From: Simon Horman Subject: Re: [RFC PATCH 05/12] ixgbe: Add SR-IOV features to main module Date: Thu, 10 Dec 2009 12:04:17 +1100 Message-ID: <20091210010416.GJ4163@verge.net.au> References: <20091208221258.28373.16663.stgit@localhost.localdomain> <20091208221435.28373.76070.stgit@localhost.localdomain> <20091210004530.GG4163@verge.net.au> <43F901BD926A4E43B106BF17856F07558C3DED4A@orsmsx508.amr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "Kirsher, Jeffrey T" , "netdev@vger.kernel.org" , "gospo@redhat.com" , "Waskiewicz Jr, Peter P" To: "Rose, Gregory V" Return-path: Received: from kirsty.vergenet.net ([202.4.237.240]:57182 "EHLO kirsty.vergenet.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751390AbZLJBEK (ORCPT ); Wed, 9 Dec 2009 20:04:10 -0500 Content-Disposition: inline In-Reply-To: <43F901BD926A4E43B106BF17856F07558C3DED4A@orsmsx508.amr.corp.intel.com> Sender: netdev-owner@vger.kernel.org List-ID: On Wed, Dec 09, 2009 at 04:49:26PM -0800, Rose, Gregory V wrote: > >-----Original Message----- > >From: Simon Horman [mailto:horms@verge.net.au] > >Sent: Wednesday, December 09, 2009 4:46 PM > >To: Kirsher, Jeffrey T > >Cc: netdev@vger.kernel.org; gospo@redhat.com; Rose, Gregory V; > >Waskiewicz Jr, Peter P > >Subject: Re: [RFC PATCH 05/12] ixgbe: Add SR-IOV features to main module > > > >> @@ -5767,6 +5954,52 @@ static int __devinit ixgbe_probe(struct pci_dev > >*pdev, > >> goto err_sw_init; > >> } > >> > >> +#ifdef CONFIG_PCI_IOV > >> + if (hw->mac.type == ixgbe_mac_82599EB) { > >> + /* The 82599 supports up to 64 VFs per physical function > >> + * but this implementation limits allocation to 63 so that > >> + * basic networking resources are still available to the > >> + * physical function > >> + */ > >> + adapter->num_vfs = (max_vfs > 63) ? 63 : max_vfs; > >> + if (adapter->num_vfs) { > >> + adapter->flags |= IXGBE_FLAG_SRIOV_ENABLED; > >> + err = pci_enable_sriov(pdev, adapter->num_vfs); > >> + if (err) { > >> + DPRINTK(PROBE, ERR, > >> + "Failed to enable PCI sriov: %d\n", > >> + err); > >> + adapter->flags &= ~IXGBE_FLAG_SRIOV_ENABLED; > >> + adapter->num_vfs = 0; > >> + } > >> + } > >> + /* If call to enable VFs succeeded then allocate memory > >> + * for per VF control structures. > >> + */ > >> + if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) { > >> + adapter->vfinfo = > >> + kcalloc(adapter->num_vfs, > >> + sizeof(struct vf_data_storage), > >> + GFP_KERNEL); > >> + if (!adapter->vfinfo) { > >> + /* Oh oh */ > >> + DPRINTK(PROBE, ERR, > >> + "Unable to allocate memory for VF " > >> + "Data Storage - SRIOV disabled\n"); > >> + adapter->flags &= ~IXGBE_FLAG_SRIOV_ENABLED; > >> + adapter->num_vfs = 0; > > > >A call to pci_disable_sriov(pdev) is needed here? > [Rose, Gregory V] > > Good catch. Thanks. Actually I only noticed while doing the error paths for my suggestion below. > >> + } else { > >> + /* Now that we're sure SR-IOV is enabled > >> + * set up the mailbox parameters > >> + */ > >> + ixgbe_init_mbx_params_pf(hw); > >> + memcpy(&hw->mbx.ops, ii->mbx_ops, > >> + sizeof(hw->mbx.ops)); > >> + } > >> + } > >> + } > >> + > >> +#endif /* CONFIG_PCI_IOV */ > >> netdev->features = NETIF_F_SG | > >> NETIF_F_IP_CSUM | > >> NETIF_F_HW_VLAN_TX | > > > >I wonder if it would be easier on the eyes to break the hunk above out > >into > >a ixgbe_probe_vf() function. Something like (compile tested only, I > >don't > >have an 82599): > [Rose, Gregory V] > > I think that was done with the igb driver. It's a good suggestion. Yes, something like that is in igb.