This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: stpd issues


On Tue, 2005-08-23 at 14:06 -0700, Hien Nguyen wrote:
> Martin Hunt wrote:
> 
> >Relayfs data arrives in big per-cpu chunks. As the chunks arrive they
> >are saved to per-cpu datafiles.  When data collection stops (after ^C),
> >stpd merges all the datafiles together and writes to stdout. It doesn't
> >really make sense to stream data to stdout as it arrives for relayfs
> >because it will be out-of-order.
> >  
> >
> Thanks for your explanation. Anyway, what is the status of the stpd? I 
> just update my src with the latest in CVS, stpd hangs my system when I 
> press ctrl-c (running the same systrace module). It was just the stpd 
> hung, and I could kill it.

That definitely should not happen. 
You have this fix?
2005-08-22  Martin Hunt  <hunt@redhat.com>

        * transport.h: Don't define _stp_transport_write.
        * transport.c (_stp_transport_write): Use
        _stp_transport_send() when in probe_exit().

I have some more bug fixes I expect to check in later today. One is a
buffer aggregation scheme that will eliminate the infinite loop caused
by each sys_read() triggering a probe on sys_read that causes another
sys_read(). Another fixes a relayfs problem. I have been testing it all
day on extreme cases and fixing any problems.  If you still have
problems after I check that in, we need to investigate.

Martin



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]