Boot failure, 2.6.10-rc2

Boot failure, 2.6.10-rc2

[Bob Gill]

Hi.  When booting 2.6.10-rc2, I get 
Warning: unable to open an initial console
(and the boot process then stalls).

My system has the following already configured:
crw-------  1 bob root 5, 1 Nov 16 10:10 /dev/console

/etc/fstab shows:
none                    /dev/pts                devpts  gid=5,mode=620
0 0

My kernel configuration includes the following:
CONFIG_UNIX98_PTYS=y
CONFIG_LEGACY_PTYS=y
CONFIG_LEGACY_PTY_COUNT=256

I have appended console=/sbin/bash to the kernel boot line (which does
not meet with success).

If it makes any difference, my system is FC3.  Is there anything special
I have to do to udev (or a particular version I have to get in order to
start a console after the kernel is loaded (and the memory is freed
after it's shuffled from himem to lomem)?

Please reply to me directly as I'm not on the list.

TIA,

Bob  
-- 
Bob Gill <gillb4@telusplanet.net>

Re: Boot failure, 2.6.10-rc2

[Jan Engelhardt]

>Hi.  When booting 2.6.10-rc2, I get
>Warning: unable to open an initial console
>(and the boot process then stalls).
>
>My system has the following already configured:
>crw-------  1 bob root 5, 1 Nov 16 10:10 /dev/console

Are you sure /dev/console exists when the kernel boots?
(It is thy duty to ask this...)

I wonder, because there is no configurator (menuconfig) option to en-/disable
the driver for /dev/console -- it's *always* in. In 2.6.8, and I have not seen
any changes to drivers/char/tty_io.c:tty_init() - where it is added - in
further kernels yet.

>My kernel configuration includes the following:
>CONFIG_UNIX98_PTYS=y
>CONFIG_LEGACY_PTYS=y
>CONFIG_LEGACY_PTY_COUNT=256

Initial console has nothing to do with ptys.
The console driver will always be compiled in, it is not wrapped into any
#if..#endif in drivers/char/tty_io.c.
But to be on the safe side, enable:

CONFIG_VT=y
CONFIG_VT_CONSOLE=y

>I have appended console=/sbin/bash to the kernel boot line (which does
>not meet with success).

The one that takes a program is init=, but that will probably also not work
because you ain't got no console.

As I read from kernel/printk.c, the console= parameter seems to set up a
-serial line-, also see Documentation/serial-console.txt and
kernel-parameters.txt.

>If it makes any difference, my system is FC3.  Is there anything special
>I have to do to udev (or a particular version I have to get in order to
>start a console after the kernel is loaded (and the memory is freed
>after it's shuffled from himem to lomem)?

How, after all, did you run into this error? Directly after upgrading (if
applicable)?



Jan Engelhardt
-- 
Gesellschaft für Wissenschaftliche Datenverarbeitung
Am Fassberg, 37077 Göttingen, www.gwdg.de

Re: Boot failure, 2.6.10-rc2

[Bob Gill]

On Tue, 2004-11-16 at 20:36 +0100, Jan Engelhardt wrote:
> >Hi.  When booting 2.6.10-rc2, I get
> >Warning: unable to open an initial console
> >(and the boot process then stalls).
> >
> >My system has the following already configured:
> >crw-------  1 bob root 5, 1 Nov 16 10:10 /dev/console
> 
> Are you sure /dev/console exists when the kernel boots?
> (It is thy duty to ask this...)
> 
> I wonder, because there is no configurator (menuconfig) option to en-/disable
> the driver for /dev/console -- it's *always* in. In 2.6.8, and I have not seen
> any changes to drivers/char/tty_io.c:tty_init() - where it is added - in
> further kernels yet.
> 
> >My kernel configuration includes the following:
> >CONFIG_UNIX98_PTYS=y
> >CONFIG_LEGACY_PTYS=y
> >CONFIG_LEGACY_PTY_COUNT=256

> But to be on the safe side, enable:
> 
> CONFIG_VT=y
> CONFIG_VT_CONSOLE=y
> 
OK, another part of my standard build script (which is now failing to
boot) includes:
#
# Character devices
#
CONFIG_VT=y
CONFIG_VT_CONSOLE=y
CONFIG_HW_CONSOLE=y
CONFIG_SERIAL_NONSTANDARD=y
# CONFIG_COMPUTONE is not set
...
> As I read from kernel/printk.c, the console= parameter seems to set up a
> -serial line-, also see Documentation/serial-console.txt and
> kernel-parameters.txt.
> 

> How, after all, did you run into this error? Directly after upgrading (if
> applicable)?
No, the standard (old) kernel that comes with Fedora Core 3 is working
ok, but I prefer to run my own custom kernels.  The official Fedora line
is that it's wrong to build your own kernel, and you are silly for
wanting to do so, but I like to build/run them anyway.
> 
> 
> Jan Engelhardt

Thanks for your reply though.  Your question as to whether /dev/console
exists at boot time is making me question whether /dev/console exists at
boot time.
-- 
Bob Gill <gillb4@telusplanet.net>

Re: Boot failure, 2.6.10-rc2

[Jan Engelhardt]

>> How, after all, did you run into this error? Directly after upgrading (if
>> applicable)?
>No, [...] but I prefer to run my own custom kernels.
(So the post here is justified)

>Thanks for your reply though.  Your question as to whether /dev/console
>exists at boot time is making me question whether /dev/console exists at
>boot time.

You could find out by taking a live distro and checking. That's because I
suspect your old kernel, which probably still works, to contain ~~ something
magical ~~ that /dev/console is created at the right time.
devfs probably?


Jan Engelhardt
-- 
Gesellschaft für Wissenschaftliche Datenverarbeitung
Am Fassberg, 37077 Göttingen, www.gwdg.de

Re: Boot failure, 2.6.10-rc2 (resolved)

[Bob Gill]

Ok, thanks, I can get 2.6.10-rc2 to boot/run.  I didn't need devfs to
build new kernels in Fedora Core 2 (and devfs is marked as obsolete),
but the new distribution -FC3- needs devfs to run.
Thanks,
Bob  

On Tue, 2004-11-16 at 21:24 +0100, Jan Engelhardt wrote:
> >> How, after all, did you run into this error? Directly after upgrading (if
> >> applicable)?
> >No, [...] but I prefer to run my own custom kernels.
> (So the post here is justified)
> 
> >Thanks for your reply though.  Your question as to whether /dev/console
> >exists at boot time is making me question whether /dev/console exists at
> >boot time.
> 
> You could find out by taking a live distro and checking. That's because I
> suspect your old kernel, which probably still works, to contain ~~ something
> magical ~~ that /dev/console is created at the right time.
> devfs probably?
> 
> 
> Jan Engelhardt
-- 
Bob Gill <gillb4@telusplanet.net>