Re: [PATCH 03/15] block: add an API to atomically update queue limits

[Christoph Hellwig <hch@lst.de>]

On Wed, Jan 24, 2024 at 07:08:58AM +0100, Hannes Reinecke wrote:
>> + * that there is outstanding I/O by other means.
>> + */
>> +static inline struct queue_limits
>> +queue_limits_start_update(struct request_queue *q)
>> +	__acquires(q->limits_lock)
>> +{
>> +	mutex_lock(&q->limits_lock);
>> +	return q->limits;
>> +}
>
> I'm slightly confused about the lifetime of the returned structure.
> By my understanding, the returned 'struct queue_limit' is allocated
> on the stack of the caller, right?
> Shouldn't we note this somewhere such that people don't start passing
> the structure around, or, worse, calling 'kfree()' on it?

queue_limits_start_update returns the structure by value.  So the
lifetime is that of whatever variable you assign it to, which better
be on the stack.  Tryign to kfree a non-pointer type will get the
compiler complain, as does passing a reference to it outside the
function these days.