From nobody Sun Oct 5 12:46:13 2025 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id C1FFC14D283; Tue, 5 Aug 2025 00:30:03 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1754353803; cv=none; b=u288CCtIzA6gdvkN9mINWaeDx8hAoOxUw/eL6uJthqyMmyYwBkfQNgT7UYCuvJPKL1g9xlypdqo97hUsFfFQaiW1KDCg8ZLp075fabbp9CTgomYKA329EQy56PIaQ4DK3behNntcXaiVXdwbVUYcAUGrq49c0hZ20XymC12c2mc= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1754353803; c=relaxed/simple; bh=dWLySnUxFT27urg9k6MCl5CiKahC8C/PdSilLgOqiRY=; h=From:To:Cc:Subject:Date:Message-Id:MIME-Version; b=WHMaBsJowF7UOijKM542tDRxDmCqV3NXnf5kUEGeYs3/pi5VRft/YDHBdxvAAAGRpXKd7f3QSG5O/Taqi+rNaotRRszjxskOB+qnASnDueDFcccOtcwPKlweihDIogBY90wwg30E3dp0FlkZ8+6EW8jM/SAfXkqHX/vMn1cudGI= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b=P9zWQFVM; arc=none smtp.client-ip=10.30.226.201 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b="P9zWQFVM" Received: by smtp.kernel.org (Postfix) with ESMTPSA id 0D17FC4CEF0; Tue, 5 Aug 2025 00:30:03 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1754353803; bh=dWLySnUxFT27urg9k6MCl5CiKahC8C/PdSilLgOqiRY=; h=From:To:Cc:Subject:Date:From; b=P9zWQFVMxDjgw4BvsamfiikNRwB7hp987Sf48hrUSAfK88m/m7S9lPUg/PIeIu5s1 mOyCUwI1VNV3XBOBtkc1lnDvY51ISTF8zMCe+V9e6x0D+nbG9N5Y2Rk8l14Z2lTb3r ke/jdezmGPlbxkzpWcybxblonTbTfY9Ae1yapa6OqQh3J4n0Xaa/30CTaFwYg51o7N X6mBCA/TQ0lwDC6KNIZX9ZExMrOfe26jrCb2sJh+B7GhNG1KHqLlPNw9YVNl/qLUA2 oVN+3sbsOcn08ctos4I0dTvz1HRfrjcdbIIMzxI2LF6LBBO+FkAedP4D5egFFcXb3b RLb036wG4whTw== From: SeongJae Park To: Cc: SeongJae Park , "Liam R. Howlett" , Andrew Morton , Chengming Zhou , David Hildenbrand , Johannes Weiner , Jonathan Corbet , Lorenzo Stoakes , Michal Hocko , Mike Rapoport , Nhat Pham , Suren Baghdasaryan , Vlastimil Babka , Yosry Ahmed , kernel-team@meta.com, linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org, linux-mm@kvack.org, Takero Funaki Subject: [RFC PATCH v2] mm/zswap: store X-Mailer: git-send-email 2.39.5 Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" When zswap writeback is enabled and it fails compressing a given page, the page is swapped out to the backing swap device. This behavior breaks the zswap's writeback LRU order, and hence users can experience unexpected latency spikes. If the page is compressed without failure, but results in a size of PAGE_SIZE, the LRU order is kept, but the decompression overhead for loading the page back on the later access is unnecessary. Keep the LRU order and optimize unnecessary decompression overheads in the cases, by storing the original content in zpool as-is. The length field of zswap_entry will be set appropriately, as PAGE_SIZE, Hence whether it is saved as-is or not (whether decompression is unnecessary) is identified by 'zswap_entry->length =3D=3D PAGE_SIZE'. So this change is not increasing per zswap entry metadata overhead. But as the number of incompressible pages increases, total zswap metadata overhead is proportionally increased. The overhead should not be problematic in usual cases, since the zswap metadata for single zswap entry is much smaller than PAGE_SIZE, and in common zswap use cases there should be a sufficient amount of compressible pages. Also it can be mitigated by the zswap writeback. When a severe memory pressure comes from memcg's memory.high, storing incompressible pages as-is may result in reducing accounted memory footprint slower, since the footprint will be reduced only after the zswap writeback kicks in. This can incur higher penalty_jiffies and degrade the performance. Arguably this is just a wrong setup, but we don't want to introduce unnecessary surprises. Add a parameter, namely 'save_incompressible_pages', to turn the feature on/off as users want. It is turned off by default. When the writeback is disabled, the additional overhead could be problematic. For the case, keep the current behavior that just returns the failure and let swap_writeout() put the page back to the active LRU list in the case. It is known to be suboptimal when the incompressible pages are cold, since the incompressible pages will continuously be tried to be zswapped out, and burn CPU cycles for compression attempts that will anyway fails. One imaginable solution for the problem is reusing the swapped-out page and its struct page to store in the zswap pool. But that's out of the scope of this patch. Tests Suggested-by tags for them. Actually Nhat first suggested this to me Suggested-by: Nhat Pham Suggested-by: Takero Funaki ----- I tested this patch using a simple self-written microbenchmark that is available at GitHub[1]. You can reproduce the test I did by executing run_tests.sh of the repo on your system. Note that the repo's documentation is not good as of this writing, so you may need to read and use the code. The basic test scenario is simple. Run a test program making artificial accesses to memory having artificial content under memory.high-set memory limit and measure how many accesses were made in given time. The test program repeatedly and randomly access three anonymous memory regions. The regions are all 500 MiB size, and accessed in the same probability. Two of those are filled up with a simple content that can easily be compressed, while the remaining one is filled up with a content that read from /dev/urandom, which is easy to fail at compressing to Suggested-by: Takero Funaki Signed-off-by: SeongJae Park --- Changes from RFC v1 (https://lore.kernel.org/20250730234059.4603-1-sj@kernel.org) - Consider PAGE_SIZE-resulting compression successes as failures. - Use zpool for storing incompressible pages. - Test with zswap shrinker enabled. - Wordsmith changelog and comments. - Add documentation of save_incompressible_pages parameter. Documentation/admin-guide/mm/zswap.rst | 9 +++++ mm/zswap.c | 53 +++++++++++++++++++++++++- 2 files changed, 61 insertions(+), 1 deletion(-) diff --git a/Documentation/admin-guide/mm/zswap.rst b/Documentation/admin-g= uide/mm/zswap.rst index c2806d051b92..20eae0734491 100644 --- a/Documentation/admin-guide/mm/zswap.rst +++ b/Documentation/admin-guide/mm/zswap.rst @@ -142,6 +142,15 @@ User can enable it as follows:: This can be enabled at the boot time if ``CONFIG_ZSWAP_SHRINKER_DEFAULT_ON= `` is selected. =20 +If a page cannot be compressed into a size smaller than PAGE_SIZE, it can = be +beneficial to save the content as is without compression, to keep the LRU +order. Users can enable this behavior, as follows:: + + echo Y > /sys/module/zswap/parameters/save_incompressible_pages + +This is disabled by default, and doesn't change behavior of zswap writeback +disabled case. + A debugfs interface is provided for various statistic about pool size, num= ber of pages stored, same-value filled pages and various counters for the reas= ons pages are rejected. diff --git a/mm/zswap.c b/mm/zswap.c index 7e02c760955f..6e196c9a4dba 100644 --- a/mm/zswap.c +++ b/mm/zswap.c @@ -129,6 +129,11 @@ static bool zswap_shrinker_enabled =3D IS_ENABLED( CONFIG_ZSWAP_SHRINKER_DEFAULT_ON); module_param_named(shrinker_enabled, zswap_shrinker_enabled, bool, 0644); =20 +/* Enable/disable incompressible pages storing */ +static bool zswap_save_incompressible_pages; +module_param_named(save_incompressible_pages, zswap_save_incompressible_pa= ges, + bool, 0644); + bool zswap_is_enabled(void) { return zswap_enabled; @@ -937,6 +942,29 @@ static void acomp_ctx_put_unlock(struct crypto_acomp_c= tx *acomp_ctx) mutex_unlock(&acomp_ctx->mutex); } =20 +/* + * Determine whether to save given page as-is. + * + * If a page cannot be compressed into a size smaller than PAGE_SIZE, it c= an be + * beneficial to saving the content as is without compression, to keep the= LRU + * order. This can increase memory overhead from metadata, but in common = zswap + * use cases where there are sufficient amount of compressible pages, the + * overhead should be not critical, and can be mitigated by the writeback. + * Also, the decompression overhead is optimized. + * + * When the writeback is disabled, however, the additional overhead could = be + * problematic. For the case, just return the failure. swap_writeout() w= ill + * put the page back to the active LRU list in the case. + */ +static bool zswap_save_as_is(int comp_ret, unsigned int dlen, + struct page *page) +{ + return zswap_save_incompressible_pages && + (comp_ret || dlen =3D=3D PAGE_SIZE) && + mem_cgroup_zswap_writeback_enabled( + folio_memcg(page_folio(page))); +} + static bool zswap_compress(struct page *page, struct zswap_entry *entry, struct zswap_pool *pool) { @@ -976,8 +1004,13 @@ static bool zswap_compress(struct page *page, struct = zswap_entry *entry, */ comp_ret =3D crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acom= p_ctx->wait); dlen =3D acomp_ctx->req->dlen; - if (comp_ret) + if (zswap_save_as_is(comp_ret, dlen, page)) { + comp_ret =3D 0; + dlen =3D PAGE_SIZE; + memcpy_from_page(dst, page, 0, dlen); + } else if (comp_ret) { goto unlock; + } =20 zpool =3D pool->zpool; gfp =3D GFP_NOWAIT | __GFP_NORETRY | __GFP_HIGHMEM | __GFP_MOVABLE; @@ -1001,6 +1034,17 @@ static bool zswap_compress(struct page *page, struct= zswap_entry *entry, return comp_ret =3D=3D 0 && alloc_ret =3D=3D 0; } =20 +/* + * If save_incompressible_pages is set and writeback is enabled, incompres= sible + * pages are saved as is without compression. For more details, refer to = the + * comments of zswap_save_as_is(). + */ +static bool zswap_saved_as_is(struct zswap_entry *entry, struct folio *fol= io) +{ + return entry->length =3D=3D PAGE_SIZE && zswap_save_incompressible_pages = && + mem_cgroup_zswap_writeback_enabled(folio_memcg(folio)); +} + static bool zswap_decompress(struct zswap_entry *entry, struct folio *foli= o) { struct zpool *zpool =3D entry->pool->zpool; @@ -1012,6 +1056,13 @@ static bool zswap_decompress(struct zswap_entry *ent= ry, struct folio *folio) acomp_ctx =3D acomp_ctx_get_cpu_lock(entry->pool); obj =3D zpool_obj_read_begin(zpool, entry->handle, acomp_ctx->buffer); =20 + if (zswap_saved_as_is(entry, folio)) { + memcpy_to_folio(folio, 0, obj, entry->length); + zpool_obj_read_end(zpool, entry->handle, obj); + acomp_ctx_put_unlock(acomp_ctx); + return true; + } + /* * zpool_obj_read_begin() might return a kmap address of highmem when * acomp_ctx->buffer is not used. However, sg_init_one() does not base-commit: d19f69751d55ef3883569c119d4b2ea3d6a0e39f --=20 2.39.5