DTrace Hangout 07 July 2016

Robert talks discusses the new audit provider. This work is likely to be committed After it gets tested within the CADETS project.
Cambridge continues to move forwards on integrating Python with DTrace for the L41 course.
George has a phabricator (D7144) for Graeme's varargs bug fix. This will go in after the freeze and be MFC'd. Mark has commented on the review with another site that needs to be updated.
Robert sat in on a PhD viva that relates to provenance but which may also relate to DTrace and tracing in FreeBSD.
We don't install the CDDL headers on FreeBSD but these are needed by the Python. Mark will put up a patch to address this.
The Juniper folks want to apply the anonymous tracing to apply to 10 as well as 11. Mark believes the code can be MFC'd but will have to look into this.
Robert has created UUIDs for the kernel that can be used in DTrace as well. He would like it to be pretty printed.
George wants to expand all of the storage sizes in DTrace to more reasonable 2016 values. Mark suggests that we add these as sysctls.
Logging directly from the kernel is not likely to work as some action end in user space.
Issues list. [gnn]
Add a BPF provider? Also perhaps an mcopy() DIF instruction to pull data out of an mbuf chain. Can we provide BPF filtering as a predicate?
How does DTrace find pid probes? This is related to making dynamic probes for things like tcpdump/BPF.