swap: add per-partition lock for swapfile
swap_lock is heavily contended when I test swap to 3 fast SSD (even
slightly slower than swap to 2 such SSD). The main contention comes
from swap_info_get(). This patch tries to fix the gap with adding a new
per-partition lock.
Global data like nr_swapfiles, total_swap_pages, least_priority and
swap_list are still protected by swap_lock.
nr_swap_pages is an atomic now, it can be changed without swap_lock. In
theory, it's possible get_swap_page() finds no swap pages but actually
there are free swap pages. But sounds not a big problem.
Accessing partition specific data (like scan_swap_map and so on) is only
protected by swap_info_struct.lock.
Changing swap_info_struct.flags need hold swap_lock and
swap_info_struct.lock, because scan_scan_map() will check it. read the
flags is ok with either the locks hold.
If both swap_lock and swap_info_struct.lock must be hold, we always hold
the former first to avoid deadlock.
swap_entry_free() can change swap_list. To delete that code, we add a
new highest_priority_index. Whenever get_swap_page() is called, we
check it. If it's valid, we use it.
It's a pity get_swap_page() still holds swap_lock(). But in practice,
swap_lock() isn't heavily contended in my test with this patch (or I can
say there are other much more heavier bottlenecks like TLB flush). And
BTW, looks get_swap_page() doesn't really need the lock. We never free
swap_info[] and we check SWAP_WRITEOK flag. The only risk without the
lock is we could swapout to some low priority swap, but we can quickly
recover after several rounds of swap, so sounds not a big deal to me.
But I'd prefer to fix this if it's a real problem.
"swap: make each swap partition have one address_space" improved the
swapout speed from 1.7G/s to 2G/s. This patch further improves the
speed to 2.3G/s, so around 15% improvement. It's a multi-process test,
so TLB flush isn't the biggest bottleneck before the patches.
Change-Id: Ic8d600653479c0e6767f04462c3f057b0fe94332
[arnd@arndb.de: fix it for nommu]
[hughd@google.com: add missing unlock]
[minchan@kernel.org: get rid of lockdep whinge on sys_swapon]
Signed-off-by: Shaohua Li <shli@fusionio.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Seth Jennings <sjenning@linux.vnet.ibm.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Dan Magenheimer <dan.magenheimer@oracle.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Git-commit: ec8acf20afb8534ed511f6613dd2226b9e301010
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
[lauraa@codeaurora.org: Various context fixups due to backport]
Signed-off-by: Laura Abbott <lauraa@codeaurora.org>
diff --git a/arch/sparc/mm/init_32.c b/arch/sparc/mm/init_32.c
index c5f9021..fbf4eeb 100644
--- a/arch/sparc/mm/init_32.c
+++ b/arch/sparc/mm/init_32.c
@@ -77,7 +77,7 @@
printk("Mem-info:\n");
show_free_areas(filter);
printk("Free swap: %6ldkB\n",
- nr_swap_pages << (PAGE_SHIFT-10));
+ get_nr_swap_pages() << (PAGE_SHIFT-10));
printk("%ld pages of RAM\n", totalram_pages);
printk("%ld free pages\n", nr_free_pages());
#if 0 /* undefined pgtable_cache_size, pgd_cache_size */
diff --git a/arch/tile/mm/pgtable.c b/arch/tile/mm/pgtable.c
index 2410aa8..ea7dd38 100644
--- a/arch/tile/mm/pgtable.c
+++ b/arch/tile/mm/pgtable.c
@@ -61,7 +61,7 @@
global_page_state(NR_PAGETABLE),
global_page_state(NR_BOUNCE),
global_page_state(NR_FILE_PAGES),
- nr_swap_pages);
+ get_nr_swap_pages());
for_each_zone(zone) {
unsigned long flags, order, total = 0, largest_order = -1;
diff --git a/include/linux/swap.h b/include/linux/swap.h
index d5bd6ee..a67c48a 100644
--- a/include/linux/swap.h
+++ b/include/linux/swap.h
@@ -197,6 +197,18 @@
struct block_device *bdev; /* swap device or bdev of swap file */
struct file *swap_file; /* seldom referenced */
unsigned int old_block_size; /* seldom referenced */
+ spinlock_t lock; /*
+ * protect map scan related fields like
+ * swap_map, lowest_bit, highest_bit,
+ * inuse_pages, cluster_next,
+ * cluster_nr, lowest_alloc and
+ * highest_alloc. other fields are only
+ * changed at swapon/swapoff, so are
+ * protected by swap_lock. changing
+ * flags need hold this lock and
+ * swap_lock. If both locks need hold,
+ * hold swap_lock first.
+ */
};
struct swap_list_t {
@@ -204,9 +216,6 @@
int next; /* swapfile to be used next */
};
-/* Swap 50% full? Release swapcache more aggressively.. */
-#define vm_swap_full() (nr_swap_pages*2 < total_swap_pages)
-
/* linux/mm/page_alloc.c */
extern unsigned long totalram_pages;
extern unsigned long totalreserve_pages;
@@ -335,8 +344,20 @@
struct vm_area_struct *vma, unsigned long addr);
/* linux/mm/swapfile.c */
-extern long nr_swap_pages;
+extern atomic_long_t nr_swap_pages;
extern long total_swap_pages;
+
+/* Swap 50% full? Release swapcache more aggressively.. */
+static inline bool vm_swap_full(void)
+{
+ return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
+}
+
+static inline long get_nr_swap_pages(void)
+{
+ return atomic_long_read(&nr_swap_pages);
+}
+
extern void si_swapinfo(struct sysinfo *);
extern swp_entry_t get_swap_page(void);
extern swp_entry_t get_swap_page_of_type(int);
@@ -368,9 +389,10 @@
#else /* CONFIG_SWAP */
-#define nr_swap_pages 0L
+#define get_nr_swap_pages() 0L
#define total_swap_pages 0L
#define total_swapcache_pages 0UL
+#define vm_swap_full() 0
#define si_swapinfo(val) \
do { (val)->freeswap = (val)->totalswap = 0; } while (0)
diff --git a/mm/mmap.c b/mm/mmap.c
index 848ef52..9932edb 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -133,7 +133,7 @@
*/
free -= global_page_state(NR_SHMEM);
- free += nr_swap_pages;
+ free += get_nr_swap_pages();
/*
* Any slabs which are created with the
diff --git a/mm/nommu.c b/mm/nommu.c
index bb8f4f0..6bb7042 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -1928,7 +1928,7 @@
*/
free -= global_page_state(NR_SHMEM);
- free += nr_swap_pages;
+ free += get_nr_swap_pages();
/*
* Any slabs which are created with the
diff --git a/mm/swap_state.c b/mm/swap_state.c
index 4c5ff7f..eb6a79c 100644
--- a/mm/swap_state.c
+++ b/mm/swap_state.c
@@ -58,7 +58,8 @@
printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
swap_cache_info.add_total, swap_cache_info.del_total,
swap_cache_info.find_success, swap_cache_info.find_total);
- printk("Free swap = %ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
+ printk("Free swap = %ldkB\n",
+ get_nr_swap_pages() << (PAGE_SHIFT - 10));
printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
}
diff --git a/mm/swapfile.c b/mm/swapfile.c
index fafc26d..5f62c5f 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -44,9 +44,11 @@
static DEFINE_SPINLOCK(swap_lock);
static unsigned int nr_swapfiles;
-long nr_swap_pages;
+atomic_long_t nr_swap_pages;
+/* protected with swap_lock. reading in vm_swap_full() doesn't need lock */
long total_swap_pages;
static int least_priority;
+static atomic_t highest_priority_index = ATOMIC_INIT(-1);
static const char Bad_file[] = "Bad swap file entry ";
static const char Unused_file[] = "Unused swap file entry ";
@@ -220,7 +222,7 @@
si->lowest_alloc = si->max;
si->highest_alloc = 0;
}
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
/*
* If seek is expensive, start searching for new cluster from
@@ -239,7 +241,7 @@
if (si->swap_map[offset])
last_in_cluster = offset + SWAPFILE_CLUSTER;
else if (offset == last_in_cluster) {
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
offset -= SWAPFILE_CLUSTER - 1;
si->cluster_next = offset;
si->cluster_nr = SWAPFILE_CLUSTER - 1;
@@ -260,7 +262,7 @@
if (si->swap_map[offset])
last_in_cluster = offset + SWAPFILE_CLUSTER;
else if (offset == last_in_cluster) {
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
offset -= SWAPFILE_CLUSTER - 1;
si->cluster_next = offset;
si->cluster_nr = SWAPFILE_CLUSTER - 1;
@@ -274,7 +276,7 @@
}
offset = scan_base;
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
si->cluster_nr = SWAPFILE_CLUSTER - 1;
si->lowest_alloc = 0;
}
@@ -290,9 +292,9 @@
/* reuse swap entry of cache-only swap if not busy. */
if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
int swap_was_freed;
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
swap_was_freed = __try_to_reclaim_swap(si, offset);
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
/* entry was freed successfully, try to use this again */
if (swap_was_freed)
goto checks;
@@ -332,13 +334,13 @@
si->lowest_alloc <= last_in_cluster)
last_in_cluster = si->lowest_alloc - 1;
si->flags |= SWP_DISCARDING;
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
if (offset < last_in_cluster)
discard_swap_cluster(si, offset,
last_in_cluster - offset + 1);
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
si->lowest_alloc = 0;
si->flags &= ~SWP_DISCARDING;
@@ -352,10 +354,10 @@
* could defer that delay until swap_writepage,
* but it's easier to keep this self-contained.
*/
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
wait_on_bit(&si->flags, ilog2(SWP_DISCARDING),
wait_for_discard, TASK_UNINTERRUPTIBLE);
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
} else {
/*
* Note pages allocated by racing tasks while
@@ -371,14 +373,14 @@
return offset;
scan:
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
while (++offset <= si->highest_bit) {
if (!si->swap_map[offset]) {
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
goto checks;
}
if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
goto checks;
}
if (unlikely(--latency_ration < 0)) {
@@ -389,11 +391,11 @@
offset = si->lowest_bit;
while (++offset < scan_base) {
if (!si->swap_map[offset]) {
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
goto checks;
}
if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
goto checks;
}
if (unlikely(--latency_ration < 0)) {
@@ -401,7 +403,7 @@
latency_ration = LATENCY_LIMIT;
}
}
- spin_lock(&swap_lock);
+ spin_lock(&si->lock);
no_page:
si->flags -= SWP_SCANNING;
@@ -414,13 +416,34 @@
pgoff_t offset;
int type, next;
int wrapped = 0;
+ int hp_index;
spin_lock(&swap_lock);
- if (nr_swap_pages <= 0)
+ if (atomic_long_read(&nr_swap_pages) <= 0)
goto noswap;
- nr_swap_pages--;
+ atomic_long_dec(&nr_swap_pages);
for (type = swap_list.next; type >= 0 && wrapped < 2; type = next) {
+ hp_index = atomic_xchg(&highest_priority_index, -1);
+ /*
+ * highest_priority_index records current highest priority swap
+ * type which just frees swap entries. If its priority is
+ * higher than that of swap_list.next swap type, we use it. It
+ * isn't protected by swap_lock, so it can be an invalid value
+ * if the corresponding swap type is swapoff. We double check
+ * the flags here. It's even possible the swap type is swapoff
+ * and swapon again and its priority is changed. In such rare
+ * case, low prority swap type might be used, but eventually
+ * high priority swap will be used after several rounds of
+ * swap.
+ */
+ if (hp_index != -1 && hp_index != type &&
+ swap_info[type]->prio < swap_info[hp_index]->prio &&
+ (swap_info[hp_index]->flags & SWP_WRITEOK)) {
+ type = hp_index;
+ swap_list.next = type;
+ }
+
si = swap_info[type];
next = si->next;
if (next < 0 ||
@@ -429,22 +452,29 @@
wrapped++;
}
- if (!si->highest_bit)
+ spin_lock(&si->lock);
+ if (!si->highest_bit) {
+ spin_unlock(&si->lock);
continue;
- if (!(si->flags & SWP_WRITEOK))
+ }
+ if (!(si->flags & SWP_WRITEOK)) {
+ spin_unlock(&si->lock);
continue;
+ }
swap_list.next = next;
+
+ spin_unlock(&swap_lock);
/* This is called for allocating swap entry for cache */
offset = scan_swap_map(si, SWAP_HAS_CACHE);
- if (offset) {
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
+ if (offset)
return swp_entry(type, offset);
- }
+ spin_lock(&swap_lock);
next = swap_list.next;
}
- nr_swap_pages++;
+ atomic_long_inc(&nr_swap_pages);
noswap:
spin_unlock(&swap_lock);
return (swp_entry_t) {0};
@@ -456,19 +486,19 @@
struct swap_info_struct *si;
pgoff_t offset;
- spin_lock(&swap_lock);
si = swap_info[type];
+ spin_lock(&si->lock);
if (si && (si->flags & SWP_WRITEOK)) {
- nr_swap_pages--;
+ atomic_long_dec(&nr_swap_pages);
/* This is called for allocating swap entry, not cache */
offset = scan_swap_map(si, 1);
if (offset) {
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
return swp_entry(type, offset);
}
- nr_swap_pages++;
+ atomic_long_inc(&nr_swap_pages);
}
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
return (swp_entry_t) {0};
}
@@ -490,7 +520,7 @@
goto bad_offset;
if (!p->swap_map[offset])
goto bad_free;
- spin_lock(&swap_lock);
+ spin_lock(&p->lock);
return p;
bad_free:
@@ -508,6 +538,27 @@
return NULL;
}
+/*
+ * This swap type frees swap entry, check if it is the highest priority swap
+ * type which just frees swap entry. get_swap_page() uses
+ * highest_priority_index to search highest priority swap type. The
+ * swap_info_struct.lock can't protect us if there are multiple swap types
+ * active, so we use atomic_cmpxchg.
+ */
+static void set_highest_priority_index(int type)
+{
+ int old_hp_index, new_hp_index;
+
+ do {
+ old_hp_index = atomic_read(&highest_priority_index);
+ if (old_hp_index != -1 &&
+ swap_info[old_hp_index]->prio >= swap_info[type]->prio)
+ break;
+ new_hp_index = type;
+ } while (atomic_cmpxchg(&highest_priority_index,
+ old_hp_index, new_hp_index) != old_hp_index);
+}
+
static unsigned char swap_entry_free(struct swap_info_struct *p,
swp_entry_t entry, unsigned char usage)
{
@@ -551,10 +602,8 @@
p->lowest_bit = offset;
if (offset > p->highest_bit)
p->highest_bit = offset;
- if (swap_list.next >= 0 &&
- p->prio > swap_info[swap_list.next]->prio)
- swap_list.next = p->type;
- nr_swap_pages++;
+ set_highest_priority_index(p->type);
+ atomic_long_inc(&nr_swap_pages);
p->inuse_pages--;
if ((p->flags & SWP_BLKDEV) &&
disk->fops->swap_slot_free_notify)
@@ -575,7 +624,7 @@
p = swap_info_get(entry);
if (p) {
swap_entry_free(p, entry, 1);
- spin_unlock(&swap_lock);
+ spin_unlock(&p->lock);
}
}
@@ -592,7 +641,7 @@
count = swap_entry_free(p, entry, SWAP_HAS_CACHE);
if (page)
mem_cgroup_uncharge_swapcache(page, entry, count != 0);
- spin_unlock(&swap_lock);
+ spin_unlock(&p->lock);
}
}
@@ -611,7 +660,7 @@
p = swap_info_get(entry);
if (p) {
count = swap_count(p->swap_map[swp_offset(entry)]);
- spin_unlock(&swap_lock);
+ spin_unlock(&p->lock);
}
return count;
}
@@ -699,7 +748,7 @@
page = NULL;
}
}
- spin_unlock(&swap_lock);
+ spin_unlock(&p->lock);
}
if (page) {
/*
@@ -828,11 +877,13 @@
if ((unsigned int)type < nr_swapfiles) {
struct swap_info_struct *sis = swap_info[type];
+ spin_lock(&sis->lock);
if (sis->flags & SWP_WRITEOK) {
n = sis->pages;
if (free)
n -= sis->inuse_pages;
}
+ spin_unlock(&sis->lock);
}
spin_unlock(&swap_lock);
return n;
@@ -1528,7 +1579,7 @@
p->prio = --least_priority;
p->swap_map = swap_map;
p->flags |= SWP_WRITEOK;
- nr_swap_pages += p->pages;
+ atomic_long_add(p->pages, &nr_swap_pages);
total_swap_pages += p->pages;
/* insert swap space into swap_list: */
@@ -1605,14 +1656,16 @@
/* just pick something that's safe... */
swap_list.next = swap_list.head;
}
+ spin_lock(&p->lock);
if (p->prio < 0) {
for (i = p->next; i >= 0; i = swap_info[i]->next)
swap_info[i]->prio = p->prio--;
least_priority++;
}
- nr_swap_pages -= p->pages;
+ atomic_long_sub(p->pages, &nr_swap_pages);
total_swap_pages -= p->pages;
p->flags &= ~SWP_WRITEOK;
+ spin_unlock(&p->lock);
spin_unlock(&swap_lock);
oom_score_adj = test_set_oom_score_adj(OOM_SCORE_ADJ_MAX);
@@ -1637,14 +1690,17 @@
mutex_lock(&swapon_mutex);
spin_lock(&swap_lock);
+ spin_lock(&p->lock);
drain_mmlist();
/* wait for anyone still in scan_swap_map */
p->highest_bit = 0; /* cuts scans short */
while (p->flags >= SWP_SCANNING) {
+ spin_unlock(&p->lock);
spin_unlock(&swap_lock);
schedule_timeout_uninterruptible(1);
spin_lock(&swap_lock);
+ spin_lock(&p->lock);
}
swap_file = p->swap_file;
@@ -1653,6 +1709,7 @@
swap_map = p->swap_map;
p->swap_map = NULL;
p->flags = 0;
+ spin_unlock(&p->lock);
spin_unlock(&swap_lock);
mutex_unlock(&swapon_mutex);
vfree(swap_map);
@@ -1856,6 +1913,7 @@
p->flags = SWP_USED;
p->next = -1;
spin_unlock(&swap_lock);
+ spin_lock_init(&p->lock);
return p;
}
@@ -2177,7 +2235,7 @@
if ((si->flags & SWP_USED) && !(si->flags & SWP_WRITEOK))
nr_to_be_unused += si->inuse_pages;
}
- val->freeswap = nr_swap_pages + nr_to_be_unused;
+ val->freeswap = atomic_long_read(&nr_swap_pages) + nr_to_be_unused;
val->totalswap = total_swap_pages + nr_to_be_unused;
spin_unlock(&swap_lock);
}
@@ -2210,7 +2268,7 @@
p = swap_info[type];
offset = swp_offset(entry);
- spin_lock(&swap_lock);
+ spin_lock(&p->lock);
if (unlikely(offset >= p->max))
goto unlock_out;
@@ -2245,7 +2303,7 @@
p->swap_map[offset] = count | has_cache;
unlock_out:
- spin_unlock(&swap_lock);
+ spin_unlock(&p->lock);
out:
return err;
@@ -2345,7 +2403,7 @@
}
if (!page) {
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
return -ENOMEM;
}
@@ -2393,7 +2451,7 @@
list_add_tail(&page->lru, &head->lru);
page = NULL; /* now it's attached, don't free it */
out:
- spin_unlock(&swap_lock);
+ spin_unlock(&si->lock);
outer:
if (page)
__free_page(page);
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 1438de9..22c215d 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1654,7 +1654,7 @@
force_scan = true;
/* If we have no swap space, do not bother scanning anon pages. */
- if (!sc->may_swap || (nr_swap_pages <= 0)) {
+ if (!sc->may_swap || (get_nr_swap_pages() <= 0)) {
noswap = 1;
fraction[0] = 0;
fraction[1] = 1;
@@ -1798,7 +1798,7 @@
*/
pages_for_compaction = (2UL << sc->order);
inactive_lru_pages = zone_nr_lru_pages(mz, LRU_INACTIVE_FILE);
- if (nr_swap_pages > 0)
+ if (get_nr_swap_pages() > 0)
inactive_lru_pages += zone_nr_lru_pages(mz, LRU_INACTIVE_ANON);
if (sc->nr_reclaimed < pages_for_compaction &&
inactive_lru_pages > pages_for_compaction)
@@ -2912,7 +2912,7 @@
nr = global_page_state(NR_ACTIVE_FILE) +
global_page_state(NR_INACTIVE_FILE);
- if (nr_swap_pages > 0)
+ if (get_nr_swap_pages() > 0)
nr += global_page_state(NR_ACTIVE_ANON) +
global_page_state(NR_INACTIVE_ANON);