I would like to point out that for item 2, killing memory hole, the 430TX at least, has hardwired the 0xa0000-0xbffff to go to PCI, so this particular hole must exist (for TX at least) even when using a linear frame buffer in PCI memory space above main memory.
----- Original Message ----- From: "Ronald G Minnich" <[EMAIL PROTECTED]> To: "Eric W. Biederman" <[EMAIL PROTECTED]> Cc: "ollie" <[EMAIL PROTECTED]>; "Jeremy Jackson" <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]> Sent: Tuesday, April 02, 2002 7:01 AM Subject: Re: important changes .... > On 1 Apr 2002, Eric W. Biederman wrote: > > I don't have a problem with that. We might have to do something interesting > > with the MP tables. But it should work just fine. You are the one who > > originally insisted the LinuxBIOS table be at 0. All of my code looks in > > boot places for the LinuxBIOS table first 0-4K then 0xf0000 - 0xFFFF. > > yeah, my fault. Sorry. The reason for 0-basing those tables was: > 1- linux doesn't use page 0 > 2- if we put them at 0, we can kill the infamous "hole in memory" > > OK, eric, before I push a potentially Very Bad Idea -- should we force > linux to look for PIRQ in the 0-4096 range instead. We just recently > discovered this crazy problem with PIRQ, DoC, and some mainboards. What's > the best thing to do? > > > Interesting. So the code to copy the pirq tables to didn't work. > > it would appear not. We have good IRQ tables in the DoC, but Linux never > finds them ... which took us a while to figure out. > > ron >
