On Sat, 21 Sep 1996, Jon Lewis wrote:
> On Sat, 21 Sep 1996, Jon Lewis wrote:
>
> > Things I turned off are:
> > DRAM Speculative Leadoff
> > Turn-Around Insertion
> > Peer Concurrency
> > Chipset Special Features
>
> Ok...I've compiled a few more kernels, played a lot with CMOS settings,
> locked up a few more times, and it seems that my problem is that my Tomcat
> II just won't run reliably with the BSD NCR driver if I have "Chipset
> Special Features" enabled. Other settings don't appear to make a
> difference, and I can run with or without CSF with the 53c7,8xx driver
> with no problems. Anyone have a clue on this one?
>
Nice!
A register dump just before the driver resets the chip may help.
The patch below should do this.
Try it once and cut the register dump from the syslog.
You must get something like that:
ncr53c825-0: regs[00...0f]: d8 00 00 03 67 00 03 0f 00 00 03 00 80 00 0f 02.
ncr53c825-0: regs[10...1f]: ff 52 08 00 00 ff ff ff 00 f0 35 21 00 00 00 00.
ncr53c825-0: regs[20...2f]: 00 00 00 00 00 00 08 98 4c fc 09 00 4c fc 09 00.
ncr53c825-0: regs[30...3f]: 0b 00 00 00 bc 50 08 00 8e 15 00 a1 57 fc 09 00.
ncr53c825-0: regs[40...4f]: 8f 06 00 00 84 00 20 0e 0c 00 80 00 07 00 00 80.
--------------------- Cut here --------------------------
--- linux/drivers/scsi/ncr53c8xx.c.01 Sat Sep 21 20:07:01 1996
+++ linux/drivers/scsi/ncr53c8xx.c Sun Sep 22 09:42:24 1996
@@ -3526,6 +3526,23 @@
np->reg = (struct ncr_reg*) np->vaddr;
+#if 1
+ /*
+ ** Dump chip registers as left by bios.
+ */
+ if (1) {
+ int i = 0;
+ do {
+ printf("%s: regs[%02x...%02x]:", ncr_name(np), i, i+15);
+ do {
+ printf (" %02x", (unsigned)INB_OFF(i));
+ ++i;
+ } while (i%16);
+ printf(".\n");
+ } while (i < 0x50);
+ }
+#endif
+
#ifndef NCR_IOMAPPED
retry_chip_init:
#endif
----------------------- Cut here ----------------------------
> All the manual says is:
>
> Chipset Special Features
> When disabled, the chipset behaves as if it were the earlier 430FX
> chipset. This option should be enabled for best performance.
> The default is enabled
>
> Anyone know what I lose by disabling this option?
>
Very probably, Stefan Esser does.
> I'm going to put it into a kernel compile loop now with 2.0.21-bsd with
> CSF disabled and see if it can make it through a hundred or so compiles
> while compiling a few other things as well.
Thanks for your help.
Gerard.