Re: [PATCH 5/6] scsi: core: don't limit per-LUN queue depth for SSD when HBA needs

[Bart Van Assche <bvanassche@acm.org>]

On 2020-01-18 23:14, Ming Lei wrote:
> +static inline bool scsi_bypass_device_busy(struct scsi_device *sdev)
> +{
> +	struct Scsi_Host *shost = sdev->host;
> +
> +	if (!shost->hostt->no_device_queue_for_ssd)
> +		return false;
> +
> +	return blk_queue_nonrot(sdev->request_queue);
> +}

In other words, sdev->device_busy is maintained for all SCSI devices
except for those SSDs controlled by a SCSI LLD driver that has
no_device_queue_for_ssd set in its host template. I'd like to see
different behavior, namely that sdev->device_busy is not maintained for
any SSD except if that SSD really needs the sdev->device_busy counter.
The blacklist mechanism may be more appropriate to mark such SSDs than
the SCSI host template.

What I also noticed is that most scsi_bypass_device_busy() calls have an
exclamation mark (!) in front of these calls. I think that inverting the
return value and renaming this function into e.g.
scsi_maintain_device_busy() would result in code that is easier to read.

Thanks,

Bart.