On Tue, 18 Jan 2005, Henry Nelson wrote:

>On Tue, Jan 18, 2005 at 12:25:27PM +0000, Christian Smith wrote:
>>
>> The early 1.6 releases exposed a bug in the sun4c MMU handling code, which
>> had always been there but just not triggered. This mat be the source of
>> your (non-disk related) problems.
>>
>> AFAIK, 1.6.2 fixed this issue, as well as being fixed in 2.0.
>>
>> See:
>> http://mail-index.netbsd.org/port-sparc/2003/07/18/0002.html
>
>Thanks for sending this URL.  I did read it, but had forgotten about it.
>I remember now that that post was the reason I waited for 1.6.2 to come out.
>That
>  `` Other MMU resouce management related problems may still surface on these
>     machines, such as the "all pmegs are gone" panic that has been reported
>     several times already. ''
>doesn't sound very encouraging, however.
>
>What is "the VM system" mentioned in that post?  Maybe there is some way
>to turn it off in the kernel?


Not advisable (or possible). VM is Virtual Memory. It implements the
mapping from memory objects to the virtual address space. Memory objects
are things such as files and anonymous memory, and are fundamentally
required. 1.6 implemented the Unified Buffer Cache with UVM (the VM model
used with NetBSD since 1.4), which is what bought out the problem with the
pmap layer on SPARC. Pmap is what implements virtual to physical address
mapping, and is the location of the bug.

Basically, memory is modelled as:

               +---------------+
               |    Process    |
               |---------------|
               | Address Space |
CPU Indepenent | - - (UVM) - - | Virtual Addresses
               | Memory Objects|
      ---------|---------------|
 CPU dependent |     PMAP      |-------------------
          -----|---------------|
      Hardware | Physical Mem. | Physical Addresses
               +---------------+

As you can see, UVM is pretty entrenched in the workings of the kernel:)

I've had some IPX crashes (also sun4c) under heavy load with NetBSD
2.0_BETA, but could not pin the problem down to hardware or software. I
may try reproducing the problem in the future, but for the time being, the
IPX is retired.

>
>henry nelson
> | day job: | http://yuba.kcn.ne.jp/biorec/nehan/henken.html
>

Christian

-- 
    /"\
    \ /    ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
     X                           - AGAINST MS ATTACHMENTS
    / \