1 |
On 01/29/2019 02:17 PM, Neil Bothwick wrote: |
2 |
> AFAIR the initramfs code is built into the kernel, not as an option. The |
3 |
> reason given for using a cpio archive is that it is simple and available |
4 |
> in the kernel. The kernel itself has an initramfs built into it which is |
5 |
> executed automatically, it's just that this initramfs is usually empty. |
6 |
> So loading an initramfs is trivial for any kernel, and loading anything |
7 |
> after that is handled by the initramfs. |
8 |
|
9 |
That may be the case now. |
10 |
|
11 |
But when I started messing with Linux nearly 20 years ago that was not |
12 |
the case. The kernel and the initramfs / initrd were two distinct |
13 |
things. I remember having to calculate where the kernel stopped on a |
14 |
floppy disk so that you could start writing the initramfs / initrd image |
15 |
after the kernel. |
16 |
|
17 |
Or for fun, modify the flag (bit?) to tell the kernel to prompt to to |
18 |
swap disks for the initramfs / initrd. |
19 |
|
20 |
Both of which needed to tell the kernel where the initramfs / initrd |
21 |
started on the medium. |
22 |
|
23 |
That was a LONG time ago. More than a few things have changed since then. |
24 |
|
25 |
> That only leaves loading the initramfs file from disk, which is handled |
26 |
> by the bootloader along with the kernel file. |
27 |
|
28 |
That assumes that there is a boot loader. There wasn't one with the old |
29 |
Slackware boot & root disks. |
30 |
|
31 |
|
32 |
|
33 |
-- |
34 |
Grant. . . . |
35 |
unix || die |