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;
Date:   Sun, 1 May 2022 07:44:00 +0900
From:   Stafford Horne <shorne@gmail.com>
To:     "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc:     linux-kernel@vger.kernel.org, linux-crypto@vger.kernel.org
Subject: Re: [PATCH v7 11/17] openrisc: account for 0 starting value in
 random_get_entropy()
Message-ID: <Ym27sFdFZEt5QV0i@antec>
References: <20220423212623.1957011-12-Jason@zx2c4.com>
 <20220429001648.1671472-1-Jason@zx2c4.com>
 <YmyOh/YbgGJ3DK79@antec>
 <YmyRATzJBQX+wnIP@zx2c4.com>
 <Ym20Ga3VsJRzXppd@antec>
 <Ym25Wn1PtpNgkKyR@zx2c4.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Ym25Wn1PtpNgkKyR@zx2c4.com>
Precedence: bulk

On Sun, May 01, 2022 at 12:34:02AM +0200, Jason A. Donenfeld wrote:
> Hi Stafford,
> 
> On Sun, May 01, 2022 at 07:11:37AM +0900, Stafford Horne wrote:
>  
> > I was thinking about this, the reason the tick timer is returing 0 is because
> > the timer is not started.  It's getting initialized right after the random
> > number generator.
> > 
> > A patch like this helps to startup the timer during intial startup, but I am not
> > sure its the best thing:
> > 
> > diff --git a/arch/openrisc/kernel/head.S b/arch/openrisc/kernel/head.S
> > index 15f1b38dfe03..a9b3b5614e13 100644
> > --- a/arch/openrisc/kernel/head.S
> > +++ b/arch/openrisc/kernel/head.S
> > @@ -521,6 +521,9 @@ _start:
> >         l.ori   r3,r0,0x1
> >         l.mtspr r0,r3,SPR_SR
> >  
> > +       l.movhi r3,hi(SPR_TTMR_CR)
> > +       l.mtspr r0,r3,SPR_TTMR
> > +
> >         CLEAR_GPR(r1)
> >         CLEAR_GPR(r2)
> >         CLEAR_GPR(r3)
> 
> Yea, great, I was thinking about doing it in assembly earlier in boot
> too, but didn't know how you'd feel about that. I like this better.
> 
> The reason I think this is a good approach is that it means the cycle
> counter includes some information about how long startup takes from the
> earliest stages -- which could involve probing various devices or
> strange things. So enabling the timer in head.S seems good to me.
> 
> > But I wonder:
> >  - Why don't any other architectures have similar issues.
> >  - Is there any more correct place to do an early timer kick off.
> 
> I think most other archs (like, say, x86) have their cycle counter
> enabled by default at boot time. I was surprised to see that the or1k
> risc cycle counter comes disabled by default actually.
> 
> I'll send a v9 incorporating your suggested assembly change.

Thanks!

-Stafford