Reducing the Kernel's Slab Allocators: Progress Report
summary
Proposal to remove duplicate allocators and standardize on SLUB, with SLOB already deprecated and removed in 6.4-rc1 removes SLOB. Reports show SLUB typically works but has ~30% memory overhead in some tests, so work targets reducing that overhead rather than keeping multiple allocators. A slab_tiny build option is added to satisfy tiny 8MB devices (fixed an OpenWRT regression), mm stress tests by Mel Gorman and recent LSF/MM discussion find no blocking regressions.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:03Linux Kernel Slab Allocators
- 13:43SLAB vs SLUB: similarities and differences
- 24:05Disadvantages of multiple implementations
- 30:28The SLOB allocator removal
- 33:17Next: SLAB allocator removal?
related talks
25mThe slab allocators of past, present, and future
01h 31mSLUB Internals for Exploit Developers
05mSlab allocators
44mHands Off and Putting SLAB/SLUB Feng Shui in a Blackbox
24mLow-overhead memory allocation tracking
42mAllocators for Compressed Pages: In-depth Comparison with z3fold in Focus
28mRefining Data Structure Implementations in the Linux Kernel for Improve..
21mCgroup Slab Memory Controller and Time Namespace