[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: jump at boot



The code is actually linked to run from DRAM,
but in the beginning it is really executed from flash
so you must use position independent code
until the code is copied to DRAM (i.e branch is ok, but not jump).
You can see the code and comment about that after the
label "basse" (I didn't come up with that name... ;-)
So after that it's fine to use jump and jsr.
So the "random" behaviour depends on what happens to be
in the DRAM after reset/power on (sometimes it takes up to half a 
minute of power off for the DRAM to loose it's contents)

/Johan


----- Original Message ----- 
From: "Ronny L Nilsson" <rln@xxxxxxx.se>
To: "Bjorn Wesen" <bjornw@xxxxxxx.com>
Sent: den 5 december 2001 21:17


> > I've noticed a strange behaviour from Etrax when fiddling with
> > early stages of boot process. 
>
> It's easier to answer if you attach the faulty code :)


ok, here it is. Only thing I've done is adding 2 lines of dummy code 
early in the kernel decompressor arch/cris/boot/compressed/head.S 
(marked with a comment). When this is flashed to the development board 
it report:
Uncompressing Linux...
bad gzip magic numbers
-- System halted
If the two lines are moved down a bit though, something else happens 
(randomly!) and at some positions far down it "MAY" actually work (if 
lucky).

/Ronny