From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: [ofa-general] Re: [PATCH v4 14/14] QLogic VNIC: sysfs Documentation Date: Thu, 12 Jun 2008 17:34:53 +0200 Message-ID: <4851421D.4030905@trash.net> References: <20080610205633.11186.45499.stgit@dale> <20080610210918.11186.64253.stgit@dale> <484F751F.7030407@trash.net> <48513E4E.4040601@trash.net> <71d336490806120829o7cac67bes384c305e03b0d746@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, rdreier@cisco.com, poornima.kamath@qlogic.com, general@lists.openfabrics.org, Amar Mudrankit To: Ramachandra K Return-path: In-Reply-To: <71d336490806120829o7cac67bes384c305e03b0d746@mail.gmail.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: general-bounces@lists.openfabrics.org Errors-To: general-bounces@lists.openfabrics.org List-Id: netdev.vger.kernel.org Ramachandra K wrote: > On Thu, Jun 12, 2008 at 8:48 PM, Patrick McHardy wrote: >> Amar Mudrankit wrote: >>>>> +/sys/class/infiniband_qlgc_vnic/interfaces// >>>>> + >>>>> + vnic_state (0444) State of the VNIC interface. >>>> This I don't understand - you seem to be registering and >>>> unregistering the net_device based some state machine, >>>> in some cases even triggered by timers. >>>> >>>> Whats the idea behind this? >>> For a given VNIC interface, unless the VNIC host driver establishes a >>> connection with EVIC >>> and completes the control and data path communication sequence with >>> the EVIC, the >>> VNIC network device is not registered on the host. Establishing a >>> connection with EVIC >>> and going through the control and data path communication involves >>> multiple packet >>> exchange on the IB side with the EVIC and can take some time. The >>> netpath statemachine is, hence, meant to PERIODICALLY check if >>> connection is fully established with EVIC or not. >>> Once connection is established, it takes care to register the VNIC >>> netdevice. >> >> We have linkstate/operstate for this. How is a user supposed >> to configure the network device when it appears at a more or >> less random time from his perspective? >> > If you are referring to IP address configuration etc, users can configure > the interfaces by setting up ifcfg files and the interfaces are automatically > configured when they are registered. Maybe they can. It gets more complicated when daemons want to bind to that device etc. But that still leaves the main question, why is it not using the standard way and simply keeps the carrier turned off until the link is ready?