The relevant code in init/main.c from my instrumented rev 1482 goes...

static int noinline init_post(void)
printk(KERN_INFO "init_post entered\n");
printk(KERN_INFO "init_post returns from free_initmem\n");
printk(KERN_INFO "init_post returns from unlock_kernel\n");
printk(KERN_INFO "init_post returns from mark_rodata_ro\n");
        system_state = SYSTEM_RUNNING;
printk(KERN_INFO "init_post returns from numa_default_policy\n");

        if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
                printk(KERN_WARNING "Warning: unable to open an initial console.\n");

        (void) sys_dup(0);
        (void) sys_dup(0);
printk(KERN_INFO "init_post returns from sys_dups\n");

So, the call to open /dev/console fails.  I don't know how the Gumstix creates /dev/console but in the staged root/ it is just a 0-length regular file.  In rev 1161 it becomes a node to device 5/1 (Linux serial console, as in all Linux), but I don't know what it is in rev 148x (since I can't get the system started).
The issue is probably how device 5/1 gets hooked up to the I/O device (serial or display) that will support the system console.

Thanks for help on this issue.

On 8/1/07, Brad House <brad@mainstreetsoftworks.com> wrote:

> I ran the init=/bin/sh scenario on my rev 1161, rev 1482, and rev 1489
> builds.
> The 1161 build did what was expected, ending up in sh...

If you're using an initrd/initramfs, you need to use
to drop to a shell, that changed sometime within the 2.6 kernel series.

Not sure if it's related, but just a thought...
I had modified some of the scripts on this page:
Please note the kernel options needed if using an initramfs...


This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>   http://get.splunk.com/
gumstix-users mailing list