linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* PCI mapping on large memory 32-bit machines
@ 2003-05-19 18:15 Timothy Miller
  2003-05-19 18:25 ` Richard B. Johnson
  2003-05-19 22:59 ` William Lee Irwin III
  0 siblings, 2 replies; 3+ messages in thread
From: Timothy Miller @ 2003-05-19 18:15 UTC (permalink / raw)
  To: Linux Kernel Mailing List

On x86 with PAE and 4 gigs of RAM or more, where do memory-mapped I/O 
devices get mapped (in the physical address space)?  Most PCI devices 
can't handle 64-bit addresses.  Can PC chipsets physically remap some of 
the RAM to above 4 gig?  Or do you just lose that much RAM?  If both RAM 
and some I/O device are mapped to the same location, isn't there a conflict?


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: PCI mapping on large memory 32-bit machines
  2003-05-19 18:15 PCI mapping on large memory 32-bit machines Timothy Miller
@ 2003-05-19 18:25 ` Richard B. Johnson
  2003-05-19 22:59 ` William Lee Irwin III
  1 sibling, 0 replies; 3+ messages in thread
From: Richard B. Johnson @ 2003-05-19 18:25 UTC (permalink / raw)
  To: Timothy Miller; +Cc: Linux Kernel Mailing List

On Mon, 19 May 2003, Timothy Miller wrote:

> On x86 with PAE and 4 gigs of RAM or more, where do memory-mapped I/O
> devices get mapped (in the physical address space)?  Most PCI devices
> can't handle 64-bit addresses.  Can PC chipsets physically remap some of
> the RAM to above 4 gig?  Or do you just lose that much RAM?  If both RAM
> and some I/O device are mapped to the same location, isn't there a conflict?
>

The answer to PC/PCI is that the I/O space set (usually by the BIOS)
into the BARs removes any RAM visibility in that area. But.... this
is BAD bacause the BIOS may still claim that there is 4 gig of RAM.
The OS may then try to use it. To "solve" this problem, Win/tell started
the "high-RAM" specification where RAM higher than XXX Megs gets
mapped with page-registers. The problem is that "XXX" is board-specific!

So, to answer your entire question... don't do it! Only use 3 gigs max
and you will not be confused by confused hardware!

Cheers,
Dick Johnson
Penguin : Linux version 2.4.20 on an i686 machine (797.90 BogoMips).
Why is the government concerned about the lunatic fringe? Think about it.


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: PCI mapping on large memory 32-bit machines
  2003-05-19 18:15 PCI mapping on large memory 32-bit machines Timothy Miller
  2003-05-19 18:25 ` Richard B. Johnson
@ 2003-05-19 22:59 ` William Lee Irwin III
  1 sibling, 0 replies; 3+ messages in thread
From: William Lee Irwin III @ 2003-05-19 22:59 UTC (permalink / raw)
  To: Timothy Miller; +Cc: Linux Kernel Mailing List

On Mon, May 19, 2003 at 02:15:23PM -0400, Timothy Miller wrote:
> On x86 with PAE and 4 gigs of RAM or more, where do memory-mapped I/O 
> devices get mapped (in the physical address space)?  Most PCI devices 
> can't handle 64-bit addresses.  Can PC chipsets physically remap some of 
> the RAM to above 4 gig?  Or do you just lose that much RAM?  If both RAM 
> and some I/O device are mapped to the same location, isn't there a conflict?

AFAIK most (if not all) of that lands below 4GB in extant chipsets/BIOS's.

Remapping above 4GB is possible but various things would probably barf.


-- wli

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2003-05-19 22:46 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-05-19 18:15 PCI mapping on large memory 32-bit machines Timothy Miller
2003-05-19 18:25 ` Richard B. Johnson
2003-05-19 22:59 ` William Lee Irwin III

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).