| 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/sched/signal.h> |
| 34 | |
| 35 | #include <linux/init.h> |
| 36 | #include <linux/seq_file.h> |
| 37 | |
| 38 | #include <linux/uaccess.h> |
| 39 | |
| 40 | #include "ipoib.h" |
| 41 | |
| 42 | static ssize_t parent_show(struct device *d, struct device_attribute *attr, |
| 43 | char *buf) |
| 44 | { |
| 45 | struct net_device *dev = to_net_dev(d); |
| 46 | struct ipoib_dev_priv *priv = ipoib_priv(dev); |
| 47 | |
| 48 | return sysfs_emit(buf, fmt: "%s\n" , priv->parent->name); |
| 49 | } |
| 50 | static DEVICE_ATTR_RO(parent); |
| 51 | |
| 52 | static bool is_child_unique(struct ipoib_dev_priv *ppriv, |
| 53 | struct ipoib_dev_priv *priv) |
| 54 | { |
| 55 | struct ipoib_dev_priv *tpriv; |
| 56 | bool result = true; |
| 57 | |
| 58 | /* |
| 59 | * Since the legacy sysfs interface uses pkey for deletion it cannot |
| 60 | * support more than one interface with the same pkey, it creates |
| 61 | * ambiguity. The RTNL interface deletes using the netdev so it does |
| 62 | * not have a problem to support duplicated pkeys. |
| 63 | */ |
| 64 | if (priv->child_type != IPOIB_LEGACY_CHILD) |
| 65 | return true; |
| 66 | |
| 67 | /* |
| 68 | * First ensure this isn't a duplicate. We check the parent device and |
| 69 | * then all of the legacy child interfaces to make sure the Pkey |
| 70 | * doesn't match. |
| 71 | */ |
| 72 | if (ppriv->pkey == priv->pkey) |
| 73 | return false; |
| 74 | |
| 75 | netdev_lock(dev: ppriv->dev); |
| 76 | list_for_each_entry(tpriv, &ppriv->child_intfs, list) { |
| 77 | if (tpriv->pkey == priv->pkey && |
| 78 | tpriv->child_type == IPOIB_LEGACY_CHILD) { |
| 79 | result = false; |
| 80 | break; |
| 81 | } |
| 82 | } |
| 83 | netdev_unlock(dev: ppriv->dev); |
| 84 | |
| 85 | return result; |
| 86 | } |
| 87 | |
| 88 | /* |
| 89 | * NOTE: If this function fails then the priv->dev will remain valid, however |
| 90 | * priv will have been freed and must not be touched by caller in the error |
| 91 | * case. |
| 92 | * |
| 93 | * If (ndev->reg_state == NETREG_UNINITIALIZED) then it is up to the caller to |
| 94 | * free the net_device (just as rtnl_newlink does) otherwise the net_device |
| 95 | * will be freed when the rtnl is unlocked. |
| 96 | */ |
| 97 | int __ipoib_vlan_add(struct ipoib_dev_priv *ppriv, struct ipoib_dev_priv *priv, |
| 98 | u16 pkey, int type) |
| 99 | { |
| 100 | struct net_device *ndev = priv->dev; |
| 101 | int result; |
| 102 | struct rdma_netdev *rn = netdev_priv(dev: ndev); |
| 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(dev: 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(dev: ndev)) |
| 146 | goto sysfs_failed; |
| 147 | if (ipoib_add_pkey_attr(dev: ndev)) |
| 148 | goto sysfs_failed; |
| 149 | if (ipoib_add_umcast_attr(dev: ndev)) |
| 150 | goto sysfs_failed; |
| 151 | |
| 152 | if (device_create_file(device: &ndev->dev, entry: &dev_attr_parent)) |
| 153 | goto sysfs_failed; |
| 154 | } |
| 155 | |
| 156 | return 0; |
| 157 | |
| 158 | sysfs_failed: |
| 159 | unregister_netdevice(dev: 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(dev: pdev); |
| 187 | |
| 188 | /* If you increase IFNAMSIZ, update snprintf below |
| 189 | * to allow longer names. |
| 190 | */ |
| 191 | BUILD_BUG_ON(IFNAMSIZ != 16); |
| 192 | snprintf(buf: intf_name, size: sizeof(intf_name), fmt: "%.10s.%04x" , ppriv->dev->name, |
| 193 | pkey); |
| 194 | |
| 195 | ndev = ipoib_intf_alloc(hca: ppriv->ca, port: ppriv->port, format: intf_name); |
| 196 | if (IS_ERR(ptr: ndev)) { |
| 197 | result = PTR_ERR(ptr: ndev); |
| 198 | goto out; |
| 199 | } |
| 200 | priv = ipoib_priv(dev: ndev); |
| 201 | |
| 202 | ndev->rtnl_link_ops = ipoib_get_link_ops(); |
| 203 | |
| 204 | result = __ipoib_vlan_add(ppriv, priv, pkey, type: IPOIB_LEGACY_CHILD); |
| 205 | |
| 206 | if (result && ndev->reg_state == NETREG_UNINITIALIZED) |
| 207 | free_netdev(dev: ndev); |
| 208 | |
| 209 | out: |
| 210 | rtnl_unlock(); |
| 211 | |
| 212 | return result; |
| 213 | } |
| 214 | |
| 215 | struct ipoib_vlan_delete_work { |
| 216 | struct work_struct work; |
| 217 | struct net_device *dev; |
| 218 | }; |
| 219 | |
| 220 | /* |
| 221 | * sysfs callbacks of a netdevice cannot obtain the rtnl lock as |
| 222 | * unregister_netdev ultimately deletes the sysfs files while holding the rtnl |
| 223 | * lock. This deadlocks the system. |
| 224 | * |
| 225 | * A callback can use rtnl_trylock to avoid the deadlock but it cannot call |
| 226 | * unregister_netdev as that internally takes and releases the rtnl_lock. So |
| 227 | * instead we find the netdev to unregister and then do the actual unregister |
| 228 | * from the global work queue where we can obtain the rtnl_lock safely. |
| 229 | */ |
| 230 | static void ipoib_vlan_delete_task(struct work_struct *work) |
| 231 | { |
| 232 | struct ipoib_vlan_delete_work *pwork = |
| 233 | container_of(work, struct ipoib_vlan_delete_work, work); |
| 234 | struct net_device *dev = pwork->dev; |
| 235 | |
| 236 | rtnl_lock(); |
| 237 | |
| 238 | /* Unregistering tasks can race with another task or parent removal */ |
| 239 | if (dev->reg_state == NETREG_REGISTERED) { |
| 240 | struct ipoib_dev_priv *priv = ipoib_priv(dev); |
| 241 | struct ipoib_dev_priv *ppriv = ipoib_priv(dev: priv->parent); |
| 242 | |
| 243 | ipoib_dbg(ppriv, "delete child vlan %s\n" , dev->name); |
| 244 | unregister_netdevice(dev); |
| 245 | } |
| 246 | |
| 247 | rtnl_unlock(); |
| 248 | |
| 249 | kfree(objp: pwork); |
| 250 | } |
| 251 | |
| 252 | int ipoib_vlan_delete(struct net_device *pdev, unsigned short pkey) |
| 253 | { |
| 254 | struct ipoib_dev_priv *ppriv, *priv, *tpriv; |
| 255 | int rc; |
| 256 | |
| 257 | if (!capable(CAP_NET_ADMIN)) |
| 258 | return -EPERM; |
| 259 | |
| 260 | if (!rtnl_trylock()) |
| 261 | return restart_syscall(); |
| 262 | |
| 263 | if (pdev->reg_state != NETREG_REGISTERED) { |
| 264 | rtnl_unlock(); |
| 265 | return -EPERM; |
| 266 | } |
| 267 | |
| 268 | ppriv = ipoib_priv(dev: pdev); |
| 269 | |
| 270 | rc = -ENODEV; |
| 271 | netdev_lock(dev: ppriv->dev); |
| 272 | list_for_each_entry_safe(priv, tpriv, &ppriv->child_intfs, list) { |
| 273 | if (priv->pkey == pkey && |
| 274 | priv->child_type == IPOIB_LEGACY_CHILD) { |
| 275 | struct ipoib_vlan_delete_work *work; |
| 276 | |
| 277 | work = kmalloc(sizeof(*work), GFP_KERNEL); |
| 278 | if (!work) { |
| 279 | rc = -ENOMEM; |
| 280 | goto out; |
| 281 | } |
| 282 | |
| 283 | list_del_init(entry: &priv->list); |
| 284 | work->dev = priv->dev; |
| 285 | INIT_WORK(&work->work, ipoib_vlan_delete_task); |
| 286 | queue_work(wq: ipoib_workqueue, work: &work->work); |
| 287 | |
| 288 | rc = 0; |
| 289 | break; |
| 290 | } |
| 291 | } |
| 292 | |
| 293 | out: |
| 294 | netdev_unlock(dev: ppriv->dev); |
| 295 | rtnl_unlock(); |
| 296 | |
| 297 | return rc; |
| 298 | } |
| 299 | |