This patchset resolves some issues I've come across while investigating nft hash sets. The first patch requires users of rhashtable to define a max_shift as suggested by Daniel Borkmann and Thomas Graf. One side-effect of not setting max_shift is that tables are not allowed to expand. I was observing this behavior with nft hash sets prior to these changes. The next patch implements this requirement for nft hash sets using desc->size as the max_shift if it's provided by the user. If not, it falls back to a newly defined default of 1024 elements. This value is somewhat arbitrary, but seems like a reasonable default to me. I used 'size' above for max_shift because it appears to be used as a ceiling for the number of elements in a set in nft_add_set_elem(). It's also used in the estimate fn. Prior to the next patch 'size' was also being used as the nelem_hint to pass to rhashtable_init(). This seems incorrect since nelem_hint is meant to provide a hint for how many hash buckets to initially allocate. Instead of using 'size' for nelem_hint, the final patch introduces a new set parameter named 'init_size'. If this approach is acceptable I can provide the userspace patches to fully implement the new parameter. The patchset is against net-next. Josh Hunt (3): rhashtable: require max_shift definition nft_hash: define max_shift rhashtable param nft_hash: introduce init_size set parameter include/net/netfilter/nf_tables.h | 4 +++- include/uapi/linux/netfilter/nf_tables.h | 2 ++ lib/rhashtable.c | 3 ++- net/netfilter/nf_tables_api.c | 4 ++++ net/netfilter/nft_hash.c | 7 ++++++- 5 files changed, 17 insertions(+), 3 deletions(-) -- 1.7.9.5 -- To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
