headers
Greg A. Woods | 2 Nov 2004 23:19
X-Face
Favicon

Re: halt after "halt".

[ On Sunday, October 17, 2004 at 21:05:45 (+0200), ragge <at> ludd.luth.se wrote: ]
> Subject: halt after "halt".
>
> I have been trying to find what the problem with the 8400 support is,
> but there is a very annoying problem that shows up: The machine do not
> go back to SRM after halt (or reboot) from ddb, it just hangs.
> It prints out the text "halt code 5", "pc = 0xfffff0000..." but then
> nothing happens, no P00>>> prompt.  Does anyone have any idea what
> the problem can be?

I added a "prom halt" command (and a few other things) to DDB and it
works on my as4000, though it may be the underlying call I use that's
not working for you on the 8400.  In any case see the attached diffs.

>  It is really annoying to have to go to the 
> computer room each time I testboot a kernel :-(

Does the 8400 have an RCM processor or equivalent?  Can you use it from
a serial console to reset (or properly halt) the system?

BTW, I've discovered that on the as4x00 and es45 the console code never
sees a BREAK signal on the serial console (presumably because of the
RCM's hooks onto the serial port) and that the only proper way to halt
the system (and thus drop into DDB in the first place) is to wake up the
RCM using its escape sequence so that it is in control of the serial
port and then to give it a "halt" command.

--

-- 
						Greg A. Woods
(Continue reading)

Permalink | Reply | 0 comments |
headers
Greg A. Woods | 2 Nov 2004 23:25
X-Face
Favicon

Re: is something wrong with either ccd(4) or isp4(4), maybe only on alpha?

[ On Monday, October 25, 2004 at 17:24:04 (-0400), Greg A. Woods wrote: ]
> Subject: is something wrong with either ccd(4) or isp4(4), maybe only on alpha?
>
> In any case I wish I did have a spare 2940UW to throw into the alpha in
> place of the QLogic 1020UW for a more direct comparison (the i386 one is
> actually a aic7880 on the motherboard so I can't swap it either....)

Well I was able to borrow an AHA-2940U2 to replace the ISP driver and
now I do see the expected througput to the CCD volume!

ahc0 at pci1 dev 4 function 0
ahc0: interrupting at kn300 irq 16
ahc0: aic7890/91: Ultra2 Wide Channel A, SCSI Id=7, 32/253 SCBs
scsibus1 at ahc0: 16 targets, 8 luns per target
...
scsibus1: waiting 2 seconds for devices to settle...
sd0 at scsibus1 target 0 lun 0: <COMPAQ, BD00965333, B917> SCSI3 0/direct fixed
sd0: 8678 MB, 5273 cyl, 20 head, 168 sec, 512 bytes/sect x 17773524 sectors
sd0: sync (50.0ns offset 31), 16-bit (40.000MB/s) transfers, tagged queueing
sd1 at scsibus1 target 2 lun 0: <COMPAQ, BD00965333, B917> SCSI3 0/direct fixed
sd1: 8678 MB, 5273 cyl, 20 head, 168 sec, 512 bytes/sect x 17773524 sectors
sd1: sync (50.0ns offset 31), 16-bit (40.000MB/s) transfers, tagged queueing
sd2 at scsibus1 target 4 lun 0: <COMPAQ, BD00965333, B917> SCSI3 0/direct fixed
sd2: 8678 MB, 5273 cyl, 20 head, 168 sec, 512 bytes/sect x 17773524 sectors
sd2: sync (50.0ns offset 31), 16-bit (40.000MB/s) transfers, tagged queueing
sd3 at scsibus1 target 6 lun 0: <COMPAQ, BD00965333, B917> SCSI3 0/direct fixed
sd3: 8678 MB, 5273 cyl, 20 head, 168 sec, 512 bytes/sect x 17773524 sectors
sd3: sync (50.0ns offset 31), 16-bit (40.000MB/s) transfers, tagged queueing

4x9 GB disks (read cache enabled) on alpha's ahc0 in ccd (with interleave 64):
(Continue reading)

Permalink | Reply | 0 comments |
headers
bart sikkes | 8 Nov 2004 20:19
Picon

somewhat ot: looking dec multia help

hello,

if there is a better place to ask for help then i havent found it, i
saw a mutlia owner list but that seemed dead for about a year now.
please point me to it if there is a better place.

i recently acquired a dec multia and it doesnt really seem to work. i
have done some searching and found it could be the battery. there
already was a non standard battery in there so it already ran out once
i assume.

but i also read the owners manual and counted the flashes the
diagnostic LED gives. that was 14 times, which appears to point into
the direction of a memory issue.

<quote>
For VX41 and VX42 models, first replace the cache SIMM (page 56). If
this does not fix the problem, continue with the recommended action
for VX40 models.
For VX40 models, replace memory modules (SIMMs) (page 55) until no
failure occurs. If replacing the SIMMs does not fix the problem,
replace the system board (page 71).
</quote>

based on the fact that i dont seem able to replace my processor i
assume i got a VX40 right? there is a heatsink on the processor
attached with two bolts but no lever on the side.

what would the wise thing to do now, just try a new battery? (i have
seen remarks on old part numbers and such, but living in the
(Continue reading)

Permalink | Reply | 3 comments |
headers
Michael Kukat | 8 Nov 2004 20:57

Re: somewhat ot: looking dec multia help

Hi !

On Mon, 8 Nov 2004, bart sikkes wrote:
> i recently acquired a dec multia and it doesnt really seem to work. i
> have done some searching and found it could be the battery. there
> already was a non standard battery in there so it already ran out once
> i assume.

Multias usually don't work :)

Okay, the battery should be your first go. It behaves _VERY_ strange, if your
battery if off limits. And off limits really has a meaning here. You need 4.5V.
Don't use 4 1.2V NiCd cells, you might have luck, but you might have the
problem, that fully charged 1.2V cells have about 1.3V, which sums to 5.2V,
which could again be off limits. Use 3 AA size batteries if you don't find the
strange 4.5V block. Hint: Motorola Mac clones (which again use some original
Apple board) use the same battery. So you may find spare parts in some Apple
spare parts store.

> but i also read the owners manual and counted the flashes the
> diagnostic LED gives. that was 14 times, which appears to point into
> the direction of a memory issue.

Whatever this means in detail... There are 2 kinds of Multia: Pentium-based and
Alpha-based. Due to the context (this list) i assume, you have the Alpha
version. The CPU is usually socketet. At least i thought so. Never had a really
closer look at it, bit maybe you must unscrew the Heatsink to see the lever for
the chip. But okay, maybe it's really soldered. Cache SIMM? This must be the
Pentium version. Never saw a cache SIMM in my Multias.
Memory... lives long. Try just 2 instead of 4 modules, or try different
(Continue reading)

Permalink | Reply | 0 comments |
headers
Greg A. Woods | 8 Nov 2004 21:16
X-Face
Favicon

Re: somewhat ot: looking dec multia help

[ On Monday, November 8, 2004 at 20:19:47 (+0100), bart sikkes wrote: ]
> Subject: somewhat ot: looking dec multia help
>
> what would the wise thing to do now, just try a new battery? (i have
> seen remarks on old part numbers and such, but living in the
> netherlands we dont have a radioshack, is it ok to just use some 4.5
> volt battery or are there other things to pay attention to?)

Yes, you definitely need a new battery first.

The old, original, battery which I removed from one of my multias and
which was still sitting at the back of my workbench is a plain little
cube-shaped Rayovac "computer clock battery" model # 840.  It's a basic
4.5 volt alkaline cell, 800 mAh according to the NetBSD "multiafaq.html"
but others say 500 mAh is more than sufficient.

The old "dead" battery still reads 4 volts even after at least five
years sitting idle, which of course means that the clock & NVRAM in the
multia need nearly the full 4.5 volts to function and maybe they'd even
be happier with a full 5.0 volts -- not sure what the chip is any more.

The replacement (which I did find at RadioShack, IIRC) still works fine
too -- I netbooted the machine a couple of weeks ago, but it has been
turned off for the vast majority of those five years....

While you've got it open don't forget to heat-sink that little 74F623
chip on the bottom (location E215, IIRC)!

--

-- 
						Greg A. Woods
(Continue reading)

Permalink | Reply | 1 comment |
headers
Havard Eidnes | 8 Nov 2004 22:21
Picon
Picon

PCI plug-in card interrupts in a CS20?

Hi,

while trying to get some more FC disks attached to an API CS20, I
installed a Qlogic ISP 2100 controller in one of the two 64-bit
PCI slots.

Initially I tried it in the bottom slot, and I could not get the
machine through the autoconfig phase; after probing the
controller as

isp0 at pci0 dev 5 function 0: QLogic FC-AL HBA
isp0: interrupting at dec 6600 irq 24
scsibus1 at isp0: 256 targets, 8 luns per target

it started spewing spurious interrupt errors later while probing
the drives:

scsibus0: waiting 2 seconds for devices to settle...
scsibus1: waiting 2 seconds for devices to settle...
stray 6600 irq 24
stray 6600 irq 24
stray 6600 irq 24
stray 6600 irq 24
sd0 at scsibus0 target 0 lun 0: <HITACHI, DX32CJ-36MC, A2T2> disk fixed
sd0: 35256 MB, 15314 cyl, 12 head, 392 sec, 512 bytes/sect x 72205440 sectors
sd0: sync (25.00ns offset 31), 16-bit (80.000MB/s) transfers, tagged queueing
atapibus0 at atabus1: 2 targets
cd0 at atapibus0 drive 0: <SAMSUNG CD-ROM SN-124, , q008> cdrom removable
cd0: 32-bit data port
cd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 2 (Ultra/33)
(Continue reading)

Permalink | Reply | 4 comments |
headers
Greg A. Woods | 9 Nov 2004 02:04
X-Face
Favicon

just a quick FYI about mozilla-1.7.3 progress (or lack thereof) on alpha 1.6.x

Just a quick FYI about mozilla-1.7.3 progress on alpha 1.6.x...

At one point during the linking of some shared library (I don't remember
which) I thought the linker had gone in a loop -- it sat there chewing
CPU without any noticable IOs, and with an RSS of at least 44MB, for a
good 10-20 CPU minutes!

Then when I wasn't looking it finished and the build carries on until it
gets here:

gmake[4]: Leaving directory `/build/package-obj/www/mozilla/work/mozilla/directory/xpcom/base/resources'
gmake[3]: Leaving directory `/build/package-obj/www/mozilla/work/mozilla/directory/xpcom/base'
gmake[2]: Leaving directory `/build/package-obj/www/mozilla/work/mozilla/directory/xpcom'
gmake[1]: Leaving directory `/build/package-obj/www/mozilla/work/mozilla'
/usr/pkg/bin/gmake tier_40
gmake: execvp: /usr/pkg/bin/gmake: Bad address
gmake: *** [default] Error 127
*** Error code 2

Stop.
make: stopped in /work/woods/m-NetBSD-pkgsrc/www/mozilla
*** Error code 1

Stop.

Anyone have any clues what that means?

Bug in gmake?

--

--
(Continue reading)

Permalink | Reply | 0 comments |
headers
bart sikkes | 9 Nov 2004 09:12
Picon

Re: somewhat ot: looking dec multia help

> > Subject: somewhat ot: looking dec multia help
> >
> > what would the wise thing to do now, just try a new battery? (i have
> > seen remarks on old part numbers and such, but living in the
> > netherlands we dont have a radioshack, is it ok to just use some 4.5
> > volt battery or are there other things to pay attention to?)
> 
> Yes, you definitely need a new battery first.
> 
> The old, original, battery which I removed from one of my multias and
> which was still sitting at the back of my workbench is a plain little
> cube-shaped Rayovac "computer clock battery" model # 840.  It's a basic
> 4.5 volt alkaline cell, 800 mAh according to the NetBSD "multiafaq.html"
> but others say 500 mAh is more than sufficient.
> 
> The old "dead" battery still reads 4 volts even after at least five
> years sitting idle, which of course means that the clock & NVRAM in the
> multia need nearly the full 4.5 volts to function and maybe they'd even
> be happier with a full 5.0 volts -- not sure what the chip is any more.
> 
> The replacement (which I did find at RadioShack, IIRC) still works fine
> too -- I netbooted the machine a couple of weeks ago, but it has been
> turned off for the vast majority of those five years....
> 
> While you've got it open don't forget to heat-sink that little 74F623
> chip on the bottom (location E215, IIRC)!

thanks for all the information so far, ill focus on getting a battery
first then and see what happens when i put that in.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Havard Eidnes | 9 Nov 2004 15:51
Picon
Picon

Re: PCI plug-in card interrupts in a CS20?

> Moving the card to the upper PCI slot made it go further, but it
> still does not work reliably.  With the card in that position, I
> can create the raidframe raid, but it will again start spewing
> the spurious interrupt messages at the moment I try to do newfs
> of a partition covering most of the raid.
>
> Why does these messages occur?  Isn't the controller / driver
> supposed to handle these interrupts?

It appears that the interrupt status register on the device
indicates that it did not trigger the interrupt.  With the
attached debugging changes applied to the isp_pci.c file, I got
these console messages when I tried newfs'ing the raidframe
device partition.

Guidance for what the actual problem might be still gratefully
accepted.

- Havard

------------------------------

raid0: RAID Level 0
raid0: Components: /dev/sd2a /dev/sd3a /dev/sd4a /dev/sd5a /dev/sd6a /dev/sd7a /dev/sd8a /dev/sd9a
/dev/sd10a /dev/sd11a
raid0: Total Sectors: 177827720 (86829 MB)
raid0: New autoconfig value is: 1
raid0: New rootpartition value is: 0
isp_pci_rd_isr: isr == 0 && sema == 0 (after mask)
isp_pci_rd_isr: isr == 0x0, sema == 0x2
(Continue reading)

Permalink | Reply | 3 comments |
headers
Havard Eidnes | 9 Nov 2004 16:34
Picon
Picon

Re: PCI plug-in card interrupts in a CS20?

Hmm...

some more thinking led me to come to think of the bus_space_barrier()
function and that it might be required between the two bus_space_read_2
calls in quick succession in isp_pci_rd_debounced().  So, modifying the
diff as shown below moved the problem to somewhere else...

I now got, while trying:

kveite# dd of=/dev/null bs=8k count=100 if=/dev/rsd2a
100+0 records in
100+0 records out
819200 bytes transferred in 0.067 secs (12226865 bytes/sec)
kveite# dd of=/dev/null bs=8k count=1000 if=/dev/rsd2a
1000+0 records in
1000+0 records out
8192000 bytes transferred in 0.556 secs (14733812 bytes/sec)
kveite# dd of=/dev/null bs=8k count=10000 if=/dev/rsd2a
dd: /dev/rsd2a: Input/output error
1062+0 records in
1062+0 records out
8699904 bytes transferred in 0.716 secs (12150703 bytes/sec)
kveite# 

on the console:

isp0: DMA error for command on 0.34.0
isp0: BOTCHED cmd for 0.34.0 cmd 0x8 datalen 8192
sd2(isp0:0:34:0): generic HBA error
(Continue reading)

Permalink | Reply | 2 comments |

Gmane