FreeBSD > 4.11 hardware issue with Asus PSCH-SR/SATA mobo
KimmoA
I just went through a FreeBSD-related "hardware hell" session, but
let's not get into that. Let me explain my problem in its entirety.
My hardware: Asus PSCH-SR/SATA mobo. This is supposed to be made for
servers and has won many "awards", according to Asus. Nothing fancy
about it as far as I can tell.
(The CPU is a P4 and the HDD is a 10 k RPM WD Raptor through S-ATA.)
I bought the hardware and assembled it originally in an 1U rackserver
chassis over a year ago. Back then, I tried to install FreeBSD 5.x, as
it was the latest stable available. It wouldn't work on the box,
though. The FreeBSD boot screen just "wrote over" itself and the
previous text already on the screen (IRQ stats and that kind of shit).
So I installed 4.11. It WORKED! Now, 4.x is considered ancient, and I
have moved on to 6.x on my other box, and really, really don't wanna
run 4.x again.
So... my problem is that my hardware seems to be giving FreeBSD 5.x and
6.x issues, but not 4.x. I find this very weird.
Tonight, I updated to the latest BIOS version, tested default settings,
and even several altered configurations ("prioritze on-board VGA" and
stuff like that). No luck. Same god damn problem...
Let me explain exactly what happens in detail: The box boots, reads the
"bootonly" 6.1 CD inserted and starts outputting text... When the first
row is done, it doesn't "feed", but starts writing over the first
(bottom) line. Then, the entire "menu" screen (with the ASCII art
saying "FreeBSD") is outputted OVER the uncleared screen!
Now, you see underscores flashing around at fixed points on the screen,
and the CD is still reading, as if stuff is happening in the
background, only I can't see it. Pressing Enter doesn't do anything,
but I seem to be able to reboot still at this point by pressing Ctrl +
Alt + Delete.
(There is NOTHING wrong with the hardware monitor!)
Alright... I'm really worried now, and I really don't wanna run Linux
or any such crap. I have come to love FreeBSD and don't want to switch
to another Unix just because of some stupid hardware issue.
What can I do? I can't even afford new hardware, and that would be
pretty silly anyway, since this is STANDARD HARDWARE!
The 4.11 installation ran for almost a year without any problems
whatsoever.
I have probably forgotten to mention something, but I did my best
explaining my problem as exactly as I possibly could.
Any help is greatly appreciated.
jpd
On 2006-06-11, KimmoA <kim...@gmail.com> wrote:
> What can I do? I can't even afford new hardware, and that would be
> pretty silly anyway, since this is STANDARD HARDWARE!
There's no such thing as standard peecee hardware. It usually works,
mostly, but apparently you ran into something weird. Ok, not fun, I
know, but shit happens, and these are, after all, just peecees, built by
the cheapest bidder, ``designed for windows'', and all that.
I must admit to not having touched recent hardware recently so I
can only give general advice, but here goes:
There are a number of things you can do. One, figure out what the
problem really is caused by. By the looks of it, it is some strange
interaction between the installer and/or the terminal driver and/or
the hardware. You could ask a relevant freebsd mailinglist (start
with freebsd-questions, see the FreeBSD project site for more info).
Note that altough you made a valiant attempt at being factual, your
frustration still shines through, even getting in the way of facts now
and then, and you might want to work on that. You need a cool head to
fix this.
Next, you could try and circumvent the problem: You could try and poke
the installer to run over the first serial, then hook up a terminal (or
other peecee with a terminal emulator). See the handbook for that.
Another way is to take the disk out of the machine, and put it in some
other machine, then install enough of FreeBSD for it to boot, then
transplant it back. There's even multiple ways to do it: As a sole
disk and do the usual install things, or as an extra disk on a working
installation. In the latter case, you have to do the entire install
process by hand, but you can do so entirely without the installer.
So, you do have a number of options; probably even more than I just
mentioned. Do report back what you tried and whether you succeeded.
--
j p d (at) d s b (dot) t u d e l f t (dot) n l .
This message was originally posted on Usenet in plain text.
Any other representation, additions, or changes do not have my
consent and may be a violation of international copyright law.
KimmoA
it wasn't the on-board VGA messing up. Exact same problem with it.
I also tried installing FreeBSD on my other computer and transfer over
the HDD. It installed correctly, but wouldn't boot on the server, due
to some mount error. Not even the "set ata.ata_dma=0" thing worked
(which I usually have to do for my boxes with FreeBSD). :(
I also burned a 7.x (bleeding edge current) CD to see if it had been
fixed. Same problem.
I eventually gave up, having tried everything within my power (I don't
own a COM cable (or whatever they are called), so I couldn't test the
terminal redirection thing), and installed Debian Linux via floppies.
It went without problems.
Now, I'm used to FreeBSD and I love(d) it. I've used it since 4.7. I
don't feel good about basically being forced to switch to Linux, but I
must say that the package handeling in this OS is far superior to the
FreeBSD ports... so far. I might be wrong, but that's my initial
thoughts.
jpd
On 2006-06-12, KimmoA <kim...@gmail.com> wrote:
> I also tried installing FreeBSD on my other computer and transfer over
> the HDD. It installed correctly, but wouldn't boot on the server, due
> to some mount error. Not even the "set ata.ata_dma=0" thing worked
> (which I usually have to do for my boxes with FreeBSD). :(
It probably means /etc/fstab contains the wrong device numbers. In which
case, changing DMA settings is a fix for something that likely isn't
wrong.
> I eventually gave up, having tried everything within my power (I don't
> own a COM cable (or whatever they are called),
``Null modem cable'', and in its simplest form it's two plugs with the
tx and rx signals crossed , and signal ground connected[1]. Layouts are
found all over the 'web[2]. Prefab cables can be bought easily as well.
> Now, I'm used to FreeBSD and I love(d) it. I've used it since 4.7. I
> don't feel good about basically being forced to switch to Linux, but I
> must say that the package handeling in this OS is far superior to the
> FreeBSD ports... so far. I might be wrong, but that's my initial
> thoughts.
Fine, whatever. I do feel you've given up too early, mainly because you
aren't thinking about this very clearly. Maybe this really is a fuckup
in FreeBSD, and maybe not. But until someone reports it, it can't be
fixed. You can report it[3], but choose not to do so. Which is a pity.
[1] Yes, that's just three wires, and for short lengths a bit of telephone
cable will do just fine.
[2] Lucky on google: http://www.loop-back.com/null-mod.html
[3] I told you to ask about it on a freebsd mailinglist; there are many
more FreeBSD developers there. Another way is to write a Problem
Report (PR). There's a writeup on the FreeBSD site on how to write
them.
Chronos
18:35, the following appeared on stdout:
> [1] Yes, that's just three wires, and for short lengths a bit of
> [telephone cable will do just fine.
*Telephone* cable [1]? Just leave your geek card with the doorman, Mr.
Jpd. All serial cables, whether they be UPS, null modem, for the EPROM
burner or otherwise shall only be made from the best quality, low-loss
category 5 cable - the solid stuff, mind you, so you have to get the
soldering iron out every other time you use them - and never a D shell
nor a piece of neoprene sleeving shall be sighted. Pure hardware
prowess and peer admiration will keep those lines from shorting.
Believe...
[1] Actually, the 3 pair stuff is just one conductor short for a full
null-modem cable. They need seven:
2 -> 3
3 -> 2
5 -> 5 *stop here if you're not using handshaking
4 -> 6
6 -> 4
7 -> 8
8 -> 7
The above is for DB9F to DB9F. From memory. I have *no* life :-)
--
Chronos
jpd
On 2006-06-12, Chronos <chr...@chronos.org.uk> wrote:
> After replacing jpd with a small shell script on Monday 12 Jun 2006
> 18:35, the following appeared on stdout:
>> [1] Yes, that's just three wires, and for short lengths a bit of
>> [telephone cable will do just fine.
>
> *Telephone* cable [1]? Just leave your geek card with the doorman, Mr.
> Jpd. All serial cables, whether they be UPS, null modem, for the EPROM
> burner or otherwise shall only be made from the best quality, low-loss
> category 5 cable - the solid stuff, mind you, so you have to get the
> soldering iron out every other time you use them - and never a D shell
> nor a piece of neoprene sleeving shall be sighted. Pure hardware
> prowess and peer admiration will keep those lines from shorting.
> Believe...
*handwave* This is not the short you're looking for! Splendid. Now,
shall we up the ante and do the same with, oh, some 380V triphase will
do, for starters?
I did plan on mentioning CAT5, then went on editing and promptly forgot
about it. Solid stuff is nice but won't stand up against movement as
well as stranded, so is not always practical. And, of course, the
simpler peecee variety doesn't use differential signals so twisting
the pairs is all but useless towards increasing cable quality for this
particular purpose. But I am now forced to ask, what's with this unfair
bias against telco geeking?
> [1] Actually, the 3 pair stuff is just one conductor short for a full
> null-modem cable. They need seven:
Most often I've seen two pair stuff[1] with an extra nekkid ground wire
thrown in. Some creative combining of signals and it'll do as a serial.
> 2 -> 3
> 3 -> 2
> 5 -> 5 *stop here if you're not using handshaking
See? Three wires. :-p
> 4 -> 6
> 6 -> 4
> 7 -> 8
> 8 -> 7
>
> The above is for DB9F to DB9F. From memory. I have *no* life :-)
I initially thought ``2,3,7, wasn't it?'', but that's for DB25.
And <ObNitPick> DB9 isn't. </> Reasons why are left as an excercise.
[1] Coloured red, blue, white, orange. This indeed is Dutch telephone
cable I'm talking about. None of that ugly red blue yellow black!
Chronos
>>> [1] Yes, that's just three wires, and for short lengths a bit of
>>> [telephone cable will do just fine.
>>
>> Telephone cable [1]? Just leave your geek card with the doorman, Mr.
>> Jpd. All serial cables, whether they be UPS, null modem, for the
>> EPROM burner or otherwise shall only be made from the best quality,
>> low-loss category 5 cable - the solid stuff, mind you, so you have to
>> get the soldering iron out every other time you use them - and never
>> a D shell nor a piece of neoprene sleeving shall be sighted. Pure
>> hardware prowess and peer admiration will keep those lines from
>> shorting. Believe...
>
> handwave This is not the short you're looking for! Splendid. Now,
> shall we up the ante and do the same with, oh, some 380V triphase will
> do, for starters?
Of course. What self-respecting BOfH hasn't got a cat5 to three phase
adaptor?
> I did plan on mentioning CAT5, then went on editing and promptly
> forgot about it. Solid stuff is nice but won't stand up against
> movement as well as stranded, so is not always practical.
But that is the whole point! It's no fun if it's reliable...
> And, of
> course, the simpler peecee variety doesn't use differential signals so
> twisting the pairs is all but useless towards increasing cable quality
> for this particular purpose. But I am now forced to ask, what's with
> this unfair bias against telco geeking?
I actually started in the telecomms industry. I can probably still
rattle off sched 0 to 50 pairs without a lookup table when faced with a
rat's nest, but I got very tired of crawling through floor-space of
factories that had never been cleaned, much less aired, since the
places were built. I could also tell you a story about me, a cable, a
large warehouse, a fork-lift truck and a very old pallet, but then I'd
have to go and take a sedative. I still get vertigo just *thinking*
about it...
--
Chronos
KimmoA
My Debian venture seems over. The stupid shit didn't even carry PHP
5... which makes it useless in my eyes. I really didn't expect that. It
had some nice things I would've liked, though...
Alright, so... FreeBSD. Same problem exists still. I don't know what to
do, except perhaps ordering one of those cables... and then it MIGHT
work.
Why must these things always happen to ME?! >:@
Chronos
22:12, the following appeared on stdout:
> Alright, so... FreeBSD. Same problem exists still. I don't know what
> to do, except perhaps ordering one of those cables... and then it
> MIGHT work.
I think jpd covered your options pretty well in his post earlier:
news:4f3kdkF...@individual.net
A null modem cable is probably the best chance you have of getting to
the bottom of this. Even if you can only get a dmesg output to a
console, you'll be a step further into understanding what is going on.
I would also remove the Raptor for now and use whatever IDE drive you
have kicking around spare just to test. The mount error was probably
due to the fstab being set up for the controller assignments on the
machine you used to set the disk up, which are not necessarily going to
be the same as the server board. At least with a bog-standard IDE
drive, you've a pretty good chance of both machines wanting to use
ad0s1a as the root partition. Once you get a system that you can at
least see the dmesg on, you can tweak the installation on the Raptor to
suit, but you've got to get the thing booted first.
--
Chronos
jpd
> a fork-lift truck and a very old pallet, but then I'd have to go and
> take a sedative. I still get vertigo just *thinking* about it...
I feel your pain. For some reason, I did a bit of standing on 4 metre
high ladders in highschool. That ladder looked like a capital lambda,
with an effectively invisible (steel) holder cable at ankle height.
There were also lots of other schoolkids being less than superstitious
enough, far too often. I have been tempted to drop the pulleys and other
heavy stuff (stage lighting and so on) involved, though.
A bit later, I had a job that also involved standing on comparably
high ladders, tending to some dancing's innards. The interesting
features there weren't so much the random unrelated surprises, or the
uncharted cabling, some right in the open and still carrying 220V, as
the occasional electrification of the overhead cabling duct itself. I
hear they finally overhauled the wiring with the specific objective of
eliminating random electrification.
Still, in some ways that job was much better than the systems
administration that followed. And not merely because, o surprise, that
also involved things like ripping at least half the cables out of the
racks, and still being left with a /kabelsalat/ worthy of some cringe
photo contest.
jpd
> do, except perhaps ordering one of those cables... and then it MIGHT
> work.
It also might work already; worse still, the serial cable idea might
land you into the exact same problem as you had before, only now over
serial. Then again, it might not, too.
However, if my earlier wild guess happens to be true, and you figure out
how to tell it where to find its root so then it will boot, so you can
fix up /etc/fstab. If it won't boot multi-user, it often can boot into
single-user mode. That is a step forward, as that gives you the tools
needed to fix it and make it boot multi-user again.
> Why must these things always happen to ME?! >:@
They don't happen to just you. In fact, lots of stuff has happened
to many systems people in the past. Yes, bad things have happened to
systems under my care, including things like the one un-backuppable
fileserver that the whole company depended on suddenly deciding to
shuffle scsi busses around or just plain not want to boot because a
certain scsi card was inserted in the wrong slot. And all of this on
``standard peecee hardware''. So I do feel your pain. However, that
doesn't help you get stuff fixed.
So, back to FreeBSD. It is highly fixable, but you might want to get
on with making a serious attempt at fixing your particular problems.
Remember; lots of stuff is documented, lots of people are willing
to help, and some of them even know lots of nitty details about the
installer and the bootprocess, and if there's a real bug people really
want to fix it. But you have to find the right place to ask, and to
provide them with less of your frustration and more facts on what goes
wrong; things like dmesg output, literal transcripts of things done as
far as possible, serial captures of the boot process, and so on.
Chronos
> Still, in some ways that job was much better than the systems
> administration that followed. And not merely because, o surprise, that
> also involved things like ripping at least half the cables out of the
> racks, and still being left with a kabelsalat worthy of some cringe
> photo contest.
Yes indeed. I still find myself standing in a comms room with a Krone
tool in one hand, half a cable tester [1] in the other, a few patch
panels without any fixings, hanging there simply because the huge
bundle of mixed brand and specification cable behind them has formed
into a solid mass which would take herculean effort to move and is
probably worth a grand in scrap copper value alone, with labels written
in what looks like a cross between traditional Chinese and Arabic and
no idea whether whoever last worked on the cabinet was a genius or a
madman. Sometimes I wonder just what I have done to deserve such a
fate.
Of course, things get steadily worse from thereon in, since as soon as
you touch one of the panels, three other faults develop almost
immediately, one of them usually affecting someone who thinks they're
important enough to be immune from a sysadmin with a headache and
probably has a wonderful singing voice if only they could control the
volume.
The best thing about telecomms, IMO, was the fact that you got to go out
of the office environment for a while. Now, I admit, I worked in that
industry before cellphones became ubiquitous, so I may be a bit
blinkered to the modern realities of some half-wit ringing you every
five minutes, but back then you were your own boss. A packet of Silk
Cut, a flask, a sandwich which may or may not have been edible when it
was made and a nice drive in the country before you got your hands
dirty. Bliss!
[1] Thank God I don't have to lug the TDR around any more. Remember
twinax? I thought I'd found my cache of N type plugs the other day.
Happily sat at the workbench, I prepped my RG213 and opened the little
plastic bag... I lost maybe an hour of my life as I blacked out in pure
terror. I think I'm going soft.
--
Chronos
KimmoA
cable thing) -- I pulled the S-ATA cable and disabled it in BIOS. Same
error.
Since a chassis fan I ordered together with a chassis and a PSU seems
to be out of stock, I might be able to tell that store to send it later
together with the precious console cable (yes, I cannot decide what to
call it!). Assuming they keep those cables...
Somehow, I strongly suspect that the output in the end of the cable
will be the same as on my screen.
This is the most expensive motherboard I have ever bought, being all
made for servers and stuff. I am totally broke now, and the ironic part
is that I depend on this machine for my new commerical project(s). My
old server simply doesn't have the processing power (800 MHz VIA C3)
for them. This is why this is more than just an annoyance and me being
impatient...
dfeu...@mindspring.com
> Alright... I'm really worried now, and I really don't wanna run Linux
> or any such crap. I have come to love FreeBSD and don't want to switch
> to another Unix just because of some stupid hardware issue.
It would be interesting to see if a OpenBSD 3.9 installation worked any
better. If you see the same video problems then you know you have bad
hardware.
--
Using OpenBSD with or without X & KDE?
See Dave's OpenBSD | X | KDE corner at
http://dfeustel.home.mindspring.com !!!
KimmoA
> better. If you see the same video problems then you know you have bad
> hardware.
Bad hardware which runs Debian Linux and FreeBSD 4.11 perfectly well? :S
dfeu...@mindspring.com
KimmoA
Bill Vermillion
Chronos <chr...@chronos.org.uk> wrote:
>After replacing jpd with a small shell script on Monday 12 Jun 2006
>18:35, the following appeared on stdout:
>> [1] Yes, that's just three wires, and for short lengths a bit of
>> [telephone cable will do just fine.
>*Telephone* cable [1]? Just leave your geek card with the doorman, Mr.
>Jpd. All serial cables, whether they be UPS, null modem, for the EPROM
>burner or otherwise shall only be made from the best quality, low-loss
>category 5 cable
Please note that JPD said 'short length'. And in that case almost
any wire will work. I speak from years of experience with all
sorts of interconnects.
> - the solid stuff, mind you, so you have to get the
>soldering iron out every other time you use them - and never a D shell
>nor a piece of neoprene sleeving shall be sighted. Pure hardware
>prowess and peer admiration will keep those lines from shorting.
>Believe...
>[1] Actually, the 3 pair stuff is just one conductor short for a full
>null-modem cable. They need seven:
>2 -> 3
>3 -> 2
>5 -> 5 *stop here if you're not using handshaking
>4 -> 6
>6 -> 4
>7 -> 8
>8 -> 7
>
>The above is for DB9F to DB9F. From memory. I have *no* life :-)
You only need 3 SINGLE wires NOT 3 PAIR. tx, rx and signal
ground.
Bill
--
Bill Vermillion - bv @ wjv . com
Bill Vermillion
I don't know the model number of the mobo on one of the new servers
I'm building up [doing it remotely].
But the Raptor comes in as ad4s1[aed] with the onboard SATA
adaptor. I'm getting over 55MB/second read/write on this drive.
It's an impressive little drive for a decent price.
This is with 6.1 So your supposition about the fstab entry is
probably correct.
Steve Ackman
> My Debian venture seems over. The stupid shit didn't even carry PHP
> 5... which makes it useless in my eyes. I really didn't expect that. It
> had some nice things I would've liked, though...
If you use older versions like Sarge (or even Woody)
then you're correct.
Both Etch and Sid ship with php 5.1.2 however.
sam
out of standard. My suggestion is get rid of them, buy something else.
S
Chronos
00:56, the following appeared on stdout:
> (yes, I cannot decide what to
> call it!)
Its proper name, as has been said many times, is a null modem cable.
Another, later, MSese name for them is a laplink cable, although that
has been out of use for ages. However, there are three types, ignoring
the difference between hardware handshaking. For modern PCs, you need a
DB9F to DB9F null modem cable. It's called a null modem cable because
it connects two bits of DTE (data terminal equipment) together without
using the normal two bits of DCE (data communications equipment, in
other words a modem) between them, hence no modem or null modem.
Honestly, nobody is pulling your leg; humour is all well and good until
it starts inconveniencing someone. If you lived in the UK I'd make you
one and send it to you, but shipping to Scandinavia is probably more
than 2 DB9Fs, two hoods and a bit of cable is really worth.
> Somehow, I strongly suspect that the output in the end of the cable
> will be the same as on my screen.
It can't. Serial terminal emulators are just that: serial, displaying
one character at a time. It can't overwrite data already displayed
without clearing what it is overwriting. It simply displays raw ASCII
coming in over a serial line. A dumb terminal does exactly the same
thing [1]. PC graphics are a totally different idea.
Be sure to match the serial port settings on both ends, though. I
*think* the defaults are 9600:8N1 (9600 baud, 8 data bits, no parity,
one stop bit) after a quick glance at ttys on this box, but ICBW. The
handbook will be the authoritative source for this. You may also need
to hit the return key a few times to get the TTY to respond.
[1] Which reminds me, I have an old dumb terminal in the loft somewhere.
I must dig it out. I never thought it would be useful again - until
now.
--
Chronos
Chronos
Jun 2006 03:15, the following appeared on stdout:
> Please note that JPD said 'short length'. And in that case almost
> any wire will work. I speak from years of experience with all
> sorts of interconnects.
Bill, it was a joke. Apologies for my warped sense of humour. Jpd got it
immediately, I think...
--
Chronos
KimmoA
I was just looking in the BIOS settings, very carefully and several
times, and there is no way to turn off the RAID part. However, after
reading the manual again, I found a jumper/switch which turns off the
RAID part. Worringly enough, it was physically located not near the
four RAID S-ATA ports, but near the two normal S-ATA ones, leading one
to believe that it would shut down S-ATA entirely...
Luckily enough, that was not the case, and guess what? IT WORKS! IT
WOOORKS! :-)
Without the RAID switch on (default), there is no post POST spam
("Adaptec embedded RAID BIOS blablabla"), and then the FreeBSD
installer isn't confused and loads up properly!
It loaded the installer perfectly fine. I'm very relieved. (Or... I was
for two seconds, until I got attacked by a monsterous spider. Having an
extreme phobia for those creatures put back my happiness level to
neutral (which is still better than normally!).)
However, there is now a new minor problem. It (the FreeBSD installer)
shows two disks for some reason, calling them "ad2" and "ar0". I only
have one physical HDD, so what's up with that? Even if this doesn't
matter, I'd like to think why it's doing that.
Chronos
15:43, the following appeared on stdout:
> It loaded the installer perfectly fine. I'm very relieved. (Or... I
> was for two seconds, until I got attacked by a monsterous spider.
> Having an extreme phobia for those creatures put back my happiness
> level to neutral (which is still better than normally!).)
:-) Yes, me too, and if I drop something in the workshop, guess where it
inevitably ends up? Yes, in the darkest, most spider infested corner of
the workshop ever.
> However, there is now a new minor problem. It (the FreeBSD installer)
> shows two disks for some reason, calling them "ad2" and "ar0". I only
> have one physical HDD, so what's up with that? Even if this doesn't
> matter, I'd like to think why it's doing that.
You've created a JBOD in the on-board RAID BIOS with the one disk as a
member. FreeBSD has read this metadata from the boot sector and created
the same array with its own ata(4) driver, giving it the name ar0.
code:
atacontrol destroy ar0
will remove the metadata and leave you with ad2 only. MAKE SURE
your /etc/fstab is using ad2 and not ar0 BEFORE you issue this command.
--
Chronos
Chronos
15:50, the following appeared on stdout:
> atacontrol destroy ar0
Sorry, apologies, my mistake!
atacontrol delete ar0
I was still in ifconfig mode. I've been messing with my IPv6 today.
--
Chronos
KimmoA
> inevitably ends up? Yes, in the darkest, most spider infested corner of
> the workshop ever.
I don't understand where the hell they come from... Do they squeeze in
from tiny cracks in the walls? I hate them, and you're never freakin'
safe, no matter where you are. :|
> You've created a JBOD in the on-board RAID BIOS with the one disk as a
> member. FreeBSD has read this metadata from the boot sector and created
> the same array with its own ata(4) driver, giving it the name ar0.
Hmm... I have? Hrm...
> code:
>
> atacontrol destroy ar0
>
> will remove the metadata and leave you with ad2 only. MAKE SURE
> your /etc/fstab is using ad2 and not ar0 BEFORE you issue this command.
Can this really be run from the primitive "command prompt" you can
select in the boot menu? Because as you (hopefully and probably) know,
there is no FreeBSD on the HDD as it is!
Or perhaps I should just get some kind of total disk clearing bootable
floppy and clean the HDD before? Any tips on that, BTW?
PS: Sheesh... I keep finding embarrasing typos and incorrect grammar in
my posts. I should really start proof-reading these messages...
Chronos
> Can this really be run from the primitive "command prompt" you can
> select in the boot menu? Because as you (hopefully and probably) know,
> there is no FreeBSD on the HDD as it is!
If your BIOS was set up to use the SATA controller as part of a software
RAID set (usually something like SATA controller mode RAID/IDE in
integrated peripherals), it may have done it automatically.
Anyway, nope, it has to be run on a set up system. Sorry, I hadn't
realised you hadn't set up yet. Of course, you can install to ad2 and
ignore ar0. ad2 will still appear in the boot sequence. ar0 is just a
pseudo-RAID device. You can safely ignore it and nuke it after you have
set up.
> Or perhaps I should just get some kind of total disk clearing bootable
> floppy and clean the HDD before? Any tips on that, BTW?
DBAN, every time. It's on the UBCD (v 3.4) or go to
http://dban.sourceforge.net/ and grab the floppy version.
--
Chronos
jpd
On 2006-06-13, Chronos <chr...@chronos.org.uk> wrote:
> Its proper name, as has been said many times, is a null modem cable.
> Another, later, MSese name for them is a laplink cable, although that
> has been out of use for ages.
<ObNitPick> Actually, the latter are ment to interconnect peecees by
blatantly abusing their parallel ports. </> I did find a conspiciously
yellow DB25M<->DB25M cable, still sealed in grubby plastic, in a grab
box somewhere, took it home, and lo and behold, I could run PLIP over
it. Runs at a fairly nice 40kByte/sec, even if it does eat cpu like
there's no tomorrow.
> Be sure to match the serial port settings on both ends, though. I
> *think* the defaults are 9600:8N1 (9600 baud, 8 data bits, no parity,
> one stop bit) after a quick glance at ttys on this box, but ICBW.
They usually are, until you hit on some crap applicance that defaults
to, say, 38400, or maybe 115200, and that almost naturally implies it
doesn't expect a glass tty but at least a vt100 (or worse, something
that understands ``pc ansi codes''), and doesn't at all deal well with
connecting to it (through a serial console server, perhaps?) well after
it has booted up and painted its screen once. Repainting the screen?
wassat? A key for that? The very thought!
FreeBSD defaults to 9600 8n1, or at least it did last time I needed it.
It is possible to change that default, be it ever so inadvisable.
To everybody who's reading and now wonders why I say that's inadvisable:
The reasons are not so much technical, as conventional. Do remember that
while 9600baud is *slow*, it is also quite fast enough for booting,
running fsck, and bringing up an ethernet interface with, it works,
and it is still the de-facto standard in serial consoles just about
everywhere. That makes it very convenient to stick to 9600 baud and not
have to work out the actual speed by trial and error whenever you really
really need at least something to just work. Such occasions do happen.
This is also why I really would prefer laptops to have a hardware serial
without having to resort to usb converters or other things that in this
context definately count as ``fancy stuff''.
Chronos
16:12, the following appeared on stdout:
> <ObNitPick> Actually, the latter are ment to interconnect peecees by
> blatantly abusing their parallel ports. </> I did find a conspiciously
> yellow DB25M<->DB25M cable, still sealed in grubby plastic, in a grab
> box somewhere, took it home, and lo and behold, I could run PLIP over
> it. Runs at a fairly nice 40kByte/sec, even if it does eat cpu like
> there's no tomorrow.
Dammit, yes, I remember those things. Now that you mention it, it
probably was parallel and yellow the last time I saw a "laplink" cable.
I also have a null modem cable with both a DB25F and a DB9F on both
ends, which neatly covers all eventualities - except <insert expletive
here> Cisco switches...
--
Chronos
jpd
> 23:46, the following appeared on stdout:
>> Still, in some ways that job was much better than the systems
>
> Yes indeed. I still find myself standing in a comms room with a Krone
> tool in one hand, half a cable tester [1] in the other, a few patch
> panels without any fixings, hanging there simply because the huge
> bundle of mixed brand and specification cable behind them has formed
> into a solid mass which would take herculean effort to move and is
> probably worth a grand in scrap copper value alone, with labels written
> in what looks like a cross between traditional Chinese and Arabic and
> no idea whether whoever last worked on the cabinet was a genius or a
> madman. Sometimes I wonder just what I have done to deserve such a
> fate.
Administrative genius, in my opinion, is setting it up such that people
afterward don't have trouble extending it the right way, but do have
trouble fscking up a good setup.
I wouldn't mind ripping out everything and do it again, or better yet,
do it from scratch. Altough while I wouldn't mind doing the terminating
and testing myself, if only to get the labeling right, I'd let someone
else pull out and put in cable. That's how I did it last time, and
despite CFO delusions of wireless grandeur[2], it worked well.
> Of course, things get steadily worse from thereon in, since as soon as
> you touch one of the panels, three other faults develop almost
> immediately, one of them usually affecting someone who thinks they're
> important enough to be immune from a sysadmin with a headache and
> probably has a wonderful singing voice if only they could control the
> volume.
You've got cable, and if not a sturdy tree, maybe a usable lamp post? I
found that if pressed enough I'm perfectly willing to kick VPs out of
the machine room while making loudly (it's *loud* in machine rooms, no?)
and perfectly clear they're not helping getting the shit mopped up and
sorted again. Of course, VPs should be wiser than going there, but hey.
> The best thing about telecomms, IMO, was the fact that you got to go out
> of the office environment for a while. Now, I admit, I worked in that
> industry before cellphones became ubiquitous, so I may be a bit
> blinkered to the modern realities of some half-wit ringing you every
> five minutes, but back then you were your own boss. A packet of Silk
> Cut, a flask, a sandwich which may or may not have been edible when it
> was made and a nice drive in the country before you got your hands
> dirty. Bliss!
Never had the pleasure, but I can see why you'd put up with a lot just
for that. Then again, even on-site one develops a bag'o'tricks to be
unavailable for up close and personal annoying as much as possible.
> [1] Thank God I don't have to lug the TDR around any more. Remember
> twinax?
No, and I'm not sorry for that. I'm not that old and the worst so far
was the students' dorm with the 250M+ cheapernet cable, part of which
was directly buried[1] between two buildings, each on different phases
or even street taps, and it wasn't grounded at either end. Oh, and of
course every single connector, whether crimped, soldered or screwed,
was rotten, with shielding sticking out enough to pinch fingers and
people walking by causing enough of a disturbance for signal loss. It
carried enough current to be seriously painful. I wonder how my poor
old 386 even survived being hooked to it. It did, but crashed anytime I
tried high-volume datatransfers. The card I'd used, a de205 I think, did
provide impressive distance-to-fault messages, though.
I later used the crimping tool I'd filched when they moved to 10baseT
to build a video distribution network elsewhere. With enough care it
*is* possible to make good crimps, and only using point-to-point links
(from/to a video amp) also helps.
> I thought I'd found my cache of N type plugs the other day.
> Happily sat at the workbench, I prepped my RG213 and opened the little
> plastic bag... I lost maybe an hour of my life as I blacked out in pure
> terror. I think I'm going soft.
Nah, just not used to that particular horror, anymore. Which isn't
altogether bad.
[1] No, the cable wasn't specced for that. Why do you ask?
[2] Wanted to go all wireless, momentarily forgetting that even if he'd
not shy away from funding adding the DECT extentions to the switch[4],
he'd still need some way to power the N laptops and the N*M other
gadgets with their numerous wallwarts, all incompatible[3].
[3] This particular company had a bit many of them. Think 3+ per executive
on that particular floor, and a fairly fast paced stream of replacements.
[4] I didn't need to ask the vendor for an offer to know he would do that.
Chronos
17:00, the following appeared on stdout:
> You've got cable, and if not a sturdy tree, maybe a usable lamp post?
You've just answered one of my "can't get my head around it" mysteries.
I always wondered why Belkin ever made those extra flexible,
easy-to-fsck-up-with-high-heels-or-office-chair silicone patch cables
that go O/C if you breathe near them, apart from a temporary increase
in sales until people realise they're crap. Now I think they must have
been designed by someone who had a run-in with an irate sysadmin,
specially made to stretch and snap before a VP's neck.
I'll bet they did all the testing in-house, too. Selfish swine...
> [2] Wanted to go all wireless, momentarily forgetting that even if
> he'd not shy away from funding adding the DECT extentions to the
> switch[4], he'd still need some way to power the N laptops and the N*M
> other gadgets with their numerous wallwarts, all incompatible[3].
...not to mention the cost of replacing said gadgets when they
got "accidentally" plugged into an incompatible power supply, which is
what I think you're hinting at in [3]. Toshibas spring instantly to
mind; the real ones use (or used to) 15VDC, the Compal rebadged stuff
uses 19VDC but both supplies have a Tosh logo on them and the same
damned plug. The glowing red LED of doom on a "real" Tosh signifies a
motherboard replacement; repair people these days rarely fault-find to
component level outside the factory.
Mind you, there are some perks to this approach, such as a near
inexhaustible supply of easily repaired (if you have a hot-air rework
station) one-year-old laptops and brand-new 2.5" HDDs. It is human
nature to plug the wrong thing into the wrong adaptor one day out of
warranty, as you probably know yourself.
--
Chronos
jpd
On 2006-06-15, Chronos <chr...@chronos.org.uk> wrote:
> I always wondered why Belkin ever made those extra flexible,
> easy-to-fsck-up-with-high-heels-or-office-chair silicone patch cables
> that go O/C if you breathe near them, apart from a temporary increase
> in sales until people realise they're crap. Now I think they must have
> been designed by someone who had a run-in with an irate sysadmin,
> specially made to stretch and snap before a VP's neck.
>
> I'll bet they did all the testing in-house, too. Selfish swine...
Isn't that just a misguided and hopefully, miserably failed, attempt at
innovating by moving to leverage synergy and resurrecting silver satin
for ISDN to the modern VoIP environment?[1] I bet the safety issues were
an un-noticed benefit^Wdrawback. Ever seen a VP with a clip-on tie?
> ...not to mention the cost of replacing said gadgets when they
> got "accidentally" plugged into an incompatible power supply, which is
> what I think you're hinting at in [3].
Apart from re-ordering stolen laptops (we used Thinkpad T series[0]) or
sending broken laptops off to warranty repair, my problems stopped at
the outlet. The smaller things and gadgets simply were SEP.
The thing was that I had *at minimum* five power outlets per person
(laptop, desk lamp, ...) and it'd not always be enough. And that was
just the laptop-using suits. In comparison, there was one network
cable[2] and one phone cable per person. But the CFO said we must reduce
cabling so therefore we must go all wireless. I wasn't the only one who
did his best to ignore him in, uhm, any non-finance matters; at least
all the rest of the board did that also.
Not to mention that going all wireless plays havoc with any ability
to resurrect broken machines, make emergency backups, and so on. Just
consider that, for example, wxp doesn't do wpa until sp2, so a fresh
install cannot hook up to get that even if you happen to have the
proper driver cd for your wifi NIC. I ran into that when my dad moved,
courtesy some local FOAF windows box shifter who was terminally opposed
to the notion of pulling a cable from the ADSL demarc to one floor
up, same side of the house, even if someone else offers to do it.
Instead he replaced the wired NICs with wireless ones and introduced a
EUR 130 ``wireless router'', got it to work on dad's gf's wxp, failed
to get it to work on dad's w95, and took off. When I was there, he'd be
unavailable to tell us the proper password for the wpa setup, which did
my progress with replacing dad's w95 installation no good. If anything
crashes, well, I left a sheet with a short writeup of what I'd done and
all passwords that had cost a day and a lot of phone calls to retrieve,
and sp2 on a cd.
Yep, wireless is easy. Thanks so much for that, people.
> Mind you, there are some perks to this approach, such as a near
> inexhaustible supply of easily repaired (if you have a hot-air rework
> station) one-year-old laptops and brand-new 2.5" HDDs. It is human
> nature to plug the wrong thing into the wrong adaptor one day out of
> warranty, as you probably know yourself.
Oh, hum. Maybe I should've moved to toshibas then. But then, we'd have
toshibas. Dunno if that'd offset having 2.5" HDDs.
[0] I think those are pretty nice, mainly because of their keyboards. The
rest is micros~1- and suit-centric as usual. *sigh* I really should get
on with my little project of writing up what I really want in a laptop
fit for unix use. Any laptop R&D depts hiring?
[1] I've been reading a book lately. And this one explains how management
*should* work, but usually doesn't. It is taking me ages to really
understand what it says, but once you do, it is actually easy to
explain why management sucks so much. So, to sum it up and in case
you haven't noticed, the paragraph referring to this footnote is a
joke. An awfully bad joke, but that is the point.
[2] Actually, there, one 8-port switch on a patchable from a wall outlet
per pair of tables[4], but only two or maybe three ports used per switch.
They weren't entirely without problem, but that's a different issue.
[4] The switches came with magnets, and they were easily stuck on the
underside of a table or on some metal cabinet or other. No big deal
clutter wise. Still, I'd rather done without but was foiled by the
layout of the floor.
Chronos
23:38, the following appeared on stdout:
> Isn't that just a misguided and hopefully, miserably failed, attempt
> at innovating by moving to leverage synergy and resurrecting silver
> satin for ISDN to the modern VoIP environment?[1] I bet the safety
> issues were an un-noticed benefit^Wdrawback. Ever seen a VP with a
> clip-on tie?
ARGH! /me runs away from the weasel-speak screaming...
I try not to look at VPs' ties. They're usually something cartoony that
is meant to portray the fact that they haven't had a humour bypass,
which is a lie in most cases.
> Apart from re-ordering stolen laptops (we used Thinkpad T series[0])
> or sending broken laptops off to warranty repair, my problems stopped
> at the outlet. The smaller things and gadgets simply were SEP.
Data protection means nothing to these people. I find myself worrying
more about where laptops and PDAs are than that they're working.
> The thing was that I had *at minimum* five power outlets per person
> (laptop, desk lamp, ...) and it'd not always be enough. And that was
> just the laptop-using suits. In comparison, there was one network
> cable[2] and one phone cable per person. But the CFO said we must
> reduce cabling so therefore we must go all wireless. I wasn't the only
> one who did his best to ignore him in, uhm, any non-finance matters;
> at least all the rest of the board did that also.
Wise. Sadly, the IT rags don't help, extolling the virtues of wireless.
The reality of wireless is somewhat different. Rogue APs, security
issues, RF screening, AP placement, channel overlap etc. And that's
just the planning phase.
> Not to mention that going all wireless plays havoc with any ability
> to resurrect broken machines, make emergency backups, and so on. Just
> consider that, for example, wxp doesn't do wpa until sp2, so a fresh
> install cannot hook up to get that even if you happen to have the
> proper driver cd for your wifi NIC.
Have you ever seen a wireless NIC that can handle PXE boot? I haven't,
and I rely on it for reloads and diagnostics. All these companies
buying into the wireless dream are going to have a shock if they ever
want to run thin clients.
> Yep, wireless is easy. Thanks so much for that, people.
Or it would be if everyone used BSD. Oh, sorry, I promised myself I
wouldn't do that this year. It sounds too COLA to be constructive...
> Oh, hum. Maybe I should've moved to toshibas then. But then, we'd have
> toshibas. Dunno if that'd offset having 2.5" HDDs.
Well, the reference to 2.5" HDDs was that they generally want their data
back, so the easiest restore method, given an identical laptop, is to
swap the HDD out with the old one. The higher voltage PSU rarely gets
past the zener on the DC input board, so the HDD is usually fine.
> [0] I think those are pretty nice, mainly because of their keyboards.
> [The
> rest is micros~1- and suit-centric as usual. sigh I really should get
> on with my little project of writing up what I really want in a laptop
> fit for unix use. Any laptop R&D depts hiring?
HPs (nx6126 Turion based) are nice little machines apart from the
graphics adaptor (Radeon Xpress 200, awful thing) and the new glossy
screen. The Synaptics touchpad actually works properly with moused,
too.
> [1] I've been reading a book lately. And this one explains how
> [management
> should work, but usually doesn't. It is taking me ages to really
> understand what it says, but once you do, it is actually easy to
> explain why management sucks so much. So, to sum it up and in case
> you haven't noticed, the paragraph referring to this footnote is a
> joke. An awfully bad joke, but that is the point.
Yes, I got the joke. I had to attend a management seminar once (never
again) which explained X managers and Y managers and such. Odd seminar.
Personally, I'm of the opinion that there are simply far too many
layers of management, in many cases with inverse relationship to
intelligence...
> [4] The switches came with magnets, and they were easily stuck on the
> underside of a table or on some metal cabinet or other. No big deal
> clutter wise. Still, I'd rather done without but was foiled by the
> layout of the floor.
Let's hope those switches don't keep corrupting their MAC tables every
five minutes like some of the cheaper ones do. One or two machines
dropping off a segment with no warning takes a lot of diagnosing when
you've never come across duff switches before. I've been there...
--
Chronos
Bill Vermillion
And about 5 years ago when the little Intel 1000?? 1RU servers came
out we put a couple in our racks. The HW defaults to 19200 on
those. They also redirect the BIOS out the serial port so if you
have a remote power switch and a serial connection into the machine
you can change the BIOS settings, before it boots. Then the serial
port switches to the OS.
First time around you have to come in a 19,200 and then switch
to the 9600 when the FreeBSD boots up.
I've not seen anything that defaults to higher than that. Though
I do recall when updating the IOS in Cisco routers, it works best
when using the supplied software that you login at the default
9600, set things up for 115,000, and then reset your program on
the PC, otherwise the transfers go really really slow.
But I do agree that leaving it at the default in the OS is best.
>This is also why I really would prefer laptops to have a hardware
>serial without having to resort to usb converters or other things
>that in this context definately count as ``fancy stuff''.
Agree.
jpd
On 2006-06-18, Bill Vermillion <b...@wjv.com> wrote:
> And about 5 years ago when the little Intel 1000?? 1RU servers came
> out we put a couple in our racks. The HW defaults to 19200 on those.
Ew, yes. There's (of course) peecee BIOSes that take something else.
Had a few opterons that did that. Luckily the first-time setup was over
vga+keyboard and then it was easily fixed. You simply cannot trust
anything peeceeish to work right the first time without vga and keyboard.
> I've not seen anything that defaults to higher than that.
Some cheap-ish external ide<->scsi raid boxes. Depending on the revision
I've seen 38400 and 115200, and not changeable either. Most of the
manual is devoted to screenshots of whatever terminal emulator they used
on windows. Really great stuff, that. It conveniently forgets to tell
you the numbering order of the disks[1] and the firmware offers no way
to force-configure anything. So if you (or your predecessor) didn't put
in a hot spare and you accidentally touch the wrong disk to replace, the
whole raid is gone in an instant. I positively hate that firmware line.
> Though I do recall when updating the IOS in Cisco routers, it works
> best when using the supplied software that you login at the default
> 9600, set things up for 115,000, and then reset your program on the
> PC, otherwise the transfers go really really slow.
Or use the serial to setup an interface and use that instead. On the
cisco switches I had a vlan separated for just that purpose. Took one
separate port on the management box (could have vlan-trunked there as
well, but hadn't) but otherwise the switches in the cluster didn't need
to sacrifice ports; their IPAs were marked for use on that vlan only.
[1] And the fancier model had three stacks of disks, not one large row.
jpd
On 2006-06-17, Chronos <chr...@chronos.org.uk> wrote:
> Data protection means nothing to these people. I find myself worrying
> more about where laptops and PDAs are than that they're working.
And that is yet another reason why for some things you want some
governemental instition, despite the bureaucracy and overhead and all
that: You can write a policy there and it'll be followed. Mostly.
> Have you ever seen a wireless NIC that can handle PXE boot? I haven't,
> and I rely on it for reloads and diagnostics. All these companies
> buying into the wireless dream are going to have a shock if they ever
> want to run thin clients.
Thin clients are pretty neat, but only if you can organize your shop
well enough. Most IT decision makers can't organize their shops out of
a paper bag if their lives depended on it. It is therefore completely
logical that PXE has no place in wifi: It's be actually useful instead
of just a lot of hot air. ;-)
> HPs (nx6126 Turion based) are nice little machines apart from the
> graphics adaptor (Radeon Xpress 200, awful thing) and the new glossy
> screen. The Synaptics touchpad actually works properly with moused,
> too.
I personally like both the lack of windows keys, and the trackpoint[tm]
thing very much, and I *don't* like touchpads, at all. Amazing how
important such little details are.
> I had to attend a management seminar once (never again) which
> explained X managers and Y managers and such. Odd seminar.
Most ``management'' is still too heavy on (often enough, imagined)
details to be truely effective. Which is ironic, for management is
the art and science of bridging the gap between doing something and
achieving a greater goal with it.
> Personally, I'm of the opinion that there are simply far too many
> layers of management, in many cases with inverse relationship to
> intelligence...
I used to think that, but it might not be true. What I fully agree
with is that too many layers of _ineffective_ management does that.
> Let's hope those switches don't keep corrupting their MAC tables every
> five minutes like some of the cheaper ones do. One or two machines
> dropping off a segment with no warning takes a lot of diagnosing when
> you've never come across duff switches before. I've been there...
I haven't had much trouble with that, but then the cheap switches were
connected only to hosts and had a cisco switch for uplink, so the number
of macs they'd see was rather limited.
One factor in that is that I habitually have my DHCP tell all lease
holders to use hybrid mode: nbns, THEN broadcast, and supply a working
netbios ns with it (if the network has one). That way I don't have every
single netbios box broadcasting while the nbns works, and if it doesn't,
well, too bad. Samba usually works, though.
On the other hand, the power was... unstable, and some switches (most
of the cheap ones) responded by still showing lights unblinkingly but
refusing to forward or even see any traffic. The last batch (with the
magnets) was even worse, for they'd insist on a specific sequence of
connecting wall-wart to wall and switch[2], otherwise they'd throw a
fit just like they'd do after a power dip.
[1] I'm not too much a fan of that, not for writing. But then, this
remark alone probably makes me terribly elitist in that regard.
[2] Wall-wart to wall THEN wall-wart to switch or the reverse, I
forget. It was highly unintuitive, at any rate.