On (25/01/24 20:03), Sergey Senozhatsky wrote: > On (25/01/24 18:30), Hillf Danton wrote: > > On (25/01/22 14:57), Sergey Senozhatsky wrote: > > > > > > - for (index = 0; index < num_pages; index++) > > > - spin_lock_init(&zram->table[index].lock); > > > + for (index = 0; index < num_ents; index++) > > > + init_rwsem(&zram->locks[index].lock); > > > > Curious if lockdep trick [1] is needed here. > > These bucket locks are not part of the v2 which I'm currnetly > working on. v2 will also come with a draft version of new zsmalloc API that does not impose atomicity restrictions in zs_map_object() (no local CPU lock and no migration rwlock involved.)
