Merge branch 'master'

14 files changed, 208 insertions, 91 deletions

diff --git a/fs/Kconfig b/fs/Kconfig
index 2524629..f9b5842 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -842,6 +842,12 @@ config TMPFS
 config HUGETLBFS
 	bool "HugeTLB file system support"
 	depends X86 || IA64 || PPC64 || SPARC64 || SUPERH || BROKEN
+	help
+	  hugetlbfs is a filesystem backing for HugeTLB pages, based on
+	  ramfs. For architectures that support it, say Y here and read
+	  <file:Documentation/vm/hugetlbpage.txt> for details.
+
+	  If unsure, say N.
 
 config HUGETLB_PAGE
 	def_bool HUGETLBFS
diff --git a/fs/exec.c b/fs/exec.c
index 4121bb5..3a79d97 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -712,7 +712,7 @@ static int de_thread(struct task_struct *tsk)
 		attach_pid(current, PIDTYPE_PID,  current->pid);
 		attach_pid(current, PIDTYPE_PGID, current->signal->pgrp);
 		attach_pid(current, PIDTYPE_SID,  current->signal->session);
-		list_add_tail(&current->tasks, &init_task.tasks);
+		list_add_tail_rcu(&current->tasks, &init_task.tasks);
 
 		current->group_leader = current;
 		leader->group_leader = current;
diff --git a/fs/ext3/resize.c b/fs/ext3/resize.c
index 14f5f6e..c5ffa85 100644
--- a/fs/ext3/resize.c
+++ b/fs/ext3/resize.c
@@ -767,6 +767,7 @@ int ext3_group_add(struct super_block *sb, struct ext3_new_group_data *input)
 	if (input->group != sbi->s_groups_count) {
 		ext3_warning(sb, __FUNCTION__,
 			     "multiple resizers run on filesystem!");
+		unlock_super(sb);
 		err = -EBUSY;
 		goto exit_journal;
 	}
diff --git a/fs/lockd/svclock.c b/fs/lockd/svclock.c
index d2b66ba..3ef7391 100644
--- a/fs/lockd/svclock.c
+++ b/fs/lockd/svclock.c
@@ -650,7 +650,7 @@ static void nlmsvc_grant_callback(struct rpc_task *task, void *data)
 	svc_wake_up(block->b_daemon);
 }
 
-void nlmsvc_grant_release(void *data)
+static void nlmsvc_grant_release(void *data)
 {
 	struct nlm_rqst		*call = data;
 
diff --git a/fs/locks.c b/fs/locks.c
index dda83d6..efad7988 100644
--- a/fs/locks.c
+++ b/fs/locks.c
@@ -2230,7 +2230,12 @@ void steal_locks(fl_owner_t from)
 
 	lock_kernel();
 	j = 0;
-	rcu_read_lock();
+
+	/*
+	 * We are not taking a ref to the file structures, so
+	 * we need to acquire ->file_lock.
+	 */
+	spin_lock(&files->file_lock);
 	fdt = files_fdtable(files);
 	for (;;) {
 		unsigned long set;
@@ -2248,7 +2253,7 @@ void steal_locks(fl_owner_t from)
 			set >>= 1;
 		}
 	}
-	rcu_read_unlock();
+	spin_unlock(&files->file_lock);
 	unlock_kernel();
 }
 EXPORT_SYMBOL(steal_locks);
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index a23f348..cae74dd 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -128,15 +128,14 @@ struct inode_operations nfs4_dir_inode_operations = {
 static int
 nfs_opendir(struct inode *inode, struct file *filp)
 {
-	int res = 0;
+	int res;
 
 	dfprintk(VFS, "NFS: opendir(%s/%ld)\n",
 			inode->i_sb->s_id, inode->i_ino);
 
 	lock_kernel();
 	/* Call generic open code in order to cache credentials */
-	if (!res)
-		res = nfs_open(inode, filp);
+	res = nfs_open(inode, filp);
 	unlock_kernel();
 	return res;
 }
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 0f583cb..3c72b0c 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -112,10 +112,9 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode
  */
 ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
 {
-	struct dentry *dentry = iocb->ki_filp->f_dentry;
-
 	dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n",
-			dentry->d_name.name, (long long) pos, nr_segs);
+			iocb->ki_filp->f_dentry->d_name.name,
+			(long long) pos, nr_segs);
 
 	return -EINVAL;
 }
@@ -468,7 +467,6 @@ static const struct rpc_call_ops nfs_commit_direct_ops = {
 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 {
 	struct nfs_write_data *data = dreq->commit_data;
-	struct rpc_task *task = &data->task;
 
 	data->inode = dreq->inode;
 	data->cred = dreq->ctx->cred;
@@ -489,7 +487,7 @@ static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 	/* Note: task.tk_ops->rpc_release will free dreq->commit_data */
 	dreq->commit_data = NULL;
 
-	dprintk("NFS: %5u initiated commit call\n", task->tk_pid);
+	dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
 
 	lock_kernel();
 	rpc_execute(&data->task);
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index f1df2c8..fade02c 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -534,10 +534,9 @@ static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
  */
 static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
 {
-	struct inode * inode = filp->f_mapping->host;
-
 	dprintk("NFS: nfs_flock(f=%s/%ld, t=%x, fl=%x)\n",
-			inode->i_sb->s_id, inode->i_ino,
+			filp->f_dentry->d_inode->i_sb->s_id,
+			filp->f_dentry->d_inode->i_ino,
 			fl->fl_type, fl->fl_flags);
 
 	/*
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 2f7656b..d0b991a 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -700,12 +700,9 @@ static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
 	/*
 	 * Display superblock I/O counters
 	 */
-	for (cpu = 0; cpu < NR_CPUS; cpu++) {
+	for_each_possible_cpu(cpu) {
 		struct nfs_iostats *stats;
 
-		if (!cpu_possible(cpu))
-			continue;
-
 		preempt_disable();
 		stats = per_cpu_ptr(nfss->io_stats, cpu);
 
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 47ece1d..d86c0db 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -1218,7 +1218,7 @@ out:
 	return status;
 }
 
-static void nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state)
+static int nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state)
 {
 	struct file *filp;
 
@@ -1227,8 +1227,10 @@ static void nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, st
 		struct nfs_open_context *ctx;
 		ctx = (struct nfs_open_context *)filp->private_data;
 		ctx->state = state;
-	} else
-		nfs4_close_state(state, nd->intent.open.flags);
+		return 0;
+	}
+	nfs4_close_state(state, nd->intent.open.flags);
+	return PTR_ERR(filp);
 }
 
 struct dentry *
@@ -1835,7 +1837,7 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 			nfs_setattr_update_inode(state->inode, sattr);
 	}
 	if (status == 0 && nd != NULL && (nd->flags & LOOKUP_OPEN))
-		nfs4_intent_set_file(nd, dentry, state);
+		status = nfs4_intent_set_file(nd, dentry, state);
 	else
 		nfs4_close_state(state, flags);
 out:
diff --git a/fs/open.c b/fs/open.c
index c32c89d..53ec28c 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -331,7 +331,10 @@ out:
 
 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
 {
-	return do_sys_ftruncate(fd, length, 1);
+	long ret = do_sys_ftruncate(fd, length, 1);
+	/* avoid REGPARM breakage on x86: */
+	prevent_tail_call(ret);
+	return ret;
 }
 
 /* LFS versions of truncate are only needed on 32 bit machines */
@@ -343,7 +346,10 @@ asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 
 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 {
-	return do_sys_ftruncate(fd, length, 0);
+	long ret = do_sys_ftruncate(fd, length, 0);
+	/* avoid REGPARM breakage on x86: */
+	prevent_tail_call(ret);
+	return ret;
 }
 #endif
 
@@ -1093,20 +1099,30 @@ long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
 
 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
 {
+	long ret;
+
 	if (force_o_largefile())
 		flags |= O_LARGEFILE;
 
-	return do_sys_open(AT_FDCWD, filename, flags, mode);
+	ret = do_sys_open(AT_FDCWD, filename, flags, mode);
+	/* avoid REGPARM breakage on x86: */
+	prevent_tail_call(ret);
+	return ret;
 }
 EXPORT_SYMBOL_GPL(sys_open);
 
 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
 			   int mode)
 {
+	long ret;
+
 	if (force_o_largefile())
 		flags |= O_LARGEFILE;
 
-	return do_sys_open(dfd, filename, flags, mode);
+	ret = do_sys_open(dfd, filename, flags, mode);
+	/* avoid REGPARM breakage on x86: */
+	prevent_tail_call(ret);
+	return ret;
 }
 EXPORT_SYMBOL_GPL(sys_openat);
 
diff --git a/fs/partitions/check.c b/fs/partitions/check.c
index f3b6af0..45ae7dd 100644
--- a/fs/partitions/check.c
+++ b/fs/partitions/check.c
@@ -372,6 +372,7 @@ static char *make_block_name(struct gendisk *disk)
 	char *name;
 	static char *block_str = "block:";
 	int size;
+	char *s;
 
 	size = strlen(block_str) + strlen(disk->disk_name) + 1;
 	name = kmalloc(size, GFP_KERNEL);
@@ -379,6 +380,10 @@ static char *make_block_name(struct gendisk *disk)
 		return NULL;
 	strcpy(name, block_str);
 	strcat(name, disk->disk_name);
+	/* ewww... some of these buggers have / in name... */
+	s = strchr(name, '/');
+	if (s)
+		*s = '!';
 	return name;
 }
 
diff --git a/fs/proc/base.c b/fs/proc/base.c
index a3a3eec..6cc77dc 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -297,16 +297,20 @@ static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsm
 
 	files = get_files_struct(task);
 	if (files) {
-		rcu_read_lock();
+		/*
+		 * We are not taking a ref to the file structure, so we must
+		 * hold ->file_lock.
+		 */
+		spin_lock(&files->file_lock);
 		file = fcheck_files(files, fd);
 		if (file) {
 			*mnt = mntget(file->f_vfsmnt);
 			*dentry = dget(file->f_dentry);
-			rcu_read_unlock();
+			spin_unlock(&files->file_lock);
 			put_files_struct(files);
 			return 0;
 		}
-		rcu_read_unlock();
+		spin_unlock(&files->file_lock);
 		put_files_struct(files);
 	}
 	return -ENOENT;
@@ -1523,7 +1527,12 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry,
 	if (!files)
 		goto out_unlock;
 	inode->i_mode = S_IFLNK;
-	rcu_read_lock();
+
+	/*
+	 * We are not taking a ref to the file structure, so we must
+	 * hold ->file_lock.
+	 */
+	spin_lock(&files->file_lock);
 	file = fcheck_files(files, fd);
 	if (!file)
 		goto out_unlock2;
@@ -1531,7 +1540,7 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry,
 		inode->i_mode |= S_IRUSR | S_IXUSR;
 	if (file->f_mode & 2)
 		inode->i_mode |= S_IWUSR | S_IXUSR;
-	rcu_read_unlock();
+	spin_unlock(&files->file_lock);
 	put_files_struct(files);
 	inode->i_op = &proc_pid_link_inode_operations;
 	inode->i_size = 64;
@@ -1541,7 +1550,7 @@ static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry,
 	return NULL;
 
 out_unlock2:
-	rcu_read_unlock();
+	spin_unlock(&files->file_lock);
 	put_files_struct(files);
 out_unlock:
 	iput(inode);
diff --git a/fs/splice.c b/fs/splice.c
index 8d57e89..0559e75 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -50,7 +50,8 @@ static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
 	struct page *page = buf->page;
 	struct address_space *mapping = page_mapping(page);
 
-	WARN_ON(!PageLocked(page));
+	lock_page(page);
+
 	WARN_ON(!PageUptodate(page));
 
 	/*
@@ -65,8 +66,10 @@ static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
 	if (PagePrivate(page))
 		try_to_release_page(page, mapping_gfp_mask(mapping));
 
-	if (!remove_mapping(mapping, page))
+	if (!remove_mapping(mapping, page)) {
+		unlock_page(page);
 		return 1;
+	}
 
 	buf->flags |= PIPE_BUF_FLAG_STOLEN | PIPE_BUF_FLAG_LRU;
 	return 0;
@@ -145,8 +148,8 @@ static struct pipe_buf_operations page_cache_pipe_buf_ops = {
  * pipe buffer operations. Otherwise very similar to the regular pipe_writev().
  */
 static ssize_t move_to_pipe(struct pipe_inode_info *pipe, struct page **pages,
-			    int nr_pages, unsigned long offset,
-			    unsigned long len, unsigned int flags)
+			    int nr_pages, unsigned long len,
+			    unsigned int offset, unsigned int flags)
 {
 	int ret, do_wakeup, i;
 
@@ -243,14 +246,16 @@ __generic_file_splice_read(struct file *in, loff_t *ppos,
 			   unsigned int flags)
 {
 	struct address_space *mapping = in->f_mapping;
-	unsigned int offset, nr_pages;
+	unsigned int loff, offset, nr_pages;
 	struct page *pages[PIPE_BUFFERS];
 	struct page *page;
-	pgoff_t index;
+	pgoff_t index, end_index;
+	loff_t isize;
+	size_t bytes;
 	int i, error;
 
 	index = *ppos >> PAGE_CACHE_SHIFT;
-	offset = *ppos & ~PAGE_CACHE_MASK;
+	loff = offset = *ppos & ~PAGE_CACHE_MASK;
 	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
 
 	if (nr_pages > PIPE_BUFFERS)
@@ -268,7 +273,17 @@ __generic_file_splice_read(struct file *in, loff_t *ppos,
 	 * Now fill in the holes:
 	 */
 	error = 0;
+	bytes = 0;
 	for (i = 0; i < nr_pages; i++, index++) {
+		unsigned int this_len;
+
+		if (!len)
+			break;
+
+		/*
+		 * this_len is the max we'll use from this page
+		 */
+		this_len = min(len, PAGE_CACHE_SIZE - loff);
 find_page:
 		/*
 		 * lookup the page for this index
@@ -276,14 +291,6 @@ find_page:
 		page = find_get_page(mapping, index);
 		if (!page) {
 			/*
-			 * If in nonblock mode then dont block on
-			 * readpage (we've kicked readahead so there
-			 * will be asynchronous progress):
-			 */
-			if (flags & SPLICE_F_NONBLOCK)
-				break;
-
-			/*
 			 * page didn't exist, allocate one
 			 */
 			page = page_cache_alloc_cold(mapping);
@@ -304,6 +311,13 @@ find_page:
 		 * If the page isn't uptodate, we may need to start io on it
 		 */
 		if (!PageUptodate(page)) {
+			/*
+			 * If in nonblock mode then dont block on waiting
+			 * for an in-flight io page
+			 */
+			if (flags & SPLICE_F_NONBLOCK)
+				break;
+
 			lock_page(page);
 
 			/*
@@ -336,13 +350,43 @@ readpage:
 					goto find_page;
 				break;
 			}
+
+			/*
+			 * i_size must be checked after ->readpage().
+			 */
+			isize = i_size_read(mapping->host);
+			end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
+			if (unlikely(!isize || index > end_index)) {
+				page_cache_release(page);
+				break;
+			}
+
+			/*
+			 * if this is the last page, see if we need to shrink
+			 * the length and stop
+			 */
+			if (end_index == index) {
+				loff = PAGE_CACHE_SIZE - (isize & ~PAGE_CACHE_MASK);
+				if (bytes + loff > isize) {
+					page_cache_release(page);
+					break;
+				}
+				/*
+				 * force quit after adding this page
+				 */
+				nr_pages = i;
+				this_len = min(this_len, loff);
+			}
 		}
 fill_it:
 		pages[i] = page;
+		bytes += this_len;
+		len -= this_len;
+		loff = 0;
 	}
 
 	if (i)
-		return move_to_pipe(pipe, pages, i, offset, len, flags);
+		return move_to_pipe(pipe, pages, i, bytes, offset, flags);
 
 	return error;
 }
@@ -369,17 +413,20 @@ ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
 	while (len) {
 		ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
 
-		if (ret <= 0)
+		if (ret < 0)
 			break;
+		else if (!ret) {
+			if (spliced)
+				break;
+			if (flags & SPLICE_F_NONBLOCK) {
+				ret = -EAGAIN;
+				break;
+			}
+		}
 
 		*ppos += ret;
 		len -= ret;
 		spliced += ret;
-
-		if (!(flags & SPLICE_F_NONBLOCK))
-			continue;
-		ret = -EAGAIN;
-		break;
 	}
 
 	if (spliced)
@@ -474,14 +521,12 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 	if (sd->flags & SPLICE_F_MOVE) {
 		/*
 		 * If steal succeeds, buf->page is now pruned from the vm
-		 * side (LRU and page cache) and we can reuse it.
+		 * side (LRU and page cache) and we can reuse it. The page
+		 * will also be looked on successful return.
 		 */
 		if (buf->ops->steal(info, buf))
 			goto find_page;
 
-		/*
-		 * this will also set the page locked
-		 */
 		page = buf->page;
 		if (add_to_page_cache(page, mapping, index, gfp_mask))
 			goto find_page;
@@ -490,15 +535,27 @@ static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
 			lru_cache_add(page);
 	} else {
 find_page:
-		ret = -ENOMEM;
-		page = find_or_create_page(mapping, index, gfp_mask);
-		if (!page)
-			goto out_nomem;
+		page = find_lock_page(mapping, index);
+		if (!page) {
+			ret = -ENOMEM;
+			page = page_cache_alloc_cold(mapping);
+			if (unlikely(!page))
+				goto out_nomem;
+
+			/*
+			 * This will also lock the page
+			 */
+			ret = add_to_page_cache_lru(page, mapping, index,
+						    gfp_mask);
+			if (unlikely(ret))
+				goto out;
+		}
 
 		/*
-		 * If the page is uptodate, it is also locked. If it isn't
-		 * uptodate, we can mark it uptodate if we are filling the
-		 * full page. Otherwise we need to read it in first...
+		 * We get here with the page locked. If the page is also
+		 * uptodate, we don't need to do more. If it isn't, we
+		 * may need to bring it in if we are not going to overwrite
+		 * the full page.
 		 */
 		if (!PageUptodate(page)) {
 			if (sd->len < PAGE_CACHE_SIZE) {
@@ -520,10 +577,8 @@ find_page:
 					ret = -EIO;
 					goto out;
 				}
-			} else {
-				WARN_ON(!PageLocked(page));
+			} else
 				SetPageUptodate(page);
-			}
 		}
 	}
 
@@ -552,10 +607,10 @@ find_page:
 	mark_page_accessed(page);
 	balance_dirty_pages_ratelimited(mapping);
 out:
-	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN)) {
+	if (!(buf->flags & PIPE_BUF_FLAG_STOLEN))
 		page_cache_release(page);
-		unlock_page(page);
-	}
+
+	unlock_page(page);
 out_nomem:
 	buf->ops->unmap(info, buf);
 	return ret;
@@ -687,22 +742,26 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
 	ssize_t ret;
 
 	ret = move_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
-
-	/*
-	 * If file or inode is SYNC and we actually wrote some data, sync it.
-	 */
-	if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(mapping->host))
-	    && ret > 0) {
+	if (ret > 0) {
 		struct inode *inode = mapping->host;
-		int err;
 
-		mutex_lock(&inode->i_mutex);
-		err = generic_osync_inode(mapping->host, mapping,
-					  OSYNC_METADATA|OSYNC_DATA);
-		mutex_unlock(&inode->i_mutex);
+		*ppos += ret;
+
+		/*
+		 * If file or inode is SYNC and we actually wrote some data,
+		 * sync it.
+		 */
+		if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(inode))) {
+			int err;
+
+			mutex_lock(&inode->i_mutex);
+			err = generic_osync_inode(inode, mapping,
+						  OSYNC_METADATA|OSYNC_DATA);
+			mutex_unlock(&inode->i_mutex);
 
-		if (err)
-			ret = err;
+			if (err)
+				ret = err;
+		}
 	}
 
 	return ret;
@@ -904,6 +963,7 @@ static long do_splice(struct file *in, loff_t __user *off_in,
 {
 	struct pipe_inode_info *pipe;
 	loff_t offset, *off;
+	long ret;
 
 	pipe = in->f_dentry->d_inode->i_pipe;
 	if (pipe) {
@@ -918,7 +978,12 @@ static long do_splice(struct file *in, loff_t __user *off_in,
 		} else
 			off = &out->f_pos;
 
-		return do_splice_from(pipe, out, off, len, flags);
+		ret = do_splice_from(pipe, out, off, len, flags);
+
+		if (off_out && copy_to_user(off_out, off, sizeof(loff_t)))
+			ret = -EFAULT;
+
+		return ret;
 	}
 
 	pipe = out->f_dentry->d_inode->i_pipe;
@@ -934,7 +999,12 @@ static long do_splice(struct file *in, loff_t __user *off_in,
 		} else
 			off = &in->f_pos;
 
-		return do_splice_to(in, off, pipe, len, flags);
+		ret = do_splice_to(in, off, pipe, len, flags);
+
+		if (off_in && copy_to_user(off_in, off, sizeof(loff_t)))
+			ret = -EFAULT;
+
+		return ret;
 	}
 
 	return -EINVAL;
@@ -979,7 +1049,9 @@ static int link_pipe(struct pipe_inode_info *ipipe,
 		     size_t len, unsigned int flags)
 {
 	struct pipe_buffer *ibuf, *obuf;
-	int ret = 0, do_wakeup = 0, i;
+	int ret, do_wakeup, i, ipipe_first;
+
+	ret = do_wakeup = ipipe_first = 0;
 
 	/*
 	 * Potential ABBA deadlock, work around it by ordering lock
@@ -987,6 +1059,7 @@ static int link_pipe(struct pipe_inode_info *ipipe,
 	 * could deadlock (one doing tee from A -> B, the other from B -> A).
 	 */
 	if (ipipe->inode < opipe->inode) {
+		ipipe_first = 1;
 		mutex_lock(&ipipe->inode->i_mutex);
 		mutex_lock(&opipe->inode->i_mutex);
 	} else {
@@ -1035,9 +1108,11 @@ static int link_pipe(struct pipe_inode_info *ipipe,
 
 			/*
 			 * We have input available, but no output room.
-			 * If we already copied data, return that.
+			 * If we already copied data, return that. If we
+			 * need to drop the opipe lock, it must be ordered
+			 * last to avoid deadlocks.
 			 */
-			if (flags & SPLICE_F_NONBLOCK) {
+			if ((flags & SPLICE_F_NONBLOCK) || !ipipe_first) {
 				if (!ret)
 					ret = -EAGAIN;
 				break;
@@ -1071,7 +1146,12 @@ static int link_pipe(struct pipe_inode_info *ipipe,
 			if (ret)
 				break;
 		}
-		if (flags & SPLICE_F_NONBLOCK) {
+		/*
+		 * pipe_wait() drops the ipipe mutex. To avoid deadlocks
+		 * with another process, we can only safely do that if
+		 * the ipipe lock is ordered last.
+		 */
+		if ((flags & SPLICE_F_NONBLOCK) || ipipe_first) {
 			if (!ret)
 				ret = -EAGAIN;
 			break;