As far as I know, it tunes specifically for dma multiword mode 2. That
is not udma, but was the best available before udma. (Note that I may
well be talking rubbishe here).
We are in the hands of the people who can decipher the buzzwords here.
What is udma anyway, if it's not busmastering dma? Is it simply a
faster physical channel, or a faster mode of using an existing channel?
Did the mode always exist, or is it new to both mobo and disks? Wups:
seems to exist! Gah .. read it and didn't learn much.
I guess it's purely the mobo and the drive's business if they use it or
not, and we simply have to tell or allow them to use it. That must
entail the kernel writing some bits somewhere. Support should be in new
2.0.3* and 2.1.* kernels. I sem to remember finding a whitelist
somewhere once in my older kernel source (triton.c?), and adding my
drives to it.
> So what's the option to turn it on?
Dunno! Done by the kernel by default (or not).
> Unless, of course, UDMA is even faster than anything I've measured so
> I've never had UDMA running. (I'm getting about 9.2MB/sec from my
> Quantum FB ST 6.4Gig).
I'm getting about 10.5MB/s from my fujitsu on the BX board. That may be
using udma, since I did nothing to turn it off, and I didn't see any
bios options that I could decipher that would æffect it either. I get
6-9Mb/s off the same kinds of disks on older boards. Older = slower.
But those speeds would be limited by the disk rotation rates. I guess
udma is only useful to get data out of the disk on-board cache real fast
over to the mobo and cpu. I see 140MB/s registered by hdparm for buffer
transfers on the BX. I see 42MB/s on TX boards, and 25MB/s on my trusty
triton 1 board.
> > Won't "ide0=noautotune" do?
> For a long while I had to use "ide0=nodma".
> Now I don't need it, I just do -X34 during a boot script.
> What would "ide0=noautotune" do?
Turn off udma, for one thing. I had to read the kernel code and ask the
> It could lock up during the boot process before the script, but I don't
> boot often enough to find out (now that the lockups don't seem to
It was inevitable when I tried it. Immediately after issuing a hdparm
-k (keep settings).
> > One lesson is that the ide boot parameters should be less crytic!
> > "ide0=udma" maybe? And hdparm needs a little extension too.
> And somewhere something could give me a clue what my system thinks it is
> now doing!
Amen!!!! That is what i meant. Actually the patched hdparm 3.3 seems
to report the disks ideas pretty well. We just don't know what the
kernel thinks about it. Only clue is the pci and other (ide?) entries
in proc. Cc:'ed to linux-kernel in the hope that someone heeds the plea for
more and clearer status information.
> -- Jamie
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to firstname.lastname@example.org
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html