gwydirsam / bsd-coredump-history

the history, present state and some ideas for the future of the core dump routines in FreeBSD

The History and Future of Core Dumps in FreeBSD

This paper details the history, present state and some ideas for the future of the core dump routines in FreeBSD. The history follows FreeBSD back through BSD to 5th Edition Research UNIX. This project would have not been possible without unix-history-repo.

This started as a talk proposal to AsiaBSD Con 2017 and BSDCan 2017 about my wants in FreeBSD but the research got out of hand and became a complete history on the topic.

1 Abstract

Crash dumps, also known as core dumps, have been a part of BSD since it’s beginnings in Research UNIX. A core dump is “a copy of memory that is saved on secondary storage by the kernel” for debugging a system failure¹. Though 38 years have passed since doadump() came about in UNIX/32V, core dumps are still needed and utilized in much the same way they were then. Given this, one might assume the core dump code changed little over time but, with some research, this assumption has proven incorrect.

What has changed over time is where core dumps are sent to and what processor architectures are supported. Previous to the advent of UNIX, core dumps were printed to a line printer or punch cards. At the birth of UNIX core dumps were made to magnetic tape and because UNIX only supported the PDP-11, it was the only architecture supported for dumps. Over time machine architecture support has evolved from different PDP-11 models to hp300, i386 up to the present day with AMD64 and ARM64. In addition the type of dump device has changed from tape, to hard disk or another machine over a LAN.

The following paper begins with a quick background on what core dumps are and why operators might need them. Following that the current state of the core dump facility and some of the more common extensions in use are examined. We conclude with a call to action for upstreaming these extensions and modularizing the core dump code such that different methods of core dump can be dynamically loaded into the kernel on demand.

In addition a complete history of core dumps in UNIX and BSD was produced as research for this paper and can be found in the appendix.

2 ‘TODO’ list

2.1 DONE Create Slides

2.2 DONE Outline Slides

2.3 DONE Flesh Out Slides

2.4 DONE Add backtrace.io features

2.5 DONE Add how to force a dump

2.6 DONE Clean up notes

2.7 DONE Finish Textdump Procedure and Contents

2.8 DONE Revise the details of the minidump procedure

2.9 DONE Add Solaris?

• Does this make sense (it’s no longer BSD based)? (no)
• It does have a ton of features I want in FreeBSD. (illumos has all of them)

2.10 DONE Add Illumos

• Could possibly replace Solaris section (it did)

2.11 DONE Finish Modularization of Core Dump Code section

2.12 TODO minidumpsz code needs publishing so I can reference it

2.13 TODO Finish Conclusion

2.14 DONE Submit Finished Paper to AsiaBSD Con 2017

2.15 DONE Submit to BSDCan

2.16 TODO Add Sandvine (projects/sv) Branch to history

2.17 DONE Add description of dumpon(8)

2.18 DONE Add thanks:

2.18.1 DONE Allan Jude - spelling

2.18.2 DONE Daniel Nowacki - spelling, grammar and syntax

2.18.3 DONE Chris Findeisen - phrasing suggestions