under ZFS, I can reliably crash my systems

[Andriy Gapon]

on 06/10/2012 19:10 Dennis Glatting said the following:
> On Sat, 2012-10-06 at 18:20 +0300, Andriy Gapon wrote:
>> on 06/10/2012 17:37 Dennis Glatting said the following:
>>> On Fri, 2012-10-05 at 19:09 +0300, Andriy Gapon wrote:
>>>> on 05/10/2012 17:33 Dennis Glatting said the following:
>>>>> swap_pager: out of swap space
>>>>> swap_pager_getswapspace(16): failed
>>>>> pid 1847 (fstat), uid 0, was killed: out of swap space
>>>>
>>>> One thing I can tell you, your kernel and userland are out of sync.
>>>>
>>>
>>> How so? If svn src and rebuild everything. Do you mean ports?
>>>
>>> Typical src build is:
>>>
>>> svn co svn://svn.pki2.com/base/stable/9/ /disk-1/src
>>> cd /usr/src; make -j65 buildworld
>>> make installworld
>>> yes | make delete-old
>>> yes | make delete-old-libs
>>> mergemaster
>>> cd /sys/amd64/conf/
>>> ./mkconfig.pl SMUNI.in 
>>> config SMUNI
>>> cd ../compile/SMUNI
>>> make cleandepend && make depend && make
>>> make install
>>>
>>> That's fairly straight forward.
>>
>> Why not use buildkernel target with KERNCONF=SMUNI?
>>
> 
> Is there a difference in the processes?

Most likely no.
But buildkernel is the official way.  If something changes in the official
procedure, then your manual procedure may miss it.

>> Anyway, well, maybe your kernel (the one that produced the crashdump) was from
>> before the upgrade.  fstat trying to allocate insane amounts of memory during
>> vmcore processing is a sign that fstat and the kernel were compiled using
>> different versions of system headers.
>>
> 
> 


-- 
Andriy Gapon