[PATCH] powerpc: Unify the 32 and 64 bit idle loops

[Kumar Gala]

I'm now seeing the following kernel panic on an 83xx system (603  
core) w/preempt on (2.6.16 is ok on this system):

Gianfar MII Bus: probed
eth0: Gianfar Ethernet Controller Version 1.2, 00:e0:db:00:04:fe
eth0: Running with NAPI disabled
eth0: 64/64 RX/TX BD ring size
e100: Intel(R) PRO/100 Network Driver, 3.5.10-k2-NAPI
e100: Copyright(c) 1999-2005 Intel Corporation
Marvell 88E1101: Registered new driver
i2c /dev entries driver
NET: Registered protocol family 2
Non-recoverable exception at PC=c0009498 MSR=8000
Oops: nonrecoverable exception, sig: 9 [#1]
Modules linked in:
NIP: C0009498 LR: C0007E78 CTR: C00093D8
REGS: c022bef0 TRAP: 0900   Not tainted  (2.6.16-g9618edab)
MSR: 00008000 <EE>  CR: 24024048  XER: 20000000
TASK = c01f5da8[0] 'swapper' THREAD: c022a000
GPR00: 00000000 C022BFA0 C01F5DA8 00800000 8090C000 00E00000 00000001  
00048000
GPR08: 00000000 C00093D8 22024022 00000010 000009D3 9D410001 0FFFF000  
007FFFFA
GPR16: 00000001 FFFFFFFF 00000000 0FFFA0E4 00000000 007FFF00 0FFA7758  
00000002
GPR24: 00000000 00800000 40000000 C0249370 C0249370 00000010 C022A024  
C022A000
NIP [C0009498] ppc6xx_idle+0xc0/0xcc
LR [C0007E78] cpu_idle+0x9c/0xf0
Call Trace:
[C022BFA0] [00800000] 0x800000 (unreliable)
[C022BFC0] [C00037A4] rest_init+0x28/0x38
[C022BFD0] [C022C6A0] start_kernel+0x1b8/0x1f8
[C022BFF0] [00003378] 0x3378
Instruction dump:
64a50080 7c842878 7c841b78 64840010 7c90fba6 60000000 60000000 60e78000
64e70004 7c0004ac 4c00012c 7ce00124 <4c00012c> 7c0004ac 4e800020  
7d70faa6
<0>Kernel panic - not syncing: Attempted to kill the idle task!
<0>Rebooting in 180 seconds..



On Mar 24, 2006, at 3:54 AM, Paul Mackerras wrote:

> This unifies the 32-bit (ARCH=ppc and ARCH=powerpc) and 64-bit idle
> loops.  It brings over the concept of having a ppc_md.power_save
> function from 32-bit to ARCH=powerpc, which lets us get rid of
> native_idle().  With that, we can also reduce the amount of special
> code for pSeries and cell because the special code has only to do what
> needs to be done when there is no work to do, rather than being a
> complete idle loop.
>
> iSeries still has its own idle loops because it wants to test
> hvlpevents_pending() as well as need_resched() etc.  I don't know if
> that is essential but I have left it as-is for now.
>
> Signed-off-by: Paul Mackerras <paulus at samba.org>