Failing to load a freshly compiled kernel 2.6.13.1 with swsusp2 4927

You will need a serial or network console (i.e., another cpu on another machine - obviously the kernel that is failing doesn't get as far as starting up any sensible logging itself, but you can watch it failing from afar). Failing that, a camera and a place to mount your snapshots.

I've never done a network console myself, though I run serial console most days. Tell me how if you get that far - I'm lazy. I must say it seems to me like some jiggery-pokery with netboot PFX must be involved! Ho hi. I guess the trick is explained in the Documentation directory in the source.

OK. It works here, and is something I have in a module. You can investigate it by loading a module later. There is no need for it to be in the kernel.

:1,.s-=y-=m

Then one would suspect you are not loading the kernel you think you are loading.

What is this ramdisk it is loading? If there is a ramdisk there ... aren't you expecting to load it? Or is it just that you don't expect to find a cramfs there, but instead something like a compressed ext2fs, as usual?

You REALLY want to boot that kernel on something it will boot on (a live cd fs? A floppy? Tomsrtbttdisk) and check whatproc-devices andproc-filesystems says (after subtractingproc-modules contribs).

OK. It ran out of file systems to try, according to the code. It seems to think it can open the root device, and that it's a CRAMFS (I may be wrong about the sequence - I am too lazy to trace through all the code). After that it should do whatever is ininitrc on the ramdisk but it seems instead not to manage a mount but instead t run out of file systems to try.

To my eye it looks as though you are confused about

a) which kernel you are loading (doesn't seem to be the one you compiled); or maybe b) which kernel you compiled; and c) if it should open a ramdisk or not.

I comment c) because it seems to me that you are not wanting it to open a ramdisk, and indeed I cannot think of a good reason for running a ramdisk when you have a DISK. May I ask how and where you have concocted the kernel boot line?

Anyway you probably want to check boot params and tick a ramdisk=no in there, or whatever does that (see Boot Param HOWTO).

No, I want it to. Once you have booted on a live cd then we can readproc-* and see what is up.

Just use any old live cd you have. We don't need to get very far in the boot. Boot from harddisk as usual with root=-dev-hdb (or whatever) as a boot parameter.

And you have not changed kernel version? The devices may have shanged support kinds with kernel version.

But that doesn't mean that they STILL will be that way on a new kernel version (I don't know, since I have no SATA disks and a 2.6.12.3 kernel at newest, but I seem to have heard that SATA support has surprises in newer kernels).

Anyway - please first rebutture yourself that you are booting (or compiling) the kernel you think you are.

And please drop that ramdisk! You have everything you need on disk. Why would you need to load modules from a ramdisk? Boot forcing no ramdisk and root=-dev-hda9 and init=-bin-sh (sync sync sync). If you keep the ramdisk in there, it will be confusing.

But first remove the cnfusion about which kernel it is.

Peter