Re: Revisiting DTrace syscall provider

[Ryota Ozaki <ozaki-r%netbsd.org@localhost>]

On Tue, Feb 24, 2015 at 4:00 AM, bch <brad.harder%gmail.com@localhost> wrote:
> Are there plans (and roadmap?) to user-space dtrace abilities in
> NetBSD -- currently we're not exporting a certain header (and required
> code support?) for userland dtrace support; kernel-only, as I
> understand.
>
> Is my understanding correct?

Yes, NetBSD now supports only tracing inside kernel activities.

AFAIK, nobody is working on the user-space support for now.

  ozaki-r

>
> -bch
>
>
> On 2/23/15, Ryota Ozaki <ozaki-r%netbsd.org@localhost> wrote:
>> Hi,
>>
>> I'm trying DTrace syscall provider (systrace) patch
>> provided by riz some years ago(*) to complete the work.
>> (I'm already ack-ed the handover by him.)
>>
>> (*) https://mail-index.netbsd.org/tech-kern/2013/06/24/msg015296.html
>>
>> I needed some tweaks to make it work with -current,
>> but it works (on amd64) anyway.
>>
>> Noticeable changes from the riz's patch are
>> - Move systrace_probe_func definition to
>>   sys/kern/kern_dtrace.c from sys/arch/amd64/amd64/trap.c
>>   to support architectures other than amd64
>> - Make it buildable for i386 and arm
>>   (some fixes to dtrace itself are already committed)
>>
>> A known todo is to support syscall emulations, although
>> it can be done later. I think only native syscall support
>> is still worthwhile to merge.
>>
>> Here is a patch:
>> http://www.netbsd.org/~ozaki-r/systrace-full.diff
>>
>> Er...the patch is too big to look at because it includes
>> mechanically generated *sysent.c and *systrace_args.c.
>> So I prepare another patch that excludes them:
>> http://www.netbsd.org/~ozaki-r/systrace.diff
>>
>> And here is a git branch. Please use it if you want to
>> see the changes as individual commits:
>> https://github.com/ozaki-r/netbsd-src/commits/dtrace-syscall-provider
>>
>> Regards,
>>   ozaki-r
>>