> Can we stop this now, and make sure that the Linux defaults are sane?
>
> Linus
I went and checked the original introduction date of triton.c. It was much
earlier than I thought -- Mark introduced it in 1.3.20, or August 1995.
Ever since its introduction, including *all* the 2.0.x series, we used to
enable DMA automatically on every drive which supported DMA mode-2 and
above. Three years have passed since, and I believe that we have to
consider all the good past experience with IDE-DMA before defaulting
to PIO mode on all drives. The data corruption reports have been *isolated*
and almost always due to a bad IDE cable or too agressive timings set
by the BIOS. PIO mode won't really fix this, at least not consistently,
just purely by luck here and there. There *have* been data corruption with
PIO mode as well.
I believe that if 2.2 ships with PIO mode by default, many users will
effectively see the Linux IDE code lagging behind several years in respect
to current IDE hardware.
IDE disks today are not much slower than the PIO interface theoretical
maximum bandwidth, and the gap is getting smaller every day. It is only
a matter of time before new disk drives will be capable of sustaining
16MB/sec on the outer cylinders, which is why Ultra-DMA raised the channel
bandwidth to 33MB/sec.
Trying to sustain 16MB/sec on a 16MB/sec channel will not mean just
that we won't be able to get high data transfer rates; since we will
be using the CPU for I/O, it means that the cpu will be wasted a lot
in the IDE driver interrupt handler, effectively preventing almost any
multi-tasking from taking place during sequential I/O.
Gadi
-
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.altern.org/andrebalsa/doc/lkml-faq.html