Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse

Pull fuse update from Miklos Szeredi:
 "This series adds cached writeback support to fuse, improving write
  throughput"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
  fuse: fix "uninitialized variable" warning
  fuse: Turn writeback cache on
  fuse: Fix O_DIRECT operations vs cached writeback misorder
  fuse: fuse_flush() should wait on writeback
  fuse: Implement write_begin/write_end callbacks
  fuse: restructure fuse_readpage()
  fuse: Flush files on wb close
  fuse: Trust kernel i_mtime only
  fuse: Trust kernel i_size only
  fuse: Connection bit for enabling writeback
  fuse: Prepare to handle short reads
  fuse: Linking file to inode helper

fs/fuse/cuse.c

@@ -95,7 +95,7 @@ static ssize_t cuse_read(struct file *file, char __user *buf, size_t count,
 	struct iovec iov = { .iov_base = buf, .iov_len = count };
 	struct fuse_io_priv io = { .async = 0, .file = file };
 
-	return fuse_direct_io(&io, &iov, 1, count, &pos, 0);
+	return fuse_direct_io(&io, &iov, 1, count, &pos, FUSE_DIO_CUSE);
 }
 
 static ssize_t cuse_write(struct file *file, const char __user *buf,
@@ -109,7 +109,8 @@ static ssize_t cuse_write(struct file *file, const char __user *buf,
 	 * No locking or generic_write_checks(), the server is
 	 * responsible for locking and sanity checks.
 	 */
-	return fuse_direct_io(&io, &iov, 1, count, &pos, 1);
+	return fuse_direct_io(&io, &iov, 1, count, &pos,
+			      FUSE_DIO_WRITE | FUSE_DIO_CUSE);
 }
 
 static int cuse_open(struct inode *inode, struct file *file)

fs/fuse/dir.c

@@ -839,6 +839,14 @@ static void fuse_fillattr(struct inode *inode, struct fuse_attr *attr,
 			  struct kstat *stat)
 {
 	unsigned int blkbits;
+	struct fuse_conn *fc = get_fuse_conn(inode);
+
+	/* see the comment in fuse_change_attributes() */
+	if (fc->writeback_cache && S_ISREG(inode->i_mode)) {
+		attr->size = i_size_read(inode);
+		attr->mtime = inode->i_mtime.tv_sec;
+		attr->mtimensec = inode->i_mtime.tv_nsec;
+	}
 
 	stat->dev = inode->i_sb->s_dev;
 	stat->ino = attr->ino;
@@ -1477,12 +1485,16 @@ static long fuse_dir_compat_ioctl(struct file *file, unsigned int cmd,
 				 FUSE_IOCTL_COMPAT | FUSE_IOCTL_DIR);
 }
 
-static bool update_mtime(unsigned ivalid)
+static bool update_mtime(unsigned ivalid, bool trust_local_mtime)
 {
 	/* Always update if mtime is explicitly set  */
 	if (ivalid & ATTR_MTIME_SET)
 		return true;
 
+	/* Or if kernel i_mtime is the official one */
+	if (trust_local_mtime)
+		return true;
+
 	/* If it's an open(O_TRUNC) or an ftruncate(), don't update */
 	if ((ivalid & ATTR_SIZE) && (ivalid & (ATTR_OPEN | ATTR_FILE)))
 		return false;
@@ -1491,7 +1503,8 @@ static bool update_mtime(unsigned ivalid)
 	return true;
 }
 
-static void iattr_to_fattr(struct iattr *iattr, struct fuse_setattr_in *arg)
+static void iattr_to_fattr(struct iattr *iattr, struct fuse_setattr_in *arg,
+			   bool trust_local_mtime)
 {
 	unsigned ivalid = iattr->ia_valid;
 
@@ -1510,11 +1523,11 @@ static void iattr_to_fattr(struct iattr *iattr, struct fuse_setattr_in *arg)
 		if (!(ivalid & ATTR_ATIME_SET))
 			arg->valid |= FATTR_ATIME_NOW;
 	}
-	if ((ivalid & ATTR_MTIME) && update_mtime(ivalid)) {
+	if ((ivalid & ATTR_MTIME) && update_mtime(ivalid, trust_local_mtime)) {
 		arg->valid |= FATTR_MTIME;
 		arg->mtime = iattr->ia_mtime.tv_sec;
 		arg->mtimensec = iattr->ia_mtime.tv_nsec;
-		if (!(ivalid & ATTR_MTIME_SET))
+		if (!(ivalid & ATTR_MTIME_SET) && !trust_local_mtime)
 			arg->valid |= FATTR_MTIME_NOW;
 	}
 }
@@ -1563,6 +1576,63 @@ void fuse_release_nowrite(struct inode *inode)
 	spin_unlock(&fc->lock);
 }
 
+static void fuse_setattr_fill(struct fuse_conn *fc, struct fuse_req *req,
+			      struct inode *inode,
+			      struct fuse_setattr_in *inarg_p,
+			      struct fuse_attr_out *outarg_p)
+{
+	req->in.h.opcode = FUSE_SETATTR;
+	req->in.h.nodeid = get_node_id(inode);
+	req->in.numargs = 1;
+	req->in.args[0].size = sizeof(*inarg_p);
+	req->in.args[0].value = inarg_p;
+	req->out.numargs = 1;
+	if (fc->minor < 9)
+		req->out.args[0].size = FUSE_COMPAT_ATTR_OUT_SIZE;
+	else
+		req->out.args[0].size = sizeof(*outarg_p);
+	req->out.args[0].value = outarg_p;
+}
+
+/*
+ * Flush inode->i_mtime to the server
+ */
+int fuse_flush_mtime(struct file *file, bool nofail)
+{
+	struct inode *inode = file->f_mapping->host;
+	struct fuse_inode *fi = get_fuse_inode(inode);
+	struct fuse_conn *fc = get_fuse_conn(inode);
+	struct fuse_req *req = NULL;
+	struct fuse_setattr_in inarg;
+	struct fuse_attr_out outarg;
+	int err;
+
+	if (nofail) {
+		req = fuse_get_req_nofail_nopages(fc, file);
+	} else {
+		req = fuse_get_req_nopages(fc);
+		if (IS_ERR(req))
+			return PTR_ERR(req);
+	}
+
+	memset(&inarg, 0, sizeof(inarg));
+	memset(&outarg, 0, sizeof(outarg));
+
+	inarg.valid |= FATTR_MTIME;
+	inarg.mtime = inode->i_mtime.tv_sec;
+	inarg.mtimensec = inode->i_mtime.tv_nsec;
+
+	fuse_setattr_fill(fc, req, inode, &inarg, &outarg);
+	fuse_request_send(fc, req);
+	err = req->out.h.error;
+	fuse_put_request(fc, req);
+
+	if (!err)
+		clear_bit(FUSE_I_MTIME_DIRTY, &fi->state);
+
+	return err;
+}
+
 /*
  * Set attributes, and at the same time refresh them.
  *
@@ -1580,8 +1650,10 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr,
 	struct fuse_setattr_in inarg;
 	struct fuse_attr_out outarg;
 	bool is_truncate = false;
+	bool is_wb = fc->writeback_cache;
 	loff_t oldsize;
 	int err;
+	bool trust_local_mtime = is_wb && S_ISREG(inode->i_mode);
 
 	if (!(fc->flags & FUSE_DEFAULT_PERMISSIONS))
 		attr->ia_valid |= ATTR_FORCE;
@@ -1610,7 +1682,7 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr,
 
 	memset(&inarg, 0, sizeof(inarg));
 	memset(&outarg, 0, sizeof(outarg));
-	iattr_to_fattr(attr, &inarg);
+	iattr_to_fattr(attr, &inarg, trust_local_mtime);
 	if (file) {
 		struct fuse_file *ff = file->private_data;
 		inarg.valid |= FATTR_FH;
@@ -1621,17 +1693,7 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr,
 		inarg.valid |= FATTR_LOCKOWNER;
 		inarg.lock_owner = fuse_lock_owner_id(fc, current->files);
 	}
-	req->in.h.opcode = FUSE_SETATTR;
-	req->in.h.nodeid = get_node_id(inode);
-	req->in.numargs = 1;
-	req->in.args[0].size = sizeof(inarg);
-	req->in.args[0].value = &inarg;
-	req->out.numargs = 1;
-	if (fc->minor < 9)
-		req->out.args[0].size = FUSE_COMPAT_ATTR_OUT_SIZE;
-	else
-		req->out.args[0].size = sizeof(outarg);
-	req->out.args[0].value = &outarg;
+	fuse_setattr_fill(fc, req, inode, &inarg, &outarg);
 	fuse_request_send(fc, req);
 	err = req->out.h.error;
 	fuse_put_request(fc, req);
@@ -1648,9 +1710,17 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr,
 	}
 
 	spin_lock(&fc->lock);
+	/* the kernel maintains i_mtime locally */
+	if (trust_local_mtime && (attr->ia_valid & ATTR_MTIME)) {
+		inode->i_mtime = attr->ia_mtime;
+		clear_bit(FUSE_I_MTIME_DIRTY, &fi->state);
+	}
+
 	fuse_change_attributes_common(inode, &outarg.attr,
 				      attr_timeout(&outarg));
 	oldsize = inode->i_size;
+	/* see the comment in fuse_change_attributes() */
+	if (!is_wb || is_truncate || !S_ISREG(inode->i_mode))
 		i_size_write(inode, outarg.attr.size);
 
 	if (is_truncate) {
@@ -1663,7 +1733,8 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr,
 	 * Only call invalidate_inode_pages2() after removing
 	 * FUSE_NOWRITE, otherwise fuse_launder_page() would deadlock.
 	 */
-	if (S_ISREG(inode->i_mode) && oldsize != outarg.attr.size) {
+	if ((is_truncate || !is_wb) &&
+	    S_ISREG(inode->i_mode) && oldsize != outarg.attr.size) {
 		truncate_pagecache(inode, outarg.attr.size);
 		invalidate_inode_pages2(inode->i_mapping);
 	}
@@ -1875,6 +1946,17 @@ static int fuse_removexattr(struct dentry *entry, const char *name)
 	return err;
 }
 
+static int fuse_update_time(struct inode *inode, struct timespec *now,
+			    int flags)
+{
+	if (flags & S_MTIME) {
+		inode->i_mtime = *now;
+		set_bit(FUSE_I_MTIME_DIRTY, &get_fuse_inode(inode)->state);
+		BUG_ON(!S_ISREG(inode->i_mode));
+	}
+	return 0;
+}
+
 static const struct inode_operations fuse_dir_inode_operations = {
 	.lookup		= fuse_lookup,
 	.mkdir		= fuse_mkdir,
@@ -1914,6 +1996,7 @@ static const struct inode_operations fuse_common_inode_operations = {
 	.getxattr	= fuse_getxattr,
 	.listxattr	= fuse_listxattr,
 	.removexattr	= fuse_removexattr,
+	.update_time	= fuse_update_time,
 };
 
 static const struct inode_operations fuse_symlink_inode_operations = {

fs/fuse/fuse_i.h

@@ -119,6 +119,8 @@ enum {
 	FUSE_I_INIT_RDPLUS,
 	/** An operation changing file size is in progress  */
 	FUSE_I_SIZE_UNSTABLE,
+	/** i_mtime has been updated locally; a flush to userspace needed */
+	FUSE_I_MTIME_DIRTY,
 };
 
 struct fuse_conn;
@@ -480,6 +482,9 @@ struct fuse_conn {
 	/** Set if bdi is valid */
 	unsigned bdi_initialized:1;
 
+	/** write-back cache policy (default is write-through) */
+	unsigned writeback_cache:1;
+
 	/*
 	 * The following bitfields are only for optimization purposes
 	 * and hence races in setting them will not cause malfunction
@@ -863,9 +868,20 @@ int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
 
 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
 		 bool isdir);
+
+/**
+ * fuse_direct_io() flags
+ */
+
+/** If set, it is WRITE; otherwise - READ */
+#define FUSE_DIO_WRITE (1 << 0)
+
+/** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
+#define FUSE_DIO_CUSE  (1 << 1)
+
 ssize_t fuse_direct_io(struct fuse_io_priv *io, const struct iovec *iov,
 		       unsigned long nr_segs, size_t count, loff_t *ppos,
-		       int write);
+		       int flags);
 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
 		   unsigned int flags);
 long fuse_ioctl_common(struct file *file, unsigned int cmd,
@@ -873,7 +889,9 @@ long fuse_ioctl_common(struct file *file, unsigned int cmd,
 unsigned fuse_file_poll(struct file *file, poll_table *wait);
 int fuse_dev_release(struct inode *inode, struct file *file);
 
-void fuse_write_update_size(struct inode *inode, loff_t pos);
+bool fuse_write_update_size(struct inode *inode, loff_t pos);
+
+int fuse_flush_mtime(struct file *file, bool nofail);
 
 int fuse_do_setattr(struct inode *inode, struct iattr *attr,
 		    struct file *file);

fs/fuse/inode.c

@@ -170,8 +170,11 @@ void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
 	inode->i_blocks  = attr->blocks;
 	inode->i_atime.tv_sec   = attr->atime;
 	inode->i_atime.tv_nsec  = attr->atimensec;
+	/* mtime from server may be stale due to local buffered write */
+	if (!fc->writeback_cache || !S_ISREG(inode->i_mode)) {
 		inode->i_mtime.tv_sec   = attr->mtime;
 		inode->i_mtime.tv_nsec  = attr->mtimensec;
+	}
 	inode->i_ctime.tv_sec   = attr->ctime;
 	inode->i_ctime.tv_nsec  = attr->ctimensec;
 
@@ -197,6 +200,7 @@ void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
 {
 	struct fuse_conn *fc = get_fuse_conn(inode);
 	struct fuse_inode *fi = get_fuse_inode(inode);
+	bool is_wb = fc->writeback_cache;
 	loff_t oldsize;
 	struct timespec old_mtime;
 
@@ -211,10 +215,16 @@ void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
 	fuse_change_attributes_common(inode, attr, attr_valid);
 
 	oldsize = inode->i_size;
+	/*
+	 * In case of writeback_cache enabled, the cached writes beyond EOF
+	 * extend local i_size without keeping userspace server in sync. So,
+	 * attr->size coming from server can be stale. We cannot trust it.
+	 */
+	if (!is_wb || !S_ISREG(inode->i_mode))
 		i_size_write(inode, attr->size);
 	spin_unlock(&fc->lock);
 
-	if (S_ISREG(inode->i_mode)) {
+	if (!is_wb && S_ISREG(inode->i_mode)) {
 		bool inval = false;
 
 		if (oldsize != attr->size) {
@@ -243,6 +253,8 @@ static void fuse_init_inode(struct inode *inode, struct fuse_attr *attr)
 {
 	inode->i_mode = attr->mode & S_IFMT;
 	inode->i_size = attr->size;
+	inode->i_mtime.tv_sec  = attr->mtime;
+	inode->i_mtime.tv_nsec = attr->mtimensec;
 	if (S_ISREG(inode->i_mode)) {
 		fuse_init_common(inode);
 		fuse_init_file_inode(inode);
@@ -289,7 +301,9 @@ struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
 		return NULL;
 
 	if ((inode->i_state & I_NEW)) {
-		inode->i_flags |= S_NOATIME|S_NOCMTIME;
+		inode->i_flags |= S_NOATIME;
+		if (!fc->writeback_cache || !S_ISREG(inode->i_mode))
+			inode->i_flags |= S_NOCMTIME;
 		inode->i_generation = generation;
 		inode->i_data.backing_dev_info = &fc->bdi;
 		fuse_init_inode(inode, attr);
@@ -873,6 +887,8 @@ static void process_init_reply(struct fuse_conn *fc, struct fuse_req *req)
 			}
 			if (arg->flags & FUSE_ASYNC_DIO)
 				fc->async_dio = 1;
+			if (arg->flags & FUSE_WRITEBACK_CACHE)
+				fc->writeback_cache = 1;
 		} else {
 			ra_pages = fc->max_read / PAGE_CACHE_SIZE;
 			fc->no_lock = 1;
@@ -900,7 +916,8 @@ static void fuse_send_init(struct fuse_conn *fc, struct fuse_req *req)
 		FUSE_EXPORT_SUPPORT | FUSE_BIG_WRITES | FUSE_DONT_MASK |
 		FUSE_SPLICE_WRITE | FUSE_SPLICE_MOVE | FUSE_SPLICE_READ |
 		FUSE_FLOCK_LOCKS | FUSE_IOCTL_DIR | FUSE_AUTO_INVAL_DATA |
-		FUSE_DO_READDIRPLUS | FUSE_READDIRPLUS_AUTO | FUSE_ASYNC_DIO;
+		FUSE_DO_READDIRPLUS | FUSE_READDIRPLUS_AUTO | FUSE_ASYNC_DIO |
+		FUSE_WRITEBACK_CACHE;
 	req->in.h.opcode = FUSE_INIT;
 	req->in.numargs = 1;
 	req->in.args[0].size = sizeof(*arg);

include/uapi/linux/fuse.h

@@ -93,6 +93,9 @@
  *
  * 7.22
  *  - add FUSE_ASYNC_DIO
+ *
+ * 7.23
+ *  - add FUSE_WRITEBACK_CACHE
  */
 
 #ifndef _LINUX_FUSE_H
@@ -128,7 +131,7 @@
 #define FUSE_KERNEL_VERSION 7
 
 /** Minor version number of this interface */
-#define FUSE_KERNEL_MINOR_VERSION 22
+#define FUSE_KERNEL_MINOR_VERSION 23
 
 /** The node ID of the root inode */
 #define FUSE_ROOT_ID 1
@@ -219,6 +222,7 @@ struct fuse_file_lock {
  * FUSE_DO_READDIRPLUS: do READDIRPLUS (READDIR+LOOKUP in one)
  * FUSE_READDIRPLUS_AUTO: adaptive readdirplus
  * FUSE_ASYNC_DIO: asynchronous direct I/O submission
+ * FUSE_WRITEBACK_CACHE: use writeback cache for buffered writes
  */
 #define FUSE_ASYNC_READ		(1 << 0)
 #define FUSE_POSIX_LOCKS	(1 << 1)
@@ -236,6 +240,7 @@ struct fuse_file_lock {
 #define FUSE_DO_READDIRPLUS	(1 << 13)
 #define FUSE_READDIRPLUS_AUTO	(1 << 14)
 #define FUSE_ASYNC_DIO		(1 << 15)
+#define FUSE_WRITEBACK_CACHE	(1 << 16)
 
 /**
  * CUSE INIT request/reply flags