Because of the size and detail of this expedition, and emphasis below by
Guy.Bormann on SCSI issues, this response is (I hope) limited to the
SCSI issues. Also, since this is apparently moot for cooker purposes,
I've moved the response to the expert list in order to provide more
suitable location for archive of issues and/or resolution for others.
Guy.Bormann wrote:
> Felix Miata wrote:
> > The whole of RH7.3 /etc/modules.conf is:
> > alias scsi_hostadapter sym53c8xx
> > alias eth0 8139too
> Could be useful to enforce ordering maybe in the IRQ assignment by the
> kernel...
> [snip : see below]
> > cat /proc/interrupts from RH7.3:
> > CPU0
> > 10: 12091 XT-PIC sym53c8xx
> !!!--> NOT happy!!! see below
> > cat /proc/interrupts from RC3, without sound in /etc/modules:
> > CPU0
> > 10: 76 XT-PIC sym53c8xx
> !!!NOT happy!!!
> > Note working mouse, but absent sound, and free IRQ 11
> Because according to your previous post you assigned IRQ 11 in the BIOS to
> the SuperVGA card in slot 3. However, this type of video card doesn't need
> an IRQ since it is a completely memory-mapped device. So, turn of the BIOS
> IRQ assignment to VGA and additionally switch off the IRQ 11 allocation to
> slot 3 (i.e. by putting it back to Auto; in fact, you should do that for
> all the slots!!!). See below...
Actually, the ET6x00 once did require an IRQ. With the native Tseng
video drivers, the OS/2 PM desktop would not come up without a BIOS IRQ
assignment for the ET6x00. Over a year ago I switched OS/2 over to the
Scitech Display Doctor Pro video drivers, and this apparently made the
need for the IRQ obsolete, as I'm now running without one. OS/2 shows
USB using IRQ 11, which is no longer explicitly assigned to anything in
the BIOS, and VGA using no IRQ, as the BIOS is now set not to assign it
one. Many moons ago, before OS/2, I used DesqView. IIRC, it also
required VGA to have an IRQ.
> > Differences between machines:
> [A world of difference, but from this I can infer that the Tyan is giving
> you the hasles????]
I thought I made that explicit earlier. The Tyan is the MVP3, the
problem child that started the thread.
> > # of SCSI HD's 1 2 !!!!
> > Boot hda sdb !!!!
!!!!? Your point? The Tyan MVP3 is booting and running hda.
The AOpen TX is booting RHL7.3 from sdb8, but mdk8.2 from hda7, mdk7.1
from sdc10, and Corel 1.2 from sdc9. sdb and sdc are the same physical
disk, 0,3,0. Whether it is sdb or sdc depends on whether SCSI 0,2,0 is
enabled during powerup.
> > That's all the differences. Same include: ET6100, sym53c875 PCI slot 1,
> > single IDE HD, CS4235 ISA slot 1, ALN-325 (Realtek 8139) PCI slot 2,
> > SCSI CD & CD-RW.
> > I tried changing BIOS settings to stop the usb/eth0 sharing on IRQ 3,
> > but nothing worked. IRQ 11 stays unused.
> No wonder, see above!!
Somewhere during the process reported in my previous post I disabled the
VGA IRQ assignment, which is how it remains.
> > Next I tried using the different sound options from from the RH7.3
> > modules.conf in rc3. That produced another no-mouse result and the
> > following cat /proc/interrupts:
> > CPU0
> > 10: 68 XT-PIC sym53c8xx
> !!!NOT happy!!!
Below . . .
> > SCSI subsystem driver Revision: 1.00
> > PCI: Found IRQ 11 for device 00:08.0
> > IRQ routing conflict for 00:08.0, have irq 10, want irq 11
> !!!!?????The MP3's don't happen to be on a SCSI disk, do they???!!!!
Mdk9/Tyan = MP3's on hda. RHL/AOpen probably = on SCSI somewhere. RHL
doesn't provide HPFS support in the stock kernel, and since HPFS is
where they were, I had to shut down RHL and boot mdk 8.2 to copy some
someplace. When I quit for that day I deleted the copies.
> > Nothing helped until I changed BIOS PCI INT 5 to Legacy ISA, leaving all
> > others at PCI/ISA PNP. That produced the following cat /proc/interrupts:
> > CPU0
> > 10: 67 XT-PIC sym53c8xx
> ^^ consistently low, compared to the RH7.3 machine
> !!!! NOT HAPPY, our usual suspect, see below
Might this be because RHL is running off SCSI and 9.0 doesn't hasn't
accessed any SCSI files? The only SCSI HD in the Tyan (mdk 9.0) machine
is used only for backing up FAT16B and HPFS data , just sitting quietly
while fooling with MP3's and running Linux. OTOH, RHL7.3 is running on
SCSI.
> > PCI: Found IRQ 11 for device 00:08.0
> > IRQ routing conflict for 00:08.0, have irq 10, want irq 11
> !!!And again, did you read over this message???
Yup! Before disabling the VGA IRQ assignment in the BIOS, I went into
the BIOS and reversed the IRQ assignments between the VGA card and the
SCSI, resulting in the following:
PCI: Found IRQ 10 for device 00:08.0
IRQ routing conflict for 00:08.0, have irq 11, want irq 10
This SCSI seems always to want the grass from the other side of the
fence. Since the BIOS change failed to remove the routing conflict
message from dmesg, I put things back like they were before proceeding.
OTOH, I'd love to know how to deal with the following later message:
sym0:2:0: tagged command queuing enabled, command queue depth 16.
Most of my SCSI disks, including the one on 0,2,0, have a maxiumum queue
depth support of 8. I'm sure this creates voodoo errors, and I'd like to
limit the depth accordingly.
> > soon as I clicked OK on the MP3 file to open, hard lockup.
> !!!In case you read over my question : do the MP3's by any chance reside
> !!!on a SCSI disk (or does anything related to the above used
> !!!programs)???????
0.
> > That worked OK for system sounds and playing CD's, but still allowed hard
> > locks trying to play MP3's off a HD. Smells like a DMA conflict.
> Do you start to see the pattern??? :-)))))
Not one involving SCSI, I think, at least not directly anyway.
> > Not knowing anything else to try, I created another partition and installed
> > RedHat 7.3. That gobbled a bunch of time, and produced essentially the
> > same result, hard lock playing MP3's off the HD, but otherwise OK, using the
[Playing MP3's off hda booted and running from hda.]
> Errr, do I need to say more?????
I think so, but lets settle the SCSI issue before proceeding with ISA
sound.
> > Is this possibly a resolvable DMA problem? After all, OS/2 plays
> > MP3's on this box without locking up.
> Well, OS/2 ignores the BIOS IRQ assignments and reprograms the IRQ
> router in case the MP3's happen to be on a SCSI device.
I don't think it does. The only IRQ it seems to handle specially is 9,
which is the cascade to 2. In other cases, if the BIOS has made an
assignment, and OS/2 wants one, then the one assigned by the BIOS is the
one used.
> If not, it still
> is an IRQ problem, not DMA one, I think(!).
> Summary : SCSI card doesn't get the IRQ it wants.
Regardless of anything I've ever done so far. To me more likely that
means the sym53c8xx module is fubar -> According to Symbios docs I read
long long long ago, SCSI is happiest on 10 or 11. I don't remember
whether one or the other is preferred, or either equally OK.
I once put a second Symbios card in this same MVP3 machine, and no
amount of BIOS fiddling or slot rearranging would result in both on the
same IRQ, though at least one was always shared with something else.
> TO DO : do the testing on the problem machine. If you get it to work but
> are experiencing random lockup (but long enough after boot), try replacing
> hardware ONE by ONE from the other (AOpen) machine.
No need to steal from the other machine. I have identical spares on the
shelf. 8^)
> Try the following setup strategy :
> * Change modules.conf to include all the top-level aliases
"top-level" means what?
"alias" in the context of modules.conf means what? Why use them at all?
--
". . . . in everything, do to others what you would have them do
to you . . . ." Matthew 7:12 NIV
Team OS/2 ** Reg. Linux User #211409
Felix Miata *** http://members.ij.net/mrmazda/
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
