On 2025-02-27 22:04:16 [+0900], Sergey Senozhatsky wrote:
> On (25/02/27 21:42), Sergey Senozhatsky wrote:
> > > ach. Got it. What about
> > >
> > > | static void zram_slot_lock_init(struct zram *zram, u32 index)
> > > | {
> > > | static struct lock_class_key __key;
> > > |
> > > | lockdep_init_map(slot_dep_map(zram, index),
> > > | "zram->table[index].lock",
> > > | &__key, 0);
> > > | }
> > >
> > > So every lock coming from zram belongs to the same class. Otherwise each
> > > lock coming from zram_slot_lock_init() would belong to a different class
> > > and for lockdep it would look like they are different locks. But they
> > > are used always in the same way.
> >
> > I see. I thought that they key was "shared" between zram meta table
> > entries because the key is per-zram device, which sort of made sense
> > (we can have different zram devices in a system - one swap, a bunch
> > mounted with various file-systems on them).
Yes. So usually you do spin_lock_init() and this creates a key at _this_
very position. So every lock initialized at this position shares the
same class/ the same pattern.
> So the lock class is registered dynamically for each zram device
>
> zram_add()
> lockdep_register_key(&zram->lock_class);
>
> and then we use that zram->lock_class to init zram->table entries.
>
> We unregister the lock_class during each zram device destruction
>
> zram_remove()
> lockdep_unregister_key(&zram->lock_class);
>
> Does this still put zram->table entries into different lock classes?
You shouldn't need to register and unregister the lock_class. What you
do should match for instance j_trans_commit_map in fs/jbd2/journal.c or
__key in include/linux/rhashtable.h & lib/rhashtable.c.
At least based on my understanding so far.
Sebastian
