xxxxx@acm.org said:
In general Linux is not a good reference on how NT or most of the
other OS’es on Intel platforms work.
I was using it to show that BIOSes do map bars. Anyhow, newer Linux
kernels take over PCI mapping, and Linux on non-Intel systems does
the mapping, too. I never intended this example to explain NT, but
to elucidate BIOS behavior. BIOSes *do* map bars. Whether that mapping
is used is a separate issue.
xxxxx@acm.org said:
Sorry you got this one wrong, Max is right. The HAL sets the base
addresses. There is a boot.ini switch /PCILOCK to indicate that the
addresses setup by the BIOS not be touched.
Fine, I can believe all that.
All the same (getting back to the original point of the thread) it
does not sound reasonable for a card to be turned into a PCI device
only after the kernel is booted. We dealt with PCI cards with reloadable
designs by including an external PCI interface chip onto the board and
keeping the FPGA on the local (non-PCI) bus of the interface chip.
If chip count is an important issue, I think Lucent has an FPGA with
a hard-wired PCI bridge on-chip along side the FPGA cells.
–
Steve Williams “The woods are lovely, dark and deep.
xxxxx@icarus.com But I have promises to keep,
xxxxx@picturel.com and lines to code before I sleep,
http://www.picturel.com And lines to code before I sleep.”