linux-kernel - Re: [PATCH v4 01/11] timekeeping: add raw clock fallback for random_get

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: <B465920D-5799-4799-A0AC-68837E159D66@zytor.com>
Date:   Thu, 14 Apr 2022 14:49:01 -0700
From:   "H. Peter Anvin" <hpa@...or.com>
To:     Rob Herring <robh@...nel.org>,
        "Jason A. Donenfeld" <Jason@...c4.com>
CC:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "open list:HARDWARE RANDOM NUMBER GENERATOR CORE" 
        <linux-crypto@...r.kernel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Arnd Bergmann <arnd@...db.de>, "Theodore Ts'o" <tytso@....edu>,
        Dominik Brodowski <linux@...inikbrodowski.net>,
        Russell King <linux@...linux.org.uk>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>,
        Geert Uytterhoeven <geert@...ux-m68k.org>,
        Thomas Bogendoerfer <tsbogend@...ha.franken.de>,
        Paul Walmsley <paul.walmsley@...ive.com>,
        Palmer Dabbelt <palmer@...belt.com>,
        Albert Ou <aou@...s.berkeley.edu>,
        "David S . Miller" <davem@...emloft.net>,
        Richard Weinberger <richard@....at>,
        Anton Ivanov <anton.ivanov@...bridgegreys.com>,
        Johannes Berg <johannes@...solutions.net>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Chris Zankel <chris@...kel.net>,
        Max Filippov <jcmvbkbc@...il.com>,
        John Stultz <john.stultz@...aro.org>,
        Stephen Boyd <sboyd@...nel.org>,
        Dinh Nguyen <dinguyen@...nel.org>,
        linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>,
        linux-m68k <linux-m68k@...r.kernel.org>,
        "open list:MIPS" <linux-mips@...r.kernel.org>,
        linux-riscv <linux-riscv@...ts.infradead.org>,
        sparclinux <sparclinux@...r.kernel.org>,
        linux-um@...ts.infradead.org, X86 ML <x86@...nel.org>,
        linux-xtensa@...ux-xtensa.org
Subject: Re: [PATCH v4 01/11] timekeeping: add raw clock fallback for random_get_entropy()

On April 14, 2022 1:41:38 PM PDT, Rob Herring <robh@...nel.org> wrote:
>On Thu, Apr 14, 2022 at 12:38:49AM +0200, Jason A. Donenfeld wrote:
>> Hi Rob,
>> 
>> On Wed, Apr 13, 2022 at 4:32 PM Rob Herring <robh@...nel.org> wrote:
>> > 'does not have a usable get_cycles(), ...' as clearly some arches have
>> > get_cycles() and yet still need a fallback.
>> >
>> > Why not handle the 'if get_cycles() returns 0 do the fallback' within
>> > a weak random_get_entropy() function? Then more arches don't need any
>> > random_get_entropy() implementation.
>> 
>> No, this doesn't really work. Actually, most archs don't need a
>> random_get_entropy() function, because it exists in asm-generic doing
>> the thing we want. So that's taken care of. But weak functions as you
>> suggested would be quite suboptimal, because on, e.g. x86, what we
>> have now gets inlined into a single rdtsc instruction. Also, the
>> relation between get_cycles() and random_get_entropy() doesn't always
>> hold; some archs may not have a working get_cycles() function but do
>> have a path for a random_get_entropy(). Etc, etc. So I'm pretty sure
>> that this commit is really the most simple and optimal thing to do. I
>> really don't want to go the weak functions route.
>
>Is random_get_entropy() a hot path?
>
>
>It doesn't have to be a weak function, but look at it this way. We have 
>the following possibilities for what random_get_entropy() does:
>
>- get_cycles()
>- get_cycles() but returns 0 sometimes
>- returns 0
>- something else
>
>You're handling the 3rd case.
>
>For the 2nd case, that's riscv, arm, nios2, and x86. That's not a lot, 
>but is 2 or 3 of the most widely used architectures. Is it really too 
>much to ask to support the 2nd case in the generic code/header?
>
>Rob

It goes into interrupts, which means it is latency critical.