Debian (Knoppix) on HP a350n

[Rick Moen]

Quoting sms@sonic.net (sms@sonic.net):

> I just mean that
> (1)
> I had found quite a few references to libata being in a 2.4.22 path, e.g...
>    /lib/modules/2.4.22-1.2140.nptlBOOT/kernel/drivers/scsi/libata.o
> & similar showing up in lots of Google-hits;
> (2)
> I hadn't seen (?m)any such with libata in a 2.4.24 path; and
> (3)
> I had found some comments from MarcelloT &/or JeffG to the effect of
> having "thought better of" making libata part of 2.4.x & those wanting
> it "should" move to 2.6.
> 
> I had concluded, perhaps wrongly, that it was inserted into the "main"
> 2.4 tree with 2.4.22 & removed again upon reflection.  It's (more than)
> possible that I was just reading too quickly/shallowly & drawing
> unwarranted conclusions.

Sorry I can't help you.  For all I know, there may well have been more
than one thing called libata.  What I've encountered is:

1.  Jeff Garzik wrote a driver set called libata that supports numerous
ATA chipsets via the kernel's SCSI layers, for 2.5.x/2.6.x.

2.  That driver set is sometimes (rarely) applied as a backport to 2.4.x.

> If it's a theoretcial boost to the ceiling, IMHO, that should be made
> explicit: 

1.  I was probably paraphrasing some other source.

2.  I might get around to rewriting that, but frankly teaching people
the fundamentals of hardware isn't a job I signed up for.  It's not 
hard to figure out that ATA has always, in every iteration of that
hardware bus standard, been bottlenecked by physical hard drive access,
given its lack of support for disconnect mode.

> I'm hardly a benchmarking expert; if you'd like to suggest a test or
> suite (reasonably-easy-to-run), and a pointer to a baseline (to which
> my numbers can be compared), I'd be pleased to run the tests & contribute
> my numbers.

Sorry, no.  I was just saying that I haven't seen your results and
therefore rationally cannot comment on them.  (You seemed to be saying
that your hdparm results proved something.  That's possible, but I
obviously cannot say much without hearing details.)

> <g>  The box has had a total of about 3h of uptime since I got Debian
> installed; "working OK" is IMHO a premature conclusion.  I'd like to
> find a few tests to apply and/or have a plan in case I find the thing
> going panic/freeze on me...

Well, OK, but my point is that, earlier, you seemed to be implying that
you needed help solving some problem:  One becomes wary over time of
leaping to assist someone who's asking for help, but who doesn't get
around to identifying what problem he's trying to solve.

When and if you have a problem and specify it, likely enough this
mailing list will be a significant resource.

> Personally, I like to actually get online with my modems...
> ;^)

That's nice, but I remember your question being (paraphrasing) "How do I
tell whether this thing is a winmodem or not?"  The way I mentioned
works and has the advantage of using simple tools, and therefore is
debugging-friendly.

However, by all means, if you're feeling lucky and prefer the optimist's
approach of firing up some dialer (e.g., wvdial) and going for broke,
then nothing's stopping you.