Hi Andrey, On Tue, Jun 19, 2018 at 7:33 PM Andrey Ryabinin <aryabi...@virtuozzo.com> wrote: > What's the status of CONFIG_SLUB_DEBUG in your config? > > AFAICS __kmem_cache_empty() is broken for CONFIG_SLUB_DEBUG=n. We use > slabs_node() there > which is always 0 for CONFIG_SLUB_DEBUG=n. > > The problem seems not limited to __kmem_cache_empty(), > __kmem_cache_shutdown() and __kmem_cache_shrink() > are also rely on correctness of the slabs_node(). Presumably this might cause > some problems while > destroying memcg kmem caches.
CONFIG_SLUB_DEBUG is not set in the crash I sent. Enabling it "fixes" the problem! This either means that KASAN+SLUB should enable SLUB_DEBUG, or the extra overhead from SLUB_DEBUG is just making the bug more rare but not actually eliminating it. Jason