linux-kernel - Re: [PATCH] crypto: scompress - call the completion callback for the completed acomp

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Sun, 21 Jun 2020 12:07:58 +1000
From:   Herbert Xu <herbert@...dor.apana.org.au>
To:     Barry Song <song.bao.hua@...ilicon.com>
Cc:     davem@...emloft.net, linux-crypto@...r.kernel.org,
        linux-kernel@...r.kernel.org, linuxarm@...wei.com
Subject: Re: [PATCH] crypto: scompress - call the completion callback for the
 completed acomp_req

On Sun, Jun 21, 2020 at 11:07:56AM +1200, Barry Song wrote:
> While using acomp_req, users may customize a completion callback, if the
> backend of acomp is actually scomp, this callback will never be called.
> This looks a bit messy to users. For one user, all scomp details should
> be hidden by acomp APIs.
> 
> But it won't cause any real problem if users simply use crypto_req_done()
> as the callback, and further use crypto_wait_req() to wait for request to
> be done since crypto_wait_req() depends on the callback only if queuing
> acomp_req returns -EINPROGRESS or -EBUSY.
> 
> This patch will make sure the customized completion callback is called
> for the first case. And it won't break the second case in which users
> use the common crypto_req_done() and crypto_wait_req() only.
> 
> Signed-off-by: Barry Song <song.bao.hua@...ilicon.com>
> ---
>  crypto/scompress.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/crypto/scompress.c b/crypto/scompress.c
> index 738f4f8f0f41..634b7ced9348 100644
> --- a/crypto/scompress.c
> +++ b/crypto/scompress.c
> @@ -158,6 +158,10 @@ static int scomp_acomp_comp_decomp(struct acomp_req *req, int dir)
>  	}
>  out:
>  	spin_unlock(&scratch->lock);
> +
> +	if (req->base.complete)
> +		acomp_request_complete(req, ret);
> +
>  	return ret;
>  }

Nack.  This is a deliberate optimisation to allow the sync code
path to execute with no overhead.

Cheers,
-- 
Email: Herbert Xu <herbert@...dor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt