Re: 2.6.18-1-sparc64 regression also on SunBlade1000

[Roland Stigge <stigge@antcom.de>]

Hi Jurij,

Jurij Smakov wrote:
> Thanks for your report. If you could figure out the last working 
> kernel, that would be very helpful.

Last working is 2.6.16, as also experienced by Uwe.

2.6.17 shows the boot errors as screenshoot at

http://people.debian.org/~stigge/2.6.17.boot.jpg

and with complete dmesg:

http://people.debian.org/~stigge/2.6.17.dmesg1.jpg
http://people.debian.org/~stigge/2.6.17.dmesg2.jpg

The qlogicfc driver used by my system with 2.6.16 has been removed for
2.6.17 in favour of qla2xxx.

With 2.6.18, an interesting change in kernel sources is:

--- linux-source-2.6.17/arch/sparc64/kernel/traps.c     2006-06-18
03:49:35.000000000 +0200
+++ linux-source-2.6.18/arch/sparc64/kernel/traps.c     2006-09-20
05:42:06.000000000 +0200
@@ -2544,7 +2545,9 @@
            (TRAP_PER_CPU_TSB_HUGE !=
             offsetof(struct trap_per_cpu, tsb_huge)) ||
            (TRAP_PER_CPU_TSB_HUGE_TEMP !=
-            offsetof(struct trap_per_cpu, tsb_huge_temp)))
+            offsetof(struct trap_per_cpu, tsb_huge_temp)) ||
+           (TRAP_PER_CPU_IRQ_WORKLIST !=
+            offsetof(struct trap_per_cpu, irq_worklist)))
                trap_per_cpu_offsets_are_bolixed_dave();

        if ((TSB_CONFIG_TSB !=

Will need to check if this is the cause.

bye,
  Roland