diff mbox series

[3/6] scsi: sd: register request queue after sd_revalidate_disk is done

Message ID 20200119071432.18558-4-ming.lei@redhat.com (mailing list archive)
State Changes Requested
Headers show
Series scsi: support bypass device busy check for some high end HBA with SSD | expand

Commit Message

Ming Lei Jan. 19, 2020, 7:14 a.m. UTC 
Prepare for improving SSD performance in the following patch, which
needs to read queue flag of QUEUE_FLAG_NONROT in IO path. So we have
to freeze queue before changing this flag in sd_revalidate_disk().

However, queue freezing becomes quite slow after the queue is registered
because RCU grace period is involved.

So delay registering queue after sd_revalidate_disk() is done for
avoiding slow queue freezing which will be added to sd_revalidate_disk()
in the following patch.

Cc: Sathya Prakash <sathya.prakash@broadcom.com>
Cc: Chaitra P B <chaitra.basappa@broadcom.com>
Cc: Suganath Prabu Subramani <suganath-prabu.subramani@broadcom.com>
Cc: Kashyap Desai <kashyap.desai@broadcom.com>
Cc: Sumit Saxena <sumit.saxena@broadcom.com>
Cc: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Cc: Ewan D. Milne <emilne@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>,
Cc: Hannes Reinecke <hare@suse.de>
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 drivers/scsi/sd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Bart Van Assche Jan. 19, 2020, 8:36 p.m. UTC | #1 
On 2020-01-18 23:14, Ming Lei wrote:
> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> index 5afb0046b12a..f401ba96dcfd 100644
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
> @@ -3380,11 +3380,12 @@ static int sd_probe(struct device *dev)
>  		pm_runtime_set_autosuspend_delay(dev,
>  			sdp->host->hostt->rpm_autosuspend_delay);
>  	}
> -	device_add_disk(dev, gd, NULL);
> +	device_add_disk_no_queue_reg(dev, gd);
>  	if (sdkp->capacity)
>  		sd_dif_config_host(sdkp);
>  
>  	sd_revalidate_disk(gd);
> +	blk_register_queue(gd);
>  
>  	if (sdkp->security) {
>  		sdkp->opal_dev = init_opal_dev(sdp, &sd_sec_submit);

The effect of changing device_add_disk() into
device_add_disk_no_queue_reg() is as follows:
- __device_add_disk() skips the elevator_init_mq() call.
- __device_add_disk() skips the blk_register_queue() call.

The above patch introduces a blk_register_queue() call but no
elevator_init_mq() call. Is that perhaps an oversight?

Thanks,

Bart.
diff mbox series

Patch

diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index 5afb0046b12a..f401ba96dcfd 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -3380,11 +3380,12 @@  static int sd_probe(struct device *dev)
 		pm_runtime_set_autosuspend_delay(dev,
 			sdp->host->hostt->rpm_autosuspend_delay);
 	}
-	device_add_disk(dev, gd, NULL);
+	device_add_disk_no_queue_reg(dev, gd);
 	if (sdkp->capacity)
 		sd_dif_config_host(sdkp);
 
 	sd_revalidate_disk(gd);
+	blk_register_queue(gd);
 
 	if (sdkp->security) {
 		sdkp->opal_dev = init_opal_dev(sdp, &sd_sec_submit);