Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <165564429738.1413209.17302088152547038708.tglx@xen13>
 <165564429887.1413209.8729918730024457105.tglx@xen13> <CAHk-=wgfrUdWBXA-Jx7ZC1x3wwsomou0L6niGgpRS2Hd5rQxyg@mail.gmail.com>
 <Yq9Q84H6HIfkJpoR@zx2c4.com>
In-Reply-To: <Yq9Q84H6HIfkJpoR@zx2c4.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sun, 19 Jun 2022 15:05:23 -0500
Message-ID: <CAHk-=whTg=MCBYCO_+KQ10qKYhZXKXQ=mdgLc=e2gxHwDwRBxQ@mail.gmail.com>
Subject: Re: [GIT pull] locking/urgent for 5.19-rc3
To:     "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        "the arch/x86 maintainers" <x86@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sun, Jun 19, 2022 at 11:38 AM Jason A. Donenfeld <Jason@zx2c4.com> wrote:
>
> This was initially my concern too, which I expressed to Sebastian, but
> he made the point that this area here is rather "special". Actually,
> randomness isn't really required here.

That wasn't really my point.

My point was that there are a lot of uses of prandom_u32() and friends
in random places. Just grepping for it, there's lots of different
drivers that use it. Who knows what locking they have.

Clearly nobody *thought* about it. This one issue is purely about RT
correctness, but how about all the uses that just want a pseudo-random
number and may have performance issues, or may be calling things so
much that a lock is just bad.

The thing is, that prandom code used to be FAST. Not just "no locks",
but also "fairly simple siphash round because its a PSEUDO random
thing and shouldn't be anything more".

The whole "make it use the same randomness" may just have been a huge
and fundamental mistake.

We've seen one actual outright bug because of it already. That was
easy to fix by avoiding the new thing that now was a mistake. What
about all the other uses with lock bouncing or whatever subtler issues
that aren't pointed out by outright correctness tests?

                  Linus