On Thu, Aug 04, 2016 at 11:43:49PM +0000, Karl Beldan wrote:
> On Thu, Aug 04, 2016 at 04:13:10PM -0500, Rob Herring wrote:
> > +Peter H
> >
> > On Thu, Aug 4, 2016 at 4:03 PM, Karl Beldan <kbeldan@xxxxxxxxxxxx> wrote:
> > > On Thu, Aug 04, 2016 at 01:29:02PM -0500, Rob Herring wrote:
> > >> On Thu, Aug 04, 2016 at 02:43:32PM +0000, Karl Beldan wrote:
> > >> > The LCDK is the successor to the late Hawkboard and has the same machine
> > >> > number.
> > >> > Among the differences are the flash (16bits vs 8bits) and some pins
> > >> > (MMC, LEDs, buttons, some external connectors), however the main
> > >> > components remain the same (eth. PHY, Audio Codec, Video decoder and
> > >> > DAC) except for the main PMIC, different and hard-wired on the LCDK (the
> > >> > LDOs and DCDCs are always ON).
> > >> > A DT-only boot with this addition gives functional uart, reboot via
> > >> > watchdog, rtc, ethernet and MMC (I added the CD gpio for the MMC
> > >> > although davinci_mmc doesn't call yet the mmc core OF facilities).
> >
> > [...]
> >
> > >> > + aliases {
> > >> > + serial2 = &serial2;
> > >> > + };
> > >> > +
> > >> > + chosen {
> > >> > + bootargs = "console=ttyS2,115200n8 earlycon";
> > >>
> > >> You don't need console with stdout-path set. And really, earlycon should
> > >> not be the default.
> > >>
> > > Ok for earlycon.
> > >
> > > For the console, I thought likewise .. until I _struggled_ to understand
> > > why the console was disappearing under my feet.
> > > It seemed even stranger when I started looking at the codepath because
> > > drivers/of/base.c would properly add_preferred_console as expected after
> > > the corresponding UART was added. But the reality was there, without it
> > > the console was not like stdout-path.
> > > Among the commits that could instill doubt as to whether stdout-path
> > > rules is f64255b5072d:
> > > "The assumption that at least 1 preferred console will be registered
> > > when the stdout-path property is set is invalid, which can result in
> > > _no_ consoles.", though it didn't answer my question.
> > > I started looking whether tty0 was not getting in the way and in the end
> > > I came up with setting console.
> > > But you seem affirmative.
> >
> > There have been some issues in this area, but I thought they were in
> > conjunction with earlycon and solved now. This is a fairly common
> > review request for dts files, so I'm surprised if it doesn't work
> > right. Maybe having tty0 or not has some influence. Or people just
> > accept review comments without testing or caring...
> >
> I can assure you I tested without early{con,printk} ;).
>
> Now I have just rechecked the codepaths and something caught my
> attention in kernel/printk/printk.c:__add_preferred_console(): if
> console_cmdline is empty selected_console gets unconditionally set to
> zero ! Knowing this I confirmed by replacing with idx (works), and also
> by using ttyS0 (i.e. alias serial0 = &uart2) (works), this would not
> affect DTs using serial0. It's late here so I'll stop there.
>
The stdout-path bug was fixed last week by Paul Burton in 05fd007e4629
("console: don't prefer first registered if DT specifies stdout-path").
Karl
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
