This is the mail archive of the 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: pre-compiled systemtap modules (try 2)

On Tue, 2006-09-12 at 22:06 -0400, Martin Hunt wrote:
> On Tue, 2006-09-12 at 16:39 -0500, David Smith wrote:
> > 1) Command-line interface.  To make this easy for a user to do, we need
> > to modify the stap command to save a kernel module and run an existing
> > one.  I've attached an updated patch that modifies the systemtap front
> > end to take 2 new options:
> > 
> >    -S DIR     Save the compiled module in DIR
> >               Run pre-compiled module
> Instead of "-P", why not have stap just realize it's been given a module
> and it automatically does the right thing?

Hmm.  You do have a point here, because with the patch it is possible to
do something like stupid like "stap -P foo.ko bar.stp".

How would you suggest stap realizing it has been given a module instead
of a script?  Looking at the file extension or something else?

> > (The original patch always saved the compiled module in '.', but Josh
> > Stone liked being able to specify a directory.  I've changed the patch
> > to reflect this.)
> Why not "-S [dir]" so that the current directory is the default?

That is certainly possible, but I'm not sure I see the advantage of "-S"
over "-S .".  It is only one character after all.

> > 2) Differing options between compile (stap) and run time (stpd).  I
> > believe that most of them should be OK, with the exception of the stap
> > '-b' relayfs option (which turns into the '-r' stpd option).  If the '-
> > b' option was specified at compile time and not at run time (or vice
> > versa), I'm not sure what will happen (since that flag appears to change
> > behavior both at compile time and run time).
> "stpd -r" just mounts the relayfs filesystem if necessary.  When loading
> a module, it known if it is using relayfs or not. It cannot be changed
> at runtime.
> > Martin Hunt was going to look into remove '-r' from stpd, but hasn't had
> > a chance yet that I know of.
> It seems obvious enough the option should be removed and the mount done
> automatically, when required.  I'll get it checked in. 


David Smith
Red Hat, Inc.
256.217.0141 (direct)
256.837.0057 (fax)

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