-----Original Message----- From: ron minnich [mailto:rminnich@gmail.com] Sent: Monday, October 12, 2009 9:37 PM To: Myles Watson Cc: Carl-Daniel Hailfinger; Stefan Reinauer; coreboot Subject: Re: [coreboot] [PATCH] more Kconfig default fixes
On Mon, Oct 12, 2009 at 7:36 PM, Myles Watson mylesgw@gmail.com wrote:
All I'm looking for is the shortest path to "not-broken". I'm open to suggestions.
So I've had this explained to me several times, let's try again.
In fact it's in the v2 code. It's just that the v2 code is so hard to read ...
- BSP core 0 starts up in CAR.
All other cores start up and put themselves to sleep ASAP.
BSP core 0 sets up DRAM programming for ALL sockets. BPS gets CBFS files in RAM. BSP zeros memory attached to BSP. 2. BSP uses IPIs to set up all AP stacks and EIP. AP core 0s start up. They do what initialization they need (e.g. zero their own DRAM). They then stop. 2a. BSP waits for all AP core 0 to either stop or error out. 3. BPS starts up core>0 and sends IPIs to get APs to start up core>0. 4. All core 0s wait for all core>0 to stop or error out. 5. BSP waits for all AP core0 to stop again. 6. BSP continues to boot.
Is there a problem with this sequence?
I don't think so.
The reason we're back on this topic is because we're having trouble nailing down the correct values for CONFIG_STACK and CONFIG_HEAP. I just noticed that they don't actually matter in CAR, so we have RAM by the time we even check those values.
Thanks, Myles