Merge tag 'ceph-for-5.19-rc7' of https://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "A folio locking fixup that Xiubo and David cooperated on, marked for
  stable. Most of it is in netfs but I picked it up into ceph tree on
  agreement with David"

* tag 'ceph-for-5.19-rc7' of https://github.com/ceph/ceph-client:
  netfs: do not unlock and put the folio twice

Documentation/filesystems/netfs_library.rst

@@ -301,7 +301,7 @@ through which it can issue requests and negotiate::
 		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 **foliop, void **_fsdata);
 		void (*done)(struct netfs_io_request *rreq);
 	};
 
@@ -381,8 +381,10 @@ The operations are as follows:
    allocated/grabbed the folio to be modified to allow the filesystem to flush
    conflicting state before allowing it to be modified.
 
-   It should return 0 if everything is now fine, -EAGAIN if the folio should be
-   regrabbed and any other error code to abort the operation.
+   It may unlock and discard the folio it was given and set the caller's folio
+   pointer to NULL.  It should return 0 if everything is now fine (``*foliop``
+   left set) or the op should be retried (``*foliop`` cleared) and any other
+   error code to abort the operation.
 
  * ``done``
 

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 **foliop, void **_fsdata)
 {
 	struct afs_vnode *vnode = AFS_FS_I(file_inode(file));
 

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 **foliop, void **_fsdata);
 
 static inline struct ceph_snap_context *page_snap_context(struct page *page)
 {
@@ -1288,18 +1288,19 @@ 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 **foliop, 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(*foliop, 0));
 	if (snapc) {
 		int r;
 
-		folio_unlock(folio);
-		folio_put(folio);
+		folio_unlock(*foliop);
+		folio_put(*foliop);
+		*foliop = NULL;
 		if (IS_ERR(snapc))
 			return PTR_ERR(snapc);
 

fs/netfs/buffered_read.c

@@ -319,8 +319,9 @@ 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 or it may return an error.  It may also unlock and put the
+ * folio, provided it sets ``*foliop`` to NULL, in which case a return of 0
+ * will cause the folio to be re-got and the process to be retried.
  *
  * The calling netfs must initialise a netfs context contiguous to the vfs
  * inode before calling this.
@@ -348,13 +349,13 @@ 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;
 		}
+		if (!folio)
+			goto retry;
 	}
 
 	if (folio_test_uptodate(folio))
@@ -416,8 +417,10 @@ int netfs_write_begin(struct netfs_inode *ctx,
 error_put:
 	netfs_put_request(rreq, false, netfs_rreq_trace_put_failed);
 error:
+	if (folio) {
 		folio_unlock(folio);
 		folio_put(folio);
+	}
 	_leave(" = %d", ret);
 	return ret;
 }

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 **foliop, void **_fsdata);
 	void (*done)(struct netfs_io_request *rreq);
 };