Memory debugging question...

Hello,

I'm working on a NIC driver for a 10 Gbps NIC and I have the
basics working,
but I clearly have a memory leak.  The thing is, I don't see
the leak when I
look at the output of 'vmstat -m'.  I see the memory climb
in the output of
'top' - it indicates that the "File" memory usage
is continually climbing
when I have transfers happening.

So the basic question is "what memory usage would be
considered 'File' memory
by 'top' but wouldn't show up in 'vmstat -m'?".  Of
course, it's possible I'm
just not seeing it in the 'vmstat -m' output, but I've
looked pretty hard!  

I have a feeling the problem is due to mismatched
"bus_dmamap_load..."/
"bus_dmamap_unload..." operations, but haven't
been able to confirm that.

Any tips on tracking this down would be most appreciated!

Thanks,

Patrick Klos
========= For LAN/WAN Protocol Analysis, check out
PacketView Pro! =========
    Patrick Klos                           Email:
patrickklos.com
    Network/Embedded Software Engineer     Web:   http://www.klos.com/
    Klos Technologies, Inc.                Phone:
603-714-0195
============================================================
================

On Mon, Jan 07, 2008 at 05:12:44AM -0800, patrickklos.com
wrote:
> Hello,
> 
> I'm working on a NIC driver for a 10 Gbps NIC and I
have the basics working,
> but I clearly have a memory leak.  The thing is, I
don't see the leak when I
> look at the output of 'vmstat -m'.  I see the memory
climb in the output of
> 'top' - it indicates that the "File" memory
usage is continually climbing
> when I have transfers happening.

Are these transfers coming from or going to a file system?

> 
> So the basic question is "what memory usage would
be considered 'File' memory
> by 'top' but wouldn't show up in 'vmstat -m'?". 
Of course, it's possible I'm
> just not seeing it in the 'vmstat -m' output, but I've
looked pretty hard!  

I'm pretty sure "File" is cached file system/block
device backing.

> 
> I have a feeling the problem is due to mismatched
"bus_dmamap_load..."/
> "bus_dmamap_unload..." operations, but
haven't been able to confirm that.
> 
> Any tips on tracking this down would be most
appreciated!
> 
> Thanks,
> 
> Patrick Klos

	Jonathan Kollasch