[parisc-linux] Re: two minor c3k bugs

[parisc-linux] Re: two minor c3k bugs

[Grant Grundler]

Helge Deller wrote:
> Forwarded to the list, since it may be of interest for others too:
> 

sure.

> On Friday 30 November 2001 08:43, Grant Grundler wrote:
...
> Hmm, I still have 2.4.16-pa4 here and it definately shows 
> "2.4.16-pa4" for me. Maybe a SMP problem ?

No. I only build UP kernels for the C3k.

> You could try to double the size of the udelay() parameter
> in lcd_print to see if it cures your problem.

After a reboot it worked fine. Could be a timing problem.
Since I can't reproduce it reliably and it's minor, I'm not
going to chase this. Someday when I get bored and ate too much
chocolate (dark, organic, Belgian from Trader Joe's), I might
try to run IRC on the LCD... ;^)

...
> Currently I use "kill_proc(1, SIGINT, 0);" to kill the init-process -
> which is normally a reboot-command and gives the above messages.

Ok. that makes sense.

> That's AFAICS not the right solution, but nfortunately I don't know
> how to do it correct right now.

./kernel/sys.c:         printk(KERN_EMERG "Power down.\n");

That chunk of code seems to do the right thing:
        case LINUX_REBOOT_CMD_POWER_OFF:
                notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
                printk(KERN_EMERG "Power down.\n");
                machine_power_off();
                do_exit(0);
                break; 

I know that msg since I regularly "powerdown" my c3k.
Perhaps replicate that code?

cheers,
grant

[parisc-linux] Re: two minor c3k bugs

[Helge Deller]

Hi Grant,

On Saturday 01 December 2001 07:11, Grant Grundler wrote:
> > Currently I use "kill_proc(1, SIGINT, 0);" to kill the init-process -
> > which is normally a reboot-command and gives the above messages.
>
> Ok. that makes sense.
>
> > That's AFAICS not the right solution, but nfortunately I don't know
> > how to do it correct right now.
>
> ./kernel/sys.c:         printk(KERN_EMERG "Power down.\n");
>
> That chunk of code seems to do the right thing:
>         case LINUX_REBOOT_CMD_POWER_OFF:
>                 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF,
> NULL); printk(KERN_EMERG "Power down.\n");
>                 machine_power_off();
>                 do_exit(0);
>                 break;
>
> I know that msg since I regularly "powerdown" my c3k.
> Perhaps replicate that code?

Yes, that's the right code-path, but sadly calling this function
would  immediately power off the computer without doing a nice
system shutdown.
I'll need to investigate how to first find the PID of the init process and
then how to tell it to shut down the system cleanly.

Helge

> cheers,
> grant

[parisc-linux] Re: two minor c3k bugs

[Helge Deller]

Forwarded to the list, since it may be of interest for others too:

On Friday 30 November 2001 08:43, Grant Grundler wrote:
> Hi helge!

Hi Grant, 

> It's totally cool that I can shutdown the c3k with just power button!
> playing with 2.4.16-pa5 today I noticed to really minor bugs.
> 1) LCD shows "Linux 2.4.1-pa5" in the top line.
>    UTS_RELEASE is definitely 2.4.16 since uname -a prints that too.
>    It wasn't obvious to me where the "6" was getting lost.
>    I only looked at led.c for a bit - just enough to understand
>    that UTS_RELEASE was used to generate the string.

Hmm, I still have 2.4.16-pa4 here and it definately shows 
"2.4.16-pa4" for me. Maybe a SMP problem ?
You could try to double the size of the udelay() parameter
in lcd_print to see if it cures your problem.

> 2) When powering off with poweroff button, the last msg is
>  "Restarting ...". Are we running the wrong event chain?
>   I haven't had a chance to look at this yet. (reviewing
>   code for bame and dug up the DIAG register access code.)

Currently I use "kill_proc(1, SIGINT, 0);" to kill the init-process -
which is normally a reboot-command and gives the above messages.
That's AFAICS not the right solution, but nfortunately I don't know
how to do it correct right now.
The other thing is, that this call will not work if the init-process
doesn't has PID 1 - this happened to me after a kernel crash, where
init needed to fsck.

> thanks!
> grant

Greetings,
Helge