag may be implemented in the
> future? I am currently porting a DMAPI application from another platform,
> and this could cause some major problems.
Dean will have to answer that one.
Steve
>
> Thanks,
>
> -James Goodwin
> Software Engineer
> IBM Global Services - Federal
> jagoodwi@xxxxxxxxxx
> Phone: (281) 336 2578
> Fax: (281) 335 4231
> T/L 260-2578
--
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx
|