Re: [PATCH v2 2/2] io_uring: Add KASAN support for alloc_caches

[Jens Axboe]

On 2/22/23 11:00?AM, Breno Leitao wrote:
> -static inline struct io_cache_entry *io_alloc_cache_get(struct io_alloc_cache *cache)
> +static inline struct io_cache_entry *io_alloc_cache_get(struct io_alloc_cache *cache,
> +							size_t size)
>  {
>  	if (cache->list.next) {
>  		struct io_cache_entry *entry;
>  		entry = container_of(cache->list.next, struct io_cache_entry, node);
> +		kasan_unpoison_range(entry, size);
>  		cache->list.next = cache->list.next->next;
>  		return entry;
>  	}

Does this generate the same code if KASAN isn't enabled? Since there's a
4-byte hole in struct io_alloc_cache(), might be cleaner to simply add
the 'size' argument to io_alloc_cache_init() and store it in the cache.
Then the above just becomes:

	kasan_unpoison_range(entry, cache->elem_size);

instead and that'd definitely generate the same code as before if KASAN
isn't enabled.

-- 
Jens Axboe