From: Eric Biggers <ebiggers@kernel.org>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
Yosry Ahmed <yosry.ahmed@linux.dev>,
Kanchana P Sridhar <kanchana.p.sridhar@intel.com>,
Sergey Senozhatsky <senozhatsky@chromium.org>
Subject: Re: [v3 PATCH 3/8] crypto: acomp - Move stream management into scomp layer
Date: Sat, 15 Mar 2025 21:36:31 -0700 [thread overview]
Message-ID: <20250316043631.GC117195@sol.localdomain> (raw)
In-Reply-To: <25f96a0e0e642e9d1c6014b12b00fd21b9f9c785.1741488107.git.herbert@gondor.apana.org.au>
On Sun, Mar 09, 2025 at 10:43:17AM +0800, Herbert Xu wrote:
> Rather than allocating the stream memoryin the request object,
> move it into a per-cpu buffer managed by scomp. This takes the
> stress off the user from having to manage large request objects
> and setting up their own per-cpu buffers in order to do so.
Well, except the workspace (which you seem to be calling a "stream" for some
reason) size depends heavily on the compression parameters, such as the maximum
input length and compression level. Zstd for example wants 1303288 (comp) +
95944 (decomp) with the parameters the crypto API is currently setting, but only
89848 + 95944 if it's properly configured with estimated_src_size=4096 which is
what most of the users actually want. So making this a per-algorithm property
is insufficiently flexible.
But of course there is also no guarantee that users want it to be per-"tfm"
either, let alone have a full set of per-CPU buffers. FWIW, this series makes
the kernel use an extra 40 MB of memory on my system if I enable
CONFIG_UBIFS_FS, which seems problematic.
I don't think the crypto API model works well for compression at all, TBH.
- Eric
next prev parent reply other threads:[~2025-03-16 4:36 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-09 2:43 [v3 PATCH 0/8] crypto: acomp - Add request chaining and virtual address support Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 1/8] crypto: api - Add cra_type->destroy hook Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 2/8] crypto: scomp - Remove tfm argument from alloc/free_ctx Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 3/8] crypto: acomp - Move stream management into scomp layer Herbert Xu
2025-03-16 4:36 ` Eric Biggers [this message]
2025-03-16 4:42 ` Herbert Xu
2025-03-16 4:46 ` Herbert Xu
2025-03-16 4:44 ` Herbert Xu
2025-03-17 8:36 ` Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 4/8] crypto: scomp - Disable BH when taking per-cpu spin lock Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 5/8] crypto: acomp - Add request chaining and virtual addresses Herbert Xu
2025-03-16 4:49 ` Eric Biggers
2025-03-16 5:43 ` Herbert Xu
2025-03-16 6:50 ` Eric Biggers
2025-03-20 17:24 ` Cabiddu, Giovanni
2025-03-21 2:33 ` Herbert Xu
2025-03-24 9:39 ` Cabiddu, Giovanni
2025-03-09 2:43 ` [v3 PATCH 6/8] crypto: testmgr - Remove NULL dst acomp tests Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 7/8] crypto: scomp - Remove support for most non-trivial destination SG lists Herbert Xu
2025-03-10 19:31 ` Dan Carpenter
2025-03-11 3:13 ` Herbert Xu
2025-03-11 3:16 ` Herbert Xu
2025-03-09 2:43 ` [v3 PATCH 8/8] crypto: scomp - Add chaining and virtual address support Herbert Xu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250316043631.GC117195@sol.localdomain \
--to=ebiggers@kernel.org \
--cc=herbert@gondor.apana.org.au \
--cc=kanchana.p.sridhar@intel.com \
--cc=linux-crypto@vger.kernel.org \
--cc=senozhatsky@chromium.org \
--cc=yosry.ahmed@linux.dev \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox