This is the mail archive of the
systemtap@sources.redhat.com
mailing list for the systemtap project.
Re: Help with netlink needed
- From: Tom Zanussi <zanussi at us dot ibm dot com>
- To: Martin Hunt <hunt at redhat dot com>
- Cc: systemtap at sources dot redhat dot com, Tom Zanussi <zanussi at us dot ibm dot com>
- Date: Tue, 29 Mar 2005 13:23:35 -0600
- Subject: Re: Help with netlink needed
- References: <1112121244.5739.13.camel@tiger>
Martin Hunt writes:
> We really need to start getting the kernel <-> user space data transport
> code into the runtime library soon. I have not done any work with this,
> so help would be greatly appreciated.
>
> We have several immediate requirements:
>
> 1. High-speed transfer of large amounts of collected data. I was
> thinking of having 2 large trace buffers, one active and one inactive.
> When the active one fills up, it is marked inactive and a task is put on
> the work queue to send it. The inactive buffer becomes active and will
> be used while the inactive one is written to user-space. A command could
> also be provided to flush the active buffer.
>
> 2. A separate channel for immediate messages. Used to signal error
> conditions, print warnings, etc. Anything that is not normal trace
> output.
>
> Any thoughts? Code?
relayfs might be useful for 1), depending on how much data you
typically have to send to user space - it's sort of tailored for
applications that need to large amounts of data. It definitely
wouldn't be good for 2). The latest version of relayfs is in the -mm
tree, and there are example applications and documention here:
http://prdownloads.sourceforge.net/dprobes/relay-apps-0.2.tar.gz?download
Tom