2023-10-24 23:45:24

by Nhat Pham

Subject: [PATCH] zswap: export compression failure stats

During a zswap store attempt, the compression algorithm could fail (for
e.g due to the page containing incompressible random data). This is not
tracked in any of existing zswap counters, making it hard to monitor for
and investigate. We have run into this problem several times in our
internal investigations on zswap store failures.

This patch adds a dedicated debugfs counter for compression algorithm
failures.

Signed-off-by: Nhat Pham <[email protected]>
---
mm/zswap.c | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/mm/zswap.c b/mm/zswap.c
index 37d2b1cb2ecb..e12e76289226 100644
--- a/mm/zswap.c
+++ b/mm/zswap.c
@@ -61,6 +61,8 @@ static u64 zswap_pool_limit_hit;
static u64 zswap_written_back_pages;
/* Store failed due to a reclaim failure after pool limit was reached */
static u64 zswap_reject_reclaim_fail;
+/* Store failed due to compression algorithm failure */
+static u64 zswap_reject_compress_fail;
/* Compressed page was too big for the allocator to (optimally) store */
static u64 zswap_reject_compress_poor;
/* Store failed because underlying allocator could not get memory */
@@ -1309,8 +1311,10 @@ bool zswap_store(struct folio *folio)
ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
dlen = acomp_ctx->req->dlen;

- if (ret)
+ if (ret) {
+ zswap_reject_compress_fail++;
goto put_dstmem;
+ }

/* store */
zpool = zswap_find_zpool(entry);
@@ -1550,6 +1554,8 @@ static int zswap_debugfs_init(void)
zswap_debugfs_root, &zswap_reject_alloc_fail);
debugfs_create_u64("reject_kmemcache_fail", 0444,
zswap_debugfs_root, &zswap_reject_kmemcache_fail);
+ debugfs_create_u64("reject_compress_fail", 0444,
+ zswap_debugfs_root, &zswap_reject_compress_fail);
debugfs_create_u64("reject_compress_poor", 0444,
zswap_debugfs_root, &zswap_reject_compress_poor);
debugfs_create_u64("written_back_pages", 0444,
--
2.34.1

2023-10-26 07:06:59

by Sergey Senozhatsky

Subject: Re: [PATCH] zswap: export compression failure stats

On (23/10/24 16:45), Nhat Pham wrote:
>
> During a zswap store attempt, the compression algorithm could fail (for
> e.g due to the page containing incompressible random data). This is not
> tracked in any of existing zswap counters, making it hard to monitor for
> and investigate. We have run into this problem several times in our
> internal investigations on zswap store failures.
>
> This patch adds a dedicated debugfs counter for compression algorithm
> failures.
>
> Signed-off-by: Nhat Pham <[email protected]>

FWIW,
Reviewed-by: Sergey Senozhatsky <[email protected]>