On Tue, 2013-08-13 at 23:22 +0200, Jan Kara wrote:
> Hi,
>
> Neither me nor linux-fsdevel has been CCed on this change. Please do that
Jan,
All the CC in the email you got were from my original email. I did CC
you and linux-fsdevel when I sent this patchset a week ago
(http://oss.sgi.com/archives/xfs/2013-08/msg00171.html).
I am confused on what happened and how you didn't get the original
email. Just now I checked linux-fsdevel archive. I do not see it there
either. Bizarre. (May be something wrong in the way I used
git-send-email)
Sorry.
> next time. Now looking into the patch in xfs mailing list archive I have
> one comment: You declare:
> struct fs_quota_statv {
> __s8 qs_version; /* version for future changes */
> __u8 qs_pad1; /* pad for 16bit alignment */
> __u16 qs_flags; /* FS_QUOTA_.* flags */
> __u32 qs_incoredqs; /* number of dquots incore */
> struct fs_qfilestatv qs_uquota; /* user quota information */
> struct fs_qfilestatv qs_gquota; /* group quota information */
> struct fs_qfilestatv qs_pquota; /* project quota information */
> __s32 qs_btimelimit; /* limit for blks timer */
> __s32 qs_itimelimit; /* limit for inodes timer */
> __s32 qs_rtbtimelimit;/* limit for rt blks timer */
> __u16 qs_bwarnlimit; /* limit for num warnings */
> __u16 qs_iwarnlimit; /* limit for num warnings */
> __u64 qs_pad2[8]; /* for future proofing */
> };
>
> Now do you really need qs_pad2 field? Since the structure is properly
> versioned now, even its size can vary between versions, cannot it?
Yes, it can.
I added the pad based on Dave Chinner's suggestion:
----------
> > Dave:
> > > > future enhancements, maybe we should add 64 bytes of empty
> > > > space at the end of the structure....
> > > Chandra:
> > > Since this version is fully backward compatible, I didn't think a
> > > future pad was needed. Do you want me to add ?
> > Dave:
> > We only really need to change the structure version when we change
> > input parameters, the size or the shape of the structure being
> > passed in from userspace. If we add padding now, then we can expand
> > output of the call without needing to bump the version of the
> > structure. Old code simply won't know (or care) about the new output
> > in the region of the structure it considers empty padding....
> Chandra:
> Ok. I will all 64 bytes of additional padding at the end.
> Otherwise the patch looks fine.
>
----------
His argument convinced me to add the padding. What do you think ?
> Honza
>
|