On Sun, Jun 27, 2021 at 02:33:31PM +0300, Leon Romanovsky wrote: > On Sun, Jun 27, 2021 at 10:40:43AM +0000, Haakon Bugge wrote: > > > > > > > On 27 Jun 2021, at 12:31, Leon Romanovsky <leon@xxxxxxxxxx> wrote: > > > > > > On Sun, Jun 27, 2021 at 12:17:51PM +0530, Anand Khoje wrote: > > >> This v6 patch series is used to read the port_attribute subnet_prefix > > >> from a valid cache entry instead of having to call > > >> device->ops.query_gid() for Infiniband link-layer devices in > > >> __ib_query_port(). > > >> > > >> In the event of a cache update, the value for subnet_prefix gets read > > >> using device->ops.query_gid() in config_non_roce_gid_cache(). > > >> > > >> Anand Khoje (2): > > >> IB/core: Updating cache for subnet_prefix in > > >> config_non_roce_gid_cache() > > >> IB/core: Read subnet_prefix in ib_query_port via cache. > > > > > > This series breaks mlx4/mlx5. You forgot to call to lock_init or > > > something like that. > > > > Thanks for catching! > > > > Sure, in ib_register_device(), setup_device() (which ends up calling __ib_query_port()) is called before ib_cache_setup_one(). Can these two calls have their order swapped? > > I don't think so, if I didn't miss anything, we are relying in gid_table_setup_one() > on some properties from setup_device(). Just reorder things enough so that the cache_lock is setup earlier, it has no business being in cache_setup_one anyhow. Jason
