On Tue, Jun 15, 2021 at 03:10:31PM +0200, Christoph Hellwig wrote:
> Replace the blk_poll interface that requires the caller to keep a queue
> and cookie from the submissions with polling based on the bio.
>
> Polling for the bio itself leads to a few advantages:
>
> - the cookie construction can made entirely private in blk-mq.c
> - the caller does not need to remember the request_queue and cookie
> separately and thus sidesteps their lifetime issues
> - keeping the device and the cookie inside the bio allows to trivially
> support polling BIOs remapping by stacking drivers
> - a lot of code to propagate the cookie back up the submission path can
> be removed entirely.
>
...
> +/**
> + * bio_poll - poll for BIO completions
> + * @bio: bio to poll for
> + * @flags: BLK_POLL_* flags that control the behavior
> + *
> + * Poll for completions on queue associated with the bio. Returns number of
> + * completed entries found.
> + *
> + * Note: the caller must either be the context that submitted @bio, or
> + * be in a RCU critical section to prevent freeing of @bio.
> + */
> +int bio_poll(struct bio *bio, unsigned int flags)
> +{
> + struct request_queue *q = bio->bi_bdev->bd_disk->queue;
> + blk_qc_t cookie = READ_ONCE(bio->bi_cookie);
> + int ret;
> +
> + if (cookie == BLK_QC_T_NONE ||
> + !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
> + return 0;
> +
> + if (current->plug)
> + blk_flush_plug_list(current->plug, false);
> +
> + if (blk_queue_enter(q, BLK_MQ_REQ_NOWAIT))
> + return 0;
> + if (WARN_ON_ONCE(!queue_is_mq(q)))
> + ret = 0; /* not yet implemented, should not happen */
> + else
> + ret = blk_mq_poll(q, cookie, flags);
> + blk_queue_exit(q);
> + return ret;
> +}
> +EXPORT_SYMBOL_GPL(bio_poll);
> +
> +/*
> + * Helper to implement file_operations.iopoll. Requires the bio to be stored
> + * in iocb->private, and cleared before freeing the bio.
> + */
> +int iocb_bio_iopoll(struct kiocb *kiocb, unsigned int flags)
> +{
> + struct bio *bio;
> + int ret = 0;
> +
> + /*
> + * Note: the bio cache only uses SLAB_TYPESAFE_BY_RCU, so bio can
> + * point to a freshly allocated bio at this point. If that happens
> + * we have a few cases to consider:
> + *
> + * 1) the bio is beeing initialized and bi_bdev is NULL. We can just
> + * simply nothing in this case
> + * 2) the bio points to a not poll enabled device. bio_poll will catch
> + * this and return 0
> + * 3) the bio points to a poll capable device, including but not
> + * limited to the one that the original bio pointed to. In this
> + * case we will call into the actual poll method and poll for I/O,
> + * even if we don't need to, but it won't cause harm either.
> + *
> + * For cases 2) and 3) above the RCU grace period ensures that bi_bdev
> + * is still allocated. Because partitions hold a reference to the whole
> + * device bdev and thus disk, the disk is also still valid. Grabbing
> + * a reference to the queue in bio_poll() ensures the hctxs and requests
> + * are still valid as well.
> + */
Not sure disk is valid, we only hold the disk when opening a bdev, but
the bdev can be closed during polling. Also disk always holds one
reference on request queue, so if disk is valid, no need to grab queue's
refcnt in bio_poll().
Thanks,
Ming
