We're using the PMC-Sierra MSP85XX Sequoia processor... (RM9000 family) The
BSP has PMON2000 and Linux 2.6.18 port.
In configuring/fixing the PMON to properly recognize single- and double-sided
RAM modules (no - the BSP did not come from PMC with 'fully' working code in
it!) I became familiar with the DDR DCR config process via RAM's I2C access.
The problem is we want to be able to dynamically read the DDR DCR's
(>0xFF000000) setup by PMON, and use the results of that (physical RAM size,
i.e. 2G) to call "add_memory_region" in place of the hard-coded 256M
"add_memory_region" already called inside the msp85XX-specific "plat_mem_setup"
(assign MSP85XX-specific function pointer used later
as "late_time_init" below, and assign others)
add_memory_region( 0x0, 0x10000000, ...) <-- hardcoded 256M
// normal "mem=" interpretation
// how can I get pointer to 0xFF000000+ DDR DCR here
// in the process? (before mem_init)
late_time_init = &py_map_ocd <-- this is where lots of
DCR's are accessed using
pointers provided by
ioremap - TOO LATE TO CALL
add_memory_region I suppose???
Using command-line parameters (used inside "parse_cmdline_early") is not an
option due to our system, and we don't want size-based linux images either.
I have successfully been able to read the DDR DCRs inside "late_time_init"
obtaining a pointer via ioremap call, but that is not a valid call earlier
within the context of the setup_arch sequence - is this true? (I get exceptions
when I do that) However in late_time_init is too late to call
Does anyone have any ideas? Comments? What other call/procedure analogous to
ioremap can I make earlier to access the DDR DCR insite setup_arch process?