On 2/1/19 9:59 AM, Bart Van Assche wrote:
> On Fri, 2019-02-01 at 08:24 -0700, Jens Axboe wrote:
>> +int iomap_dio_iopoll(struct kiocb *kiocb, bool spin)
>> +{
>> + struct request_queue *q = READ_ONCE(kiocb->private);
>> +
>> + if (!q)
>> + return 0;
>> + return blk_poll(q, READ_ONCE(kiocb->ki_cookie), spin);
>> +}
>> +EXPORT_SYMBOL_GPL(iomap_dio_iopoll);
>
> How does this interact with block device removal? My understanding is that this
> function can get called after a request has completed. Which mechanism, if any,
> prevents that request queue 'q' is removed after kiocb->private has been read
> and before blk_poll() is called?
The completion event is split from the reap, the latter being the one that
would put the file. So we're never going to call into ->iopoll() for a
kiocb where we've done the file put.
--
Jens Axboe
