[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aa6c291a-d9f0-1e2a-63df-e53dfcefba53@huawei.com>
Date: Thu, 31 Aug 2023 17:39:08 +0800
From: Yang Shen <shenyang39@...wei.com>
To: Herbert Xu <herbert@...dor.apana.org.au>
CC: Linus Torvalds <torvalds@...ux-foundation.org>, <ardb@...nel.org>,
<kees@...nel.org>, <linux-kernel@...r.kernel.org>,
<enlin.mu@...soc.com>, <ebiggers@...gle.com>,
<gpiccoli@...lia.com>, <willy@...radead.org>,
<yunlong.xing@...soc.com>, <yuxiaozhang@...gle.com>,
Giovanni Cabiddu <giovanni.cabiddu@...el.com>,
<qat-linux@...el.com>,
Linux Crypto Mailing List <linux-crypto@...r.kernel.org>,
Zhou Wang <wangzhou1@...ilicon.com>
Subject: Re: [PATCH 1/4] crypto: hisilicon/zip - Remove driver
在 2023/8/31 13:42, Herbert Xu 写道:
> On Thu, Aug 31, 2023 at 10:21:52AM +0800, Yang Shen wrote:
>> It's a pity to see that there is no user in the kernel of zlib-deflate.
>> However, there may still be hidden
>> users in the current kernel who may be using the zlib-deflate algorithm.
>> Such as zswap, it can use
>> user-specified algorithm. So there are still some benefits to be gained from
>> zlib hardware.
> Perhaps you should try reconstructing the zlib header in your
> driver so that it becomes capable of handling "deflate" data as
> is rather than adding the non-standard "zlib-deflate" algorithm?
>
> There is no way of getting the checksum without decompressing
> the data first but perhaps your hardware could ignore checksum
> errors?
OK, I'm sorry I got missunderstand about the purpose of the patch. The
hardware also
support the deflate algorithm. I can change the driver to register the
deflate interface and
remove the zlib-deflate. Is this plan acceptable? If yes, please don't
remove the hisilicon
zip driver and I'll send the fix patch as soon as possible. It may take
me a week to adapt
and verify the driver.
Thanks,
Yang
>
>> What's more, hisilicon zip driver also does other work besides supporting
>> the zlib-deflate:
>> 1.Support gzip algorithm.
> We don't even have a generic "gzip" implementation so this should
> never have gone into the kernel.
>
>> 2.Support a user space cdev hisi-zip which can accelerate user space process
>> via uacce subsystem.
> Feel free to resubmit this as a new driver but it doesn't belong
> in drivers/crypto.
>
> Cheers,
Powered by blists - more mailing lists