Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 18 Jul 2022 20:47:15 -0700
From:   Eric Biggers <ebiggers@kernel.org>
To:     "Guozihua (Scott)" <guozihua@huawei.com>
Cc:     linux-crypto@vger.kernel.org, luto@kernel.org, tytso@mit.edu,
        "Jason A. Donenfeld" <Jason@zx2c4.com>
Subject: Re: Inquiry about the removal of flag O_NONBLOCK on /dev/random
Message-ID: <YtYpQ1gf23mxIiYH@sol.localdomain>
References: <13e1fa9d-4df8-1a99-ca22-d9d655f2d023@huawei.com>
 <d49d35f5-8eaf-d5e8-7443-ac896a946db7@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <d49d35f5-8eaf-d5e8-7443-ac896a946db7@huawei.com>
Precedence: bulk

On Mon, Jul 18, 2022 at 04:52:59PM +0800, Guozihua (Scott) wrote:
> On 2022/7/14 15:33, Guozihua (Scott) wrote:
> > Hi Community,
> > 
> > Recently we noticed the removal of flag O_NONBLOCK on /dev/random by
> > commit 30c08efec888 ("random: make /dev/random be almost like
> > /dev/urandom"), it seems that some of the open_source packages e.g.
> > random_get_fd() of util-linux and __getrandom() of glibc. The man page
> > for random() is not updated either.
> 
> Correction: I mean various open source packages are still using O_NONBLOCK
> flag while accessing /dev/random
> > 
> > Would anyone please kindly provide some background knowledge of this
> > flag and it's removal? Thanks!
> > 

This was changed a while ago, in v5.6.  /dev/random no longer recognizes
O_NONBLOCK because it no longer blocks, except before the entropy pool has been
initialized.

(I don't know why O_NONBLOCK stopped being recognized *before* the entropy pool
has been initialized; it's either an oversight, or it was decided it doesn't
matter.  Probably the latter, since I can't think of a real use case for using
O_NONBLOCK on /dev/random.)

The random(4) man page is indeed in need of an update, not just for this reason
but for some other reasons too.

The util-linux code which you mentioned is opening /dev/random with O_NONBLOCK
if opening /dev/urandom fails, which is pretty much pointless.  Perhaps the
author thought that /dev/random with O_NONBLOCK is equivalent to /dev/urandom
(it's not).  The glibc code, if you mean sysdeps/mach/hurd/getrandom.c, is
actually code written for GNU Hurd, not for Linux, so it's not relevant.

- Eric