[ANNOUNCEMENT] Wiki for discussing P35/IHC9(R)/SATA issues set
韓家標 Bill Hacker
askbill at conducive.net
Mon Nov 5 23:40:30 PST 2007
Aryeh M. Friedman wrote:
>> Presuming a good cable, this looks very much like a hardware (and its
>> onboard firmware) issue from where I sit.
> Suprising the urge to hurl my keyboard across the room at people's lack
> of research (it is not just you bill so take no personal offense):
I *AM* aware of those, and I don't take offense. Hear me out:
> 1. (I will dig the refs up in a few mins) this has been reported on
> multiple drives and ihc9 based controllers. So it is *NOT* a HW
> issue... see the wiki.
Versteh. But *I* get grumpy when it is *so* easy to 'close the door firmly' by
slapping in last-year's scrap IDE CD or a diferent SATA device, and *then* post
with a great deal more certainty that those options produced this or that
difference. Or NOT. Ergo the problem is more clearly ID'ed - or reconfirmed -
Surprises DO happen. Remember the CMD-640 ATA controller? Those folks call
themselves SiliconImage nowadays, but it was the tier-one Japanese firm that
'foundered' the silicon.
> 2. The only difference between different hardware is the implications of
> sw issues... for example I get timeouts but no infinite cycle when
> booting from cd
ACK - and developers of software very seldom have access to even the tiniest
fraction of the hardware that the community at large have their hands on.
WTF - as predominantly a 'hardware' R&D guy, I personally seldom have fewer than
a dozen MB within arm's length, though rarely are more than 2 or 3 'current'
items. Had a lot more before shipping a chunk of it off to developers.
So - yes - what I am asking is or *should be* redundant work and may sound
useless repitition at first.
But it places a very light load on 'many hands' of experimenters - most of whom
cannot code anyway, but *can* attach a cable - so as to reduce the load on 'few
hands' of developers. Very few.
Not as efficient as an aircraft hangar full of every make of parts known with
all developers in the same building, but placing the 'extra' work on hundreds of
testers vs few developers is the only tool we can afford to shorten the time
needed to solve the problem before the parts go obsolete for a NEW set of
On which score we ARE falling behind... Accelerating pace of 'new' is the nature
of the market.
Another reason is that while it is very inexpensive to source all manner of MB &
controllers here in HKG, to name one (cheaper than in Taipei by far..) it is
deucedly expensive in time & money to gin up proper customs paperwork, ship them
to, for example the EC, and manage their incoming VAT w/o harm to the recipient.
The VAT alone on a 'free' box of hardware can cause serious heatburn on a
student budget, 'coz the national authority takes a dim view of even 'bona fide
IOW - a 'donation' across a national border just doesn't fly well.
And the key developers are not only all over the world now, some of the best of
the team are in countries that may be quite modern - but do NOT have good,
cheap, *low tax* parts markets.
More information about the freebsd-current