linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
* pages not marked as accessed on non-page boundaries
@ 2004-12-05 14:13 Miquel van Smoorenburg
  2004-12-05 14:44 ` Rik van Riel
  0 siblings, 1 reply; 3+ messages in thread
From: Miquel van Smoorenburg @ 2004-12-05 14:13 UTC (permalink / raw)
  To: linux-mm

(Not sure if this should go to linux-kernel or linux-mm,
 I'll try the latter first)

In the current kernel (I used 2.6.9), pages read into memory
through read() are only marked as accessed if the read
started at offset 0 of the page.

When you have a database accessing small amounts of data
in an index file randomly, then most of those pages will
not be marked as read and will be thrown out too soon.

I noticed this when I was writing a patch for something else-
posix_fadvise(LINUX_FADV_STICKY) support, with which you can
ask the kernel to try to keep the pages of a file in core
a bit more aggressively than normal. I'l probably post that later.

Would it be a good thing to fix this ? Patch is below.


=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

PATCH: mark_page_accessed() for read()s on non-page boundaries

When reading a (partial) page from disk using read(), the kernel only
marks the page as "accessed" if the read started at a page boundary.
This means that files that are accessed randomly at non-page boundaries
(usually database style files) will not be cached properly.

The patch below uses the readahead state instead. If a page is read(),
it is marked as "accessed" if the previous read() was for a different
page, whatever the offset in the page.

Signed-Off-By: Miquel van Smoorenburg <miquels@cistron.nl>

diff --exclude-from=exclude -ruN linux-2.6.9-rc4-tw.ORIG/mm/filemap.c linux-2.6.9-rc4-tw/mm/filemap.c
--- linux-2.6.9-rc4.ORIG/mm/filemap.c	2004-10-23 22:21:18.000000000 +0200
+++ linux-2.6.9-rc4/mm/filemap.c	2004-10-25 12:58:26.000000000 +0200
@@ -718,6 +718,7 @@
 {
 	struct inode *inode = mapping->host;
 	unsigned long index, end_index, offset;
+	unsigned long prev_page;
 	loff_t isize;
 	struct page *cached_page;
 	int error;
@@ -748,6 +749,8 @@
 		}
 		nr = nr - offset;
 
+		prev_page = ra.next_size ? ra.prev_page : (unsigned long)-1;
+
 		cond_resched();
 		page_cache_readahead(mapping, &ra, filp, index);
 
@@ -755,10 +758,13 @@
 		page = find_get_page(mapping, index);
 		if (unlikely(page == NULL)) {
 			handle_ra_miss(mapping, &ra, index);
+			prev_page = (unsigned long)-1;
 			goto no_cached_page;
 		}
-		if (!PageUptodate(page))
+		if (!PageUptodate(page)) {
+			prev_page = (unsigned long)-1;
 			goto page_not_up_to_date;
+		}
 page_ok:
 
 		/* If users can be writing to this page using arbitrary
@@ -769,9 +775,10 @@
 			flush_dcache_page(page);
 
 		/*
-		 * Mark the page accessed if we read the beginning.
+		 * Mark the page accessed only if this was the initial
+		 * read, not for subsequential sub-page sized reads.
 		 */
-		if (!offset)
+		if (prev_page != index)
 			mark_page_accessed(page);
 
 		/*
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>

^ permalink raw reply	[flat|nested] 3+ messages in thread
end of thread, other threads:[~2004-12-06 14:44 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-12-05 14:13 pages not marked as accessed on non-page boundaries Miquel van Smoorenburg
2004-12-05 14:44 ` Rik van Riel
2004-12-06 14:44   ` Miquel van Smoorenburg

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox