[PATCH 10/17] hyper-v: trace vmbus_open()

Greg KH gregkh at linuxfoundation.org
Sat Nov 4 11:05:20 UTC 2017


On Wed, Nov 01, 2017 at 05:43:08PM +0000, KY Srinivasan wrote:
> 
> 
> > -----Original Message-----
> > From: devel [mailto:driverdev-devel-bounces at linuxdriverproject.org] On
> > Behalf Of Steven Rostedt
> > Sent: Tuesday, October 31, 2017 6:10 AM
> > To: Greg KH <gregkh at linuxfoundation.org>
> > Cc: olaf at aepfle.de; Stephen Hemminger <sthemmin at microsoft.com>;
> > jasowang at redhat.com; linux-kernel at vger.kernel.org; apw at canonical.com;
> > marcelo.cerri at canonical.com; devel at linuxdriverproject.org; Vitaly
> > Kuznetsov <vkuznets at redhat.com>; leann.ogasawara at canonical.com
> > Subject: Re: [PATCH 10/17] hyper-v: trace vmbus_open()
> > 
> > On Tue, 31 Oct 2017 13:48:00 +0100
> > Greg KH <gregkh at linuxfoundation.org> wrote:
> > 
> > > > I don't see how that information can be extracted easily without a
> > > > tracepoint here. Am I missing something?
> > >
> > > Wasn't one of the outcomes of the conference last week the fact that for
> > > ftrace + ebpf we could get access to the structures of the function
> > > parameters?  Or that work would soon be showing up?
> > 
> > I told Linus that I'll start building an infrastructure on function
> > tracing to see what we can do. But it may be very limited in features.
> > I don't believe eBPF can follow arbitrary data structure pointers
> > without helper functions. Which doesn't exist for this type of code yet.
> > 
> > >
> > > It just feels "wrong" to add a tracepoint for a function call, like it
> > > is a duplication of work/functionality we already have.
> > 
> > We don't already have it. We may have something in a year (or two) that
> > may be able to get all the data that is requested here. But it's going
> > to take lots of RFC patch sets and brain storming to come up with
> > something that everyone is satisfied with.
> > 
> > In other words, the functionality is currently in vaporware state.
> 
> Greg,
> 
> The added memory overhead is very minimal, and the added tracing support is extremely useful for
> our debugging. If you don't mind, I would like to have this tracing support.

Ok, it's your subsystem, you get to handle the long-term maintenance of
these :)

Now merged,

greg k-h


More information about the devel mailing list