[PATCH 11/13] afs: Copy local writes to the cache when writing to the server

From: David Howells <dhowells@redhat.com>
To: Trond Myklebust <trondmy@hammerspace.com>,
	Anna Schumaker <anna.schumaker@netapp.com>,
	Steve French <sfrench@samba.org>,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	Matthew Wilcox <willy@infradead.org>
Cc: Jeff Layton <jlayton@redhat.com>,
	Dave Wysochanski <dwysocha@redhat.com>,
	dhowells@redhat.com, linux-cachefs@redhat.com,
	linux-afs@lists.infradead.org, linux-nfs@vger.kernel.org,
	linux-cifs@vger.kernel.org, ceph-devel@vger.kernel.org,
	v9fs-developer@lists.sourceforge.net,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: [PATCH 11/13] afs: Copy local writes to the cache when writing to the server
Date: Mon, 13 Jul 2020 17:39:07 +0100	[thread overview]
Message-ID: <159465834740.1377938.16996503111636748338.stgit@warthog.procyon.org.uk> (raw)
In-Reply-To: <159465821598.1377938.2046362270225008168.stgit@warthog.procyon.org.uk>

When writing to the server from afs_writepage() or afs_writepages(), copy
the data to the cache object too.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/write.c |  132 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 128 insertions(+), 4 deletions(-)

diff --git a/fs/afs/write.c b/fs/afs/write.c
index cb27027c06bb..10c60837775e 100644
--- a/fs/afs/write.c
+++ b/fs/afs/write.c
@@ -13,6 +13,9 @@
 #include <linux/pagevec.h>
 #include "internal.h"
 
+static void afs_write_to_cache(struct afs_vnode *vnode,
+			       pgoff_t start, pgoff_t end, loff_t a, loff_t b);
+
 /*
  * mark a page as having been made dirty and thus needing writeback
  */
@@ -427,6 +430,8 @@ static int afs_write_back_from_locked_page(struct address_space *mapping,
 	count = 1;
 	if (test_set_page_writeback(primary_page))
 		BUG();
+	if (TestSetPageFsCache(primary_page))
+		BUG();
 
 	/* Find all consecutive lockable dirty pages that have contiguous
 	 * written regions, stopping when we find a page that is not
@@ -475,7 +480,8 @@ static int afs_write_back_from_locked_page(struct address_space *mapping,
 				break;
 			if (!trylock_page(page))
 				break;
-			if (!PageDirty(page) || PageWriteback(page)) {
+			if (!PageDirty(page) || PageWriteback(page) ||
+			    PageFsCache(page)) {
 				unlock_page(page);
 				break;
 			}
@@ -497,6 +503,8 @@ static int afs_write_back_from_locked_page(struct address_space *mapping,
 				BUG();
 			if (test_set_page_writeback(page))
 				BUG();
+			if (TestSetPageFsCache(page))
+				BUG();
 			unlock_page(page);
 			put_page(page);
 		}
@@ -534,6 +542,11 @@ static int afs_write_back_from_locked_page(struct address_space *mapping,
 		end = i_size;
 
 	if (pos < i_size) {
+		/* Speculatively write to the cache.  We have to fix this up
+		 * later if the store fails.
+		 */
+		afs_write_to_cache(vnode, first, last, pos, end);
+
 		iov_iter_mapping(&iter, WRITE, mapping, pos, end - pos);
 		ret = afs_store_data(vnode, &iter, pos, first, last);
 	} else {
@@ -592,6 +605,10 @@ int afs_writepage(struct page *page, struct writeback_control *wbc)
 
 	_enter("{%lx},", page->index);
 
+#ifdef CONFIG_AFS_FSCACHE
+	wait_on_page_fscache(page);
+#endif
+
 	ret = afs_write_back_from_locked_page(page->mapping, wbc, page,
 					      wbc->range_end >> PAGE_SHIFT);
 	if (ret < 0) {
@@ -619,7 +636,7 @@ static int afs_writepages_region(struct address_space *mapping,
 
 	do {
 		n = find_get_pages_range_tag(mapping, &index, end,
-					PAGECACHE_TAG_DIRTY, 1, &page);
+					     PAGECACHE_TAG_DIRTY, 1, &page);
 		if (!n)
 			break;
 
@@ -644,10 +661,14 @@ static int afs_writepages_region(struct address_space *mapping,
 			continue;
 		}
 
-		if (PageWriteback(page)) {
+		if (PageWriteback(page) || PageFsCache(page)) {
 			unlock_page(page);
-			if (wbc->sync_mode != WB_SYNC_NONE)
+			if (wbc->sync_mode != WB_SYNC_NONE) {
 				wait_on_page_writeback(page);
+#ifdef CONFIG_AFS_FSCACHE
+				wait_on_page_fscache(page);
+#endif
+			}
 			put_page(page);
 			continue;
 		}
@@ -868,3 +889,106 @@ int afs_launder_page(struct page *page)
 	ClearPagePrivate(page);
 	return ret;
 }
+
+/*
+ * Clear the PG_fscache flag from a sequence of pages and wake up anyone who's
+ * waiting.  The last page is included in the sequence.
+ */
+static void afs_clear_fscache_bits(struct address_space *mapping,
+				   pgoff_t start, pgoff_t last)
+{
+	struct page *page;
+
+	XA_STATE(xas, &mapping->i_pages, start);
+
+	rcu_read_lock();
+	xas_for_each(&xas, page, last) {
+		unlock_page_fscache(page);
+	}
+	rcu_read_unlock();
+}
+
+/*
+ * Deal with the completion of writing the data to the cache.
+ */
+static void afs_write_to_cache_done(struct fscache_io_request *_req)
+{
+	struct afs_read *req = container_of(_req, struct afs_read, cache);
+	pgoff_t index = req->cache.pos >> PAGE_SHIFT;
+	pgoff_t last = index + req->cache.nr_pages - 1;
+
+	_enter("%lx,%x,%llx", index, req->cache.nr_pages, req->cache.transferred);
+
+	afs_clear_fscache_bits(req->cache.mapping, index, last);
+
+	if (req->cache.error && req->cache.error != -ENOBUFS) {
+		struct afs_vnode *vnode = req->vnode;
+		struct afs_vnode_cache_aux aux = {
+			.data_version = vnode->status.data_version,
+		};
+		_debug("inval wr %d", req->cache.error);
+		fscache_invalidate(req->cache.cookie, &aux,
+				   i_size_read(&vnode->vfs_inode), 0);
+	}
+}
+
+static const struct fscache_io_request_ops afs_write_req_ops = {
+	.get		= afs_req_get,
+	.put		= afs_req_put,
+};
+
+/*
+ * Save the write to the cache also.
+ */
+static void afs_write_to_cache(struct afs_vnode *vnode,
+			       pgoff_t start, pgoff_t last, loff_t a, loff_t b)
+{
+	struct afs_read *req;
+	struct iov_iter iter;
+
+	struct fscache_request_shape shape = {
+		.proposed_start		= start,
+		.proposed_nr_pages	= last - start + 1,
+		.max_io_pages		= UINT_MAX,
+		.i_size			= i_size_read(&vnode->vfs_inode),
+		.for_write		= true,
+	};
+
+	_enter("%lx,%lx,%llx,%llx", start, last, a, b);
+
+	fscache_shape_request(afs_vnode_cache(vnode), &shape);
+	if (!(shape.to_be_done & FSCACHE_WRITE_TO_CACHE) ||
+	    shape.actual_nr_pages == 0 ||
+	    shape.actual_start != start)
+		goto abandon;
+
+	if (shape.actual_nr_pages < shape.proposed_nr_pages) {
+		afs_clear_fscache_bits(vnode->vfs_inode.i_mapping,
+				       start + shape.actual_nr_pages,
+				       start + shape.proposed_nr_pages - 1);
+		last = start + shape.actual_nr_pages - 1;
+		b = (loff_t)(last + 1) << PAGE_SHIFT;
+	}
+
+	req = afs_alloc_read(GFP_NOFS);
+	if (!req)
+		goto abandon;
+
+	fscache_init_io_request(&req->cache, afs_vnode_cache(vnode),
+				&afs_write_req_ops);
+	req->vnode		= vnode;
+	req->cache.pos		= round_down(a, shape.dio_block_size);
+	req->cache.len		= round_up(b, shape.dio_block_size) - req->cache.pos;
+	req->cache.nr_pages	= shape.actual_nr_pages;
+	req->cache.mapping	= vnode->vfs_inode.i_mapping;
+	req->cache.io_done	= &afs_write_to_cache_done;
+
+	iov_iter_mapping(&iter, WRITE, req->cache.mapping,
+			 req->cache.pos, req->cache.len);
+	fscache_write(&req->cache, &iter);
+	afs_put_read(req);
+	return;
+
+abandon:
+	afs_clear_fscache_bits(vnode->vfs_inode.i_mapping, start, last);
+}


