netfs: Implement a write-through caching option

Provide a flag whereby a filesystem may request that cifs_perform_write()
perform write-through caching.  This involves putting pages directly into
writeback rather than dirty and attaching them to a write operation as we
go.

Further, the writes being made are limited to the byte range being written
rather than whole folios being written.  This can be used by cifs, for
example, to deal with strict byte-range locking.

This can't be used with content encryption as that may require expansion of
the write RPC beyond the write being made.

This doesn't affect writes via mmap - those are written back in the normal
way; similarly failed writethrough writes are marked dirty and left to
writeback to retry.  Another option would be to simply invalidate them, but
the contents can be simultaneously accessed by read() and through mmap.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: linux-cachefs@redhat.com
cc: linux-fsdevel@vger.kernel.org
cc: linux-mm@kvack.org

1 files changed, 59 insertions, 10 deletions

diff --git a/fs/netfs/buffered_write.c b/fs/netfs/buffered_write.c
index 50be8fe3ca43..6ca6c4bde5eb 100644
--- a/fs/netfs/buffered_write.c
+++ b/fs/netfs/buffered_write.c
@@ -26,6 +26,8 @@ enum netfs_how_to_modify {
 	NETFS_FLUSH_CONTENT,		/* Flush incompatible content. */
 };
 
+static void netfs_cleanup_buffered_write(struct netfs_io_request *wreq);
+
 static void netfs_set_group(struct folio *folio, struct netfs_group *netfs_group)
 {
 	if (netfs_group && !folio_get_private(folio))
@@ -133,6 +135,14 @@ ssize_t netfs_perform_write(struct kiocb *iocb, struct iov_iter *iter,
 	struct inode *inode = file_inode(file);
 	struct address_space *mapping = inode->i_mapping;
 	struct netfs_inode *ctx = netfs_inode(inode);
+	struct writeback_control wbc = {
+		.sync_mode	= WB_SYNC_NONE,
+		.for_sync	= true,
+		.nr_to_write	= LONG_MAX,
+		.range_start	= iocb->ki_pos,
+		.range_end	= iocb->ki_pos + iter->count,
+	};
+	struct netfs_io_request *wreq = NULL;
 	struct netfs_folio *finfo;
 	struct folio *folio;
 	enum netfs_how_to_modify howto;
@@ -143,6 +153,30 @@ ssize_t netfs_perform_write(struct kiocb *iocb, struct iov_iter *iter,
 	size_t max_chunk = PAGE_SIZE << MAX_PAGECACHE_ORDER;
 	bool maybe_trouble = false;
 
+	if (unlikely(test_bit(NETFS_ICTX_WRITETHROUGH, &ctx->flags) ||
+		     iocb->ki_flags & (IOCB_DSYNC | IOCB_SYNC))
+	    ) {
+		if (pos < i_size_read(inode)) {
+			ret = filemap_write_and_wait_range(mapping, pos, pos + iter->count);
+			if (ret < 0) {
+				goto out;
+			}
+		}
+
+		wbc_attach_fdatawrite_inode(&wbc, mapping->host);
+
+		wreq = netfs_begin_writethrough(iocb, iter->count);
+		if (IS_ERR(wreq)) {
+			wbc_detach_inode(&wbc);
+			ret = PTR_ERR(wreq);
+			wreq = NULL;
+			goto out;
+		}
+		if (!is_sync_kiocb(iocb))
+			wreq->iocb = iocb;
+		wreq->cleanup = netfs_cleanup_buffered_write;
+	}
+
 	do {
 		size_t flen;
 		size_t offset;	/* Offset into pagecache folio */
@@ -315,7 +349,25 @@ ssize_t netfs_perform_write(struct kiocb *iocb, struct iov_iter *iter,
 		}
 		written += copied;
 
-		folio_mark_dirty(folio);
+		if (likely(!wreq)) {
+			folio_mark_dirty(folio);
+		} else {
+			if (folio_test_dirty(folio))
+				/* Sigh.  mmap. */
+				folio_clear_dirty_for_io(folio);
+			/* We make multiple writes to the folio... */
+			if (!folio_test_writeback(folio)) {
+				folio_wait_fscache(folio);
+				folio_start_writeback(folio);
+				folio_start_fscache(folio);
+				if (wreq->iter.count == 0)
+					trace_netfs_folio(folio, netfs_folio_trace_wthru);
+				else
+					trace_netfs_folio(folio, netfs_folio_trace_wthru_plus);
+			}
+			netfs_advance_writethrough(wreq, copied,
+						   offset + copied == flen);
+		}
 	retry:
 		folio_unlock(folio);
 		folio_put(folio);
@@ -325,17 +377,14 @@ ssize_t netfs_perform_write(struct kiocb *iocb, struct iov_iter *iter,
 	} while (iov_iter_count(iter));
 
 out:
-	if (likely(written)) {
-		/* Flush and wait for a write that requires immediate synchronisation. */
-		if (iocb->ki_flags & (IOCB_DSYNC | IOCB_SYNC)) {
-			_debug("dsync");
-			ret = filemap_fdatawait_range(mapping, iocb->ki_pos,
-						      iocb->ki_pos + written);
-		}
-
-		iocb->ki_pos += written;
+	if (unlikely(wreq)) {
+		ret = netfs_end_writethrough(wreq, iocb);
+		wbc_detach_inode(&wbc);
+		if (ret == -EIOCBQUEUED)
+			return ret;
 	}
 
+	iocb->ki_pos += written;
 	_leave(" = %zd [%zd]", written, ret);
 	return written ? written : ret;