Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 5 Sep 2023 18:23:02 +0200
From:   David Sterba <dsterba@suse.cz>
To:     Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Cc:     Herbert Xu <herbert@gondor.apana.org.au>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        ardb@kernel.org, kees@kernel.org, linux-kernel@vger.kernel.org,
        enlin.mu@unisoc.com, ebiggers@google.com, gpiccoli@igalia.com,
        willy@infradead.org, yunlong.xing@unisoc.com,
        yuxiaozhang@google.com, qat-linux@intel.com,
        Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
        Yang Shen <shenyang39@huawei.com>,
        Zhou Wang <wangzhou1@hisilicon.com>,
        Weigang Li <weigang.li@intel.com>, Chris Mason <clm@meta.com>,
        Brian Will <brian.will@intel.com>, linux-btrfs@vger.kernel.org
Subject: Re: [PATCH 2/4] crypto: qat - Remove zlib-deflate
Message-ID: <20230905162302.GD14420@twin.jikos.cz>
Reply-To: dsterba@suse.cz
References: <ZO8ULhlJSrJ0Mcsx@gondor.apana.org.au>
 <E1qbI7x-009Bvo-IM@formenos.hmeau.com>
 <ZPcqALQ0Ck/3lF0U@gcabiddu-mobl1.ger.corp.intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZPcqALQ0Ck/3lF0U@gcabiddu-mobl1.ger.corp.intel.com>
User-Agent: Mutt/1.5.23.1-rc1 (2014-03-12)
Precedence: bulk

On Tue, Sep 05, 2023 at 02:15:44PM +0100, Giovanni Cabiddu wrote:
> Hi Herbert,
> 
> On Wed, Aug 30, 2023 at 06:08:47PM +0800, Herbert Xu wrote:
> > Remove the implementation of zlib-deflate because it is completely
> > unused in the kernel.
> We are working at a new revision of [1] which enables BTRFS to use acomp
> for offloading zlib-deflate. We see that there is value in using QAT for
> such use case in terms of throughput / CPU utilization / compression ratio
> compared to software.
> Zlib-deflate is preferred to deflate since BTRFS already uses that
> format.
> 
> We expect to send this patch for 6.7.
> Can we keep zlib-deflate in the kernel?
> 
> Thanks,
> 
> [1] https://patchwork.kernel.org/project/linux-btrfs/patch/1467083180-111750-1-git-send-email-weigang.li@intel.com/

The patch is from 2016 and zlib though still supported has been
superseded by zstd that is from 2017. It would be good to see numbers
comparing zlib (cpu), zlib (qat) against relevant zstd levels. The
offloading might be an improvement and worth adding the support
otherwise I don't see much reason to add it unless there are users.

I can see there's QAT support for zstd too,
https://github.com/intel/QAT-ZSTD-Plugin, can't find one for lzo but in
case ther's QAT for all 3 algorithms used by btrfs I wouldn't mind
keeping the QAT support for zlib for parity.