Re: panic pmap_tlb_intersecting_active_p on RPI2

[Michael Plass <mfpnb%plass-family.net@localhost>]

On Jul 25, 2016, at 5:28 PM, Michael Plass wrote:
> 
> On Jul 25, 2016, at 12:12 AM, Michael Plass wrote:
> 
>> On Jul 24, 2016, at 2:09 AM, Nick Hudson wrote:
>> 
>>> On 07/17/16 01:10, Michael Plass wrote:
>>>> I'm seeing a panic with a recently built -current (a couple of instances here):
>>> 
>>>> panic: kernel diagnostic assertion "pmap_tlb_intersecting_active_p(PAI_PMAP(pai,
>>>> ti), ti)" failed: file "/usr/home/michael/netbsd/usr/src/sys/uvm/pmap/pmap_tlb.
>>>> c", line 220
>>> 
>>> Can you try with src/sys/uvm/pmap/pmap_tlb.c:1.18 ?
>>> 
>>> Thanks,
>>> Nick
>> 
>> It's been running now for over an hour.
>> 
>> Thanks,
>> - Michael
>> 
>> 
>> 
> ...

> I saw there was another commit about pmap locking - I have not yet pulled that into my tree.
> 

Nick,

Sorry - I just noticed that I didn't yet have 1.18 in my kernel yet for that last panic.

With pmap_tlb.c,v 1.18, my machine *almost* survived running a -j5 kernel build, but panicked
during or soon after the final link step:

panic: kernel diagnostic assertion "kcpuset_isset(pm->pm_onproc, cpu_index(ci))" failed: file "/nb/usr/src/sys/uvm/pmap/pmap_tlb.c", line 976 pmap_tlb_asid_deactivate: pmap 0xba76da48 onproc 0xa6e532c8 doesn't include cpu 0 (0x80550fd0)
cpu0: Begin traceback...
0x8636fbcc: netbsd:db_panic+0xc
0x8636fbe4: netbsd:vpanic+0x1b4
0x8636fbfc: netbsd:__udivmoddi4
0x8636fc3c: netbsd:pmap_tlb_asid_deactivate+0x11c
0x8636fc54: netbsd:pmap_deactivate+0x40
0x8636fcac: netbsd:mi_switch+0x328
0x8636fcdc: netbsd:sleepq_block+0xb4
0x8636fd1c: netbsd:turnstile_block+0x33c
0x8636fd8c: netbsd:rw_enter+0x3e8
0x8636fdac: netbsd:vm_map_lock+0x68
0x8636fe04: netbsd:uvm_map_prepare+0x154
0x8636fe6c: netbsd:uvm_map+0x88
0x8636feb4: netbsd:uvm_mmap.part.0+0x18c
0x8636ff34: netbsd:sys_mmap+0x3e4
0x8636ffac: netbsd:syscall+0x98
cpu0: End traceback...

- Michael