Date: Thu, 27 Jul 2000 18:21:16 +0200
From: Abramo Bagnara <abramo@alsa-project.org>
> PCI mmap to user space can be done (obvious coherency and other issues left
> to the user mode sufferer) but that isnt done by ioremap
I suppose you mean I need to use remap_page_range with the address
obtained from ioremap, don't you?
No, the ioremap address is not even really defined to be a kernel
virtual address, which is why you must use readw and can't just
dereference the pointer.
But, that apart, is a device driver that give available a part of the
PCI IO space of controlled card via mmap portable to all platforms that
have PCI?
If you do it correctly. I agree it's currently not all that aparent
how to do this correctly. On sparc64, we actually store the physical
addresses in the pci_dev->resource[xxx] values, but nobody should know
that, it's implementation defined what these resource values mean.
The only part of them which is defined is how you go about making use
of them, that is all.
And if this is true, the user application that use this mmaped area may
be written in a platform independant way?
I'd like that you show a little example of what you call "suffering".
Frankly, if userspace wants to get at PCI I/O ports (as opposed to PCI
MEM areas) it must know if it is x86/ia64 vs. some other
architecture. Because on the former you just feed the raw base
address register values + some offset into inb etc. instructions,
whereas on the latter you must mmap the I/O space area desired and
then go about accessing them. It gets even uglier on older Alpha's
which need to access things specially since these cpus lack byte sized
load and store instructions.
Later,
David S. Miller
davem@redhat.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jul 31 2000 - 21:00:25 EST