Weekly Edition
Return to the Kernel page
| |||||||||||||
hot-n-cold pages: free and allocate hints
ChangeSet 1.911, 2002/10/30 15:36:13-08:00, akpm@digeo.com
[PATCH] hot-n-cold pages: free and allocate hints
Add a `cold' hint to struct pagevec, and teach truncate and page
reclaim to use it.
Empirical testing showed that truncate's pages tend to be hot. And page
reclaim's are certainly cold.
# This patch includes the following deltas:
# ChangeSet 1.910 -> 1.911
# include/linux/pagemap.h 1.29 -> 1.30
# mm/readahead.c 1.19 -> 1.20
# mm/page_alloc.c 1.120 -> 1.121
# mm/vmscan.c 1.123 -> 1.124
# fs/mpage.c 1.27 -> 1.28
# mm/filemap.c 1.152 -> 1.153
# mm/swap.c 1.38 -> 1.39
# mm/swap_state.c 1.48 -> 1.49
# mm/truncate.c 1.3 -> 1.4
# include/linux/pagevec.h 1.5 -> 1.6
#
fs/mpage.c | 4 ++--
include/linux/pagemap.h | 2 +-
include/linux/pagevec.h | 10 +++++++++-
mm/filemap.c | 2 +-
mm/page_alloc.c | 2 +-
mm/readahead.c | 2 +-
mm/swap.c | 15 ++++++++-------
mm/swap_state.c | 2 +-
mm/truncate.c | 10 +++++++---
mm/vmscan.c | 6 +++---
10 files changed, 34 insertions(+), 21 deletions(-)
diff -Nru a/fs/mpage.c b/fs/mpage.c
--- a/fs/mpage.c Wed Oct 30 16:22:01 2002
+++ b/fs/mpage.c Wed Oct 30 16:22:01 2002
@@ -263,7 +263,7 @@
sector_t last_block_in_bio = 0;
struct pagevec lru_pvec;
- pagevec_init(&lru_pvec);
+ pagevec_init(&lru_pvec, 0);
for (page_idx = 0; page_idx < nr_pages; page_idx++) {
struct page *page = list_entry(pages->prev, struct page, list);
@@ -560,7 +560,7 @@
if (get_block == NULL)
writepage = mapping->a_ops->writepage;
- pagevec_init(&pvec);
+ pagevec_init(&pvec, 0);
write_lock(&mapping->page_lock);
list_splice_init(&mapping->dirty_pages, &mapping->io_pages);
diff -Nru a/include/linux/pagemap.h b/include/linux/pagemap.h
--- a/include/linux/pagemap.h Wed Oct 30 16:22:01 2002
+++ b/include/linux/pagemap.h Wed Oct 30 16:22:01 2002
@@ -24,7 +24,7 @@
#define page_cache_get(page) get_page(page)
#define page_cache_release(page) put_page(page)
-void release_pages(struct page **pages, int nr);
+void release_pages(struct page **pages, int nr, int cold);
static inline struct page *page_cache_alloc(struct address_space *x)
{
diff -Nru a/include/linux/pagevec.h b/include/linux/pagevec.h
--- a/include/linux/pagevec.h Wed Oct 30 16:22:01 2002
+++ b/include/linux/pagevec.h Wed Oct 30 16:22:01 2002
@@ -12,6 +12,7 @@
struct pagevec {
unsigned nr;
+ int cold;
struct page *pages[PAGEVEC_SIZE];
};
@@ -25,7 +26,13 @@
unsigned int pagevec_lookup(struct pagevec *pvec, struct address_space *mapping,
pgoff_t start, unsigned int nr_pages);
-static inline void pagevec_init(struct pagevec *pvec)
+static inline void pagevec_init(struct pagevec *pvec, int cold)
+{
+ pvec->nr = 0;
+ pvec->cold = cold;
+}
+
+static inline void pagevec_reinit(struct pagevec *pvec)
{
pvec->nr = 0;
}
@@ -48,6 +55,7 @@
pvec->pages[pvec->nr++] = page;
return pagevec_space(pvec);
}
+
static inline void pagevec_release(struct pagevec *pvec)
{
diff -Nru a/mm/filemap.c b/mm/filemap.c
--- a/mm/filemap.c Wed Oct 30 16:22:01 2002
+++ b/mm/filemap.c Wed Oct 30 16:22:01 2002
@@ -1449,7 +1449,7 @@
/* We can write back this queue in page reclaim */
current->backing_dev_info = mapping->backing_dev_info;
- pagevec_init(&lru_pvec);
+ pagevec_init(&lru_pvec, 0);
if (unlikely(file->f_error)) {
err = file->f_error;
diff -Nru a/mm/page_alloc.c b/mm/page_alloc.c
--- a/mm/page_alloc.c Wed Oct 30 16:22:01 2002
+++ b/mm/page_alloc.c Wed Oct 30 16:22:01 2002
@@ -548,7 +548,7 @@
int i = pagevec_count(pvec);
while (--i >= 0)
- free_hot_page(pvec->pages[i]);
+ free_hot_cold_page(pvec->pages[i], pvec->cold);
}
void __free_pages(struct page *page, unsigned int order)
diff -Nru a/mm/readahead.c b/mm/readahead.c
--- a/mm/readahead.c Wed Oct 30 16:22:01 2002
+++ b/mm/readahead.c Wed Oct 30 16:22:01 2002
@@ -49,7 +49,7 @@
unsigned page_idx;
struct pagevec lru_pvec;
- pagevec_init(&lru_pvec);
+ pagevec_init(&lru_pvec, 0);
if (mapping->a_ops->readpages)
return mapping->a_ops->readpages(mapping, pages, nr_pages);
diff -Nru a/mm/swap.c b/mm/swap.c
--- a/mm/swap.c Wed Oct 30 16:22:01 2002
+++ b/mm/swap.c Wed Oct 30 16:22:01 2002
@@ -99,13 +99,13 @@
* page count inside the lock to see whether shrink_cache grabbed the page
* via the LRU. If it did, give up: shrink_cache will free it.
*/
-void release_pages(struct page **pages, int nr)
+void release_pages(struct page **pages, int nr, int cold)
{
int i;
struct pagevec pages_to_free;
struct zone *zone = NULL;
- pagevec_init(&pages_to_free);
+ pagevec_init(&pages_to_free, cold);
for (i = 0; i < nr; i++) {
struct page *page = pages[i];
struct zone *pagezone;
@@ -126,7 +126,7 @@
if (!pagevec_add(&pages_to_free, page)) {
spin_unlock_irq(&zone->lru_lock);
__pagevec_free(&pages_to_free);
- pagevec_init(&pages_to_free);
+ pagevec_reinit(&pages_to_free);
zone = NULL; /* No lock is held */
}
}
@@ -139,8 +139,8 @@
void __pagevec_release(struct pagevec *pvec)
{
- release_pages(pvec->pages, pagevec_count(pvec));
- pagevec_init(pvec);
+ release_pages(pvec->pages, pagevec_count(pvec), pvec->cold);
+ pagevec_reinit(pvec);
}
/*
@@ -153,7 +153,8 @@
int i;
struct pagevec pages_to_free;
- pagevec_init(&pages_to_free);
+ pagevec_init(&pages_to_free, pvec->cold);
+ pages_to_free.cold = pvec->cold;
for (i = 0; i < pagevec_count(pvec); i++) {
struct page *page = pvec->pages[i];
@@ -162,7 +163,7 @@
pagevec_add(&pages_to_free, page);
}
pagevec_free(&pages_to_free);
- pagevec_init(pvec);
+ pagevec_reinit(pvec);
}
/*
diff -Nru a/mm/swap_state.c b/mm/swap_state.c
--- a/mm/swap_state.c Wed Oct 30 16:22:01 2002
+++ b/mm/swap_state.c Wed Oct 30 16:22:01 2002
@@ -301,7 +301,7 @@
for (i = 0; i < todo; i++)
free_swap_cache(pagep[i]);
- release_pages(pagep, todo);
+ release_pages(pagep, todo, 0);
pagep += todo;
nr -= todo;
}
diff -Nru a/mm/truncate.c b/mm/truncate.c
--- a/mm/truncate.c Wed Oct 30 16:22:01 2002
+++ b/mm/truncate.c Wed Oct 30 16:22:01 2002
@@ -100,6 +100,10 @@
* When looking at page->index outside the page lock we need to be careful to
* copy it into a local to avoid races (it could change at any time).
*
+ * We pass down the cache-hot hint to the page freeing code. Even if the
+ * mapping is large, it is probably the case that the final pages are the most
+ * recently touched, and freeing happens in ascending file offset order.
+ *
* Called under (and serialised by) inode->i_sem.
*/
void truncate_inode_pages(struct address_space *mapping, loff_t lstart)
@@ -110,7 +114,7 @@
pgoff_t next;
int i;
- pagevec_init(&pvec);
+ pagevec_init(&pvec, 0);
next = start;
while (pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
for (i = 0; i < pagevec_count(&pvec); i++) {
@@ -185,7 +189,7 @@
pgoff_t next = 0;
int i;
- pagevec_init(&pvec);
+ pagevec_init(&pvec, 0);
while (pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
for (i = 0; i < pagevec_count(&pvec); i++) {
struct page *page = pvec.pages[i];
@@ -226,7 +230,7 @@
pgoff_t next = 0;
int i;
- pagevec_init(&pvec);
+ pagevec_init(&pvec, 0);
while (pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
for (i = 0; i < pagevec_count(&pvec); i++) {
struct page *page = pvec.pages[i];
diff -Nru a/mm/vmscan.c b/mm/vmscan.c
--- a/mm/vmscan.c Wed Oct 30 16:22:01 2002
+++ b/mm/vmscan.c Wed Oct 30 16:22:01 2002
@@ -203,7 +203,7 @@
int pgactivate = 0;
int ret = 0;
- pagevec_init(&freed_pvec);
+ pagevec_init(&freed_pvec, 1);
while (!list_empty(page_list)) {
struct page *page;
int may_enter_fs;
@@ -433,7 +433,7 @@
if (nr_to_process < SWAP_CLUSTER_MAX)
nr_to_process = SWAP_CLUSTER_MAX;
- pagevec_init(&pvec);
+ pagevec_init(&pvec, 1);
lru_add_drain();
spin_lock_irq(&zone->lru_lock);
@@ -617,7 +617,7 @@
pgdeactivate++;
}
- pagevec_init(&pvec);
+ pagevec_init(&pvec, 1);
spin_lock_irq(&zone->lru_lock);
while (!list_empty(&l_inactive)) {
page = list_entry(l_inactive.prev, struct page, lru);
-
To unsubscribe from this list: send the line "unsubscribe bk-commits-head" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
|
|||||||||||||