| Message ID | 20241119154656.774395-2-bfoster@redhat.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | iomap: zero range folio batch processing prototype | expand |
On Tue, Nov 19, 2024 at 10:46:53AM -0500, Brian Foster wrote: > To facilitate batch processing of dirty folios for zero range, tweak > the write begin path to allow the caller to optionally pass in its > own folio/pos combination. > > Signed-off-by: Brian Foster <bfoster@redhat.com> > --- > fs/iomap/buffered-io.c | 20 +++++++++++++------- > 1 file changed, 13 insertions(+), 7 deletions(-) > > diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c > index ce73d2a48c1e..d1a86aea1a7a 100644 > --- a/fs/iomap/buffered-io.c > +++ b/fs/iomap/buffered-io.c > @@ -781,7 +781,7 @@ static int iomap_write_begin(struct iomap_iter *iter, loff_t pos, > { > const struct iomap_folio_ops *folio_ops = iter->iomap.folio_ops; > const struct iomap *srcmap = iomap_iter_srcmap(iter); > - struct folio *folio; > + struct folio *folio = *foliop; > int status = 0; > > BUG_ON(pos + len > iter->iomap.offset + iter->iomap.length); > @@ -794,9 +794,15 @@ static int iomap_write_begin(struct iomap_iter *iter, loff_t pos, > if (!mapping_large_folio_support(iter->inode->i_mapping)) > len = min_t(size_t, len, PAGE_SIZE - offset_in_page(pos)); > > - folio = __iomap_get_folio(iter, pos, len); > - if (IS_ERR(folio)) > - return PTR_ERR(folio); > + /* > + * XXX: Might want to plumb batch handling down through here. For now > + * let the caller do it. Yeah, plumbing in the batch here would be nicer. I suspect doing batch processing might actually be a neat thing for the normal write patch as well.
On Wed, Nov 20, 2024 at 12:38:41AM -0800, Christoph Hellwig wrote: > On Tue, Nov 19, 2024 at 10:46:53AM -0500, Brian Foster wrote: > > To facilitate batch processing of dirty folios for zero range, tweak > > the write begin path to allow the caller to optionally pass in its > > own folio/pos combination. > > > > Signed-off-by: Brian Foster <bfoster@redhat.com> > > --- > > fs/iomap/buffered-io.c | 20 +++++++++++++------- > > 1 file changed, 13 insertions(+), 7 deletions(-) > > > > diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c > > index ce73d2a48c1e..d1a86aea1a7a 100644 > > --- a/fs/iomap/buffered-io.c > > +++ b/fs/iomap/buffered-io.c > > @@ -781,7 +781,7 @@ static int iomap_write_begin(struct iomap_iter *iter, loff_t pos, > > { > > const struct iomap_folio_ops *folio_ops = iter->iomap.folio_ops; > > const struct iomap *srcmap = iomap_iter_srcmap(iter); > > - struct folio *folio; > > + struct folio *folio = *foliop; > > int status = 0; > > > > BUG_ON(pos + len > iter->iomap.offset + iter->iomap.length); > > @@ -794,9 +794,15 @@ static int iomap_write_begin(struct iomap_iter *iter, loff_t pos, > > if (!mapping_large_folio_support(iter->inode->i_mapping)) > > len = min_t(size_t, len, PAGE_SIZE - offset_in_page(pos)); > > > > - folio = __iomap_get_folio(iter, pos, len); > > - if (IS_ERR(folio)) > > - return PTR_ERR(folio); > > + /* > > + * XXX: Might want to plumb batch handling down through here. For now > > + * let the caller do it. > > Yeah, plumbing in the batch here would be nicer. > Ok, I'll take a closer look at that. IIRC I punted on it initially because we'll also have to fix up pos/len down in this path when the next folio is not contiguous. This was easier to just get the basics working. Brian > I suspect doing batch processing might actually be a neat thing for > the normal write patch as well. >
diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c index ce73d2a48c1e..d1a86aea1a7a 100644 --- a/fs/iomap/buffered-io.c +++ b/fs/iomap/buffered-io.c @@ -781,7 +781,7 @@ static int iomap_write_begin(struct iomap_iter *iter, loff_t pos, { const struct iomap_folio_ops *folio_ops = iter->iomap.folio_ops; const struct iomap *srcmap = iomap_iter_srcmap(iter); - struct folio *folio; + struct folio *folio = *foliop; int status = 0; BUG_ON(pos + len > iter->iomap.offset + iter->iomap.length); @@ -794,9 +794,15 @@ static int iomap_write_begin(struct iomap_iter *iter, loff_t pos, if (!mapping_large_folio_support(iter->inode->i_mapping)) len = min_t(size_t, len, PAGE_SIZE - offset_in_page(pos)); - folio = __iomap_get_folio(iter, pos, len); - if (IS_ERR(folio)) - return PTR_ERR(folio); + /* + * XXX: Might want to plumb batch handling down through here. For now + * let the caller do it. + */ + if (!folio) { + folio = __iomap_get_folio(iter, pos, len); + if (IS_ERR(folio)) + return PTR_ERR(folio); + } /* * Now we have a locked folio, before we do anything with it we need to @@ -918,7 +924,7 @@ static loff_t iomap_write_iter(struct iomap_iter *iter, struct iov_iter *i) unsigned int bdp_flags = (iter->flags & IOMAP_NOWAIT) ? BDP_ASYNC : 0; do { - struct folio *folio; + struct folio *folio = NULL; loff_t old_size; size_t offset; /* Offset into folio */ size_t bytes; /* Bytes to write to folio */ @@ -1281,7 +1287,7 @@ static loff_t iomap_unshare_iter(struct iomap_iter *iter) return length; do { - struct folio *folio; + struct folio *folio = NULL; int status; size_t offset; size_t bytes = min_t(u64, SIZE_MAX, length); @@ -1385,7 +1391,7 @@ static loff_t iomap_zero_iter(struct iomap_iter *iter, bool *did_zero, } do { - struct folio *folio; + struct folio *folio = NULL; int status; size_t offset; size_t bytes = min_t(u64, SIZE_MAX, length);
To facilitate batch processing of dirty folios for zero range, tweak the write begin path to allow the caller to optionally pass in its own folio/pos combination. Signed-off-by: Brian Foster <bfoster@redhat.com> --- fs/iomap/buffered-io.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-)