| 1 |
// SPDX-License-Identifier: GPL-2.0-only |
| 2 |
/* |
| 3 |
* linux/fs/proc/net.c |
| 4 |
* |
| 5 |
* Copyright (C) 2007 |
| 6 |
* |
| 7 |
* Author: Eric Biederman <[email protected]> |
| 8 |
* |
| 9 |
* proc net directory handling functions |
| 10 |
*/ |
| 11 |
#include <linux/errno.h> |
| 12 |
#include <linux/time.h> |
| 13 |
#include <linux/proc_fs.h> |
| 14 |
#include <linux/stat.h> |
| 15 |
#include <linux/slab.h> |
| 16 |
#include <linux/init.h> |
| 17 |
#include <linux/sched.h> |
| 18 |
#include <linux/sched/task.h> |
| 19 |
#include <linux/module.h> |
| 20 |
#include <linux/bitops.h> |
| 21 |
#include <linux/mount.h> |
| 22 |
#include <linux/nsproxy.h> |
| 23 |
#include <linux/uidgid.h> |
| 24 |
#include <net/net_namespace.h> |
| 25 |
#include <linux/seq_file.h> |
| 26 |
|
| 27 |
#include "internal.h" |
| 28 |
|
| 29 |
static inline struct net *PDE_NET(struct proc_dir_entry *pde) |
| 30 |
{ |
| 31 |
return pde->parent->data; |
| 32 |
} |
| 33 |
|
| 34 |
static struct net *get_proc_net(const struct inode *inode) |
| 35 |
{ |
| 36 |
return maybe_get_net(PDE_NET(PDE(inode))); |
| 37 |
} |
| 38 |
|
| 39 |
static int seq_open_net(struct inode *inode, struct file *file) |
| 40 |
{ |
| 41 |
unsigned int state_size = PDE(inode)->state_size; |
| 42 |
struct seq_net_private *p; |
| 43 |
struct net *net; |
| 44 |
|
| 45 |
WARN_ON_ONCE(state_size < sizeof(*p)); |
| 46 |
|
| 47 |
if (file->f_mode & FMODE_WRITE && !PDE(inode)->write) |
| 48 |
return -EACCES; |
| 49 |
|
| 50 |
net = get_proc_net(inode); |
| 51 |
if (!net) |
| 52 |
return -ENXIO; |
| 53 |
|
| 54 |
p = __seq_open_private(file, PDE(inode)->seq_ops, state_size); |
| 55 |
if (!p) { |
| 56 |
put_net(net); |
| 57 |
return -ENOMEM; |
| 58 |
} |
| 59 |
#ifdef CONFIG_NET_NS |
| 60 |
p->net = net; |
| 61 |
netns_tracker_alloc(net, &p->ns_tracker, GFP_KERNEL); |
| 62 |
#endif |
| 63 |
return 0; |
| 64 |
} |
| 65 |
|
| 66 |
static void seq_file_net_put_net(struct seq_file *seq) |
| 67 |
{ |
| 68 |
#ifdef CONFIG_NET_NS |
| 69 |
struct seq_net_private *priv = seq->private; |
| 70 |
|
| 71 |
put_net_track(priv->net, &priv->ns_tracker); |
| 72 |
#else |
| 73 |
put_net(&init_net); |
| 74 |
#endif |
| 75 |
} |
| 76 |
|
| 77 |
static int seq_release_net(struct inode *ino, struct file *f) |
| 78 |
{ |
| 79 |
struct seq_file *seq = f->private_data; |
| 80 |
|
| 81 |
seq_file_net_put_net(seq); |
| 82 |
seq_release_private(ino, f); |
| 83 |
return 0; |
| 84 |
} |
| 85 |
|
| 86 |
static const struct proc_ops proc_net_seq_ops = { |
| 87 |
.proc_open = seq_open_net, |
| 88 |
.proc_read = seq_read, |
| 89 |
.proc_write = proc_simple_write, |
| 90 |
.proc_lseek = seq_lseek, |
| 91 |
.proc_release = seq_release_net, |
| 92 |
}; |
| 93 |
|
| 94 |
int bpf_iter_init_seq_net(void *priv_data, struct bpf_iter_aux_info *aux) |
| 95 |
{ |
| 96 |
#ifdef CONFIG_NET_NS |
| 97 |
struct seq_net_private *p = priv_data; |
| 98 |
|
| 99 |
p->net = get_net_track(current->nsproxy->net_ns, &p->ns_tracker, |
| 100 |
GFP_KERNEL); |
| 101 |
#endif |
| 102 |
return 0; |
| 103 |
} |
| 104 |
|
| 105 |
void bpf_iter_fini_seq_net(void *priv_data) |
| 106 |
{ |
| 107 |
#ifdef CONFIG_NET_NS |
| 108 |
struct seq_net_private *p = priv_data; |
| 109 |
|
| 110 |
put_net_track(p->net, &p->ns_tracker); |
| 111 |
#endif |
| 112 |
} |
| 113 |
|
| 114 |
struct proc_dir_entry *proc_create_net_data(const char *name, umode_t mode, |
| 115 |
struct proc_dir_entry *parent, const struct seq_operations *ops, |
| 116 |
unsigned int state_size, void *data) |
| 117 |
{ |
| 118 |
struct proc_dir_entry *p; |
| 119 |
|
| 120 |
p = proc_create_reg(name, mode, &parent, data); |
| 121 |
if (!p) |
| 122 |
return NULL; |
| 123 |
pde_force_lookup(p); |
| 124 |
p->proc_ops = &proc_net_seq_ops; |
| 125 |
p->seq_ops = ops; |
| 126 |
p->state_size = state_size; |
| 127 |
return proc_register(parent, p); |
| 128 |
} |
| 129 |
EXPORT_SYMBOL_GPL(proc_create_net_data); |
| 130 |
|
| 131 |
/** |
| 132 |
* proc_create_net_data_write - Create a writable net_ns-specific proc file |
| 133 |
* @name: The name of the file. |
| 134 |
* @mode: The file's access mode. |
| 135 |
* @parent: The parent directory in which to create. |
| 136 |
* @ops: The seq_file ops with which to read the file. |
| 137 |
* @write: The write method with which to 'modify' the file. |
| 138 |
* @state_size: The size of the per-file private state to allocate. |
| 139 |
* @data: Data for retrieval by pde_data(). |
| 140 |
* |
| 141 |
* Create a network namespaced proc file in the @parent directory with the |
| 142 |
* specified @name and @mode that allows reading of a file that displays a |
| 143 |
* series of elements and also provides for the file accepting writes that have |
| 144 |
* some arbitrary effect. |
| 145 |
* |
| 146 |
* The functions in the @ops table are used to iterate over items to be |
| 147 |
* presented and extract the readable content using the seq_file interface. |
| 148 |
* |
| 149 |
* The @write function is called with the data copied into a kernel space |
| 150 |
* scratch buffer and has a NUL appended for convenience. The buffer may be |
| 151 |
* modified by the @write function. @write should return 0 on success. |
| 152 |
* |
| 153 |
* The @data value is accessible from the @show and @write functions by calling |
| 154 |
* pde_data() on the file inode. The network namespace must be accessed by |
| 155 |
* calling seq_file_net() on the seq_file struct. |
| 156 |
*/ |
| 157 |
struct proc_dir_entry *proc_create_net_data_write(const char *name, umode_t mode, |
| 158 |
struct proc_dir_entry *parent, |
| 159 |
const struct seq_operations *ops, |
| 160 |
proc_write_t write, |
| 161 |
unsigned int state_size, void *data) |
| 162 |
{ |
| 163 |
struct proc_dir_entry *p; |
| 164 |
|
| 165 |
p = proc_create_reg(name, mode, &parent, data); |
| 166 |
if (!p) |
| 167 |
return NULL; |
| 168 |
pde_force_lookup(p); |
| 169 |
p->proc_ops = &proc_net_seq_ops; |
| 170 |
p->seq_ops = ops; |
| 171 |
p->state_size = state_size; |
| 172 |
p->write = write; |
| 173 |
return proc_register(parent, p); |
| 174 |
} |
| 175 |
EXPORT_SYMBOL_GPL(proc_create_net_data_write); |
| 176 |
|
| 177 |
static int single_open_net(struct inode *inode, struct file *file) |
| 178 |
{ |
| 179 |
struct proc_dir_entry *de = PDE(inode); |
| 180 |
struct net *net; |
| 181 |
int err; |
| 182 |
|
| 183 |
net = get_proc_net(inode); |
| 184 |
if (!net) |
| 185 |
return -ENXIO; |
| 186 |
|
| 187 |
err = single_open(file, de->single_show, net); |
| 188 |
if (err) |
| 189 |
put_net(net); |
| 190 |
return err; |
| 191 |
} |
| 192 |
|
| 193 |
static int single_release_net(struct inode *ino, struct file *f) |
| 194 |
{ |
| 195 |
struct seq_file *seq = f->private_data; |
| 196 |
put_net(seq->private); |
| 197 |
return single_release(ino, f); |
| 198 |
} |
| 199 |
|
| 200 |
static const struct proc_ops proc_net_single_ops = { |
| 201 |
.proc_open = single_open_net, |
| 202 |
.proc_read = seq_read, |
| 203 |
.proc_write = proc_simple_write, |
| 204 |
.proc_lseek = seq_lseek, |
| 205 |
.proc_release = single_release_net, |
| 206 |
}; |
| 207 |
|
| 208 |
struct proc_dir_entry *proc_create_net_single(const char *name, umode_t mode, |
| 209 |
struct proc_dir_entry *parent, |
| 210 |
int (*show)(struct seq_file *, void *), void *data) |
| 211 |
{ |
| 212 |
struct proc_dir_entry *p; |
| 213 |
|
| 214 |
p = proc_create_reg(name, mode, &parent, data); |
| 215 |
if (!p) |
| 216 |
return NULL; |
| 217 |
pde_force_lookup(p); |
| 218 |
p->proc_ops = &proc_net_single_ops; |
| 219 |
p->single_show = show; |
| 220 |
return proc_register(parent, p); |
| 221 |
} |
| 222 |
EXPORT_SYMBOL_GPL(proc_create_net_single); |
| 223 |
|
| 224 |
/** |
| 225 |
* proc_create_net_single_write - Create a writable net_ns-specific proc file |
| 226 |
* @name: The name of the file. |
| 227 |
* @mode: The file's access mode. |
| 228 |
* @parent: The parent directory in which to create. |
| 229 |
* @show: The seqfile show method with which to read the file. |
| 230 |
* @write: The write method with which to 'modify' the file. |
| 231 |
* @data: Data for retrieval by pde_data(). |
| 232 |
* |
| 233 |
* Create a network-namespaced proc file in the @parent directory with the |
| 234 |
* specified @name and @mode that allows reading of a file that displays a |
| 235 |
* single element rather than a series and also provides for the file accepting |
| 236 |
* writes that have some arbitrary effect. |
| 237 |
* |
| 238 |
* The @show function is called to extract the readable content via the |
| 239 |
* seq_file interface. |
| 240 |
* |
| 241 |
* The @write function is called with the data copied into a kernel space |
| 242 |
* scratch buffer and has a NUL appended for convenience. The buffer may be |
| 243 |
* modified by the @write function. @write should return 0 on success. |
| 244 |
* |
| 245 |
* The @data value is accessible from the @show and @write functions by calling |
| 246 |
* pde_data() on the file inode. The network namespace must be accessed by |
| 247 |
* calling seq_file_single_net() on the seq_file struct. |
| 248 |
*/ |
| 249 |
struct proc_dir_entry *proc_create_net_single_write(const char *name, umode_t mode, |
| 250 |
struct proc_dir_entry *parent, |
| 251 |
int (*show)(struct seq_file *, void *), |
| 252 |
proc_write_t write, |
| 253 |
void *data) |
| 254 |
{ |
| 255 |
struct proc_dir_entry *p; |
| 256 |
|
| 257 |
p = proc_create_reg(name, mode, &parent, data); |
| 258 |
if (!p) |
| 259 |
return NULL; |
| 260 |
pde_force_lookup(p); |
| 261 |
p->proc_ops = &proc_net_single_ops; |
| 262 |
p->single_show = show; |
| 263 |
p->write = write; |
| 264 |
return proc_register(parent, p); |
| 265 |
} |
| 266 |
EXPORT_SYMBOL_GPL(proc_create_net_single_write); |
| 267 |
|
| 268 |
static struct net *get_proc_task_net(struct inode *dir) |
| 269 |
{ |
| 270 |
struct task_struct *task; |
| 271 |
struct nsproxy *ns; |
| 272 |
struct net *net = NULL; |
| 273 |
|
| 274 |
rcu_read_lock(); |
| 275 |
task = pid_task(proc_pid(dir), PIDTYPE_PID); |
| 276 |
if (task != NULL) { |
| 277 |
task_lock(task); |
| 278 |
ns = task->nsproxy; |
| 279 |
if (ns != NULL) |
| 280 |
net = get_net(ns->net_ns); |
| 281 |
task_unlock(task); |
| 282 |
} |
| 283 |
rcu_read_unlock(); |
| 284 |
|
| 285 |
return net; |
| 286 |
} |
| 287 |
|
| 288 |
static struct dentry *proc_tgid_net_lookup(struct inode *dir, |
| 289 |
struct dentry *dentry, unsigned int flags) |
| 290 |
{ |
| 291 |
struct dentry *de; |
| 292 |
struct net *net; |
| 293 |
|
| 294 |
de = ERR_PTR(-ENOENT); |
| 295 |
net = get_proc_task_net(dir); |
| 296 |
if (net != NULL) { |
| 297 |
de = proc_lookup_de(dir, dentry, net->proc_net); |
| 298 |
put_net(net); |
| 299 |
} |
| 300 |
return de; |
| 301 |
} |
| 302 |
|
| 303 |
static int proc_tgid_net_getattr(struct mnt_idmap *idmap, |
| 304 |
const struct path *path, struct kstat *stat, |
| 305 |
u32 request_mask, unsigned int query_flags) |
| 306 |
{ |
| 307 |
struct inode *inode = d_inode(path->dentry); |
| 308 |
struct net *net; |
| 309 |
|
| 310 |
net = get_proc_task_net(inode); |
| 311 |
|
| 312 |
generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
| 313 |
|
| 314 |
if (net != NULL) { |
| 315 |
stat->nlink = net->proc_net->nlink; |
| 316 |
put_net(net); |
| 317 |
} |
| 318 |
|
| 319 |
return 0; |
| 320 |
} |
| 321 |
|
| 322 |
const struct inode_operations proc_net_inode_operations = { |
| 323 |
.lookup = proc_tgid_net_lookup, |
| 324 |
.getattr = proc_tgid_net_getattr, |
| 325 |
.setattr = proc_setattr, |
| 326 |
}; |
| 327 |
|
| 328 |
static int proc_tgid_net_readdir(struct file *file, struct dir_context *ctx) |
| 329 |
{ |
| 330 |
int ret; |
| 331 |
struct net *net; |
| 332 |
|
| 333 |
ret = -EINVAL; |
| 334 |
net = get_proc_task_net(file_inode(file)); |
| 335 |
if (net != NULL) { |
| 336 |
ret = proc_readdir_de(file, ctx, net->proc_net); |
| 337 |
put_net(net); |
| 338 |
} |
| 339 |
return ret; |
| 340 |
} |
| 341 |
|
| 342 |
const struct file_operations proc_net_operations = { |
| 343 |
.llseek = generic_file_llseek, |
| 344 |
.read = generic_read_dir, |
| 345 |
.iterate_shared = proc_tgid_net_readdir, |
| 346 |
}; |
| 347 |
|
| 348 |
static __net_init int proc_net_ns_init(struct net *net) |
| 349 |
{ |
| 350 |
struct proc_dir_entry *netd, *net_statd; |
| 351 |
kuid_t uid; |
| 352 |
kgid_t gid; |
| 353 |
int err; |
| 354 |
|
| 355 |
/* |
| 356 |
* This PDE acts only as an anchor for /proc/${pid}/net hierarchy. |
| 357 |
* Corresponding inode (PDE(inode) == net->proc_net) is never |
| 358 |
* instantiated therefore blanket zeroing is fine. |
| 359 |
* net->proc_net_stat inode is instantiated normally. |
| 360 |
*/ |
| 361 |
err = -ENOMEM; |
| 362 |
netd = kmem_cache_zalloc(proc_dir_entry_cache, GFP_KERNEL); |
| 363 |
if (!netd) |
| 364 |
goto out; |
| 365 |
|
| 366 |
netd->subdir = RB_ROOT; |
| 367 |
netd->data = net; |
| 368 |
netd->nlink = 2; |
| 369 |
netd->namelen = 3; |
| 370 |
netd->parent = &proc_root; |
| 371 |
netd->name = netd->inline_name; |
| 372 |
memcpy(netd->name, "net", 4); |
| 373 |
|
| 374 |
uid = make_kuid(net->user_ns, 0); |
| 375 |
if (!uid_valid(uid)) |
| 376 |
uid = netd->uid; |
| 377 |
|
| 378 |
gid = make_kgid(net->user_ns, 0); |
| 379 |
if (!gid_valid(gid)) |
| 380 |
gid = netd->gid; |
| 381 |
|
| 382 |
proc_set_user(netd, uid, gid); |
| 383 |
|
| 384 |
/* Seed dentry revalidation for /proc/${pid}/net */ |
| 385 |
pde_force_lookup(netd); |
| 386 |
|
| 387 |
err = -EEXIST; |
| 388 |
net_statd = proc_net_mkdir(net, "stat", netd); |
| 389 |
if (!net_statd) |
| 390 |
goto free_net; |
| 391 |
|
| 392 |
net->proc_net = netd; |
| 393 |
net->proc_net_stat = net_statd; |
| 394 |
return 0; |
| 395 |
|
| 396 |
free_net: |
| 397 |
pde_free(netd); |
| 398 |
out: |
| 399 |
return err; |
| 400 |
} |
| 401 |
|
| 402 |
static __net_exit void proc_net_ns_exit(struct net *net) |
| 403 |
{ |
| 404 |
remove_proc_entry("stat", net->proc_net); |
| 405 |
pde_free(net->proc_net); |
| 406 |
} |
| 407 |
|
| 408 |
static struct pernet_operations __net_initdata proc_net_ns_ops = { |
| 409 |
.init = proc_net_ns_init, |
| 410 |
.exit = proc_net_ns_exit, |
| 411 |
}; |
| 412 |
|
| 413 |
int __init proc_net_init(void) |
| 414 |
{ |
| 415 |
proc_symlink("net", NULL, "self/net"); |
| 416 |
|
| 417 |
return register_pernet_subsys(&proc_net_ns_ops); |
| 418 |
} |