[patch, 2.5] highmem.c HASHED_PAGE_VIRTUAL unnecessary pool_lock

Marcus Alanen (maalanen@ra.abo.fi)
Wed, 16 Oct 2002 22:11:07 +0300 (EEST)


pool_lock is equivalent to kmap_lock, so remove pool_lock.

When kmap_lock is taken, so will pool_lock be taken in
set_page_address => we are already protected by kmap_lock.
set_page_address must be made static due to this, but this is no
problem since all callers are in highmem.c anyway.

Note, flush_all_zero_pkmaps and especially map_new_virtual are a bit
slow due to linear scanning of pkmap_count. map_new_virtual could be
done in O(1). But perhaps it's not worth the trouble to create two
sets of very different highmem.c routines for some speedup in the
HASHED_PAGE_VIRTUAL configuration...

Marcus

diff -Naurd --exclude-from=/home/maalanen/linux/base/diff_exclude linus-2.5.43/include/linux/mm.h msa-2.5.43/include/linux/mm.h
--- linus-2.5.43/include/linux/mm.h Wed Oct 16 16:31:16 2002
+++ msa-2.5.43/include/linux/mm.h Wed Oct 16 17:55:58 2002
@@ -308,7 +308,6 @@

#if defined(HASHED_PAGE_VIRTUAL)
void *page_address(struct page *page);
-void set_page_address(struct page *page, void *virtual);
void page_address_init(void);
#endif

diff -Naurd --exclude-from=/home/maalanen/linux/base/diff_exclude linus-2.5.43/mm/highmem.c msa-2.5.43/mm/highmem.c
--- linus-2.5.43/mm/highmem.c Mon Oct 7 22:25:30 2002
+++ msa-2.5.43/mm/highmem.c Wed Oct 16 18:51:35 2002
@@ -48,9 +48,18 @@
* since the last TLB flush - so we can't use it.
* n means that there are (n-1) current users of it.
*/
+
+#if defined(HASHED_PAGE_VIRTUAL)
+void set_page_address(struct page *page, void *virtual);
+#endif
+
#ifdef CONFIG_HIGHMEM
static int pkmap_count[LAST_PKMAP];
static unsigned int last_pkmap_nr;
+/*
+ * protects pkmap_count
+ * protects page_address_pool if hashed virtual pages
+ */
static spinlock_t kmap_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;

pte_t * pkmap_page_table;
@@ -503,7 +512,6 @@
* page_address_map freelist, allocated from page_address_maps.
*/
static struct list_head page_address_pool; /* freelist */
-static spinlock_t pool_lock; /* protects page_address_pool */

/*
* Hash table bucket
@@ -545,7 +553,7 @@
return ret;
}

-void set_page_address(struct page *page, void *virtual)
+static void set_page_address(struct page *page, void *virtual)
{
unsigned long flags;
struct page_address_slot *pas;
@@ -557,11 +565,9 @@
if (virtual) { /* Add */
BUG_ON(list_empty(&page_address_pool));

- spin_lock_irqsave(&pool_lock, flags);
pam = list_entry(page_address_pool.next,
struct page_address_map, list);
list_del(&pam->list);
- spin_unlock_irqrestore(&pool_lock, flags);

pam->page = page;
pam->virtual = virtual;
@@ -575,9 +581,7 @@
if (pam->page == page) {
list_del(&pam->list);
spin_unlock_irqrestore(&pas->lock, flags);
- spin_lock_irqsave(&pool_lock, flags);
list_add_tail(&pam->list, &page_address_pool);
- spin_unlock_irqrestore(&pool_lock, flags);
goto done;
}
}
@@ -600,7 +604,6 @@
INIT_LIST_HEAD(&page_address_htable[i].lh);
spin_lock_init(&page_address_htable[i].lock);
}
- spin_lock_init(&pool_lock);
}

#endif /* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/