Pete Zaitcev wrote:
> Is that true?!! If they are physical (by Linus) then sparc arch is
> into a serious rewrite because Gleb made them virtual! His pcic.c
> silently replaces these addresses with virtual ones and drivers
> "just use" them.
Yes, it's true in some sense. They are PCI addresses, so they are
neither physical nor virtual. Fortunately, we seem to use transparent
mapping and it should work while
1. we never write to real PCI regs and
2. prom initializes all PCI devices (or each device initializes itself)
It seems not to be true with new JavaStation and cp1200, because you
might plug any PCI device in.
The proper PCI initialization should include
Identification of PCI IO and MEM ranges for each register of each device
Mapping those ranges to the physical address space. PCIC has several
windows for that.
Mapping physical range(s) to the virtual address space.
You have to implement PCI specific ioremap and, possibly, sparc_alloc_io
to do this.
Regards,
Gleb.
-
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of the message to [EMAIL PROTECTED]
