Re: libmemstat(3) - Library for monitoring kernel memory use

    0n Tue, Jun 06, 2006 at 09:11:10AM +0100, Robert Watson wrote: 

    >
    >On Tue, 6 Jun 2006, Wilkinson, Alex wrote:
    >
    >>   0n Tue, Jun 06, 2006 at 07:20:39AM +0100, Robert Watson wrote:
    >>
    >>   >
    >>   >On Tue, 6 Jun 2006 bhuvan.kumarmital_at_wipro.com wrote:
    >>   >
    >>   >>   Saw your tool (memtop) for monitoring kernel memory. I'd like to 
    >>   use a
    >>   >>similar tool for linux, i believe your tool is bsd based. Could you 
    >>   tell
    >>   >>me a similar tool, or perhaps another version of memtop built for 
    >>   linux.
    >>   >>I'd really appreciate you help. Please reply on my email address.
    >>   >
    >>   >You are correct that libmemstat and derived tools currently rely on
    >>   >features present in the FreeBSD kernel.  The library provides a general
    >>   >monitoring abstraction over a set of specific kernel memory allocators 
    >>   --
    >>   >specifically, the FreeBSD malloc(9) and uma(9) allocators.  It is
    >>   >relatively straight forward to implement that abstraction for other 
    >>   memory
    >>   >allocators, such as user space allocators or kernel allocators from 
    >>   other
    >>   >platforms, but that work has not been done (as far as I know).  I'm not
    >>   >aware of specific monitoring tools for the Linux operating system that 
    >>   are
    >>   >able to perform this type of profiling/monitoring, although I presume 
    >>   some
    >>   >sort of kernel memory profiling tool does exist.
    >>
    >>Erm, Robert, where does memtop live ? I can find it in ports nor base 
    >>system.
    >
    >memtop is an experimental monitoring tool based on libmemstat, you can find 
    >the source here:
    >
    >    http://www.watson.org/~robert/freebsd/libmemstat/
    >
    >Possibly something like this could be integrated into systat, but my 
    >ncurses knowledge is a bit weak, and I've not had a chance to investigate 
    >further. As with vmstat, the interpretation of the output requires a 
    >moderate amount of insight into how the kernel works, so I've been a bit 
    >reluctant to push it as a debugging tool without some more refinement.  I 
    >think it would be neat if someone picked it up and did something useful 
    >with it, though :-).

I assume this only works with -CURRENT ?

 -aW