From mboxrd@z Thu Jan  1 00:00:00 1970
From: David Vrabel <david.vrabel@citrix.com>
Subject: Re: [PATCH] introduce grant copy for user land
Date: Tue, 11 Nov 2014 13:34:18 +0000
Message-ID: <5462105A.6020100@citrix.com>
References: <1412262916-22596-1-git-send-email-thanos.makatos@citrix.com>
	<5457C35F.50504@citrix.com>
	<2368A3FCF9F7214298E53C823B0A48EC04289299@AMSPEX01CL02.citrite.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xen.org>
Received: from mail6.bemta3.messagelabs.com ([195.245.230.39])
	by lists.xen.org with esmtp (Exim 4.72)
	(envelope-from <david.vrabel@citrix.com>) id 1XoBaQ-0001E3-Tw
	for xen-devel@lists.xenproject.org; Tue, 11 Nov 2014 13:34:23 +0000
In-Reply-To: <2368A3FCF9F7214298E53C823B0A48EC04289299@AMSPEX01CL02.citrite.net>
List-Unsubscribe: <http://lists.xen.org/cgi-bin/mailman/options/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xen.org>
List-Help: <mailto:xen-devel-request@lists.xen.org?subject=help>
List-Subscribe: <http://lists.xen.org/cgi-bin/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=subscribe>
Sender: xen-devel-bounces@lists.xen.org
Errors-To: xen-devel-bounces@lists.xen.org
To: Thanos Makatos <thanos.makatos@citrix.com>, "xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>
Cc: "boris.ostrovsky@oracle.com" <boris.ostrovsky@oracle.com>
List-Id: xen-devel@lists.xenproject.org

On 11/11/14 12:27, Thanos Makatos wrote:
>> The arbitrary limitations in number of ops and page alignment should be
>> removed.  I think both can be removed relatively easily by consuming page
>> aligned chunks of segments and doign the hypercall when a batch of ops is
>> filled.
> 
> Wouldn't this lead to multiple calls to gnttab_batch_copy(), potentially hurting performance?

The incremental benefits of batching diminishes as the batch size increase.

We also don't want multi-page allocations in this driver, so the batch
size should be set accordingly.

>>> +#define IOCTL_GNTDEV_GRANT_COPY \
>>> +_IOC(_IOC_NONE, 'G', 8, sizeof(struct ioctl_gntdev_grant_copy))
>>> +struct ioctl_gntdev_grant_copy {
>>> +	/*
>>> +	 * copy direction: 0 to copy to guest, 1 to copy from guest
>>> +	 */
>>> +	int dir;
>>
>> I think this dir should be per-segment and use the GNTCPY_source_gref and
>> GNTCOPY_dest_gref flags, since per-op direction is what the hypercall
>> provides.
> 
> OK.

The interface should also support grant to grant copies.

David