Boot LFS from compact flash into ramdrive?

David Farrell dfarrell at protoparts.com
Fri Jan 30 15:15:47 PST 2004


Mike Asher wrote:
> ----- Original Message ----- 
> From: "David Farrell" <dfarrell at protoparts.com>
> Newsgroups: lfs.support
> To: <lfs-support at linuxfromscratch.org>
> Sent: Thursday, January 29, 2004 8:08 PM
> Subject: Re: Boot LFS from compact flash into ramdrive?
> 
> 
> 
>>Ken Moffat wrote:
>>
>>>On Mon, 26 Jan 2004, BoB wrote:
>>>
>>>
>>>
>>>>I'd like to boot from the CF card in a read-only fashion, loading an LFS
>>>>system to a ramdrive.  Are there any docs that describe this more in
> 
> depth?
> 
>>>>Can someone offer a pointer or some assistance?  FWIW, I already have a
> 
> dev
> 
>>>>box running SuSE, so I'll only be running LFS on the other device.
>>>>
>>>
>>>
>>> Not quite what you asked, but there's the readonly_rootfs.txt hint at
>>>your local mirror (i.e. '/' is read only, so much less needs to be in
>>>your ramdrive, but access to the CF is reputed to be slow).
>>>
>>>Ken
>>
>>I am running a EPIA ME-6000 with a Sandisk UltraII series compact flash.
>>These CF cards are not slow, quite the contrary.  The only thing I tried
>>  to eliminate was a swap partition.  I figured heavy activity was bad
>>for the card.  When I tested with a swap partition, there was very
>>little virutal memory activity, so this assumption was probably not a
>>good one.  I do leave the root ext2 R/O and have a separate ext3
>>partition for /var and /tmp.  I build the image with the uname hack for
>>a i486 processor.  I use the same card with a Tsystems TS-5500 which
>>uses the AMD elan processor.  I have a number of these boards "in the
>>field" for the past several years running LFS with Interbase SQL.
>>
> 
> 
> If you want a bit of extra performance, you can get around the i486
> limitation on the Epias.  The problem is a bug in GCC (they lack one of the
> optional cmpxchg cmovs, but gcc assumes they have it).  With the newest
> gccs, you can specify C3, with older ones you can just hack the opcode out
> of gcc's md rules table.
> 
> BTW, the Nehemiah+ cores are supposed to implement this; I've not verified
> it myself.
> 
> -
> Mike
> 
Thanks for the info, I really needed the i486 for the elan based board 
but for testing I used the EPIA.  I like the ME6000 since I am looking
to replace the TS-5500 with the EPIA and I really wanted fanless.
I mispoke and said I build the image, of course what I really meant is
I built LFS with the uname hack.  Perhaps a rebuild specifying a
C3 in the uname_hack is in order.  The hack I used was the loadable
module version.  Everything built fine with LFS5.0+ up to the point
Greg changed glibc, I have yet to rebuild with the new lib.
David.




More information about the lfs-support mailing list