linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "Roger Pau Monné" <roger.pau@citrix.com>
To: Bob Liu <bob.liu@oracle.com>, <xen-devel@lists.xen.org>
Cc: <david.vrabel@citrix.com>, <linux-kernel@vger.kernel.org>,
	<konrad.wilk@oracle.com>, <felipe.franciosi@citrix.com>,
	<axboe@fb.com>, <hch@infradead.org>, <avanzini.arianna@gmail.com>,
	<rafal.mielniczuk@citrix.com>, <boris.ostrovsky@oracle.com>,
	<jonathan.davies@citrix.com>
Subject: Re: [PATCH v3 9/9] xen/blkback: get number of hardware queues/rings from blkfront
Date: Mon, 5 Oct 2015 17:15:05 +0200	[thread overview]
Message-ID: <561293F9.1050006@citrix.com> (raw)
In-Reply-To: <1441456782-31318-10-git-send-email-bob.liu@oracle.com>

El 05/09/15 a les 14.39, Bob Liu ha escrit:
> Backend advertises "multi-queue-max-queues" to front, and then read back the
> final negotiated queues/rings from "multi-queue-num-queues" which is wrote by
> blkfront.
> 
> Signed-off-by: Bob Liu <bob.liu@oracle.com>
> ---
>  drivers/block/xen-blkback/blkback.c |    8 ++++++++
>  drivers/block/xen-blkback/xenbus.c  |   36 ++++++++++++++++++++++++++++++-----
>  2 files changed, 39 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
> index fd02240..b904fe05f0 100644
> --- a/drivers/block/xen-blkback/blkback.c
> +++ b/drivers/block/xen-blkback/blkback.c
> @@ -83,6 +83,11 @@ module_param_named(max_persistent_grants, xen_blkif_max_pgrants, int, 0644);
>  MODULE_PARM_DESC(max_persistent_grants,
>                   "Maximum number of grants to map persistently");
>  
> +unsigned int xenblk_max_queues;
> +module_param_named(max_queues, xenblk_max_queues, uint, 0644);
> +MODULE_PARM_DESC(max_queues,
> +		 "Maximum number of hardware queues per virtual disk");
> +
>  /*
>   * Maximum order of pages to be used for the shared ring between front and
>   * backend, 4KB page granularity is used.
> @@ -1458,6 +1463,9 @@ static int __init xen_blkif_init(void)
>  		xen_blkif_max_ring_order = XENBUS_MAX_RING_PAGE_ORDER;
>  	}
>  
> +	/* Allow as many queues as there are CPUs, by default */
> +	xenblk_max_queues = num_online_cpus();

Hm, I'm not sure of the best way to set a default value for this.
Consider for example a scenario were Dom0 is limited to 2vCPUs, but DomU
has 8 vCPUs. Are we going to limit the number of queues to two? Is that
the most appropriate value from a performance PoV?

I have to admit I don't have a clear idea of a default value for this
field, and maybe the number of CPUs on the backend is indeed what works
better, but there needs to be a comment explaining the reasoning behind
this setting.

>  	rc = xen_blkif_interface_init();
>  	if (rc)
>  		goto failed_init;
> diff --git a/drivers/block/xen-blkback/xenbus.c b/drivers/block/xen-blkback/xenbus.c
> index 04b8d0d..aa97ea5 100644
> --- a/drivers/block/xen-blkback/xenbus.c
> +++ b/drivers/block/xen-blkback/xenbus.c
> @@ -28,6 +28,8 @@
>  #define RINGREF_NAME_LEN (20)
>  #define RINGREF_NAME_LEN (20)
>  
> +extern unsigned int xenblk_max_queues;

This should live in blkback/common.h

> +
>  struct backend_info {
>  	struct xenbus_device	*dev;
>  	struct xen_blkif	*blkif;
> @@ -191,11 +193,6 @@ static struct xen_blkif *xen_blkif_alloc(domid_t domid)
>  	atomic_set(&blkif->drain, 0);
>  	INIT_WORK(&blkif->free_work, xen_blkif_deferred_free);
>  
> -	blkif->nr_rings = 1;
> -	if (xen_blkif_alloc_rings(blkif)) {
> -		kmem_cache_free(xen_blkif_cachep, blkif);
> -		return ERR_PTR(-ENOMEM);
> -	}
>  	return blkif;
>  }
>  
> @@ -618,6 +615,14 @@ static int xen_blkbk_probe(struct xenbus_device *dev,
>  		goto fail;
>  	}
>  
> +	/* Multi-queue: wrte how many queues backend supported. */
                        ^ write how many queues are supported by the
backend.
> +	err = xenbus_printf(XBT_NIL, dev->nodename,
> +			    "multi-queue-max-queues", "%u", xenblk_max_queues);
> +	if (err) {
> +		pr_debug("Error writing multi-queue-num-queues\n");
                ^ pr_warn at least.
> +		goto fail;
> +	}
> +
>  	/* setup back pointer */
>  	be->blkif->be = be;
>  
> @@ -1008,6 +1013,7 @@ static int connect_ring(struct backend_info *be)
>  	char *xspath;
>  	size_t xspathsize;
>  	const size_t xenstore_path_ext_size = 11; /* sufficient for "/queue-NNN" */
> +	unsigned int requested_num_queues = 0;
>  
>  	pr_debug("%s %s\n", __func__, dev->otherend);
>  
> @@ -1035,6 +1041,26 @@ static int connect_ring(struct backend_info *be)
>  	be->blkif->vbd.feature_gnt_persistent = pers_grants;
>  	be->blkif->vbd.overflow_max_grants = 0;
>  
> +	/*
> +	 * Read the number of hardware queus from frontend.
                                       ^ queues
> +	 */
> +	err = xenbus_scanf(XBT_NIL, dev->otherend, "multi-queue-num-queues", "%u", &requested_num_queues);
> +	if (err < 0) {
> +		requested_num_queues = 1;
> +	} else {
> +		if (requested_num_queues > xenblk_max_queues
> +		    || requested_num_queues == 0) {
> +			/* buggy or malicious guest */
> +			xenbus_dev_fatal(dev, err,
> +					"guest requested %u queues, exceeding the maximum of %u.",
> +					requested_num_queues, xenblk_max_queues);
> +			return -1;
> +		}
> +	}
> +	be->blkif->nr_rings = requested_num_queues;
> +	if (xen_blkif_alloc_rings(be->blkif))
> +		return -ENOMEM;
> +
>  	pr_info("nr_rings:%d protocol %d (%s) %s\n", be->blkif->nr_rings,
>  		 be->blkif->blk_protocol, protocol,
>  		 pers_grants ? "persistent grants" : "");
> 


  reply	other threads:[~2015-10-05 15:15 UTC|newest]

Thread overview: 42+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-09-05 12:39 [PATCH v3 0/9] xen-block: support multi hardware-queues/rings Bob Liu
2015-09-05 12:39 ` [PATCH v3 1/9] xen-blkfront: convert to blk-mq APIs Bob Liu
2015-09-23 20:31   ` Konrad Rzeszutek Wilk
2015-09-23 21:12     ` Konrad Rzeszutek Wilk
2015-09-05 12:39 ` [PATCH v3 2/9] xen-block: add document for mutli hardware queues/rings Bob Liu
2015-09-23 20:32   ` Konrad Rzeszutek Wilk
2015-10-02 16:04   ` Roger Pau Monné
2015-10-02 16:12     ` [Xen-devel] " Wei Liu
2015-10-02 16:22       ` Roger Pau Monné
2015-10-02 23:55         ` Bob Liu
2015-09-05 12:39 ` [PATCH v3 3/9] xen/blkfront: separate per ring information out of device info Bob Liu
2015-10-02 17:02   ` Roger Pau Monné
2015-10-03  0:34     ` Bob Liu
2015-10-05 15:17       ` Roger Pau Monné
2015-10-10  8:30     ` Bob Liu
2015-10-19  9:42       ` Roger Pau Monné
2015-09-05 12:39 ` [PATCH v3 4/9] xen/blkfront: pseudo support for multi hardware queues/rings Bob Liu
2015-10-05 10:52   ` Roger Pau Monné
2015-10-07 10:28     ` Bob Liu
2015-09-05 12:39 ` [PATCH v3 5/9] xen/blkfront: convert per device io_lock to per ring ring_lock Bob Liu
2015-10-05 14:13   ` Roger Pau Monné
2015-10-07 10:34     ` Bob Liu
2015-09-05 12:39 ` [PATCH v3 6/9] xen/blkfront: negotiate the number of hw queues/rings with backend Bob Liu
2015-10-05 14:40   ` Roger Pau Monné
2015-10-07 10:39     ` Bob Liu
2015-10-07 11:46       ` Roger Pau Monné
2015-10-07 12:19         ` Bob Liu
2015-09-05 12:39 ` [PATCH v3 7/9] xen/blkback: separate ring information out of struct xen_blkif Bob Liu
2015-10-05 14:55   ` Roger Pau Monné
2015-10-07 10:41     ` Bob Liu
2015-10-10  4:08     ` Bob Liu
2015-10-19  9:36       ` Roger Pau Monné
2015-10-19 10:03         ` Bob Liu
2015-10-05 14:55   ` Roger Pau Monné
2015-09-05 12:39 ` [PATCH v3 8/9] xen/blkback: pseudo support for multi hardware queues/rings Bob Liu
2015-10-05 15:08   ` Roger Pau Monné
2015-10-07 10:50     ` Bob Liu
2015-10-07 11:49       ` Roger Pau Monné
2015-09-05 12:39 ` [PATCH v3 9/9] xen/blkback: get number of hardware queues/rings from blkfront Bob Liu
2015-10-05 15:15   ` Roger Pau Monné [this message]
2015-10-07 10:54     ` Bob Liu
2015-10-02  9:57 ` [PATCH v3 0/9] xen-block: support multi hardware-queues/rings Rafal Mielniczuk

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=561293F9.1050006@citrix.com \
    --to=roger.pau@citrix.com \
    --cc=avanzini.arianna@gmail.com \
    --cc=axboe@fb.com \
    --cc=bob.liu@oracle.com \
    --cc=boris.ostrovsky@oracle.com \
    --cc=david.vrabel@citrix.com \
    --cc=felipe.franciosi@citrix.com \
    --cc=hch@infradead.org \
    --cc=jonathan.davies@citrix.com \
    --cc=konrad.wilk@oracle.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=rafal.mielniczuk@citrix.com \
    --cc=xen-devel@lists.xen.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).