Thursday 17 November 2011

Dtrace .. new release

I just put out a new release. Hopefully this fixes a number
of stability issues and some silly bugs. I have rewritten the mutex
code to decouple from the kernel, and avoid reentrancy issues.

Just as I was about to release this a few days back I found a regression
where on repeated driver reloads, the kernel would crash or complain
of interrupts being disabled when they shouldnt.

I spent ages trying to drill down to where I had screwed up, and
in the end, it was a really stupid typo (sizeof(mp) vs sizeo(*mp)).

Glad to fix that and get back to a degree of sanity.

So this release has been tested exhaustively on Fedora Core 15 and
Ubuntu 11.10 (Linux kernel 3.0). I may have broken the build on
earlier kernels, and am looking to try and test on the earlier releases.

Please give it a try.

As an aside, I did get an email from someone trying out Dtrace on an
Oracle Linux host, and getting some strange complaints when /usr/bin/dtrace
didnt like the command line. /usr/bin/dtrace is *Oracle*s dtrace, not mine,
and hopefully the person is in a better position now running a working
dtrace on the system.

Lets see how much damage people can do to this. Many thanks to Nigel Smith
who kept harping on at me for "this doesnt work" which lead me to deep dive
a number of corner cases.

I have temporarily disabled some aspects of pid tracing/shadowing
and will need to fix that before evaluating the whole pid tracing space.


Post created by CRiSP v10.0.17a-b6103


No comments:

Post a Comment