Staging: ramzswap: Use lock for 64-bit stats
64-bit stats corruption was observed when ramzswap was
used on SMP systems. To prevent this, use separate spinlock
to protect these stats.
Also, replace stat_*() with rzs_stat*() to avoid possible
conflict with core kernel code.
Eventually, these will be converted to per-cpu counters
if this driver finds use on large scale systems and this
locking is found to affect scalability.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
diff --git a/drivers/staging/ramzswap/ramzswap_drv.c b/drivers/staging/ramzswap/ramzswap_drv.c
index baf7572..05273c0 100644
--- a/drivers/staging/ramzswap/ramzswap_drv.c
+++ b/drivers/staging/ramzswap/ramzswap_drv.c
@@ -24,7 +24,6 @@
#include <linux/genhd.h>
#include <linux/highmem.h>
#include <linux/lzo.h>
-#include <linux/mutex.h>
#include <linux/string.h>
#include <linux/swap.h>
#include <linux/swapops.h>
@@ -239,7 +238,8 @@
mem_used = xv_get_total_size_bytes(rzs->mem_pool)
+ (rs->pages_expand << PAGE_SHIFT);
- succ_writes = rs->num_writes - rs->failed_writes;
+ succ_writes = rzs_stat64_read(rzs, &rs->num_writes) -
+ rzs_stat64_read(rzs, &rs->failed_writes);
if (succ_writes && rs->pages_stored) {
good_compress_perc = rs->good_compress * 100
@@ -248,11 +248,12 @@
/ rs->pages_stored;
}
- s->num_reads = rs->num_reads;
- s->num_writes = rs->num_writes;
- s->failed_reads = rs->failed_reads;
- s->failed_writes = rs->failed_writes;
- s->invalid_io = rs->invalid_io;
+ s->num_reads = rzs_stat64_read(rzs, &rs->num_reads);
+ s->num_writes = rzs_stat64_read(rzs, &rs->num_writes);
+ s->failed_reads = rzs_stat64_read(rzs, &rs->failed_reads);
+ s->failed_writes = rzs_stat64_read(rzs, &rs->failed_writes);
+ s->invalid_io = rzs_stat64_read(rzs, &rs->invalid_io);
+ s->notify_free = rzs_stat64_read(rzs, &rs->notify_free);
s->pages_zero = rs->pages_zero;
s->good_compress_pct = good_compress_perc;
@@ -264,8 +265,8 @@
s->compr_data_size = rs->compr_size;
s->mem_used_total = mem_used;
- s->bdev_num_reads = rs->bdev_num_reads;
- s->bdev_num_writes = rs->bdev_num_writes;
+ s->bdev_num_reads = rzs_stat64_read(rzs, &rs->bdev_num_reads);
+ s->bdev_num_writes = rzs_stat64_read(rzs, &rs->bdev_num_writes);
}
#endif /* CONFIG_RAMZSWAP_STATS */
}
@@ -594,7 +595,7 @@
if (unlikely(!page)) {
if (rzs_test_flag(rzs, index, RZS_ZERO)) {
rzs_clear_flag(rzs, index, RZS_ZERO);
- stat_dec(rzs->stats.pages_zero);
+ rzs_stat_dec(&rzs->stats.pages_zero);
}
return;
}
@@ -603,7 +604,7 @@
clen = PAGE_SIZE;
__free_page(page);
rzs_clear_flag(rzs, index, RZS_UNCOMPRESSED);
- stat_dec(rzs->stats.pages_expand);
+ rzs_stat_dec(&rzs->stats.pages_expand);
goto out;
}
@@ -613,11 +614,11 @@
xv_free(rzs->mem_pool, page, offset);
if (clen <= PAGE_SIZE / 2)
- stat_dec(rzs->stats.good_compress);
+ rzs_stat_dec(&rzs->stats.good_compress);
out:
rzs->stats.compr_size -= clen;
- stat_dec(rzs->stats.pages_stored);
+ rzs_stat_dec(&rzs->stats.pages_stored);
rzs->table[index].page = NULL;
rzs->table[index].offset = 0;
@@ -679,8 +680,8 @@
*/
if (rzs->backing_swap) {
u32 pagenum;
- stat_dec(rzs->stats.num_reads);
- stat_inc(rzs->stats.bdev_num_reads);
+ rzs_stat64_dec(rzs, &rzs->stats.num_reads);
+ rzs_stat64_inc(rzs, &rzs->stats.bdev_num_reads);
bio->bi_bdev = rzs->backing_swap;
/*
@@ -718,7 +719,7 @@
struct zobj_header *zheader;
unsigned char *user_mem, *cmem;
- stat_inc(rzs->stats.num_reads);
+ rzs_stat64_inc(rzs, &rzs->stats.num_reads);
page = bio->bi_io_vec[0].bv_page;
index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;
@@ -752,7 +753,7 @@
if (unlikely(ret != LZO_E_OK)) {
pr_err("Decompression failed! err=%d, page=%u\n",
ret, index);
- stat_inc(rzs->stats.failed_reads);
+ rzs_stat64_inc(rzs, &rzs->stats.failed_reads);
goto out;
}
@@ -776,7 +777,7 @@
struct page *page, *page_store;
unsigned char *user_mem, *cmem, *src;
- stat_inc(rzs->stats.num_writes);
+ rzs_stat64_inc(rzs, &rzs->stats.num_writes);
page = bio->bi_io_vec[0].bv_page;
index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT;
@@ -796,7 +797,7 @@
* Simply clear zero page flag.
*/
if (rzs_test_flag(rzs, index, RZS_ZERO)) {
- stat_dec(rzs->stats.pages_zero);
+ rzs_stat_dec(&rzs->stats.pages_zero);
rzs_clear_flag(rzs, index, RZS_ZERO);
}
@@ -806,7 +807,7 @@
if (page_zero_filled(user_mem)) {
kunmap_atomic(user_mem, KM_USER0);
mutex_unlock(&rzs->lock);
- stat_inc(rzs->stats.pages_zero);
+ rzs_stat_inc(&rzs->stats.pages_zero);
rzs_set_flag(rzs, index, RZS_ZERO);
set_bit(BIO_UPTODATE, &bio->bi_flags);
@@ -830,7 +831,7 @@
if (unlikely(ret != LZO_E_OK)) {
mutex_unlock(&rzs->lock);
pr_err("Compression failed! err=%d\n", ret);
- stat_inc(rzs->stats.failed_writes);
+ rzs_stat64_inc(rzs, &rzs->stats.failed_writes);
goto out;
}
@@ -853,13 +854,13 @@
mutex_unlock(&rzs->lock);
pr_info("Error allocating memory for incompressible "
"page: %u\n", index);
- stat_inc(rzs->stats.failed_writes);
+ rzs_stat64_inc(rzs, &rzs->stats.failed_writes);
goto out;
}
offset = 0;
rzs_set_flag(rzs, index, RZS_UNCOMPRESSED);
- stat_inc(rzs->stats.pages_expand);
+ rzs_stat_inc(&rzs->stats.pages_expand);
rzs->table[index].page = page_store;
src = kmap_atomic(page, KM_USER0);
goto memstore;
@@ -871,7 +872,7 @@
mutex_unlock(&rzs->lock);
pr_info("Error allocating memory for compressed "
"page: %u, size=%zu\n", index, clen);
- stat_inc(rzs->stats.failed_writes);
+ rzs_stat64_inc(rzs, &rzs->stats.failed_writes);
if (rzs->backing_swap)
fwd_write_request = 1;
goto out;
@@ -900,9 +901,9 @@
/* Update stats */
rzs->stats.compr_size += clen;
- stat_inc(rzs->stats.pages_stored);
+ rzs_stat_inc(&rzs->stats.pages_stored);
if (clen <= PAGE_SIZE / 2)
- stat_inc(rzs->stats.good_compress);
+ rzs_stat_inc(&rzs->stats.good_compress);
mutex_unlock(&rzs->lock);
@@ -912,7 +913,7 @@
out:
if (fwd_write_request) {
- stat_inc(rzs->stats.bdev_num_writes);
+ rzs_stat64_inc(rzs, &rzs->stats.bdev_num_writes);
bio->bi_bdev = rzs->backing_swap;
#if 0
/*
@@ -974,7 +975,7 @@
}
if (!valid_swap_request(rzs, bio)) {
- stat_inc(rzs->stats.invalid_io);
+ rzs_stat64_inc(rzs, &rzs->stats.invalid_io);
bio_io_error(bio);
return 0;
}
@@ -1295,6 +1296,7 @@
static int create_device(struct ramzswap *rzs, int device_id)
{
mutex_init(&rzs->lock);
+ spin_lock_init(&rzs->stat64_lock);
INIT_LIST_HEAD(&rzs->backing_swap_extent_list);
rzs->queue = blk_alloc_queue(GFP_KERNEL);