From mboxrd@z Thu Jan 1 00:00:00 1970 From: Darren Hart Subject: Re: [PATCH] pch_gbe: Use a randomly generated MAC instead of failing probe Date: Mon, 16 Jan 2012 07:42:55 -0800 Message-ID: <4F14457F.8060509@linux.intel.com> References: <132d2a41a089905de3147b4656e350608aa7fd6f.1326523495.git.dvhart@linux.intel.com> <20120114.001430.787918662083526597.davem@davemloft.net> <4F11A533.4040406@linux.intel.com> <20120114.115604.2101782124431552110.davem@davemloft.net> <20120114214658.4ddfec30@pyramind.ukuu.org.uk> <4F120357.3020808@linux.intel.com> <20120116123110.7ce91f2d@bob.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: Tomoya MORINAGA , Alan Cox , David Miller , linux-kernel@vger.kernel.org, arjan@linux.intel.com, jeffrey.t.kirsher@intel.com, paul.gortmaker@windriver.com, jdmason@kudzu.us, netdev@vger.kernel.org To: Alan Cox Return-path: In-Reply-To: <20120116123110.7ce91f2d@bob.linux.org.uk> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On 01/16/2012 04:31 AM, Alan Cox wrote: >> Saving MAC address into external ROM is generic method, I think. >> Though I know the ROM-less system using eg20t-pch, however I think >> this system is not common. >> So, I think pch_gbe shouldn't have auto-mac address assignment. > > The problem is the module load fails for those cases. You cannot load > the module and use the standard ifconfig eth0 hw aa:bb:cc:dd:ee:ff > interface. The better fix might be to make sure it loads. > > So change from > > memcpy(netdev->dev_addr, adapter->hw.mac.addr, > netdev->addr_len); > if (!is_valid_ether_addr(netdev->dev_addr)) { > dev_err(&pdev->dev, "Invalid MAC Address\n"); > ret = -EIO; > goto err_free_adapter; > } > > to just printing a warning, and check the current address when a user > tries to ifconfig it up and refuse to allow the port to go active. I can go this route I suppose. I don't really understand the objection to the use of a random mac addr in the special case given the prevalence of this approach within existing drivers. One reason I don't care for this alternative approach is that this particular hardware is targeted at embedded use where we can't assume a full init system is available, etc. It can be made to work of course, it just isn't as automated. David, would you orefer/accept an alternative patch which allows the driver to load without a MAC address so the user can set it via ifconfig after boot? -- Darren Hart Intel Open Source Technology Center Yocto Project - Linux Kernel