[9fans] sd53c8xx: bios scntl3(00) stest2(00)

[9fans] sd53c8xx: bios scntl3(00) stest2(00)

[Frank Lenaerts]

Hi,

I want to learn Plan 9, so I downloaded the latest[*] version of the
live CD. 

I want to use the following PC hardware for this: 

- mainboard, CPU, memory: MSI K7T Turbo 2 (VIA VT82C686, KT133) with
  an AMD Duron 850 and 768 MB RAM
- storage: Tekram DC-390U2W SCSI controller (LSI Logic / Symbios Logic
  53c895 (rev 01)) with 2 x 36GB disks
- ethernet controllers: 2 x VT6102

When I boot Plan 9 from the CD, the system seems to hang:

--- begin ---
Plan 9
E820: 00000000 0009fc00 memory
E820: 0009fc00 000a0000 memory
E820: 000f0000 00100000 reserved
E820: 00100000 2fff0000 memory
E820: 2fff0000 2fff3000 acpi nvs
E820: 2fff3000 30000000 acpi reclaim
E820: ffff0000 10000000 reserved
126 holes free
00017000 0009f000 557056
00264000 10000000 265928704
266485760 bytes free
cpu0: 850MHz AuthenticAMD AMD-Athlon (cpuid: ...)
ELCR: 1E20
pcirouting: BIOS workaround: PCI.0.7.2 at pin 4 link 5 irq 11 -> 5
pcirouting: BIOS workaround: PCI.0.7.5 at pin 3 link 3 irq 10 -> 11
sd53c8xx: SYM53C895 rev. 0x01 intr=11 command=0007
768M memory: 256M kernel data, 511M user, 1136M swap
root is from (...)[local!#S/sdD0/data]:
sd53c8xx: bios scntl3(00) stest2(00)
---  end  ---

After this line, the system seems to hang i.e. it doesn't continue.

When I choose to install Plan 9 from the CD, the output is similar:

--- begin ---
...
768M memory: ...
kfs...version...time...

init: starting /bin/rc
sd53c8xx: bios scntl3(00) stest2(00)
---  end  ---

Before I get the menu to choose either 1 (install from CD) or 2 (boot
the CD), I also noticed the following line:

--- begin ---
sd53c8xx: bios scntl3(9f) stest2(00)
---  end  ---

It takes several seconds before this line "returns" and the menu is
presented.

I searched the web to see if this is a known issue, but didn't get any
usefull result. [The SYM53C895 seems to be supported well, but I'm not
sure if this is also valid for "rev. 0x01". In addition to that, I
also wonder if there might be a problem with irq sharing. According to
the ESCD data (from the BIOS), the SCSI controller and the "Multimedia
Device" (Bus 0, Dev 7, Func 5) share irq 11 (and the pcirouting (2nd
line) seems to be ok). Note that all PCI slots share irq's with one or
another slot/device.]

Can somebody explain me or point me to a document explaining me how to
debug or work around this issue? Can I somehow boot with more debug
information? If so, how can I do this?


[*] Mon May 21 04:50:24 EDT 2007 md5 e189331bf8f56370881f49efb85d9423
    sha1 68dce6328cffdd623f2b4fc47b3822a6fbc4302e plan9.iso.bz2


Kind regards,

-- 
Frank Lenaerts ---------------------------------------- frank@inua.be

Re: [9fans] sd53c8xx: bios scntl3(00) stest2(00)

[Dave Eckhardt]

I am successfully using one of those cards on my
fossil server.  It emits that message every time
it boots (the last time it booted the values were
00 and 00; I can't claim they are every time).

If I can be of use in debugging, let me know...
I should have time starting Wednesday.

Have you checked that your SCSI bus is terminated?

Dave Eckhardt

Re: [9fans] sd53c8xx: bios scntl3(00) stest2(00)

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 57 bytes --]

i'd suspect that the interrupts aren't getting through.

[-- Attachment #2: Type: message/rfc822, Size: 2992 bytes --]

From: Dave Eckhardt <davide+p9@cs.cmu.edu>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] sd53c8xx: bios scntl3(00) stest2(00)
Date: Mon, 21 May 2007 20:16:24 -0400
Message-ID: <9158.1179792984@lunacy.ugrad.cs.cmu.edu>

I am successfully using one of those cards on my
fossil server.  It emits that message every time
it boots (the last time it booted the values were
00 and 00; I can't claim they are every time).

If I can be of use in debugging, let me know...
I should have time starting Wednesday.

Have you checked that your SCSI bus is terminated?

Dave Eckhardt

Re: [9fans] sd53c8xx: bios scntl3(00) stest2(00)

[Frank Lenaerts]

On Tue, May 22, 2007 at 10:49:35AM +0100, Charles Forsyth wrote:
> i'd suspect that the interrupts aren't getting through.

That's also what I first thought. Before posting my question to the
list, I therefore also checked the IRQ assignments to the PCI slots
and put the card in the one PCI slot that didn't share an IRQ with
another PCI slot.

Apart from the PCI slots however, there are still some other devices
(like AC97 audio and USB ports) sharing IRQs so I started playing with
them i.e. disabling devices, checking IRQ assignment, ... 

All this trial and error, with lots of reboots, finally resulted in 1
BIOS setting being crucial namely "Assignment of IRQ to USB". If this
is enabled, the system "hangs", if it is disabled, the system
boots[*].

Note that I also removed a dual serial port PCI card. I will add this
one again tomorrow and test with the latest Plan 9 CD.

[*] With an older Plan 9 CD I had because the one from May 21 2007
    seemed to be broken (also already mentionned by someone else on
    the list).

-- 
Frank Lenaerts ---------------------------------------- frank@inua.be

Re: [9fans] sd53c8xx: bios scntl3(00) stest2(00)

[Frank Lenaerts]

On Mon, May 21, 2007 at 08:16:24PM -0400, Dave Eckhardt wrote:
> I am successfully using one of those cards on my
> fossil server.  It emits that message every time
> it boots (the last time it booted the values were
> 00 and 00; I can't claim they are every time).
> 
> If I can be of use in debugging, let me know...
> I should have time starting Wednesday.
> 
> Have you checked that your SCSI bus is terminated?

As this system has always been running Debian GNU/Linux, I would say
yes.

Nevertheless, I also checked the user manual of the card and found out
that I should actually use the special twisted pair ribbon cable with
a SCSI terminator (LVD) to use Ultra2 SCSI (my disks are U2W) instead
of the flat cable (SE) which is only for Ultra/Wide SCSI.

I didn't know there was a difference between the 2 cables and used the
flat cable (having 3 connectors) as the machine only has 2 SCSI disks
anyway (so all 3 connectors would be used). The ribbon cable would
have connectors left over: it has 6 connectors: 1 for the controller,
1 for an LVD terminator and 4 for disks. I always thought that the 2
unused connectors would leave the SCSI bus open.

After switching the cables (not the connector), the Linux on the
harddisk worked as before (with no noticable difference but I didn't
investigate throughput or so any further). The Plan 9 CD didn't get
any further however.

> Dave Eckhardt

-- 
Frank Lenaerts ---------------------------------------- frank@inua.be

Re: [9fans] sd53c8xx: bios scntl3(00) stest2(00)

[Steve Simon]

It doesn't help you much but I use symbios SCSI cards in two servers.
The only problems I have had have been related to a dubious drive
which was sitting on the bus. This I worked around by hacking the driver,
however even I when I had this problem I got further than you are.

Sorry, but it should work.

-Steve