1 |
Thanks to Chris Smart, I now have a bootable kernel! |
2 |
|
3 |
The New Behavior: |
4 |
|
5 |
The kernel gets as far as mounting /dev/hda1 on /newroot, and it |
6 |
appears to raidstart /dev/md0, but /dev/md0 doesn't have a |
7 |
superblock. When I looked around, I discovered that all the drives |
8 |
on my PCI IDE controller are not discovered. |
9 |
|
10 |
Since the LiveCD uses the siimage module, I tried compiling with |
11 |
CONFIG_BLK_DEV_SIIMAGE=y. But after doing this, the problem is that |
12 |
my hda is not discovered properly. |
13 |
|
14 |
My guess was that the PCI IDE controller was somehow messing up the |
15 |
device order. Since the LiveCD boot uses siimage as a module, I |
16 |
tried compiling the kernel with CONFIG_BLK_DEV_SIIMAGE=m. |
17 |
|
18 |
But now the kernel doesn't discover the PCI IDE controller's drives. |
19 |
The module isn't getting loaded. When I typed "modprobe siimage; |
20 |
raidstart /dev/md0; mount /dev/md0 /mnt/md0" everything worked fine. |
21 |
|
22 |
The new problem: |
23 |
|
24 |
I generate initrd using "genkernel initrd". The resulting initrd |
25 |
doesn't seem to be loading the siimage module during boot. |
26 |
|
27 |
When I compile the siimage driver in permanently (not as a module), |
28 |
the drive order gets messed up (that's just a HYPOTHESIS). |
29 |
|
30 |
The new question: |
31 |
|
32 |
How can I get siimage to load early enough for raid to properly |
33 |
start, but late enough that it doesn't mess up the drive order? |
34 |
|
35 |
Thanks, |
36 |
Dave |