Kirkwood: no console messages from init_sequence?
I rebased my dockstar patches on master, after the recent ARM relocation support for kirkwood had been merged.
I no longer see any of the early console messages that should result from display_banner, print_cpuinfo, etc. The first thing I see is the NAND information:
Marvell>> reset
resetting ...
NAND: 256 MiB
In: serial
Out: serial
Err: serial
Net: egiga0
[...](I see the same behavior on a cold start, not just a reset.) Can someone with another kirkwood board please check whether they see this behavior also, or did I somehow break something when I rebased? Thanks.
Signature · 1 line
Eric Cooper e c c @ c m u . e d u
Quoted 4 lines · Eric Cooper
I no longer see any of the early console messages that should result from
display_banner,print_cpuinfo, etc. The first thing I see is the NAND information:
[...]
But if I load the same u-boot image into memory and execute it with "go", I see the entire banner, DRAM message, etc. So the problem only occurs when it is booting from NAND flash.
Signature · 1 line
Eric Cooper e c c @ c m u . e d u
Dear Eric Cooper,
Quoted 9 lines · Eric Cooper
Quoted 4 lines
I no longer see any of the early console messages that should result from
display_banner,print_cpuinfo, etc. The first thing I see is the NAND information:
[...]But if I load the same u-boot image into memory and execute it with "go", I see the entire banner, DRAM message, etc. So the problem only occurs when it is booting from NAND flash.
Looks to me like the console does not get initialized early or that the console driver does not work before relocation. Check for possible static data usage in the driver.
Best Regards
Reinhard
Quoted 5 lines
Dear Eric Cooper,
Quoted 3 lines · Eric Cooper
Quoted 2 lines
I no longer see any of the early console messages that should result from
display_banner,print_cpuinfo, etc. The first thing I see is
the
Quoted 9 lines
Quoted 6 lines
Quoted 2 lines
NAND information:
[...]But if I load the same u-boot image into memory and execute it with "go", I see the entire banner, DRAM message, etc. So the problem only occurs when it is booting from NAND flash.
Looks to me like the console does not get initialized early or that the console driver does not work before relocation. Check for possible
static
Quoted 1 line
data usage in the driver.
The init_sequence is the first variable to hold initialized static data, possibly
its pointers are wrong even before relocation. Will be if your link address!=load address