Date: Wed, 29 Nov 1995 23:18:50 +0100 From: se@zpr.uni-koeln.de (Stefan Esser) To: iain@nwpeople.demon.co.uk Cc: hackers@freebsd.org Subject: Re: NCR815 problem Message-ID: <199511292218.AA18436@Sysiphos> In-Reply-To: iain@nwpeople.demon.co.uk (Iain Baird) "Re: NCR815 problem" (Nov 29, 19:17)
next in thread | raw e-mail | index | archive | help
On Nov 29, 19:17, Iain Baird wrote:
} Subject: Re: NCR815 problem
} > Your system uses much lower addresses: 0xc0000 is an
} > address below 1MB, i.e. in the ISA peripheral address
} > space.
}
} What causes this? Is it my motherboard/BIOS, or the NCR 815 card?
} I have been thinking about replacing the motherboard.
The PCI BIOS seems to have choosen that address.
But it is valid. You just have to make sure that
BIOS shadowing is OFF for that memory region ...
} I've turned caching and shadowing off.
} I haven't seen any cache test failures since.
That is a strong indication, that there really
was a coherency problem :)
} I have experimented with this, using other IRQs makes no difference.
} I have only one IDE controller, it uses IRQ 14.
}
} > } pci0:5: vendor=0x1c1c, device=0x0001, class=old (misc) [no driver assigned]
} > ^^^^^^
} > any idea, what kind of device that is ?
}
} Yes: it's a DTC PCI IDE Controller card. My motherboard has no
} on-board IDE controller, the system came with the DTC card.
Ok. Thanks ...
} Well, I got a little further, but I have been unable to get sd0
} to work under FreeBSD. It works fine under DOS. I have attempted to
} disklabel the sd0 manually, but I think I'm tripping over the new
} slice support. I consulted the FAQ, it didn't help - the FAQ was
} clear enough, but nothing it suggested actually worked.
}
} Finally I used the installation floppy to create a FreeBSD
} partition on sd0. I had previously created a DOS partition so
} that the correct BIOS geometries would be picked up (1021/31/63).
Hmm, I'm not sure that is right. I believe, I once
tried a similar geometry, and found 1021/63/31 made
things work ...
Anyway, I just started a DOS install, and stopped it
when it got to "formatting C:". I kept the geometry
that had been put into the MBR by DOS.
} I chose "Dedicate whole disk to FreeBSD" and "Don't cooperate
} with other OS's". I forced a write from the partition menu.
}
} I then booted off IDE, verbose boot messages follow below.
} Note the last message:
}
} sd0s1: C/H/S end 1021/0/24 (24527) != end 1994036: invalid
^^^
The "head" field comes out as "0", and that is obviously
wrong ...
} Attempts to disklabel the partition failed with the same error.
} I don't know whether I'm doing something wrong, or hitting a hardware
} problem or a bug. Something seems to be trashing the geometry.
} I'll be grateful for any further assistance you can give.
You most probably have only this geometry problem
left. I'd guess you should just start a DOS install,
and then use the geometry the FreeBSD sysinstall
offers as default.
This is not a NCR driver problem anymore, you got
beyond that :)
} The following boot messages were obtained with:
} External cache disabled
It surely should work with external cache enabled!
} Caching of C800-CBFF,CC00-CFFF disabled
} (Should this make any difference, if external cache is disabled?
} What about internal L1 cache?)
This memory region is generally only accessed by
the BIOS during the first two boot stages. But
since the NCR seems to have been mapped into that
memory region, you should make sure that caching
and shadowing is off.
The L1 cache is write-through on the i486DX2, so
there are no coherency problems.
In fact, you should have L2 enabled and set to
write-through, too, for performance reasons.
} Shadowing of C800-CBFF,CC00-CFFF disabled
} IRQ 15 the only IRQ left free for PCI
Ok.
} FreeBSD 2.1.0-951026-SNAP #0: Mon Nov 27 02:28:15 GMT 1995
} root@nomad:/usr/src/sys/compile/NOMAD.NCR
} CPU: i486 DX2 (486-class CPU)
} Origin = "GenuineIntel" Id = 0x435 Stepping=5
} Features=0x3<FPU,VME>
} chip0 <Intel 82424ZX (Saturn) cache DRAM controller> rev 4 on pci0:0
} CPU: 486DX2 or 486DX4, bus=33MHz, CPU->Memory posting ON
} Warning: NO DRAM parity!
} Cache: None
} DRAM: page mode code fetch, read and write, memory clocks=X-1-2-1
} CPU->PCI: posting ON, burst mode ON
} PCI->Memory: posting ON
Ok. The cache could be ON, and the PCI performance
features should all work reliably with that revision
of the Saturn chip set.
} ncr0 <ncr 53c815 scsi> rev 3 int a irq 15 on pci0:3
} mapreg[10] type=1 addr=0000ff00 size=0100.
} mapreg[14] type=0 addr=000cc000 size=0100.
} pci_map_mem failed: device's memrange 0xcc000-0xcc0ff is incompatible with its bridge's memrange 0x4000000-0xffffffff
The warning just indicates that the PCI code is
surprised anybody maps PCI devices into the ISA
address space. Be surprised, too, but ignore it
otherwise :)
} reg20: virtual=0xf2a4f000 physical=0xcc000 size=0x100
Ok.
} ncr0 scanning for targets 0..6 (V2 pl23 95/09/07)
} ncr0 waiting for scsi devices to settle
You can remove the 'options "SCSI_DELAY=15"' line from
the kernel config file, this will make your system boot
much faster.
} (ncr0:0:0): "SEAGATE ST31250N 0006" type 0 fixed SCSI 2
} sd0(ncr0:0:0): Direct-Access
} sd0(ncr0:0:0): FAST SCSI-2 100ns (10 Mb/sec) offset 8.
} 973MB (1994037 512 byte sectors)
} sd0(ncr0:0:0): with 3712 cyls, 5 heads, and an average 107 sectors/track
The drive is correctly identified and synchronous
transfers at a rate of 10MHz are negotiated. There
have been a number of SCSI transactions already,
proof that the NCR driver works on your system.
} BIOS Geometries:
} 0:03fe0f3f 0..1022=1023 cylinders, 0..15=16 heads, 1..63=63 sectors
} 1:03fc1e3f 0..1020=1021 cylinders, 0..30=31 heads, 1..63=63 sectors
} 0 accounted for
} wd0s1: type 0x6, start 63, end = 409247, size 409185 : OK
} wd0s2: type 0xa5, start 409248, end = 1056383, size 647136 : OK
} sd0s1: type 0xa5, start 0, end = 1994036, size 1994037
} sd0s1: C/H/S end 1021/0/24 (24527) != end 1994036: invalid
If you get the number of heads right, then you are
ready to install FreeBSD ...
Regards, STefan
--
Stefan Esser, Zentrum fuer Paralleles Rechnen Tel: +49 221 4706021
Universitaet zu Koeln, Weyertal 80, 50931 Koeln FAX: +49 221 4705160
==============================================================================
http://www.zpr.uni-koeln.de/~se <se@ZPR.Uni-Koeln.DE>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199511292218.AA18436>