| 1 |
On 10/21/2009 11:25 AM, Maxim Wexler wrote: |
| 2 |
> Hi group, |
| 3 |
> |
| 4 |
> Did linux#make menuconfig followed by linux# make && make |
| 5 |
> modules_install on the .2.6.30-gentoo-r7 sources. And copied over the |
| 6 |
> new kernel and rebooted. |
| 7 |
> |
| 8 |
> The kernel panicked. The relevant messages are: |
| 9 |
> |
| 10 |
> <...> |
| 11 |
> VFS: Mounted root (ext2 filesystem) readonly on device 8:1, |
| 12 |
> Freeing unused kernel memory: 276k freed |
| 13 |
> Warning: unable to open an initial console |
| 14 |
|
| 15 |
That warning is interesting. This is from the kernel sources: |
| 16 |
if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) |
| 17 |
printk(KERN_WARNING "Warning: unable to open an initial console.\n"); |
| 18 |
|
| 19 |
In the dark ages, before udev, /dev/console had to be created in the bare |
| 20 |
/dev directory (i.e. before anything is mounted on /dev) to prevent this |
| 21 |
sort of error. I thought it had been solved, but maybe I'm wrong about |
| 22 |
that. |
| 23 |
|
| 24 |
Of course, I'm not sure I don't get the same warning, but if I do it |
| 25 |
flashes by so quickly I can't read it. |
| 26 |
|
| 27 |
The message clearly states that / is mounted readonly, but that test for |
| 28 |
/dev/console is asking for read/write. Maybe that's why the test fails, |
| 29 |
dunno. And I don't know if that warning is important. |
| 30 |
|
| 31 |
Nowadays udev uses a tmpfs in RAM to populate /dev with devices, but |
| 32 |
at the point your kernel panics, that tmpfs doesn't exist yet, so you |
| 33 |
are stuck with the /dev directory on / including its readonly status, |
| 34 |
I'm assuming. |
| 35 |
|
| 36 |
Anyone else know? |
Archives