On Tue, 2018-04-24 at 16:58 +0200, Roman Pen wrote:
> +void ib_pool_dev_init(enum ib_pd_flags pd_flags,
> + struct ib_pool_device *(*alloc)(void),
> + void (*free)(struct ib_pool_device *),
> + int (*init)(struct ib_pool_device *),
> + struct ib_device_pool *pool)
> +{
> + INIT_LIST_HEAD(&pool->list);
> + mutex_init(&pool->mutex);
> +
> + pool->pd_flags = pd_flags;
> +
> + BUG_ON(!alloc ^ !free);
Most of this file looks generally OK, but this is a hard No. No new
BUG_ON()s in any code unless it truly is a machine ending event where
stopping is better than corrupting. That's not the case here, this was
just used as a lazy means of not doing a graceful failure during early
prototype. I get that. But it's got to go before this patch can even
be considered.
--
Doug Ledford <dledford@xxxxxxxxxx>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDDAttachment:
signature.asc
Description: This is a digitally signed message part
