Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   xiubli@redhat.com
To:     dhowells@redhat.com, idryomov@gmail.com, jlayton@kernel.org
Cc:     marc.dionne@auristor.com, willy@infradead.org,
        keescook@chromium.org, kirill.shutemov@linux.intel.com,
        william.kucharski@oracle.com, linux-afs@lists.infradead.org,
        linux-kernel@vger.kernel.org, ceph-devel@vger.kernel.org,
        linux-cachefs@redhat.com, vshankar@redhat.com,
        Xiubo Li <xiubli@redhat.com>
Subject: [PATCH v3 1/2] netfs: do not unlock and put the folio twice
Date:   Thu,  7 Jul 2022 12:51:11 +0800
Message-Id: <20220707045112.10177-2-xiubli@redhat.com>
In-Reply-To: <20220707045112.10177-1-xiubli@redhat.com>
References: <20220707045112.10177-1-xiubli@redhat.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Xiubo Li <xiubli@redhat.com>

check_write_begin() will unlock and put the folio when return
non-zero. So we should avoid unlocking and putting it twice in
netfs layer.

At the same time pass a **folio to check_write_begin(), in which
the folio should be cleared after being put.

URL: https://tracker.ceph.com/issues/56423
Signed-off-by: Xiubo Li <xiubli@redhat.com>
---
 fs/afs/file.c            |  2 +-
 fs/ceph/addr.c           | 11 ++++++-----
 fs/netfs/buffered_read.c | 18 ++++++++++--------
 include/linux/netfs.h    |  2 +-
 4 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/fs/afs/file.c b/fs/afs/file.c
index 42118a4f3383..afacce797fb9 100644
--- a/fs/afs/file.c
+++ b/fs/afs/file.c
@@ -375,7 +375,7 @@ static int afs_begin_cache_operation(struct netfs_io_request *rreq)
 }
 
 static int afs_check_write_begin(struct file *file, loff_t pos, unsigned len,
-				 struct folio *folio, void **_fsdata)
+				 struct folio **folio, void **_fsdata)
 {
 	struct afs_vnode *vnode = AFS_FS_I(file_inode(file));
 
diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index 8095fc47230e..cf909b08907f 100644
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -63,7 +63,7 @@
 	 (CONGESTION_ON_THRESH(congestion_kb) >> 2))
 
 static int ceph_netfs_check_write_begin(struct file *file, loff_t pos, unsigned int len,
-					struct folio *folio, void **_fsdata);
+					struct folio **folio, void **_fsdata);
 
 static inline struct ceph_snap_context *page_snap_context(struct page *page)
 {
@@ -1280,18 +1280,19 @@ ceph_find_incompatible(struct page *page)
 }
 
 static int ceph_netfs_check_write_begin(struct file *file, loff_t pos, unsigned int len,
-					struct folio *folio, void **_fsdata)
+					struct folio **folio, void **_fsdata)
 {
 	struct inode *inode = file_inode(file);
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	struct ceph_snap_context *snapc;
 
-	snapc = ceph_find_incompatible(folio_page(folio, 0));
+	snapc = ceph_find_incompatible(folio_page(*folio, 0));
 	if (snapc) {
 		int r;
 
-		folio_unlock(folio);
-		folio_put(folio);
+		folio_unlock(*folio);
+		folio_put(*folio);
+		*folio = NULL;
 		if (IS_ERR(snapc))
 			return PTR_ERR(snapc);
 
diff --git a/fs/netfs/buffered_read.c b/fs/netfs/buffered_read.c
index 42f892c5712e..308c2ad4da8e 100644
--- a/fs/netfs/buffered_read.c
+++ b/fs/netfs/buffered_read.c
@@ -319,8 +319,8 @@ static bool netfs_skip_folio_read(struct folio *folio, loff_t pos, size_t len,
  * conflicting writes once the folio is grabbed and locked.  It is passed a
  * pointer to the fsdata cookie that gets returned to the VM to be passed to
  * write_end.  It is permitted to sleep.  It should return 0 if the request
- * should go ahead; unlock the folio and return -EAGAIN to cause the folio to
- * be regot; or return an error.
+ * should go ahead; otherwise unlock, put and clear the folio and then return
+ * an error, -EAGAIN will cause the folio to be regot.
  *
  * The calling netfs must initialise a netfs context contiguous to the vfs
  * inode before calling this.
@@ -348,13 +348,14 @@ int netfs_write_begin(struct netfs_inode *ctx,
 
 	if (ctx->ops->check_write_begin) {
 		/* Allow the netfs (eg. ceph) to flush conflicts. */
-		ret = ctx->ops->check_write_begin(file, pos, len, folio, _fsdata);
+		ret = ctx->ops->check_write_begin(file, pos, len, &folio, _fsdata);
 		if (ret < 0) {
 			trace_netfs_failure(NULL, NULL, ret, netfs_fail_check_write_begin);
 			if (ret == -EAGAIN)
 				goto retry;
-			goto error;
+			goto error_unlocked;
 		}
+		BUG_ON(!folio);
 	}
 
 	if (folio_test_uptodate(folio))
@@ -375,7 +376,7 @@ int netfs_write_begin(struct netfs_inode *ctx,
 				   NETFS_READ_FOR_WRITE);
 	if (IS_ERR(rreq)) {
 		ret = PTR_ERR(rreq);
-		goto error;
+		goto error_locked;
 	}
 	rreq->no_unlock_folio	= folio_index(folio);
 	__set_bit(NETFS_RREQ_NO_UNLOCK_FOLIO, &rreq->flags);
@@ -402,12 +403,12 @@ int netfs_write_begin(struct netfs_inode *ctx,
 
 	ret = netfs_begin_read(rreq, true);
 	if (ret < 0)
-		goto error;
+		goto error_locked;
 
 have_folio:
 	ret = folio_wait_fscache_killable(folio);
 	if (ret < 0)
-		goto error;
+		goto error_locked;
 have_folio_no_wait:
 	*_folio = folio;
 	_leave(" = 0");
@@ -415,9 +416,10 @@ int netfs_write_begin(struct netfs_inode *ctx,
 
 error_put:
 	netfs_put_request(rreq, false, netfs_rreq_trace_put_failed);
-error:
+error_locked:
 	folio_unlock(folio);
 	folio_put(folio);
+error_unlocked:
 	_leave(" = %d", ret);
 	return ret;
 }
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index 1773e5df8e65..88c4eb85c6f7 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -214,7 +214,7 @@ struct netfs_request_ops {
 	void (*issue_read)(struct netfs_io_subrequest *subreq);
 	bool (*is_still_valid)(struct netfs_io_request *rreq);
 	int (*check_write_begin)(struct file *file, loff_t pos, unsigned len,
-				 struct folio *folio, void **_fsdata);
+				 struct folio **folio, void **_fsdata);
 	void (*done)(struct netfs_io_request *rreq);
 };
 
-- 
2.36.0.rc1