Date:	Wed, 07 Oct 2015 18:39:56 +0800
From:	Bob Liu <bob.liu@...cle.com>
To:	Roger Pau Monné <roger.pau@...rix.com>
CC:	xen-devel@...ts.xen.org, david.vrabel@...rix.com,
	linux-kernel@...r.kernel.org, konrad.wilk@...cle.com,
	felipe.franciosi@...rix.com, axboe@...com, hch@...radead.org,
	avanzini.arianna@...il.com, rafal.mielniczuk@...rix.com,
	boris.ostrovsky@...cle.com, jonathan.davies@...rix.com
Subject: Re: [PATCH v3 6/9] xen/blkfront: negotiate the number of hw queues/rings
 with backend


On 10/05/2015 10:40 PM, Roger Pau Monné wrote:
> El 05/09/15 a les 14.39, Bob Liu ha escrit:
>> The max number of hardware queues for xen/blkfront is set by parameter
>> 'max_queues', while the number xen/blkback supported is notified through
>> xenstore("multi-queue-max-queues").
>>
>> The negotiated number was the smaller one, and was written back to
>> xen/blkback as "multi-queue-num-queues".
> 
> I would write this in present instead of past, I think it's clearer:
> 
> The negotiated number _is_ the smaller...

Will update.

> 
>>
>> Signed-off-by: Bob Liu <bob.liu@...cle.com>
>> ---
>>  drivers/block/xen-blkfront.c |  142 ++++++++++++++++++++++++++++++++----------
>>  1 file changed, 108 insertions(+), 34 deletions(-)
>>
>> diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c
>> index 1cae76b..1aa66c9 100644
>> --- a/drivers/block/xen-blkfront.c
>> +++ b/drivers/block/xen-blkfront.c
>> @@ -107,6 +107,10 @@ static unsigned int xen_blkif_max_ring_order;
>>  module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, S_IRUGO);
>>  MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring");
>>  
>> +static unsigned int xen_blkif_max_queues;
>> +module_param_named(max_queues, xen_blkif_max_queues, uint, S_IRUGO);
>> +MODULE_PARM_DESC(max_queues, "Maximum number of hardware queues/rings per virtual disk");
>> +
>>  #define BLK_RING_SIZE(dinfo) __CONST_RING_SIZE(blkif, PAGE_SIZE * (dinfo)->pages_per_ring)
>>  #define BLK_MAX_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE * XENBUS_MAX_RING_PAGES)
>>  /*
>> @@ -114,6 +118,7 @@ MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the
>>   * characters are enough. Define to 20 to keep consist with backend.
>>   */
>>  #define RINGREF_NAME_LEN (20)
>> +#define QUEUE_NAME_LEN (12)
>>  
>>  /*
>>   *  Per-ring info.
>> @@ -687,7 +692,7 @@ static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size,
>>  
>>  	memset(&dinfo->tag_set, 0, sizeof(dinfo->tag_set));
>>  	dinfo->tag_set.ops = &blkfront_mq_ops;
>> -	dinfo->tag_set.nr_hw_queues = 1;
>> +	dinfo->tag_set.nr_hw_queues = dinfo->nr_rings;
>>  	dinfo->tag_set.queue_depth =  BLK_RING_SIZE(dinfo);
>>  	dinfo->tag_set.numa_node = NUMA_NO_NODE;
>>  	dinfo->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE;
>> @@ -1350,6 +1355,51 @@ fail:
>>  	return err;
>>  }
>>  
>> +static int write_per_ring_nodes(struct xenbus_transaction xbt,
>> +				struct blkfront_ring_info *rinfo, const char *dir)
>> +{
>> +	int err, i;
>> +	const char *message = NULL;
>> +	struct blkfront_dev_info *dinfo = rinfo->dinfo;
>> +
>> +	if (dinfo->pages_per_ring == 1) {
>> +		err = xenbus_printf(xbt, dir, "ring-ref", "%u", rinfo->ring_ref[0]);
>> +		if (err) {
>> +			message = "writing ring-ref";
>> +			goto abort_transaction;
>> +		}
>> +		pr_info("%s: write ring-ref:%d\n", dir, rinfo->ring_ref[0]);
>> +	} else {
>> +		for (i = 0; i < dinfo->pages_per_ring; i++) {
>> +			char ring_ref_name[RINGREF_NAME_LEN];
>> +
>> +			snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i);
>> +			err = xenbus_printf(xbt, dir, ring_ref_name,
>> +					    "%u", rinfo->ring_ref[i]);
>> +			if (err) {
>> +				message = "writing ring-ref";
>> +				goto abort_transaction;
>> +			}
>> +			pr_info("%s: write ring-ref:%d\n", dir, rinfo->ring_ref[i]);
>> +		}
>> +	}
>> +
>> +	err = xenbus_printf(xbt, dir, "event-channel", "%u", rinfo->evtchn);
>> +	if (err) {
>> +		message = "writing event-channel";
>> +		goto abort_transaction;
>> +	}
>> +	pr_info("%s: write event-channel:%d\n", dir, rinfo->evtchn);
>> +
>> +	return 0;
>> +
>> +abort_transaction:
>> +	xenbus_transaction_end(xbt, 1);
> 
> The transaction is not started inside of the helper, so I would prefer
> it to be ended in the same function where it is started. This should
> just return the error and the caller should end the transaction.
> 

The difficult part is how to return the right message?

>> +	if (message)
>> +		xenbus_dev_fatal(dinfo->xbdev, err, "%s", message);
>> +
>> +	return err;
>> +}
>>  
>>  /* Common code used when first setting up, and when resuming. */
>>  static int talk_to_blkback(struct xenbus_device *dev,
>> @@ -1386,45 +1436,51 @@ again:
>>  		goto out;
>>  	}
>>  
>> +	if (dinfo->pages_per_ring > 1) {
>> +		err = xenbus_printf(xbt, dev->nodename, "ring-page-order", "%u",
>> +				    ring_page_order);
>> +		if (err) {
>> +			message = "writing ring-page-order";
>> +			goto abort_transaction;
>> +		}
>> +	}
>> +
>> +	/* We already got the number of queues in _probe */
>>  	if (dinfo->nr_rings == 1) {
>>  		rinfo = &dinfo->rinfo[0];
>>  
>> -		if (dinfo->pages_per_ring == 1) {
>> -			err = xenbus_printf(xbt, dev->nodename,
>> -					    "ring-ref", "%u", rinfo->ring_ref[0]);
>> -			if (err) {
>> -				message = "writing ring-ref";
>> -				goto abort_transaction;
>> -			}
>> -		} else {
>> -			err = xenbus_printf(xbt, dev->nodename,
>> -					    "ring-page-order", "%u", ring_page_order);
>> -			if (err) {
>> -				message = "writing ring-page-order";
>> -				goto abort_transaction;
>> -			}
>> -
>> -			for (i = 0; i < dinfo->pages_per_ring; i++) {
>> -				char ring_ref_name[RINGREF_NAME_LEN];
>> +		err = write_per_ring_nodes(xbt, &dinfo->rinfo[0], dev->nodename);
>> +		if (err)
>> +			goto out;
>> +	} else {
>> +		char *path;
>> +		size_t pathsize;
>>  
>> -				snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i);
>> -				err = xenbus_printf(xbt, dev->nodename, ring_ref_name,
>> -						    "%u", rinfo->ring_ref[i]);
>> -				if (err) {
>> -					message = "writing ring-ref";
>> -					goto abort_transaction;
>> -				}
>> -			}
>> -		}
>> -		err = xenbus_printf(xbt, dev->nodename,
>> -				    "event-channel", "%u", rinfo->evtchn);
>> +		err = xenbus_printf(xbt, dev->nodename, "multi-queue-num-queues", "%u",
>> +				    dinfo->nr_rings);
>>  		if (err) {
>> -			message = "writing event-channel";
>> +			message = "writing multi-queue-num-queues";
>>  			goto abort_transaction;
>>  		}
>> -	} else {
>> -		/* Not supported at this stage */
>> -		goto abort_transaction;
>> +
>> +		pathsize = strlen(dev->nodename) + QUEUE_NAME_LEN;
>> +		path = kzalloc(pathsize, GFP_KERNEL);
> 
> kmalloc seems better here, since you are memseting it below in order to
> reuse it.
> 
>> +		if (!path) {
>> +			err = -ENOMEM;
>> +			message = "ENOMEM while writing ring references";
>> +			goto abort_transaction;
>> +		}
>> +
>> +		for (i = 0; i < dinfo->nr_rings; i++) {
>> +			memset(path, 0, pathsize);
>> +			snprintf(path, pathsize, "%s/queue-%u", dev->nodename, i);
>> +			err = write_per_ring_nodes(xbt, &dinfo->rinfo[i], path);
>> +			if (err) {
>> +				kfree(path);
>> +				goto out;
>> +			}
>> +		}
>> +		kfree(path);
>>  	}
>>  
>>  	err = xenbus_printf(xbt, dev->nodename, "protocol", "%s",
>> @@ -1480,6 +1536,7 @@ static int blkfront_probe(struct xenbus_device *dev,
>>  	int err, vdevice, r_index;
>>  	struct blkfront_dev_info *dinfo;
>>  	struct blkfront_ring_info *rinfo;
>> +	unsigned int back_max_queues = 0;
>>  
>>  	/* FIXME: Use dynamic device id if this is not set. */
>>  	err = xenbus_scanf(XBT_NIL, dev->nodename,
>> @@ -1534,7 +1591,17 @@ static int blkfront_probe(struct xenbus_device *dev,
>>  	dinfo->vdevice = vdevice;
>>  	dinfo->connected = BLKIF_STATE_DISCONNECTED;
>>  
>> -	dinfo->nr_rings = 1;
>> +	/* Check if backend supports multiple queues */
>> +	err = xenbus_scanf(XBT_NIL, dinfo->xbdev->otherend,
>> +			   "multi-queue-max-queues", "%u", &back_max_queues);
>> +	if (err < 0)
>> +		back_max_queues = 1;
>> +
>> +	dinfo->nr_rings = min(back_max_queues, xen_blkif_max_queues);
>> +	if (dinfo->nr_rings <= 0)
>> +		dinfo->nr_rings = 1;
>> +	pr_info("dinfo->nr_rings:%u, backend support max-queues:%u\n", dinfo->nr_rings, back_max_queues);
>         ^ pr_debug                           ^ supports
> 
> And I would also print the number of queues that are going to be used.
> 

Will update all above comments.

>> +
>>  	dinfo->rinfo = kzalloc(sizeof(*rinfo) * dinfo->nr_rings, GFP_KERNEL);
>>  	if (!dinfo->rinfo) {
>>  		xenbus_dev_fatal(dev, -ENOMEM, "allocating ring_info structure");
>> @@ -2257,6 +2324,7 @@ static struct xenbus_driver blkfront_driver = {
>>  static int __init xlblk_init(void)
>>  {
>>  	int ret;
>> +	int nr_cpus = num_online_cpus();
>>  
>>  	if (!xen_domain())
>>  		return -ENODEV;
>> @@ -2267,6 +2335,12 @@ static int __init xlblk_init(void)
>>  		xen_blkif_max_ring_order = 0;
>>  	}
>>  
>> +	if (xen_blkif_max_queues > nr_cpus) {
> 
> Shouldn't there be a default value for xen_blkif_max_queues if the user
> hasn't set the parameter on the command line?
> 

Then the default value is 0, multi-queue isn't enabled by default.

>> +		pr_info("Invalid max_queues (%d), will use default max: %d.\n",
>> +			xen_blkif_max_queues, nr_cpus);
>> +		xen_blkif_max_queues = nr_cpus;
>> +	}
>> +
>>  	if (!xen_has_pv_disk_devices())
>>  		return -ENODEV;
>>  
>>
> 

-- 
Regards,
-Bob
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists