> > I will get started on the groundwork patches and do x86 to set the > > example. I think I can get this going in the next few days (next few > > business days anyway, not sure about the holiday weekend right now). > > Looking forward to this...
See the thread starting at http://lkml.org/lkml/2007/11/25/75 for what I've just submitted. This covers step 1 for x86 and powerpc. Note the highly incremental approach, and how I took the opportunity to clean up the arch code along the way. All this amount of detailed work is not really required for each arch (27 patches just for the easiest step!). But it shows how there is a lot of cruft in the arch code that can be cleaned up. When we do it in little steps, it is more likely that it will be integrated upstream, and sooner. (The x86 patches I posted upstream last week relating to TLS are part of the same cleanup and groundwork too. Now those are already integrated in the x86 maintainers' staging tree.) > At what step do you envisage the core utrace bits (engine, etc), be > merged? From a pure utrace client (!ptrace) perspective, it'd be > important to be able to do things as attach an engine, quiesce a thread > and the like. Of course. That's a large subject, really. The point of the stage of the work I've discussed so far is to get the arch porting issue out of the way of that. This will get us to a place where the gatekeepers of core kernel code are the only people preventing the incorporation of the fundamental work. There will be plenty of work to do on that front too. I have more detailed thoughts on that, but they are not fully-fleshed. I don't want to focus on that until we have really gotten the ball rolling on the arch work. The question I hope many people will have is, "What can I do right away to help get this arch work done and merged?" The answer is to be an arch advocate for this cleanup work. (There is no need even to refer to utrace per se in upstream discussion of this stage.) I've got the ball rolling with step 1 for x86 and powerpc. The basis for the other steps I laid out will be out in a few more days. I will be poking each person who has worked on utrace arch support. But even for x86 and powerpc, there is no reason that I need to be the person driving the arch part of this work. Chime in upstream with the arch maintainers. Take charge of getting a cleanup patch integrated, updating it as needed from upstream feedback, and keeping an eye on it in the arch code. Another answer is to do the more formal write-up of a howto for arch maintainers, as I suggested (working from utrace-arch-porting-howto.txt and my postings). The answer to, "Are we there yet?" is, "No." The answer to, "When will we get there?" is, "When you get out and push." Thanks, Roland
