[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <a8d7b051-8f43-3e21-714f-50d7faa53e0e@linux.vnet.ibm.com>
Date: Wed, 20 Sep 2017 16:37:04 +0200
From: Ursula Braun <ubraun@...ux.vnet.ibm.com>
To: Leon Romanovsky <leon@...nel.org>
Cc: davem@...emloft.net, netdev@...r.kernel.org,
linux-rdma@...r.kernel.org, linux-s390@...r.kernel.org,
jwi@...ux.vnet.ibm.com, schwidefsky@...ibm.com,
heiko.carstens@...ibm.com, raspl@...ux.vnet.ibm.com
Subject: Re: [PATCH net-next 08/10] net/smc: introduce a delay
On 09/20/2017 04:03 PM, Leon Romanovsky wrote:
> On Wed, Sep 20, 2017 at 01:58:11PM +0200, Ursula Braun wrote:
>> The number of outstanding work requests is limited. If all work
>> requests are in use, tx processing is postponed to another scheduling
>> of the tx worker. Switch to a delayed worker to have a gap for tx
>> completion queue events before the next retry.
>>
>
> How will delay prevent and protect the resource exhausting?
>
> Thanks
>
SMC runs with a fixed number of in-flight work requests per QP (constant
SMC_WR_BUF_CNT) to prevent resource exhausting. If all work requests are
currently in use, sending of another work request has to wait till some
outstanding work request is confirmed via send completion queue. If sending
is done in a context which is not allowed to wait, the tx_worker is
scheduled instead.
With this patch a small delay is added to avoid too many unsuccessful send
retries due to a still ongoing "all work requests in use" condition.
Download attachment "0xC5ED6645.asc" of type "application/pgp-keys" (9796 bytes)
Download attachment "signature.asc" of type "application/pgp-signature" (837 bytes)
Powered by blists - more mailing lists