From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by dpdk.org (Postfix) with ESMTP id 545387CFC for ; Wed, 6 Dec 2017 03:48:41 +0100 (CET) Received: from orsmga006.jf.intel.com ([10.7.209.51]) by orsmga101.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 05 Dec 2017 18:48:40 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.45,366,1508828400"; d="scan'208";a="489246" Received: from tanjianf-mobl.ccr.corp.intel.com (HELO [10.67.64.58]) ([10.67.64.58]) by orsmga006.jf.intel.com with ESMTP; 05 Dec 2017 18:48:40 -0800 To: Thomas Monjalon References: <20171201003642.19827-1-thomas@monjalon.net> <5374836.XgGFu4empV@xps> <1712827.SmiPysL3o0@xps> Cc: dev@dpdk.org From: "Tan, Jianfeng" Message-ID: <484ecb30-1eae-c1c8-adce-2318dab2f4b1@intel.com> Date: Wed, 6 Dec 2017 10:48:39 +0800 User-Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0 MIME-Version: 1.0 In-Reply-To: <1712827.SmiPysL3o0@xps> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [dpdk-dev] [PATCH] bus/vdev: add custom scan hook X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 06 Dec 2017 02:48:42 -0000 > From: Thomas Monjalon [mailto:thomas@monjalon.net] > Sent: Tuesday, December 5, 2017 11:21 PM > To: Tan, Jianfeng > Cc: dev@dpdk.org > Subject: Re: [PATCH] bus/vdev: add custom scan hook > > 05/12/2017 14:56, Tan, Jianfeng: > > > > On 12/5/2017 4:41 PM, Thomas Monjalon wrote: > > > > > > 05/12/2017 09:27, Tan, Jianfeng: > > > > > > > > > > > > > > On 12/4/2017 5:31 PM, Thomas Monjalon wrote: > > > > > > > > > > > > > > > > The hook is in bus->scan(). > > > > > > > > I think we should launch a bus scan when there is a new device event. > > > > > > > > > > > > > > That's what I'm trying to say. We finally need to execute a handler as > > > > > > > of a device event to finish the job. > > > > > > Please be more specific, I am not sure to understand. > > > > > > > By the handler, I mean when we monitor the udev by select/poll/epoll and > > device uevents come, the application will execute a handler (or just a > > function) for each of such uevent. Then why not adding the vdev there? > > Because it must work for hotplug, and initial scan too. > We can also think to application requiring a manual scan. > The bus scan is the right place to have every scans called. That's simple. Yes, the logic is simpler; anything changes, appication just invokes a scan.Then we will have a somewhat complex hook that: (1) scan all devices and identify added/removed devices; (2) and add vdev for added device; (3) remove vdev for removed device. Compared to that, we can separate the implementation for hotplug and initial scan: * For initial scan, iterate all devices to add vdev for each of them. * For hotplug, based on which device is added/removed, we just add/remove the corresponding vdev for that device. Anyway, since this patch does provide a new way; and will not affect the other way. So I think it's OK to add this. Thank you for your clarification. Thanks, Jianfeng