[PATCH v2 3/4] zram: make deduplication feature optional
From: js1304
Date: Thu Mar 30 2017 - 01:40:16 EST
From: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
Benefit of deduplication is dependent on the workload so it's not
preferable to always enable. Therefore, make it optional in Kconfig
and device param. Default is 'off'. This option will be beneficial
for users who use the zram as blockdev and stores build output to it.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
---
Documentation/ABI/testing/sysfs-block-zram | 10 +++++
Documentation/blockdev/zram.txt | 1 +
drivers/block/zram/Kconfig | 14 +++++++
drivers/block/zram/Makefile | 5 ++-
drivers/block/zram/zram_dedup.c | 31 ++++++++++++++-
drivers/block/zram/zram_dedup.h | 33 +++++++++++++++-
drivers/block/zram/zram_drv.c | 62 ++++++++++++++++++++++++++----
drivers/block/zram/zram_drv.h | 1 +
8 files changed, 146 insertions(+), 11 deletions(-)
diff --git a/Documentation/ABI/testing/sysfs-block-zram b/Documentation/ABI/testing/sysfs-block-zram
index 451b6d8..3c1945f 100644
--- a/Documentation/ABI/testing/sysfs-block-zram
+++ b/Documentation/ABI/testing/sysfs-block-zram
@@ -90,3 +90,13 @@ Description:
device's debugging info useful for kernel developers. Its
format is not documented intentionally and may change
anytime without any notice.
+
+What: /sys/block/zram<id>/use_dedup
+Date: March 2017
+Contact: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
+Description:
+ The use_dedup file is read-write and specifies deduplication
+ feature is used or not. If enabled, duplicated data is
+ managed by reference count and will not be stored in memory
+ twice. Benefit of this feature largely depends on the workload
+ so keep attention when use.
diff --git a/Documentation/blockdev/zram.txt b/Documentation/blockdev/zram.txt
index 2cdc303..cbbe39b 100644
--- a/Documentation/blockdev/zram.txt
+++ b/Documentation/blockdev/zram.txt
@@ -168,6 +168,7 @@ max_comp_streams RW the number of possible concurrent compress operations
comp_algorithm RW show and change the compression algorithm
compact WO trigger memory compaction
debug_stat RO this file is used for zram debugging purposes
+use_dedup RW show and set deduplication feature
User space is advised to use the following files to read the device statistics.
diff --git a/drivers/block/zram/Kconfig b/drivers/block/zram/Kconfig
index b8ecba6..2f3dd1f 100644
--- a/drivers/block/zram/Kconfig
+++ b/drivers/block/zram/Kconfig
@@ -13,3 +13,17 @@ config ZRAM
disks and maybe many more.
See zram.txt for more information.
+
+config ZRAM_DEDUP
+ bool "Deduplication support for ZRAM data"
+ depends on ZRAM
+ default n
+ help
+ Deduplicate ZRAM data to reduce amount of memory consumption.
+ Advantage largely depends on the workload. In some cases, this
+ option reduces memory usage to the half. However, if there is no
+ duplicated data, the amount of memory consumption would be
+ increased due to additional metadata usage. And, there is
+ computation time trade-off. Please check the benefit before
+ enabling this option. Experiment shows the positive effect when
+ the zram is used as blockdev and is used to store build output.
diff --git a/drivers/block/zram/Makefile b/drivers/block/zram/Makefile
index 29cb008..1f6fecd 100644
--- a/drivers/block/zram/Makefile
+++ b/drivers/block/zram/Makefile
@@ -1,3 +1,4 @@
-zram-y := zcomp.o zram_drv.o zram_dedup.o
+zram-y := zcomp.o zram_drv.o
-obj-$(CONFIG_ZRAM) += zram.o
+obj-$(CONFIG_ZRAM) += zram.o
+obj-$(CONFIG_ZRAM_DEDUP) += zram_dedup.o
diff --git a/drivers/block/zram/zram_dedup.c b/drivers/block/zram/zram_dedup.c
index d313fc8..1df1ce1 100644
--- a/drivers/block/zram/zram_dedup.c
+++ b/drivers/block/zram/zram_dedup.c
@@ -27,6 +27,19 @@ u64 zram_dedup_meta_size(struct zram *zram)
return (u64)atomic64_read(&zram->stats.meta_data_size);
}
+static inline bool zram_dedup_enabled(struct zram_meta *meta)
+{
+ return meta->hash;
+}
+
+unsigned long zram_dedup_handle(struct zram *zram, struct zram_entry *entry)
+{
+ if (!zram_dedup_enabled(zram->meta))
+ return (unsigned long)entry;
+
+ return entry->handle;
+}
+
static u32 zram_dedup_checksum(unsigned char *mem)
{
return jhash(mem, PAGE_SIZE, 0);
@@ -41,6 +54,9 @@ void zram_dedup_insert(struct zram *zram, struct zram_entry *new,
struct rb_node **rb_node, *parent = NULL;
struct zram_entry *entry;
+ if (!zram_dedup_enabled(zram->meta))
+ return;
+
new->checksum = checksum;
hash = &meta->hash[checksum % meta->hash_size];
rb_root = &hash->rb_root;
@@ -149,6 +165,9 @@ static struct zram_entry *zram_dedup_get(struct zram *zram,
struct zram_entry *zram_dedup_find(struct zram *zram, unsigned char *mem,
u32 *checksum)
{
+ if (!zram_dedup_enabled(zram->meta))
+ return NULL;
+
*checksum = zram_dedup_checksum(mem);
return zram_dedup_get(zram, mem, *checksum);
@@ -160,6 +179,9 @@ struct zram_entry *zram_dedup_alloc(struct zram *zram,
{
struct zram_entry *entry;
+ if (!zram_dedup_enabled(zram->meta))
+ return (struct zram_entry *)handle;
+
entry = kzalloc(sizeof(*entry),
flags & ~(__GFP_HIGHMEM|__GFP_MOVABLE));
if (!entry)
@@ -180,6 +202,9 @@ unsigned long zram_dedup_free(struct zram *zram, struct zram_meta *meta,
{
unsigned long handle;
+ if (!zram_dedup_enabled(meta))
+ return (unsigned long)entry;
+
if (zram_dedup_put(zram, meta, entry))
return 0;
@@ -193,11 +218,14 @@ unsigned long zram_dedup_free(struct zram *zram, struct zram_meta *meta,
return handle;
}
-int zram_dedup_init(struct zram_meta *meta, size_t num_pages)
+int zram_dedup_init(struct zram *zram, struct zram_meta *meta, size_t num_pages)
{
int i;
struct zram_hash *hash;
+ if (!zram->use_dedup)
+ return 0;
+
meta->hash_size = num_pages >> ZRAM_HASH_SHIFT;
meta->hash_size = min_t(size_t, ZRAM_HASH_SIZE_MAX, meta->hash_size);
meta->hash_size = max_t(size_t, ZRAM_HASH_SIZE_MIN, meta->hash_size);
@@ -219,4 +247,5 @@ int zram_dedup_init(struct zram_meta *meta, size_t num_pages)
void zram_dedup_fini(struct zram_meta *meta)
{
vfree(meta->hash);
+ meta->hash = NULL;
}
diff --git a/drivers/block/zram/zram_dedup.h b/drivers/block/zram/zram_dedup.h
index 7071f32..470009a 100644
--- a/drivers/block/zram/zram_dedup.h
+++ b/drivers/block/zram/zram_dedup.h
@@ -5,9 +5,11 @@ struct zram;
struct zram_meta;
struct zram_entry;
+#ifdef CONFIG_ZRAM_DEDUP
u64 zram_dedup_dup_size(struct zram *zram);
u64 zram_dedup_meta_size(struct zram *zram);
+unsigned long zram_dedup_handle(struct zram *zram, struct zram_entry *entry);
void zram_dedup_insert(struct zram *zram, struct zram_entry *new,
u32 checksum);
struct zram_entry *zram_dedup_find(struct zram *zram, unsigned char *mem,
@@ -19,7 +21,36 @@ struct zram_entry *zram_dedup_alloc(struct zram *zram,
unsigned long zram_dedup_free(struct zram *zram, struct zram_meta *meta,
struct zram_entry *entry);
-int zram_dedup_init(struct zram_meta *meta, size_t num_pages);
+int zram_dedup_init(struct zram *zram, struct zram_meta *meta,
+ size_t num_pages);
void zram_dedup_fini(struct zram_meta *meta);
+#else
+
+static inline u64 zram_dedup_dup_size(struct zram *zram) { return 0; }
+static inline u64 zram_dedup_meta_size(struct zram *zram) { return 0; }
+
+static inline unsigned long zram_dedup_handle(struct zram *zram,
+ struct zram_entry *entry) { return (unsigned long)entry; }
+static inline void zram_dedup_insert(struct zram *zram, struct zram_entry *new,
+ u32 checksum) { }
+static inline struct zram_entry *zram_dedup_find(struct zram *zram,
+ unsigned char *mem, u32 *checksum) { return NULL; }
+
+static inline struct zram_entry *zram_dedup_alloc(struct zram *zram,
+ unsigned long handle, unsigned int len, gfp_t flags)
+{
+ return (struct zram_entry *)handle;
+}
+static inline unsigned long zram_dedup_free(struct zram *zram,
+ struct zram_meta *meta, struct zram_entry *entry)
+{
+ return (unsigned long)entry;
+}
+
+static inline int zram_dedup_init(struct zram *zram, struct zram_meta *meta,
+ size_t num_pages) { return 0; }
+void zram_dedup_fini(struct zram_meta *meta) { }
+
+#endif
#endif /* _ZRAM_DEDUP_H_ */
diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index 15cecd6..016b2ee 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -327,6 +327,43 @@ static ssize_t comp_algorithm_store(struct device *dev,
return len;
}
+static ssize_t use_dedup_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ bool val;
+ struct zram *zram = dev_to_zram(dev);
+
+ down_read(&zram->init_lock);
+ val = zram->use_dedup;
+ up_read(&zram->init_lock);
+
+ return scnprintf(buf, PAGE_SIZE, "%d\n", (int)val);
+}
+
+static ssize_t use_dedup_store(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t len)
+{
+#ifdef CONFIG_ZRAM_DEDUP
+ int val;
+ struct zram *zram = dev_to_zram(dev);
+
+ if (kstrtoint(buf, 10, &val) || (val != 0 && val != 1))
+ return -EINVAL;
+
+ down_write(&zram->init_lock);
+ if (init_done(zram)) {
+ up_write(&zram->init_lock);
+ pr_info("Can't change dedup usage for initialized device\n");
+ return -EBUSY;
+ }
+ zram->use_dedup = val;
+ up_write(&zram->init_lock);
+ return len;
+#else
+ return -EINVAL;
+#endif
+}
+
static ssize_t compact_store(struct device *dev,
struct device_attribute *attr, const char *buf, size_t len)
{
@@ -421,6 +458,12 @@ static DEVICE_ATTR_RO(io_stat);
static DEVICE_ATTR_RO(mm_stat);
static DEVICE_ATTR_RO(debug_stat);
+static unsigned long zram_entry_handle(struct zram *zram,
+ struct zram_entry *entry)
+{
+ return zram_dedup_handle(zram, entry);
+}
+
static struct zram_entry *zram_entry_alloc(struct zram *zram,
unsigned int len, gfp_t flags)
{
@@ -475,9 +518,10 @@ static void zram_meta_free(struct zram_meta *meta, u64 disksize)
kfree(meta);
}
-static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
+static struct zram_meta *zram_meta_alloc(struct zram *zram, u64 disksize)
{
size_t num_pages;
+ char *pool_name = zram->disk->disk_name;
struct zram_meta *meta = kzalloc(sizeof(*meta), GFP_KERNEL);
if (!meta)
@@ -490,7 +534,7 @@ static struct zram_meta *zram_meta_alloc(char *pool_name, u64 disksize)
goto out_error;
}
- if (zram_dedup_init(meta, num_pages)) {
+ if (zram_dedup_init(zram, meta, num_pages)) {
pr_err("Error initializing zram entry hash\n");
goto out_error;
}
@@ -562,7 +606,8 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
return 0;
}
- cmem = zs_map_object(meta->mem_pool, entry->handle, ZS_MM_RO);
+ cmem = zs_map_object(meta->mem_pool,
+ zram_entry_handle(zram, entry), ZS_MM_RO);
if (size == PAGE_SIZE) {
copy_page(mem, cmem);
} else {
@@ -571,7 +616,7 @@ static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
ret = zcomp_decompress(zstrm, cmem, size, mem);
zcomp_stream_put(zram->comp);
}
- zs_unmap_object(meta->mem_pool, entry->handle);
+ zs_unmap_object(meta->mem_pool, zram_entry_handle(zram, entry));
bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value);
/* Should NEVER happen. Return bio error if it does. */
@@ -771,7 +816,8 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
goto out;
}
- cmem = zs_map_object(meta->mem_pool, entry->handle, ZS_MM_WO);
+ cmem = zs_map_object(meta->mem_pool,
+ zram_entry_handle(zram, entry), ZS_MM_WO);
if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) {
src = kmap_atomic(page);
@@ -783,7 +829,7 @@ static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
zcomp_stream_put(zram->comp);
zstrm = NULL;
- zs_unmap_object(meta->mem_pool, entry->handle);
+ zs_unmap_object(meta->mem_pool, zram_entry_handle(zram, entry));
zram_dedup_insert(zram, entry, checksum);
found_dup:
@@ -1055,7 +1101,7 @@ static ssize_t disksize_store(struct device *dev,
return -EINVAL;
disksize = PAGE_ALIGN(disksize);
- meta = zram_meta_alloc(zram->disk->disk_name, disksize);
+ meta = zram_meta_alloc(zram, disksize);
if (!meta)
return -ENOMEM;
@@ -1166,6 +1212,7 @@ static DEVICE_ATTR_WO(mem_limit);
static DEVICE_ATTR_WO(mem_used_max);
static DEVICE_ATTR_RW(max_comp_streams);
static DEVICE_ATTR_RW(comp_algorithm);
+static DEVICE_ATTR_RW(use_dedup);
static struct attribute *zram_disk_attrs[] = {
&dev_attr_disksize.attr,
@@ -1176,6 +1223,7 @@ static struct attribute *zram_disk_attrs[] = {
&dev_attr_mem_used_max.attr,
&dev_attr_max_comp_streams.attr,
&dev_attr_comp_algorithm.attr,
+ &dev_attr_use_dedup.attr,
&dev_attr_io_stat.attr,
&dev_attr_mm_stat.attr,
&dev_attr_debug_stat.attr,
diff --git a/drivers/block/zram/zram_drv.h b/drivers/block/zram/zram_drv.h
index 968e269..53be62a 100644
--- a/drivers/block/zram/zram_drv.h
+++ b/drivers/block/zram/zram_drv.h
@@ -138,6 +138,7 @@ struct zram {
* zram is claimed so open request will be failed
*/
bool claim; /* Protected by bdev->bd_mutex */
+ bool use_dedup;
};
void zram_entry_free(struct zram *zram, struct zram_meta *meta,
--
2.7.4