Re: [PATCH 2/2] crypto: caam - fix DMA mapping of stack memory

From: Horia Geanta <horia.geanta@nxp.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Roland Hieber <rhi@pengutronix.de>,
	Aymen Sghaier <aymen.sghaier@nxp.com>,
	"linux-crypto@vger.kernel.org" <linux-crypto@vger.kernel.org>,
	Jesper Nilsson <jesper.nilsson@axis.com>,
	Lars Persson <lars.persson@axis.com>,
	Christian Lamparter <chunkeey@gmail.com>,
	Roland Hieber <r.hieber@pengutronix.de>,
	"David S. Miller" <davem@davemloft.net>,
	"kernel@pengutronix.de" <kernel@pengutronix.de>
Subject: Re: [PATCH 2/2] crypto: caam - fix DMA mapping of stack memory
Date: Fri, 25 Jan 2019 14:26:27 +0000	[thread overview]
Message-ID: <VI1PR0402MB34859DF7A32E650C9E2E1A34989B0@VI1PR0402MB3485.eurprd04.prod.outlook.com> (raw)
In-Reply-To: 20190125134756.5f5e2uxmwvtnc6bb@gondor.apana.org.au

On 1/25/2019 3:48 PM, Herbert Xu wrote:
> On Fri, Jan 25, 2019 at 12:43:21PM +0000, Horia Geanta wrote:
>>
>> Herbert, is there something we could do to avoid this?
> 
> Async crypto requests cannot be allocated off the stack.  So
> whoever is doing this needs to stop.
> 
IIUC, the crypto request is allocated in ima_calc_file_hash_atfm() using
ahash_request_alloc(), so this seems to be fine.

However, further below the req->result pointer is set to a location on stack:
out2:
	if (!rc) {
		ahash_request_set_crypt(req, NULL, hash->digest, 0);

More exactly, the call sequence is:
ima_collect_measurement() -> ima_calc_file_hash() -> ima_calc_file_ahash() ->
ima_calc_file_hash_atfm()
and "hash" is allocated on stack in ima_collect_measurement().

Thus, it seems the problem lies in the fact that ahash_request structure defines
the digest buffer as "u8 *result" - thus the memory area might not be DMAable.

Thanks,
Horia