Re: kernel panic with memory disks

Kris Kennaway wrote:
> On Tue, Aug 21, 2007 at 10:32:45AM +0300, Stefan Lambrev wrote:
>   
>> Hello,
>>
>> Eric Kjeldergaard wrote:
>>     
>>> On 20/08/07, Stefan Lambrev <stefan.lambrev_at_moneybookers.com> wrote:
>>>  
>>>       
>>>> Hello,
>>>>
>>>> I do not know if this is know behavior, and I know that 6.2 panic if the
>>>> memory disk got full,
>>>> but on 7-current the panic is before the disk got full.
>>>>
>>>> Here is what I do:
>>>>
>>>> mdconfig -a -t malloc -s 800m
>>>> newfs /dev/md0
>>>> mount /dev/md0 /mnt
>>>> cp 600mb.file /mnt
>>>>    
>>>>         
>>>     -t type
>>>             Select the type of the memory disk.
>>>
>>>             malloc   Storage for this type of memory disk is allocated 
>>>             with
>>>                      malloc(9).  This limits the size to the malloc bucket
>>>                      limit in the kernel.  If the -o reserve option is not
>>>                      set, creating and filling a large malloc-backed 
>>>                      memory
>>>                      disk is a very easy way to panic a system.
>>> -- mdconfig(8)
>>>
>>>  
>>>       
>> I really should read manuals more carefully :) Thanks for the information.
>>
>> Just one more question - to prevent panic I should use "-o reserve" and 
>> have to increase:
>> vfs.maxmallocbufspace
>> vfs.bufmallocspace
>>     
>
> No, you should use -o swap.  Where did it tell you to change the
> sysctls?
>
> Kris
>   
Nowhere just guessing.

I just needed one big file in the memory to ignore the slowness of hard 
drives, to run few small benchmarks :)
I did this using tmpfs, but it act just like "-t swap" :)

Btw the confusion comes from the manual of mdconfig where it states:

swap     Swap space is used to back this memory disk.

and I thought that type swap is always stored on the hard drives.

and md(4) explains it a lot better:

swap     Backing store is allocated from buffer memory.  Pages get pushed
              out to the swap when the system is under memory pressure, 
other-
              wise they stay in the operating memory.  Using swap backing is
              generally preferable over malloc backing.


-- 

Best Wishes,
Stefan Lambrev
ICQ# 24134177