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, 21 Nov 2022 04:23:20 +0100
From:   "Jason A. Donenfeld" <Jason@zx2c4.com>
To:     Eric Biggers <ebiggers@kernel.org>
Cc:     linux-kernel@vger.kernel.org, patches@lists.linux.dev,
        linux-crypto@vger.kernel.org, x86@kernel.org,
        Thomas Gleixner <tglx@linutronix.de>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>,
        Carlos O'Donell <carlos@redhat.com>
Subject: Re: [PATCH v5 2/3] random: introduce generic vDSO getrandom()
 implementation
Message-ID: <Y3rvKEeSAq6548Sz@zx2c4.com>
References: <20221119120929.2963813-1-Jason@zx2c4.com>
 <20221119120929.2963813-3-Jason@zx2c4.com>
 <Y3liVEdYByF2gZZU@sol.localdomain>
 <Y3l6ocn1dTN0+1GK@zx2c4.com>
 <Y3l9MXCO02/rOpFx@zx2c4.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <Y3l9MXCO02/rOpFx@zx2c4.com>
Precedence: bulk

On Sun, Nov 20, 2022 at 02:04:49AM +0100, Jason A. Donenfeld wrote:
> On Sun, Nov 20, 2022 at 01:53:53AM +0100, Jason A. Donenfeld wrote:
> > I'm not quite sure what the best approach here is. One idea would be to
> > just note that libcs should wait until vgetrandom() has returned
> > everywhere before forking, using its atfork functionality. 
> 
> To elaborate on this idea a bit, the way this looks is:
> 
>   rwlock_t l;
>   pid_t fork(void)
>   {
>     pid_t pid;
>     write_lock(&l);
>     pid = syscall_fork();
>     write_unlock(&l);
>     return pid;
>   }
>   ssize_t getrandom(...)
>   {
>     ssize_t ret;
>     ...
>     if (!read_try_lock(&l))
>     	return syscall_getrandom(...);
>     ret = vdso_getrandom(...);
>     read_unlock(&l);
>     return ret;
>   }
> 
> So maybe that doesn't seem that bad, especially considering libc already
> has the kind of infrastructure in place to do that somewhat easily.
> Maybe there's a priority locking thing to get right here -- the writer
> should immediately starve out all future readers, so it's not unbound --
> but that seems par for the course.

Fortunately none of this was necessary, and I've got things implemented
without needing to resort to that, for v+1.

Jason