Ktap almost gets into 3.13

Posted Nov 7, 2013 13:14 UTC (Thu) by deater (subscriber, #11746)
In reply to: Ktap almost gets into 3.13 by karim
Parent article: Ktap almost gets into 3.13

> Seriously, making this part of perf makes it much less likely
> that I'll use it.

As long as Ingo maintains perf it's unlikely that any other profiling tool will get into the kernel unless it's through perf_event, even if it doesn't make sense.

It's why things like AMD Lightweight Profiling never made it into the kernel

And it's why things like ftrace tracing are part of perf, complicating the interface, which has led to recent bugs where a normal user can lock the kernel using the perf interface.

to post comments

Ktap almost gets into 3.13

Posted Nov 7, 2013 14:24 UTC (Thu) by karim (subscriber, #114) [Link]

Yes, I'm fully aware of that. And while I respect Ingo's work a great deal, there are areas where it's my personal belief that his well-intentioned stewardship of key functionality is actually counter-productive. Specifically, his clout makes it so that any suggestion that runs counter to his own designs results in that proposal to be doomed to failure.

At this stage many of those that have tried are staying away. Personally I've given up on Linux ever providing meaningful online diagnostic to rival any of the other OSes out there. There's just too much kernel politics in the way. This recent idea that all new performance/monitoring stuff has to go through perf craziness is just one more on the list.

FWIW I think ftrace's user interface is kickass. It's not perfect and there are other ways of doing things, but ftrace certainly shows that it was architectured based on deep, real-life experience with tracing lots of data. Smarts just aren't enough.