With an IMGate partner in Dallas, we sold a 1U co-lo box to one of his 
clients who does big mailings, as outbound gateway.

We had the box built to our specs by Tiger Direct (their Vision brand), 
with a Promise TX2000 and ATA133/7200 RPM disks.

First off, the box came with TX2 ATA66!  So Tiger overnighted us a retail 
box of TX2000.

The TX2000 at $80, not for RAID but for ATA133 to support fastest IDE disk 
i/o for postfix mailq.  The TX2 is $40, but only does only does 66, and 
somehow I missed the "Ultra133 TX2" which does 133, so next time, it will 
be an non-RAID Ultra133 TX2 to save a few bucks vs. the TX2000.

Anyway, FreeBSD dmesg hardware probe showed this:

ad4: DMA limited to UDMA33, non-ATA66 cable or device
ad6: DMA limited to UDMA33, non-ATA66 cable or device

ar0: 39083MB <ATA SPAN array> [4982/255/63] status: READY subdisks:
  0 READY ad4: 39083MB <Maxtor 6Y040L0> [79408/16/63] at ata2-master UDMA33

ar1: 39083MB <ATA SPAN array> [4982/255/63] status: READY subdisks:
  0 READY ad6: 39083MB <Maxtor 6Y040L0> [79408/16/63] at ata3-master UDMA33


UDMA33 ? what about our UDMA133 ?? and what's this "non-ATA66 cable or 
device" stuff.

We could boot the CDROM (hooked to the mobo IDE i/f) and install FreeBSD 
through the TX2000 onto the disk with no problems, so obviously thing were 
"working".

Note: don't hook ATA33 devices like cdrom on the same ATA bus as ATA133 
disk, since that will choke the bus down to ATA33.

So what's with with the slow 33 after we bought all these expensive pieces 
we so carefully specified?

The TX2000 on-board setup utility showed both drives to be mode 
"UMDA6"  (aka ATA133), so why did FreeBSD choke itself down to 33??

The short story: the integrator apparently had difficulty (as my partner 
did when he found and fixed the pb) in routing the Promise 18" ATA133 cable 
because the mid-cable 3M connector fell right on a too-small pass-thru hole 
in an transversal  bulkhead, between the drive bays and the mobo.

Welll, it was a flat cable, like all ATA cables,  obviously indifferent to 
which end was which, so the integrator reversed the cable end-to-end, and 
he then could fold the cable through the hole now that the mid-cable 
connector fell elsewhere.

Ship it!!   :))

A FreeBSD software guy suspected the TX200 was somehow different in how it 
reports cable and device status, and was sure we had to patch the ATA 
driver code NOT to report so we could FreeBSD to run at 133.  We were 
almost decided to give up on the TX2000 and hook the cable to the mobo ATA 
sockets to get at least ATA100, tons better that 33 through thru controller.

Well, Soeren Schmidt in Denmark, "Mr ATA driver" for FreeBSD, suspected the 
cable, but we swore it was the official Promise cable rated ATA133 that 
came with the controller.

 From somewhere came the idea that the cable was not end-to-end symmetric, 
and sure enough, ONE SENTENCE in the Promise doc said "blue end to the 
controller", but the 1U case was with cable doing blue end .... to the 
disk!!!

With great patience and difficulty and millmeters of luck, my partner got 
the cable + connector through the hole and bingo:

ar0: 39083MB <ATA SPAN array> [4982/255/63] status: READY subdisks:
  0 READY ad4: 39083MB <Maxtor 6Y040L0> [79408/16/63] at ata2-master UDMA133

ar1: 39083MB <ATA SPAN array> [4982/255/63] status: READY subdisks:
  0 READY ad6: 39083MB <Maxtor 6Y040L0> [79408/16/63] at ata3-master UDMA133

:)))

============

btw, even earlier than the above story, since we knew we didn't want no 
stinkin RAID, JBOD was what we wanted, but the TX2000 didn't talk about 
JBOD, we just skipped the "define array" in the TX2000 utility.  The disks 
worked fine. Installed FreeBSD perfectly, but ... no boot!! not a peep from 
FreeBSD boot block.

Then we had the idea, reading between the doc's lines, that "span" array 
was what Promise called JBOD.  So we ran the TX2000 utility to set up both 
disks, on separate ATA channels, in "SPAN" mode (see dmesg above) with one 
disk, but now FreeBSD would  boot! yeah!! but reported some "error while 
mounting..."

So we guessed the FreeBSD installed as non-SPAN wasn't exactly where it was 
expected to be after we activated SPAN mode, so we installed FreeBSD, yet 
again, and bingo, it booted.

Len


Reply via email to