splice: adjust balance_dirty_pages_ratelimited() call
authorJens Axboe <jens.axboe@oracle.com>
Fri, 15 Jun 2007 11:10:37 +0000 (13:10 +0200)
committerJens Axboe <jens.axboe@oracle.com>
Fri, 15 Jun 2007 11:10:37 +0000 (13:10 +0200)
As we have potentially dirtied more than 1 page, we should indicate as
such to the dirty page balancing. So call
balance_dirty_pages_ratelimited_nr() and pass in the approximate number
of pages we dirtied.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
fs/splice.c

index cb21136..12d247f 100644 (file)
@@ -811,7 +811,10 @@ generic_file_splice_write_nolock(struct pipe_inode_info *pipe, struct file *out,
 
        ret = __splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
        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,
@@ -824,7 +827,7 @@ generic_file_splice_write_nolock(struct pipe_inode_info *pipe, struct file *out,
                        if (err)
                                ret = err;
                }
-               balance_dirty_pages_ratelimited(mapping);
+               balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
        }
 
        return ret;
@@ -863,7 +866,10 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
 
        ret = splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
        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,
@@ -878,7 +884,7 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
                        if (err)
                                ret = err;
                }
-               balance_dirty_pages_ratelimited(mapping);
+               balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
        }
 
        return ret;