Linux Audio

Check our new training course

Loading...
v4.10.11
 
   1/*
   2 * Syscall interface to knfsd.
   3 *
   4 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
   5 */
   6
   7#include <linux/slab.h>
   8#include <linux/namei.h>
   9#include <linux/ctype.h>
 
  10
  11#include <linux/sunrpc/svcsock.h>
  12#include <linux/lockd/lockd.h>
  13#include <linux/sunrpc/addr.h>
  14#include <linux/sunrpc/gss_api.h>
  15#include <linux/sunrpc/gss_krb5_enctypes.h>
  16#include <linux/sunrpc/rpc_pipe_fs.h>
  17#include <linux/module.h>
 
  18
  19#include "idmap.h"
  20#include "nfsd.h"
  21#include "cache.h"
  22#include "state.h"
  23#include "netns.h"
  24#include "pnfs.h"
 
 
 
  25
  26/*
  27 *	We have a single directory with several nodes in it.
  28 */
  29enum {
  30	NFSD_Root = 1,
  31	NFSD_List,
 
  32	NFSD_Export_features,
  33	NFSD_Fh,
  34	NFSD_FO_UnlockIP,
  35	NFSD_FO_UnlockFS,
  36	NFSD_Threads,
  37	NFSD_Pool_Threads,
  38	NFSD_Pool_Stats,
  39	NFSD_Reply_Cache_Stats,
  40	NFSD_Versions,
  41	NFSD_Ports,
  42	NFSD_MaxBlkSize,
  43	NFSD_MaxConnections,
  44	NFSD_SupportedEnctypes,
  45	/*
  46	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
  47	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
  48	 */
  49#ifdef CONFIG_NFSD_V4
  50	NFSD_Leasetime,
  51	NFSD_Gracetime,
  52	NFSD_RecoveryDir,
  53	NFSD_V4EndGrace,
  54#endif
 
  55};
  56
  57/*
  58 * write() for these nodes.
  59 */
  60static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
  61static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
  62static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
  63static ssize_t write_threads(struct file *file, char *buf, size_t size);
  64static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
  65static ssize_t write_versions(struct file *file, char *buf, size_t size);
  66static ssize_t write_ports(struct file *file, char *buf, size_t size);
  67static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
  68static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
  69#ifdef CONFIG_NFSD_V4
  70static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
  71static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
 
  72static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
 
  73static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
  74#endif
  75
  76static ssize_t (*write_op[])(struct file *, char *, size_t) = {
  77	[NFSD_Fh] = write_filehandle,
  78	[NFSD_FO_UnlockIP] = write_unlock_ip,
  79	[NFSD_FO_UnlockFS] = write_unlock_fs,
  80	[NFSD_Threads] = write_threads,
  81	[NFSD_Pool_Threads] = write_pool_threads,
  82	[NFSD_Versions] = write_versions,
  83	[NFSD_Ports] = write_ports,
  84	[NFSD_MaxBlkSize] = write_maxblksize,
  85	[NFSD_MaxConnections] = write_maxconn,
  86#ifdef CONFIG_NFSD_V4
  87	[NFSD_Leasetime] = write_leasetime,
  88	[NFSD_Gracetime] = write_gracetime,
 
  89	[NFSD_RecoveryDir] = write_recoverydir,
 
  90	[NFSD_V4EndGrace] = write_v4_end_grace,
  91#endif
  92};
  93
  94static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
  95{
  96	ino_t ino =  file_inode(file)->i_ino;
  97	char *data;
  98	ssize_t rv;
  99
 100	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
 101		return -EINVAL;
 102
 103	data = simple_transaction_get(file, buf, size);
 104	if (IS_ERR(data))
 105		return PTR_ERR(data);
 106
 107	rv =  write_op[ino](file, data, size);
 108	if (rv >= 0) {
 109		simple_transaction_set(file, rv);
 110		rv = size;
 111	}
 112	return rv;
 113}
 114
 115static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 116{
 117	if (! file->private_data) {
 118		/* An attempt to read a transaction file without writing
 119		 * causes a 0-byte write so that the file can return
 120		 * state information
 121		 */
 122		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
 123		if (rv < 0)
 124			return rv;
 125	}
 126	return simple_transaction_read(file, buf, size, pos);
 127}
 128
 129static const struct file_operations transaction_ops = {
 130	.write		= nfsctl_transaction_write,
 131	.read		= nfsctl_transaction_read,
 132	.release	= simple_transaction_release,
 133	.llseek		= default_llseek,
 134};
 135
 136static int exports_net_open(struct net *net, struct file *file)
 137{
 138	int err;
 139	struct seq_file *seq;
 140	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 141
 142	err = seq_open(file, &nfs_exports_op);
 143	if (err)
 144		return err;
 145
 146	seq = file->private_data;
 147	seq->private = nn->svc_export_cache;
 148	return 0;
 149}
 150
 151static int exports_proc_open(struct inode *inode, struct file *file)
 152{
 153	return exports_net_open(current->nsproxy->net_ns, file);
 154}
 155
 156static const struct file_operations exports_proc_operations = {
 157	.open		= exports_proc_open,
 158	.read		= seq_read,
 159	.llseek		= seq_lseek,
 160	.release	= seq_release,
 161};
 162
 163static int exports_nfsd_open(struct inode *inode, struct file *file)
 164{
 165	return exports_net_open(inode->i_sb->s_fs_info, file);
 166}
 167
 168static const struct file_operations exports_nfsd_operations = {
 169	.open		= exports_nfsd_open,
 170	.read		= seq_read,
 171	.llseek		= seq_lseek,
 172	.release	= seq_release,
 173};
 174
 175static int export_features_show(struct seq_file *m, void *v)
 176{
 177	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
 178	return 0;
 179}
 180
 181static int export_features_open(struct inode *inode, struct file *file)
 182{
 183	return single_open(file, export_features_show, NULL);
 184}
 185
 186static const struct file_operations export_features_operations = {
 187	.open		= export_features_open,
 188	.read		= seq_read,
 189	.llseek		= seq_lseek,
 190	.release	= single_release,
 191};
 192
 193#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
 194static int supported_enctypes_show(struct seq_file *m, void *v)
 195{
 196	seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
 197	return 0;
 198}
 199
 200static int supported_enctypes_open(struct inode *inode, struct file *file)
 201{
 202	return single_open(file, supported_enctypes_show, NULL);
 203}
 204
 205static const struct file_operations supported_enctypes_ops = {
 206	.open		= supported_enctypes_open,
 207	.read		= seq_read,
 208	.llseek		= seq_lseek,
 209	.release	= single_release,
 210};
 211#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
 212
 213static const struct file_operations pool_stats_operations = {
 214	.open		= nfsd_pool_stats_open,
 215	.read		= seq_read,
 216	.llseek		= seq_lseek,
 217	.release	= nfsd_pool_stats_release,
 218};
 219
 220static const struct file_operations reply_cache_stats_operations = {
 221	.open		= nfsd_reply_cache_stats_open,
 222	.read		= seq_read,
 223	.llseek		= seq_lseek,
 224	.release	= single_release,
 225};
 226
 227/*----------------------------------------------------------------------------*/
 228/*
 229 * payload - write methods
 230 */
 231
 232static inline struct net *netns(struct file *file)
 233{
 234	return file_inode(file)->i_sb->s_fs_info;
 235}
 236
 237/**
 238 * write_unlock_ip - Release all locks used by a client
 239 *
 240 * Experimental.
 241 *
 242 * Input:
 243 *			buf:	'\n'-terminated C string containing a
 244 *				presentation format IP address
 245 *			size:	length of C string in @buf
 246 * Output:
 247 *	On success:	returns zero if all specified locks were released;
 248 *			returns one if one or more locks were not released
 249 *	On error:	return code is negative errno value
 250 */
 251static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
 252{
 253	struct sockaddr_storage address;
 254	struct sockaddr *sap = (struct sockaddr *)&address;
 255	size_t salen = sizeof(address);
 256	char *fo_path;
 257	struct net *net = netns(file);
 258
 259	/* sanity check */
 260	if (size == 0)
 261		return -EINVAL;
 262
 263	if (buf[size-1] != '\n')
 264		return -EINVAL;
 265
 266	fo_path = buf;
 267	if (qword_get(&buf, fo_path, size) < 0)
 268		return -EINVAL;
 269
 270	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
 271		return -EINVAL;
 272
 
 273	return nlmsvc_unlock_all_by_ip(sap);
 274}
 275
 276/**
 277 * write_unlock_fs - Release all locks on a local file system
 278 *
 279 * Experimental.
 280 *
 281 * Input:
 282 *			buf:	'\n'-terminated C string containing the
 283 *				absolute pathname of a local file system
 284 *			size:	length of C string in @buf
 285 * Output:
 286 *	On success:	returns zero if all specified locks were released;
 287 *			returns one if one or more locks were not released
 288 *	On error:	return code is negative errno value
 289 */
 290static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
 291{
 292	struct path path;
 293	char *fo_path;
 294	int error;
 295
 296	/* sanity check */
 297	if (size == 0)
 298		return -EINVAL;
 299
 300	if (buf[size-1] != '\n')
 301		return -EINVAL;
 302
 303	fo_path = buf;
 304	if (qword_get(&buf, fo_path, size) < 0)
 305		return -EINVAL;
 306
 307	error = kern_path(fo_path, 0, &path);
 308	if (error)
 309		return error;
 310
 311	/*
 312	 * XXX: Needs better sanity checking.  Otherwise we could end up
 313	 * releasing locks on the wrong file system.
 314	 *
 315	 * For example:
 316	 * 1.  Does the path refer to a directory?
 317	 * 2.  Is that directory a mount point, or
 318	 * 3.  Is that directory the root of an exported file system?
 319	 */
 320	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
 321
 322	path_put(&path);
 323	return error;
 324}
 325
 326/**
 327 * write_filehandle - Get a variable-length NFS file handle by path
 328 *
 329 * On input, the buffer contains a '\n'-terminated C string comprised of
 330 * three alphanumeric words separated by whitespace.  The string may
 331 * contain escape sequences.
 332 *
 333 * Input:
 334 *			buf:
 335 *				domain:		client domain name
 336 *				path:		export pathname
 337 *				maxsize:	numeric maximum size of
 338 *						@buf
 339 *			size:	length of C string in @buf
 340 * Output:
 341 *	On success:	passed-in buffer filled with '\n'-terminated C
 342 *			string containing a ASCII hex text version
 343 *			of the NFS file handle;
 344 *			return code is the size in bytes of the string
 345 *	On error:	return code is negative errno value
 346 */
 347static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
 348{
 349	char *dname, *path;
 350	int uninitialized_var(maxsize);
 351	char *mesg = buf;
 352	int len;
 353	struct auth_domain *dom;
 354	struct knfsd_fh fh;
 355
 356	if (size == 0)
 357		return -EINVAL;
 358
 359	if (buf[size-1] != '\n')
 360		return -EINVAL;
 361	buf[size-1] = 0;
 362
 363	dname = mesg;
 364	len = qword_get(&mesg, dname, size);
 365	if (len <= 0)
 366		return -EINVAL;
 367	
 368	path = dname+len+1;
 369	len = qword_get(&mesg, path, size);
 370	if (len <= 0)
 371		return -EINVAL;
 372
 373	len = get_int(&mesg, &maxsize);
 374	if (len)
 375		return len;
 376
 377	if (maxsize < NFS_FHSIZE)
 378		return -EINVAL;
 379	maxsize = min(maxsize, NFS3_FHSIZE);
 380
 381	if (qword_get(&mesg, mesg, size)>0)
 382		return -EINVAL;
 383
 
 
 384	/* we have all the words, they are in buf.. */
 385	dom = unix_domain_find(dname);
 386	if (!dom)
 387		return -ENOMEM;
 388
 389	len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
 390	auth_domain_put(dom);
 391	if (len)
 392		return len;
 393	
 394	mesg = buf;
 395	len = SIMPLE_TRANSACTION_LIMIT;
 396	qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
 397	mesg[-1] = '\n';
 398	return mesg - buf;	
 399}
 400
 401/**
 402 * write_threads - Start NFSD, or report the current number of running threads
 403 *
 404 * Input:
 405 *			buf:		ignored
 406 *			size:		zero
 407 * Output:
 408 *	On success:	passed-in buffer filled with '\n'-terminated C
 409 *			string numeric value representing the number of
 410 *			running NFSD threads;
 411 *			return code is the size in bytes of the string
 412 *	On error:	return code is zero
 413 *
 414 * OR
 415 *
 416 * Input:
 417 *			buf:		C string containing an unsigned
 418 *					integer value representing the
 419 *					number of NFSD threads to start
 420 *			size:		non-zero length of C string in @buf
 421 * Output:
 422 *	On success:	NFS service is started;
 423 *			passed-in buffer filled with '\n'-terminated C
 424 *			string numeric value representing the number of
 425 *			running NFSD threads;
 426 *			return code is the size in bytes of the string
 427 *	On error:	return code is zero or a negative errno value
 428 */
 429static ssize_t write_threads(struct file *file, char *buf, size_t size)
 430{
 431	char *mesg = buf;
 432	int rv;
 433	struct net *net = netns(file);
 434
 435	if (size > 0) {
 436		int newthreads;
 437		rv = get_int(&mesg, &newthreads);
 438		if (rv)
 439			return rv;
 440		if (newthreads < 0)
 441			return -EINVAL;
 442		rv = nfsd_svc(newthreads, net);
 
 443		if (rv < 0)
 444			return rv;
 445	} else
 446		rv = nfsd_nrthreads(net);
 447
 448	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
 449}
 450
 451/**
 452 * write_pool_threads - Set or report the current number of threads per pool
 453 *
 454 * Input:
 455 *			buf:		ignored
 456 *			size:		zero
 457 *
 458 * OR
 459 *
 460 * Input:
 461 * 			buf:		C string containing whitespace-
 462 * 					separated unsigned integer values
 463 *					representing the number of NFSD
 464 *					threads to start in each pool
 465 *			size:		non-zero length of C string in @buf
 466 * Output:
 467 *	On success:	passed-in buffer filled with '\n'-terminated C
 468 *			string containing integer values representing the
 469 *			number of NFSD threads in each pool;
 470 *			return code is the size in bytes of the string
 471 *	On error:	return code is zero or a negative errno value
 472 */
 473static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
 474{
 475	/* if size > 0, look for an array of number of threads per node
 476	 * and apply them  then write out number of threads per node as reply
 477	 */
 478	char *mesg = buf;
 479	int i;
 480	int rv;
 481	int len;
 482	int npools;
 483	int *nthreads;
 484	struct net *net = netns(file);
 485
 486	mutex_lock(&nfsd_mutex);
 487	npools = nfsd_nrpools(net);
 488	if (npools == 0) {
 489		/*
 490		 * NFS is shut down.  The admin can start it by
 491		 * writing to the threads file but NOT the pool_threads
 492		 * file, sorry.  Report zero threads.
 493		 */
 494		mutex_unlock(&nfsd_mutex);
 495		strcpy(buf, "0\n");
 496		return strlen(buf);
 497	}
 498
 499	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
 500	rv = -ENOMEM;
 501	if (nthreads == NULL)
 502		goto out_free;
 503
 504	if (size > 0) {
 505		for (i = 0; i < npools; i++) {
 506			rv = get_int(&mesg, &nthreads[i]);
 507			if (rv == -ENOENT)
 508				break;		/* fewer numbers than pools */
 509			if (rv)
 510				goto out_free;	/* syntax error */
 511			rv = -EINVAL;
 512			if (nthreads[i] < 0)
 513				goto out_free;
 
 514		}
 515		rv = nfsd_set_nrthreads(i, nthreads, net);
 516		if (rv)
 517			goto out_free;
 518	}
 519
 520	rv = nfsd_get_nrthreads(npools, nthreads, net);
 521	if (rv)
 522		goto out_free;
 523
 524	mesg = buf;
 525	size = SIMPLE_TRANSACTION_LIMIT;
 526	for (i = 0; i < npools && size > 0; i++) {
 527		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
 528		len = strlen(mesg);
 529		size -= len;
 530		mesg += len;
 531	}
 532	rv = mesg - buf;
 533out_free:
 534	kfree(nthreads);
 535	mutex_unlock(&nfsd_mutex);
 536	return rv;
 537}
 538
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 539static ssize_t __write_versions(struct file *file, char *buf, size_t size)
 540{
 541	char *mesg = buf;
 542	char *vers, *minorp, sign;
 543	int len, num, remaining;
 544	unsigned minor;
 545	ssize_t tlen = 0;
 546	char *sep;
 547	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 548
 549	if (size>0) {
 550		if (nn->nfsd_serv)
 551			/* Cannot change versions without updating
 552			 * nn->nfsd_serv->sv_xdrsize, and reallocing
 553			 * rq_argp and rq_resp
 554			 */
 555			return -EBUSY;
 556		if (buf[size-1] != '\n')
 557			return -EINVAL;
 558		buf[size-1] = 0;
 
 559
 560		vers = mesg;
 561		len = qword_get(&mesg, vers, size);
 562		if (len <= 0) return -EINVAL;
 563		do {
 
 
 564			sign = *vers;
 565			if (sign == '+' || sign == '-')
 566				num = simple_strtol((vers+1), &minorp, 0);
 567			else
 568				num = simple_strtol(vers, &minorp, 0);
 569			if (*minorp == '.') {
 570				if (num != 4)
 571					return -EINVAL;
 572				minor = simple_strtoul(minorp+1, NULL, 0);
 573				if (minor == 0)
 574					return -EINVAL;
 575				if (nfsd_minorversion(minor, sign == '-' ?
 576						     NFSD_CLEAR : NFSD_SET) < 0)
 577					return -EINVAL;
 578				goto next;
 579			}
 
 
 580			switch(num) {
 
 581			case 2:
 
 582			case 3:
 
 
 583			case 4:
 584				nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
 
 
 
 
 
 
 
 
 
 
 
 
 585				break;
 586			default:
 587				return -EINVAL;
 
 
 588			}
 589		next:
 590			vers += len + 1;
 591		} while ((len = qword_get(&mesg, vers, size)) > 0);
 592		/* If all get turned off, turn them back on, as
 593		 * having no versions is BAD
 594		 */
 595		nfsd_reset_versions();
 596	}
 597
 598	/* Now write current state into reply buffer */
 599	len = 0;
 600	sep = "";
 601	remaining = SIMPLE_TRANSACTION_LIMIT;
 602	for (num=2 ; num <= 4 ; num++)
 603		if (nfsd_vers(num, NFSD_AVAIL)) {
 604			len = snprintf(buf, remaining, "%s%c%d", sep,
 605				       nfsd_vers(num, NFSD_TEST)?'+':'-',
 606				       num);
 607			sep = " ";
 608
 609			if (len >= remaining)
 610				break;
 611			remaining -= len;
 612			buf += len;
 613			tlen += len;
 614		}
 615	if (nfsd_vers(4, NFSD_AVAIL))
 616		for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
 617		     minor++) {
 618			len = snprintf(buf, remaining, " %c4.%u",
 619					(nfsd_vers(4, NFSD_TEST) &&
 620					 nfsd_minorversion(minor, NFSD_TEST)) ?
 621						'+' : '-',
 622					minor);
 623
 
 
 
 
 624			if (len >= remaining)
 625				break;
 626			remaining -= len;
 627			buf += len;
 628			tlen += len;
 629		}
 630
 
 
 
 
 631	len = snprintf(buf, remaining, "\n");
 632	if (len >= remaining)
 633		return -EINVAL;
 634	return tlen + len;
 635}
 636
 637/**
 638 * write_versions - Set or report the available NFS protocol versions
 639 *
 640 * Input:
 641 *			buf:		ignored
 642 *			size:		zero
 643 * Output:
 644 *	On success:	passed-in buffer filled with '\n'-terminated C
 645 *			string containing positive or negative integer
 646 *			values representing the current status of each
 647 *			protocol version;
 648 *			return code is the size in bytes of the string
 649 *	On error:	return code is zero or a negative errno value
 650 *
 651 * OR
 652 *
 653 * Input:
 654 * 			buf:		C string containing whitespace-
 655 * 					separated positive or negative
 656 * 					integer values representing NFS
 657 * 					protocol versions to enable ("+n")
 658 * 					or disable ("-n")
 659 *			size:		non-zero length of C string in @buf
 660 * Output:
 661 *	On success:	status of zero or more protocol versions has
 662 *			been updated; passed-in buffer filled with
 663 *			'\n'-terminated C string containing positive
 664 *			or negative integer values representing the
 665 *			current status of each protocol version;
 666 *			return code is the size in bytes of the string
 667 *	On error:	return code is zero or a negative errno value
 668 */
 669static ssize_t write_versions(struct file *file, char *buf, size_t size)
 670{
 671	ssize_t rv;
 672
 673	mutex_lock(&nfsd_mutex);
 674	rv = __write_versions(file, buf, size);
 675	mutex_unlock(&nfsd_mutex);
 676	return rv;
 677}
 678
 679/*
 680 * Zero-length write.  Return a list of NFSD's current listener
 681 * transports.
 682 */
 683static ssize_t __write_ports_names(char *buf, struct net *net)
 684{
 685	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 686
 687	if (nn->nfsd_serv == NULL)
 688		return 0;
 689	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
 690}
 691
 692/*
 693 * A single 'fd' number was written, in which case it must be for
 694 * a socket of a supported family/protocol, and we use it as an
 695 * nfsd listener.
 696 */
 697static ssize_t __write_ports_addfd(char *buf, struct net *net)
 698{
 699	char *mesg = buf;
 700	int fd, err;
 701	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 
 702
 703	err = get_int(&mesg, &fd);
 704	if (err != 0 || fd < 0)
 705		return -EINVAL;
 706
 707	if (svc_alien_sock(net, fd)) {
 708		printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
 709		return -EINVAL;
 710	}
 711
 712	err = nfsd_create_serv(net);
 713	if (err != 0)
 714		return err;
 715
 716	err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
 717	if (err < 0) {
 718		nfsd_destroy(net);
 719		return err;
 720	}
 721
 722	/* Decrease the count, but don't shut down the service */
 723	nn->nfsd_serv->sv_nrthreads--;
 724	return err;
 725}
 726
 727/*
 728 * A transport listener is added by writing it's transport name and
 729 * a port number.
 730 */
 731static ssize_t __write_ports_addxprt(char *buf, struct net *net)
 732{
 733	char transport[16];
 734	struct svc_xprt *xprt;
 735	int port, err;
 736	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 
 737
 738	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
 739		return -EINVAL;
 740
 741	if (port < 1 || port > USHRT_MAX)
 742		return -EINVAL;
 
 743
 744	err = nfsd_create_serv(net);
 745	if (err != 0)
 746		return err;
 747
 748	err = svc_create_xprt(nn->nfsd_serv, transport, net,
 749				PF_INET, port, SVC_SOCK_ANONYMOUS);
 
 750	if (err < 0)
 751		goto out_err;
 752
 753	err = svc_create_xprt(nn->nfsd_serv, transport, net,
 754				PF_INET6, port, SVC_SOCK_ANONYMOUS);
 755	if (err < 0 && err != -EAFNOSUPPORT)
 756		goto out_close;
 757
 758	/* Decrease the count, but don't shut down the service */
 759	nn->nfsd_serv->sv_nrthreads--;
 760	return 0;
 761out_close:
 762	xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
 763	if (xprt != NULL) {
 764		svc_close_xprt(xprt);
 765		svc_xprt_put(xprt);
 766	}
 767out_err:
 768	nfsd_destroy(net);
 
 
 769	return err;
 770}
 771
 772static ssize_t __write_ports(struct file *file, char *buf, size_t size,
 773			     struct net *net)
 774{
 775	if (size == 0)
 776		return __write_ports_names(buf, net);
 777
 778	if (isdigit(buf[0]))
 779		return __write_ports_addfd(buf, net);
 780
 781	if (isalpha(buf[0]))
 782		return __write_ports_addxprt(buf, net);
 783
 784	return -EINVAL;
 785}
 786
 787/**
 788 * write_ports - Pass a socket file descriptor or transport name to listen on
 789 *
 790 * Input:
 791 *			buf:		ignored
 792 *			size:		zero
 793 * Output:
 794 *	On success:	passed-in buffer filled with a '\n'-terminated C
 795 *			string containing a whitespace-separated list of
 796 *			named NFSD listeners;
 797 *			return code is the size in bytes of the string
 798 *	On error:	return code is zero or a negative errno value
 799 *
 800 * OR
 801 *
 802 * Input:
 803 *			buf:		C string containing an unsigned
 804 *					integer value representing a bound
 805 *					but unconnected socket that is to be
 806 *					used as an NFSD listener; listen(3)
 807 *					must be called for a SOCK_STREAM
 808 *					socket, otherwise it is ignored
 809 *			size:		non-zero length of C string in @buf
 810 * Output:
 811 *	On success:	NFS service is started;
 812 *			passed-in buffer filled with a '\n'-terminated C
 813 *			string containing a unique alphanumeric name of
 814 *			the listener;
 815 *			return code is the size in bytes of the string
 816 *	On error:	return code is a negative errno value
 817 *
 818 * OR
 819 *
 820 * Input:
 821 *			buf:		C string containing a transport
 822 *					name and an unsigned integer value
 823 *					representing the port to listen on,
 824 *					separated by whitespace
 825 *			size:		non-zero length of C string in @buf
 826 * Output:
 827 *	On success:	returns zero; NFS service is started
 828 *	On error:	return code is a negative errno value
 829 */
 830static ssize_t write_ports(struct file *file, char *buf, size_t size)
 831{
 832	ssize_t rv;
 833
 834	mutex_lock(&nfsd_mutex);
 835	rv = __write_ports(file, buf, size, netns(file));
 836	mutex_unlock(&nfsd_mutex);
 837	return rv;
 838}
 839
 840
 841int nfsd_max_blksize;
 842
 843/**
 844 * write_maxblksize - Set or report the current NFS blksize
 845 *
 846 * Input:
 847 *			buf:		ignored
 848 *			size:		zero
 849 *
 850 * OR
 851 *
 852 * Input:
 853 * 			buf:		C string containing an unsigned
 854 * 					integer value representing the new
 855 * 					NFS blksize
 856 *			size:		non-zero length of C string in @buf
 857 * Output:
 858 *	On success:	passed-in buffer filled with '\n'-terminated C string
 859 *			containing numeric value of the current NFS blksize
 860 *			setting;
 861 *			return code is the size in bytes of the string
 862 *	On error:	return code is zero or a negative errno value
 863 */
 864static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
 865{
 866	char *mesg = buf;
 867	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 868
 869	if (size > 0) {
 870		int bsize;
 871		int rv = get_int(&mesg, &bsize);
 872		if (rv)
 873			return rv;
 
 
 874		/* force bsize into allowed range and
 875		 * required alignment.
 876		 */
 877		bsize = max_t(int, bsize, 1024);
 878		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
 879		bsize &= ~(1024-1);
 880		mutex_lock(&nfsd_mutex);
 881		if (nn->nfsd_serv) {
 882			mutex_unlock(&nfsd_mutex);
 883			return -EBUSY;
 884		}
 885		nfsd_max_blksize = bsize;
 886		mutex_unlock(&nfsd_mutex);
 887	}
 888
 889	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
 890							nfsd_max_blksize);
 891}
 892
 893/**
 894 * write_maxconn - Set or report the current max number of connections
 895 *
 896 * Input:
 897 *			buf:		ignored
 898 *			size:		zero
 899 * OR
 900 *
 901 * Input:
 902 * 			buf:		C string containing an unsigned
 903 * 					integer value representing the new
 904 * 					number of max connections
 905 *			size:		non-zero length of C string in @buf
 906 * Output:
 907 *	On success:	passed-in buffer filled with '\n'-terminated C string
 908 *			containing numeric value of max_connections setting
 909 *			for this net namespace;
 910 *			return code is the size in bytes of the string
 911 *	On error:	return code is zero or a negative errno value
 912 */
 913static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
 914{
 915	char *mesg = buf;
 916	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 917	unsigned int maxconn = nn->max_connections;
 918
 919	if (size > 0) {
 920		int rv = get_uint(&mesg, &maxconn);
 921
 922		if (rv)
 923			return rv;
 
 924		nn->max_connections = maxconn;
 925	}
 926
 927	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
 928}
 929
 930#ifdef CONFIG_NFSD_V4
 931static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
 932				  time_t *time, struct nfsd_net *nn)
 933{
 
 934	char *mesg = buf;
 935	int rv, i;
 936
 937	if (size > 0) {
 938		if (nn->nfsd_serv)
 939			return -EBUSY;
 940		rv = get_int(&mesg, &i);
 941		if (rv)
 942			return rv;
 
 
 
 943		/*
 944		 * Some sanity checking.  We don't have a reason for
 945		 * these particular numbers, but problems with the
 946		 * extremes are:
 947		 *	- Too short: the briefest network outage may
 948		 *	  cause clients to lose all their locks.  Also,
 949		 *	  the frequent polling may be wasteful.
 950		 *	- Too long: do you really want reboot recovery
 951		 *	  to take more than an hour?  Or to make other
 952		 *	  clients wait an hour before being able to
 953		 *	  revoke a dead client's locks?
 954		 */
 955		if (i < 10 || i > 3600)
 956			return -EINVAL;
 957		*time = i;
 958	}
 959
 960	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
 961}
 962
 963static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
 964				time_t *time, struct nfsd_net *nn)
 965{
 966	ssize_t rv;
 967
 968	mutex_lock(&nfsd_mutex);
 969	rv = __nfsd4_write_time(file, buf, size, time, nn);
 970	mutex_unlock(&nfsd_mutex);
 971	return rv;
 972}
 973
 974/**
 975 * write_leasetime - Set or report the current NFSv4 lease time
 976 *
 977 * Input:
 978 *			buf:		ignored
 979 *			size:		zero
 980 *
 981 * OR
 982 *
 983 * Input:
 984 *			buf:		C string containing an unsigned
 985 *					integer value representing the new
 986 *					NFSv4 lease expiry time
 987 *			size:		non-zero length of C string in @buf
 988 * Output:
 989 *	On success:	passed-in buffer filled with '\n'-terminated C
 990 *			string containing unsigned integer value of the
 991 *			current lease expiry time;
 992 *			return code is the size in bytes of the string
 993 *	On error:	return code is zero or a negative errno value
 994 */
 995static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
 996{
 997	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 998	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
 999}
1000
1001/**
1002 * write_gracetime - Set or report current NFSv4 grace period time
1003 *
1004 * As above, but sets the time of the NFSv4 grace period.
1005 *
1006 * Note this should never be set to less than the *previous*
1007 * lease-period time, but we don't try to enforce this.  (In the common
1008 * case (a new boot), we don't know what the previous lease time was
1009 * anyway.)
1010 */
1011static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1012{
1013	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1014	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1015}
1016
 
1017static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1018				   struct nfsd_net *nn)
1019{
1020	char *mesg = buf;
1021	char *recdir;
1022	int len, status;
1023
1024	if (size > 0) {
1025		if (nn->nfsd_serv)
1026			return -EBUSY;
1027		if (size > PATH_MAX || buf[size-1] != '\n')
1028			return -EINVAL;
1029		buf[size-1] = 0;
1030
1031		recdir = mesg;
1032		len = qword_get(&mesg, recdir, size);
1033		if (len <= 0)
1034			return -EINVAL;
 
1035
1036		status = nfs4_reset_recoverydir(recdir);
1037		if (status)
1038			return status;
1039	}
1040
1041	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1042							nfs4_recoverydir());
1043}
1044
1045/**
1046 * write_recoverydir - Set or report the pathname of the recovery directory
1047 *
1048 * Input:
1049 *			buf:		ignored
1050 *			size:		zero
1051 *
1052 * OR
1053 *
1054 * Input:
1055 *			buf:		C string containing the pathname
1056 *					of the directory on a local file
1057 *					system containing permanent NFSv4
1058 *					recovery data
1059 *			size:		non-zero length of C string in @buf
1060 * Output:
1061 *	On success:	passed-in buffer filled with '\n'-terminated C string
1062 *			containing the current recovery pathname setting;
1063 *			return code is the size in bytes of the string
1064 *	On error:	return code is zero or a negative errno value
1065 */
1066static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1067{
1068	ssize_t rv;
1069	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1070
1071	mutex_lock(&nfsd_mutex);
1072	rv = __write_recoverydir(file, buf, size, nn);
1073	mutex_unlock(&nfsd_mutex);
1074	return rv;
1075}
 
1076
1077/**
1078 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1079 *
1080 * Input:
1081 *			buf:		ignored
1082 *			size:		zero
1083 * OR
1084 *
1085 * Input:
1086 * 			buf:		any value
1087 *			size:		non-zero length of C string in @buf
1088 * Output:
1089 *			passed-in buffer filled with "Y" or "N" with a newline
1090 *			and NULL-terminated C string. This indicates whether
1091 *			the grace period has ended in the current net
1092 *			namespace. Return code is the size in bytes of the
1093 *			string. Writing a string that starts with 'Y', 'y', or
1094 *			'1' to the file will end the grace period for nfsd's v4
1095 *			lock manager.
1096 */
1097static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1098{
1099	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1100
1101	if (size > 0) {
1102		switch(buf[0]) {
1103		case 'Y':
1104		case 'y':
1105		case '1':
 
 
 
1106			nfsd4_end_grace(nn);
1107			break;
1108		default:
1109			return -EINVAL;
1110		}
1111	}
1112
1113	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1114			 nn->grace_ended ? 'Y' : 'N');
1115}
1116
1117#endif
1118
1119/*----------------------------------------------------------------------------*/
1120/*
1121 *	populating the filesystem.
1122 */
1123
1124static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1125{
1126	static struct tree_descr nfsd_files[] = {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1127		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
 
 
1128		[NFSD_Export_features] = {"export_features",
1129					&export_features_operations, S_IRUGO},
1130		[NFSD_FO_UnlockIP] = {"unlock_ip",
1131					&transaction_ops, S_IWUSR|S_IRUSR},
1132		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1133					&transaction_ops, S_IWUSR|S_IRUSR},
1134		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1135		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1136		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1137		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1138		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
 
1139		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1140		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1141		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1142		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1143#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1144		[NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1145#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1146#ifdef CONFIG_NFSD_V4
1147		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1148		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1149		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1150		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1151#endif
1152		/* last one */ {""}
1153	};
1154	get_net(sb->s_fs_info);
1155	return simple_fill_super(sb, 0x6e667364, nfsd_files);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1156}
1157
1158static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1159	int flags, const char *dev_name, void *data)
 
 
 
 
 
 
 
 
 
 
1160{
1161	struct net *net = current->nsproxy->net_ns;
1162	return mount_ns(fs_type, flags, data, net, net->user_ns, nfsd_fill_super);
 
 
1163}
1164
1165static void nfsd_umount(struct super_block *sb)
1166{
1167	struct net *net = sb->s_fs_info;
1168
 
 
1169	kill_litter_super(sb);
1170	put_net(net);
1171}
1172
1173static struct file_system_type nfsd_fs_type = {
1174	.owner		= THIS_MODULE,
1175	.name		= "nfsd",
1176	.mount		= nfsd_mount,
1177	.kill_sb	= nfsd_umount,
1178};
1179MODULE_ALIAS_FS("nfsd");
1180
1181#ifdef CONFIG_PROC_FS
 
 
 
 
 
 
 
 
 
 
 
 
 
1182static int create_proc_exports_entry(void)
1183{
1184	struct proc_dir_entry *entry;
1185
1186	entry = proc_mkdir("fs/nfs", NULL);
1187	if (!entry)
1188		return -ENOMEM;
1189	entry = proc_create("exports", 0, entry,
1190				 &exports_proc_operations);
1191	if (!entry) {
1192		remove_proc_entry("fs/nfs", NULL);
1193		return -ENOMEM;
1194	}
1195	return 0;
1196}
1197#else /* CONFIG_PROC_FS */
1198static int create_proc_exports_entry(void)
1199{
1200	return 0;
1201}
1202#endif
1203
1204unsigned int nfsd_net_id;
1205
1206static __net_init int nfsd_init_net(struct net *net)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1207{
1208	int retval;
1209	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1210
1211	retval = nfsd_export_init(net);
1212	if (retval)
1213		goto out_export_error;
1214	retval = nfsd_idmap_init(net);
1215	if (retval)
1216		goto out_idmap_error;
1217	nn->nfsd4_lease = 90;	/* default lease time */
1218	nn->nfsd4_grace = 90;
1219	nn->clverifier_counter = prandom_u32();
1220	nn->clientid_counter = prandom_u32();
 
 
 
 
 
1221	return 0;
1222
 
 
1223out_idmap_error:
1224	nfsd_export_shutdown(net);
1225out_export_error:
1226	return retval;
1227}
1228
1229static __net_exit void nfsd_exit_net(struct net *net)
 
 
 
 
 
1230{
 
 
 
1231	nfsd_idmap_shutdown(net);
1232	nfsd_export_shutdown(net);
 
1233}
1234
1235static struct pernet_operations nfsd_net_ops = {
1236	.init = nfsd_init_net,
1237	.exit = nfsd_exit_net,
1238	.id   = &nfsd_net_id,
1239	.size = sizeof(struct nfsd_net),
1240};
1241
1242static int __init init_nfsd(void)
1243{
1244	int retval;
1245	printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1246
1247	retval = register_pernet_subsys(&nfsd_net_ops);
1248	if (retval < 0)
1249		return retval;
1250	retval = register_cld_notifier();
1251	if (retval)
1252		goto out_unregister_pernet;
1253	retval = nfsd4_init_slabs();
1254	if (retval)
1255		goto out_unregister_notifier;
1256	retval = nfsd4_init_pnfs();
1257	if (retval)
1258		goto out_free_slabs;
1259	retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1260	if (retval)
1261		goto out_exit_pnfs;
1262	nfsd_stat_init();	/* Statistics */
1263	retval = nfsd_reply_cache_init();
1264	if (retval)
1265		goto out_free_stat;
1266	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1267	retval = create_proc_exports_entry();
1268	if (retval)
1269		goto out_free_lockd;
 
 
 
 
 
 
 
 
 
1270	retval = register_filesystem(&nfsd_fs_type);
1271	if (retval)
1272		goto out_free_all;
 
 
 
 
1273	return 0;
1274out_free_all:
 
 
 
 
 
 
1275	remove_proc_entry("fs/nfs/exports", NULL);
1276	remove_proc_entry("fs/nfs", NULL);
1277out_free_lockd:
1278	nfsd_lockd_shutdown();
1279	nfsd_reply_cache_shutdown();
1280out_free_stat:
1281	nfsd_stat_shutdown();
1282	nfsd_fault_inject_cleanup();
1283out_exit_pnfs:
1284	nfsd4_exit_pnfs();
1285out_free_slabs:
1286	nfsd4_free_slabs();
1287out_unregister_notifier:
1288	unregister_cld_notifier();
1289out_unregister_pernet:
1290	unregister_pernet_subsys(&nfsd_net_ops);
1291	return retval;
1292}
1293
1294static void __exit exit_nfsd(void)
1295{
1296	nfsd_reply_cache_shutdown();
 
 
 
 
 
1297	remove_proc_entry("fs/nfs/exports", NULL);
1298	remove_proc_entry("fs/nfs", NULL);
1299	nfsd_stat_shutdown();
1300	nfsd_lockd_shutdown();
1301	nfsd4_free_slabs();
1302	nfsd4_exit_pnfs();
1303	nfsd_fault_inject_cleanup();
1304	unregister_filesystem(&nfsd_fs_type);
1305	unregister_cld_notifier();
1306	unregister_pernet_subsys(&nfsd_net_ops);
1307}
1308
1309MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
 
1310MODULE_LICENSE("GPL");
1311module_init(init_nfsd)
1312module_exit(exit_nfsd)
v6.8
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * Syscall interface to knfsd.
   4 *
   5 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
   6 */
   7
   8#include <linux/slab.h>
   9#include <linux/namei.h>
  10#include <linux/ctype.h>
  11#include <linux/fs_context.h>
  12
  13#include <linux/sunrpc/svcsock.h>
  14#include <linux/lockd/lockd.h>
  15#include <linux/sunrpc/addr.h>
  16#include <linux/sunrpc/gss_api.h>
 
  17#include <linux/sunrpc/rpc_pipe_fs.h>
  18#include <linux/module.h>
  19#include <linux/fsnotify.h>
  20
  21#include "idmap.h"
  22#include "nfsd.h"
  23#include "cache.h"
  24#include "state.h"
  25#include "netns.h"
  26#include "pnfs.h"
  27#include "filecache.h"
  28#include "trace.h"
  29#include "netlink.h"
  30
  31/*
  32 *	We have a single directory with several nodes in it.
  33 */
  34enum {
  35	NFSD_Root = 1,
  36	NFSD_List,
  37	NFSD_Export_Stats,
  38	NFSD_Export_features,
  39	NFSD_Fh,
  40	NFSD_FO_UnlockIP,
  41	NFSD_FO_UnlockFS,
  42	NFSD_Threads,
  43	NFSD_Pool_Threads,
  44	NFSD_Pool_Stats,
  45	NFSD_Reply_Cache_Stats,
  46	NFSD_Versions,
  47	NFSD_Ports,
  48	NFSD_MaxBlkSize,
  49	NFSD_MaxConnections,
  50	NFSD_Filecache,
 
 
 
 
  51#ifdef CONFIG_NFSD_V4
  52	NFSD_Leasetime,
  53	NFSD_Gracetime,
  54	NFSD_RecoveryDir,
  55	NFSD_V4EndGrace,
  56#endif
  57	NFSD_MaxReserved
  58};
  59
  60/*
  61 * write() for these nodes.
  62 */
  63static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
  64static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
  65static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
  66static ssize_t write_threads(struct file *file, char *buf, size_t size);
  67static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
  68static ssize_t write_versions(struct file *file, char *buf, size_t size);
  69static ssize_t write_ports(struct file *file, char *buf, size_t size);
  70static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
  71static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
  72#ifdef CONFIG_NFSD_V4
  73static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
  74static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
  75#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
  76static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
  77#endif
  78static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
  79#endif
  80
  81static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
  82	[NFSD_Fh] = write_filehandle,
  83	[NFSD_FO_UnlockIP] = write_unlock_ip,
  84	[NFSD_FO_UnlockFS] = write_unlock_fs,
  85	[NFSD_Threads] = write_threads,
  86	[NFSD_Pool_Threads] = write_pool_threads,
  87	[NFSD_Versions] = write_versions,
  88	[NFSD_Ports] = write_ports,
  89	[NFSD_MaxBlkSize] = write_maxblksize,
  90	[NFSD_MaxConnections] = write_maxconn,
  91#ifdef CONFIG_NFSD_V4
  92	[NFSD_Leasetime] = write_leasetime,
  93	[NFSD_Gracetime] = write_gracetime,
  94#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
  95	[NFSD_RecoveryDir] = write_recoverydir,
  96#endif
  97	[NFSD_V4EndGrace] = write_v4_end_grace,
  98#endif
  99};
 100
 101static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
 102{
 103	ino_t ino =  file_inode(file)->i_ino;
 104	char *data;
 105	ssize_t rv;
 106
 107	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
 108		return -EINVAL;
 109
 110	data = simple_transaction_get(file, buf, size);
 111	if (IS_ERR(data))
 112		return PTR_ERR(data);
 113
 114	rv = write_op[ino](file, data, size);
 115	if (rv < 0)
 116		return rv;
 117
 118	simple_transaction_set(file, rv);
 119	return size;
 120}
 121
 122static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 123{
 124	if (! file->private_data) {
 125		/* An attempt to read a transaction file without writing
 126		 * causes a 0-byte write so that the file can return
 127		 * state information
 128		 */
 129		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
 130		if (rv < 0)
 131			return rv;
 132	}
 133	return simple_transaction_read(file, buf, size, pos);
 134}
 135
 136static const struct file_operations transaction_ops = {
 137	.write		= nfsctl_transaction_write,
 138	.read		= nfsctl_transaction_read,
 139	.release	= simple_transaction_release,
 140	.llseek		= default_llseek,
 141};
 142
 143static int exports_net_open(struct net *net, struct file *file)
 144{
 145	int err;
 146	struct seq_file *seq;
 147	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 148
 149	err = seq_open(file, &nfs_exports_op);
 150	if (err)
 151		return err;
 152
 153	seq = file->private_data;
 154	seq->private = nn->svc_export_cache;
 155	return 0;
 156}
 157
 
 
 
 
 
 
 
 
 
 
 
 
 158static int exports_nfsd_open(struct inode *inode, struct file *file)
 159{
 160	return exports_net_open(inode->i_sb->s_fs_info, file);
 161}
 162
 163static const struct file_operations exports_nfsd_operations = {
 164	.open		= exports_nfsd_open,
 165	.read		= seq_read,
 166	.llseek		= seq_lseek,
 167	.release	= seq_release,
 168};
 169
 170static int export_features_show(struct seq_file *m, void *v)
 171{
 172	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
 173	return 0;
 174}
 175
 176DEFINE_SHOW_ATTRIBUTE(export_features);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 177
 178static const struct file_operations pool_stats_operations = {
 179	.open		= nfsd_pool_stats_open,
 180	.read		= seq_read,
 181	.llseek		= seq_lseek,
 182	.release	= seq_release,
 183};
 184
 185DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
 186
 187DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
 
 
 
 188
 189/*----------------------------------------------------------------------------*/
 190/*
 191 * payload - write methods
 192 */
 193
 194static inline struct net *netns(struct file *file)
 195{
 196	return file_inode(file)->i_sb->s_fs_info;
 197}
 198
 199/*
 200 * write_unlock_ip - Release all locks used by a client
 201 *
 202 * Experimental.
 203 *
 204 * Input:
 205 *			buf:	'\n'-terminated C string containing a
 206 *				presentation format IP address
 207 *			size:	length of C string in @buf
 208 * Output:
 209 *	On success:	returns zero if all specified locks were released;
 210 *			returns one if one or more locks were not released
 211 *	On error:	return code is negative errno value
 212 */
 213static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
 214{
 215	struct sockaddr_storage address;
 216	struct sockaddr *sap = (struct sockaddr *)&address;
 217	size_t salen = sizeof(address);
 218	char *fo_path;
 219	struct net *net = netns(file);
 220
 221	/* sanity check */
 222	if (size == 0)
 223		return -EINVAL;
 224
 225	if (buf[size-1] != '\n')
 226		return -EINVAL;
 227
 228	fo_path = buf;
 229	if (qword_get(&buf, fo_path, size) < 0)
 230		return -EINVAL;
 231
 232	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
 233		return -EINVAL;
 234
 235	trace_nfsd_ctl_unlock_ip(net, buf);
 236	return nlmsvc_unlock_all_by_ip(sap);
 237}
 238
 239/*
 240 * write_unlock_fs - Release all locks on a local file system
 241 *
 242 * Experimental.
 243 *
 244 * Input:
 245 *			buf:	'\n'-terminated C string containing the
 246 *				absolute pathname of a local file system
 247 *			size:	length of C string in @buf
 248 * Output:
 249 *	On success:	returns zero if all specified locks were released;
 250 *			returns one if one or more locks were not released
 251 *	On error:	return code is negative errno value
 252 */
 253static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
 254{
 255	struct path path;
 256	char *fo_path;
 257	int error;
 258
 259	/* sanity check */
 260	if (size == 0)
 261		return -EINVAL;
 262
 263	if (buf[size-1] != '\n')
 264		return -EINVAL;
 265
 266	fo_path = buf;
 267	if (qword_get(&buf, fo_path, size) < 0)
 268		return -EINVAL;
 269	trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
 270	error = kern_path(fo_path, 0, &path);
 271	if (error)
 272		return error;
 273
 274	/*
 275	 * XXX: Needs better sanity checking.  Otherwise we could end up
 276	 * releasing locks on the wrong file system.
 277	 *
 278	 * For example:
 279	 * 1.  Does the path refer to a directory?
 280	 * 2.  Is that directory a mount point, or
 281	 * 3.  Is that directory the root of an exported file system?
 282	 */
 283	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
 284
 285	path_put(&path);
 286	return error;
 287}
 288
 289/*
 290 * write_filehandle - Get a variable-length NFS file handle by path
 291 *
 292 * On input, the buffer contains a '\n'-terminated C string comprised of
 293 * three alphanumeric words separated by whitespace.  The string may
 294 * contain escape sequences.
 295 *
 296 * Input:
 297 *			buf:
 298 *				domain:		client domain name
 299 *				path:		export pathname
 300 *				maxsize:	numeric maximum size of
 301 *						@buf
 302 *			size:	length of C string in @buf
 303 * Output:
 304 *	On success:	passed-in buffer filled with '\n'-terminated C
 305 *			string containing a ASCII hex text version
 306 *			of the NFS file handle;
 307 *			return code is the size in bytes of the string
 308 *	On error:	return code is negative errno value
 309 */
 310static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
 311{
 312	char *dname, *path;
 313	int maxsize;
 314	char *mesg = buf;
 315	int len;
 316	struct auth_domain *dom;
 317	struct knfsd_fh fh;
 318
 319	if (size == 0)
 320		return -EINVAL;
 321
 322	if (buf[size-1] != '\n')
 323		return -EINVAL;
 324	buf[size-1] = 0;
 325
 326	dname = mesg;
 327	len = qword_get(&mesg, dname, size);
 328	if (len <= 0)
 329		return -EINVAL;
 330
 331	path = dname+len+1;
 332	len = qword_get(&mesg, path, size);
 333	if (len <= 0)
 334		return -EINVAL;
 335
 336	len = get_int(&mesg, &maxsize);
 337	if (len)
 338		return len;
 339
 340	if (maxsize < NFS_FHSIZE)
 341		return -EINVAL;
 342	maxsize = min(maxsize, NFS3_FHSIZE);
 343
 344	if (qword_get(&mesg, mesg, size) > 0)
 345		return -EINVAL;
 346
 347	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
 348
 349	/* we have all the words, they are in buf.. */
 350	dom = unix_domain_find(dname);
 351	if (!dom)
 352		return -ENOMEM;
 353
 354	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
 355	auth_domain_put(dom);
 356	if (len)
 357		return len;
 358
 359	mesg = buf;
 360	len = SIMPLE_TRANSACTION_LIMIT;
 361	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
 362	mesg[-1] = '\n';
 363	return mesg - buf;
 364}
 365
 366/*
 367 * write_threads - Start NFSD, or report the current number of running threads
 368 *
 369 * Input:
 370 *			buf:		ignored
 371 *			size:		zero
 372 * Output:
 373 *	On success:	passed-in buffer filled with '\n'-terminated C
 374 *			string numeric value representing the number of
 375 *			running NFSD threads;
 376 *			return code is the size in bytes of the string
 377 *	On error:	return code is zero
 378 *
 379 * OR
 380 *
 381 * Input:
 382 *			buf:		C string containing an unsigned
 383 *					integer value representing the
 384 *					number of NFSD threads to start
 385 *			size:		non-zero length of C string in @buf
 386 * Output:
 387 *	On success:	NFS service is started;
 388 *			passed-in buffer filled with '\n'-terminated C
 389 *			string numeric value representing the number of
 390 *			running NFSD threads;
 391 *			return code is the size in bytes of the string
 392 *	On error:	return code is zero or a negative errno value
 393 */
 394static ssize_t write_threads(struct file *file, char *buf, size_t size)
 395{
 396	char *mesg = buf;
 397	int rv;
 398	struct net *net = netns(file);
 399
 400	if (size > 0) {
 401		int newthreads;
 402		rv = get_int(&mesg, &newthreads);
 403		if (rv)
 404			return rv;
 405		if (newthreads < 0)
 406			return -EINVAL;
 407		trace_nfsd_ctl_threads(net, newthreads);
 408		rv = nfsd_svc(newthreads, net, file->f_cred);
 409		if (rv < 0)
 410			return rv;
 411	} else
 412		rv = nfsd_nrthreads(net);
 413
 414	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
 415}
 416
 417/*
 418 * write_pool_threads - Set or report the current number of threads per pool
 419 *
 420 * Input:
 421 *			buf:		ignored
 422 *			size:		zero
 423 *
 424 * OR
 425 *
 426 * Input:
 427 *			buf:		C string containing whitespace-
 428 *					separated unsigned integer values
 429 *					representing the number of NFSD
 430 *					threads to start in each pool
 431 *			size:		non-zero length of C string in @buf
 432 * Output:
 433 *	On success:	passed-in buffer filled with '\n'-terminated C
 434 *			string containing integer values representing the
 435 *			number of NFSD threads in each pool;
 436 *			return code is the size in bytes of the string
 437 *	On error:	return code is zero or a negative errno value
 438 */
 439static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
 440{
 441	/* if size > 0, look for an array of number of threads per node
 442	 * and apply them  then write out number of threads per node as reply
 443	 */
 444	char *mesg = buf;
 445	int i;
 446	int rv;
 447	int len;
 448	int npools;
 449	int *nthreads;
 450	struct net *net = netns(file);
 451
 452	mutex_lock(&nfsd_mutex);
 453	npools = nfsd_nrpools(net);
 454	if (npools == 0) {
 455		/*
 456		 * NFS is shut down.  The admin can start it by
 457		 * writing to the threads file but NOT the pool_threads
 458		 * file, sorry.  Report zero threads.
 459		 */
 460		mutex_unlock(&nfsd_mutex);
 461		strcpy(buf, "0\n");
 462		return strlen(buf);
 463	}
 464
 465	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
 466	rv = -ENOMEM;
 467	if (nthreads == NULL)
 468		goto out_free;
 469
 470	if (size > 0) {
 471		for (i = 0; i < npools; i++) {
 472			rv = get_int(&mesg, &nthreads[i]);
 473			if (rv == -ENOENT)
 474				break;		/* fewer numbers than pools */
 475			if (rv)
 476				goto out_free;	/* syntax error */
 477			rv = -EINVAL;
 478			if (nthreads[i] < 0)
 479				goto out_free;
 480			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
 481		}
 482		rv = nfsd_set_nrthreads(i, nthreads, net);
 483		if (rv)
 484			goto out_free;
 485	}
 486
 487	rv = nfsd_get_nrthreads(npools, nthreads, net);
 488	if (rv)
 489		goto out_free;
 490
 491	mesg = buf;
 492	size = SIMPLE_TRANSACTION_LIMIT;
 493	for (i = 0; i < npools && size > 0; i++) {
 494		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
 495		len = strlen(mesg);
 496		size -= len;
 497		mesg += len;
 498	}
 499	rv = mesg - buf;
 500out_free:
 501	kfree(nthreads);
 502	mutex_unlock(&nfsd_mutex);
 503	return rv;
 504}
 505
 506static ssize_t
 507nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
 508		const char *sep, unsigned vers, int minor)
 509{
 510	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
 511	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
 512
 513	if (vers == 4 && minor >= 0 &&
 514	    !nfsd_minorversion(nn, minor, NFSD_TEST))
 515		supported = false;
 516	if (minor == 0 && supported)
 517		/*
 518		 * special case for backward compatability.
 519		 * +4.0 is never reported, it is implied by
 520		 * +4, unless -4.0 is present.
 521		 */
 522		return 0;
 523	return snprintf(buf, remaining, format, sep,
 524			supported ? '+' : '-', vers, minor);
 525}
 526
 527static ssize_t __write_versions(struct file *file, char *buf, size_t size)
 528{
 529	char *mesg = buf;
 530	char *vers, *minorp, sign;
 531	int len, num, remaining;
 
 532	ssize_t tlen = 0;
 533	char *sep;
 534	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 535
 536	if (size > 0) {
 537		if (nn->nfsd_serv)
 538			/* Cannot change versions without updating
 539			 * nn->nfsd_serv->sv_xdrsize, and reallocing
 540			 * rq_argp and rq_resp
 541			 */
 542			return -EBUSY;
 543		if (buf[size-1] != '\n')
 544			return -EINVAL;
 545		buf[size-1] = 0;
 546		trace_nfsd_ctl_version(netns(file), buf);
 547
 548		vers = mesg;
 549		len = qword_get(&mesg, vers, size);
 550		if (len <= 0) return -EINVAL;
 551		do {
 552			enum vers_op cmd;
 553			unsigned minor;
 554			sign = *vers;
 555			if (sign == '+' || sign == '-')
 556				num = simple_strtol((vers+1), &minorp, 0);
 557			else
 558				num = simple_strtol(vers, &minorp, 0);
 559			if (*minorp == '.') {
 560				if (num != 4)
 561					return -EINVAL;
 562				if (kstrtouint(minorp+1, 0, &minor) < 0)
 
 563					return -EINVAL;
 
 
 
 
 564			}
 565
 566			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
 567			switch(num) {
 568#ifdef CONFIG_NFSD_V2
 569			case 2:
 570#endif
 571			case 3:
 572				nfsd_vers(nn, num, cmd);
 573				break;
 574			case 4:
 575				if (*minorp == '.') {
 576					if (nfsd_minorversion(nn, minor, cmd) < 0)
 577						return -EINVAL;
 578				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
 579					/*
 580					 * Either we have +4 and no minors are enabled,
 581					 * or we have -4 and at least one minor is enabled.
 582					 * In either case, propagate 'cmd' to all minors.
 583					 */
 584					minor = 0;
 585					while (nfsd_minorversion(nn, minor, cmd) >= 0)
 586						minor++;
 587				}
 588				break;
 589			default:
 590				/* Ignore requests to disable non-existent versions */
 591				if (cmd == NFSD_SET)
 592					return -EINVAL;
 593			}
 
 594			vers += len + 1;
 595		} while ((len = qword_get(&mesg, vers, size)) > 0);
 596		/* If all get turned off, turn them back on, as
 597		 * having no versions is BAD
 598		 */
 599		nfsd_reset_versions(nn);
 600	}
 601
 602	/* Now write current state into reply buffer */
 
 603	sep = "";
 604	remaining = SIMPLE_TRANSACTION_LIMIT;
 605	for (num=2 ; num <= 4 ; num++) {
 606		int minor;
 607		if (!nfsd_vers(nn, num, NFSD_AVAIL))
 608			continue;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 609
 610		minor = -1;
 611		do {
 612			len = nfsd_print_version_support(nn, buf, remaining,
 613					sep, num, minor);
 614			if (len >= remaining)
 615				goto out;
 616			remaining -= len;
 617			buf += len;
 618			tlen += len;
 619			minor++;
 620			if (len)
 621				sep = " ";
 622		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
 623	}
 624out:
 625	len = snprintf(buf, remaining, "\n");
 626	if (len >= remaining)
 627		return -EINVAL;
 628	return tlen + len;
 629}
 630
 631/*
 632 * write_versions - Set or report the available NFS protocol versions
 633 *
 634 * Input:
 635 *			buf:		ignored
 636 *			size:		zero
 637 * Output:
 638 *	On success:	passed-in buffer filled with '\n'-terminated C
 639 *			string containing positive or negative integer
 640 *			values representing the current status of each
 641 *			protocol version;
 642 *			return code is the size in bytes of the string
 643 *	On error:	return code is zero or a negative errno value
 644 *
 645 * OR
 646 *
 647 * Input:
 648 *			buf:		C string containing whitespace-
 649 *					separated positive or negative
 650 *					integer values representing NFS
 651 *					protocol versions to enable ("+n")
 652 *					or disable ("-n")
 653 *			size:		non-zero length of C string in @buf
 654 * Output:
 655 *	On success:	status of zero or more protocol versions has
 656 *			been updated; passed-in buffer filled with
 657 *			'\n'-terminated C string containing positive
 658 *			or negative integer values representing the
 659 *			current status of each protocol version;
 660 *			return code is the size in bytes of the string
 661 *	On error:	return code is zero or a negative errno value
 662 */
 663static ssize_t write_versions(struct file *file, char *buf, size_t size)
 664{
 665	ssize_t rv;
 666
 667	mutex_lock(&nfsd_mutex);
 668	rv = __write_versions(file, buf, size);
 669	mutex_unlock(&nfsd_mutex);
 670	return rv;
 671}
 672
 673/*
 674 * Zero-length write.  Return a list of NFSD's current listener
 675 * transports.
 676 */
 677static ssize_t __write_ports_names(char *buf, struct net *net)
 678{
 679	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 680
 681	if (nn->nfsd_serv == NULL)
 682		return 0;
 683	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
 684}
 685
 686/*
 687 * A single 'fd' number was written, in which case it must be for
 688 * a socket of a supported family/protocol, and we use it as an
 689 * nfsd listener.
 690 */
 691static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
 692{
 693	char *mesg = buf;
 694	int fd, err;
 695	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 696	struct svc_serv *serv;
 697
 698	err = get_int(&mesg, &fd);
 699	if (err != 0 || fd < 0)
 700		return -EINVAL;
 701	trace_nfsd_ctl_ports_addfd(net, fd);
 
 
 
 
 702
 703	err = nfsd_create_serv(net);
 704	if (err != 0)
 705		return err;
 706
 707	serv = nn->nfsd_serv;
 708	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
 709
 710	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
 711		nfsd_destroy_serv(net);
 712
 
 
 713	return err;
 714}
 715
 716/*
 717 * A transport listener is added by writing its transport name and
 718 * a port number.
 719 */
 720static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
 721{
 722	char transport[16];
 723	struct svc_xprt *xprt;
 724	int port, err;
 725	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 726	struct svc_serv *serv;
 727
 728	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
 729		return -EINVAL;
 730
 731	if (port < 1 || port > USHRT_MAX)
 732		return -EINVAL;
 733	trace_nfsd_ctl_ports_addxprt(net, transport, port);
 734
 735	err = nfsd_create_serv(net);
 736	if (err != 0)
 737		return err;
 738
 739	serv = nn->nfsd_serv;
 740	err = svc_xprt_create(serv, transport, net,
 741			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
 742	if (err < 0)
 743		goto out_err;
 744
 745	err = svc_xprt_create(serv, transport, net,
 746			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
 747	if (err < 0 && err != -EAFNOSUPPORT)
 748		goto out_close;
 749
 
 
 750	return 0;
 751out_close:
 752	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
 753	if (xprt != NULL) {
 754		svc_xprt_close(xprt);
 755		svc_xprt_put(xprt);
 756	}
 757out_err:
 758	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
 759		nfsd_destroy_serv(net);
 760
 761	return err;
 762}
 763
 764static ssize_t __write_ports(struct file *file, char *buf, size_t size,
 765			     struct net *net)
 766{
 767	if (size == 0)
 768		return __write_ports_names(buf, net);
 769
 770	if (isdigit(buf[0]))
 771		return __write_ports_addfd(buf, net, file->f_cred);
 772
 773	if (isalpha(buf[0]))
 774		return __write_ports_addxprt(buf, net, file->f_cred);
 775
 776	return -EINVAL;
 777}
 778
 779/*
 780 * write_ports - Pass a socket file descriptor or transport name to listen on
 781 *
 782 * Input:
 783 *			buf:		ignored
 784 *			size:		zero
 785 * Output:
 786 *	On success:	passed-in buffer filled with a '\n'-terminated C
 787 *			string containing a whitespace-separated list of
 788 *			named NFSD listeners;
 789 *			return code is the size in bytes of the string
 790 *	On error:	return code is zero or a negative errno value
 791 *
 792 * OR
 793 *
 794 * Input:
 795 *			buf:		C string containing an unsigned
 796 *					integer value representing a bound
 797 *					but unconnected socket that is to be
 798 *					used as an NFSD listener; listen(3)
 799 *					must be called for a SOCK_STREAM
 800 *					socket, otherwise it is ignored
 801 *			size:		non-zero length of C string in @buf
 802 * Output:
 803 *	On success:	NFS service is started;
 804 *			passed-in buffer filled with a '\n'-terminated C
 805 *			string containing a unique alphanumeric name of
 806 *			the listener;
 807 *			return code is the size in bytes of the string
 808 *	On error:	return code is a negative errno value
 809 *
 810 * OR
 811 *
 812 * Input:
 813 *			buf:		C string containing a transport
 814 *					name and an unsigned integer value
 815 *					representing the port to listen on,
 816 *					separated by whitespace
 817 *			size:		non-zero length of C string in @buf
 818 * Output:
 819 *	On success:	returns zero; NFS service is started
 820 *	On error:	return code is a negative errno value
 821 */
 822static ssize_t write_ports(struct file *file, char *buf, size_t size)
 823{
 824	ssize_t rv;
 825
 826	mutex_lock(&nfsd_mutex);
 827	rv = __write_ports(file, buf, size, netns(file));
 828	mutex_unlock(&nfsd_mutex);
 829	return rv;
 830}
 831
 832
 833int nfsd_max_blksize;
 834
 835/*
 836 * write_maxblksize - Set or report the current NFS blksize
 837 *
 838 * Input:
 839 *			buf:		ignored
 840 *			size:		zero
 841 *
 842 * OR
 843 *
 844 * Input:
 845 *			buf:		C string containing an unsigned
 846 *					integer value representing the new
 847 *					NFS blksize
 848 *			size:		non-zero length of C string in @buf
 849 * Output:
 850 *	On success:	passed-in buffer filled with '\n'-terminated C string
 851 *			containing numeric value of the current NFS blksize
 852 *			setting;
 853 *			return code is the size in bytes of the string
 854 *	On error:	return code is zero or a negative errno value
 855 */
 856static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
 857{
 858	char *mesg = buf;
 859	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 860
 861	if (size > 0) {
 862		int bsize;
 863		int rv = get_int(&mesg, &bsize);
 864		if (rv)
 865			return rv;
 866		trace_nfsd_ctl_maxblksize(netns(file), bsize);
 867
 868		/* force bsize into allowed range and
 869		 * required alignment.
 870		 */
 871		bsize = max_t(int, bsize, 1024);
 872		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
 873		bsize &= ~(1024-1);
 874		mutex_lock(&nfsd_mutex);
 875		if (nn->nfsd_serv) {
 876			mutex_unlock(&nfsd_mutex);
 877			return -EBUSY;
 878		}
 879		nfsd_max_blksize = bsize;
 880		mutex_unlock(&nfsd_mutex);
 881	}
 882
 883	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
 884							nfsd_max_blksize);
 885}
 886
 887/*
 888 * write_maxconn - Set or report the current max number of connections
 889 *
 890 * Input:
 891 *			buf:		ignored
 892 *			size:		zero
 893 * OR
 894 *
 895 * Input:
 896 *			buf:		C string containing an unsigned
 897 *					integer value representing the new
 898 *					number of max connections
 899 *			size:		non-zero length of C string in @buf
 900 * Output:
 901 *	On success:	passed-in buffer filled with '\n'-terminated C string
 902 *			containing numeric value of max_connections setting
 903 *			for this net namespace;
 904 *			return code is the size in bytes of the string
 905 *	On error:	return code is zero or a negative errno value
 906 */
 907static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
 908{
 909	char *mesg = buf;
 910	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 911	unsigned int maxconn = nn->max_connections;
 912
 913	if (size > 0) {
 914		int rv = get_uint(&mesg, &maxconn);
 915
 916		if (rv)
 917			return rv;
 918		trace_nfsd_ctl_maxconn(netns(file), maxconn);
 919		nn->max_connections = maxconn;
 920	}
 921
 922	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
 923}
 924
 925#ifdef CONFIG_NFSD_V4
 926static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
 927				  time64_t *time, struct nfsd_net *nn)
 928{
 929	struct dentry *dentry = file_dentry(file);
 930	char *mesg = buf;
 931	int rv, i;
 932
 933	if (size > 0) {
 934		if (nn->nfsd_serv)
 935			return -EBUSY;
 936		rv = get_int(&mesg, &i);
 937		if (rv)
 938			return rv;
 939		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
 940				    dentry->d_name.len, i);
 941
 942		/*
 943		 * Some sanity checking.  We don't have a reason for
 944		 * these particular numbers, but problems with the
 945		 * extremes are:
 946		 *	- Too short: the briefest network outage may
 947		 *	  cause clients to lose all their locks.  Also,
 948		 *	  the frequent polling may be wasteful.
 949		 *	- Too long: do you really want reboot recovery
 950		 *	  to take more than an hour?  Or to make other
 951		 *	  clients wait an hour before being able to
 952		 *	  revoke a dead client's locks?
 953		 */
 954		if (i < 10 || i > 3600)
 955			return -EINVAL;
 956		*time = i;
 957	}
 958
 959	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
 960}
 961
 962static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
 963				time64_t *time, struct nfsd_net *nn)
 964{
 965	ssize_t rv;
 966
 967	mutex_lock(&nfsd_mutex);
 968	rv = __nfsd4_write_time(file, buf, size, time, nn);
 969	mutex_unlock(&nfsd_mutex);
 970	return rv;
 971}
 972
 973/*
 974 * write_leasetime - Set or report the current NFSv4 lease time
 975 *
 976 * Input:
 977 *			buf:		ignored
 978 *			size:		zero
 979 *
 980 * OR
 981 *
 982 * Input:
 983 *			buf:		C string containing an unsigned
 984 *					integer value representing the new
 985 *					NFSv4 lease expiry time
 986 *			size:		non-zero length of C string in @buf
 987 * Output:
 988 *	On success:	passed-in buffer filled with '\n'-terminated C
 989 *			string containing unsigned integer value of the
 990 *			current lease expiry time;
 991 *			return code is the size in bytes of the string
 992 *	On error:	return code is zero or a negative errno value
 993 */
 994static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
 995{
 996	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 997	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
 998}
 999
1000/*
1001 * write_gracetime - Set or report current NFSv4 grace period time
1002 *
1003 * As above, but sets the time of the NFSv4 grace period.
1004 *
1005 * Note this should never be set to less than the *previous*
1006 * lease-period time, but we don't try to enforce this.  (In the common
1007 * case (a new boot), we don't know what the previous lease time was
1008 * anyway.)
1009 */
1010static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1011{
1012	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1013	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1014}
1015
1016#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1017static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1018				   struct nfsd_net *nn)
1019{
1020	char *mesg = buf;
1021	char *recdir;
1022	int len, status;
1023
1024	if (size > 0) {
1025		if (nn->nfsd_serv)
1026			return -EBUSY;
1027		if (size > PATH_MAX || buf[size-1] != '\n')
1028			return -EINVAL;
1029		buf[size-1] = 0;
1030
1031		recdir = mesg;
1032		len = qword_get(&mesg, recdir, size);
1033		if (len <= 0)
1034			return -EINVAL;
1035		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1036
1037		status = nfs4_reset_recoverydir(recdir);
1038		if (status)
1039			return status;
1040	}
1041
1042	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1043							nfs4_recoverydir());
1044}
1045
1046/*
1047 * write_recoverydir - Set or report the pathname of the recovery directory
1048 *
1049 * Input:
1050 *			buf:		ignored
1051 *			size:		zero
1052 *
1053 * OR
1054 *
1055 * Input:
1056 *			buf:		C string containing the pathname
1057 *					of the directory on a local file
1058 *					system containing permanent NFSv4
1059 *					recovery data
1060 *			size:		non-zero length of C string in @buf
1061 * Output:
1062 *	On success:	passed-in buffer filled with '\n'-terminated C string
1063 *			containing the current recovery pathname setting;
1064 *			return code is the size in bytes of the string
1065 *	On error:	return code is zero or a negative errno value
1066 */
1067static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1068{
1069	ssize_t rv;
1070	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1071
1072	mutex_lock(&nfsd_mutex);
1073	rv = __write_recoverydir(file, buf, size, nn);
1074	mutex_unlock(&nfsd_mutex);
1075	return rv;
1076}
1077#endif
1078
1079/*
1080 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1081 *
1082 * Input:
1083 *			buf:		ignored
1084 *			size:		zero
1085 * OR
1086 *
1087 * Input:
1088 *			buf:		any value
1089 *			size:		non-zero length of C string in @buf
1090 * Output:
1091 *			passed-in buffer filled with "Y" or "N" with a newline
1092 *			and NULL-terminated C string. This indicates whether
1093 *			the grace period has ended in the current net
1094 *			namespace. Return code is the size in bytes of the
1095 *			string. Writing a string that starts with 'Y', 'y', or
1096 *			'1' to the file will end the grace period for nfsd's v4
1097 *			lock manager.
1098 */
1099static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1100{
1101	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1102
1103	if (size > 0) {
1104		switch(buf[0]) {
1105		case 'Y':
1106		case 'y':
1107		case '1':
1108			if (!nn->nfsd_serv)
1109				return -EBUSY;
1110			trace_nfsd_end_grace(netns(file));
1111			nfsd4_end_grace(nn);
1112			break;
1113		default:
1114			return -EINVAL;
1115		}
1116	}
1117
1118	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1119			 nn->grace_ended ? 'Y' : 'N');
1120}
1121
1122#endif
1123
1124/*----------------------------------------------------------------------------*/
1125/*
1126 *	populating the filesystem.
1127 */
1128
1129/* Basically copying rpc_get_inode. */
1130static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1131{
1132	struct inode *inode = new_inode(sb);
1133	if (!inode)
1134		return NULL;
1135	/* Following advice from simple_fill_super documentation: */
1136	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1137	inode->i_mode = mode;
1138	simple_inode_init_ts(inode);
1139	switch (mode & S_IFMT) {
1140	case S_IFDIR:
1141		inode->i_fop = &simple_dir_operations;
1142		inode->i_op = &simple_dir_inode_operations;
1143		inc_nlink(inode);
1144		break;
1145	case S_IFLNK:
1146		inode->i_op = &simple_symlink_inode_operations;
1147		break;
1148	default:
1149		break;
1150	}
1151	return inode;
1152}
1153
1154static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1155{
1156	struct inode *inode;
1157
1158	inode = nfsd_get_inode(dir->i_sb, mode);
1159	if (!inode)
1160		return -ENOMEM;
1161	if (ncl) {
1162		inode->i_private = ncl;
1163		kref_get(&ncl->cl_ref);
1164	}
1165	d_add(dentry, inode);
1166	inc_nlink(dir);
1167	fsnotify_mkdir(dir, dentry);
1168	return 0;
1169}
1170
1171static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1172{
1173	struct inode *dir = parent->d_inode;
1174	struct dentry *dentry;
1175	int ret = -ENOMEM;
1176
1177	inode_lock(dir);
1178	dentry = d_alloc_name(parent, name);
1179	if (!dentry)
1180		goto out_err;
1181	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1182	if (ret)
1183		goto out_err;
1184out:
1185	inode_unlock(dir);
1186	return dentry;
1187out_err:
1188	dput(dentry);
1189	dentry = ERR_PTR(ret);
1190	goto out;
1191}
1192
1193#if IS_ENABLED(CONFIG_SUNRPC_GSS)
1194static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1195			  umode_t mode, const char *content)
1196{
1197	struct inode *inode;
1198
1199	inode = nfsd_get_inode(dir->i_sb, mode);
1200	if (!inode)
1201		return -ENOMEM;
1202
1203	inode->i_link = (char *)content;
1204	inode->i_size = strlen(content);
1205
1206	d_add(dentry, inode);
1207	inc_nlink(dir);
1208	fsnotify_create(dir, dentry);
1209	return 0;
1210}
1211
1212/*
1213 * @content is assumed to be a NUL-terminated string that lives
1214 * longer than the symlink itself.
1215 */
1216static void _nfsd_symlink(struct dentry *parent, const char *name,
1217			  const char *content)
1218{
1219	struct inode *dir = parent->d_inode;
1220	struct dentry *dentry;
1221	int ret;
1222
1223	inode_lock(dir);
1224	dentry = d_alloc_name(parent, name);
1225	if (!dentry)
1226		goto out;
1227	ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1228	if (ret)
1229		dput(dentry);
1230out:
1231	inode_unlock(dir);
1232}
1233#else
1234static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1235				 const char *content)
1236{
1237}
1238
1239#endif
1240
1241static void clear_ncl(struct dentry *dentry)
1242{
1243	struct inode *inode = d_inode(dentry);
1244	struct nfsdfs_client *ncl = inode->i_private;
1245
1246	spin_lock(&inode->i_lock);
1247	inode->i_private = NULL;
1248	spin_unlock(&inode->i_lock);
1249	kref_put(&ncl->cl_ref, ncl->cl_release);
1250}
1251
1252struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1253{
1254	struct nfsdfs_client *nc;
1255
1256	spin_lock(&inode->i_lock);
1257	nc = inode->i_private;
1258	if (nc)
1259		kref_get(&nc->cl_ref);
1260	spin_unlock(&inode->i_lock);
1261	return nc;
1262}
1263
1264/* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1265 * code instead. */
1266static  int nfsdfs_create_files(struct dentry *root,
1267				const struct tree_descr *files,
1268				struct nfsdfs_client *ncl,
1269				struct dentry **fdentries)
1270{
1271	struct inode *dir = d_inode(root);
1272	struct inode *inode;
1273	struct dentry *dentry;
1274	int i;
1275
1276	inode_lock(dir);
1277	for (i = 0; files->name && files->name[0]; i++, files++) {
1278		dentry = d_alloc_name(root, files->name);
1279		if (!dentry)
1280			goto out;
1281		inode = nfsd_get_inode(d_inode(root)->i_sb,
1282					S_IFREG | files->mode);
1283		if (!inode) {
1284			dput(dentry);
1285			goto out;
1286		}
1287		kref_get(&ncl->cl_ref);
1288		inode->i_fop = files->ops;
1289		inode->i_private = ncl;
1290		d_add(dentry, inode);
1291		fsnotify_create(dir, dentry);
1292		if (fdentries)
1293			fdentries[i] = dentry;
1294	}
1295	inode_unlock(dir);
1296	return 0;
1297out:
1298	inode_unlock(dir);
1299	return -ENOMEM;
1300}
1301
1302/* on success, returns positive number unique to that client. */
1303struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1304				 struct nfsdfs_client *ncl, u32 id,
1305				 const struct tree_descr *files,
1306				 struct dentry **fdentries)
1307{
1308	struct dentry *dentry;
1309	char name[11];
1310	int ret;
1311
1312	sprintf(name, "%u", id);
1313
1314	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1315	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1316		return NULL;
1317	ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1318	if (ret) {
1319		nfsd_client_rmdir(dentry);
1320		return NULL;
1321	}
1322	return dentry;
1323}
1324
1325/* Taken from __rpc_rmdir: */
1326void nfsd_client_rmdir(struct dentry *dentry)
1327{
1328	simple_recursive_removal(dentry, clear_ncl);
1329}
1330
1331static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1332{
1333	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1334							nfsd_net_id);
1335	struct dentry *dentry;
1336	int ret;
1337
1338	static const struct tree_descr nfsd_files[] = {
1339		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1340		/* Per-export io stats use same ops as exports file */
1341		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1342		[NFSD_Export_features] = {"export_features",
1343					&export_features_fops, S_IRUGO},
1344		[NFSD_FO_UnlockIP] = {"unlock_ip",
1345					&transaction_ops, S_IWUSR|S_IRUSR},
1346		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1347					&transaction_ops, S_IWUSR|S_IRUSR},
1348		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1349		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1350		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1351		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1352		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1353					&nfsd_reply_cache_stats_fops, S_IRUGO},
1354		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1355		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1356		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1357		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1358		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
 
 
1359#ifdef CONFIG_NFSD_V4
1360		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1361		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1362		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1363		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1364#endif
1365		/* last one */ {""}
1366	};
1367
1368	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1369	if (ret)
1370		return ret;
1371	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1372		      "/proc/net/rpc/gss_krb5_enctypes");
1373	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1374	if (IS_ERR(dentry))
1375		return PTR_ERR(dentry);
1376	nn->nfsd_client_dir = dentry;
1377	return 0;
1378}
1379
1380static int nfsd_fs_get_tree(struct fs_context *fc)
1381{
1382	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1383}
1384
1385static void nfsd_fs_free_fc(struct fs_context *fc)
1386{
1387	if (fc->s_fs_info)
1388		put_net(fc->s_fs_info);
1389}
1390
1391static const struct fs_context_operations nfsd_fs_context_ops = {
1392	.free		= nfsd_fs_free_fc,
1393	.get_tree	= nfsd_fs_get_tree,
1394};
1395
1396static int nfsd_init_fs_context(struct fs_context *fc)
1397{
1398	put_user_ns(fc->user_ns);
1399	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1400	fc->ops = &nfsd_fs_context_ops;
1401	return 0;
1402}
1403
1404static void nfsd_umount(struct super_block *sb)
1405{
1406	struct net *net = sb->s_fs_info;
1407
1408	nfsd_shutdown_threads(net);
1409
1410	kill_litter_super(sb);
1411	put_net(net);
1412}
1413
1414static struct file_system_type nfsd_fs_type = {
1415	.owner		= THIS_MODULE,
1416	.name		= "nfsd",
1417	.init_fs_context = nfsd_init_fs_context,
1418	.kill_sb	= nfsd_umount,
1419};
1420MODULE_ALIAS_FS("nfsd");
1421
1422#ifdef CONFIG_PROC_FS
1423
1424static int exports_proc_open(struct inode *inode, struct file *file)
1425{
1426	return exports_net_open(current->nsproxy->net_ns, file);
1427}
1428
1429static const struct proc_ops exports_proc_ops = {
1430	.proc_open	= exports_proc_open,
1431	.proc_read	= seq_read,
1432	.proc_lseek	= seq_lseek,
1433	.proc_release	= seq_release,
1434};
1435
1436static int create_proc_exports_entry(void)
1437{
1438	struct proc_dir_entry *entry;
1439
1440	entry = proc_mkdir("fs/nfs", NULL);
1441	if (!entry)
1442		return -ENOMEM;
1443	entry = proc_create("exports", 0, entry, &exports_proc_ops);
 
1444	if (!entry) {
1445		remove_proc_entry("fs/nfs", NULL);
1446		return -ENOMEM;
1447	}
1448	return 0;
1449}
1450#else /* CONFIG_PROC_FS */
1451static int create_proc_exports_entry(void)
1452{
1453	return 0;
1454}
1455#endif
1456
1457unsigned int nfsd_net_id;
1458
1459/**
1460 * nfsd_nl_rpc_status_get_start - Prepare rpc_status_get dumpit
1461 * @cb: netlink metadata and command arguments
1462 *
1463 * Return values:
1464 *   %0: The rpc_status_get command may proceed
1465 *   %-ENODEV: There is no NFSD running in this namespace
1466 */
1467int nfsd_nl_rpc_status_get_start(struct netlink_callback *cb)
1468{
1469	struct nfsd_net *nn = net_generic(sock_net(cb->skb->sk), nfsd_net_id);
1470	int ret = -ENODEV;
1471
1472	mutex_lock(&nfsd_mutex);
1473	if (nn->nfsd_serv)
1474		ret = 0;
1475	else
1476		mutex_unlock(&nfsd_mutex);
1477
1478	return ret;
1479}
1480
1481static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1482					    struct netlink_callback *cb,
1483					    struct nfsd_genl_rqstp *rqstp)
1484{
1485	void *hdr;
1486	u32 i;
1487
1488	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1489			  &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1490	if (!hdr)
1491		return -ENOBUFS;
1492
1493	if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1494	    nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1495	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1496	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1497	    nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1498	    nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1499			ktime_to_us(rqstp->rq_stime),
1500			NFSD_A_RPC_STATUS_PAD))
1501		return -ENOBUFS;
1502
1503	switch (rqstp->rq_saddr.sa_family) {
1504	case AF_INET: {
1505		const struct sockaddr_in *s_in, *d_in;
1506
1507		s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1508		d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1509		if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1510				    s_in->sin_addr.s_addr) ||
1511		    nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1512				    d_in->sin_addr.s_addr) ||
1513		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1514				 s_in->sin_port) ||
1515		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1516				 d_in->sin_port))
1517			return -ENOBUFS;
1518		break;
1519	}
1520	case AF_INET6: {
1521		const struct sockaddr_in6 *s_in, *d_in;
1522
1523		s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1524		d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1525		if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1526				     &s_in->sin6_addr) ||
1527		    nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1528				     &d_in->sin6_addr) ||
1529		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1530				 s_in->sin6_port) ||
1531		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1532				 d_in->sin6_port))
1533			return -ENOBUFS;
1534		break;
1535	}
1536	}
1537
1538	for (i = 0; i < rqstp->rq_opcnt; i++)
1539		if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1540				rqstp->rq_opnum[i]))
1541			return -ENOBUFS;
1542
1543	genlmsg_end(skb, hdr);
1544	return 0;
1545}
1546
1547/**
1548 * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1549 * @skb: reply buffer
1550 * @cb: netlink metadata and command arguments
1551 *
1552 * Returns the size of the reply or a negative errno.
1553 */
1554int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1555				  struct netlink_callback *cb)
1556{
1557	struct nfsd_net *nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1558	int i, ret, rqstp_index = 0;
1559
1560	rcu_read_lock();
1561
1562	for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1563		struct svc_rqst *rqstp;
1564
1565		if (i < cb->args[0]) /* already consumed */
1566			continue;
1567
1568		rqstp_index = 0;
1569		list_for_each_entry_rcu(rqstp,
1570				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
1571				rq_all) {
1572			struct nfsd_genl_rqstp genl_rqstp;
1573			unsigned int status_counter;
1574
1575			if (rqstp_index++ < cb->args[1]) /* already consumed */
1576				continue;
1577			/*
1578			 * Acquire rq_status_counter before parsing the rqst
1579			 * fields. rq_status_counter is set to an odd value in
1580			 * order to notify the consumers the rqstp fields are
1581			 * meaningful.
1582			 */
1583			status_counter =
1584				smp_load_acquire(&rqstp->rq_status_counter);
1585			if (!(status_counter & 1))
1586				continue;
1587
1588			genl_rqstp.rq_xid = rqstp->rq_xid;
1589			genl_rqstp.rq_flags = rqstp->rq_flags;
1590			genl_rqstp.rq_vers = rqstp->rq_vers;
1591			genl_rqstp.rq_prog = rqstp->rq_prog;
1592			genl_rqstp.rq_proc = rqstp->rq_proc;
1593			genl_rqstp.rq_stime = rqstp->rq_stime;
1594			genl_rqstp.rq_opcnt = 0;
1595			memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1596			       sizeof(struct sockaddr));
1597			memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1598			       sizeof(struct sockaddr));
1599
1600#ifdef CONFIG_NFSD_V4
1601			if (rqstp->rq_vers == NFS4_VERSION &&
1602			    rqstp->rq_proc == NFSPROC4_COMPOUND) {
1603				/* NFSv4 compound */
1604				struct nfsd4_compoundargs *args;
1605				int j;
1606
1607				args = rqstp->rq_argp;
1608				genl_rqstp.rq_opcnt = args->opcnt;
1609				for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1610					genl_rqstp.rq_opnum[j] =
1611						args->ops[j].opnum;
1612			}
1613#endif /* CONFIG_NFSD_V4 */
1614
1615			/*
1616			 * Acquire rq_status_counter before reporting the rqst
1617			 * fields to the user.
1618			 */
1619			if (smp_load_acquire(&rqstp->rq_status_counter) !=
1620			    status_counter)
1621				continue;
1622
1623			ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1624							       &genl_rqstp);
1625			if (ret)
1626				goto out;
1627		}
1628	}
1629
1630	cb->args[0] = i;
1631	cb->args[1] = rqstp_index;
1632	ret = skb->len;
1633out:
1634	rcu_read_unlock();
1635
1636	return ret;
1637}
1638
1639/**
1640 * nfsd_nl_rpc_status_get_done - rpc_status_get dumpit post-processing
1641 * @cb: netlink metadata and command arguments
1642 *
1643 * Return values:
1644 *   %0: Success
1645 */
1646int nfsd_nl_rpc_status_get_done(struct netlink_callback *cb)
1647{
1648	mutex_unlock(&nfsd_mutex);
1649
1650	return 0;
1651}
1652
1653/**
1654 * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
1655 * @net: a freshly-created network namespace
1656 *
1657 * This information stays around as long as the network namespace is
1658 * alive whether or not there is an NFSD instance running in the
1659 * namespace.
1660 *
1661 * Returns zero on success, or a negative errno otherwise.
1662 */
1663static __net_init int nfsd_net_init(struct net *net)
1664{
1665	int retval;
1666	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1667
1668	retval = nfsd_export_init(net);
1669	if (retval)
1670		goto out_export_error;
1671	retval = nfsd_idmap_init(net);
1672	if (retval)
1673		goto out_idmap_error;
1674	retval = nfsd_net_reply_cache_init(nn);
1675	if (retval)
1676		goto out_repcache_error;
1677	nn->nfsd_versions = NULL;
1678	nn->nfsd4_minorversions = NULL;
1679	nfsd4_init_leases_net(nn);
1680	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1681	seqlock_init(&nn->writeverf_lock);
1682
1683	return 0;
1684
1685out_repcache_error:
1686	nfsd_idmap_shutdown(net);
1687out_idmap_error:
1688	nfsd_export_shutdown(net);
1689out_export_error:
1690	return retval;
1691}
1692
1693/**
1694 * nfsd_net_exit - Release the nfsd_net portion of a net namespace
1695 * @net: a network namespace that is about to be destroyed
1696 *
1697 */
1698static __net_exit void nfsd_net_exit(struct net *net)
1699{
1700	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1701
1702	nfsd_net_reply_cache_destroy(nn);
1703	nfsd_idmap_shutdown(net);
1704	nfsd_export_shutdown(net);
1705	nfsd_netns_free_versions(nn);
1706}
1707
1708static struct pernet_operations nfsd_net_ops = {
1709	.init = nfsd_net_init,
1710	.exit = nfsd_net_exit,
1711	.id   = &nfsd_net_id,
1712	.size = sizeof(struct nfsd_net),
1713};
1714
1715static int __init init_nfsd(void)
1716{
1717	int retval;
 
1718
 
 
 
 
 
 
1719	retval = nfsd4_init_slabs();
1720	if (retval)
1721		return retval;
1722	retval = nfsd4_init_pnfs();
1723	if (retval)
1724		goto out_free_slabs;
1725	retval = nfsd_stat_init();	/* Statistics */
1726	if (retval)
1727		goto out_free_pnfs;
1728	retval = nfsd_drc_slab_create();
 
1729	if (retval)
1730		goto out_free_stat;
1731	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1732	retval = create_proc_exports_entry();
1733	if (retval)
1734		goto out_free_lockd;
1735	retval = register_pernet_subsys(&nfsd_net_ops);
1736	if (retval < 0)
1737		goto out_free_exports;
1738	retval = register_cld_notifier();
1739	if (retval)
1740		goto out_free_subsys;
1741	retval = nfsd4_create_laundry_wq();
1742	if (retval)
1743		goto out_free_cld;
1744	retval = register_filesystem(&nfsd_fs_type);
1745	if (retval)
1746		goto out_free_all;
1747	retval = genl_register_family(&nfsd_nl_family);
1748	if (retval)
1749		goto out_free_all;
1750
1751	return 0;
1752out_free_all:
1753	nfsd4_destroy_laundry_wq();
1754out_free_cld:
1755	unregister_cld_notifier();
1756out_free_subsys:
1757	unregister_pernet_subsys(&nfsd_net_ops);
1758out_free_exports:
1759	remove_proc_entry("fs/nfs/exports", NULL);
1760	remove_proc_entry("fs/nfs", NULL);
1761out_free_lockd:
1762	nfsd_lockd_shutdown();
1763	nfsd_drc_slab_free();
1764out_free_stat:
1765	nfsd_stat_shutdown();
1766out_free_pnfs:
 
1767	nfsd4_exit_pnfs();
1768out_free_slabs:
1769	nfsd4_free_slabs();
 
 
 
 
1770	return retval;
1771}
1772
1773static void __exit exit_nfsd(void)
1774{
1775	genl_unregister_family(&nfsd_nl_family);
1776	unregister_filesystem(&nfsd_fs_type);
1777	nfsd4_destroy_laundry_wq();
1778	unregister_cld_notifier();
1779	unregister_pernet_subsys(&nfsd_net_ops);
1780	nfsd_drc_slab_free();
1781	remove_proc_entry("fs/nfs/exports", NULL);
1782	remove_proc_entry("fs/nfs", NULL);
1783	nfsd_stat_shutdown();
1784	nfsd_lockd_shutdown();
1785	nfsd4_free_slabs();
1786	nfsd4_exit_pnfs();
 
 
 
 
1787}
1788
1789MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1790MODULE_DESCRIPTION("In-kernel NFS server");
1791MODULE_LICENSE("GPL");
1792module_init(init_nfsd)
1793module_exit(exit_nfsd)