-----Original Message----- From: Patrick Georgi [mailto:patrick@georgi-clan.de] Sent: Thursday, May 14, 2009 11:13 AM To: Myles Watson Cc: coreboot Subject: Re: [coreboot] [PATCH] Table code cleanup
Am 14.05.2009 18:46, schrieb Myles Watson:
On Thu, May 14, 2009 at 9:33 AM, Myles Watsonmylesgw@gmail.com wrote:
Sorry to be dense. Could we do this in terms of which tables we want
in
each place?
Page0 (low_table_start): COREBOOT TABLE
0xf0000 (rom_table): PIRQ MPTABLE (with floating table) ACPI bulk
0x7fff0000 (high_tables): PIRQ MPTABLE ACPI bulk
That was less than clear :)
In general, I agree with that table. The constraints are:
MPTABLE must be two complete copies (because of Linux)
ACPI must be one copy (because there needs to be a single address for
some objects) with two RSDPs (see below).
OK.
A problem that I see with the code right now is that it depends on the implementation of acpi_write_tables, which is per-mainboard right now.
If someone decides to align their rsdp at 64 bytes instead of 16, it might break.
Right. I thought it works out, until the generic parts of ACPI aren't copied into every mainboard tree anymore.
You also end up with two rsdp structures, one in 0xf0000 and one in high_tables. Is that really what we want?
SeaBIOS overwrites the fseg with itself, then copies the RSDP in high_tables down into fseg again.
So the low RSDP is used if the payload doesn't overwrite it, the high one is used as "backup" in case the payload claims the same memory location.
In that case I think we could add an extra parameter to write_acpi_tables and write the "backup" rsdp at that location. I'll put that together for Kontron and qemu. I'm assuming that's where you'd like to test.
This is probably the time to unify write_acpi_tables, but I don't want to muddy the waters too much.
I'll test your patches tonight, and look how to get stable RSDP handling in there.
Thanks, Myles