linux-kernel - Re: [PATCH] /dev/random: Insufficient of entropy on many architectures

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]

Message-ID: <CAMuHMdUNPp9a_dAoeJvrZFGFxStU96Ws+WSp2gRocMngZrNSAA@mail.gmail.com>
Date:	Thu, 12 Sep 2013 14:47:23 +0200
From:	Geert Uytterhoeven <geert@...ux-m68k.org>
To:	Stephan Mueller <smueller@...onox.de>
Cc:	"Theodore Ts'o" <tytso@....edu>,
	LKML <linux-kernel@...r.kernel.org>, dave.taht@...ferbloat.net
Subject: Re: [PATCH] /dev/random: Insufficient of entropy on many architectures

On Thu, Sep 12, 2013 at 2:35 PM, Stephan Mueller <smueller@...onox.de> wrote:
>>On Thu, Sep 12, 2013 at 2:08 PM, Stephan Mueller <smueller@...onox.de>
> wrote:
>>>>BTW, I prefer a different name than "random_get_fast_cycles()", as
>>>>it's better to have something that returns different and
>>>>unpredictable numbers than an actual monotonic cycle counter.
>>>>
>>> A monotonic counter is fully ok. Note, for /dev/random, the
>>> occurrence
>>> of events delivers entropy. Thus, we have to be able to precisely
>>> measure that occurrence. The timer itself does not need to deliver
>>> any
>>> entropy as long as it is fast.
>>
>>Well, in my specific case (m68k/Amiga) I can use:
>>  - a 24-bit counter running at only ca. 15 or 31 kHz (actual
>>frequency may vary),
>>  - a 16-bit counter running at ca. 700 kHz.
>>
>>That is, if they have to be monotonic cycle counters.
>>
>>If not, I can mix the two (e.g. "a << 8 | (b & 0xff)") to get a 32-bit
>>value. That result would be fine for /dev/random, I guess, but it's
>>not really "get_cycles()".
>
> Note, get_cycles should return an u64.

Currently cycles_t is 64-bit (defined as a 64-bit type, that still doesn't say
anything about the actual values) on blackfin, c6x, cris, ia64, m32r, parisc64,
ppc64, s390x, tile, x86, and xtensa.
On all other architectures cycles_t is 32 bit.

> Not sure what a and b here is, but if a is the 24 bit value and b the
> faster 16 bit value, wouldn't there be a gap?

No, 24 + 8 = 32 (cycles_t is 32-bit on m68k).

> I.e. wouldn't it be better to use the full 16 bit counter as low value
> and OR the 24 bit on bits 48 to 17?
>
> Yet, there is a break in that counter: the 16 low bits rotate several
> times (around 10 times) before bit 17 is changed once.

Sure, but it's also slower, as this will be called for every interrupt, and the
counters (accessed by byte!) are on a slow bus. Think of Ted's L3-cache-miss
story.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@...ux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/