readahead: combine file_ra_state.prev_index/prev_offset into prev_pos

Combine the file_ra_state members
				unsigned long prev_index
				unsigned int prev_offset
into
				loff_t prev_pos

It is more consistent and better supports huge files.

Thanks to Peter for the nice proposal!

[[email protected]: fix shift overflow]
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Fengguang Wu <[email protected]>
Cc: Rusty Russell <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fs/ext3/dir.c

@@ -143,7 +143,7 @@ static int ext3_readdir(struct file * filp,
 					sb->s_bdev->bd_inode->i_mapping,
 					&filp->f_ra, filp,
 					index, 1);
-			filp->f_ra.prev_index = index;
+			filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
 			bh = ext3_bread(NULL, inode, blk, 0, &err);
 		}
 

fs/ext4/dir.c

@@ -142,7 +142,7 @@ static int ext4_readdir(struct file * filp,
 					sb->s_bdev->bd_inode->i_mapping,
 					&filp->f_ra, filp,
 					index, 1);
-			filp->f_ra.prev_index = index;
+			filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
 			bh = ext4_bread(NULL, inode, blk, 0, &err);
 		}
 

fs/splice.c

@@ -447,7 +447,7 @@ __generic_file_splice_read(struct file *in, loff_t *ppos,
 	 */
 	while (page_nr < nr_pages)
 		page_cache_release(pages[page_nr++]);
-	in->f_ra.prev_index = index;
+	in->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
 
 	if (spd.nr_pages)
 		return splice_to_pipe(pipe, &spd);

include/linux/fs.h

@@ -704,8 +704,7 @@ struct file_ra_state {
 
 	unsigned int ra_pages;		/* Maximum readahead window */
 	int mmap_miss;			/* Cache miss stat for mmap accesses */
-	unsigned long prev_index;	/* Cache last read() position */
-	unsigned int prev_offset;	/* Offset where last read() ended in a page */
+	loff_t prev_pos;		/* Cache last read() position */
 };
 
 /*

mm/filemap.c

@@ -879,8 +879,8 @@ void do_generic_mapping_read(struct address_space *mapping,
 	cached_page = NULL;
 	index = *ppos >> PAGE_CACHE_SHIFT;
 	next_index = index;
-	prev_index = ra.prev_index;
-	prev_offset = ra.prev_offset;
+	prev_index = ra.prev_pos >> PAGE_CACHE_SHIFT;
+	prev_offset = ra.prev_pos & (PAGE_CACHE_SIZE-1);
 	last_index = (*ppos + desc->count + PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT;
 	offset = *ppos & ~PAGE_CACHE_MASK;
 
@@ -966,7 +966,6 @@ void do_generic_mapping_read(struct address_space *mapping,
 		index += offset >> PAGE_CACHE_SHIFT;
 		offset &= ~PAGE_CACHE_MASK;
 		prev_offset = offset;
-		ra.prev_offset = offset;
 
 		page_cache_release(page);
 		if (ret == nr && desc->count)
@@ -1056,9 +1055,11 @@ void do_generic_mapping_read(struct address_space *mapping,
 
 out:
 	*_ra = ra;
-	_ra->prev_index = prev_index;
+	_ra->prev_pos = prev_index;
+	_ra->prev_pos <<= PAGE_CACHE_SHIFT;
+	_ra->prev_pos |= prev_offset;
 
-	*ppos = ((loff_t) index << PAGE_CACHE_SHIFT) + offset;
+	*ppos = ((loff_t)index << PAGE_CACHE_SHIFT) + offset;
 	if (cached_page)
 		page_cache_release(cached_page);
 	if (filp)
@@ -1396,7 +1397,7 @@ int filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	 * Found the page and have a reference on it.
 	 */
 	mark_page_accessed(page);
-	ra->prev_index = page->index;
+	ra->prev_pos = (loff_t)page->index << PAGE_CACHE_SHIFT;
 	vmf->page = page;
 	return ret | VM_FAULT_LOCKED;
 

mm/readahead.c

@@ -46,7 +46,7 @@ void
 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping)
 {
 	ra->ra_pages = mapping->backing_dev_info->ra_pages;
-	ra->prev_index = -1;
+	ra->prev_pos = -1;
 }
 EXPORT_SYMBOL_GPL(file_ra_state_init);
 
@@ -327,7 +327,7 @@ static unsigned long get_next_ra_size(struct file_ra_state *ra,
  * indicator. The flag won't be set on already cached pages, to avoid the
  * readahead-for-nothing fuss, saving pointless page cache lookups.
  *
- * prev_index tracks the last visited page in the _previous_ read request.
+ * prev_pos tracks the last visited byte in the _previous_ read request.
  * It should be maintained by the caller, and will be used for detecting
  * small random reads. Note that the readahead algorithm checks loosely
  * for sequential patterns. Hence interleaved reads might be served as
@@ -351,11 +351,9 @@ ondemand_readahead(struct address_space *mapping,
 		   bool hit_readahead_marker, pgoff_t offset,
 		   unsigned long req_size)
 {
-	int max;	/* max readahead pages */
-	int sequential;
-
-	max = ra->ra_pages;
-	sequential = (offset - ra->prev_index <= 1UL) || (req_size > max);
+	int	max = ra->ra_pages;	/* max readahead pages */
+	pgoff_t prev_offset;
+	int	sequential;
 
 	/*
 	 * It's the expected callback offset, assume sequential access.
@@ -369,6 +367,9 @@ ondemand_readahead(struct address_space *mapping,
 		goto readit;
 	}
 
+	prev_offset = ra->prev_pos >> PAGE_CACHE_SHIFT;
+	sequential = offset - prev_offset <= 1UL || req_size > max;
+
 	/*
 	 * Standalone, small read.
 	 * Read as is, and do not pollute the readahead state.