Linux Developers Look At Dropping SLOB
The Simple List Of Blocks "SLOB" memory allocator has been around a long time as an alternative to the SLAB and SLUB allocators. Linux developers are also considering the possibility of dropping SLAB in the future too and just leaving SLUB, but as a first target they are looking at dropping SLOB. SLOB is known for its high fragmentation though has stood up well at times on embedded systems.
Dropping SLOB is being considered in the name of the code maintenance burden, some features only compatible with the other allocator(s), blocking API improvement work within the kernel, and SLUB memory requirements can be potentially lowered to that of SLOB if needed.
The discussion over deprecating and removal of SLOB can be found via this mailing list thread. Other upstream developers so far are in favor of deprecating and removing SLOB while the removal of SLAB may take more time. It's likely there will be a period of a few kernel releases where SLOB is marked as broken or otherwise prominently indicated that it is deprecated before ultimately being stripped out of the kernel source tree.
13 Comments
