Debian (Knoppix) on HP a350n

Rick Moen rick@linuxmafia.com
Tue, 24 Feb 2004 15:20:57 -0800


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

> Another bit of evidence is that we never diddled with the BIOS (as
> prior SATA-on-Linux doc's implied was required).

Steve --

Let's review the Linux/SATA topic for a minute.

Starting last year, about ten or so chipmakers started producing ATA
("IDE") host adapter cards -- and adapter circuitry embedded into
motherboards -- compliant with an improved ATA spec called Serial ATA 
("SATA").  The prior ATA spec in all its sundry versions was thereupon
dubbed Parallel ATA ("PATA").

When that happened, the existing Linux 2.4.x kernel family's ATA drivers
automatically kinda-sorta supported the new chipsets, because by and
large they were modest extensions to existing ones.  But some entirely
new driver code was needed, plus some revisions of existing drivers
became necessary to make the new hardware work reliably in SATA mode
(under Linux).

One very new set of drivers, Jeff Garzik's libata set, got introduced
only in the (then-experimental) 2.5.x/2.6.x kernel series.  It's capable
of being applied (backported) to the earlier and still more-common 2.4.x 
kernel series, but seldom is found that way in commonly available
precompiled 2.4.x kernels, such as those used by Linux distribution
installers.

Absent Garzik's libata having been backported to a Linux distribution's
installation kernel, that kernel's ability to support arbitrary SATA
chipsets is kind of... limited and (in some cases) prone to freeze-ups.
Thus, whether that distribution will successfully install onto your
_particular_ SATA chipset -- if the chipset is running in SATA mode -- 
may depend on which such chip it is.  

If memory serves, your HP Pavilion A350N sports an Intel ICH5 SATA
chipset.  The reason your hot-off-the-press Knoppix had no problem with
that is twofold:

1.  The ICH5 is sufficiently similar to the earlier ICH4 and prior PATA
chipsets (and thus the longtime ATA driver's piix support) that it
doesn't (usually) have serious problems.

2.  Happily, one of the advantages of using hot-off-the-press Knoppix is
that it gets you easy access to a very recent 2.4.x kernel.  Very recent
2.4.x kernels include tweaks to the piix support that cure freeze-ups
encountered on ICH5 when using earlier 2.4.x kernels.  If you'd been
using, say, an installer running 2.4.18, the installer might well have
frozen up.

_If_ a particular distro's installer fails to see block devices -- or
freezes up -- and if one suspects that it's because of paltry SATA
support, then there are several possible workarounds.  One of those is
(where permitted by the system BIOS Setup program) to temporarily switch
the chipset into "legacy PATA mode" via the BIOS Setup.  Then you run
through the installer.  Next, you fetch a kernel with better SATA
support.  Last, you re-enable SATA mode in the BIOS.  So, to revisit
your statement:

> Another bit of evidence is that we never diddled with the BIOS (as
> prior SATA-on-Linux doc's implied was required).

Um, no.  Not exactly _required_.  Simply often helpful.

I.e., _if_ your installer chokes on recognising block devices at all,
when they're running in SATA mode, try running them in _non-SATA_ mode
long enough to get the distribution running.

> My "helper" at the Installfest did a "hdparm -t" for some testing, and
> suggests that the speed "looks like" it's running fully as a SATA
> disk.

Now, _that's_ a bunch of rubbish.  SATA raises the theoretical maximum
bus bandwidth of an ATA bus, but it does nothing at all to make an ATA
drive actually pump out data faster.  Remember:  ATA chains support only
one device being active at a time.  The bottleneck of ATA is still, as
it always has been, the physical access limits of hard drives, and it's
still completely impossible to even max out an antique ATA/100 bus's bus
ceiling, let alone the alleged 150 MB/sec bus limit of SATA.  Hell, I'd
give no better than even money as to whether ATA/66 can yet be
saturated.

Anyhow, if you want to find what communications mode your ATA chain(s)
is (are) set to use, look in the BIOS Setup screens.  But, if you think
SATA vs. PATA makes a speed difference, think again.  It really doesn't.
(SATA _does_ allow you to use slim-profile cables without violating the
ATA spec, thus letting you use the things in small system boxes without
impeding airflow horribly.  That's probably the biggest benefit.)

> Unfortunately, I *was* time-crunched, so I didn't quite have time to
> finish all the config issues.  In particular, I haven't checked if I'm
> stuck on a "poxy winmodem".  :-/    "lspci" baby... lspci all the way,
> with a visit to Linmodem!

Only you can decide how much your time is worth.  Me, I'd try to set up
dial-up.  Then, if that seemed to fail for mysterious reasons, I'd give
it up as a probable winmodem, and go buy a real (external, serial-port)
modem to use instead.

More on that at:  http://linuxmafia.com/~rick/faq/

-- 
Cheers,                   The cynics among us might say:   "We laugh, 
Rick Moen                 monkeyboys -- Linux IS the mainstream UNIX now!
rick@linuxmafia.com       MuaHaHaHa!" but that would be rude. -- Jim Dennis