From mboxrd@z Thu Jan  1 00:00:00 1970
From: annie li <annie.li@oracle.com>
Subject: Re: [PATCH] fix netfront alloc_page error handling bug,
 need to raise up rx_refill timer if rx request buffer not big
 enough for backend
Date: Fri, 15 Nov 2013 11:09:08 +0800
Message-ID: <52859054.8080109@oracle.com>
References: <1384359348-2086-1-git-send-email-jieyue.majy@aliyun-inc.com>
	<528437E4.4040603@oracle.com> <5284B61B.5030604@citrix.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; Format="flowed"
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xen.org>
In-Reply-To: <5284B61B.5030604@citrix.com>
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: David Vrabel <david.vrabel@citrix.com>
Cc: Ma JieYue <jieyue.majy@aliyun-inc.com>, xen-devel@lists.xen.org
List-Id: xen-devel@lists.xenproject.org


On 2013-11-14 19:38, David Vrabel wrote:
> On 14/11/13 02:39, annie li wrote:
>> With this assumption, is it easier to fix this issue by moving mod_timer
>> directly
>>              mod_timer(&np->rx_refill_timer,
>>                    jiffies + (HZ/10));
>> above
>>              if (i != 0)
>>                  goto refill;
>>
>> in xennet_alloc_rx_buffers?
> Yes, always setting the timer if we didn't reach the target is a better
> fix I think.
>
> However, we probably want to reduce the fill target if we are under
> memory pressure.
>
> Something like:
>
> no_skb:
>                 /* Memory pressure, reduce fill target. */
> 	       if (--np->rx_target < np->rx_target_min)
>                     np->rx_target = np->rx_target_min;
>                 mod_timer(...);
>                 goto refill;
>
> perhaps?

Sounds reasonable.
When less memory is availble, then reduce rx_target to decrease 
requirement of memory. And rx_target can be adjusted back after memory 
is enough in following code,

     if (((req_prod - np->rx.sring->rsp_prod) < (np->rx_target / 4)) &&
         ((np->rx_target *= 2) > np->rx_max_target))
         np->rx_target = np->rx_max_target;

Thanks
Annie
>
> David