Preemption on MV78100 (ARMv5TE)?

[Ian Lepore]

On Mon, 2013-03-25 at 09:36 -0700, MagnusNilsson wrote:
> > Oh yeah, here's another fix that isn't in 8.2 that's pretty important,
> > at least on kirkwood series chips... Enabling cache write-allocate
> > causes data corruption.
> 
> Applying that patch made for a proper crash instead of a silent hang (which
> I consider a step forward):
> "Entropy harvesting: interrupts ethernet point_to_point
> vm_fault(0xc4dfc4e4, 0, 2, 0) -> 1
> Fatal kernel mode data abort: 'Translation Fault (P)'
> trapframe: 0xd8ea2b64
> FSR=00000017, FAR=00000018, spsr=600000d3
> r0 =000002a0, r1 =00000000, r2 =00000380, r3 =0000001c
> r4 =c51115e0, r5 =00000000, r6 =00000000, r7 =00000000
> r8 =cbec9c50, r9 =c0bc5da0, r10=c0b4f908, r11=c0bb1178
> r12=c0bca52c, ssp=d8ea2bb0, slr=c09f25fc, pc =c09f26f0"
> 
> pc is in sleepq_add(), which I assume is reasonable on a system with broken
> preemption?

Hrm.  Yeah, progress but not all good.

Preemption shouldn't be broken.  I enabled it on one of my atmel-based
units here that runs 8.2-plus-some-fixes and it's working fine.

I guess I need to get 8.2 running on a marvell chip to see if I can
reproduce the preemption problem.  That may take a couple days, it's
been quite a while since I've run anything but -current on marvell.

-- Ian