Count me, and most/all of the SystemTap team, among the folks who would like to see utrace accepted into Linus's kernel. Upstream acceptance seems to be bogged down on the following points: 1) On some architectures, ptrace still hasn't been successfully adapted to use utrace. 2) The current utrace patch set may be too big a mouthful for the kernel community to digest all at once.
This suggests an incremental approach to pushing utrace upstream. The enclosed email from Roland suggests that he's been thinking in those terms and may be open to such an approach. I'd like to (re)open this topic for ideas and suggestions. Plainly, if Roland can find time to elaborate on his ideas, that'd be a good starting place. For the SystemTap crowd, the utrace kernel API is what's most important. We're using it for kprobes-style tracing/probing of user apps (uprobes) and also user-space instruction tracing. Jim Keniston IBM Linux Technology Center -------- Forwarded Message -------- From: Roland McGrath <[EMAIL PROTECTED]> To: Jim Keniston <[EMAIL PROTECTED]> Cc: [EMAIL PROTECTED], Maneesh Soni <[EMAIL PROTECTED]>, Frank Eigler <[EMAIL PROTECTED]> Subject: Re: utrace refactoring... Date: Sun, 11 Nov 2007 16:17:17 -0800 (PST) > ... The utrace discussion on our [SystemTap team] call was fairly > brief, but I understood Frank to say that utrace received a lot of > attention at the recent Red Hat planning summit. Apparently the > coexistence idea came up at that meeting, and Roland felt strongly that > coexistence of legacy ptrace code with utrace isn't feasible. That's true. > I'm cc-ing Frank and Roland so they can correct me if I misunderstood. > This was the idea of getting past the current utrace/ptrace roadblocks > by allowing legacy (utrace-less) ptrace to remain, perhaps as a > configurable option, at least on architectures where the > ptrace-over-utrace work is not satisfactorily completed. I'm still not sure that can be accomplished exactly as just described. But I've been thinking more about strategies lately. I'm a bit more optimistic about some kinds of incremental approach than I was before. > We can move this discussion to the utrace list if you think it'd be > useful. Yes, please do. I think some other folks (who are hopefully on the list) have been keen on this in the past, when I was unfortunately more vigorously skeptical of the approach than I might have been. Thanks, Roland
