From mboxrd@z Thu Jan  1 00:00:00 1970
From: Heiko Stuebner <heiko@sntech.de>
Date: Mon, 27 Mar 2017 22:11:57 +0200
Subject: [U-Boot] [PATCH 0/8] rockchip: rk3188: fixups and armclk speedup
In-Reply-To: <CAPnjgZ22dTCpEJpfGF=_ggDJyr46EjBbh9mqPYXPNAJDA7-eLQ@mail.gmail.com>
References: <20170320114036.21475-1-heiko@sntech.de> <4763928.6redEziD3S@phil>
 <CAPnjgZ22dTCpEJpfGF=_ggDJyr46EjBbh9mqPYXPNAJDA7-eLQ@mail.gmail.com>
Message-ID: <12321742.1bqC5QCCX6@phil>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 8bit
To: u-boot@lists.denx.de

Am Montag, 27. März 2017, 12:36:00 CEST schrieb Simon Glass:
> Hi Heiko,
> 
> On 24 March 2017 at 10:04, Heiko Stuebner <heiko@sntech.de> wrote:
> > Am Donnerstag, 23. März 2017, 21:28:01 CET schrieb Simon Glass:
> >> Hi Heiko,
> >>
> >> On 20 March 2017 at 05:40, Heiko Stuebner <heiko@sntech.de> wrote:
> >> > The ARMCLK starts at 24MHz on the rk3188 which makes u-boot startup
> >> > unnecessary slow. We can easily switch to 600MHz without involving
> >> > the pmic and thus do this in the SPL to also make the rc4-decoding
> >> > of the U-Boot image faster.
> >> >
> >> > Some smaller fixes also turned up while adding the ARMCLK-support.
> >> >
> >> > It's currently based on Simon's spl-working branch and Kever's
> >> > spl_early_init patch, as that includes the last missing rk3188
> >> > patches and also keeps uboot starting on rk3188.
> >>
> >> This should be in mainline now so can you also please test against that?
> >
> > yep, all good.
> >
> > As stated in my fixup series yesterday, this should of course go on
> > top of the final radxarock addition.
> 
> I sent a few patches which get it building for me. Please take a look
> and see what you think.

Patches 1+2 look very nice, see replies.

I'm not sure about the third patch though, as I don't see it affecting the
rock-tpl to much. When looking at the diassembly of the tpl-binary, the
only driver-related functions still present there, are serial_post_probe
and serial_pre_remove.

Yesterday I was playing around with serial options a bit but trying to
disable the default DM_SERIAL seemed to create just more mayhem and
would probably need even more header config settings when saving 26 bytes.