Subject: Re: what is the status of the ktrace-lwp branch?
To: None <tech-kern@netbsd.org>
From: Christos Zoulas <christos@astron.com>
List: tech-kern
Date: 10/30/2005 05:01:59
In article <20051029210343.99717564FA@rebar.astron.com>,
Christos Zoulas <christos@zoulas.com> wrote:
>On Oct 29,  8:21pm, skrll@netbsd.org (Nick Hudson) wrote:
>-- Subject: Re: what is the status of the ktrace-lwp branch?
>
>| On Saturday 29 October 2005 18:21, Christos Zoulas wrote:
>| > It would be nice to merge it in at somepoint.
>| 
>| Two things needs to be resolved
>| 
>| 1) the format of struct ktr_header - chuq made some suggestions to make it MI
>|    (iirc)
>| 
>| 2) the uio_proc -> uio_lwp change in struct uio. chuq wasn't keen. I traced
>|    through the reason behind this and its difficult not to do it.
>
>Can we please have a summary here? I am almost done merging it to head
>(~70 files with conflicts).

Ok, now that I've finished merging it, and have it all working including
userland changes I understand a little better what's going on here:

1. We could encode the version inside the type field to avoid the endianness
   issue, but I don't see this as a big deal. I think I prefer what nick has
   now.
2. I understand where chuq is coming from and I think I agree with him. I
   think that the uio_ is a process wide struct and not an lwp specific one.
   For example, what happens when the lwp exits before the uio is completed?
   On the other hand, keeping it as uio_proc, would require us to call
   proc_representative_lwp(uio->uio_proc) in a bunch of places; not very
   attractive either.

This is a fairly large patch and each time we let it go stale it requires
a lot of work to get it going again (this merge took me almost 4 hours).
We should decide what we want to do for those two things and commit the
patch sooner than later.

christos