System Calls for the Linux Security Module Infrastructure
summary
Linux introduces three new syscalls to replace messy /proc and securityfs interfaces: lsm_list_modules, lsm_self_get_attr and lsm_self_set_attr. They use a binary LSMCTX structure with a numeric LSM ID (IDs 1–99 reserved), flags (including LSM_FLAG_SINGLE), total and context lengths, and potentially binary data. get_attr returns contexts for every active LSM; set_attr accepts one attribute at a time to avoid rollback complexity. System calls give atomic ops, ABI stability, no special mounts, and enable LSM stacking and common user-space handling.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:15History of LSM
- 4:10LSM mechanisms
- 7:36The current conflict
- 12:39Moving to syscalls
- 15:57LSM context design
- 21:44New system calls
- 26:53Future considerations
- 32:25Conclusion
- 34:01Q&A session
related talks
23mStacked LSMs and User Space
15mSubsystem Update: LSM Stacking - What You Can Do Now and What's Next
33mSecurity Module Stacks That Don't Fall Over
41mThe State and Direction of LSM Stacking
30mSecurity Module Stacks that Don't Fall Over
01h 35mTutorial: How to Write a Linux Security Module
30mLSM Stacking - What You Can Do Now and What's Next
28mOverview of the Linux Kernel Security Subsystem