]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
[PATCH] [MTD] block2mtd: remove readahead code
authorJoern Engel <joern@wh.fh-wedel.de>
Tue, 20 Feb 2007 19:22:22 +0000 (20:22 +0100)
committerJoern Engel <joern@wh.fh-wedel.de>
Tue, 20 Feb 2007 19:22:22 +0000 (20:22 +0100)
Over the years there was a slow trickle of complaints against the readahead
code.  Most of them concerned performance, Peter Zijlstra stumbled over it
when working unrelated changes and I believe there was an actual bug report.
Oh, Andrew Morton also complained about duplicating code from mm/readahead.c.

It is just not worth it.  On flash media like usb sticks, readahead will
make things go slow - very slow.  On spinning disks, readahead may be a
win, but this is definitely not the place to add it.

Signed-off-by: Jörn Engel <joern@lazybastard.org>
drivers/mtd/devices/block2mtd.c

index 3bc92153e2b697450fe7ac1e6cb17eb05388fff5..ce47544dc120adda76a28ee4babba0269aed875e 100644 (file)
@@ -40,56 +40,9 @@ struct block2mtd_dev {
 static LIST_HEAD(blkmtd_device_list);
 
 
-#define PAGE_READAHEAD 64
-static void cache_readahead(struct address_space *mapping, int index)
+static struct page* page_read(struct address_space *mapping, int index)
 {
        filler_t *filler = (filler_t*)mapping->a_ops->readpage;
-       int i, pagei;
-       unsigned ret = 0;
-       unsigned long end_index;
-       struct page *page;
-       LIST_HEAD(page_pool);
-       struct inode *inode = mapping->host;
-       loff_t isize = i_size_read(inode);
-
-       if (!isize) {
-               INFO("iSize=0 in cache_readahead\n");
-               return;
-       }
-
-       end_index = ((isize - 1) >> PAGE_CACHE_SHIFT);
-
-       read_lock_irq(&mapping->tree_lock);
-       for (i = 0; i < PAGE_READAHEAD; i++) {
-               pagei = index + i;
-               if (pagei > end_index) {
-                       INFO("Overrun end of disk in cache readahead\n");
-                       break;
-               }
-               page = radix_tree_lookup(&mapping->page_tree, pagei);
-               if (page && (!i))
-                       break;
-               if (page)
-                       continue;
-               read_unlock_irq(&mapping->tree_lock);
-               page = page_cache_alloc_cold(mapping);
-               read_lock_irq(&mapping->tree_lock);
-               if (!page)
-                       break;
-               page->index = pagei;
-               list_add(&page->lru, &page_pool);
-               ret++;
-       }
-       read_unlock_irq(&mapping->tree_lock);
-       if (ret)
-               read_cache_pages(mapping, &page_pool, filler, NULL);
-}
-
-
-static struct page* page_readahead(struct address_space *mapping, int index)
-{
-       filler_t *filler = (filler_t*)mapping->a_ops->readpage;
-       cache_readahead(mapping, index);
        return read_cache_page(mapping, index, filler, NULL);
 }
 
@@ -105,7 +58,7 @@ static int _block2mtd_erase(struct block2mtd_dev *dev, loff_t to, size_t len)
        u_long *max;
 
        while (pages) {
-               page = page_readahead(mapping, index);
+               page = page_read(mapping, index);
                if (!page)
                        return -ENOMEM;
                if (IS_ERR(page))
@@ -174,8 +127,7 @@ static int block2mtd_read(struct mtd_info *mtd, loff_t from, size_t len,
                        cpylen = len;   // this page
                len = len - cpylen;
 
-               //      Get page
-               page = page_readahead(dev->blkdev->bd_inode->i_mapping, index);
+               page = page_read(dev->blkdev->bd_inode->i_mapping, index);
                if (!page)
                        return -ENOMEM;
                if (IS_ERR(page))
@@ -213,8 +165,7 @@ static int _block2mtd_write(struct block2mtd_dev *dev, const u_char *buf,
                        cpylen = len;                   // this page
                len = len - cpylen;
 
-               //      Get page
-               page = page_readahead(mapping, index);
+               page = page_read(mapping, index);
                if (!page)
                        return -ENOMEM;
                if (IS_ERR(page))