On Wed, Jan 22, 2020 at 02:57:55PM +0000, David Howells wrote: > An alternative could be to have an "ITER_ARRAY" that just has the page > pointers and not the offset/length info. This decreases the redundancy and > increases the max payload-per-array-page to 2M. We could also have an ITER_XARRAY which you just pass &mapping->i_pages to. I don't think you use any other part of the mapping, so that would be a more generic version that is equally efficient. > + rcu_read_lock(); \ > + for (page = xas_load(&xas); page; page = xas_next(&xas)) { \ > + if (xas_retry(&xas, page)) \ > + continue; \ > + if (xa_is_value(page)) \ > + break; \ Do you also want to check for !page? That would be a bug in the caller. > + if (PageCompound(page)) \ > + break; \ It's perfectly legal to have compound pages in the page cache. Call find_subpage(page, xas.xa_index) unconditionally. > + if (page_to_pgoff(page) != xas.xa_index) \ > + break; \ ... and you can ditch this if the pages are pinned as find_subpage() will bug in this case. > + __v.bv_page = page; \ > + offset = (i->mapping_start + skip) & ~PAGE_MASK; \ > + seg = PAGE_SIZE - offset; \ > + __v.bv_offset = offset; \ > + __v.bv_len = min(n, seg); \ > + (void)(STEP); \ > + n -= __v.bv_len; \ > + skip += __v.bv_len; \ Do we want STEP to be called with PAGE_SIZE chunks, or if they have a THP, can we have it called with larger than a PAGE_SIZE chunk? > +#define iterate_all_kinds(i, n, v, I, B, K, M) { \ > if (likely(n)) { \ > size_t skip = i->iov_offset; \ > if (unlikely(i->type & ITER_BVEC)) { \ > @@ -86,6 +119,9 @@ > struct kvec v; \ > iterate_kvec(i, n, v, kvec, skip, (K)) \ > } else if (unlikely(i->type & ITER_DISCARD)) { \ > + } else if (unlikely(i->type & ITER_MAPPING)) { \ > + struct bio_vec v; \ > + iterate_mapping(i, n, v, skip, (M)); \ bio_vec? > -#define iterate_and_advance(i, n, v, I, B, K) { \ > +#define iterate_and_advance(i, n, v, I, B, K, M) { \ > if (unlikely(i->count < n)) \ > n = i->count; \ > if (i->count) { \ > @@ -119,6 +155,9 @@ > i->kvec = kvec; \ > } else if (unlikely(i->type & ITER_DISCARD)) { \ > skip += n; \ > + } else if (unlikely(i->type & ITER_MAPPING)) { \ > + struct bio_vec v; \ > + iterate_mapping(i, n, v, skip, (M)) \ again?