[clug] Kernel without initramfs
Robert Edwards
bob at cs.anu.edu.au
Wed Mar 25 23:44:12 GMT 2009
David,
I notice in one of your earlier posts that you may be using ext3 on your
CF device. May I suggest that it is a bad idea to use ext3 in particular
and most other journalling FS's in general on a "flash"-based device?
There are special FS's to use such as jffs, but I would recommend
dividing up your file-system structure into read-only bits (eg. /usr)
and writable parts. For the writable parts work out if you can afford
to write to a RAM-base FS and then copy important data back to
non-volatile storage. Logging etc., if possible, should be to an
external logging server.
Just some ideas that I have had to address in the past.
Bob Edwards.
David Cottrill wrote:
> Thanks all, the big delay was coming from the debian-live initramfs
> bootscripts (I'm using Debian as host too).
>
> I believe scsi was the right place to look but it is still failing at
> the same point. Have tried all variations of root=/dev/hda1, /dev/sda1
> LABEL=...
>
> Inserting a generic kernel into the same image is a serious performance
> hit - these SoC motherboards are not like a standard computer, so P2
> 200MHz /128MB ram is actually more like 486dx 200MHz / 100MB ram. I've
> built PII 400 desktops (Gentoo and Debian) and they are vastly different
> beast.
>
> Even so, the boot time has been cut down enough using debootstrap
> instead of cdebootstrap that I can deal with the idea of having a
> generic kernel (but I will change modules=most to modules=dep).
> debootstrap more accurately assesses my idea of a 'minimal' system.
>
> When I say generic kernel - I mean Debian, not vanilla.
>
> The setting for CONFIG_RAMFS I believe is 'y' (but support for
> initrd/initramfs is 'n') so I'll give that one last crack
>
> Now that I'm settled on not using debian-live, I have to figure out a
> way to reduce the number of writes to disk. CF cards don't react well to
> that over the short term and it's positively destructive medium to long
> term.
>
> Interestingly I've noticed that many flash disk drive manufacturers are
> pitching at the server market. Flash drives can take maybe a million
> hits (cheap ones 100,000) per bit before they hit MTBF. Even with load
> levelling, I'm betting most people out there could rack that up on a
> serious server in a day or so.
>
> Thanks again,
> David
>
> On Wed, Mar 25, 2009 at 11:52 AM, Robert Edwards <bob at cs.anu.edu.au
> <mailto:bob at cs.anu.edu.au>> wrote:
>
> David Cottrill wrote:
>
> I'm working at getting a P2 300MHz to boot in a reasonable time
> - the
> initramfs runtime on a generic kernel is about 2 minutes.To fix
> this I've
> built and will be rebuilding a custom kernel with all the
> required modules
> built in.
>
> What I've done - lsmod from the generic kernel, made sure all
> modules listed
> got included in the kernel in my custom version.
>
> The remaining problem: After finding all my devices (including
> my hard disk)
> the boot sequence fails to mount the root partition (local).
>
> The first and last error message is:
> VFS: can't mount (insert disk label or /dev/hda1 or /dev/sda1 or
> anything
> else here) or block (0,0)
> Please append a correct root=... Choose from the following
> partitions:
>
> There isn't a list of partitions to choose from so I guess it is
> the problem
> is in the part of the kernel that recognises partitions but my
> most recent
> build included almost all of the partition options so I'm a
> little bit
> mystified.
>
> Any suggestions?
>
>
> Thanks,
> David
>
>
> Are you compiling a kernel.org <http://kernel.org> kernel or one
> from the Ubuntu kernel
> source packages? The Ubuntu kernel sources include patches to find
> the root file system by disk label. You probably should identify your
> root fs by block device instead (eg. root=/dev/sda1 etc.) (and in
> fstab as well...).
>
> Cheers,
>
> Bob Edwards.
>
>
More information about the linux
mailing list