AW: [U-Boot-Users] USB booting...

AW: [U-Boot-Users] USB booting...

[Martin Krause]

Hi all,

u-boot-users-admin at lists.sourceforge.net wrote on :
> In message <1120664027.6064.99.camel@daroca> you wrote:
> > I am attempting to boot Linux on a target platform using a
> > USB mass storage device (flash disk) to hold the kernel and the
> > root filesystem.
> > 
> > I thought it may be a good idea to replicate the same steps
> > used when booting from a tftp server, i.e. I copy the root
> > filesystem (using an initrd image) to an address in RAM, then
> > copy the kernel to a different address in RAM.  Finally bootm with
> > the start address of the kernel. Bootargs will be set during this
> > process.
> 
> You can do it this way, but it is a waste of system  RAM.  Why  don't
> you use a (read-only) file system mounted on the USB device?

I tried this with a read-only ext2 root file system on a linux partition
on an USB memory stick some time ago - and failed. I hadn't much time
to analyze what's going wrong but I think the problem was that the
kernel (linuxppc_2_4_devel from denx) tried to mount the root file 
system _before_ initializing of the usb sub system has completed. At
least the console messages looked so - but perhaps I missed something.

Has anybody else tried this? I would be happy to hear about a 
successful way to do this.

Regards,
Martin

AW: [U-Boot-Users] USB booting...

[Peter Korsgaard]

>>>>> "Martin" == Martin Krause <Martin.Krause@tqs.de> writes:

 Martin> I tried this with a read-only ext2 root file system on a
 Martin> linux partition on an USB memory stick some time ago - and
 Martin> failed. I hadn't much time to analyze what's going wrong but
 Martin> I think the problem was that the kernel (linuxppc_2_4_devel
 Martin> from denx) tried to mount the root file system _before_
 Martin> initializing of the usb sub system has completed. At least
 Martin> the console messages looked so - but perhaps I missed
 Martin> something.

 Martin> Has anybody else tried this? I would be happy to hear about a
 Martin> successful way to do this.

Yes, there's a race condition between the kernel mounting the rootfs
and USB detection / partition scan - I have solved it with a small
initrd (~100k) which sits in a loop trying to mount the USB stick and
then does a pivot_root to it.

-- 
Bye, Peter Korsgaard