Duncan posted <pan.2006.03.26.10.34.46.613270@...>, excerpted below,
on Sun, 26 Mar 2006 03:34:47 -0700:
> With kernel 2.6.15, configuring IOMMU won't let me boot (without a mem=
> kernel command line). Configuring it in but forcing it off (iommu=off on
> the kernel command line) gives me some IOMMU needed errors and it still
> won't boot (again, without a mem=). Configuring it out, the system boots
> just fine.
> With kernel 2.6.16, configuring IOMMU in or out, on or off, won't boot
> (without a mem=).
For those that may be interested, I think I may have a defective hardware
IOMMU. Setting iommu=soft on the kernel command line works just fine.
That's a bit disappointing for AMD hardware, where the hardware IOMMU is
supposed to be better, but I guess it's no worse than the Intel chips,
which don't have an IOMMU so emulate it in software.
It's still fairly likely that I have something not quite configured
correctly, however, and that when I figure it out, it'll work the way
it's supposed to. I have some research to do tho, to grok that.
Meanwhile, with the iommu=soft switch, I can run 2.6.16 kernels with full
memory, just as I could .15 kernels. (The .15 and apparently earlier
kernels had bounce-buffers built-in for SCSI, while .16 took them out
since there's the IOMMU in either software for Intel, or hardware for
normal AMD, anyway.)
BTW, merged KDE 3.5.2 last nite, upgraded from 3.5.1. That's the first
time I've run a major emerge session with $PORTAGE_TMPDIR on a RAM backed
tmpfs. VERY NICE! =8^) I /still/ haven't used the full 8 gigabit yet,
tho. I got upto 6.5-ish when kmail was compiling, with four other compile
jobs running in parallel (5 total), each of which was running
MAKEOPTS="-j6 -l5". The -l5 limited load some, and KDE doesn't make as
good a use of multiple compile jobs as it could (the reason I was running
5 in parallel since the -j6 -l5 doesn't work so well with KDE), but I
/was/ running 7-10 load average for several hours (on a dual Opteron 242).
Hopefully by the time I'm compiling KDE4, I'll be doing it with dual
cores, a bit faster than the 1.6 GHz of the Opteron 242s, as well.
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman in
email@example.com mailing list