linux-kernel - Re: [PATCH v1] crypto: caam

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] [thread-next>] [day] [month] [year] [list]
Date:   Wed, 2 Aug 2017 14:03:14 +0000
From:   Horia Geantă <horia.geanta@....com>
To:     Harald Freudenberger <freude@...ux.vnet.ibm.com>,
        Oleksij Rempel <ore@...gutronix.de>
CC:     Herbert Xu <herbert@...dor.apana.org.au>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Oleksij Rempel <o.rempel@...gutronix.de>,
        "Dan Douglass" <dan.douglass@....com>,
        "linux-crypto@...r.kernel.org" <linux-crypto@...r.kernel.org>,
        "kernel@...gutronix.de" <kernel@...gutronix.de>,
        Martin Schwidefsky <schwidefsky@...ibm.com>,
        "David S. Miller" <davem@...emloft.net>
Subject: Re: [PATCH v1] crypto: caam - set hwrng quality level

On 7/20/2017 4:08 PM, Harald Freudenberger wrote:
> On 07/19/2017 08:13 PM, Oleksij Rempel wrote:
>> On Wed, Jul 19, 2017 at 04:53:21PM +0000, Horia Geantă wrote:
>>> On 7/19/2017 7:32 PM, Oleksij Rempel wrote:
>>>> On Wed, Jul 19, 2017 at 12:49:47PM +0000, Horia Geantă wrote:
>>>>> On 7/19/2017 10:45 AM, Oleksij Rempel wrote:
>>>>>> According documentation, it is NIST certified TRNG.
>>>>>> So, set high quality to let the HWRNG framework automatically use it.
>>>>>>
>>>>>> Signed-off-by: Oleksij Rempel <o.rempel@...gutronix.de>
>>>>>> ---
>>>>>>  drivers/crypto/caam/caamrng.c | 6 ++++++
>>>>>>  1 file changed, 6 insertions(+)
>>>>>>
>>>>>> diff --git a/drivers/crypto/caam/caamrng.c b/drivers/crypto/caam/caamrng.c
>>>>>> index 41398da3edf4..684c0bc88dfd 100644
>>>>>> --- a/drivers/crypto/caam/caamrng.c
>>>>>> +++ b/drivers/crypto/caam/caamrng.c
>>>>>> @@ -292,10 +292,16 @@ static int caam_init_rng(struct caam_rng_ctx *ctx, struct device *jrdev)
>>>>>>  	return 0;
>>>>>>  }
>>>>>>  
>>>>>> +/*
>>>>>> + * hwrng register struct
>>>>>> + * The trng is suppost to have 100% entropy, and thus
>>>>>> + * we register with a very high quality value.
>>>>>> + */
>>>>>>  static struct hwrng caam_rng = {
>>>>>>  	.name		= "rng-caam",
>>>>>>  	.cleanup	= caam_cleanup,
>>>>>>  	.read		= caam_read,
>>>>>> +	.quality	= 999,
>>>>> Why not 1024, i.e. where is 999 coming from?
>>>> It comes from s390-trng.c driver.
>>>> Should I use 1024 instead?
>>>>
>>> AFAICT the range for quality is from 0 to 1024 (no entropy -> perfect
>>> entropy).
>>>
>>> 1024 should be used since I'd expect a HW TRNG to provide perfect
>>> entropy unless otherwise stated.
>> I assume 1024 can be given only on verified HW with accessible verilog
>> files and compared with resulting chip :)
>> May be this would be a good example https://www.sifive.com/
>>
> Well, the header file says:
> ...
> /**
>  * struct hwrng - Hardware Random Number Generator driver
>  * @name:               Unique RNG name.
>  * @init:               Initialization callback (can be NULL).
>  * @cleanup:            Cleanup callback (can be NULL).
>  * @data_present:       Callback to determine if data is available
>  *                      on the RNG. If NULL, it is assumed that
>  *                      there is always data available.  *OBSOLETE*
>  * @data_read:          Read data from the RNG device.
>  *                      Returns the number of lower random bytes in "data".
>  *                      Must not be NULL.    *OBSOLETE*
>  * @read:               New API. drivers can fill up to max bytes of data
>  *                      into the buffer. The buffer is aligned for any type
>  *                      and max is a multiple of 4 and >= 32 bytes.
>  * @priv:               Private data, for use by the RNG driver.
>  * @quality:            Estimation of true entropy in RNG's bitstream
>  *                      (per mill).
>  */
> ...
> quality = estimation of true entropy per mill.
"per mill as in https://en.wikipedia.org/wiki/Mill_(currency) ?
I consider it rather unfortunate, as already noticed here:
https://lkml.org/lkml/2014/3/27/210

And isn't this inaccurate, since the (de)rating factor is quality/1024,
not quality/1000?

> I understand this as quality=1000 meaning 100% entropy.
> However, the core code currently does not really check this value.
> When more than one hwrng sources do register, simple the one with
> the higher quality value wins :-) The value is not even checked
> to be in a given range.
> 
> I searched through some device drivers which do register at
> the hwrng and it looks like most of the drivers do not even
> set this value. My feeling is, you should use 999 when your

Maybe this is because it's not clear how to determine quality's value?

Take CAAM's engine HWRNG: it can work both as a TRNG and as a
TRNG-seeded DRBG (that's how it's currently configured).
IIUC, both setups are fit as source for the entropy pool.

Do I have to set quality value comparing the two cases?
(It's a bit like comparing the quality of entropy offered by RDSEED vs.
RDRAND.)
Meaning: give full credit - maximum quality - for the TRNG setup, and
provide a lower value for quality in the case of TRNG-seeded DRBG?

> hardware provides 'perfect' random. So there is a chance for
> an even 'more perfect' hardware coming up later to overrule
> your 'perfect' hardware.

I am not sure why the hwrng with the best quality wins, instead of using
all available resources, as suggested here:
https://lkml.org/lkml/2014/3/27/210

Thanks,
Horia