On 2025-02-22 07:25:42 [+0900], Sergey Senozhatsky wrote:
> index 1424ee73cbb5..03710d71d022 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c
> @@ -226,6 +226,9 @@ struct zs_pool {
> /* protect zspage migration/compaction */
> rwlock_t lock;
> atomic_t compaction_in_progress;
> +#ifdef CONFIG_DEBUG_LOCK_ALLOC
> + struct lock_class_key lock_class;
> +#endif
No ifdef
> };
>
> static inline void zpdesc_set_first(struct zpdesc *zpdesc)
> @@ -279,6 +294,93 @@ struct mapping_area {
> enum zs_mapmode vm_mm; /* mapping mode */
> };
>
> +#ifdef CONFIG_DEBUG_LOCK_ALLOC
> +#define zsl_dep_map(zsl) (&(zsl)->dep_map)
> +#define zspool_lock_class(pool) (&(pool)->lock_class)
> +#else
> +#define zsl_dep_map(zsl) NULL
> +#define zspool_lock_class(pool) NULL
> +#endif
> +
> +static void zspage_lock_init(struct zspage *zspage)
> +{
> + struct zspage_lock *zsl = &zspage->zsl;
> +
> + lockdep_init_map(zsl_dep_map(zsl), "zspage->lock",
> + zspool_lock_class(zspage->pool), 0);
> + spin_lock_init(&zsl->lock);
> + zsl->cnt = ZS_PAGE_UNLOCKED;
> +}
> +
> +/*
> + * The zspage lock can be held from atomic contexts, but it needs to remain
> + * preemptible when held for reading because it remains held outside of those
> + * atomic contexts, otherwise we unnecessarily lose preemptibility.
> + *
> + * To achieve this, the following rules are enforced on readers and writers:
> + *
> + * - Writers are blocked by both writers and readers, while readers are only
> + * blocked by writers (i.e. normal rwlock semantics).
> + *
> + * - Writers are always atomic (to allow readers to spin waiting for them).
> + *
> + * - Writers always use trylock (as the lock may be held be sleeping readers).
> + *
> + * - Readers may spin on the lock (as they can only wait for atomic writers).
> + *
> + * - Readers may sleep while holding the lock (as writes only use trylock).
> + */
> +static void zspage_read_lock(struct zspage *zspage)
> +{
> + struct zspage_lock *zsl = &zspage->zsl;
> +
> + rwsem_acquire_read(zsl_dep_map(zsl), 0, 0, _RET_IP_);
> +
> + spin_lock(&zsl->lock);
> + zsl->cnt++;
> + spin_unlock(&zsl->lock);
How is this working given that a read_lock always increments the
counter? If it is write_locked then a read_lock makes it UNLOCKED.
migrate_read_lock() did block if a writer was pending, this does not or
at least it is not obvious how.
> + lock_acquired(zsl_dep_map(zsl), _RET_IP_);
> +}
> +
Sebastian
