Re: [Qemu-devel] [RFC 0/3] Tracing framework for QEMU

[Prerna Saxena]

Hi Stefan,


Interesting to see your patches, tracepoint definitions/declarations
look similar to in-kernel tracepoints :).


The choice of a kernel-tracepoint approach had advantages since it 
places no inherent limitation on maximum number of arguments that can be 
logged. Also, it is flexible enough to dump entire arrays if need be -- 
which is not as easily accomplished in the proposed approach as with 
kernel tracepoints.


--
Prerna Saxena

Linux Technology Centre,
IBM Systems and Technology Lab,
Bangalore, India

Re: [Qemu-devel] [RFC 0/3] Tracing framework for QEMU

[Stefan Hajnoczi]

Interesting to see your patches, tracepoint definitions/declarations
look similar to in-kernel tracepoints :).

Please post future patches inline to the email so reviewing and
replying is easy (e.g. use git-send-email to send patches).

Stefan

[Qemu-devel] [RFC 0/3] Tracing framework for QEMU

[Prerna Saxena]

Hi all,
The following patchset includes a simple implementation for qemu 
tracing. This introduces a framework for :

i) Internal buffers for QEMU, and API for logging traces therein.
	ii) Tracepoint framework that logs traces to the buffer and also 
interprets already logged traces.
	iii) Monitor commands to display logged traces in tracepoint-specific 
formats, and also for enabling/disabling tracepoints individually.


This is *work in progress*. There are known issues that I'm chasing, 
which includes segfault that happens while transitioning from monitor to 
guest with tracing enabled. (Will appreciate any pointers if I'm missing 
the obvious somewhere :-))


Stefan,
	Thanks for your ideas on tracing ! I'm just posting out a rough cut of 
whatever I have ready ; We can take the best from both your patches and 
mine and have something running for qemu tracing. I'll take care of 
merging pieces together.


Looking forward to suggestions..
--
Prerna Saxena

Linux Technology Centre,
IBM Systems and Technology Lab,
Bangalore, India