[SerialICE] Readback errors during RAM filling (writel)

[Idwer Vollering]

Hello,

I have a 'minor' problem with what it seems is memory filling.
During this process, there are two "readback errors", qemu and gdb
seem to die (go in a S+ state) where writel shows 000201b8.

Would this be a good time to use ram_is_initialized and/or
SerialICE_register_physical() ? Setting ram_is_initialized at this
point seems too early to me.

The memory region that is written to when the readback errors show up
should be covered by this line [1] from serialice.lua: "elseif  addr
>= 0x00000000 and addr <= 0x000dffff then", right?
Is one (or both) of the write(s) overwriting serialice on the target machine?

Relevant logfile snippet (the whole log is over one megabyte):
MEM: writel 00000000 <= 00000000 *
MEM: writel 00000004 <= 00000000 *
MEM: writel 00000008 <= 00000000 *
MEM: writel 0000000c <= 00000000 *
MEM: writel 00000010 <= 00000000 *

// abbreviated

MEM: writel 0001d3b0 <= 00000000 *
MEM: writel 0001d3b4 <= 00000000 *
MEM: writel 0001d3b8 <= 00000000 *
MEM: writel 0001d3bc <= 00000000 *
Readback error! fffffff0/30
MEM: writel 0001d3c0 <= 00000000 *
MEM: writel 0001d3c4 <= 00000000 *
MEM: writel 0001d3c8 <= 00000000 *
MEM: writel 0001d3cc <= 00000000 *

//abbreviated

MEM: writel 000201b0 <= 00000000 *
MEM: writel 000201b4 <= 00000000 *
MEM: writel 000201b8 <= 00000000 *
Readback error! fffffff8/30


Idwer

[1] http://review.coreboot.org/gitweb?p=serialice.git;a=blob;f=SerialICE/scripts/serialice.lua;h=a5dfa62d9fbb9bf3b9b443b108c6733730c7f122;hb=HEAD#l416