Re: The Slab Allocator (Was: Re: New dcache not using slab allocator?)

David S. Miller (
Tue, 5 Aug 1997 13:44:22 -0400

Can you implement front ends to SLAB which need not grab any locks nor
disable interrupts? This is why the per-cpu magazine aspect of SIMP
was a no brainer for Thomas to implement, and why it is perfect for
inode and dentry allocation, outside of the fragmentation issues you

If you can make SLAB have these wins as well (hard to do without
entirely different entry points for free/alloc and also
try_to_free_page() liberation) then SIMP is indeed not worth while,
but until such time it is and should stay.