>After I saw the original post about irqtune here, I decided to ask about
>it on the kernel list. Linus is looking into it now.
Funny thing is, the following message was posted in that very same kernel
list on August 7th, and even though I and a couple of others chimed in with
confirmation of the problem, at that time Linus did not appear to notice the
thread, or chose not to respond (we all have the same 24hours/day to work
with, after all, so I understand....)
On August 7, [EMAIL PROTECTED] (Wallace Poon) wrote in
linux.dev.kernel:
>I used Linux 1.2.13 and then up to 2.0.3 and found no problem in
>using rz in Minicom for downloading. Once I upgraded to
>2.0.9 and beyond, there is ALWAYS CRC errors when the IDE hardisk
>read/write. There is also packet dropped in PPP session.
>So could someone help me with this problem ?
>The Quantum 1.7G IDE is connected to the on-board IDE controller.
>Wallace
> Apparently, irqtune only helps with some (probably buggy) interrupt
>controllers.
I politely beg to differ, in that I think you are generalizing excessively
and prematurely. In Wallace's case above, the same hardware did not
exhibit the problem with a slightly earlier version of the kernel (2.0.3),
so there's some food for thought.
Furthermore, in the case of my own target machine (a modest 386-40 with 8MB
RAM), I can and *have* done problem-free high-speed (V34) downloads under
DOS, Windows95, OS/2 Warp Connect and older Linux kernels (e.g. 1.2.13 as
found in Slackware 3.0 and RedHat 3.3.3). Now of course Wallace and I only
represent two data points, but shouldn't that be enough to caution you
against prematurely declaring all affected hardware to be buggy?
> Below is
>Linus' reply along with a kernel patch that *may* possibly have the same
>effect as irqtune on systems where irqtune does make a difference. On my
>system, I see no improvement whatsover with irqtune. But, others have
>seen substantial improvement. Someone told me he saw some improvement but
>at the cost of slowing down his SCSI drive.
Better to slow down disk I/O a bit than to suffer frequent overruns (and
their consequences such as block resends, PPP stalls, etc...) on incoming
serial traffic, no? Anyway, it's all a matter of priority (which is exactly
what this irqtune gizmo is supposed to allow us to tune...).
I'm glad this is being discussed and tested by a lot of knowledgeable
people. That's all I hoped to achieve by bringing this issue up in here.
Thank you for doing your share by managing to attract Linus' attention to it
on the kernel mailing list...
