On Fri, 10 Sep 2010 15:42:32 -0700 "Ira W. Snyder" iws@ovro.caltech.edu wrote:
Newer Linux kernels can overrun the initial memory window used for booting with their BSS area. When this happens, they overwrite the FDT and silently fail to boot.
On e300 CPUs, the Linux kernel uses an initial BAT covering the first 256MB of RAM. See arch/powerpc/kernel/head_32.S for details. Increase the value of CONFIG_SYS_BOOTMAPSZ to accommodate the maximum value allowed by Linux. This will allow very large kernels to boot.
Signed-off-by: Ira W. Snyder iws@ovro.caltech.edu
Wolfgang,
I did both mpc83xx and mpc51xx (CONFIG_E300) at the same time. There were fewer boards than I expected. A quick read through head_fsl_booke.S shows that CONFIG_E500 could be increased to 64MB. I'll leave that to that arch's maintainer. They have already increased theirs to 16MB, after finding this problem.
This patch doesn't attempt to add support for actually loading a kernel larger than 8MB uncompressed (code + data, but not bss). That would mean adding #define CONFIG_SYS_BOOTM_LEN SOME_VALUE_GREATER_THAN_8MB.
This boots up fine on my MPC8349EMDS.
Thanks, Ira
include/configs/MPC8308RDB.h | 4 ++-- include/configs/MPC8313ERDB.h | 4 ++-- include/configs/MPC8315ERDB.h | 4 ++-- include/configs/MPC8323ERDB.h | 4 ++-- include/configs/MPC832XEMDS.h | 4 ++-- include/configs/MPC8349EMDS.h | 4 ++-- include/configs/MPC8349ITX.h | 4 ++-- include/configs/MPC8360EMDS.h | 4 ++-- include/configs/MPC8360ERDK.h | 4 ++-- include/configs/MPC837XEMDS.h | 4 ++-- include/configs/MPC837XERDB.h | 4 ++-- include/configs/MVBLM7.h | 4 ++-- include/configs/SIMPC8313.h | 4 ++-- include/configs/TQM834x.h | 4 ++-- include/configs/aria.h | 4 ++-- include/configs/kmeter1.h | 4 ++-- include/configs/mecp5123.h | 4 ++-- include/configs/mpc5121ads.h | 4 ++-- include/configs/pdm360ng.h | 4 ++-- include/configs/sbc8349.h | 4 ++-- include/configs/ve8313.h | 4 ++-- include/configs/vme8349.h | 4 ++-- 22 files changed, 44 insertions(+), 44 deletions(-)
applied to u-boot-mpc83xx/next. Wolfgang, hope you don't mind.
Thanks,
Kim