1 /* 2 * Copyright (c) 2004 Topspin Communications. All rights reserved. 3 * 4 * This software is available to you under a choice of one of two 5 * licenses. You may choose to be licensed under the terms of the GNU 6 * General Public License (GPL) Version 2, available from the file 7 * COPYING in the main directory of this source tree, or the 8 * OpenIB.org BSD license below: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * 18 * - Redistributions in binary form must reproduce the above 19 * copyright notice, this list of conditions and the following 20 * disclaimer in the documentation and/or other materials 21 * provided with the distribution. 22 * 23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 30 * SOFTWARE. 31 */ 32 33 #include <linux/module.h> 34 #include <linux/sched/signal.h> 35 36 #include <linux/init.h> 37 #include <linux/seq_file.h> 38 39 #include <linux/uaccess.h> 40 41 #include "ipoib.h" 42 43 static ssize_t show_parent(struct device *d, struct device_attribute *attr, 44 char *buf) 45 { 46 struct net_device *dev = to_net_dev(d); 47 struct ipoib_dev_priv *priv = ipoib_priv(dev); 48 49 return sysfs_emit(buf, "%s\n", priv->parent->name); 50 } 51 static DEVICE_ATTR(parent, S_IRUGO, show_parent, NULL); 52 53 static bool is_child_unique(struct ipoib_dev_priv *ppriv, 54 struct ipoib_dev_priv *priv) 55 { 56 struct ipoib_dev_priv *tpriv; 57 58 ASSERT_RTNL(); 59 60 /* 61 * Since the legacy sysfs interface uses pkey for deletion it cannot 62 * support more than one interface with the same pkey, it creates 63 * ambiguity. The RTNL interface deletes using the netdev so it does 64 * not have a problem to support duplicated pkeys. 65 */ 66 if (priv->child_type != IPOIB_LEGACY_CHILD) 67 return true; 68 69 /* 70 * First ensure this isn't a duplicate. We check the parent device and 71 * then all of the legacy child interfaces to make sure the Pkey 72 * doesn't match. 73 */ 74 if (ppriv->pkey == priv->pkey) 75 return false; 76 77 list_for_each_entry(tpriv, &ppriv->child_intfs, list) { 78 if (tpriv->pkey == priv->pkey && 79 tpriv->child_type == IPOIB_LEGACY_CHILD) 80 return false; 81 } 82 83 return true; 84 } 85 86 /* 87 * NOTE: If this function fails then the priv->dev will remain valid, however 88 * priv will have been freed and must not be touched by caller in the error 89 * case. 90 * 91 * If (ndev->reg_state == NETREG_UNINITIALIZED) then it is up to the caller to 92 * free the net_device (just as rtnl_newlink does) otherwise the net_device 93 * will be freed when the rtnl is unlocked. 94 */ 95 int __ipoib_vlan_add(struct ipoib_dev_priv *ppriv, struct ipoib_dev_priv *priv, 96 u16 pkey, int type) 97 { 98 struct net_device *ndev = priv->dev; 99 int result; 100 struct rdma_netdev *rn = netdev_priv(ndev); 101 102 ASSERT_RTNL(); 103 104 /* 105 * We do not need to touch priv if register_netdevice fails, so just 106 * always use this flow. 107 */ 108 ndev->priv_destructor = ipoib_intf_free; 109 110 /* 111 * Racing with unregister of the parent must be prevented by the 112 * caller. 113 */ 114 WARN_ON(ppriv->dev->reg_state != NETREG_REGISTERED); 115 116 if (pkey == 0 || pkey == 0x8000) { 117 result = -EINVAL; 118 goto out_early; 119 } 120 121 rn->mtu = priv->mcast_mtu; 122 123 priv->parent = ppriv->dev; 124 priv->pkey = pkey; 125 priv->child_type = type; 126 127 if (!is_child_unique(ppriv, priv)) { 128 result = -ENOTUNIQ; 129 goto out_early; 130 } 131 132 result = register_netdevice(ndev); 133 if (result) { 134 ipoib_warn(priv, "failed to initialize; error %i", result); 135 136 /* 137 * register_netdevice sometimes calls priv_destructor, 138 * sometimes not. Make sure it was done. 139 */ 140 goto out_early; 141 } 142 143 /* RTNL childs don't need proprietary sysfs entries */ 144 if (type == IPOIB_LEGACY_CHILD) { 145 if (ipoib_cm_add_mode_attr(ndev)) 146 goto sysfs_failed; 147 if (ipoib_add_pkey_attr(ndev)) 148 goto sysfs_failed; 149 if (ipoib_add_umcast_attr(ndev)) 150 goto sysfs_failed; 151 152 if (device_create_file(&ndev->dev, &dev_attr_parent)) 153 goto sysfs_failed; 154 } 155 156 return 0; 157 158 sysfs_failed: 159 unregister_netdevice(priv->dev); 160 return -ENOMEM; 161 162 out_early: 163 if (ndev->priv_destructor) 164 ndev->priv_destructor(ndev); 165 return result; 166 } 167 168 int ipoib_vlan_add(struct net_device *pdev, unsigned short pkey) 169 { 170 struct ipoib_dev_priv *ppriv, *priv; 171 char intf_name[IFNAMSIZ]; 172 struct net_device *ndev; 173 int result; 174 175 if (!capable(CAP_NET_ADMIN)) 176 return -EPERM; 177 178 if (!rtnl_trylock()) 179 return restart_syscall(); 180 181 if (pdev->reg_state != NETREG_REGISTERED) { 182 rtnl_unlock(); 183 return -EPERM; 184 } 185 186 ppriv = ipoib_priv(pdev); 187 188 snprintf(intf_name, sizeof(intf_name), "%s.%04x", 189 ppriv->dev->name, pkey); 190 191 ndev = ipoib_intf_alloc(ppriv->ca, ppriv->port, intf_name); 192 if (IS_ERR(ndev)) { 193 result = PTR_ERR(ndev); 194 goto out; 195 } 196 priv = ipoib_priv(ndev); 197 198 ndev->rtnl_link_ops = ipoib_get_link_ops(); 199 200 result = __ipoib_vlan_add(ppriv, priv, pkey, IPOIB_LEGACY_CHILD); 201 202 if (result && ndev->reg_state == NETREG_UNINITIALIZED) 203 free_netdev(ndev); 204 205 out: 206 rtnl_unlock(); 207 208 return result; 209 } 210 211 struct ipoib_vlan_delete_work { 212 struct work_struct work; 213 struct net_device *dev; 214 }; 215 216 /* 217 * sysfs callbacks of a netdevice cannot obtain the rtnl lock as 218 * unregister_netdev ultimately deletes the sysfs files while holding the rtnl 219 * lock. This deadlocks the system. 220 * 221 * A callback can use rtnl_trylock to avoid the deadlock but it cannot call 222 * unregister_netdev as that internally takes and releases the rtnl_lock. So 223 * instead we find the netdev to unregister and then do the actual unregister 224 * from the global work queue where we can obtain the rtnl_lock safely. 225 */ 226 static void ipoib_vlan_delete_task(struct work_struct *work) 227 { 228 struct ipoib_vlan_delete_work *pwork = 229 container_of(work, struct ipoib_vlan_delete_work, work); 230 struct net_device *dev = pwork->dev; 231 232 rtnl_lock(); 233 234 /* Unregistering tasks can race with another task or parent removal */ 235 if (dev->reg_state == NETREG_REGISTERED) { 236 struct ipoib_dev_priv *priv = ipoib_priv(dev); 237 struct ipoib_dev_priv *ppriv = ipoib_priv(priv->parent); 238 239 ipoib_dbg(ppriv, "delete child vlan %s\n", dev->name); 240 unregister_netdevice(dev); 241 } 242 243 rtnl_unlock(); 244 245 kfree(pwork); 246 } 247 248 int ipoib_vlan_delete(struct net_device *pdev, unsigned short pkey) 249 { 250 struct ipoib_dev_priv *ppriv, *priv, *tpriv; 251 int rc; 252 253 if (!capable(CAP_NET_ADMIN)) 254 return -EPERM; 255 256 if (!rtnl_trylock()) 257 return restart_syscall(); 258 259 if (pdev->reg_state != NETREG_REGISTERED) { 260 rtnl_unlock(); 261 return -EPERM; 262 } 263 264 ppriv = ipoib_priv(pdev); 265 266 rc = -ENODEV; 267 list_for_each_entry_safe(priv, tpriv, &ppriv->child_intfs, list) { 268 if (priv->pkey == pkey && 269 priv->child_type == IPOIB_LEGACY_CHILD) { 270 struct ipoib_vlan_delete_work *work; 271 272 work = kmalloc(sizeof(*work), GFP_KERNEL); 273 if (!work) { 274 rc = -ENOMEM; 275 goto out; 276 } 277 278 down_write(&ppriv->vlan_rwsem); 279 list_del_init(&priv->list); 280 up_write(&ppriv->vlan_rwsem); 281 work->dev = priv->dev; 282 INIT_WORK(&work->work, ipoib_vlan_delete_task); 283 queue_work(ipoib_workqueue, &work->work); 284 285 rc = 0; 286 break; 287 } 288 } 289 290 out: 291 rtnl_unlock(); 292 293 return rc; 294 } 295