Hide Forgot
Description of problem: RHEL7 has replaced its traditional memory allocator from CONFIG_SLAB to CONFIG_SLUB. The "kmem -S" command displays statistics about the objects in each per-cpu kmem_cache_cpu slab and about all of the per-node, partially-used slabs. A change in the manner of bookkeeping was introduced in Linux 3.1 that causes the object and in-use count of those slabs to display incorrect data. Version-Release number of selected component (if applicable): crash-7.0.2-2.el7 How reproducible: Always. Steps to Reproduce: 1. Run "kmem -S" command. 2. 3. Actual results: For example, the slabs in the mm_struct kmem_cache each have 15 total objects per-slab, but the "NODE 0 PARTIAL" and "NODE 1 PARTIAL" listings show slabs with "TOTAL" values that are equal to the "ALLOCATED" values, and as a result, show "FREE" values of 0: crash> kmem -S mm_struct ... [ cut ] ... KMEM_CACHE_NODE NODE SLABS PARTIAL PER-CPU ffff88022709b400 0 5 5 0 NODE 0 PARTIAL: SLAB MEMORY NODE TOTAL ALLOCATED FREE ffffea0008713e00 ffff88021c4f8000 0 2 2 0 ffffea000877e000 ffff88021df80000 0 0 0 0 ffffea000874b400 ffff88021d2d0000 0 0 0 0 ffffea0008742200 ffff88021d088000 0 5 5 0 ffffea000878ba00 ffff88021e2e8000 0 7 7 0 NODE 0 FULL: (empty) KMEM_CACHE_NODE NODE SLABS PARTIAL PER-CPU ffff880426807400 1 4 4 0 NODE 1 PARTIAL: SLAB MEMORY NODE TOTAL ALLOCATED FREE ffffea00107a6e00 ffff88041e9b8000 1 4 4 0 ffffea0010802000 ffff880420080000 1 0 0 0 ffffea00107d5800 ffff88041f560000 1 3 3 0 ffffea00107cbc00 ffff88041f2f0000 1 5 5 0 NODE 1 FULL: (empty) crash> Expected results: The TOTAL values should all be 15, with corrected FREE counts. Additional info:
This request was resolved in Red Hat Enterprise Linux 7.0. Contact your manager or support representative in case you have further questions about the request.