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;
Message-ID: <f7c52ed1-8061-8147-f676-86190118cc56@gmail.com>
Date:   Tue, 4 Oct 2022 15:01:41 +0900
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.11.0
Subject: Re: [PATCH] crypto: x86: Do not acquire fpu context for too long
Content-Language: en-US
To:     Herbert Xu <herbert@gondor.apana.org.au>
Cc:     linux-crypto@vger.kernel.org, davem@davemloft.net,
        tglx@linutronix.de, mingo@redhat.com, bp@alien8.de,
        dave.hansen@linux.intel.com, x86@kernel.org, hpa@zytor.com,
        ardb@kernel.org, ebiggers@google.com
References: <20221004044912.24770-1-ap420073@gmail.com>
 <Yzu8Kd2botr3eegj@gondor.apana.org.au>
From:   Taehee Yoo <ap420073@gmail.com>
In-Reply-To: <Yzu8Kd2botr3eegj@gondor.apana.org.au>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi Herbert,
Thanks a lot for your review!

On 10/4/22 13:52, Herbert Xu wrote:
 > On Tue, Oct 04, 2022 at 04:49:12AM +0000, Taehee Yoo wrote:
 >>
 >>   #define ECB_WALK_START(req, bsize, fpu_blocks) do {			\
 >>   	void *ctx = crypto_skcipher_ctx(crypto_skcipher_reqtfm(req));	\
 >> +	unsigned int walked_bytes = 0;					\
 >>   	const int __bsize = (bsize);					\
 >>   	struct skcipher_walk walk;					\
 >> -	int err = skcipher_walk_virt(&walk, (req), false);		\
 >> +	int err;							\
 >> +									\
 >> +	err = skcipher_walk_virt(&walk, (req), false);			\
 >>   	while (walk.nbytes > 0) {					\
 >> -		unsigned int nbytes = walk.nbytes;			\
 >> -		bool do_fpu = (fpu_blocks) != -1 &&			\
 >> -			      nbytes >= (fpu_blocks) * __bsize;		\
 >>   		const u8 *src = walk.src.virt.addr;			\
 >> -		u8 *dst = walk.dst.virt.addr;				\
 >>   		u8 __maybe_unused buf[(bsize)];				\
 >> -		if (do_fpu) kernel_fpu_begin()
 >> +		u8 *dst = walk.dst.virt.addr;				\
 >> +		unsigned int nbytes;					\
 >> +		bool do_fpu;						\
 >> +									\
 >> +		if (walk.nbytes - walked_bytes > ECB_CBC_WALK_MAX) {	\
 >> +			nbytes = ECB_CBC_WALK_MAX;			\
 >> +			walked_bytes += ECB_CBC_WALK_MAX;		\
 >> +		} else {						\
 >> +			nbytes = walk.nbytes - walked_bytes;		\
 >> +			walked_bytes = walk.nbytes;			\
 >> +		}							\
 >> +									\
 >> +		do_fpu = (fpu_blocks) != -1 &&				\
 >> +			 nbytes >= (fpu_blocks) * __bsize;		\
 >> +		if (do_fpu)						\
 >> +			kernel_fpu_begin()
 >>
 >>   #define CBC_WALK_START(req, bsize, fpu_blocks)				\
 >>   	ECB_WALK_START(req, bsize, fpu_blocks)
 >> @@ -65,8 +81,12 @@
 >>   } while (0)
 >>
 >>   #define ECB_WALK_END()							\
 >> -		if (do_fpu) kernel_fpu_end();				\
 >> +		if (do_fpu)						\
 >> +			kernel_fpu_end();				\
 >> +		if (walked_bytes < walk.nbytes)				\
 >> +			continue;					\
 >>   		err = skcipher_walk_done(&walk, nbytes);		\
 >> +		walked_bytes = 0;					\
 >>   	}								\
 >>   	return err;							\
 >>   } while (0)
 >
 > skcipher_walk_* is supposed to return at most a page.  Why is this
 > necessary?
 >
 > Cheers,

I referred to below link.
https://lore.kernel.org/all/MW5PR84MB18426EBBA3303770A8BC0BDFAB759@MW5PR84MB1842.NAMPRD84.PROD.OUTLOOK.COM/

Sorry for that I didn't check that skcipher_walk_* returns only under 4K 
sizes.
So, I thought fpu context would be too long.
But, I just checked the skcipher_walk_*, and it's right, it returns 
under 4K sizes.
So, there are no problems as you mentioned.

Thank you so much!
Taehee Yoo