On Thu, Jul 27, 2023 at 9:23 AM Johannes Weiner <hannes@xxxxxxxxxxx> wrote:
>
> Removing a zswap entry from the tree is tied to an explicit operation
> that's supposed to drop the base reference: swap invalidation,
> exclusive load, duplicate store. Don't silently remove the entry on
> final put, but instead warn if an entry is in tree without reference.
>
> While in that diff context, convert a BUG_ON to a WARN_ON_ONCE. No
> need to crash on a refcount underflow.
>
> Signed-off-by: Johannes Weiner <hannes@xxxxxxxxxxx>
I have always found it confusing that we explicitly remove the zswap
entry from the entry in the contexts you mentioned, yet we have
zswap_rb_erase() called in zswap_entry_put(). In fact, I think in some
contexts this leads to zswap_rb_erase() being called unnecessarily
twice on the same entry (e.g. once from invalidation, then once again
when an outstanding local ref is dropped). It's probably harmless with
the current implementation, but such a design can easily go wrong.
Thanks for the cleanup, it would be interesting to see if this warning
is actually fired.
Reviewed-by: Yosry Ahmed <yosryahmed@xxxxxxxxxx>
> ---
> mm/zswap.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/mm/zswap.c b/mm/zswap.c
> index e123b1c7981c..e34ac89e6098 100644
> --- a/mm/zswap.c
> +++ b/mm/zswap.c
> @@ -402,9 +402,9 @@ static void zswap_entry_put(struct zswap_tree *tree,
> {
> int refcount = --entry->refcount;
>
> - BUG_ON(refcount < 0);
> + WARN_ON_ONCE(refcount < 0);
> if (refcount == 0) {
> - zswap_rb_erase(&tree->rbroot, entry);
> + WARN_ON_ONCE(!RB_EMPTY_NODE(&entry->rbnode));
> zswap_free_entry(entry);
> }
> }
> --
> 2.41.0
>
