ocfs2: fix i_mutex locking in ocfs2_splice_to_file()
Miklos Szeredi [Tue, 14 Apr 2009 17:48:39 +0000 (19:48 +0200)]
Rearrange locking of i_mutex on destination and call to
ocfs2_rw_lock() so locks are only held while buffers are copied with
the pipe_to_file() actor, and not while waiting for more data on the
pipe.

Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

fs/ocfs2/file.c
fs/splice.c
include/linux/splice.h

index 8672b95..c2a87c8 100644 (file)
@@ -1912,6 +1912,22 @@ out_sems:
        return written ? written : ret;
 }
 
+static int ocfs2_splice_to_file(struct pipe_inode_info *pipe,
+                               struct file *out,
+                               struct splice_desc *sd)
+{
+       int ret;
+
+       ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, &sd->pos,
+                                           sd->total_len, 0, NULL);
+       if (ret < 0) {
+               mlog_errno(ret);
+               return ret;
+       }
+
+       return splice_from_pipe_feed(pipe, sd, pipe_to_file);
+}
+
 static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
                                       struct file *out,
                                       loff_t *ppos,
@@ -1919,38 +1935,76 @@ static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
                                       unsigned int flags)
 {
        int ret;
-       struct inode *inode = out->f_path.dentry->d_inode;
+       struct address_space *mapping = out->f_mapping;
+       struct inode *inode = mapping->host;
+       struct splice_desc sd = {
+               .total_len = len,
+               .flags = flags,
+               .pos = *ppos,
+               .u.file = out,
+       };
 
        mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", out, pipe,
                   (unsigned int)len,
                   out->f_path.dentry->d_name.len,
                   out->f_path.dentry->d_name.name);
 
-       mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
+       if (pipe->inode)
+               mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_PARENT);
 
-       ret = ocfs2_rw_lock(inode, 1);
-       if (ret < 0) {
-               mlog_errno(ret);
-               goto out;
-       }
+       splice_from_pipe_begin(&sd);
+       do {
+               ret = splice_from_pipe_next(pipe, &sd);
+               if (ret <= 0)
+                       break;
 
-       ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, ppos, len, 0,
-                                           NULL);
-       if (ret < 0) {
-               mlog_errno(ret);
-               goto out_unlock;
-       }
+               mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
+               ret = ocfs2_rw_lock(inode, 1);
+               if (ret < 0)
+                       mlog_errno(ret);
+               else {
+                       ret = ocfs2_splice_to_file(pipe, out, &sd);
+                       ocfs2_rw_unlock(inode, 1);
+               }
+               mutex_unlock(&inode->i_mutex);
+       } while (ret > 0);
+       splice_from_pipe_end(pipe, &sd);
 
        if (pipe->inode)
-               mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_CHILD);
-       ret = generic_file_splice_write_nolock(pipe, out, ppos, len, flags);
-       if (pipe->inode)
                mutex_unlock(&pipe->inode->i_mutex);
 
-out_unlock:
-       ocfs2_rw_unlock(inode, 1);
-out:
-       mutex_unlock(&inode->i_mutex);
+       if (sd.num_spliced)
+               ret = sd.num_spliced;
+
+       if (ret > 0) {
+               unsigned long nr_pages;
+
+               *ppos += ret;
+               nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+
+               /*
+                * 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 = ocfs2_rw_lock(inode, 1);
+                       if (err < 0) {
+                               mlog_errno(err);
+                       } else {
+                               err = generic_osync_inode(inode, mapping,
+                                                 OSYNC_METADATA|OSYNC_DATA);
+                               ocfs2_rw_unlock(inode, 1);
+                       }
+                       mutex_unlock(&inode->i_mutex);
+
+                       if (err)
+                               ret = err;
+               }
+               balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
+       }
 
        mlog_exit(ret);
        return ret;
index a1f595b..584b2b7 100644 (file)
@@ -555,8 +555,8 @@ static int pipe_to_sendpage(struct pipe_inode_info *pipe,
  * SPLICE_F_MOVE isn't set, or we cannot move the page, we simply create
  * a new page in the output file page cache and fill/dirty that.
  */
-static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
-                       struct splice_desc *sd)
+int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
+                struct splice_desc *sd)
 {
        struct file *file = sd->u.file;
        struct address_space *mapping = file->f_mapping;
@@ -600,6 +600,7 @@ static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
 out:
        return ret;
 }
+EXPORT_SYMBOL(pipe_to_file);
 
 static void wakeup_pipe_writers(struct pipe_inode_info *pipe)
 {
index 8fc2a63..5f3faa9 100644 (file)
@@ -75,6 +75,8 @@ extern int splice_from_pipe_next(struct pipe_inode_info *,
 extern void splice_from_pipe_begin(struct splice_desc *);
 extern void splice_from_pipe_end(struct pipe_inode_info *,
                                 struct splice_desc *);
+extern int pipe_to_file(struct pipe_inode_info *, struct pipe_buffer *,
+                       struct splice_desc *);
 
 extern ssize_t splice_to_pipe(struct pipe_inode_info *,
                              struct splice_pipe_desc *);