The patch titled
Subject: mm: zswap: multiple zpools support
has been added to the -mm mm-unstable branch. Its filename is
mm-zswap-multiple-zpools-support.patch
This patch will shortly appear at
https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/mm-zswap-multiple-zpools-support.patch
This patch will later appear in the mm-unstable branch at
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next via the mm-everything
branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there every 2-3 working days
------------------------------------------------------
From: Yosry Ahmed <yosryahmed@xxxxxxxxxx>
Subject: mm: zswap: multiple zpools support
Date: Wed, 31 May 2023 22:13:01 +0000
Support using multiple zpools of the same type in zswap, for concurrency
purposes. Add CONFIG_ZSWAP_NR_ZPOOLS_ORDER to control the number of
zpools. The order is specific by the config rather than the absolute
number to guarantee a power of 2. This is useful so that we can use
deterministically link each entry to a zpool by hashing the zswap_entry
pointer.
On a setup with zswap and zsmalloc, comparing a single zpool (current
default) to 32 zpools (by setting CONFIG_ZSWAP_NR_ZPOOLS_ORDER=32) shows
improvements in the zsmalloc lock contention, especially on the swap out
path.
The following shows the perf analysis of the swapout path when 10
workloads are simultaneously reclaiming and refaulting tmpfs pages. There
are some improvements on the swap in path as well, but much less
significant.
1 zpool:
|--28.99%--zswap_frontswap_store
|
<snip>
|
|--8.98%--zpool_map_handle
| |
| --8.98%--zs_zpool_map
| |
| --8.95%--zs_map_object
| |
| --8.38%--_raw_spin_lock
| |
| --7.39%--queued_spin_lock_slowpath
|
|--8.82%--zpool_malloc
| |
| --8.82%--zs_zpool_malloc
| |
| --8.80%--zs_malloc
| |
| |--7.21%--_raw_spin_lock
| | |
| | --6.81%--queued_spin_lock_slowpath
<snip>
32 zpools:
|--16.73%--zswap_frontswap_store
|
<snip>
|
|--1.81%--zpool_malloc
| |
| --1.81%--zs_zpool_malloc
| |
| --1.79%--zs_malloc
| |
| --0.73%--obj_malloc
|
|--1.06%--zswap_update_total_size
|
|--0.59%--zpool_map_handle
| |
| --0.59%--zs_zpool_map
| |
| --0.57%--zs_map_object
| |
| --0.51%--_raw_spin_lock
<snip>
Link: https://lkml.kernel.org/r/20230531221302.1664613-1-yosryahmed@xxxxxxxxxx
Signed-off-by: Yosry Ahmed <yosryahmed@xxxxxxxxxx>
Suggested-by: Yu Zhao <yuzhao@xxxxxxxxxx>
Cc: Dan Streetman <ddstreet@xxxxxxxx>
Cc: Domenico Cerasuolo <cerasuolodomenico@xxxxxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
Cc: Nhat Pham <nphamcs@xxxxxxxxx>
Cc: Seth Jennings <sjenning@xxxxxxxxxx>
Cc: Vitaly Wool <vitaly.wool@xxxxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/Kconfig | 12 ++++++
mm/zswap.c | 93 ++++++++++++++++++++++++++++++++++-----------------
2 files changed, 74 insertions(+), 31 deletions(-)
--- a/mm/Kconfig~mm-zswap-multiple-zpools-support
+++ a/mm/Kconfig
@@ -59,6 +59,18 @@ config ZSWAP_EXCLUSIVE_LOADS
The cost is that if the page was never dirtied and needs to be
swapped out again, it will be re-compressed.
+config ZSWAP_NR_ZPOOLS_ORDER
+ int "Number of zpools in zswap, as power of 2"
+ default 0
+ depends on ZSWAP
+ help
+ This options determines the number of zpools to use for zswap, it
+ will be 1 << CONFIG_ZSWAP_NR_ZPOOLS_ORDER.
+
+ Having multiple zpools helps with concurrency and lock contention
+ on the swap in and swap out paths, but uses a little bit of extra
+ space.
+
choice
prompt "Default compressor"
depends on ZSWAP
--- a/mm/zswap.c~mm-zswap-multiple-zpools-support
+++ a/mm/zswap.c
@@ -137,6 +137,9 @@ static bool zswap_non_same_filled_pages_
module_param_named(non_same_filled_pages_enabled, zswap_non_same_filled_pages_enabled,
bool, 0644);
+/* Number of zpools in zswap_pool */
+#define ZSWAP_NR_ZPOOLS (1 << CONFIG_ZSWAP_NR_ZPOOLS_ORDER)
+
/*********************************
* data structures
**********************************/
@@ -150,7 +153,6 @@ struct crypto_acomp_ctx {
};
struct zswap_pool {
- struct zpool *zpool;
struct crypto_acomp_ctx __percpu *acomp_ctx;
struct kref kref;
struct list_head list;
@@ -158,6 +160,7 @@ struct zswap_pool {
struct work_struct shrink_work;
struct hlist_node node;
char tfm_name[CRYPTO_MAX_ALG_NAME];
+ struct zpool *zpools[ZSWAP_NR_ZPOOLS];
};
/*
@@ -236,7 +239,7 @@ static bool zswap_has_pool;
#define zswap_pool_debug(msg, p) \
pr_debug("%s pool %s/%s\n", msg, (p)->tfm_name, \
- zpool_get_type((p)->zpool))
+ zpool_get_type((p)->zpools[0]))
static int zswap_writeback_entry(struct zpool *pool, unsigned long handle);
static int zswap_pool_get(struct zswap_pool *pool);
@@ -263,11 +266,13 @@ static void zswap_update_total_size(void
{
struct zswap_pool *pool;
u64 total = 0;
+ int i;
rcu_read_lock();
list_for_each_entry_rcu(pool, &zswap_pools, list)
- total += zpool_get_total_size(pool->zpool);
+ for (i = 0; i < ZSWAP_NR_ZPOOLS; i++)
+ total += zpool_get_total_size(pool->zpools[i]);
rcu_read_unlock();
@@ -350,6 +355,16 @@ static void zswap_rb_erase(struct rb_roo
}
}
+static struct zpool *zswap_find_zpool(struct zswap_entry *entry)
+{
+ int i = 0;
+
+ if (ZSWAP_NR_ZPOOLS > 1)
+ i = hash_ptr(entry, ilog2(ZSWAP_NR_ZPOOLS));
+
+ return entry->pool->zpools[i];
+}
+
/*
* Carries out the common pattern of freeing and entry's zpool allocation,
* freeing the entry itself, and decrementing the number of stored pages.
@@ -363,7 +378,7 @@ static void zswap_free_entry(struct zswa
if (!entry->length)
atomic_dec(&zswap_same_filled_pages);
else {
- zpool_free(entry->pool->zpool, entry->handle);
+ zpool_free(zswap_find_zpool(entry), entry->handle);
zswap_pool_put(entry->pool);
}
zswap_entry_cache_free(entry);
@@ -572,7 +587,8 @@ static struct zswap_pool *zswap_pool_fin
list_for_each_entry_rcu(pool, &zswap_pools, list) {
if (strcmp(pool->tfm_name, compressor))
continue;
- if (strcmp(zpool_get_type(pool->zpool), type))
+ /* all zpools share the same type */
+ if (strcmp(zpool_get_type(pool->zpools[0]), type))
continue;
/* if we can't get it, it's about to be destroyed */
if (!zswap_pool_get(pool))
@@ -587,14 +603,17 @@ static void shrink_worker(struct work_st
{
struct zswap_pool *pool = container_of(w, typeof(*pool),
shrink_work);
+ int i;
- if (zpool_shrink(pool->zpool, 1, NULL))
- zswap_reject_reclaim_fail++;
+ for (i = 0; i < ZSWAP_NR_ZPOOLS; i++)
+ if (zpool_shrink(pool->zpools[i], 1, NULL))
+ zswap_reject_reclaim_fail++;
zswap_pool_put(pool);
}
static struct zswap_pool *zswap_pool_create(char *type, char *compressor)
{
+ int i;
struct zswap_pool *pool;
char name[38]; /* 'zswap' + 32 char (max) num + \0 */
gfp_t gfp = __GFP_NORETRY | __GFP_NOWARN | __GFP_KSWAPD_RECLAIM;
@@ -615,15 +634,19 @@ static struct zswap_pool *zswap_pool_cre
if (!pool)
return NULL;
- /* unique name for each pool specifically required by zsmalloc */
- snprintf(name, 38, "zswap%x", atomic_inc_return(&zswap_pools_count));
-
- pool->zpool = zpool_create_pool(type, name, gfp, &zswap_zpool_ops);
- if (!pool->zpool) {
- pr_err("%s zpool not available\n", type);
- goto error;
+ for (i = 0; i < ZSWAP_NR_ZPOOLS; i++) {
+ /* unique name for each pool specifically required by zsmalloc */
+ snprintf(name, 38, "zswap%x",
+ atomic_inc_return(&zswap_pools_count));
+
+ pool->zpools[i] = zpool_create_pool(type, name, gfp,
+ &zswap_zpool_ops);
+ if (!pool->zpools[i]) {
+ pr_err("%s zpool not available\n", type);
+ goto error;
+ }
}
- pr_debug("using %s zpool\n", zpool_get_type(pool->zpool));
+ pr_debug("using %s zpool\n", zpool_get_type(pool->zpools[0]));
strscpy(pool->tfm_name, compressor, sizeof(pool->tfm_name));
@@ -653,8 +676,8 @@ static struct zswap_pool *zswap_pool_cre
error:
if (pool->acomp_ctx)
free_percpu(pool->acomp_ctx);
- if (pool->zpool)
- zpool_destroy_pool(pool->zpool);
+ while (i--)
+ zpool_destroy_pool(pool->zpools[i]);
kfree(pool);
return NULL;
}
@@ -703,11 +726,14 @@ static struct zswap_pool *__zswap_pool_c
static void zswap_pool_destroy(struct zswap_pool *pool)
{
+ int i;
+
zswap_pool_debug("destroying", pool);
cpuhp_state_remove_instance(CPUHP_MM_ZSWP_POOL_PREPARE, &pool->node);
free_percpu(pool->acomp_ctx);
- zpool_destroy_pool(pool->zpool);
+ for (i = 0; i < ZSWAP_NR_ZPOOLS; i++)
+ zpool_destroy_pool(pool->zpools[i]);
kfree(pool);
}
@@ -1160,6 +1186,7 @@ static int zswap_frontswap_store(unsigne
unsigned long handle, value;
char *buf;
u8 *src, *dst;
+ struct zpool *zpool;
struct zswap_header zhdr = { .swpentry = swp_entry(type, offset) };
gfp_t gfp;
@@ -1264,11 +1291,13 @@ static int zswap_frontswap_store(unsigne
}
/* store */
- hlen = zpool_evictable(entry->pool->zpool) ? sizeof(zhdr) : 0;
+ zpool = zswap_find_zpool(entry);
+ hlen = zpool_evictable(zpool) ? sizeof(zhdr) : 0;
gfp = __GFP_NORETRY | __GFP_NOWARN | __GFP_KSWAPD_RECLAIM;
- if (zpool_malloc_support_movable(entry->pool->zpool))
+ if (zpool_malloc_support_movable(zpool))
gfp |= __GFP_HIGHMEM | __GFP_MOVABLE;
- ret = zpool_malloc(entry->pool->zpool, hlen + dlen, gfp, &handle);
+ ret = zpool_malloc(zpool, hlen + dlen, gfp, &handle);
+
if (ret == -ENOSPC) {
zswap_reject_compress_poor++;
goto put_dstmem;
@@ -1277,10 +1306,10 @@ static int zswap_frontswap_store(unsigne
zswap_reject_alloc_fail++;
goto put_dstmem;
}
- buf = zpool_map_handle(entry->pool->zpool, handle, ZPOOL_MM_WO);
+ buf = zpool_map_handle(zpool, handle, ZPOOL_MM_WO);
memcpy(buf, &zhdr, hlen);
memcpy(buf + hlen, dst, dlen);
- zpool_unmap_handle(entry->pool->zpool, handle);
+ zpool_unmap_handle(zpool, handle);
mutex_unlock(acomp_ctx->mutex);
/* populate entry */
@@ -1358,6 +1387,7 @@ static int zswap_frontswap_load(unsigned
u8 *src, *dst, *tmp;
unsigned int dlen;
int ret;
+ struct zpool *zpool;
/* find */
spin_lock(&tree->lock);
@@ -1377,7 +1407,8 @@ static int zswap_frontswap_load(unsigned
goto stats;
}
- if (!zpool_can_sleep_mapped(entry->pool->zpool)) {
+ zpool = zswap_find_zpool(entry);
+ if (!zpool_can_sleep_mapped(zpool)) {
tmp = kmalloc(entry->length, GFP_KERNEL);
if (!tmp) {
ret = -ENOMEM;
@@ -1387,14 +1418,14 @@ static int zswap_frontswap_load(unsigned
/* decompress */
dlen = PAGE_SIZE;
- src = zpool_map_handle(entry->pool->zpool, entry->handle, ZPOOL_MM_RO);
- if (zpool_evictable(entry->pool->zpool))
+ src = zpool_map_handle(zpool, entry->handle, ZPOOL_MM_RO);
+ if (zpool_evictable(zpool))
src += sizeof(struct zswap_header);
- if (!zpool_can_sleep_mapped(entry->pool->zpool)) {
+ if (!zpool_can_sleep_mapped(zpool)) {
memcpy(tmp, src, entry->length);
src = tmp;
- zpool_unmap_handle(entry->pool->zpool, entry->handle);
+ zpool_unmap_handle(zpool, entry->handle);
}
acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
@@ -1406,8 +1437,8 @@ static int zswap_frontswap_load(unsigned
ret = crypto_wait_req(crypto_acomp_decompress(acomp_ctx->req), &acomp_ctx->wait);
mutex_unlock(acomp_ctx->mutex);
- if (zpool_can_sleep_mapped(entry->pool->zpool))
- zpool_unmap_handle(entry->pool->zpool, entry->handle);
+ if (zpool_can_sleep_mapped(zpool))
+ zpool_unmap_handle(zpool, entry->handle);
else
kfree(tmp);
@@ -1563,7 +1594,7 @@ static int zswap_setup(void)
pool = __zswap_pool_create_fallback();
if (pool) {
pr_info("loaded using pool %s/%s\n", pool->tfm_name,
- zpool_get_type(pool->zpool));
+ zpool_get_type(pool->zpools[0]));
list_add(&pool->list, &zswap_pools);
zswap_has_pool = true;
} else {
_
Patches currently in -mm which might be from yosryahmed@xxxxxxxxxx are
memcg-use-seq_buf_do_printk-with-mem_cgroup_print_oom_meminfo.patch
memcg-dump-memorystat-during-cgroup-oom-for-v1.patch
writeback-move-wb_over_bg_thresh-call-outside-lock-section.patch
memcg-flush-stats-non-atomically-in-mem_cgroup_wb_stats.patch
memcg-calculate-root-usage-from-global-state.patch
memcg-remove-mem_cgroup_flush_stats_atomic.patch
cgroup-remove-cgroup_rstat_flush_atomic.patch
mm-zswap-support-exclusive-loads.patch
mm-zswap-multiple-zpools-support.patch
