David Chinner wrote:
> It's a good first cut - comments are inline in the patch below.
>
>
>   
> Looks like some whitespace problems there (mixing spaces and tabs).
> Also, can you include empty lines either side of a unique hunk of code
> like this?
>
> I wonder how many other places we are going to have to put this check?
> I haven't looked myself, but this is a good place to start ;)
>
>   

I'm still wondering how should I educate emacs to introduce alsways tabs ;)
> You shouldn't renumber the existing ioctls - that changes the
> interfaces to userspace and so will break lots of stuff :( Just put
> them at the end as 126/127. (Oh, you've got two "111" ioctls in
> there, anyway ;)
>   

> IOWs, the flag passed to xfs_trans_alloc() defines the type of the
> transaction and the flag passed to xfs_alloc_log_agf() defines what
> has been modified within the transaction.
>
>   
Ok, thank you for the explanation, I think that now I got it right.
Attached is a new patch.

There is one question that I would like to ask: when you sketched the 
xfs_alloc_set_flag_ag function, you put inside it the call to the 
funcintion xfs_alloc_log_agf (see next code snippet).

STATIC void
xfs_alloc_set_flag_ag(
 xfs_trans_t *tp,
 xfs_buf_t *agbp, /* buffer for a.g. freelist header */
 xfs_perag_t *pag,
 int flag)
{
 xfs_agf_t *agf; /* a.g. freespace structure */

 agf = XFS_BUF_TO_AGF(agbp);
 pag->pagf_flags |= flag;
 agf->agf_flags = cpu_to_be32(pag->pagf_flags);

 xfs_alloc_log_agf(tp, agbp, XFS_AGF_FLAGS); <-- ***** FROM HERE
}

is it required to do the transaction log right after the change or can it be
done in the caller function right after calling xfs_alloc_set_flag_ag?

For example

caller(...)

{
xfs_alloc_set_flag_ag(tp, bp, pag, XFS_AGFLAG_ALLOC_DENY);

<-- **** TO HERE

xfs_trans_set_sync(tp);
xfs_trans_commit(tp, 0);
}

Thanks

-- 
Rubén Porras
LinWorks GmbH