Received-SPF: pass (google.com: best guess record for domain of linux-crypto-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20191017122549.4634-1-t-kristo@ti.com> <20191017122549.4634-10-t-kristo@ti.com>
In-Reply-To: <20191017122549.4634-10-t-kristo@ti.com>
From:   Gilad Ben-Yossef <gilad@benyossef.com>
Date:   Wed, 6 Nov 2019 08:39:16 +0200
Message-ID: <CAOtvUMeBXjDBhSVgMOW=hshEx_AkNPg-Zk2c2jCDzY8vyXWW5g@mail.gmail.com>
Subject: Re: [PATCH 09/10] crypto: add timeout to crypto_wait_req
To:     Tero Kristo <t-kristo@ti.com>
Cc:     Herbert Xu <herbert@gondor.apana.org.au>,
        Eric Biggers <ebiggers@google.com>,
        David Miller <davem@davemloft.net>,
        Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
        Ard Biesheuvel <ard.biesheuvel@linaro.org>,
        linux-omap@vger.kernel.org,
        Linux ARM <linux-arm-kernel@lists.infradead.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-crypto-owner@vger.kernel.org
Precedence: bulk

Hi,


On Thu, Oct 17, 2019 at 3:26 PM Tero Kristo <t-kristo@ti.com> wrote:
>
> Currently crypto_wait_req waits indefinitely for an async crypto request
> to complete. This is bad as it can cause for example the crypto test
> manager to hang without any notification as to why it has happened.
> Instead of waiting indefinitely, add a 1 second timeout to the call,
> and provide a warning print if a timeout happens.

While the incentive is clear and positive, this suggested solution
creates problems of its own.
In many (most?) cases where we are waiting here, we are waiting for a
DMA operation to finish from hardware.
Exiting while this pending DMA operation is not finished, even with a
proper error return value, is dangerous because
unless the calling code takes great care to not release the memory the
DMA is being done from/to, this can have disastrous effects.

As Eric has already mentioned, one second might seem like a long time,
but we don't really know if it is enough.

How about adding a second API (ig. crypto_wait_req_timeout) which
supports a calee specified timeout where
the calle knows how to correctly deal with timeout and port the
relevant call sites to use this?

Thanks!
Gilad


>
> Signed-off-by: Tero Kristo <t-kristo@ti.com>
> ---
>  include/linux/crypto.h | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/include/linux/crypto.h b/include/linux/crypto.h
> index 19ea3a371d7b..b8f0e5c3cc0c 100644
> --- a/include/linux/crypto.h
> +++ b/include/linux/crypto.h
> @@ -682,8 +682,15 @@ static inline int crypto_wait_req(int err, struct cr=
ypto_wait *wait)
>         switch (err) {
>         case -EINPROGRESS:
>         case -EBUSY:
> -               wait_for_completion(&wait->completion);
> +               err =3D wait_for_completion_timeout(&wait->completion,
> +                                                 msecs_to_jiffies(1000))=
;
>                 reinit_completion(&wait->completion);
> +               if (!err) {
> +                       pr_err("%s: timeout for %p\n", __func__, wait);
> +                       err =3D -ETIMEDOUT;
> +                       break;
> +               }
> +
>                 err =3D wait->err;
>                 break;
>         };
> --
> 2.17.1
>
> --
> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/=
Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki


--
Gilad Ben-Yossef
Chief Coffee Drinker

values of =CE=B2 will give rise to dom!