On Fri, Oct 20, 2023 at 08:20:23AM +0200, Christoph Hellwig wrote:
> On Thu, Oct 19, 2023 at 09:27:48AM -0700, Darrick J. Wong wrote:
> > + size_t first, last;
> > +
> > + first = (void *)xfs_rsumblock_infoptr(bp, infoword) - bp->b_addr;
> > + last = first + sizeof(xfs_suminfo_t) - 1;
>
> > + size_t first, last;
> > +
> > + first = (void *)xfs_rbmblock_wordptr(bp, from) - bp->b_addr;
> > + last = ((void *)xfs_rbmblock_wordptr(bp, next) - 1) - bp->b_addr;
> > +
> > + xfs_trans_log_buf(tp, bp, first, last);
>
> Going to pointers and back looks a bit confusing and rather inefficient
> to me. But given how late we are in the cycle I don't want to derail
> your series, so let's keep this as-is for now, and I'll add a TODO
> list item to my ever growing list to eventually lean this up.
<nod> I think this function ultimately becomes:
/* Log rtbitmap block from the word @from to the byte before @next. */
static inline void
xfs_trans_log_rtbitmap(
struct xfs_rtalloc_args *args,
unsigned int from,
unsigned int next)
{
struct xfs_buf *bp = args->rbmbp;
size_t first = from * sizeof(xfs_rtword_t);
size_t last = next * sizeof(xfs_rtword_t) - 1;
if (xfs_has_rtgroup(args->mp)) {
first += sizeof(struct xfs_rtbuf_blkinfo);
last += sizeof(struct xfs_rtbuf_blkinfo);
}
xfs_trans_log_buf(args->tp, bp, first, last);
}
I'll go play with the compiler to see what asm it generates.
--D
