Linux Audio

Check our new training course

Loading...
v3.1
 
   1/*
   2 *  linux/fs/proc/base.c
   3 *
   4 *  Copyright (C) 1991, 1992 Linus Torvalds
   5 *
   6 *  proc base directory handling functions
   7 *
   8 *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
   9 *  Instead of using magical inumbers to determine the kind of object
  10 *  we allocate and fill in-core inodes upon lookup. They don't even
  11 *  go into icache. We cache the reference to task_struct upon lookup too.
  12 *  Eventually it should become a filesystem in its own. We don't use the
  13 *  rest of procfs anymore.
  14 *
  15 *
  16 *  Changelog:
  17 *  17-Jan-2005
  18 *  Allan Bezerra
  19 *  Bruna Moreira <bruna.moreira@indt.org.br>
  20 *  Edjard Mota <edjard.mota@indt.org.br>
  21 *  Ilias Biris <ilias.biris@indt.org.br>
  22 *  Mauricio Lin <mauricio.lin@indt.org.br>
  23 *
  24 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  25 *
  26 *  A new process specific entry (smaps) included in /proc. It shows the
  27 *  size of rss for each memory area. The maps entry lacks information
  28 *  about physical memory size (rss) for each mapped file, i.e.,
  29 *  rss information for executables and library files.
  30 *  This additional information is useful for any tools that need to know
  31 *  about physical memory consumption for a process specific library.
  32 *
  33 *  Changelog:
  34 *  21-Feb-2005
  35 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  36 *  Pud inclusion in the page table walking.
  37 *
  38 *  ChangeLog:
  39 *  10-Mar-2005
  40 *  10LE Instituto Nokia de Tecnologia - INdT:
  41 *  A better way to walks through the page table as suggested by Hugh Dickins.
  42 *
  43 *  Simo Piiroinen <simo.piiroinen@nokia.com>:
  44 *  Smaps information related to shared, private, clean and dirty pages.
  45 *
  46 *  Paul Mundt <paul.mundt@nokia.com>:
  47 *  Overall revision about smaps.
  48 */
  49
  50#include <asm/uaccess.h>
  51
  52#include <linux/errno.h>
  53#include <linux/time.h>
  54#include <linux/proc_fs.h>
  55#include <linux/stat.h>
  56#include <linux/task_io_accounting_ops.h>
  57#include <linux/init.h>
  58#include <linux/capability.h>
  59#include <linux/file.h>
  60#include <linux/fdtable.h>
  61#include <linux/string.h>
  62#include <linux/seq_file.h>
  63#include <linux/namei.h>
  64#include <linux/mnt_namespace.h>
  65#include <linux/mm.h>
  66#include <linux/swap.h>
  67#include <linux/rcupdate.h>
  68#include <linux/kallsyms.h>
  69#include <linux/stacktrace.h>
  70#include <linux/resource.h>
  71#include <linux/module.h>
  72#include <linux/mount.h>
  73#include <linux/security.h>
  74#include <linux/ptrace.h>
  75#include <linux/tracehook.h>
 
  76#include <linux/cgroup.h>
  77#include <linux/cpuset.h>
  78#include <linux/audit.h>
  79#include <linux/poll.h>
  80#include <linux/nsproxy.h>
  81#include <linux/oom.h>
  82#include <linux/elf.h>
  83#include <linux/pid_namespace.h>
 
 
  84#include <linux/fs_struct.h>
  85#include <linux/slab.h>
  86#ifdef CONFIG_HARDWALL
  87#include <asm/hardwall.h>
  88#endif
 
 
 
 
 
 
 
 
 
  89#include "internal.h"
 
 
 
  90
  91/* NOTE:
  92 *	Implementing inode permission operations in /proc is almost
  93 *	certainly an error.  Permission checks need to happen during
  94 *	each system call not at open time.  The reason is that most of
  95 *	what we wish to check for permissions in /proc varies at runtime.
  96 *
  97 *	The classic example of a problem is opening file descriptors
  98 *	in /proc for a task before it execs a suid executable.
  99 */
 100
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 101struct pid_entry {
 102	char *name;
 103	int len;
 104	mode_t mode;
 105	const struct inode_operations *iop;
 106	const struct file_operations *fop;
 107	union proc_op op;
 108};
 109
 110#define NOD(NAME, MODE, IOP, FOP, OP) {			\
 111	.name = (NAME),					\
 112	.len  = sizeof(NAME) - 1,			\
 113	.mode = MODE,					\
 114	.iop  = IOP,					\
 115	.fop  = FOP,					\
 116	.op   = OP,					\
 117}
 118
 119#define DIR(NAME, MODE, iops, fops)	\
 120	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
 121#define LNK(NAME, get_link)					\
 122	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
 123		&proc_pid_link_inode_operations, NULL,		\
 124		{ .proc_get_link = get_link } )
 125#define REG(NAME, MODE, fops)				\
 126	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
 127#define INF(NAME, MODE, read)				\
 128	NOD(NAME, (S_IFREG|(MODE)), 			\
 129		NULL, &proc_info_file_operations,	\
 130		{ .proc_read = read } )
 131#define ONE(NAME, MODE, show)				\
 132	NOD(NAME, (S_IFREG|(MODE)), 			\
 133		NULL, &proc_single_file_operations,	\
 134		{ .proc_show = show } )
 
 
 
 
 135
 136/*
 137 * Count the number of hardlinks for the pid_entry table, excluding the .
 138 * and .. links.
 139 */
 140static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
 141	unsigned int n)
 142{
 143	unsigned int i;
 144	unsigned int count;
 145
 146	count = 0;
 147	for (i = 0; i < n; ++i) {
 148		if (S_ISDIR(entries[i].mode))
 149			++count;
 150	}
 151
 152	return count;
 153}
 154
 155static int get_task_root(struct task_struct *task, struct path *root)
 156{
 157	int result = -ENOENT;
 158
 159	task_lock(task);
 160	if (task->fs) {
 161		get_fs_root(task->fs, root);
 162		result = 0;
 163	}
 164	task_unlock(task);
 165	return result;
 166}
 167
 168static int proc_cwd_link(struct inode *inode, struct path *path)
 169{
 170	struct task_struct *task = get_proc_task(inode);
 171	int result = -ENOENT;
 172
 173	if (task) {
 174		task_lock(task);
 175		if (task->fs) {
 176			get_fs_pwd(task->fs, path);
 177			result = 0;
 178		}
 179		task_unlock(task);
 180		put_task_struct(task);
 181	}
 182	return result;
 183}
 184
 185static int proc_root_link(struct inode *inode, struct path *path)
 186{
 187	struct task_struct *task = get_proc_task(inode);
 188	int result = -ENOENT;
 189
 190	if (task) {
 191		result = get_task_root(task, path);
 192		put_task_struct(task);
 193	}
 194	return result;
 195}
 196
 197static struct mm_struct *__check_mem_permission(struct task_struct *task)
 
 
 
 
 
 
 198{
 199	struct mm_struct *mm;
 
 200
 201	mm = get_task_mm(task);
 202	if (!mm)
 203		return ERR_PTR(-EINVAL);
 204
 205	/*
 206	 * A task can always look at itself, in case it chooses
 207	 * to use system calls instead of load instructions.
 208	 */
 209	if (task == current)
 210		return mm;
 211
 212	/*
 213	 * If current is actively ptrace'ing, and would also be
 214	 * permitted to freshly attach with ptrace now, permit it.
 215	 */
 216	if (task_is_stopped_or_traced(task)) {
 217		int match;
 218		rcu_read_lock();
 219		match = (ptrace_parent(task) == current);
 220		rcu_read_unlock();
 221		if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
 222			return mm;
 
 
 
 
 
 
 
 223	}
 224
 225	/*
 226	 * No one else is allowed.
 227	 */
 228	mmput(mm);
 229	return ERR_PTR(-EPERM);
 230}
 231
 232/*
 233 * If current may access user memory in @task return a reference to the
 234 * corresponding mm, otherwise ERR_PTR.
 235 */
 236static struct mm_struct *check_mem_permission(struct task_struct *task)
 237{
 238	struct mm_struct *mm;
 239	int err;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 240
 241	/*
 242	 * Avoid racing if task exec's as we might get a new mm but validate
 243	 * against old credentials.
 
 244	 */
 245	err = mutex_lock_killable(&task->signal->cred_guard_mutex);
 246	if (err)
 247		return ERR_PTR(err);
 
 
 
 
 
 
 
 
 
 248
 249	mm = __check_mem_permission(task);
 250	mutex_unlock(&task->signal->cred_guard_mutex);
 
 
 
 
 
 
 
 251
 252	return mm;
 253}
 
 
 
 
 
 
 
 254
 255struct mm_struct *mm_for_maps(struct task_struct *task)
 256{
 257	struct mm_struct *mm;
 258	int err;
 259
 260	err =  mutex_lock_killable(&task->signal->cred_guard_mutex);
 261	if (err)
 262		return ERR_PTR(err);
 
 263
 264	mm = get_task_mm(task);
 265	if (mm && mm != current->mm &&
 266			!ptrace_may_access(task, PTRACE_MODE_READ)) {
 267		mmput(mm);
 268		mm = ERR_PTR(-EACCES);
 
 
 
 
 
 
 
 
 269	}
 270	mutex_unlock(&task->signal->cred_guard_mutex);
 271
 272	return mm;
 
 273}
 274
 275static int proc_pid_cmdline(struct task_struct *task, char * buffer)
 
 276{
 277	int res = 0;
 278	unsigned int len;
 279	struct mm_struct *mm = get_task_mm(task);
 
 280	if (!mm)
 281		goto out;
 282	if (!mm->arg_end)
 283		goto out_mm;	/* Shh! No looking before we're done */
 284
 285 	len = mm->arg_end - mm->arg_start;
 286 
 287	if (len > PAGE_SIZE)
 288		len = PAGE_SIZE;
 289 
 290	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
 291
 292	// If the nul at the end of args has been overwritten, then
 293	// assume application is using setproctitle(3).
 294	if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
 295		len = strnlen(buffer, res);
 296		if (len < res) {
 297		    res = len;
 298		} else {
 299			len = mm->env_end - mm->env_start;
 300			if (len > PAGE_SIZE - res)
 301				len = PAGE_SIZE - res;
 302			res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
 303			res = strnlen(buffer, res);
 304		}
 305	}
 306out_mm:
 307	mmput(mm);
 308out:
 309	return res;
 310}
 311
 312static int proc_pid_auxv(struct task_struct *task, char *buffer)
 
 313{
 314	struct mm_struct *mm = mm_for_maps(task);
 315	int res = PTR_ERR(mm);
 316	if (mm && !IS_ERR(mm)) {
 317		unsigned int nwords = 0;
 318		do {
 319			nwords += 2;
 320		} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
 321		res = nwords * sizeof(mm->saved_auxv[0]);
 322		if (res > PAGE_SIZE)
 323			res = PAGE_SIZE;
 324		memcpy(buffer, mm->saved_auxv, res);
 325		mmput(mm);
 326	}
 327	return res;
 328}
 329
 
 
 
 
 330
 331#ifdef CONFIG_KALLSYMS
 332/*
 333 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 334 * Returns the resolved symbol.  If that fails, simply return the address.
 335 */
 336static int proc_pid_wchan(struct task_struct *task, char *buffer)
 
 337{
 338	unsigned long wchan;
 339	char symname[KSYM_NAME_LEN];
 340
 
 
 
 341	wchan = get_wchan(task);
 
 
 
 
 342
 343	if (lookup_symbol_name(wchan, symname) < 0)
 344		if (!ptrace_may_access(task, PTRACE_MODE_READ))
 345			return 0;
 346		else
 347			return sprintf(buffer, "%lu", wchan);
 348	else
 349		return sprintf(buffer, "%s", symname);
 350}
 351#endif /* CONFIG_KALLSYMS */
 352
 353static int lock_trace(struct task_struct *task)
 354{
 355	int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
 356	if (err)
 357		return err;
 358	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
 359		mutex_unlock(&task->signal->cred_guard_mutex);
 360		return -EPERM;
 361	}
 362	return 0;
 363}
 364
 365static void unlock_trace(struct task_struct *task)
 366{
 367	mutex_unlock(&task->signal->cred_guard_mutex);
 368}
 369
 370#ifdef CONFIG_STACKTRACE
 371
 372#define MAX_STACK_TRACE_DEPTH	64
 373
 374static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
 375			  struct pid *pid, struct task_struct *task)
 376{
 377	struct stack_trace trace;
 378	unsigned long *entries;
 379	int err;
 380	int i;
 381
 382	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 383	if (!entries)
 384		return -ENOMEM;
 385
 386	trace.nr_entries	= 0;
 387	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
 388	trace.entries		= entries;
 389	trace.skip		= 0;
 390
 391	err = lock_trace(task);
 392	if (!err) {
 393		save_stack_trace_tsk(task, &trace);
 
 
 
 394
 395		for (i = 0; i < trace.nr_entries; i++) {
 396			seq_printf(m, "[<%pK>] %pS\n",
 397				   (void *)entries[i], (void *)entries[i]);
 398		}
 
 399		unlock_trace(task);
 400	}
 401	kfree(entries);
 402
 403	return err;
 404}
 405#endif
 406
 407#ifdef CONFIG_SCHEDSTATS
 408/*
 409 * Provides /proc/PID/schedstat
 410 */
 411static int proc_pid_schedstat(struct task_struct *task, char *buffer)
 
 412{
 413	return sprintf(buffer, "%llu %llu %lu\n",
 414			(unsigned long long)task->se.sum_exec_runtime,
 415			(unsigned long long)task->sched_info.run_delay,
 416			task->sched_info.pcount);
 
 
 
 
 
 417}
 418#endif
 419
 420#ifdef CONFIG_LATENCYTOP
 421static int lstats_show_proc(struct seq_file *m, void *v)
 422{
 423	int i;
 424	struct inode *inode = m->private;
 425	struct task_struct *task = get_proc_task(inode);
 426
 427	if (!task)
 428		return -ESRCH;
 429	seq_puts(m, "Latency Top version : v0.1\n");
 430	for (i = 0; i < 32; i++) {
 431		struct latency_record *lr = &task->latency_record[i];
 432		if (lr->backtrace[0]) {
 433			int q;
 434			seq_printf(m, "%i %li %li",
 435				   lr->count, lr->time, lr->max);
 436			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
 437				unsigned long bt = lr->backtrace[q];
 
 438				if (!bt)
 439					break;
 440				if (bt == ULONG_MAX)
 441					break;
 442				seq_printf(m, " %ps", (void *)bt);
 443			}
 444			seq_putc(m, '\n');
 445		}
 446
 447	}
 448	put_task_struct(task);
 449	return 0;
 450}
 451
 452static int lstats_open(struct inode *inode, struct file *file)
 453{
 454	return single_open(file, lstats_show_proc, inode);
 455}
 456
 457static ssize_t lstats_write(struct file *file, const char __user *buf,
 458			    size_t count, loff_t *offs)
 459{
 460	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
 461
 462	if (!task)
 463		return -ESRCH;
 464	clear_all_latency_tracing(task);
 465	put_task_struct(task);
 466
 467	return count;
 468}
 469
 470static const struct file_operations proc_lstats_operations = {
 471	.open		= lstats_open,
 472	.read		= seq_read,
 473	.write		= lstats_write,
 474	.llseek		= seq_lseek,
 475	.release	= single_release,
 476};
 477
 478#endif
 479
 480static int proc_oom_score(struct task_struct *task, char *buffer)
 
 481{
 
 482	unsigned long points = 0;
 
 
 
 
 
 
 
 
 
 
 
 
 483
 484	read_lock(&tasklist_lock);
 485	if (pid_alive(task))
 486		points = oom_badness(task, NULL, NULL,
 487					totalram_pages + total_swap_pages);
 488	read_unlock(&tasklist_lock);
 489	return sprintf(buffer, "%lu\n", points);
 490}
 491
 492struct limit_names {
 493	char *name;
 494	char *unit;
 495};
 496
 497static const struct limit_names lnames[RLIM_NLIMITS] = {
 498	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
 499	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
 500	[RLIMIT_DATA] = {"Max data size", "bytes"},
 501	[RLIMIT_STACK] = {"Max stack size", "bytes"},
 502	[RLIMIT_CORE] = {"Max core file size", "bytes"},
 503	[RLIMIT_RSS] = {"Max resident set", "bytes"},
 504	[RLIMIT_NPROC] = {"Max processes", "processes"},
 505	[RLIMIT_NOFILE] = {"Max open files", "files"},
 506	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
 507	[RLIMIT_AS] = {"Max address space", "bytes"},
 508	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
 509	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
 510	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
 511	[RLIMIT_NICE] = {"Max nice priority", NULL},
 512	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
 513	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
 514};
 515
 516/* Display limits for a process */
 517static int proc_pid_limits(struct task_struct *task, char *buffer)
 
 518{
 519	unsigned int i;
 520	int count = 0;
 521	unsigned long flags;
 522	char *bufptr = buffer;
 523
 524	struct rlimit rlim[RLIM_NLIMITS];
 525
 526	if (!lock_task_sighand(task, &flags))
 527		return 0;
 528	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
 529	unlock_task_sighand(task, &flags);
 530
 531	/*
 532	 * print the file header
 533	 */
 534	count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
 535			"Limit", "Soft Limit", "Hard Limit", "Units");
 
 
 536
 537	for (i = 0; i < RLIM_NLIMITS; i++) {
 538		if (rlim[i].rlim_cur == RLIM_INFINITY)
 539			count += sprintf(&bufptr[count], "%-25s %-20s ",
 540					 lnames[i].name, "unlimited");
 541		else
 542			count += sprintf(&bufptr[count], "%-25s %-20lu ",
 543					 lnames[i].name, rlim[i].rlim_cur);
 544
 545		if (rlim[i].rlim_max == RLIM_INFINITY)
 546			count += sprintf(&bufptr[count], "%-20s ", "unlimited");
 547		else
 548			count += sprintf(&bufptr[count], "%-20lu ",
 549					 rlim[i].rlim_max);
 550
 551		if (lnames[i].unit)
 552			count += sprintf(&bufptr[count], "%-10s\n",
 553					 lnames[i].unit);
 554		else
 555			count += sprintf(&bufptr[count], "\n");
 556	}
 557
 558	return count;
 559}
 560
 561#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 562static int proc_pid_syscall(struct task_struct *task, char *buffer)
 
 563{
 564	long nr;
 565	unsigned long args[6], sp, pc;
 566	int res = lock_trace(task);
 
 
 567	if (res)
 568		return res;
 569
 570	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
 571		res = sprintf(buffer, "running\n");
 572	else if (nr < 0)
 573		res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
 
 574	else
 575		res = sprintf(buffer,
 576		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
 577		       nr,
 578		       args[0], args[1], args[2], args[3], args[4], args[5],
 579		       sp, pc);
 580	unlock_trace(task);
 581	return res;
 
 582}
 583#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
 584
 585/************************************************************************/
 586/*                       Here the fs part begins                        */
 587/************************************************************************/
 588
 589/* permission checks */
 590static int proc_fd_access_allowed(struct inode *inode)
 591{
 592	struct task_struct *task;
 593	int allowed = 0;
 594	/* Allow access to a task's file descriptors if it is us or we
 595	 * may use ptrace attach to the process and find out that
 596	 * information.
 597	 */
 598	task = get_proc_task(inode);
 599	if (task) {
 600		allowed = ptrace_may_access(task, PTRACE_MODE_READ);
 601		put_task_struct(task);
 602	}
 603	return allowed;
 604}
 605
 606int proc_setattr(struct dentry *dentry, struct iattr *attr)
 
 607{
 608	int error;
 609	struct inode *inode = dentry->d_inode;
 610
 611	if (attr->ia_valid & ATTR_MODE)
 612		return -EPERM;
 613
 614	error = inode_change_ok(inode, attr);
 615	if (error)
 616		return error;
 617
 618	if ((attr->ia_valid & ATTR_SIZE) &&
 619	    attr->ia_size != i_size_read(inode)) {
 620		error = vmtruncate(inode, attr->ia_size);
 621		if (error)
 622			return error;
 623	}
 624
 625	setattr_copy(inode, attr);
 626	mark_inode_dirty(inode);
 627	return 0;
 628}
 629
 630static const struct inode_operations proc_def_inode_operations = {
 631	.setattr	= proc_setattr,
 632};
 633
 634static int mounts_open_common(struct inode *inode, struct file *file,
 635			      const struct seq_operations *op)
 636{
 637	struct task_struct *task = get_proc_task(inode);
 638	struct nsproxy *nsp;
 639	struct mnt_namespace *ns = NULL;
 640	struct path root;
 641	struct proc_mounts *p;
 642	int ret = -EINVAL;
 643
 644	if (task) {
 645		rcu_read_lock();
 646		nsp = task_nsproxy(task);
 647		if (nsp) {
 648			ns = nsp->mnt_ns;
 649			if (ns)
 650				get_mnt_ns(ns);
 651		}
 652		rcu_read_unlock();
 653		if (ns && get_task_root(task, &root) == 0)
 654			ret = 0;
 655		put_task_struct(task);
 656	}
 657
 658	if (!ns)
 659		goto err;
 660	if (ret)
 661		goto err_put_ns;
 662
 663	ret = -ENOMEM;
 664	p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
 665	if (!p)
 666		goto err_put_path;
 667
 668	file->private_data = &p->m;
 669	ret = seq_open(file, op);
 670	if (ret)
 671		goto err_free;
 672
 673	p->m.private = p;
 674	p->ns = ns;
 675	p->root = root;
 676	p->m.poll_event = ns->event;
 677
 678	return 0;
 679
 680 err_free:
 681	kfree(p);
 682 err_put_path:
 683	path_put(&root);
 684 err_put_ns:
 685	put_mnt_ns(ns);
 686 err:
 687	return ret;
 688}
 689
 690static int mounts_release(struct inode *inode, struct file *file)
 691{
 692	struct proc_mounts *p = file->private_data;
 693	path_put(&p->root);
 694	put_mnt_ns(p->ns);
 695	return seq_release(inode, file);
 696}
 697
 698static unsigned mounts_poll(struct file *file, poll_table *wait)
 699{
 700	struct proc_mounts *p = file->private_data;
 701	unsigned res = POLLIN | POLLRDNORM;
 702
 703	poll_wait(file, &p->ns->poll, wait);
 704	if (mnt_had_events(p))
 705		res |= POLLERR | POLLPRI;
 706
 707	return res;
 708}
 709
 710static int mounts_open(struct inode *inode, struct file *file)
 711{
 712	return mounts_open_common(inode, file, &mounts_op);
 713}
 714
 715static const struct file_operations proc_mounts_operations = {
 716	.open		= mounts_open,
 717	.read		= seq_read,
 718	.llseek		= seq_lseek,
 719	.release	= mounts_release,
 720	.poll		= mounts_poll,
 721};
 722
 723static int mountinfo_open(struct inode *inode, struct file *file)
 724{
 725	return mounts_open_common(inode, file, &mountinfo_op);
 726}
 727
 728static const struct file_operations proc_mountinfo_operations = {
 729	.open		= mountinfo_open,
 730	.read		= seq_read,
 731	.llseek		= seq_lseek,
 732	.release	= mounts_release,
 733	.poll		= mounts_poll,
 734};
 735
 736static int mountstats_open(struct inode *inode, struct file *file)
 737{
 738	return mounts_open_common(inode, file, &mountstats_op);
 
 
 739}
 740
 741static const struct file_operations proc_mountstats_operations = {
 742	.open		= mountstats_open,
 743	.read		= seq_read,
 744	.llseek		= seq_lseek,
 745	.release	= mounts_release,
 746};
 747
 748#define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
 749
 750static ssize_t proc_info_read(struct file * file, char __user * buf,
 751			  size_t count, loff_t *ppos)
 752{
 753	struct inode * inode = file->f_path.dentry->d_inode;
 754	unsigned long page;
 755	ssize_t length;
 756	struct task_struct *task = get_proc_task(inode);
 757
 758	length = -ESRCH;
 759	if (!task)
 760		goto out_no_task;
 
 
 761
 762	if (count > PROC_BLOCK_SIZE)
 763		count = PROC_BLOCK_SIZE;
 
 
 
 
 
 
 
 
 764
 765	length = -ENOMEM;
 766	if (!(page = __get_free_page(GFP_TEMPORARY)))
 767		goto out;
 
 768
 769	length = PROC_I(inode)->op.proc_read(task, (char*)page);
 770
 771	if (length >= 0)
 772		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
 773	free_page(page);
 774out:
 775	put_task_struct(task);
 776out_no_task:
 777	return length;
 778}
 779
 780static const struct file_operations proc_info_file_operations = {
 781	.read		= proc_info_read,
 782	.llseek		= generic_file_llseek,
 783};
 784
 785static int proc_single_show(struct seq_file *m, void *v)
 786{
 787	struct inode *inode = m->private;
 788	struct pid_namespace *ns;
 789	struct pid *pid;
 790	struct task_struct *task;
 791	int ret;
 792
 793	ns = inode->i_sb->s_fs_info;
 794	pid = proc_pid(inode);
 795	task = get_pid_task(pid, PIDTYPE_PID);
 796	if (!task)
 797		return -ESRCH;
 798
 799	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
 800
 801	put_task_struct(task);
 802	return ret;
 803}
 804
 805static int proc_single_open(struct inode *inode, struct file *filp)
 806{
 807	return single_open(filp, proc_single_show, inode);
 808}
 809
 810static const struct file_operations proc_single_file_operations = {
 811	.open		= proc_single_open,
 812	.read		= seq_read,
 813	.llseek		= seq_lseek,
 814	.release	= single_release,
 815};
 816
 817static int mem_open(struct inode* inode, struct file* file)
 818{
 819	file->private_data = (void*)((long)current->self_exec_id);
 820	/* OK to pass negative loff_t, we can catch out-of-range */
 821	file->f_mode |= FMODE_UNSIGNED_OFFSET;
 822	return 0;
 823}
 824
 825static ssize_t mem_read(struct file * file, char __user * buf,
 826			size_t count, loff_t *ppos)
 827{
 828	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
 829	char *page;
 830	unsigned long src = *ppos;
 831	int ret = -ESRCH;
 832	struct mm_struct *mm;
 833
 834	if (!task)
 835		goto out_no_task;
 836
 837	ret = -ENOMEM;
 838	page = (char *)__get_free_page(GFP_TEMPORARY);
 839	if (!page)
 840		goto out;
 841
 842	mm = check_mem_permission(task);
 843	ret = PTR_ERR(mm);
 844	if (IS_ERR(mm))
 845		goto out_free;
 846
 847	ret = -EIO;
 848 
 849	if (file->private_data != (void*)((long)current->self_exec_id))
 850		goto out_put;
 851
 852	ret = 0;
 853 
 854	while (count > 0) {
 855		int this_len, retval;
 856
 857		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
 858		retval = access_remote_vm(mm, src, page, this_len, 0);
 859		if (!retval) {
 860			if (!ret)
 861				ret = -EIO;
 862			break;
 863		}
 864
 865		if (copy_to_user(buf, page, retval)) {
 866			ret = -EFAULT;
 867			break;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 868		}
 869 
 870		ret += retval;
 871		src += retval;
 872		buf += retval;
 873		count -= retval;
 874	}
 875	*ppos = src;
 876
 877out_put:
 878	mmput(mm);
 879out_free:
 880	free_page((unsigned long) page);
 881out:
 882	put_task_struct(task);
 883out_no_task:
 884	return ret;
 885}
 886
 887static ssize_t mem_write(struct file * file, const char __user *buf,
 888			 size_t count, loff_t *ppos)
 889{
 890	int copied;
 
 
 891	char *page;
 892	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
 893	unsigned long dst = *ppos;
 894	struct mm_struct *mm;
 895
 896	copied = -ESRCH;
 897	if (!task)
 898		goto out_no_task;
 899
 900	copied = -ENOMEM;
 901	page = (char *)__get_free_page(GFP_TEMPORARY);
 902	if (!page)
 903		goto out_task;
 904
 905	mm = check_mem_permission(task);
 906	copied = PTR_ERR(mm);
 907	if (IS_ERR(mm))
 908		goto out_free;
 909
 910	copied = -EIO;
 911	if (file->private_data != (void *)((long)current->self_exec_id))
 912		goto out_mm;
 913
 914	copied = 0;
 915	while (count > 0) {
 916		int this_len, retval;
 917
 918		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
 919		if (copy_from_user(page, buf, this_len)) {
 920			copied = -EFAULT;
 921			break;
 922		}
 923		retval = access_remote_vm(mm, dst, page, this_len, 1);
 924		if (!retval) {
 
 925			if (!copied)
 926				copied = -EIO;
 927			break;
 928		}
 929		copied += retval;
 930		buf += retval;
 931		dst += retval;
 932		count -= retval;			
 
 
 
 
 
 
 933	}
 934	*ppos = dst;
 935
 936out_mm:
 937	mmput(mm);
 938out_free:
 939	free_page((unsigned long) page);
 940out_task:
 941	put_task_struct(task);
 942out_no_task:
 943	return copied;
 944}
 945
 
 
 
 
 
 
 
 
 
 
 
 
 946loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 947{
 948	switch (orig) {
 949	case 0:
 950		file->f_pos = offset;
 951		break;
 952	case 1:
 953		file->f_pos += offset;
 954		break;
 955	default:
 956		return -EINVAL;
 957	}
 958	force_successful_syscall_return();
 959	return file->f_pos;
 960}
 961
 
 
 
 
 
 
 
 
 962static const struct file_operations proc_mem_operations = {
 963	.llseek		= mem_lseek,
 964	.read		= mem_read,
 965	.write		= mem_write,
 966	.open		= mem_open,
 
 
 967};
 968
 
 
 
 
 
 969static ssize_t environ_read(struct file *file, char __user *buf,
 970			size_t count, loff_t *ppos)
 971{
 972	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
 973	char *page;
 974	unsigned long src = *ppos;
 975	int ret = -ESRCH;
 976	struct mm_struct *mm;
 
 977
 978	if (!task)
 979		goto out_no_task;
 
 980
 981	ret = -ENOMEM;
 982	page = (char *)__get_free_page(GFP_TEMPORARY);
 983	if (!page)
 984		goto out;
 985
 
 
 
 986
 987	mm = mm_for_maps(task);
 988	ret = PTR_ERR(mm);
 989	if (!mm || IS_ERR(mm))
 990		goto out_free;
 991
 992	ret = 0;
 993	while (count > 0) {
 994		int this_len, retval, max_len;
 995
 996		this_len = mm->env_end - (mm->env_start + src);
 997
 998		if (this_len <= 0)
 999			break;
1000
1001		max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
1002		this_len = (this_len > max_len) ? max_len : this_len;
1003
1004		retval = access_process_vm(task, (mm->env_start + src),
1005			page, this_len, 0);
 
 
1006
1007		if (retval <= 0) {
1008			ret = retval;
1009			break;
1010		}
1011
1012		if (copy_to_user(buf, page, retval)) {
1013			ret = -EFAULT;
1014			break;
1015		}
1016
1017		ret += retval;
1018		src += retval;
1019		buf += retval;
1020		count -= retval;
1021	}
1022	*ppos = src;
1023
1024	mmput(mm);
1025out_free:
 
1026	free_page((unsigned long) page);
1027out:
1028	put_task_struct(task);
1029out_no_task:
1030	return ret;
1031}
1032
1033static const struct file_operations proc_environ_operations = {
 
1034	.read		= environ_read,
1035	.llseek		= generic_file_llseek,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1036};
1037
1038static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1039				size_t count, loff_t *ppos)
1040{
1041	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1042	char buffer[PROC_NUMBUF];
 
1043	size_t len;
1044	int oom_adjust = OOM_DISABLE;
1045	unsigned long flags;
1046
1047	if (!task)
1048		return -ESRCH;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1049
1050	if (lock_task_sighand(task, &flags)) {
1051		oom_adjust = task->signal->oom_adj;
1052		unlock_task_sighand(task, &flags);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1053	}
1054
1055	put_task_struct(task);
 
 
 
 
 
 
1056
1057	len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
 
 
 
 
 
 
 
1058
1059	return simple_read_from_buffer(buf, count, ppos, buffer, len);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1060}
1061
1062static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1063				size_t count, loff_t *ppos)
 
 
 
 
 
 
 
 
 
 
1064{
1065	struct task_struct *task;
1066	char buffer[PROC_NUMBUF];
1067	int oom_adjust;
1068	unsigned long flags;
1069	int err;
1070
1071	memset(buffer, 0, sizeof(buffer));
1072	if (count > sizeof(buffer) - 1)
1073		count = sizeof(buffer) - 1;
1074	if (copy_from_user(buffer, buf, count)) {
1075		err = -EFAULT;
1076		goto out;
1077	}
1078
1079	err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
1080	if (err)
1081		goto out;
1082	if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1083	     oom_adjust != OOM_DISABLE) {
1084		err = -EINVAL;
1085		goto out;
1086	}
1087
1088	task = get_proc_task(file->f_path.dentry->d_inode);
1089	if (!task) {
1090		err = -ESRCH;
1091		goto out;
1092	}
1093
1094	task_lock(task);
1095	if (!task->mm) {
1096		err = -EINVAL;
1097		goto err_task_lock;
1098	}
1099
1100	if (!lock_task_sighand(task, &flags)) {
1101		err = -ESRCH;
1102		goto err_task_lock;
1103	}
1104
1105	if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1106		err = -EACCES;
1107		goto err_sighand;
1108	}
1109
1110	if (oom_adjust != task->signal->oom_adj) {
1111		if (oom_adjust == OOM_DISABLE)
1112			atomic_inc(&task->mm->oom_disable_count);
1113		if (task->signal->oom_adj == OOM_DISABLE)
1114			atomic_dec(&task->mm->oom_disable_count);
1115	}
1116
1117	/*
1118	 * Warn that /proc/pid/oom_adj is deprecated, see
1119	 * Documentation/feature-removal-schedule.txt.
1120	 */
1121	printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1122		  current->comm, task_pid_nr(current), task_pid_nr(task),
1123		  task_pid_nr(task));
1124	task->signal->oom_adj = oom_adjust;
1125	/*
1126	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1127	 * value is always attainable.
1128	 */
1129	if (task->signal->oom_adj == OOM_ADJUST_MAX)
1130		task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1131	else
1132		task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1133								-OOM_DISABLE;
1134err_sighand:
1135	unlock_task_sighand(task, &flags);
1136err_task_lock:
1137	task_unlock(task);
1138	put_task_struct(task);
1139out:
1140	return err < 0 ? err : count;
1141}
1142
1143static const struct file_operations proc_oom_adjust_operations = {
1144	.read		= oom_adjust_read,
1145	.write		= oom_adjust_write,
1146	.llseek		= generic_file_llseek,
1147};
1148
1149static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1150					size_t count, loff_t *ppos)
1151{
1152	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1153	char buffer[PROC_NUMBUF];
1154	int oom_score_adj = OOM_SCORE_ADJ_MIN;
1155	unsigned long flags;
1156	size_t len;
1157
1158	if (!task)
1159		return -ESRCH;
1160	if (lock_task_sighand(task, &flags)) {
1161		oom_score_adj = task->signal->oom_score_adj;
1162		unlock_task_sighand(task, &flags);
1163	}
1164	put_task_struct(task);
1165	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1166	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1167}
1168
1169static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1170					size_t count, loff_t *ppos)
1171{
1172	struct task_struct *task;
1173	char buffer[PROC_NUMBUF];
1174	unsigned long flags;
1175	int oom_score_adj;
1176	int err;
1177
1178	memset(buffer, 0, sizeof(buffer));
1179	if (count > sizeof(buffer) - 1)
1180		count = sizeof(buffer) - 1;
1181	if (copy_from_user(buffer, buf, count)) {
1182		err = -EFAULT;
1183		goto out;
1184	}
1185
1186	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1187	if (err)
1188		goto out;
1189	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1190			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1191		err = -EINVAL;
1192		goto out;
1193	}
1194
1195	task = get_proc_task(file->f_path.dentry->d_inode);
1196	if (!task) {
1197		err = -ESRCH;
1198		goto out;
1199	}
1200
1201	task_lock(task);
1202	if (!task->mm) {
1203		err = -EINVAL;
1204		goto err_task_lock;
1205	}
1206
1207	if (!lock_task_sighand(task, &flags)) {
1208		err = -ESRCH;
1209		goto err_task_lock;
1210	}
1211
1212	if (oom_score_adj < task->signal->oom_score_adj_min &&
1213			!capable(CAP_SYS_RESOURCE)) {
1214		err = -EACCES;
1215		goto err_sighand;
1216	}
1217
1218	if (oom_score_adj != task->signal->oom_score_adj) {
1219		if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1220			atomic_inc(&task->mm->oom_disable_count);
1221		if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1222			atomic_dec(&task->mm->oom_disable_count);
1223	}
1224	task->signal->oom_score_adj = oom_score_adj;
1225	if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1226		task->signal->oom_score_adj_min = oom_score_adj;
1227	/*
1228	 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1229	 * always attainable.
1230	 */
1231	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1232		task->signal->oom_adj = OOM_DISABLE;
1233	else
1234		task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1235							OOM_SCORE_ADJ_MAX;
1236err_sighand:
1237	unlock_task_sighand(task, &flags);
1238err_task_lock:
1239	task_unlock(task);
1240	put_task_struct(task);
1241out:
1242	return err < 0 ? err : count;
1243}
1244
1245static const struct file_operations proc_oom_score_adj_operations = {
1246	.read		= oom_score_adj_read,
1247	.write		= oom_score_adj_write,
1248	.llseek		= default_llseek,
1249};
1250
1251#ifdef CONFIG_AUDITSYSCALL
1252#define TMPBUFLEN 21
1253static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1254				  size_t count, loff_t *ppos)
1255{
1256	struct inode * inode = file->f_path.dentry->d_inode;
1257	struct task_struct *task = get_proc_task(inode);
1258	ssize_t length;
1259	char tmpbuf[TMPBUFLEN];
1260
1261	if (!task)
1262		return -ESRCH;
1263	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1264				audit_get_loginuid(task));
 
1265	put_task_struct(task);
1266	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1267}
1268
1269static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1270				   size_t count, loff_t *ppos)
1271{
1272	struct inode * inode = file->f_path.dentry->d_inode;
1273	char *page, *tmp;
1274	ssize_t length;
1275	uid_t loginuid;
 
 
1276
1277	if (!capable(CAP_AUDIT_CONTROL))
 
1278		return -EPERM;
1279
1280	rcu_read_lock();
1281	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1282		rcu_read_unlock();
1283		return -EPERM;
1284	}
1285	rcu_read_unlock();
1286
1287	if (count >= PAGE_SIZE)
1288		count = PAGE_SIZE - 1;
1289
1290	if (*ppos != 0) {
1291		/* No partial writes. */
1292		return -EINVAL;
1293	}
1294	page = (char*)__get_free_page(GFP_TEMPORARY);
1295	if (!page)
1296		return -ENOMEM;
1297	length = -EFAULT;
1298	if (copy_from_user(page, buf, count))
1299		goto out_free_page;
1300
1301	page[count] = '\0';
1302	loginuid = simple_strtoul(page, &tmp, 10);
1303	if (tmp == page) {
1304		length = -EINVAL;
1305		goto out_free_page;
1306
1307	}
1308	length = audit_set_loginuid(current, loginuid);
1309	if (likely(length == 0))
1310		length = count;
1311
1312out_free_page:
1313	free_page((unsigned long) page);
1314	return length;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1315}
1316
1317static const struct file_operations proc_loginuid_operations = {
1318	.read		= proc_loginuid_read,
1319	.write		= proc_loginuid_write,
1320	.llseek		= generic_file_llseek,
1321};
1322
1323static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1324				  size_t count, loff_t *ppos)
1325{
1326	struct inode * inode = file->f_path.dentry->d_inode;
1327	struct task_struct *task = get_proc_task(inode);
1328	ssize_t length;
1329	char tmpbuf[TMPBUFLEN];
1330
1331	if (!task)
1332		return -ESRCH;
1333	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1334				audit_get_sessionid(task));
1335	put_task_struct(task);
1336	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1337}
1338
1339static const struct file_operations proc_sessionid_operations = {
1340	.read		= proc_sessionid_read,
1341	.llseek		= generic_file_llseek,
1342};
1343#endif
1344
1345#ifdef CONFIG_FAULT_INJECTION
1346static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1347				      size_t count, loff_t *ppos)
1348{
1349	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1350	char buffer[PROC_NUMBUF];
1351	size_t len;
1352	int make_it_fail;
1353
1354	if (!task)
1355		return -ESRCH;
1356	make_it_fail = task->make_it_fail;
1357	put_task_struct(task);
1358
1359	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1360
1361	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1362}
1363
1364static ssize_t proc_fault_inject_write(struct file * file,
1365			const char __user * buf, size_t count, loff_t *ppos)
1366{
1367	struct task_struct *task;
1368	char buffer[PROC_NUMBUF], *end;
1369	int make_it_fail;
 
1370
1371	if (!capable(CAP_SYS_RESOURCE))
1372		return -EPERM;
1373	memset(buffer, 0, sizeof(buffer));
1374	if (count > sizeof(buffer) - 1)
1375		count = sizeof(buffer) - 1;
1376	if (copy_from_user(buffer, buf, count))
1377		return -EFAULT;
1378	make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1379	if (*end)
 
 
1380		return -EINVAL;
1381	task = get_proc_task(file->f_dentry->d_inode);
 
1382	if (!task)
1383		return -ESRCH;
1384	task->make_it_fail = make_it_fail;
1385	put_task_struct(task);
1386
1387	return count;
1388}
1389
1390static const struct file_operations proc_fault_inject_operations = {
1391	.read		= proc_fault_inject_read,
1392	.write		= proc_fault_inject_write,
1393	.llseek		= generic_file_llseek,
1394};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1395#endif
1396
1397
1398#ifdef CONFIG_SCHED_DEBUG
1399/*
1400 * Print out various scheduling related per-task fields:
1401 */
1402static int sched_show(struct seq_file *m, void *v)
1403{
1404	struct inode *inode = m->private;
 
1405	struct task_struct *p;
1406
1407	p = get_proc_task(inode);
1408	if (!p)
1409		return -ESRCH;
1410	proc_sched_show_task(p, m);
1411
1412	put_task_struct(p);
1413
1414	return 0;
1415}
1416
1417static ssize_t
1418sched_write(struct file *file, const char __user *buf,
1419	    size_t count, loff_t *offset)
1420{
1421	struct inode *inode = file->f_path.dentry->d_inode;
1422	struct task_struct *p;
1423
1424	p = get_proc_task(inode);
1425	if (!p)
1426		return -ESRCH;
1427	proc_sched_set_task(p);
1428
1429	put_task_struct(p);
1430
1431	return count;
1432}
1433
1434static int sched_open(struct inode *inode, struct file *filp)
1435{
1436	return single_open(filp, sched_show, inode);
1437}
1438
1439static const struct file_operations proc_pid_sched_operations = {
1440	.open		= sched_open,
1441	.read		= seq_read,
1442	.write		= sched_write,
1443	.llseek		= seq_lseek,
1444	.release	= single_release,
1445};
1446
1447#endif
1448
1449#ifdef CONFIG_SCHED_AUTOGROUP
1450/*
1451 * Print out autogroup related information:
1452 */
1453static int sched_autogroup_show(struct seq_file *m, void *v)
1454{
1455	struct inode *inode = m->private;
1456	struct task_struct *p;
1457
1458	p = get_proc_task(inode);
1459	if (!p)
1460		return -ESRCH;
1461	proc_sched_autogroup_show_task(p, m);
1462
1463	put_task_struct(p);
1464
1465	return 0;
1466}
1467
1468static ssize_t
1469sched_autogroup_write(struct file *file, const char __user *buf,
1470	    size_t count, loff_t *offset)
1471{
1472	struct inode *inode = file->f_path.dentry->d_inode;
1473	struct task_struct *p;
1474	char buffer[PROC_NUMBUF];
1475	int nice;
1476	int err;
1477
1478	memset(buffer, 0, sizeof(buffer));
1479	if (count > sizeof(buffer) - 1)
1480		count = sizeof(buffer) - 1;
1481	if (copy_from_user(buffer, buf, count))
1482		return -EFAULT;
1483
1484	err = kstrtoint(strstrip(buffer), 0, &nice);
1485	if (err < 0)
1486		return err;
1487
1488	p = get_proc_task(inode);
1489	if (!p)
1490		return -ESRCH;
1491
1492	err = nice;
1493	err = proc_sched_autogroup_set_nice(p, &err);
1494	if (err)
1495		count = err;
1496
1497	put_task_struct(p);
1498
1499	return count;
1500}
1501
1502static int sched_autogroup_open(struct inode *inode, struct file *filp)
1503{
1504	int ret;
1505
1506	ret = single_open(filp, sched_autogroup_show, NULL);
1507	if (!ret) {
1508		struct seq_file *m = filp->private_data;
1509
1510		m->private = inode;
1511	}
1512	return ret;
1513}
1514
1515static const struct file_operations proc_pid_sched_autogroup_operations = {
1516	.open		= sched_autogroup_open,
1517	.read		= seq_read,
1518	.write		= sched_autogroup_write,
1519	.llseek		= seq_lseek,
1520	.release	= single_release,
1521};
1522
1523#endif /* CONFIG_SCHED_AUTOGROUP */
1524
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1525static ssize_t comm_write(struct file *file, const char __user *buf,
1526				size_t count, loff_t *offset)
1527{
1528	struct inode *inode = file->f_path.dentry->d_inode;
1529	struct task_struct *p;
1530	char buffer[TASK_COMM_LEN];
 
1531
1532	memset(buffer, 0, sizeof(buffer));
1533	if (count > sizeof(buffer) - 1)
1534		count = sizeof(buffer) - 1;
1535	if (copy_from_user(buffer, buf, count))
1536		return -EFAULT;
1537
1538	p = get_proc_task(inode);
1539	if (!p)
1540		return -ESRCH;
1541
1542	if (same_thread_group(current, p))
1543		set_task_comm(p, buffer);
 
 
1544	else
1545		count = -EINVAL;
1546
1547	put_task_struct(p);
1548
1549	return count;
1550}
1551
1552static int comm_show(struct seq_file *m, void *v)
1553{
1554	struct inode *inode = m->private;
1555	struct task_struct *p;
1556
1557	p = get_proc_task(inode);
1558	if (!p)
1559		return -ESRCH;
1560
1561	task_lock(p);
1562	seq_printf(m, "%s\n", p->comm);
1563	task_unlock(p);
1564
1565	put_task_struct(p);
1566
1567	return 0;
1568}
1569
1570static int comm_open(struct inode *inode, struct file *filp)
1571{
1572	return single_open(filp, comm_show, inode);
1573}
1574
1575static const struct file_operations proc_pid_set_comm_operations = {
1576	.open		= comm_open,
1577	.read		= seq_read,
1578	.write		= comm_write,
1579	.llseek		= seq_lseek,
1580	.release	= single_release,
1581};
1582
1583static int proc_exe_link(struct inode *inode, struct path *exe_path)
1584{
1585	struct task_struct *task;
1586	struct mm_struct *mm;
1587	struct file *exe_file;
1588
1589	task = get_proc_task(inode);
1590	if (!task)
1591		return -ENOENT;
1592	mm = get_task_mm(task);
1593	put_task_struct(task);
1594	if (!mm)
1595		return -ENOENT;
1596	exe_file = get_mm_exe_file(mm);
1597	mmput(mm);
1598	if (exe_file) {
1599		*exe_path = exe_file->f_path;
1600		path_get(&exe_file->f_path);
1601		fput(exe_file);
1602		return 0;
1603	} else
1604		return -ENOENT;
1605}
1606
1607static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
 
 
1608{
1609	struct inode *inode = dentry->d_inode;
1610	int error = -EACCES;
1611
1612	/* We don't need a base pointer in the /proc filesystem */
1613	path_put(&nd->path);
1614
1615	/* Are we allowed to snoop on the tasks file descriptors? */
1616	if (!proc_fd_access_allowed(inode))
1617		goto out;
1618
1619	error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
 
 
 
 
1620out:
1621	return ERR_PTR(error);
1622}
1623
1624static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1625{
1626	char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1627	char *pathname;
1628	int len;
1629
1630	if (!tmp)
1631		return -ENOMEM;
1632
1633	pathname = d_path(path, tmp, PAGE_SIZE);
1634	len = PTR_ERR(pathname);
1635	if (IS_ERR(pathname))
1636		goto out;
1637	len = tmp + PAGE_SIZE - 1 - pathname;
1638
1639	if (len > buflen)
1640		len = buflen;
1641	if (copy_to_user(buffer, pathname, len))
1642		len = -EFAULT;
1643 out:
1644	free_page((unsigned long)tmp);
1645	return len;
1646}
1647
1648static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1649{
1650	int error = -EACCES;
1651	struct inode *inode = dentry->d_inode;
1652	struct path path;
1653
1654	/* Are we allowed to snoop on the tasks file descriptors? */
1655	if (!proc_fd_access_allowed(inode))
1656		goto out;
1657
1658	error = PROC_I(inode)->op.proc_get_link(inode, &path);
1659	if (error)
1660		goto out;
1661
1662	error = do_proc_readlink(&path, buffer, buflen);
1663	path_put(&path);
1664out:
1665	return error;
1666}
1667
1668static const struct inode_operations proc_pid_link_inode_operations = {
1669	.readlink	= proc_pid_readlink,
1670	.follow_link	= proc_pid_follow_link,
1671	.setattr	= proc_setattr,
1672};
1673
1674
1675/* building an inode */
1676
1677static int task_dumpable(struct task_struct *task)
 
1678{
1679	int dumpable = 0;
1680	struct mm_struct *mm;
 
 
 
 
1681
1682	task_lock(task);
1683	mm = task->mm;
1684	if (mm)
1685		dumpable = get_dumpable(mm);
1686	task_unlock(task);
1687	if(dumpable == 1)
1688		return 1;
1689	return 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1690}
1691
1692struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
 
1693{
1694	struct inode * inode;
1695	struct proc_inode *ei;
1696	const struct cred *cred;
1697
1698	/* We need a new inode */
1699
1700	inode = new_inode(sb);
1701	if (!inode)
1702		goto out;
1703
1704	/* Common stuff */
1705	ei = PROC_I(inode);
 
1706	inode->i_ino = get_next_ino();
1707	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1708	inode->i_op = &proc_def_inode_operations;
1709
1710	/*
1711	 * grab the reference to task.
1712	 */
1713	ei->pid = get_task_pid(task, PIDTYPE_PID);
1714	if (!ei->pid)
1715		goto out_unlock;
1716
1717	if (task_dumpable(task)) {
1718		rcu_read_lock();
1719		cred = __task_cred(task);
1720		inode->i_uid = cred->euid;
1721		inode->i_gid = cred->egid;
1722		rcu_read_unlock();
1723	}
1724	security_task_to_inode(task, inode);
1725
1726out:
1727	return inode;
1728
1729out_unlock:
1730	iput(inode);
1731	return NULL;
1732}
1733
1734int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1735{
1736	struct inode *inode = dentry->d_inode;
 
1737	struct task_struct *task;
1738	const struct cred *cred;
1739
1740	generic_fillattr(inode, stat);
1741
 
 
1742	rcu_read_lock();
1743	stat->uid = 0;
1744	stat->gid = 0;
1745	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1746	if (task) {
1747		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1748		    task_dumpable(task)) {
1749			cred = __task_cred(task);
1750			stat->uid = cred->euid;
1751			stat->gid = cred->egid;
 
 
1752		}
 
1753	}
1754	rcu_read_unlock();
1755	return 0;
1756}
1757
1758/* dentry stuff */
1759
1760/*
1761 *	Exceptional case: normally we are not allowed to unhash a busy
1762 * directory. In this case, however, we can do it - no aliasing problems
1763 * due to the way we treat inodes.
1764 *
 
 
 
 
 
 
 
1765 * Rewrite the inode's ownerships here because the owning task may have
1766 * performed a setuid(), etc.
1767 *
1768 * Before the /proc/pid/status file was created the only way to read
1769 * the effective uid of a /process was to stat /proc/pid.  Reading
1770 * /proc/pid/status is slow enough that procps and other packages
1771 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1772 * made this apply to all per process world readable and executable
1773 * directories.
1774 */
1775int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1776{
1777	struct inode *inode;
1778	struct task_struct *task;
1779	const struct cred *cred;
1780
1781	if (nd && nd->flags & LOOKUP_RCU)
1782		return -ECHILD;
1783
1784	inode = dentry->d_inode;
1785	task = get_proc_task(inode);
 
 
 
1786
1787	if (task) {
1788		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1789		    task_dumpable(task)) {
1790			rcu_read_lock();
1791			cred = __task_cred(task);
1792			inode->i_uid = cred->euid;
1793			inode->i_gid = cred->egid;
1794			rcu_read_unlock();
1795		} else {
1796			inode->i_uid = 0;
1797			inode->i_gid = 0;
1798		}
1799		inode->i_mode &= ~(S_ISUID | S_ISGID);
1800		security_task_to_inode(task, inode);
1801		put_task_struct(task);
1802		return 1;
1803	}
1804	d_drop(dentry);
1805	return 0;
 
 
 
 
 
 
1806}
1807
1808static int pid_delete_dentry(const struct dentry * dentry)
1809{
1810	/* Is the task we represent dead?
1811	 * If so, then don't put the dentry on the lru list,
1812	 * kill it immediately.
1813	 */
1814	return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1815}
1816
1817const struct dentry_operations pid_dentry_operations =
1818{
1819	.d_revalidate	= pid_revalidate,
1820	.d_delete	= pid_delete_dentry,
1821};
1822
1823/* Lookups */
1824
1825/*
1826 * Fill a directory entry.
1827 *
1828 * If possible create the dcache entry and derive our inode number and
1829 * file type from dcache entry.
1830 *
1831 * Since all of the proc inode numbers are dynamically generated, the inode
1832 * numbers do not exist until the inode is cache.  This means creating the
1833 * the dcache entry in readdir is necessary to keep the inode numbers
1834 * reported by readdir in sync with the inode numbers reported
1835 * by stat.
1836 */
1837int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1838	const char *name, int len,
1839	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1840{
1841	struct dentry *child, *dir = filp->f_path.dentry;
 
1842	struct inode *inode;
1843	struct qstr qname;
1844	ino_t ino = 0;
1845	unsigned type = DT_UNKNOWN;
 
1846
1847	qname.name = name;
1848	qname.len  = len;
1849	qname.hash = full_name_hash(name, len);
1850
1851	child = d_lookup(dir, &qname);
1852	if (!child) {
1853		struct dentry *new;
1854		new = d_alloc(dir, &qname);
1855		if (new) {
1856			child = instantiate(dir->d_inode, new, task, ptr);
1857			if (child)
1858				dput(new);
1859			else
1860				child = new;
1861		}
1862	}
1863	if (!child || IS_ERR(child) || !child->d_inode)
1864		goto end_instantiate;
1865	inode = child->d_inode;
1866	if (inode) {
1867		ino = inode->i_ino;
1868		type = inode->i_mode >> 12;
1869	}
 
 
 
1870	dput(child);
1871end_instantiate:
1872	if (!ino)
1873		ino = find_inode_number(dir, &qname);
1874	if (!ino)
1875		ino = 1;
1876	return filldir(dirent, name, len, filp->f_pos, ino, type);
1877}
1878
1879static unsigned name_to_int(struct dentry *dentry)
 
 
 
 
 
1880{
1881	const char *name = dentry->d_name.name;
1882	int len = dentry->d_name.len;
1883	unsigned n = 0;
1884
1885	if (len > 1 && *name == '0')
1886		goto out;
1887	while (len-- > 0) {
1888		unsigned c = *name++ - '0';
1889		if (c > 9)
1890			goto out;
1891		if (n >= (~0U-9)/10)
1892			goto out;
1893		n *= 10;
1894		n += c;
1895	}
1896	return n;
1897out:
1898	return ~0U;
1899}
1900
1901#define PROC_FDINFO_MAX 64
 
 
1902
1903static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1904{
1905	struct task_struct *task = get_proc_task(inode);
1906	struct files_struct *files = NULL;
1907	struct file *file;
1908	int fd = proc_fd(inode);
 
 
1909
1910	if (task) {
1911		files = get_files_struct(task);
1912		put_task_struct(task);
1913	}
1914	if (files) {
1915		/*
1916		 * We are not taking a ref to the file structure, so we must
1917		 * hold ->file_lock.
1918		 */
1919		spin_lock(&files->file_lock);
1920		file = fcheck_files(files, fd);
1921		if (file) {
1922			unsigned int f_flags;
1923			struct fdtable *fdt;
1924
1925			fdt = files_fdtable(files);
1926			f_flags = file->f_flags & ~O_CLOEXEC;
1927			if (FD_ISSET(fd, fdt->close_on_exec))
1928				f_flags |= O_CLOEXEC;
1929
1930			if (path) {
1931				*path = file->f_path;
1932				path_get(&file->f_path);
1933			}
1934			if (info)
1935				snprintf(info, PROC_FDINFO_MAX,
1936					 "pos:\t%lli\n"
1937					 "flags:\t0%o\n",
1938					 (long long) file->f_pos,
1939					 f_flags);
1940			spin_unlock(&files->file_lock);
1941			put_files_struct(files);
1942			return 0;
1943		}
1944		spin_unlock(&files->file_lock);
1945		put_files_struct(files);
1946	}
1947	return -ENOENT;
1948}
1949
1950static int proc_fd_link(struct inode *inode, struct path *path)
1951{
1952	return proc_fd_info(inode, path, NULL);
 
1953}
1954
1955static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1956{
1957	struct inode *inode;
 
 
1958	struct task_struct *task;
1959	int fd;
1960	struct files_struct *files;
1961	const struct cred *cred;
1962
1963	if (nd && nd->flags & LOOKUP_RCU)
1964		return -ECHILD;
1965
1966	inode = dentry->d_inode;
1967	task = get_proc_task(inode);
1968	fd = proc_fd(inode);
 
1969
1970	if (task) {
1971		files = get_files_struct(task);
1972		if (files) {
1973			rcu_read_lock();
1974			if (fcheck_files(files, fd)) {
1975				rcu_read_unlock();
1976				put_files_struct(files);
1977				if (task_dumpable(task)) {
1978					rcu_read_lock();
1979					cred = __task_cred(task);
1980					inode->i_uid = cred->euid;
1981					inode->i_gid = cred->egid;
1982					rcu_read_unlock();
1983				} else {
1984					inode->i_uid = 0;
1985					inode->i_gid = 0;
1986				}
1987				inode->i_mode &= ~(S_ISUID | S_ISGID);
1988				security_task_to_inode(task, inode);
1989				put_task_struct(task);
1990				return 1;
1991			}
1992			rcu_read_unlock();
1993			put_files_struct(files);
1994		}
1995		put_task_struct(task);
1996	}
1997	d_drop(dentry);
1998	return 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
1999}
2000
2001static const struct dentry_operations tid_fd_dentry_operations =
2002{
2003	.d_revalidate	= tid_fd_revalidate,
2004	.d_delete	= pid_delete_dentry,
2005};
2006
2007static struct dentry *proc_fd_instantiate(struct inode *dir,
2008	struct dentry *dentry, struct task_struct *task, const void *ptr)
2009{
2010	unsigned fd = *(const unsigned *)ptr;
2011	struct file *file;
2012	struct files_struct *files;
2013 	struct inode *inode;
2014 	struct proc_inode *ei;
2015	struct dentry *error = ERR_PTR(-ENOENT);
2016
2017	inode = proc_pid_make_inode(dir->i_sb, task);
2018	if (!inode)
 
2019		goto out;
2020	ei = PROC_I(inode);
2021	ei->fd = fd;
2022	files = get_files_struct(task);
2023	if (!files)
2024		goto out_iput;
2025	inode->i_mode = S_IFLNK;
2026
2027	/*
2028	 * We are not taking a ref to the file structure, so we must
2029	 * hold ->file_lock.
2030	 */
2031	spin_lock(&files->file_lock);
2032	file = fcheck_files(files, fd);
2033	if (!file)
2034		goto out_unlock;
2035	if (file->f_mode & FMODE_READ)
2036		inode->i_mode |= S_IRUSR | S_IXUSR;
2037	if (file->f_mode & FMODE_WRITE)
2038		inode->i_mode |= S_IWUSR | S_IXUSR;
2039	spin_unlock(&files->file_lock);
2040	put_files_struct(files);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2041
2042	inode->i_op = &proc_pid_link_inode_operations;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2043	inode->i_size = 64;
2044	ei->op.proc_get_link = proc_fd_link;
2045	d_set_d_op(dentry, &tid_fd_dentry_operations);
2046	d_add(dentry, inode);
2047	/* Close the race of the process dying before we return the dentry */
2048	if (tid_fd_revalidate(dentry, NULL))
2049		error = NULL;
2050
2051 out:
2052	return error;
2053out_unlock:
2054	spin_unlock(&files->file_lock);
2055	put_files_struct(files);
2056out_iput:
2057	iput(inode);
2058	goto out;
2059}
2060
2061static struct dentry *proc_lookupfd_common(struct inode *dir,
2062					   struct dentry *dentry,
2063					   instantiate_t instantiate)
2064{
2065	struct task_struct *task = get_proc_task(dir);
2066	unsigned fd = name_to_int(dentry);
2067	struct dentry *result = ERR_PTR(-ENOENT);
 
 
2068
 
 
2069	if (!task)
2070		goto out_no_task;
2071	if (fd == ~0U)
2072		goto out;
2073
2074	result = instantiate(dir, dentry, task, &fd);
2075out:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2076	put_task_struct(task);
2077out_no_task:
2078	return result;
2079}
2080
2081static int proc_readfd_common(struct file * filp, void * dirent,
2082			      filldir_t filldir, instantiate_t instantiate)
 
 
 
 
 
 
2083{
2084	struct dentry *dentry = filp->f_path.dentry;
2085	struct inode *inode = dentry->d_inode;
2086	struct task_struct *p = get_proc_task(inode);
2087	unsigned int fd, ino;
2088	int retval;
2089	struct files_struct * files;
 
 
2090
2091	retval = -ENOENT;
2092	if (!p)
2093		goto out_no_task;
2094	retval = 0;
2095
2096	fd = filp->f_pos;
2097	switch (fd) {
2098		case 0:
2099			if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2100				goto out;
2101			filp->f_pos++;
2102		case 1:
2103			ino = parent_ino(dentry);
2104			if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2105				goto out;
2106			filp->f_pos++;
2107		default:
2108			files = get_files_struct(p);
2109			if (!files)
2110				goto out;
2111			rcu_read_lock();
2112			for (fd = filp->f_pos-2;
2113			     fd < files_fdtable(files)->max_fds;
2114			     fd++, filp->f_pos++) {
2115				char name[PROC_NUMBUF];
2116				int len;
2117
2118				if (!fcheck_files(files, fd))
2119					continue;
2120				rcu_read_unlock();
2121
2122				len = snprintf(name, sizeof(name), "%d", fd);
2123				if (proc_fill_cache(filp, dirent, filldir,
2124						    name, len, instantiate,
2125						    p, &fd) < 0) {
2126					rcu_read_lock();
2127					break;
2128				}
2129				rcu_read_lock();
2130			}
2131			rcu_read_unlock();
2132			put_files_struct(files);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2133	}
 
 
 
2134out:
2135	put_task_struct(p);
2136out_no_task:
2137	return retval;
2138}
2139
2140static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2141				    struct nameidata *nd)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2142{
2143	return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
 
 
 
 
 
 
 
 
 
 
2144}
2145
2146static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2147{
2148	return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
 
2149}
2150
2151static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2152				      size_t len, loff_t *ppos)
2153{
2154	char tmp[PROC_FDINFO_MAX];
2155	int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2156	if (!err)
2157		err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2158	return err;
 
 
 
 
 
 
2159}
2160
2161static const struct file_operations proc_fdinfo_file_operations = {
2162	.open           = nonseekable_open,
2163	.read		= proc_fdinfo_read,
2164	.llseek		= no_llseek,
2165};
 
 
 
 
 
2166
2167static const struct file_operations proc_fd_operations = {
2168	.read		= generic_read_dir,
2169	.readdir	= proc_readfd,
2170	.llseek		= default_llseek,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2171};
2172
2173/*
2174 * /proc/pid/fd needs a special permission handler so that a process can still
2175 * access /proc/self/fd after it has executed a setuid().
2176 */
2177static int proc_fd_permission(struct inode *inode, int mask)
2178{
2179	int rv = generic_permission(inode, mask);
2180	if (rv == 0)
2181		return 0;
2182	if (task_pid(current) == proc_pid(inode))
2183		rv = 0;
2184	return rv;
 
 
 
 
2185}
2186
2187/*
2188 * proc directories can do almost nothing..
2189 */
2190static const struct inode_operations proc_fd_inode_operations = {
2191	.lookup		= proc_lookupfd,
2192	.permission	= proc_fd_permission,
2193	.setattr	= proc_setattr,
2194};
 
2195
2196static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2197	struct dentry *dentry, struct task_struct *task, const void *ptr)
2198{
2199	unsigned fd = *(unsigned *)ptr;
2200 	struct inode *inode;
2201 	struct proc_inode *ei;
2202	struct dentry *error = ERR_PTR(-ENOENT);
2203
2204	inode = proc_pid_make_inode(dir->i_sb, task);
2205	if (!inode)
2206		goto out;
2207	ei = PROC_I(inode);
2208	ei->fd = fd;
2209	inode->i_mode = S_IFREG | S_IRUSR;
2210	inode->i_fop = &proc_fdinfo_file_operations;
2211	d_set_d_op(dentry, &tid_fd_dentry_operations);
2212	d_add(dentry, inode);
2213	/* Close the race of the process dying before we return the dentry */
2214	if (tid_fd_revalidate(dentry, NULL))
2215		error = NULL;
2216
2217 out:
2218	return error;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2219}
2220
2221static struct dentry *proc_lookupfdinfo(struct inode *dir,
2222					struct dentry *dentry,
2223					struct nameidata *nd)
2224{
2225	return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2226}
2227
2228static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2229{
2230	return proc_readfd_common(filp, dirent, filldir,
2231				  proc_fdinfo_instantiate);
2232}
2233
2234static const struct file_operations proc_fdinfo_operations = {
2235	.read		= generic_read_dir,
2236	.readdir	= proc_readfdinfo,
2237	.llseek		= default_llseek,
2238};
2239
2240/*
2241 * proc directories can do almost nothing..
2242 */
2243static const struct inode_operations proc_fdinfo_inode_operations = {
2244	.lookup		= proc_lookupfdinfo,
2245	.setattr	= proc_setattr,
2246};
2247
2248
2249static struct dentry *proc_pident_instantiate(struct inode *dir,
2250	struct dentry *dentry, struct task_struct *task, const void *ptr)
2251{
2252	const struct pid_entry *p = ptr;
2253	struct inode *inode;
2254	struct proc_inode *ei;
2255	struct dentry *error = ERR_PTR(-ENOENT);
2256
2257	inode = proc_pid_make_inode(dir->i_sb, task);
2258	if (!inode)
2259		goto out;
2260
2261	ei = PROC_I(inode);
2262	inode->i_mode = p->mode;
2263	if (S_ISDIR(inode->i_mode))
2264		inode->i_nlink = 2;	/* Use getattr to fix if necessary */
2265	if (p->iop)
2266		inode->i_op = p->iop;
2267	if (p->fop)
2268		inode->i_fop = p->fop;
2269	ei->op = p->op;
 
2270	d_set_d_op(dentry, &pid_dentry_operations);
2271	d_add(dentry, inode);
2272	/* Close the race of the process dying before we return the dentry */
2273	if (pid_revalidate(dentry, NULL))
2274		error = NULL;
2275out:
2276	return error;
2277}
2278
2279static struct dentry *proc_pident_lookup(struct inode *dir, 
2280					 struct dentry *dentry,
2281					 const struct pid_entry *ents,
2282					 unsigned int nents)
2283{
2284	struct dentry *error;
2285	struct task_struct *task = get_proc_task(dir);
2286	const struct pid_entry *p, *last;
2287
2288	error = ERR_PTR(-ENOENT);
2289
2290	if (!task)
2291		goto out_no_task;
2292
2293	/*
2294	 * Yes, it does not scale. And it should not. Don't add
2295	 * new entries into /proc/<tgid>/ without very good reasons.
2296	 */
2297	last = &ents[nents - 1];
2298	for (p = ents; p <= last; p++) {
2299		if (p->len != dentry->d_name.len)
2300			continue;
2301		if (!memcmp(dentry->d_name.name, p->name, p->len))
 
2302			break;
 
2303	}
2304	if (p > last)
2305		goto out;
2306
2307	error = proc_pident_instantiate(dir, dentry, task, p);
2308out:
2309	put_task_struct(task);
2310out_no_task:
2311	return error;
2312}
2313
2314static int proc_pident_fill_cache(struct file *filp, void *dirent,
2315	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2316{
2317	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2318				proc_pident_instantiate, task, p);
2319}
2320
2321static int proc_pident_readdir(struct file *filp,
2322		void *dirent, filldir_t filldir,
2323		const struct pid_entry *ents, unsigned int nents)
2324{
2325	int i;
2326	struct dentry *dentry = filp->f_path.dentry;
2327	struct inode *inode = dentry->d_inode;
2328	struct task_struct *task = get_proc_task(inode);
2329	const struct pid_entry *p, *last;
2330	ino_t ino;
2331	int ret;
2332
2333	ret = -ENOENT;
2334	if (!task)
2335		goto out_no_task;
2336
2337	ret = 0;
2338	i = filp->f_pos;
2339	switch (i) {
2340	case 0:
2341		ino = inode->i_ino;
2342		if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2343			goto out;
2344		i++;
2345		filp->f_pos++;
2346		/* fall through */
2347	case 1:
2348		ino = parent_ino(dentry);
2349		if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2350			goto out;
2351		i++;
2352		filp->f_pos++;
2353		/* fall through */
2354	default:
2355		i -= 2;
2356		if (i >= nents) {
2357			ret = 1;
2358			goto out;
2359		}
2360		p = ents + i;
2361		last = &ents[nents - 1];
2362		while (p <= last) {
2363			if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2364				goto out;
2365			filp->f_pos++;
2366			p++;
2367		}
2368	}
2369
2370	ret = 1;
 
 
 
 
 
2371out:
2372	put_task_struct(task);
2373out_no_task:
2374	return ret;
2375}
2376
2377#ifdef CONFIG_SECURITY
 
 
 
 
 
 
 
2378static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2379				  size_t count, loff_t *ppos)
2380{
2381	struct inode * inode = file->f_path.dentry->d_inode;
2382	char *p = NULL;
2383	ssize_t length;
2384	struct task_struct *task = get_proc_task(inode);
2385
2386	if (!task)
2387		return -ESRCH;
2388
2389	length = security_getprocattr(task,
2390				      (char*)file->f_path.dentry->d_name.name,
2391				      &p);
2392	put_task_struct(task);
2393	if (length > 0)
2394		length = simple_read_from_buffer(buf, count, ppos, p, length);
2395	kfree(p);
2396	return length;
2397}
2398
2399static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2400				   size_t count, loff_t *ppos)
2401{
2402	struct inode * inode = file->f_path.dentry->d_inode;
2403	char *page;
2404	ssize_t length;
2405	struct task_struct *task = get_proc_task(inode);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2406
2407	length = -ESRCH;
2408	if (!task)
2409		goto out_no_task;
2410	if (count > PAGE_SIZE)
2411		count = PAGE_SIZE;
2412
2413	/* No partial writes. */
2414	length = -EINVAL;
2415	if (*ppos != 0)
2416		goto out;
2417
2418	length = -ENOMEM;
2419	page = (char*)__get_free_page(GFP_TEMPORARY);
2420	if (!page)
2421		goto out;
2422
2423	length = -EFAULT;
2424	if (copy_from_user(page, buf, count))
2425		goto out_free;
2426
2427	/* Guard against adverse ptrace interaction */
2428	length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2429	if (length < 0)
2430		goto out_free;
2431
2432	length = security_setprocattr(task,
2433				      (char*)file->f_path.dentry->d_name.name,
2434				      (void*)page, count);
2435	mutex_unlock(&task->signal->cred_guard_mutex);
2436out_free:
2437	free_page((unsigned long) page);
2438out:
2439	put_task_struct(task);
2440out_no_task:
2441	return length;
2442}
2443
2444static const struct file_operations proc_pid_attr_operations = {
 
2445	.read		= proc_pid_attr_read,
2446	.write		= proc_pid_attr_write,
2447	.llseek		= generic_file_llseek,
 
2448};
2449
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2450static const struct pid_entry attr_dir_stuff[] = {
2451	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2452	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2453	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2454	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2455	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2456	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
 
 
 
 
 
 
 
 
2457};
2458
2459static int proc_attr_dir_readdir(struct file * filp,
2460			     void * dirent, filldir_t filldir)
2461{
2462	return proc_pident_readdir(filp,dirent,filldir,
2463				   attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2464}
2465
2466static const struct file_operations proc_attr_dir_operations = {
2467	.read		= generic_read_dir,
2468	.readdir	= proc_attr_dir_readdir,
2469	.llseek		= default_llseek,
2470};
2471
2472static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2473				struct dentry *dentry, struct nameidata *nd)
2474{
2475	return proc_pident_lookup(dir, dentry,
2476				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
 
2477}
2478
2479static const struct inode_operations proc_attr_dir_inode_operations = {
2480	.lookup		= proc_attr_dir_lookup,
2481	.getattr	= pid_getattr,
2482	.setattr	= proc_setattr,
2483};
2484
2485#endif
2486
2487#ifdef CONFIG_ELF_CORE
2488static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2489					 size_t count, loff_t *ppos)
2490{
2491	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2492	struct mm_struct *mm;
2493	char buffer[PROC_NUMBUF];
2494	size_t len;
2495	int ret;
2496
2497	if (!task)
2498		return -ESRCH;
2499
2500	ret = 0;
2501	mm = get_task_mm(task);
2502	if (mm) {
2503		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2504			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2505				MMF_DUMP_FILTER_SHIFT));
2506		mmput(mm);
2507		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2508	}
2509
2510	put_task_struct(task);
2511
2512	return ret;
2513}
2514
2515static ssize_t proc_coredump_filter_write(struct file *file,
2516					  const char __user *buf,
2517					  size_t count,
2518					  loff_t *ppos)
2519{
2520	struct task_struct *task;
2521	struct mm_struct *mm;
2522	char buffer[PROC_NUMBUF], *end;
2523	unsigned int val;
2524	int ret;
2525	int i;
2526	unsigned long mask;
2527
2528	ret = -EFAULT;
2529	memset(buffer, 0, sizeof(buffer));
2530	if (count > sizeof(buffer) - 1)
2531		count = sizeof(buffer) - 1;
2532	if (copy_from_user(buffer, buf, count))
2533		goto out_no_task;
2534
2535	ret = -EINVAL;
2536	val = (unsigned int)simple_strtoul(buffer, &end, 0);
2537	if (*end == '\n')
2538		end++;
2539	if (end - buffer == 0)
2540		goto out_no_task;
2541
2542	ret = -ESRCH;
2543	task = get_proc_task(file->f_dentry->d_inode);
2544	if (!task)
2545		goto out_no_task;
2546
2547	ret = end - buffer;
2548	mm = get_task_mm(task);
2549	if (!mm)
2550		goto out_no_mm;
 
2551
2552	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2553		if (val & mask)
2554			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2555		else
2556			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2557	}
2558
2559	mmput(mm);
2560 out_no_mm:
2561	put_task_struct(task);
2562 out_no_task:
2563	return ret;
 
 
2564}
2565
2566static const struct file_operations proc_coredump_filter_operations = {
2567	.read		= proc_coredump_filter_read,
2568	.write		= proc_coredump_filter_write,
2569	.llseek		= generic_file_llseek,
2570};
2571#endif
2572
2573/*
2574 * /proc/self:
2575 */
2576static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2577			      int buflen)
2578{
2579	struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2580	pid_t tgid = task_tgid_nr_ns(current, ns);
2581	char tmp[PROC_NUMBUF];
2582	if (!tgid)
2583		return -ENOENT;
2584	sprintf(tmp, "%d", tgid);
2585	return vfs_readlink(dentry,buffer,buflen,tmp);
2586}
2587
2588static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2589{
2590	struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2591	pid_t tgid = task_tgid_nr_ns(current, ns);
2592	char *name = ERR_PTR(-ENOENT);
2593	if (tgid) {
2594		name = __getname();
2595		if (!name)
2596			name = ERR_PTR(-ENOMEM);
2597		else
2598			sprintf(name, "%d", tgid);
2599	}
2600	nd_set_link(nd, name);
2601	return NULL;
2602}
2603
2604static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2605				void *cookie)
2606{
2607	char *s = nd_get_link(nd);
2608	if (!IS_ERR(s))
2609		__putname(s);
2610}
2611
2612static const struct inode_operations proc_self_inode_operations = {
2613	.readlink	= proc_self_readlink,
2614	.follow_link	= proc_self_follow_link,
2615	.put_link	= proc_self_put_link,
2616};
2617
2618/*
2619 * proc base
2620 *
2621 * These are the directory entries in the root directory of /proc
2622 * that properly belong to the /proc filesystem, as they describe
2623 * describe something that is process related.
2624 */
2625static const struct pid_entry proc_base_stuff[] = {
2626	NOD("self", S_IFLNK|S_IRWXUGO,
2627		&proc_self_inode_operations, NULL, {}),
2628};
2629
2630static struct dentry *proc_base_instantiate(struct inode *dir,
2631	struct dentry *dentry, struct task_struct *task, const void *ptr)
2632{
2633	const struct pid_entry *p = ptr;
2634	struct inode *inode;
2635	struct proc_inode *ei;
2636	struct dentry *error;
2637
2638	/* Allocate the inode */
2639	error = ERR_PTR(-ENOMEM);
2640	inode = new_inode(dir->i_sb);
2641	if (!inode)
2642		goto out;
 
 
 
 
 
 
 
 
 
 
 
2643
2644	/* Initialize the inode */
2645	ei = PROC_I(inode);
2646	inode->i_ino = get_next_ino();
2647	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2648
2649	/*
2650	 * grab the reference to the task.
2651	 */
2652	ei->pid = get_task_pid(task, PIDTYPE_PID);
2653	if (!ei->pid)
2654		goto out_iput;
2655
2656	inode->i_mode = p->mode;
2657	if (S_ISDIR(inode->i_mode))
2658		inode->i_nlink = 2;
2659	if (S_ISLNK(inode->i_mode))
2660		inode->i_size = 64;
2661	if (p->iop)
2662		inode->i_op = p->iop;
2663	if (p->fop)
2664		inode->i_fop = p->fop;
2665	ei->op = p->op;
2666	d_add(dentry, inode);
2667	error = NULL;
2668out:
2669	return error;
2670out_iput:
2671	iput(inode);
2672	goto out;
2673}
 
2674
2675static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
 
 
2676{
2677	struct dentry *error;
2678	struct task_struct *task = get_proc_task(dir);
2679	const struct pid_entry *p, *last;
 
2680
2681	error = ERR_PTR(-ENOENT);
 
 
 
 
 
 
 
 
2682
2683	if (!task)
2684		goto out_no_task;
 
2685
2686	/* Lookup the directory entry */
2687	last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2688	for (p = proc_base_stuff; p <= last; p++) {
2689		if (p->len != dentry->d_name.len)
2690			continue;
2691		if (!memcmp(dentry->d_name.name, p->name, p->len))
2692			break;
2693	}
2694	if (p > last)
2695		goto out;
2696
2697	error = proc_base_instantiate(dir, dentry, task, p);
 
 
 
 
 
 
2698
2699out:
2700	put_task_struct(task);
2701out_no_task:
2702	return error;
2703}
2704
2705static int proc_base_fill_cache(struct file *filp, void *dirent,
2706	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2707{
2708	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2709				proc_base_instantiate, task, p);
2710}
2711
2712#ifdef CONFIG_TASK_IO_ACCOUNTING
2713static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2714{
2715	struct task_io_accounting acct = task->ioac;
2716	unsigned long flags;
2717	int result;
2718
2719	result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2720	if (result)
2721		return result;
 
 
 
 
2722
2723	if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2724		result = -EACCES;
2725		goto out_unlock;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2726	}
 
 
2727
2728	if (whole && lock_task_sighand(task, &flags)) {
2729		struct task_struct *t = task;
 
 
 
2730
2731		task_io_accounting_add(&acct, &task->signal->ioac);
2732		while_each_thread(task, t)
2733			task_io_accounting_add(&acct, &t->ioac);
2734
2735		unlock_task_sighand(task, &flags);
2736	}
2737	result = sprintf(buffer,
2738			"rchar: %llu\n"
2739			"wchar: %llu\n"
2740			"syscr: %llu\n"
2741			"syscw: %llu\n"
2742			"read_bytes: %llu\n"
2743			"write_bytes: %llu\n"
2744			"cancelled_write_bytes: %llu\n",
2745			(unsigned long long)acct.rchar,
2746			(unsigned long long)acct.wchar,
2747			(unsigned long long)acct.syscr,
2748			(unsigned long long)acct.syscw,
2749			(unsigned long long)acct.read_bytes,
2750			(unsigned long long)acct.write_bytes,
2751			(unsigned long long)acct.cancelled_write_bytes);
2752out_unlock:
2753	mutex_unlock(&task->signal->cred_guard_mutex);
2754	return result;
2755}
2756
2757static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2758{
2759	return do_io_accounting(task, buffer, 0);
 
 
2760}
2761
2762static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2763{
2764	return do_io_accounting(task, buffer, 1);
 
 
 
 
2765}
2766#endif /* CONFIG_TASK_IO_ACCOUNTING */
 
 
 
 
 
 
 
 
2767
2768static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2769				struct pid *pid, struct task_struct *task)
2770{
2771	int err = lock_trace(task);
2772	if (!err) {
2773		seq_printf(m, "%08x\n", task->personality);
2774		unlock_trace(task);
2775	}
2776	return err;
2777}
2778
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2779/*
2780 * Thread groups
2781 */
2782static const struct file_operations proc_task_operations;
2783static const struct inode_operations proc_task_inode_operations;
2784
2785static const struct pid_entry tgid_base_stuff[] = {
2786	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2787	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2788	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
 
2789	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2790#ifdef CONFIG_NET
2791	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2792#endif
2793	REG("environ",    S_IRUSR, proc_environ_operations),
2794	INF("auxv",       S_IRUSR, proc_pid_auxv),
2795	ONE("status",     S_IRUGO, proc_pid_status),
2796	ONE("personality", S_IRUGO, proc_pid_personality),
2797	INF("limits",	  S_IRUGO, proc_pid_limits),
2798#ifdef CONFIG_SCHED_DEBUG
2799	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2800#endif
2801#ifdef CONFIG_SCHED_AUTOGROUP
2802	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2803#endif
 
 
 
2804	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2805#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2806	INF("syscall",    S_IRUGO, proc_pid_syscall),
2807#endif
2808	INF("cmdline",    S_IRUGO, proc_pid_cmdline),
2809	ONE("stat",       S_IRUGO, proc_tgid_stat),
2810	ONE("statm",      S_IRUGO, proc_pid_statm),
2811	REG("maps",       S_IRUGO, proc_maps_operations),
2812#ifdef CONFIG_NUMA
2813	REG("numa_maps",  S_IRUGO, proc_numa_maps_operations),
2814#endif
2815	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2816	LNK("cwd",        proc_cwd_link),
2817	LNK("root",       proc_root_link),
2818	LNK("exe",        proc_exe_link),
2819	REG("mounts",     S_IRUGO, proc_mounts_operations),
2820	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2821	REG("mountstats", S_IRUSR, proc_mountstats_operations),
2822#ifdef CONFIG_PROC_PAGE_MONITOR
2823	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2824	REG("smaps",      S_IRUGO, proc_smaps_operations),
2825	REG("pagemap",    S_IRUGO, proc_pagemap_operations),
 
2826#endif
2827#ifdef CONFIG_SECURITY
2828	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2829#endif
2830#ifdef CONFIG_KALLSYMS
2831	INF("wchan",      S_IRUGO, proc_pid_wchan),
2832#endif
2833#ifdef CONFIG_STACKTRACE
2834	ONE("stack",      S_IRUGO, proc_pid_stack),
2835#endif
2836#ifdef CONFIG_SCHEDSTATS
2837	INF("schedstat",  S_IRUGO, proc_pid_schedstat),
2838#endif
2839#ifdef CONFIG_LATENCYTOP
2840	REG("latency",  S_IRUGO, proc_lstats_operations),
2841#endif
2842#ifdef CONFIG_PROC_PID_CPUSET
2843	REG("cpuset",     S_IRUGO, proc_cpuset_operations),
2844#endif
2845#ifdef CONFIG_CGROUPS
2846	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
 
 
 
2847#endif
2848	INF("oom_score",  S_IRUGO, proc_oom_score),
2849	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2850	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2851#ifdef CONFIG_AUDITSYSCALL
2852	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2853	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2854#endif
2855#ifdef CONFIG_FAULT_INJECTION
2856	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
 
2857#endif
2858#ifdef CONFIG_ELF_CORE
2859	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2860#endif
2861#ifdef CONFIG_TASK_IO_ACCOUNTING
2862	INF("io",	S_IRUSR, proc_tgid_io_accounting),
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2863#endif
2864#ifdef CONFIG_HARDWALL
2865	INF("hardwall",   S_IRUGO, proc_pid_hardwall),
 
 
 
 
2866#endif
2867};
2868
2869static int proc_tgid_base_readdir(struct file * filp,
2870			     void * dirent, filldir_t filldir)
2871{
2872	return proc_pident_readdir(filp,dirent,filldir,
2873				   tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2874}
2875
2876static const struct file_operations proc_tgid_base_operations = {
2877	.read		= generic_read_dir,
2878	.readdir	= proc_tgid_base_readdir,
2879	.llseek		= default_llseek,
2880};
2881
2882static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
 
 
 
 
 
 
 
 
 
2883	return proc_pident_lookup(dir, dentry,
2884				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
 
2885}
2886
2887static const struct inode_operations proc_tgid_base_inode_operations = {
2888	.lookup		= proc_tgid_base_lookup,
2889	.getattr	= pid_getattr,
2890	.setattr	= proc_setattr,
 
2891};
2892
2893static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2894{
2895	struct dentry *dentry, *leader, *dir;
2896	char buf[PROC_NUMBUF];
2897	struct qstr name;
2898
2899	name.name = buf;
2900	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2901	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2902	if (dentry) {
2903		shrink_dcache_parent(dentry);
2904		d_drop(dentry);
2905		dput(dentry);
2906	}
2907
2908	name.name = buf;
2909	name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2910	leader = d_hash_and_lookup(mnt->mnt_root, &name);
2911	if (!leader)
2912		goto out;
2913
2914	name.name = "task";
2915	name.len = strlen(name.name);
2916	dir = d_hash_and_lookup(leader, &name);
2917	if (!dir)
2918		goto out_put_leader;
2919
2920	name.name = buf;
2921	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2922	dentry = d_hash_and_lookup(dir, &name);
2923	if (dentry) {
2924		shrink_dcache_parent(dentry);
2925		d_drop(dentry);
2926		dput(dentry);
2927	}
2928
2929	dput(dir);
2930out_put_leader:
2931	dput(leader);
2932out:
2933	return;
2934}
2935
2936/**
2937 * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
2938 * @task: task that should be flushed.
2939 *
2940 * When flushing dentries from proc, one needs to flush them from global
2941 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2942 * in. This call is supposed to do all of this job.
2943 *
2944 * Looks in the dcache for
2945 * /proc/@pid
2946 * /proc/@tgid/task/@pid
2947 * if either directory is present flushes it and all of it'ts children
2948 * from the dcache.
2949 *
2950 * It is safe and reasonable to cache /proc entries for a task until
2951 * that task exits.  After that they just clog up the dcache with
2952 * useless entries, possibly causing useful dcache entries to be
2953 * flushed instead.  This routine is proved to flush those useless
2954 * dcache entries at process exit time.
2955 *
2956 * NOTE: This routine is just an optimization so it does not guarantee
2957 *       that no dcache entries will exist at process exit time it
2958 *       just makes it very unlikely that any will persist.
2959 */
2960
2961void proc_flush_task(struct task_struct *task)
2962{
2963	int i;
2964	struct pid *pid, *tgid;
2965	struct upid *upid;
2966
2967	pid = task_pid(task);
2968	tgid = task_tgid(task);
2969
2970	for (i = 0; i <= pid->level; i++) {
2971		upid = &pid->numbers[i];
2972		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2973					tgid->numbers[i].nr);
2974	}
2975
2976	upid = &pid->numbers[pid->level];
2977	if (upid->nr == 1)
2978		pid_ns_release_proc(upid->ns);
2979}
2980
2981static struct dentry *proc_pid_instantiate(struct inode *dir,
2982					   struct dentry * dentry,
2983					   struct task_struct *task, const void *ptr)
2984{
2985	struct dentry *error = ERR_PTR(-ENOENT);
2986	struct inode *inode;
2987
2988	inode = proc_pid_make_inode(dir->i_sb, task);
 
2989	if (!inode)
2990		goto out;
2991
2992	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2993	inode->i_op = &proc_tgid_base_inode_operations;
2994	inode->i_fop = &proc_tgid_base_operations;
2995	inode->i_flags|=S_IMMUTABLE;
2996
2997	inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2998		ARRAY_SIZE(tgid_base_stuff));
2999
3000	d_set_d_op(dentry, &pid_dentry_operations);
3001
3002	d_add(dentry, inode);
3003	/* Close the race of the process dying before we return the dentry */
3004	if (pid_revalidate(dentry, NULL))
3005		error = NULL;
3006out:
3007	return error;
3008}
3009
3010struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3011{
3012	struct dentry *result;
3013	struct task_struct *task;
3014	unsigned tgid;
 
3015	struct pid_namespace *ns;
 
3016
3017	result = proc_base_lookup(dir, dentry);
3018	if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3019		goto out;
3020
3021	tgid = name_to_int(dentry);
3022	if (tgid == ~0U)
3023		goto out;
3024
3025	ns = dentry->d_sb->s_fs_info;
 
3026	rcu_read_lock();
3027	task = find_task_by_pid_ns(tgid, ns);
3028	if (task)
3029		get_task_struct(task);
3030	rcu_read_unlock();
3031	if (!task)
3032		goto out;
3033
3034	result = proc_pid_instantiate(dir, dentry, task, NULL);
 
 
 
 
 
 
 
3035	put_task_struct(task);
3036out:
3037	return result;
3038}
3039
3040/*
3041 * Find the first task with tgid >= tgid
3042 *
3043 */
3044struct tgid_iter {
3045	unsigned int tgid;
3046	struct task_struct *task;
3047};
3048static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3049{
3050	struct pid *pid;
3051
3052	if (iter.task)
3053		put_task_struct(iter.task);
3054	rcu_read_lock();
3055retry:
3056	iter.task = NULL;
3057	pid = find_ge_pid(iter.tgid, ns);
3058	if (pid) {
3059		iter.tgid = pid_nr_ns(pid, ns);
3060		iter.task = pid_task(pid, PIDTYPE_PID);
3061		/* What we to know is if the pid we have find is the
3062		 * pid of a thread_group_leader.  Testing for task
3063		 * being a thread_group_leader is the obvious thing
3064		 * todo but there is a window when it fails, due to
3065		 * the pid transfer logic in de_thread.
3066		 *
3067		 * So we perform the straight forward test of seeing
3068		 * if the pid we have found is the pid of a thread
3069		 * group leader, and don't worry if the task we have
3070		 * found doesn't happen to be a thread group leader.
3071		 * As we don't care in the case of readdir.
3072		 */
3073		if (!iter.task || !has_group_leader_pid(iter.task)) {
3074			iter.tgid += 1;
3075			goto retry;
3076		}
3077		get_task_struct(iter.task);
3078	}
3079	rcu_read_unlock();
3080	return iter;
3081}
3082
3083#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3084
3085static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3086	struct tgid_iter iter)
3087{
3088	char name[PROC_NUMBUF];
3089	int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3090	return proc_fill_cache(filp, dirent, filldir, name, len,
3091				proc_pid_instantiate, iter.task, NULL);
3092}
3093
3094/* for the /proc/ directory itself, after non-process stuff has been done */
3095int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3096{
3097	unsigned int nr;
3098	struct task_struct *reaper;
3099	struct tgid_iter iter;
3100	struct pid_namespace *ns;
3101
3102	if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3103		goto out_no_task;
3104	nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3105
3106	reaper = get_proc_task(filp->f_path.dentry->d_inode);
3107	if (!reaper)
3108		goto out_no_task;
3109
3110	for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3111		const struct pid_entry *p = &proc_base_stuff[nr];
3112		if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3113			goto out;
 
3114	}
3115
3116	ns = filp->f_dentry->d_sb->s_fs_info;
 
 
 
 
 
3117	iter.task = NULL;
3118	iter.tgid = filp->f_pos - TGID_OFFSET;
3119	for (iter = next_tgid(ns, iter);
3120	     iter.task;
3121	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3122		filp->f_pos = iter.tgid + TGID_OFFSET;
3123		if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
 
 
 
 
 
 
 
 
 
3124			put_task_struct(iter.task);
3125			goto out;
3126		}
3127	}
3128	filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3129out:
3130	put_task_struct(reaper);
3131out_no_task:
3132	return 0;
3133}
3134
3135/*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3136 * Tasks
3137 */
3138static const struct pid_entry tid_base_stuff[] = {
3139	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3140	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3141	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
 
 
 
3142	REG("environ",   S_IRUSR, proc_environ_operations),
3143	INF("auxv",      S_IRUSR, proc_pid_auxv),
3144	ONE("status",    S_IRUGO, proc_pid_status),
3145	ONE("personality", S_IRUGO, proc_pid_personality),
3146	INF("limits",	 S_IRUGO, proc_pid_limits),
3147#ifdef CONFIG_SCHED_DEBUG
3148	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3149#endif
3150	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
 
 
3151#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3152	INF("syscall",   S_IRUGO, proc_pid_syscall),
3153#endif
3154	INF("cmdline",   S_IRUGO, proc_pid_cmdline),
3155	ONE("stat",      S_IRUGO, proc_tid_stat),
3156	ONE("statm",     S_IRUGO, proc_pid_statm),
3157	REG("maps",      S_IRUGO, proc_maps_operations),
 
 
 
3158#ifdef CONFIG_NUMA
3159	REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3160#endif
3161	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3162	LNK("cwd",       proc_cwd_link),
3163	LNK("root",      proc_root_link),
3164	LNK("exe",       proc_exe_link),
3165	REG("mounts",    S_IRUGO, proc_mounts_operations),
3166	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3167#ifdef CONFIG_PROC_PAGE_MONITOR
3168	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3169	REG("smaps",     S_IRUGO, proc_smaps_operations),
3170	REG("pagemap",    S_IRUGO, proc_pagemap_operations),
 
3171#endif
3172#ifdef CONFIG_SECURITY
3173	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3174#endif
3175#ifdef CONFIG_KALLSYMS
3176	INF("wchan",     S_IRUGO, proc_pid_wchan),
3177#endif
3178#ifdef CONFIG_STACKTRACE
3179	ONE("stack",      S_IRUGO, proc_pid_stack),
3180#endif
3181#ifdef CONFIG_SCHEDSTATS
3182	INF("schedstat", S_IRUGO, proc_pid_schedstat),
3183#endif
3184#ifdef CONFIG_LATENCYTOP
3185	REG("latency",  S_IRUGO, proc_lstats_operations),
3186#endif
3187#ifdef CONFIG_PROC_PID_CPUSET
3188	REG("cpuset",    S_IRUGO, proc_cpuset_operations),
3189#endif
3190#ifdef CONFIG_CGROUPS
3191	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
 
 
 
3192#endif
3193	INF("oom_score", S_IRUGO, proc_oom_score),
3194	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3195	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3196#ifdef CONFIG_AUDITSYSCALL
3197	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3198	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3199#endif
3200#ifdef CONFIG_FAULT_INJECTION
3201	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
 
3202#endif
3203#ifdef CONFIG_TASK_IO_ACCOUNTING
3204	INF("io",	S_IRUSR, proc_tid_io_accounting),
 
 
 
 
 
 
 
 
 
 
 
 
3205#endif
3206#ifdef CONFIG_HARDWALL
3207	INF("hardwall",   S_IRUGO, proc_pid_hardwall),
 
 
 
 
3208#endif
3209};
3210
3211static int proc_tid_base_readdir(struct file * filp,
3212			     void * dirent, filldir_t filldir)
3213{
3214	return proc_pident_readdir(filp,dirent,filldir,
3215				   tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3216}
3217
3218static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
 
3219	return proc_pident_lookup(dir, dentry,
3220				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
 
3221}
3222
3223static const struct file_operations proc_tid_base_operations = {
3224	.read		= generic_read_dir,
3225	.readdir	= proc_tid_base_readdir,
3226	.llseek		= default_llseek,
3227};
3228
3229static const struct inode_operations proc_tid_base_inode_operations = {
3230	.lookup		= proc_tid_base_lookup,
3231	.getattr	= pid_getattr,
3232	.setattr	= proc_setattr,
3233};
3234
3235static struct dentry *proc_task_instantiate(struct inode *dir,
3236	struct dentry *dentry, struct task_struct *task, const void *ptr)
3237{
3238	struct dentry *error = ERR_PTR(-ENOENT);
3239	struct inode *inode;
3240	inode = proc_pid_make_inode(dir->i_sb, task);
3241
3242	if (!inode)
3243		goto out;
3244	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3245	inode->i_op = &proc_tid_base_inode_operations;
3246	inode->i_fop = &proc_tid_base_operations;
3247	inode->i_flags|=S_IMMUTABLE;
3248
3249	inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3250		ARRAY_SIZE(tid_base_stuff));
3251
3252	d_set_d_op(dentry, &pid_dentry_operations);
3253
3254	d_add(dentry, inode);
3255	/* Close the race of the process dying before we return the dentry */
3256	if (pid_revalidate(dentry, NULL))
3257		error = NULL;
3258out:
3259	return error;
3260}
3261
3262static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3263{
3264	struct dentry *result = ERR_PTR(-ENOENT);
3265	struct task_struct *task;
3266	struct task_struct *leader = get_proc_task(dir);
3267	unsigned tid;
 
3268	struct pid_namespace *ns;
 
3269
3270	if (!leader)
3271		goto out_no_task;
3272
3273	tid = name_to_int(dentry);
3274	if (tid == ~0U)
3275		goto out;
3276
3277	ns = dentry->d_sb->s_fs_info;
 
3278	rcu_read_lock();
3279	task = find_task_by_pid_ns(tid, ns);
3280	if (task)
3281		get_task_struct(task);
3282	rcu_read_unlock();
3283	if (!task)
3284		goto out;
3285	if (!same_thread_group(leader, task))
3286		goto out_drop_task;
3287
3288	result = proc_task_instantiate(dir, dentry, task, NULL);
3289out_drop_task:
3290	put_task_struct(task);
3291out:
3292	put_task_struct(leader);
3293out_no_task:
3294	return result;
3295}
3296
3297/*
3298 * Find the first tid of a thread group to return to user space.
3299 *
3300 * Usually this is just the thread group leader, but if the users
3301 * buffer was too small or there was a seek into the middle of the
3302 * directory we have more work todo.
3303 *
3304 * In the case of a short read we start with find_task_by_pid.
3305 *
3306 * In the case of a seek we start with the leader and walk nr
3307 * threads past it.
3308 */
3309static struct task_struct *first_tid(struct task_struct *leader,
3310		int tid, int nr, struct pid_namespace *ns)
3311{
3312	struct task_struct *pos;
 
 
 
 
3313
3314	rcu_read_lock();
3315	/* Attempt to start with the pid of a thread */
3316	if (tid && (nr > 0)) {
 
 
 
 
3317		pos = find_task_by_pid_ns(tid, ns);
3318		if (pos && (pos->group_leader == leader))
3319			goto found;
3320	}
3321
3322	/* If nr exceeds the number of threads there is nothing todo */
3323	pos = NULL;
3324	if (nr && nr >= get_nr_threads(leader))
3325		goto out;
3326
3327	/* If we haven't found our starting place yet start
3328	 * with the leader and walk nr threads forward.
3329	 */
3330	for (pos = leader; nr > 0; --nr) {
3331		pos = next_thread(pos);
3332		if (pos == leader) {
3333			pos = NULL;
3334			goto out;
3335		}
3336	}
 
 
 
3337found:
3338	get_task_struct(pos);
3339out:
3340	rcu_read_unlock();
3341	return pos;
3342}
3343
3344/*
3345 * Find the next thread in the thread list.
3346 * Return NULL if there is an error or no next thread.
3347 *
3348 * The reference to the input task_struct is released.
3349 */
3350static struct task_struct *next_tid(struct task_struct *start)
3351{
3352	struct task_struct *pos = NULL;
3353	rcu_read_lock();
3354	if (pid_alive(start)) {
3355		pos = next_thread(start);
3356		if (thread_group_leader(pos))
3357			pos = NULL;
3358		else
3359			get_task_struct(pos);
3360	}
3361	rcu_read_unlock();
3362	put_task_struct(start);
3363	return pos;
3364}
3365
3366static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3367	struct task_struct *task, int tid)
3368{
3369	char name[PROC_NUMBUF];
3370	int len = snprintf(name, sizeof(name), "%d", tid);
3371	return proc_fill_cache(filp, dirent, filldir, name, len,
3372				proc_task_instantiate, task, NULL);
3373}
3374
3375/* for the /proc/TGID/task/ directories */
3376static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3377{
3378	struct dentry *dentry = filp->f_path.dentry;
3379	struct inode *inode = dentry->d_inode;
3380	struct task_struct *leader = NULL;
3381	struct task_struct *task;
3382	int retval = -ENOENT;
3383	ino_t ino;
3384	int tid;
3385	struct pid_namespace *ns;
 
3386
3387	task = get_proc_task(inode);
3388	if (!task)
3389		goto out_no_task;
3390	rcu_read_lock();
3391	if (pid_alive(task)) {
3392		leader = task->group_leader;
3393		get_task_struct(leader);
3394	}
3395	rcu_read_unlock();
3396	put_task_struct(task);
3397	if (!leader)
3398		goto out_no_task;
3399	retval = 0;
3400
3401	switch ((unsigned long)filp->f_pos) {
3402	case 0:
3403		ino = inode->i_ino;
3404		if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3405			goto out;
3406		filp->f_pos++;
3407		/* fall through */
3408	case 1:
3409		ino = parent_ino(dentry);
3410		if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3411			goto out;
3412		filp->f_pos++;
3413		/* fall through */
3414	}
3415
3416	/* f_version caches the tgid value that the last readdir call couldn't
3417	 * return. lseek aka telldir automagically resets f_version to 0.
3418	 */
3419	ns = filp->f_dentry->d_sb->s_fs_info;
3420	tid = (int)filp->f_version;
3421	filp->f_version = 0;
3422	for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3423	     task;
3424	     task = next_tid(task), filp->f_pos++) {
 
 
 
3425		tid = task_pid_nr_ns(task, ns);
3426		if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
 
 
 
 
3427			/* returning this tgid failed, save it as the first
3428			 * pid for the next readir call */
3429			filp->f_version = (u64)tid;
3430			put_task_struct(task);
3431			break;
3432		}
3433	}
3434out:
3435	put_task_struct(leader);
3436out_no_task:
3437	return retval;
3438}
3439
3440static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 
 
3441{
3442	struct inode *inode = dentry->d_inode;
3443	struct task_struct *p = get_proc_task(inode);
3444	generic_fillattr(inode, stat);
3445
3446	if (p) {
3447		stat->nlink += get_nr_threads(p);
3448		put_task_struct(p);
3449	}
3450
3451	return 0;
3452}
3453
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3454static const struct inode_operations proc_task_inode_operations = {
3455	.lookup		= proc_task_lookup,
3456	.getattr	= proc_task_getattr,
3457	.setattr	= proc_setattr,
 
3458};
3459
3460static const struct file_operations proc_task_operations = {
3461	.read		= generic_read_dir,
3462	.readdir	= proc_task_readdir,
3463	.llseek		= default_llseek,
3464};
v6.13.7
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 *  linux/fs/proc/base.c
   4 *
   5 *  Copyright (C) 1991, 1992 Linus Torvalds
   6 *
   7 *  proc base directory handling functions
   8 *
   9 *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
  10 *  Instead of using magical inumbers to determine the kind of object
  11 *  we allocate and fill in-core inodes upon lookup. They don't even
  12 *  go into icache. We cache the reference to task_struct upon lookup too.
  13 *  Eventually it should become a filesystem in its own. We don't use the
  14 *  rest of procfs anymore.
  15 *
  16 *
  17 *  Changelog:
  18 *  17-Jan-2005
  19 *  Allan Bezerra
  20 *  Bruna Moreira <bruna.moreira@indt.org.br>
  21 *  Edjard Mota <edjard.mota@indt.org.br>
  22 *  Ilias Biris <ilias.biris@indt.org.br>
  23 *  Mauricio Lin <mauricio.lin@indt.org.br>
  24 *
  25 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  26 *
  27 *  A new process specific entry (smaps) included in /proc. It shows the
  28 *  size of rss for each memory area. The maps entry lacks information
  29 *  about physical memory size (rss) for each mapped file, i.e.,
  30 *  rss information for executables and library files.
  31 *  This additional information is useful for any tools that need to know
  32 *  about physical memory consumption for a process specific library.
  33 *
  34 *  Changelog:
  35 *  21-Feb-2005
  36 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  37 *  Pud inclusion in the page table walking.
  38 *
  39 *  ChangeLog:
  40 *  10-Mar-2005
  41 *  10LE Instituto Nokia de Tecnologia - INdT:
  42 *  A better way to walks through the page table as suggested by Hugh Dickins.
  43 *
  44 *  Simo Piiroinen <simo.piiroinen@nokia.com>:
  45 *  Smaps information related to shared, private, clean and dirty pages.
  46 *
  47 *  Paul Mundt <paul.mundt@nokia.com>:
  48 *  Overall revision about smaps.
  49 */
  50
  51#include <linux/uaccess.h>
  52
  53#include <linux/errno.h>
  54#include <linux/time.h>
  55#include <linux/proc_fs.h>
  56#include <linux/stat.h>
  57#include <linux/task_io_accounting_ops.h>
  58#include <linux/init.h>
  59#include <linux/capability.h>
  60#include <linux/file.h>
  61#include <linux/generic-radix-tree.h>
  62#include <linux/string.h>
  63#include <linux/seq_file.h>
  64#include <linux/namei.h>
  65#include <linux/mnt_namespace.h>
  66#include <linux/mm.h>
  67#include <linux/swap.h>
  68#include <linux/rcupdate.h>
  69#include <linux/kallsyms.h>
  70#include <linux/stacktrace.h>
  71#include <linux/resource.h>
  72#include <linux/module.h>
  73#include <linux/mount.h>
  74#include <linux/security.h>
  75#include <linux/ptrace.h>
  76#include <linux/printk.h>
  77#include <linux/cache.h>
  78#include <linux/cgroup.h>
  79#include <linux/cpuset.h>
  80#include <linux/audit.h>
  81#include <linux/poll.h>
  82#include <linux/nsproxy.h>
  83#include <linux/oom.h>
  84#include <linux/elf.h>
  85#include <linux/pid_namespace.h>
  86#include <linux/user_namespace.h>
  87#include <linux/fs_parser.h>
  88#include <linux/fs_struct.h>
  89#include <linux/slab.h>
  90#include <linux/sched/autogroup.h>
  91#include <linux/sched/mm.h>
  92#include <linux/sched/coredump.h>
  93#include <linux/sched/debug.h>
  94#include <linux/sched/stat.h>
  95#include <linux/posix-timers.h>
  96#include <linux/time_namespace.h>
  97#include <linux/resctrl.h>
  98#include <linux/cn_proc.h>
  99#include <linux/ksm.h>
 100#include <uapi/linux/lsm.h>
 101#include <trace/events/oom.h>
 102#include "internal.h"
 103#include "fd.h"
 104
 105#include "../../lib/kstrtox.h"
 106
 107/* NOTE:
 108 *	Implementing inode permission operations in /proc is almost
 109 *	certainly an error.  Permission checks need to happen during
 110 *	each system call not at open time.  The reason is that most of
 111 *	what we wish to check for permissions in /proc varies at runtime.
 112 *
 113 *	The classic example of a problem is opening file descriptors
 114 *	in /proc for a task before it execs a suid executable.
 115 */
 116
 117static u8 nlink_tid __ro_after_init;
 118static u8 nlink_tgid __ro_after_init;
 119
 120enum proc_mem_force {
 121	PROC_MEM_FORCE_ALWAYS,
 122	PROC_MEM_FORCE_PTRACE,
 123	PROC_MEM_FORCE_NEVER
 124};
 125
 126static enum proc_mem_force proc_mem_force_override __ro_after_init =
 127	IS_ENABLED(CONFIG_PROC_MEM_NO_FORCE) ? PROC_MEM_FORCE_NEVER :
 128	IS_ENABLED(CONFIG_PROC_MEM_FORCE_PTRACE) ? PROC_MEM_FORCE_PTRACE :
 129	PROC_MEM_FORCE_ALWAYS;
 130
 131static const struct constant_table proc_mem_force_table[] __initconst = {
 132	{ "always", PROC_MEM_FORCE_ALWAYS },
 133	{ "ptrace", PROC_MEM_FORCE_PTRACE },
 134	{ "never", PROC_MEM_FORCE_NEVER },
 135	{ }
 136};
 137
 138static int __init early_proc_mem_force_override(char *buf)
 139{
 140	if (!buf)
 141		return -EINVAL;
 142
 143	/*
 144	 * lookup_constant() defaults to proc_mem_force_override to preseve
 145	 * the initial Kconfig choice in case an invalid param gets passed.
 146	 */
 147	proc_mem_force_override = lookup_constant(proc_mem_force_table,
 148						  buf, proc_mem_force_override);
 149
 150	return 0;
 151}
 152early_param("proc_mem.force_override", early_proc_mem_force_override);
 153
 154struct pid_entry {
 155	const char *name;
 156	unsigned int len;
 157	umode_t mode;
 158	const struct inode_operations *iop;
 159	const struct file_operations *fop;
 160	union proc_op op;
 161};
 162
 163#define NOD(NAME, MODE, IOP, FOP, OP) {			\
 164	.name = (NAME),					\
 165	.len  = sizeof(NAME) - 1,			\
 166	.mode = MODE,					\
 167	.iop  = IOP,					\
 168	.fop  = FOP,					\
 169	.op   = OP,					\
 170}
 171
 172#define DIR(NAME, MODE, iops, fops)	\
 173	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
 174#define LNK(NAME, get_link)					\
 175	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
 176		&proc_pid_link_inode_operations, NULL,		\
 177		{ .proc_get_link = get_link } )
 178#define REG(NAME, MODE, fops)				\
 179	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
 
 
 
 
 180#define ONE(NAME, MODE, show)				\
 181	NOD(NAME, (S_IFREG|(MODE)),			\
 182		NULL, &proc_single_file_operations,	\
 183		{ .proc_show = show } )
 184#define ATTR(LSMID, NAME, MODE)				\
 185	NOD(NAME, (S_IFREG|(MODE)),			\
 186		NULL, &proc_pid_attr_operations,	\
 187		{ .lsmid = LSMID })
 188
 189/*
 190 * Count the number of hardlinks for the pid_entry table, excluding the .
 191 * and .. links.
 192 */
 193static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
 194	unsigned int n)
 195{
 196	unsigned int i;
 197	unsigned int count;
 198
 199	count = 2;
 200	for (i = 0; i < n; ++i) {
 201		if (S_ISDIR(entries[i].mode))
 202			++count;
 203	}
 204
 205	return count;
 206}
 207
 208static int get_task_root(struct task_struct *task, struct path *root)
 209{
 210	int result = -ENOENT;
 211
 212	task_lock(task);
 213	if (task->fs) {
 214		get_fs_root(task->fs, root);
 215		result = 0;
 216	}
 217	task_unlock(task);
 218	return result;
 219}
 220
 221static int proc_cwd_link(struct dentry *dentry, struct path *path)
 222{
 223	struct task_struct *task = get_proc_task(d_inode(dentry));
 224	int result = -ENOENT;
 225
 226	if (task) {
 227		task_lock(task);
 228		if (task->fs) {
 229			get_fs_pwd(task->fs, path);
 230			result = 0;
 231		}
 232		task_unlock(task);
 233		put_task_struct(task);
 234	}
 235	return result;
 236}
 237
 238static int proc_root_link(struct dentry *dentry, struct path *path)
 239{
 240	struct task_struct *task = get_proc_task(d_inode(dentry));
 241	int result = -ENOENT;
 242
 243	if (task) {
 244		result = get_task_root(task, path);
 245		put_task_struct(task);
 246	}
 247	return result;
 248}
 249
 250/*
 251 * If the user used setproctitle(), we just get the string from
 252 * user space at arg_start, and limit it to a maximum of one page.
 253 */
 254static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
 255				size_t count, unsigned long pos,
 256				unsigned long arg_start)
 257{
 258	char *page;
 259	int ret, got;
 260
 261	if (pos >= PAGE_SIZE)
 262		return 0;
 
 263
 264	page = (char *)__get_free_page(GFP_KERNEL);
 265	if (!page)
 266		return -ENOMEM;
 
 
 
 267
 268	ret = 0;
 269	got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
 270	if (got > 0) {
 271		int len = strnlen(page, got);
 272
 273		/* Include the NUL character if it was found */
 274		if (len < got)
 275			len++;
 276
 277		if (len > pos) {
 278			len -= pos;
 279			if (len > count)
 280				len = count;
 281			len -= copy_to_user(buf, page+pos, len);
 282			if (!len)
 283				len = -EFAULT;
 284			ret = len;
 285		}
 286	}
 287	free_page((unsigned long)page);
 288	return ret;
 
 
 
 
 289}
 290
 291static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
 292			      size_t count, loff_t *ppos)
 
 
 
 293{
 294	unsigned long arg_start, arg_end, env_start, env_end;
 295	unsigned long pos, len;
 296	char *page, c;
 297
 298	/* Check if process spawned far enough to have cmdline. */
 299	if (!mm->env_end)
 300		return 0;
 301
 302	spin_lock(&mm->arg_lock);
 303	arg_start = mm->arg_start;
 304	arg_end = mm->arg_end;
 305	env_start = mm->env_start;
 306	env_end = mm->env_end;
 307	spin_unlock(&mm->arg_lock);
 308
 309	if (arg_start >= arg_end)
 310		return 0;
 311
 312	/*
 313	 * We allow setproctitle() to overwrite the argument
 314	 * strings, and overflow past the original end. But
 315	 * only when it overflows into the environment area.
 316	 */
 317	if (env_start != arg_end || env_end < env_start)
 318		env_start = env_end = arg_end;
 319	len = env_end - arg_start;
 320
 321	/* We're not going to care if "*ppos" has high bits set */
 322	pos = *ppos;
 323	if (pos >= len)
 324		return 0;
 325	if (count > len - pos)
 326		count = len - pos;
 327	if (!count)
 328		return 0;
 329
 330	/*
 331	 * Magical special case: if the argv[] end byte is not
 332	 * zero, the user has overwritten it with setproctitle(3).
 333	 *
 334	 * Possible future enhancement: do this only once when
 335	 * pos is 0, and set a flag in the 'struct file'.
 336	 */
 337	if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
 338		return get_mm_proctitle(mm, buf, count, pos, arg_start);
 339
 340	/*
 341	 * For the non-setproctitle() case we limit things strictly
 342	 * to the [arg_start, arg_end[ range.
 343	 */
 344	pos += arg_start;
 345	if (pos < arg_start || pos >= arg_end)
 346		return 0;
 347	if (count > arg_end - pos)
 348		count = arg_end - pos;
 349
 350	page = (char *)__get_free_page(GFP_KERNEL);
 351	if (!page)
 352		return -ENOMEM;
 
 353
 354	len = 0;
 355	while (count) {
 356		int got;
 357		size_t size = min_t(size_t, PAGE_SIZE, count);
 358
 359		got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
 360		if (got <= 0)
 361			break;
 362		got -= copy_to_user(buf, page, got);
 363		if (unlikely(!got)) {
 364			if (!len)
 365				len = -EFAULT;
 366			break;
 367		}
 368		pos += got;
 369		buf += got;
 370		len += got;
 371		count -= got;
 372	}
 
 373
 374	free_page((unsigned long)page);
 375	return len;
 376}
 377
 378static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
 379				size_t count, loff_t *pos)
 380{
 381	struct mm_struct *mm;
 382	ssize_t ret;
 383
 384	mm = get_task_mm(tsk);
 385	if (!mm)
 386		return 0;
 
 
 387
 388	ret = get_mm_cmdline(mm, buf, count, pos);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 389	mmput(mm);
 390	return ret;
 
 391}
 392
 393static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 394				     size_t count, loff_t *pos)
 395{
 396	struct task_struct *tsk;
 397	ssize_t ret;
 398
 399	BUG_ON(*pos < 0);
 400
 401	tsk = get_proc_task(file_inode(file));
 402	if (!tsk)
 403		return -ESRCH;
 404	ret = get_task_cmdline(tsk, buf, count, pos);
 405	put_task_struct(tsk);
 406	if (ret > 0)
 407		*pos += ret;
 408	return ret;
 
 409}
 410
 411static const struct file_operations proc_pid_cmdline_ops = {
 412	.read	= proc_pid_cmdline_read,
 413	.llseek	= generic_file_llseek,
 414};
 415
 416#ifdef CONFIG_KALLSYMS
 417/*
 418 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 419 * Returns the resolved symbol.  If that fails, simply return the address.
 420 */
 421static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
 422			  struct pid *pid, struct task_struct *task)
 423{
 424	unsigned long wchan;
 425	char symname[KSYM_NAME_LEN];
 426
 427	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
 428		goto print0;
 429
 430	wchan = get_wchan(task);
 431	if (wchan && !lookup_symbol_name(wchan, symname)) {
 432		seq_puts(m, symname);
 433		return 0;
 434	}
 435
 436print0:
 437	seq_putc(m, '0');
 438	return 0;
 
 
 
 
 439}
 440#endif /* CONFIG_KALLSYMS */
 441
 442static int lock_trace(struct task_struct *task)
 443{
 444	int err = down_read_killable(&task->signal->exec_update_lock);
 445	if (err)
 446		return err;
 447	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
 448		up_read(&task->signal->exec_update_lock);
 449		return -EPERM;
 450	}
 451	return 0;
 452}
 453
 454static void unlock_trace(struct task_struct *task)
 455{
 456	up_read(&task->signal->exec_update_lock);
 457}
 458
 459#ifdef CONFIG_STACKTRACE
 460
 461#define MAX_STACK_TRACE_DEPTH	64
 462
 463static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
 464			  struct pid *pid, struct task_struct *task)
 465{
 
 466	unsigned long *entries;
 467	int err;
 
 468
 469	/*
 470	 * The ability to racily run the kernel stack unwinder on a running task
 471	 * and then observe the unwinder output is scary; while it is useful for
 472	 * debugging kernel issues, it can also allow an attacker to leak kernel
 473	 * stack contents.
 474	 * Doing this in a manner that is at least safe from races would require
 475	 * some work to ensure that the remote task can not be scheduled; and
 476	 * even then, this would still expose the unwinder as local attack
 477	 * surface.
 478	 * Therefore, this interface is restricted to root.
 479	 */
 480	if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
 481		return -EACCES;
 482
 483	entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
 484				GFP_KERNEL);
 485	if (!entries)
 486		return -ENOMEM;
 487
 
 
 
 
 
 488	err = lock_trace(task);
 489	if (!err) {
 490		unsigned int i, nr_entries;
 491
 492		nr_entries = stack_trace_save_tsk(task, entries,
 493						  MAX_STACK_TRACE_DEPTH, 0);
 494
 495		for (i = 0; i < nr_entries; i++) {
 496			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
 
 497		}
 498
 499		unlock_trace(task);
 500	}
 501	kfree(entries);
 502
 503	return err;
 504}
 505#endif
 506
 507#ifdef CONFIG_SCHED_INFO
 508/*
 509 * Provides /proc/PID/schedstat
 510 */
 511static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
 512			      struct pid *pid, struct task_struct *task)
 513{
 514	if (unlikely(!sched_info_on()))
 515		seq_puts(m, "0 0 0\n");
 516	else
 517		seq_printf(m, "%llu %llu %lu\n",
 518		   (unsigned long long)task->se.sum_exec_runtime,
 519		   (unsigned long long)task->sched_info.run_delay,
 520		   task->sched_info.pcount);
 521
 522	return 0;
 523}
 524#endif
 525
 526#ifdef CONFIG_LATENCYTOP
 527static int lstats_show_proc(struct seq_file *m, void *v)
 528{
 529	int i;
 530	struct inode *inode = m->private;
 531	struct task_struct *task = get_proc_task(inode);
 532
 533	if (!task)
 534		return -ESRCH;
 535	seq_puts(m, "Latency Top version : v0.1\n");
 536	for (i = 0; i < LT_SAVECOUNT; i++) {
 537		struct latency_record *lr = &task->latency_record[i];
 538		if (lr->backtrace[0]) {
 539			int q;
 540			seq_printf(m, "%i %li %li",
 541				   lr->count, lr->time, lr->max);
 542			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
 543				unsigned long bt = lr->backtrace[q];
 544
 545				if (!bt)
 546					break;
 
 
 547				seq_printf(m, " %ps", (void *)bt);
 548			}
 549			seq_putc(m, '\n');
 550		}
 551
 552	}
 553	put_task_struct(task);
 554	return 0;
 555}
 556
 557static int lstats_open(struct inode *inode, struct file *file)
 558{
 559	return single_open(file, lstats_show_proc, inode);
 560}
 561
 562static ssize_t lstats_write(struct file *file, const char __user *buf,
 563			    size_t count, loff_t *offs)
 564{
 565	struct task_struct *task = get_proc_task(file_inode(file));
 566
 567	if (!task)
 568		return -ESRCH;
 569	clear_tsk_latency_tracing(task);
 570	put_task_struct(task);
 571
 572	return count;
 573}
 574
 575static const struct file_operations proc_lstats_operations = {
 576	.open		= lstats_open,
 577	.read		= seq_read,
 578	.write		= lstats_write,
 579	.llseek		= seq_lseek,
 580	.release	= single_release,
 581};
 582
 583#endif
 584
 585static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 586			  struct pid *pid, struct task_struct *task)
 587{
 588	unsigned long totalpages = totalram_pages() + total_swap_pages;
 589	unsigned long points = 0;
 590	long badness;
 591
 592	badness = oom_badness(task, totalpages);
 593	/*
 594	 * Special case OOM_SCORE_ADJ_MIN for all others scale the
 595	 * badness value into [0, 2000] range which we have been
 596	 * exporting for a long time so userspace might depend on it.
 597	 */
 598	if (badness != LONG_MIN)
 599		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
 600
 601	seq_printf(m, "%lu\n", points);
 602
 603	return 0;
 
 
 
 
 
 604}
 605
 606struct limit_names {
 607	const char *name;
 608	const char *unit;
 609};
 610
 611static const struct limit_names lnames[RLIM_NLIMITS] = {
 612	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
 613	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
 614	[RLIMIT_DATA] = {"Max data size", "bytes"},
 615	[RLIMIT_STACK] = {"Max stack size", "bytes"},
 616	[RLIMIT_CORE] = {"Max core file size", "bytes"},
 617	[RLIMIT_RSS] = {"Max resident set", "bytes"},
 618	[RLIMIT_NPROC] = {"Max processes", "processes"},
 619	[RLIMIT_NOFILE] = {"Max open files", "files"},
 620	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
 621	[RLIMIT_AS] = {"Max address space", "bytes"},
 622	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
 623	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
 624	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
 625	[RLIMIT_NICE] = {"Max nice priority", NULL},
 626	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
 627	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
 628};
 629
 630/* Display limits for a process */
 631static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
 632			   struct pid *pid, struct task_struct *task)
 633{
 634	unsigned int i;
 
 635	unsigned long flags;
 
 636
 637	struct rlimit rlim[RLIM_NLIMITS];
 638
 639	if (!lock_task_sighand(task, &flags))
 640		return 0;
 641	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
 642	unlock_task_sighand(task, &flags);
 643
 644	/*
 645	 * print the file header
 646	 */
 647	seq_puts(m, "Limit                     "
 648		"Soft Limit           "
 649		"Hard Limit           "
 650		"Units     \n");
 651
 652	for (i = 0; i < RLIM_NLIMITS; i++) {
 653		if (rlim[i].rlim_cur == RLIM_INFINITY)
 654			seq_printf(m, "%-25s %-20s ",
 655				   lnames[i].name, "unlimited");
 656		else
 657			seq_printf(m, "%-25s %-20lu ",
 658				   lnames[i].name, rlim[i].rlim_cur);
 659
 660		if (rlim[i].rlim_max == RLIM_INFINITY)
 661			seq_printf(m, "%-20s ", "unlimited");
 662		else
 663			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
 
 664
 665		if (lnames[i].unit)
 666			seq_printf(m, "%-10s\n", lnames[i].unit);
 
 667		else
 668			seq_putc(m, '\n');
 669	}
 670
 671	return 0;
 672}
 673
 674#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 675static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
 676			    struct pid *pid, struct task_struct *task)
 677{
 678	struct syscall_info info;
 679	u64 *args = &info.data.args[0];
 680	int res;
 681
 682	res = lock_trace(task);
 683	if (res)
 684		return res;
 685
 686	if (task_current_syscall(task, &info))
 687		seq_puts(m, "running\n");
 688	else if (info.data.nr < 0)
 689		seq_printf(m, "%d 0x%llx 0x%llx\n",
 690			   info.data.nr, info.sp, info.data.instruction_pointer);
 691	else
 692		seq_printf(m,
 693		       "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
 694		       info.data.nr,
 695		       args[0], args[1], args[2], args[3], args[4], args[5],
 696		       info.sp, info.data.instruction_pointer);
 697	unlock_trace(task);
 698
 699	return 0;
 700}
 701#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
 702
 703/************************************************************************/
 704/*                       Here the fs part begins                        */
 705/************************************************************************/
 706
 707/* permission checks */
 708static bool proc_fd_access_allowed(struct inode *inode)
 709{
 710	struct task_struct *task;
 711	bool allowed = false;
 712	/* Allow access to a task's file descriptors if it is us or we
 713	 * may use ptrace attach to the process and find out that
 714	 * information.
 715	 */
 716	task = get_proc_task(inode);
 717	if (task) {
 718		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 719		put_task_struct(task);
 720	}
 721	return allowed;
 722}
 723
 724int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
 725		 struct iattr *attr)
 726{
 727	int error;
 728	struct inode *inode = d_inode(dentry);
 729
 730	if (attr->ia_valid & ATTR_MODE)
 731		return -EPERM;
 732
 733	error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
 734	if (error)
 735		return error;
 736
 737	setattr_copy(&nop_mnt_idmap, inode, attr);
 
 
 
 
 
 
 
 
 738	return 0;
 739}
 740
 741/*
 742 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
 743 * or euid/egid (for hide_pid_min=2)?
 744 */
 745static bool has_pid_permissions(struct proc_fs_info *fs_info,
 746				 struct task_struct *task,
 747				 enum proc_hidepid hide_pid_min)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 748{
 749	/*
 750	 * If 'hidpid' mount option is set force a ptrace check,
 751	 * we indicate that we are using a filesystem syscall
 752	 * by passing PTRACE_MODE_READ_FSCREDS
 753	 */
 754	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
 755		return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 
 
 
 756
 757	if (fs_info->hide_pid < hide_pid_min)
 758		return true;
 759	if (in_group_p(fs_info->pid_gid))
 760		return true;
 761	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 762}
 763
 
 
 
 
 
 
 
 
 764
 765static int proc_pid_permission(struct mnt_idmap *idmap,
 766			       struct inode *inode, int mask)
 767{
 768	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
 769	struct task_struct *task;
 770	bool has_perms;
 
 771
 772	task = get_proc_task(inode);
 773	if (!task)
 774		return -ESRCH;
 775	has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
 776	put_task_struct(task);
 777
 778	if (!has_perms) {
 779		if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
 780			/*
 781			 * Let's make getdents(), stat(), and open()
 782			 * consistent with each other.  If a process
 783			 * may not stat() a file, it shouldn't be seen
 784			 * in procfs at all.
 785			 */
 786			return -ENOENT;
 787		}
 788
 789		return -EPERM;
 790	}
 791	return generic_permission(&nop_mnt_idmap, inode, mask);
 792}
 793
 
 794
 
 
 
 
 
 
 
 
 795
 796static const struct inode_operations proc_def_inode_operations = {
 797	.setattr	= proc_setattr,
 
 798};
 799
 800static int proc_single_show(struct seq_file *m, void *v)
 801{
 802	struct inode *inode = m->private;
 803	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
 804	struct pid *pid = proc_pid(inode);
 805	struct task_struct *task;
 806	int ret;
 807
 
 
 808	task = get_pid_task(pid, PIDTYPE_PID);
 809	if (!task)
 810		return -ESRCH;
 811
 812	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
 813
 814	put_task_struct(task);
 815	return ret;
 816}
 817
 818static int proc_single_open(struct inode *inode, struct file *filp)
 819{
 820	return single_open(filp, proc_single_show, inode);
 821}
 822
 823static const struct file_operations proc_single_file_operations = {
 824	.open		= proc_single_open,
 825	.read		= seq_read,
 826	.llseek		= seq_lseek,
 827	.release	= single_release,
 828};
 829
 
 
 
 
 
 
 
 830
 831struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
 
 832{
 833	struct task_struct *task = get_proc_task(inode);
 
 
 
 834	struct mm_struct *mm;
 835
 836	if (!task)
 837		return ERR_PTR(-ESRCH);
 838
 839	mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
 840	put_task_struct(task);
 
 
 841
 
 
 842	if (IS_ERR(mm))
 843		return mm == ERR_PTR(-ESRCH) ? NULL : mm;
 844
 845	/* ensure this mm_struct can't be freed */
 846	mmgrab(mm);
 847	/* but do not pin its memory */
 848	mmput(mm);
 849
 850	return mm;
 851}
 
 
 852
 853static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
 854{
 855	struct mm_struct *mm = proc_mem_open(inode, mode);
 
 
 
 
 856
 857	if (IS_ERR(mm))
 858		return PTR_ERR(mm);
 859
 860	file->private_data = mm;
 861	return 0;
 862}
 863
 864static int mem_open(struct inode *inode, struct file *file)
 865{
 866	if (WARN_ON_ONCE(!(file->f_op->fop_flags & FOP_UNSIGNED_OFFSET)))
 867		return -EINVAL;
 868	return __mem_open(inode, file, PTRACE_MODE_ATTACH);
 869}
 870
 871static bool proc_mem_foll_force(struct file *file, struct mm_struct *mm)
 872{
 873	struct task_struct *task;
 874	bool ptrace_active = false;
 875
 876	switch (proc_mem_force_override) {
 877	case PROC_MEM_FORCE_NEVER:
 878		return false;
 879	case PROC_MEM_FORCE_PTRACE:
 880		task = get_proc_task(file_inode(file));
 881		if (task) {
 882			ptrace_active =	READ_ONCE(task->ptrace) &&
 883					READ_ONCE(task->mm) == mm &&
 884					READ_ONCE(task->parent) == current;
 885			put_task_struct(task);
 886		}
 887		return ptrace_active;
 888	default:
 889		return true;
 
 
 890	}
 
 
 
 
 
 
 
 
 
 
 891}
 892
 893static ssize_t mem_rw(struct file *file, char __user *buf,
 894			size_t count, loff_t *ppos, int write)
 895{
 896	struct mm_struct *mm = file->private_data;
 897	unsigned long addr = *ppos;
 898	ssize_t copied;
 899	char *page;
 900	unsigned int flags;
 
 
 901
 902	if (!mm)
 903		return 0;
 
 904
 905	page = (char *)__get_free_page(GFP_KERNEL);
 
 906	if (!page)
 907		return -ENOMEM;
 908
 909	copied = 0;
 910	if (!mmget_not_zero(mm))
 911		goto free;
 
 912
 913	flags = write ? FOLL_WRITE : 0;
 914	if (proc_mem_foll_force(file, mm))
 915		flags |= FOLL_FORCE;
 916
 
 917	while (count > 0) {
 918		size_t this_len = min_t(size_t, count, PAGE_SIZE);
 919
 920		if (write && copy_from_user(page, buf, this_len)) {
 
 921			copied = -EFAULT;
 922			break;
 923		}
 924
 925		this_len = access_remote_vm(mm, addr, page, this_len, flags);
 926		if (!this_len) {
 927			if (!copied)
 928				copied = -EIO;
 929			break;
 930		}
 931
 932		if (!write && copy_to_user(buf, page, this_len)) {
 933			copied = -EFAULT;
 934			break;
 935		}
 936
 937		buf += this_len;
 938		addr += this_len;
 939		copied += this_len;
 940		count -= this_len;
 941	}
 942	*ppos = addr;
 943
 
 944	mmput(mm);
 945free:
 946	free_page((unsigned long) page);
 
 
 
 947	return copied;
 948}
 949
 950static ssize_t mem_read(struct file *file, char __user *buf,
 951			size_t count, loff_t *ppos)
 952{
 953	return mem_rw(file, buf, count, ppos, 0);
 954}
 955
 956static ssize_t mem_write(struct file *file, const char __user *buf,
 957			 size_t count, loff_t *ppos)
 958{
 959	return mem_rw(file, (char __user*)buf, count, ppos, 1);
 960}
 961
 962loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 963{
 964	switch (orig) {
 965	case 0:
 966		file->f_pos = offset;
 967		break;
 968	case 1:
 969		file->f_pos += offset;
 970		break;
 971	default:
 972		return -EINVAL;
 973	}
 974	force_successful_syscall_return();
 975	return file->f_pos;
 976}
 977
 978static int mem_release(struct inode *inode, struct file *file)
 979{
 980	struct mm_struct *mm = file->private_data;
 981	if (mm)
 982		mmdrop(mm);
 983	return 0;
 984}
 985
 986static const struct file_operations proc_mem_operations = {
 987	.llseek		= mem_lseek,
 988	.read		= mem_read,
 989	.write		= mem_write,
 990	.open		= mem_open,
 991	.release	= mem_release,
 992	.fop_flags	= FOP_UNSIGNED_OFFSET,
 993};
 994
 995static int environ_open(struct inode *inode, struct file *file)
 996{
 997	return __mem_open(inode, file, PTRACE_MODE_READ);
 998}
 999
1000static ssize_t environ_read(struct file *file, char __user *buf,
1001			size_t count, loff_t *ppos)
1002{
 
1003	char *page;
1004	unsigned long src = *ppos;
1005	int ret = 0;
1006	struct mm_struct *mm = file->private_data;
1007	unsigned long env_start, env_end;
1008
1009	/* Ensure the process spawned far enough to have an environment. */
1010	if (!mm || !mm->env_end)
1011		return 0;
1012
1013	page = (char *)__get_free_page(GFP_KERNEL);
 
1014	if (!page)
1015		return -ENOMEM;
1016
1017	ret = 0;
1018	if (!mmget_not_zero(mm))
1019		goto free;
1020
1021	spin_lock(&mm->arg_lock);
1022	env_start = mm->env_start;
1023	env_end = mm->env_end;
1024	spin_unlock(&mm->arg_lock);
1025
 
1026	while (count > 0) {
1027		size_t this_len, max_len;
1028		int retval;
 
1029
1030		if (src >= (env_end - env_start))
1031			break;
1032
1033		this_len = env_end - (env_start + src);
 
1034
1035		max_len = min_t(size_t, PAGE_SIZE, count);
1036		this_len = min(max_len, this_len);
1037
1038		retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
1039
1040		if (retval <= 0) {
1041			ret = retval;
1042			break;
1043		}
1044
1045		if (copy_to_user(buf, page, retval)) {
1046			ret = -EFAULT;
1047			break;
1048		}
1049
1050		ret += retval;
1051		src += retval;
1052		buf += retval;
1053		count -= retval;
1054	}
1055	*ppos = src;
 
1056	mmput(mm);
1057
1058free:
1059	free_page((unsigned long) page);
 
 
 
1060	return ret;
1061}
1062
1063static const struct file_operations proc_environ_operations = {
1064	.open		= environ_open,
1065	.read		= environ_read,
1066	.llseek		= generic_file_llseek,
1067	.release	= mem_release,
1068};
1069
1070static int auxv_open(struct inode *inode, struct file *file)
1071{
1072	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1073}
1074
1075static ssize_t auxv_read(struct file *file, char __user *buf,
1076			size_t count, loff_t *ppos)
1077{
1078	struct mm_struct *mm = file->private_data;
1079	unsigned int nwords = 0;
1080
1081	if (!mm)
1082		return 0;
1083	do {
1084		nwords += 2;
1085	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1086	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1087				       nwords * sizeof(mm->saved_auxv[0]));
1088}
1089
1090static const struct file_operations proc_auxv_operations = {
1091	.open		= auxv_open,
1092	.read		= auxv_read,
1093	.llseek		= generic_file_llseek,
1094	.release	= mem_release,
1095};
1096
1097static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1098			    loff_t *ppos)
1099{
1100	struct task_struct *task = get_proc_task(file_inode(file));
1101	char buffer[PROC_NUMBUF];
1102	int oom_adj = OOM_ADJUST_MIN;
1103	size_t len;
 
 
1104
1105	if (!task)
1106		return -ESRCH;
1107	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1108		oom_adj = OOM_ADJUST_MAX;
1109	else
1110		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1111			  OOM_SCORE_ADJ_MAX;
1112	put_task_struct(task);
1113	if (oom_adj > OOM_ADJUST_MAX)
1114		oom_adj = OOM_ADJUST_MAX;
1115	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1116	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1117}
1118
1119static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1120{
1121	struct mm_struct *mm = NULL;
1122	struct task_struct *task;
1123	int err = 0;
1124
1125	task = get_proc_task(file_inode(file));
1126	if (!task)
1127		return -ESRCH;
1128
1129	mutex_lock(&oom_adj_mutex);
1130	if (legacy) {
1131		if (oom_adj < task->signal->oom_score_adj &&
1132				!capable(CAP_SYS_RESOURCE)) {
1133			err = -EACCES;
1134			goto err_unlock;
1135		}
1136		/*
1137		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1138		 * /proc/pid/oom_score_adj instead.
1139		 */
1140		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1141			  current->comm, task_pid_nr(current), task_pid_nr(task),
1142			  task_pid_nr(task));
1143	} else {
1144		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1145				!capable(CAP_SYS_RESOURCE)) {
1146			err = -EACCES;
1147			goto err_unlock;
1148		}
1149	}
1150
1151	/*
1152	 * Make sure we will check other processes sharing the mm if this is
1153	 * not vfrok which wants its own oom_score_adj.
1154	 * pin the mm so it doesn't go away and get reused after task_unlock
1155	 */
1156	if (!task->vfork_done) {
1157		struct task_struct *p = find_lock_task_mm(task);
1158
1159		if (p) {
1160			if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1161				mm = p->mm;
1162				mmgrab(mm);
1163			}
1164			task_unlock(p);
1165		}
1166	}
1167
1168	task->signal->oom_score_adj = oom_adj;
1169	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1170		task->signal->oom_score_adj_min = (short)oom_adj;
1171	trace_oom_score_adj_update(task);
1172
1173	if (mm) {
1174		struct task_struct *p;
1175
1176		rcu_read_lock();
1177		for_each_process(p) {
1178			if (same_thread_group(task, p))
1179				continue;
1180
1181			/* do not touch kernel threads or the global init */
1182			if (p->flags & PF_KTHREAD || is_global_init(p))
1183				continue;
1184
1185			task_lock(p);
1186			if (!p->vfork_done && process_shares_mm(p, mm)) {
1187				p->signal->oom_score_adj = oom_adj;
1188				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1189					p->signal->oom_score_adj_min = (short)oom_adj;
1190			}
1191			task_unlock(p);
1192		}
1193		rcu_read_unlock();
1194		mmdrop(mm);
1195	}
1196err_unlock:
1197	mutex_unlock(&oom_adj_mutex);
1198	put_task_struct(task);
1199	return err;
1200}
1201
1202/*
1203 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1204 * kernels.  The effective policy is defined by oom_score_adj, which has a
1205 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1206 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1207 * Processes that become oom disabled via oom_adj will still be oom disabled
1208 * with this implementation.
1209 *
1210 * oom_adj cannot be removed since existing userspace binaries use it.
1211 */
1212static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1213			     size_t count, loff_t *ppos)
1214{
1215	char buffer[PROC_NUMBUF] = {};
1216	int oom_adj;
 
 
1217	int err;
1218
 
1219	if (count > sizeof(buffer) - 1)
1220		count = sizeof(buffer) - 1;
1221	if (copy_from_user(buffer, buf, count)) {
1222		err = -EFAULT;
1223		goto out;
1224	}
1225
1226	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1227	if (err)
1228		goto out;
1229	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1230	     oom_adj != OOM_DISABLE) {
1231		err = -EINVAL;
1232		goto out;
1233	}
1234
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1235	/*
1236	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1237	 * value is always attainable.
1238	 */
1239	if (oom_adj == OOM_ADJUST_MAX)
1240		oom_adj = OOM_SCORE_ADJ_MAX;
1241	else
1242		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1243
1244	err = __set_oom_adj(file, oom_adj, true);
 
 
 
 
1245out:
1246	return err < 0 ? err : count;
1247}
1248
1249static const struct file_operations proc_oom_adj_operations = {
1250	.read		= oom_adj_read,
1251	.write		= oom_adj_write,
1252	.llseek		= generic_file_llseek,
1253};
1254
1255static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1256					size_t count, loff_t *ppos)
1257{
1258	struct task_struct *task = get_proc_task(file_inode(file));
1259	char buffer[PROC_NUMBUF];
1260	short oom_score_adj = OOM_SCORE_ADJ_MIN;
 
1261	size_t len;
1262
1263	if (!task)
1264		return -ESRCH;
1265	oom_score_adj = task->signal->oom_score_adj;
 
 
 
1266	put_task_struct(task);
1267	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1268	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1269}
1270
1271static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1272					size_t count, loff_t *ppos)
1273{
1274	char buffer[PROC_NUMBUF] = {};
 
 
1275	int oom_score_adj;
1276	int err;
1277
 
1278	if (count > sizeof(buffer) - 1)
1279		count = sizeof(buffer) - 1;
1280	if (copy_from_user(buffer, buf, count)) {
1281		err = -EFAULT;
1282		goto out;
1283	}
1284
1285	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1286	if (err)
1287		goto out;
1288	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1289			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1290		err = -EINVAL;
1291		goto out;
1292	}
1293
1294	err = __set_oom_adj(file, oom_score_adj, false);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1295out:
1296	return err < 0 ? err : count;
1297}
1298
1299static const struct file_operations proc_oom_score_adj_operations = {
1300	.read		= oom_score_adj_read,
1301	.write		= oom_score_adj_write,
1302	.llseek		= default_llseek,
1303};
1304
1305#ifdef CONFIG_AUDIT
1306#define TMPBUFLEN 11
1307static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1308				  size_t count, loff_t *ppos)
1309{
1310	struct inode * inode = file_inode(file);
1311	struct task_struct *task = get_proc_task(inode);
1312	ssize_t length;
1313	char tmpbuf[TMPBUFLEN];
1314
1315	if (!task)
1316		return -ESRCH;
1317	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1318			   from_kuid(file->f_cred->user_ns,
1319				     audit_get_loginuid(task)));
1320	put_task_struct(task);
1321	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1322}
1323
1324static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1325				   size_t count, loff_t *ppos)
1326{
1327	struct inode * inode = file_inode(file);
 
 
1328	uid_t loginuid;
1329	kuid_t kloginuid;
1330	int rv;
1331
1332	/* Don't let kthreads write their own loginuid */
1333	if (current->flags & PF_KTHREAD)
1334		return -EPERM;
1335
1336	rcu_read_lock();
1337	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1338		rcu_read_unlock();
1339		return -EPERM;
1340	}
1341	rcu_read_unlock();
1342
 
 
 
1343	if (*ppos != 0) {
1344		/* No partial writes. */
1345		return -EINVAL;
1346	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1347
1348	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1349	if (rv < 0)
1350		return rv;
1351
1352	/* is userspace tring to explicitly UNSET the loginuid? */
1353	if (loginuid == AUDIT_UID_UNSET) {
1354		kloginuid = INVALID_UID;
1355	} else {
1356		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1357		if (!uid_valid(kloginuid))
1358			return -EINVAL;
1359	}
1360
1361	rv = audit_set_loginuid(kloginuid);
1362	if (rv < 0)
1363		return rv;
1364	return count;
1365}
1366
1367static const struct file_operations proc_loginuid_operations = {
1368	.read		= proc_loginuid_read,
1369	.write		= proc_loginuid_write,
1370	.llseek		= generic_file_llseek,
1371};
1372
1373static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1374				  size_t count, loff_t *ppos)
1375{
1376	struct inode * inode = file_inode(file);
1377	struct task_struct *task = get_proc_task(inode);
1378	ssize_t length;
1379	char tmpbuf[TMPBUFLEN];
1380
1381	if (!task)
1382		return -ESRCH;
1383	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1384				audit_get_sessionid(task));
1385	put_task_struct(task);
1386	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1387}
1388
1389static const struct file_operations proc_sessionid_operations = {
1390	.read		= proc_sessionid_read,
1391	.llseek		= generic_file_llseek,
1392};
1393#endif
1394
1395#ifdef CONFIG_FAULT_INJECTION
1396static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1397				      size_t count, loff_t *ppos)
1398{
1399	struct task_struct *task = get_proc_task(file_inode(file));
1400	char buffer[PROC_NUMBUF];
1401	size_t len;
1402	int make_it_fail;
1403
1404	if (!task)
1405		return -ESRCH;
1406	make_it_fail = task->make_it_fail;
1407	put_task_struct(task);
1408
1409	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1410
1411	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1412}
1413
1414static ssize_t proc_fault_inject_write(struct file * file,
1415			const char __user * buf, size_t count, loff_t *ppos)
1416{
1417	struct task_struct *task;
1418	char buffer[PROC_NUMBUF] = {};
1419	int make_it_fail;
1420	int rv;
1421
1422	if (!capable(CAP_SYS_RESOURCE))
1423		return -EPERM;
1424
1425	if (count > sizeof(buffer) - 1)
1426		count = sizeof(buffer) - 1;
1427	if (copy_from_user(buffer, buf, count))
1428		return -EFAULT;
1429	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1430	if (rv < 0)
1431		return rv;
1432	if (make_it_fail < 0 || make_it_fail > 1)
1433		return -EINVAL;
1434
1435	task = get_proc_task(file_inode(file));
1436	if (!task)
1437		return -ESRCH;
1438	task->make_it_fail = make_it_fail;
1439	put_task_struct(task);
1440
1441	return count;
1442}
1443
1444static const struct file_operations proc_fault_inject_operations = {
1445	.read		= proc_fault_inject_read,
1446	.write		= proc_fault_inject_write,
1447	.llseek		= generic_file_llseek,
1448};
1449
1450static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1451				   size_t count, loff_t *ppos)
1452{
1453	struct task_struct *task;
1454	int err;
1455	unsigned int n;
1456
1457	err = kstrtouint_from_user(buf, count, 0, &n);
1458	if (err)
1459		return err;
1460
1461	task = get_proc_task(file_inode(file));
1462	if (!task)
1463		return -ESRCH;
1464	task->fail_nth = n;
1465	put_task_struct(task);
1466
1467	return count;
1468}
1469
1470static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1471				  size_t count, loff_t *ppos)
1472{
1473	struct task_struct *task;
1474	char numbuf[PROC_NUMBUF];
1475	ssize_t len;
1476
1477	task = get_proc_task(file_inode(file));
1478	if (!task)
1479		return -ESRCH;
1480	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1481	put_task_struct(task);
1482	return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1483}
1484
1485static const struct file_operations proc_fail_nth_operations = {
1486	.read		= proc_fail_nth_read,
1487	.write		= proc_fail_nth_write,
1488};
1489#endif
1490
1491
1492#ifdef CONFIG_SCHED_DEBUG
1493/*
1494 * Print out various scheduling related per-task fields:
1495 */
1496static int sched_show(struct seq_file *m, void *v)
1497{
1498	struct inode *inode = m->private;
1499	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1500	struct task_struct *p;
1501
1502	p = get_proc_task(inode);
1503	if (!p)
1504		return -ESRCH;
1505	proc_sched_show_task(p, ns, m);
1506
1507	put_task_struct(p);
1508
1509	return 0;
1510}
1511
1512static ssize_t
1513sched_write(struct file *file, const char __user *buf,
1514	    size_t count, loff_t *offset)
1515{
1516	struct inode *inode = file_inode(file);
1517	struct task_struct *p;
1518
1519	p = get_proc_task(inode);
1520	if (!p)
1521		return -ESRCH;
1522	proc_sched_set_task(p);
1523
1524	put_task_struct(p);
1525
1526	return count;
1527}
1528
1529static int sched_open(struct inode *inode, struct file *filp)
1530{
1531	return single_open(filp, sched_show, inode);
1532}
1533
1534static const struct file_operations proc_pid_sched_operations = {
1535	.open		= sched_open,
1536	.read		= seq_read,
1537	.write		= sched_write,
1538	.llseek		= seq_lseek,
1539	.release	= single_release,
1540};
1541
1542#endif
1543
1544#ifdef CONFIG_SCHED_AUTOGROUP
1545/*
1546 * Print out autogroup related information:
1547 */
1548static int sched_autogroup_show(struct seq_file *m, void *v)
1549{
1550	struct inode *inode = m->private;
1551	struct task_struct *p;
1552
1553	p = get_proc_task(inode);
1554	if (!p)
1555		return -ESRCH;
1556	proc_sched_autogroup_show_task(p, m);
1557
1558	put_task_struct(p);
1559
1560	return 0;
1561}
1562
1563static ssize_t
1564sched_autogroup_write(struct file *file, const char __user *buf,
1565	    size_t count, loff_t *offset)
1566{
1567	struct inode *inode = file_inode(file);
1568	struct task_struct *p;
1569	char buffer[PROC_NUMBUF] = {};
1570	int nice;
1571	int err;
1572
 
1573	if (count > sizeof(buffer) - 1)
1574		count = sizeof(buffer) - 1;
1575	if (copy_from_user(buffer, buf, count))
1576		return -EFAULT;
1577
1578	err = kstrtoint(strstrip(buffer), 0, &nice);
1579	if (err < 0)
1580		return err;
1581
1582	p = get_proc_task(inode);
1583	if (!p)
1584		return -ESRCH;
1585
1586	err = proc_sched_autogroup_set_nice(p, nice);
 
1587	if (err)
1588		count = err;
1589
1590	put_task_struct(p);
1591
1592	return count;
1593}
1594
1595static int sched_autogroup_open(struct inode *inode, struct file *filp)
1596{
1597	int ret;
1598
1599	ret = single_open(filp, sched_autogroup_show, NULL);
1600	if (!ret) {
1601		struct seq_file *m = filp->private_data;
1602
1603		m->private = inode;
1604	}
1605	return ret;
1606}
1607
1608static const struct file_operations proc_pid_sched_autogroup_operations = {
1609	.open		= sched_autogroup_open,
1610	.read		= seq_read,
1611	.write		= sched_autogroup_write,
1612	.llseek		= seq_lseek,
1613	.release	= single_release,
1614};
1615
1616#endif /* CONFIG_SCHED_AUTOGROUP */
1617
1618#ifdef CONFIG_TIME_NS
1619static int timens_offsets_show(struct seq_file *m, void *v)
1620{
1621	struct task_struct *p;
1622
1623	p = get_proc_task(file_inode(m->file));
1624	if (!p)
1625		return -ESRCH;
1626	proc_timens_show_offsets(p, m);
1627
1628	put_task_struct(p);
1629
1630	return 0;
1631}
1632
1633static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1634				    size_t count, loff_t *ppos)
1635{
1636	struct inode *inode = file_inode(file);
1637	struct proc_timens_offset offsets[2];
1638	char *kbuf = NULL, *pos, *next_line;
1639	struct task_struct *p;
1640	int ret, noffsets;
1641
1642	/* Only allow < page size writes at the beginning of the file */
1643	if ((*ppos != 0) || (count >= PAGE_SIZE))
1644		return -EINVAL;
1645
1646	/* Slurp in the user data */
1647	kbuf = memdup_user_nul(buf, count);
1648	if (IS_ERR(kbuf))
1649		return PTR_ERR(kbuf);
1650
1651	/* Parse the user data */
1652	ret = -EINVAL;
1653	noffsets = 0;
1654	for (pos = kbuf; pos; pos = next_line) {
1655		struct proc_timens_offset *off = &offsets[noffsets];
1656		char clock[10];
1657		int err;
1658
1659		/* Find the end of line and ensure we don't look past it */
1660		next_line = strchr(pos, '\n');
1661		if (next_line) {
1662			*next_line = '\0';
1663			next_line++;
1664			if (*next_line == '\0')
1665				next_line = NULL;
1666		}
1667
1668		err = sscanf(pos, "%9s %lld %lu", clock,
1669				&off->val.tv_sec, &off->val.tv_nsec);
1670		if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1671			goto out;
1672
1673		clock[sizeof(clock) - 1] = 0;
1674		if (strcmp(clock, "monotonic") == 0 ||
1675		    strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1676			off->clockid = CLOCK_MONOTONIC;
1677		else if (strcmp(clock, "boottime") == 0 ||
1678			 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1679			off->clockid = CLOCK_BOOTTIME;
1680		else
1681			goto out;
1682
1683		noffsets++;
1684		if (noffsets == ARRAY_SIZE(offsets)) {
1685			if (next_line)
1686				count = next_line - kbuf;
1687			break;
1688		}
1689	}
1690
1691	ret = -ESRCH;
1692	p = get_proc_task(inode);
1693	if (!p)
1694		goto out;
1695	ret = proc_timens_set_offset(file, p, offsets, noffsets);
1696	put_task_struct(p);
1697	if (ret)
1698		goto out;
1699
1700	ret = count;
1701out:
1702	kfree(kbuf);
1703	return ret;
1704}
1705
1706static int timens_offsets_open(struct inode *inode, struct file *filp)
1707{
1708	return single_open(filp, timens_offsets_show, inode);
1709}
1710
1711static const struct file_operations proc_timens_offsets_operations = {
1712	.open		= timens_offsets_open,
1713	.read		= seq_read,
1714	.write		= timens_offsets_write,
1715	.llseek		= seq_lseek,
1716	.release	= single_release,
1717};
1718#endif /* CONFIG_TIME_NS */
1719
1720static ssize_t comm_write(struct file *file, const char __user *buf,
1721				size_t count, loff_t *offset)
1722{
1723	struct inode *inode = file_inode(file);
1724	struct task_struct *p;
1725	char buffer[TASK_COMM_LEN] = {};
1726	const size_t maxlen = sizeof(buffer) - 1;
1727
1728	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
 
 
 
1729		return -EFAULT;
1730
1731	p = get_proc_task(inode);
1732	if (!p)
1733		return -ESRCH;
1734
1735	if (same_thread_group(current, p)) {
1736		set_task_comm(p, buffer);
1737		proc_comm_connector(p);
1738	}
1739	else
1740		count = -EINVAL;
1741
1742	put_task_struct(p);
1743
1744	return count;
1745}
1746
1747static int comm_show(struct seq_file *m, void *v)
1748{
1749	struct inode *inode = m->private;
1750	struct task_struct *p;
1751
1752	p = get_proc_task(inode);
1753	if (!p)
1754		return -ESRCH;
1755
1756	proc_task_name(m, p, false);
1757	seq_putc(m, '\n');
 
1758
1759	put_task_struct(p);
1760
1761	return 0;
1762}
1763
1764static int comm_open(struct inode *inode, struct file *filp)
1765{
1766	return single_open(filp, comm_show, inode);
1767}
1768
1769static const struct file_operations proc_pid_set_comm_operations = {
1770	.open		= comm_open,
1771	.read		= seq_read,
1772	.write		= comm_write,
1773	.llseek		= seq_lseek,
1774	.release	= single_release,
1775};
1776
1777static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1778{
1779	struct task_struct *task;
 
1780	struct file *exe_file;
1781
1782	task = get_proc_task(d_inode(dentry));
1783	if (!task)
1784		return -ENOENT;
1785	exe_file = get_task_exe_file(task);
1786	put_task_struct(task);
 
 
 
 
1787	if (exe_file) {
1788		*exe_path = exe_file->f_path;
1789		path_get(&exe_file->f_path);
1790		fput(exe_file);
1791		return 0;
1792	} else
1793		return -ENOENT;
1794}
1795
1796static const char *proc_pid_get_link(struct dentry *dentry,
1797				     struct inode *inode,
1798				     struct delayed_call *done)
1799{
1800	struct path path;
1801	int error = -EACCES;
1802
1803	if (!dentry)
1804		return ERR_PTR(-ECHILD);
1805
1806	/* Are we allowed to snoop on the tasks file descriptors? */
1807	if (!proc_fd_access_allowed(inode))
1808		goto out;
1809
1810	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1811	if (error)
1812		goto out;
1813
1814	error = nd_jump_link(&path);
1815out:
1816	return ERR_PTR(error);
1817}
1818
1819static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen)
1820{
1821	char *tmp = kmalloc(PATH_MAX, GFP_KERNEL);
1822	char *pathname;
1823	int len;
1824
1825	if (!tmp)
1826		return -ENOMEM;
1827
1828	pathname = d_path(path, tmp, PATH_MAX);
1829	len = PTR_ERR(pathname);
1830	if (IS_ERR(pathname))
1831		goto out;
1832	len = tmp + PATH_MAX - 1 - pathname;
1833
1834	if (len > buflen)
1835		len = buflen;
1836	if (copy_to_user(buffer, pathname, len))
1837		len = -EFAULT;
1838 out:
1839	kfree(tmp);
1840	return len;
1841}
1842
1843static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1844{
1845	int error = -EACCES;
1846	struct inode *inode = d_inode(dentry);
1847	struct path path;
1848
1849	/* Are we allowed to snoop on the tasks file descriptors? */
1850	if (!proc_fd_access_allowed(inode))
1851		goto out;
1852
1853	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1854	if (error)
1855		goto out;
1856
1857	error = do_proc_readlink(&path, buffer, buflen);
1858	path_put(&path);
1859out:
1860	return error;
1861}
1862
1863const struct inode_operations proc_pid_link_inode_operations = {
1864	.readlink	= proc_pid_readlink,
1865	.get_link	= proc_pid_get_link,
1866	.setattr	= proc_setattr,
1867};
1868
1869
1870/* building an inode */
1871
1872void task_dump_owner(struct task_struct *task, umode_t mode,
1873		     kuid_t *ruid, kgid_t *rgid)
1874{
1875	/* Depending on the state of dumpable compute who should own a
1876	 * proc file for a task.
1877	 */
1878	const struct cred *cred;
1879	kuid_t uid;
1880	kgid_t gid;
1881
1882	if (unlikely(task->flags & PF_KTHREAD)) {
1883		*ruid = GLOBAL_ROOT_UID;
1884		*rgid = GLOBAL_ROOT_GID;
1885		return;
1886	}
1887
1888	/* Default to the tasks effective ownership */
1889	rcu_read_lock();
1890	cred = __task_cred(task);
1891	uid = cred->euid;
1892	gid = cred->egid;
1893	rcu_read_unlock();
1894
1895	/*
1896	 * Before the /proc/pid/status file was created the only way to read
1897	 * the effective uid of a /process was to stat /proc/pid.  Reading
1898	 * /proc/pid/status is slow enough that procps and other packages
1899	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1900	 * made this apply to all per process world readable and executable
1901	 * directories.
1902	 */
1903	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1904		struct mm_struct *mm;
1905		task_lock(task);
1906		mm = task->mm;
1907		/* Make non-dumpable tasks owned by some root */
1908		if (mm) {
1909			if (get_dumpable(mm) != SUID_DUMP_USER) {
1910				struct user_namespace *user_ns = mm->user_ns;
1911
1912				uid = make_kuid(user_ns, 0);
1913				if (!uid_valid(uid))
1914					uid = GLOBAL_ROOT_UID;
1915
1916				gid = make_kgid(user_ns, 0);
1917				if (!gid_valid(gid))
1918					gid = GLOBAL_ROOT_GID;
1919			}
1920		} else {
1921			uid = GLOBAL_ROOT_UID;
1922			gid = GLOBAL_ROOT_GID;
1923		}
1924		task_unlock(task);
1925	}
1926	*ruid = uid;
1927	*rgid = gid;
1928}
1929
1930void proc_pid_evict_inode(struct proc_inode *ei)
1931{
1932	struct pid *pid = ei->pid;
1933
1934	if (S_ISDIR(ei->vfs_inode.i_mode)) {
1935		spin_lock(&pid->lock);
1936		hlist_del_init_rcu(&ei->sibling_inodes);
1937		spin_unlock(&pid->lock);
1938	}
1939}
1940
1941struct inode *proc_pid_make_inode(struct super_block *sb,
1942				  struct task_struct *task, umode_t mode)
1943{
1944	struct inode * inode;
1945	struct proc_inode *ei;
1946	struct pid *pid;
1947
1948	/* We need a new inode */
1949
1950	inode = new_inode(sb);
1951	if (!inode)
1952		goto out;
1953
1954	/* Common stuff */
1955	ei = PROC_I(inode);
1956	inode->i_mode = mode;
1957	inode->i_ino = get_next_ino();
1958	simple_inode_init_ts(inode);
1959	inode->i_op = &proc_def_inode_operations;
1960
1961	/*
1962	 * grab the reference to task.
1963	 */
1964	pid = get_task_pid(task, PIDTYPE_PID);
1965	if (!pid)
1966		goto out_unlock;
1967
1968	/* Let the pid remember us for quick removal */
1969	ei->pid = pid;
1970
1971	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
 
 
 
1972	security_task_to_inode(task, inode);
1973
1974out:
1975	return inode;
1976
1977out_unlock:
1978	iput(inode);
1979	return NULL;
1980}
1981
1982/*
1983 * Generating an inode and adding it into @pid->inodes, so that task will
1984 * invalidate inode's dentry before being released.
1985 *
1986 * This helper is used for creating dir-type entries under '/proc' and
1987 * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>'
1988 * can be released by invalidating '/proc/<tgid>' dentry.
1989 * In theory, dentries under '/proc/<tgid>/task' can also be released by
1990 * invalidating '/proc/<tgid>' dentry, we reserve it to handle single
1991 * thread exiting situation: Any one of threads should invalidate its
1992 * '/proc/<tgid>/task/<pid>' dentry before released.
1993 */
1994static struct inode *proc_pid_make_base_inode(struct super_block *sb,
1995				struct task_struct *task, umode_t mode)
1996{
1997	struct inode *inode;
1998	struct proc_inode *ei;
1999	struct pid *pid;
2000
2001	inode = proc_pid_make_inode(sb, task, mode);
2002	if (!inode)
2003		return NULL;
2004
2005	/* Let proc_flush_pid find this directory inode */
2006	ei = PROC_I(inode);
2007	pid = ei->pid;
2008	spin_lock(&pid->lock);
2009	hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
2010	spin_unlock(&pid->lock);
2011
2012	return inode;
2013}
2014
2015int pid_getattr(struct mnt_idmap *idmap, const struct path *path,
2016		struct kstat *stat, u32 request_mask, unsigned int query_flags)
2017{
2018	struct inode *inode = d_inode(path->dentry);
2019	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
2020	struct task_struct *task;
 
2021
2022	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
2023
2024	stat->uid = GLOBAL_ROOT_UID;
2025	stat->gid = GLOBAL_ROOT_GID;
2026	rcu_read_lock();
 
 
2027	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2028	if (task) {
2029		if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
2030			rcu_read_unlock();
2031			/*
2032			 * This doesn't prevent learning whether PID exists,
2033			 * it only makes getattr() consistent with readdir().
2034			 */
2035			return -ENOENT;
2036		}
2037		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
2038	}
2039	rcu_read_unlock();
2040	return 0;
2041}
2042
2043/* dentry stuff */
2044
2045/*
2046 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
2047 */
2048void pid_update_inode(struct task_struct *task, struct inode *inode)
2049{
2050	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
2051
2052	inode->i_mode &= ~(S_ISUID | S_ISGID);
2053	security_task_to_inode(task, inode);
2054}
2055
2056/*
2057 * Rewrite the inode's ownerships here because the owning task may have
2058 * performed a setuid(), etc.
2059 *
 
 
 
 
 
 
2060 */
2061static int pid_revalidate(struct dentry *dentry, unsigned int flags)
2062{
2063	struct inode *inode;
2064	struct task_struct *task;
2065	int ret = 0;
 
 
 
2066
2067	rcu_read_lock();
2068	inode = d_inode_rcu(dentry);
2069	if (!inode)
2070		goto out;
2071	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2072
2073	if (task) {
2074		pid_update_inode(task, inode);
2075		ret = 1;
 
 
 
 
 
 
 
 
 
 
 
 
 
2076	}
2077out:
2078	rcu_read_unlock();
2079	return ret;
2080}
2081
2082static inline bool proc_inode_is_dead(struct inode *inode)
2083{
2084	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
2085}
2086
2087int pid_delete_dentry(const struct dentry *dentry)
2088{
2089	/* Is the task we represent dead?
2090	 * If so, then don't put the dentry on the lru list,
2091	 * kill it immediately.
2092	 */
2093	return proc_inode_is_dead(d_inode(dentry));
2094}
2095
2096const struct dentry_operations pid_dentry_operations =
2097{
2098	.d_revalidate	= pid_revalidate,
2099	.d_delete	= pid_delete_dentry,
2100};
2101
2102/* Lookups */
2103
2104/*
2105 * Fill a directory entry.
2106 *
2107 * If possible create the dcache entry and derive our inode number and
2108 * file type from dcache entry.
2109 *
2110 * Since all of the proc inode numbers are dynamically generated, the inode
2111 * numbers do not exist until the inode is cache.  This means creating
2112 * the dcache entry in readdir is necessary to keep the inode numbers
2113 * reported by readdir in sync with the inode numbers reported
2114 * by stat.
2115 */
2116bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2117	const char *name, unsigned int len,
2118	instantiate_t instantiate, struct task_struct *task, const void *ptr)
2119{
2120	struct dentry *child, *dir = file->f_path.dentry;
2121	struct qstr qname = QSTR_INIT(name, len);
2122	struct inode *inode;
 
 
2123	unsigned type = DT_UNKNOWN;
2124	ino_t ino = 1;
2125
2126	child = d_hash_and_lookup(dir, &qname);
 
 
 
 
2127	if (!child) {
2128		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2129		child = d_alloc_parallel(dir, &qname, &wq);
2130		if (IS_ERR(child))
2131			goto end_instantiate;
2132		if (d_in_lookup(child)) {
2133			struct dentry *res;
2134			res = instantiate(child, task, ptr);
2135			d_lookup_done(child);
2136			if (unlikely(res)) {
2137				dput(child);
2138				child = res;
2139				if (IS_ERR(child))
2140					goto end_instantiate;
2141			}
2142		}
 
2143	}
2144	inode = d_inode(child);
2145	ino = inode->i_ino;
2146	type = inode->i_mode >> 12;
2147	dput(child);
2148end_instantiate:
2149	return dir_emit(ctx, name, len, ino, type);
 
 
 
 
2150}
2151
2152/*
2153 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2154 * which represent vma start and end addresses.
2155 */
2156static int dname_to_vma_addr(struct dentry *dentry,
2157			     unsigned long *start, unsigned long *end)
2158{
2159	const char *str = dentry->d_name.name;
2160	unsigned long long sval, eval;
2161	unsigned int len;
2162
2163	if (str[0] == '0' && str[1] != '-')
2164		return -EINVAL;
2165	len = _parse_integer(str, 16, &sval);
2166	if (len & KSTRTOX_OVERFLOW)
2167		return -EINVAL;
2168	if (sval != (unsigned long)sval)
2169		return -EINVAL;
2170	str += len;
 
 
 
 
 
 
 
2171
2172	if (*str != '-')
2173		return -EINVAL;
2174	str++;
2175
2176	if (str[0] == '0' && str[1])
2177		return -EINVAL;
2178	len = _parse_integer(str, 16, &eval);
2179	if (len & KSTRTOX_OVERFLOW)
2180		return -EINVAL;
2181	if (eval != (unsigned long)eval)
2182		return -EINVAL;
2183	str += len;
2184
2185	if (*str != '\0')
2186		return -EINVAL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2187
2188	*start = sval;
2189	*end = eval;
2190
2191	return 0;
2192}
2193
2194static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2195{
2196	unsigned long vm_start, vm_end;
2197	bool exact_vma_exists = false;
2198	struct mm_struct *mm = NULL;
2199	struct task_struct *task;
2200	struct inode *inode;
2201	int status = 0;
 
2202
2203	if (flags & LOOKUP_RCU)
2204		return -ECHILD;
2205
2206	inode = d_inode(dentry);
2207	task = get_proc_task(inode);
2208	if (!task)
2209		goto out_notask;
2210
2211	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2212	if (IS_ERR(mm))
2213		goto out;
2214
2215	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2216		status = mmap_read_lock_killable(mm);
2217		if (!status) {
2218			exact_vma_exists = !!find_exact_vma(mm, vm_start,
2219							    vm_end);
2220			mmap_read_unlock(mm);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2221		}
 
2222	}
2223
2224	mmput(mm);
2225
2226	if (exact_vma_exists) {
2227		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2228
2229		security_task_to_inode(task, inode);
2230		status = 1;
2231	}
2232
2233out:
2234	put_task_struct(task);
2235
2236out_notask:
2237	return status;
2238}
2239
2240static const struct dentry_operations tid_map_files_dentry_operations = {
2241	.d_revalidate	= map_files_d_revalidate,
 
2242	.d_delete	= pid_delete_dentry,
2243};
2244
2245static int map_files_get_link(struct dentry *dentry, struct path *path)
 
2246{
2247	unsigned long vm_start, vm_end;
2248	struct vm_area_struct *vma;
2249	struct task_struct *task;
2250	struct mm_struct *mm;
2251	int rc;
 
2252
2253	rc = -ENOENT;
2254	task = get_proc_task(d_inode(dentry));
2255	if (!task)
2256		goto out;
 
 
 
 
 
 
2257
2258	mm = get_task_mm(task);
2259	put_task_struct(task);
2260	if (!mm)
2261		goto out;
2262
2263	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2264	if (rc)
2265		goto out_mmput;
2266
2267	rc = mmap_read_lock_killable(mm);
2268	if (rc)
2269		goto out_mmput;
2270
2271	rc = -ENOENT;
2272	vma = find_exact_vma(mm, vm_start, vm_end);
2273	if (vma && vma->vm_file) {
2274		*path = *file_user_path(vma->vm_file);
2275		path_get(path);
2276		rc = 0;
2277	}
2278	mmap_read_unlock(mm);
2279
2280out_mmput:
2281	mmput(mm);
2282out:
2283	return rc;
2284}
2285
2286struct map_files_info {
2287	unsigned long	start;
2288	unsigned long	end;
2289	fmode_t		mode;
2290};
2291
2292/*
2293 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2294 * to concerns about how the symlinks may be used to bypass permissions on
2295 * ancestor directories in the path to the file in question.
2296 */
2297static const char *
2298proc_map_files_get_link(struct dentry *dentry,
2299			struct inode *inode,
2300		        struct delayed_call *done)
2301{
2302	if (!checkpoint_restore_ns_capable(&init_user_ns))
2303		return ERR_PTR(-EPERM);
2304
2305	return proc_pid_get_link(dentry, inode, done);
2306}
2307
2308/*
2309 * Identical to proc_pid_link_inode_operations except for get_link()
2310 */
2311static const struct inode_operations proc_map_files_link_inode_operations = {
2312	.readlink	= proc_pid_readlink,
2313	.get_link	= proc_map_files_get_link,
2314	.setattr	= proc_setattr,
2315};
2316
2317static struct dentry *
2318proc_map_files_instantiate(struct dentry *dentry,
2319			   struct task_struct *task, const void *ptr)
2320{
2321	fmode_t mode = (fmode_t)(unsigned long)ptr;
2322	struct proc_inode *ei;
2323	struct inode *inode;
2324
2325	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2326				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2327				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2328	if (!inode)
2329		return ERR_PTR(-ENOENT);
2330
2331	ei = PROC_I(inode);
2332	ei->op.proc_get_link = map_files_get_link;
2333
2334	inode->i_op = &proc_map_files_link_inode_operations;
2335	inode->i_size = 64;
 
 
 
 
 
 
2336
2337	return proc_splice_unmountable(inode, dentry,
2338				       &tid_map_files_dentry_operations);
 
 
 
 
 
 
2339}
2340
2341static struct dentry *proc_map_files_lookup(struct inode *dir,
2342		struct dentry *dentry, unsigned int flags)
 
2343{
2344	unsigned long vm_start, vm_end;
2345	struct vm_area_struct *vma;
2346	struct task_struct *task;
2347	struct dentry *result;
2348	struct mm_struct *mm;
2349
2350	result = ERR_PTR(-ENOENT);
2351	task = get_proc_task(dir);
2352	if (!task)
 
 
2353		goto out;
2354
2355	result = ERR_PTR(-EACCES);
2356	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2357		goto out_put_task;
2358
2359	result = ERR_PTR(-ENOENT);
2360	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2361		goto out_put_task;
2362
2363	mm = get_task_mm(task);
2364	if (!mm)
2365		goto out_put_task;
2366
2367	result = ERR_PTR(-EINTR);
2368	if (mmap_read_lock_killable(mm))
2369		goto out_put_mm;
2370
2371	result = ERR_PTR(-ENOENT);
2372	vma = find_exact_vma(mm, vm_start, vm_end);
2373	if (!vma)
2374		goto out_no_vma;
2375
2376	if (vma->vm_file)
2377		result = proc_map_files_instantiate(dentry, task,
2378				(void *)(unsigned long)vma->vm_file->f_mode);
2379
2380out_no_vma:
2381	mmap_read_unlock(mm);
2382out_put_mm:
2383	mmput(mm);
2384out_put_task:
2385	put_task_struct(task);
2386out:
2387	return result;
2388}
2389
2390static const struct inode_operations proc_map_files_inode_operations = {
2391	.lookup		= proc_map_files_lookup,
2392	.permission	= proc_fd_permission,
2393	.setattr	= proc_setattr,
2394};
2395
2396static int
2397proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2398{
2399	struct vm_area_struct *vma;
2400	struct task_struct *task;
2401	struct mm_struct *mm;
2402	unsigned long nr_files, pos, i;
2403	GENRADIX(struct map_files_info) fa;
2404	struct map_files_info *p;
2405	int ret;
2406	struct vma_iterator vmi;
2407
2408	genradix_init(&fa);
 
 
 
2409
2410	ret = -ENOENT;
2411	task = get_proc_task(file_inode(file));
2412	if (!task)
2413		goto out;
2414
2415	ret = -EACCES;
2416	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2417		goto out_put_task;
2418
2419	ret = 0;
2420	if (!dir_emit_dots(file, ctx))
2421		goto out_put_task;
2422
2423	mm = get_task_mm(task);
2424	if (!mm)
2425		goto out_put_task;
2426
2427	ret = mmap_read_lock_killable(mm);
2428	if (ret) {
2429		mmput(mm);
2430		goto out_put_task;
2431	}
2432
2433	nr_files = 0;
2434
2435	/*
2436	 * We need two passes here:
2437	 *
2438	 *  1) Collect vmas of mapped files with mmap_lock taken
2439	 *  2) Release mmap_lock and instantiate entries
2440	 *
2441	 * otherwise we get lockdep complained, since filldir()
2442	 * routine might require mmap_lock taken in might_fault().
2443	 */
2444
2445	pos = 2;
2446	vma_iter_init(&vmi, mm, 0);
2447	for_each_vma(vmi, vma) {
2448		if (!vma->vm_file)
2449			continue;
2450		if (++pos <= ctx->pos)
2451			continue;
2452
2453		p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2454		if (!p) {
2455			ret = -ENOMEM;
2456			mmap_read_unlock(mm);
2457			mmput(mm);
2458			goto out_put_task;
2459		}
2460
2461		p->start = vma->vm_start;
2462		p->end = vma->vm_end;
2463		p->mode = vma->vm_file->f_mode;
2464	}
2465	mmap_read_unlock(mm);
2466	mmput(mm);
2467
2468	for (i = 0; i < nr_files; i++) {
2469		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2470		unsigned int len;
2471
2472		p = genradix_ptr(&fa, i);
2473		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2474		if (!proc_fill_cache(file, ctx,
2475				      buf, len,
2476				      proc_map_files_instantiate,
2477				      task,
2478				      (void *)(unsigned long)p->mode))
2479			break;
2480		ctx->pos++;
2481	}
2482
2483out_put_task:
2484	put_task_struct(task);
2485out:
2486	genradix_free(&fa);
2487	return ret;
 
2488}
2489
2490static const struct file_operations proc_map_files_operations = {
2491	.read		= generic_read_dir,
2492	.iterate_shared	= proc_map_files_readdir,
2493	.llseek		= generic_file_llseek,
2494};
2495
2496#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2497struct timers_private {
2498	struct pid *pid;
2499	struct task_struct *task;
2500	struct sighand_struct *sighand;
2501	struct pid_namespace *ns;
2502	unsigned long flags;
2503};
2504
2505static void *timers_start(struct seq_file *m, loff_t *pos)
2506{
2507	struct timers_private *tp = m->private;
2508
2509	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2510	if (!tp->task)
2511		return ERR_PTR(-ESRCH);
2512
2513	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2514	if (!tp->sighand)
2515		return ERR_PTR(-ESRCH);
2516
2517	return seq_hlist_start(&tp->task->signal->posix_timers, *pos);
2518}
2519
2520static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2521{
2522	struct timers_private *tp = m->private;
2523	return seq_hlist_next(v, &tp->task->signal->posix_timers, pos);
2524}
2525
2526static void timers_stop(struct seq_file *m, void *v)
 
2527{
2528	struct timers_private *tp = m->private;
2529
2530	if (tp->sighand) {
2531		unlock_task_sighand(tp->task, &tp->flags);
2532		tp->sighand = NULL;
2533	}
2534
2535	if (tp->task) {
2536		put_task_struct(tp->task);
2537		tp->task = NULL;
2538	}
2539}
2540
2541static int show_timer(struct seq_file *m, void *v)
2542{
2543	struct k_itimer *timer;
2544	struct timers_private *tp = m->private;
2545	int notify;
2546	static const char * const nstr[] = {
2547		[SIGEV_SIGNAL] = "signal",
2548		[SIGEV_NONE] = "none",
2549		[SIGEV_THREAD] = "thread",
2550	};
2551
2552	timer = hlist_entry((struct hlist_node *)v, struct k_itimer, list);
2553	notify = timer->it_sigev_notify;
2554
2555	seq_printf(m, "ID: %d\n", timer->it_id);
2556	seq_printf(m, "signal: %d/%px\n",
2557		   timer->sigq.info.si_signo,
2558		   timer->sigq.info.si_value.sival_ptr);
2559	seq_printf(m, "notify: %s/%s.%d\n",
2560		   nstr[notify & ~SIGEV_THREAD_ID],
2561		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2562		   pid_nr_ns(timer->it_pid, tp->ns));
2563	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2564
2565	return 0;
2566}
2567
2568static const struct seq_operations proc_timers_seq_ops = {
2569	.start	= timers_start,
2570	.next	= timers_next,
2571	.stop	= timers_stop,
2572	.show	= show_timer,
2573};
2574
2575static int proc_timers_open(struct inode *inode, struct file *file)
 
 
 
 
2576{
2577	struct timers_private *tp;
2578
2579	tp = __seq_open_private(file, &proc_timers_seq_ops,
2580			sizeof(struct timers_private));
2581	if (!tp)
2582		return -ENOMEM;
2583
2584	tp->pid = proc_pid(inode);
2585	tp->ns = proc_pid_ns(inode->i_sb);
2586	return 0;
2587}
2588
2589static const struct file_operations proc_timers_operations = {
2590	.open		= proc_timers_open,
2591	.read		= seq_read,
2592	.llseek		= seq_lseek,
2593	.release	= seq_release_private,
 
 
2594};
2595#endif
2596
2597static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2598					size_t count, loff_t *offset)
2599{
2600	struct inode *inode = file_inode(file);
2601	struct task_struct *p;
2602	u64 slack_ns;
2603	int err;
2604
2605	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2606	if (err < 0)
2607		return err;
 
 
 
 
 
 
 
 
 
2608
2609	p = get_proc_task(inode);
2610	if (!p)
2611		return -ESRCH;
2612
2613	if (p != current) {
2614		rcu_read_lock();
2615		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2616			rcu_read_unlock();
2617			count = -EPERM;
2618			goto out;
2619		}
2620		rcu_read_unlock();
2621
2622		err = security_task_setscheduler(p);
2623		if (err) {
2624			count = err;
2625			goto out;
2626		}
2627	}
2628
2629	task_lock(p);
2630	if (rt_or_dl_task_policy(p))
2631		slack_ns = 0;
2632	else if (slack_ns == 0)
2633		slack_ns = p->default_timer_slack_ns;
2634	p->timer_slack_ns = slack_ns;
2635	task_unlock(p);
2636
2637out:
2638	put_task_struct(p);
2639
2640	return count;
2641}
2642
2643static int timerslack_ns_show(struct seq_file *m, void *v)
 
 
2644{
2645	struct inode *inode = m->private;
2646	struct task_struct *p;
2647	int err = 0;
2648
2649	p = get_proc_task(inode);
2650	if (!p)
2651		return -ESRCH;
2652
2653	if (p != current) {
2654		rcu_read_lock();
2655		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2656			rcu_read_unlock();
2657			err = -EPERM;
2658			goto out;
2659		}
2660		rcu_read_unlock();
2661
2662		err = security_task_getscheduler(p);
2663		if (err)
2664			goto out;
2665	}
2666
2667	task_lock(p);
2668	seq_printf(m, "%llu\n", p->timer_slack_ns);
2669	task_unlock(p);
2670
2671out:
2672	put_task_struct(p);
2673
2674	return err;
2675}
2676
2677static int timerslack_ns_open(struct inode *inode, struct file *filp)
2678{
2679	return single_open(filp, timerslack_ns_show, inode);
 
2680}
2681
2682static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2683	.open		= timerslack_ns_open,
2684	.read		= seq_read,
2685	.write		= timerslack_ns_write,
2686	.llseek		= seq_lseek,
2687	.release	= single_release,
 
 
 
 
 
 
2688};
2689
2690static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2691	struct task_struct *task, const void *ptr)
 
2692{
2693	const struct pid_entry *p = ptr;
2694	struct inode *inode;
2695	struct proc_inode *ei;
 
2696
2697	inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2698	if (!inode)
2699		return ERR_PTR(-ENOENT);
2700
2701	ei = PROC_I(inode);
 
2702	if (S_ISDIR(inode->i_mode))
2703		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2704	if (p->iop)
2705		inode->i_op = p->iop;
2706	if (p->fop)
2707		inode->i_fop = p->fop;
2708	ei->op = p->op;
2709	pid_update_inode(task, inode);
2710	d_set_d_op(dentry, &pid_dentry_operations);
2711	return d_splice_alias(inode, dentry);
 
 
 
 
 
2712}
2713
2714static struct dentry *proc_pident_lookup(struct inode *dir, 
2715					 struct dentry *dentry,
2716					 const struct pid_entry *p,
2717					 const struct pid_entry *end)
2718{
 
2719	struct task_struct *task = get_proc_task(dir);
2720	struct dentry *res = ERR_PTR(-ENOENT);
 
 
2721
2722	if (!task)
2723		goto out_no_task;
2724
2725	/*
2726	 * Yes, it does not scale. And it should not. Don't add
2727	 * new entries into /proc/<tgid>/ without very good reasons.
2728	 */
2729	for (; p < end; p++) {
 
2730		if (p->len != dentry->d_name.len)
2731			continue;
2732		if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2733			res = proc_pident_instantiate(dentry, task, p);
2734			break;
2735		}
2736	}
 
 
 
 
 
2737	put_task_struct(task);
2738out_no_task:
2739	return res;
 
 
 
 
 
 
 
2740}
2741
2742static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
 
2743		const struct pid_entry *ents, unsigned int nents)
2744{
2745	struct task_struct *task = get_proc_task(file_inode(file));
2746	const struct pid_entry *p;
 
 
 
 
 
2747
 
2748	if (!task)
2749		return -ENOENT;
2750
2751	if (!dir_emit_dots(file, ctx))
2752		goto out;
2753
2754	if (ctx->pos >= nents + 2)
2755		goto out;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2756
2757	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2758		if (!proc_fill_cache(file, ctx, p->name, p->len,
2759				proc_pident_instantiate, task, p))
2760			break;
2761		ctx->pos++;
2762	}
2763out:
2764	put_task_struct(task);
2765	return 0;
 
2766}
2767
2768#ifdef CONFIG_SECURITY
2769static int proc_pid_attr_open(struct inode *inode, struct file *file)
2770{
2771	file->private_data = NULL;
2772	__mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2773	return 0;
2774}
2775
2776static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2777				  size_t count, loff_t *ppos)
2778{
2779	struct inode * inode = file_inode(file);
2780	char *p = NULL;
2781	ssize_t length;
2782	struct task_struct *task = get_proc_task(inode);
2783
2784	if (!task)
2785		return -ESRCH;
2786
2787	length = security_getprocattr(task, PROC_I(inode)->op.lsmid,
2788				      file->f_path.dentry->d_name.name,
2789				      &p);
2790	put_task_struct(task);
2791	if (length > 0)
2792		length = simple_read_from_buffer(buf, count, ppos, p, length);
2793	kfree(p);
2794	return length;
2795}
2796
2797static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2798				   size_t count, loff_t *ppos)
2799{
2800	struct inode * inode = file_inode(file);
2801	struct task_struct *task;
2802	void *page;
2803	int rv;
2804
2805	/* A task may only write when it was the opener. */
2806	if (file->private_data != current->mm)
2807		return -EPERM;
2808
2809	rcu_read_lock();
2810	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2811	if (!task) {
2812		rcu_read_unlock();
2813		return -ESRCH;
2814	}
2815	/* A task may only write its own attributes. */
2816	if (current != task) {
2817		rcu_read_unlock();
2818		return -EACCES;
2819	}
2820	/* Prevent changes to overridden credentials. */
2821	if (current_cred() != current_real_cred()) {
2822		rcu_read_unlock();
2823		return -EBUSY;
2824	}
2825	rcu_read_unlock();
2826
 
 
 
2827	if (count > PAGE_SIZE)
2828		count = PAGE_SIZE;
2829
2830	/* No partial writes. */
 
2831	if (*ppos != 0)
2832		return -EINVAL;
2833
2834	page = memdup_user(buf, count);
2835	if (IS_ERR(page)) {
2836		rv = PTR_ERR(page);
2837		goto out;
2838	}
 
 
 
2839
2840	/* Guard against adverse ptrace interaction */
2841	rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2842	if (rv < 0)
2843		goto out_free;
2844
2845	rv = security_setprocattr(PROC_I(inode)->op.lsmid,
2846				  file->f_path.dentry->d_name.name, page,
2847				  count);
2848	mutex_unlock(&current->signal->cred_guard_mutex);
2849out_free:
2850	kfree(page);
2851out:
2852	return rv;
 
 
2853}
2854
2855static const struct file_operations proc_pid_attr_operations = {
2856	.open		= proc_pid_attr_open,
2857	.read		= proc_pid_attr_read,
2858	.write		= proc_pid_attr_write,
2859	.llseek		= generic_file_llseek,
2860	.release	= mem_release,
2861};
2862
2863#define LSM_DIR_OPS(LSM) \
2864static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2865			     struct dir_context *ctx) \
2866{ \
2867	return proc_pident_readdir(filp, ctx, \
2868				   LSM##_attr_dir_stuff, \
2869				   ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2870} \
2871\
2872static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2873	.read		= generic_read_dir, \
2874	.iterate_shared	= proc_##LSM##_attr_dir_iterate, \
2875	.llseek		= default_llseek, \
2876}; \
2877\
2878static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2879				struct dentry *dentry, unsigned int flags) \
2880{ \
2881	return proc_pident_lookup(dir, dentry, \
2882				  LSM##_attr_dir_stuff, \
2883				  LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2884} \
2885\
2886static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2887	.lookup		= proc_##LSM##_attr_dir_lookup, \
2888	.getattr	= pid_getattr, \
2889	.setattr	= proc_setattr, \
2890}
2891
2892#ifdef CONFIG_SECURITY_SMACK
2893static const struct pid_entry smack_attr_dir_stuff[] = {
2894	ATTR(LSM_ID_SMACK, "current",	0666),
2895};
2896LSM_DIR_OPS(smack);
2897#endif
2898
2899#ifdef CONFIG_SECURITY_APPARMOR
2900static const struct pid_entry apparmor_attr_dir_stuff[] = {
2901	ATTR(LSM_ID_APPARMOR, "current",	0666),
2902	ATTR(LSM_ID_APPARMOR, "prev",		0444),
2903	ATTR(LSM_ID_APPARMOR, "exec",		0666),
2904};
2905LSM_DIR_OPS(apparmor);
2906#endif
2907
2908static const struct pid_entry attr_dir_stuff[] = {
2909	ATTR(LSM_ID_UNDEF, "current",	0666),
2910	ATTR(LSM_ID_UNDEF, "prev",		0444),
2911	ATTR(LSM_ID_UNDEF, "exec",		0666),
2912	ATTR(LSM_ID_UNDEF, "fscreate",	0666),
2913	ATTR(LSM_ID_UNDEF, "keycreate",	0666),
2914	ATTR(LSM_ID_UNDEF, "sockcreate",	0666),
2915#ifdef CONFIG_SECURITY_SMACK
2916	DIR("smack",			0555,
2917	    proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2918#endif
2919#ifdef CONFIG_SECURITY_APPARMOR
2920	DIR("apparmor",			0555,
2921	    proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2922#endif
2923};
2924
2925static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
 
2926{
2927	return proc_pident_readdir(file, ctx, 
2928				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2929}
2930
2931static const struct file_operations proc_attr_dir_operations = {
2932	.read		= generic_read_dir,
2933	.iterate_shared	= proc_attr_dir_readdir,
2934	.llseek		= generic_file_llseek,
2935};
2936
2937static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2938				struct dentry *dentry, unsigned int flags)
2939{
2940	return proc_pident_lookup(dir, dentry,
2941				  attr_dir_stuff,
2942				  attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2943}
2944
2945static const struct inode_operations proc_attr_dir_inode_operations = {
2946	.lookup		= proc_attr_dir_lookup,
2947	.getattr	= pid_getattr,
2948	.setattr	= proc_setattr,
2949};
2950
2951#endif
2952
2953#ifdef CONFIG_ELF_CORE
2954static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2955					 size_t count, loff_t *ppos)
2956{
2957	struct task_struct *task = get_proc_task(file_inode(file));
2958	struct mm_struct *mm;
2959	char buffer[PROC_NUMBUF];
2960	size_t len;
2961	int ret;
2962
2963	if (!task)
2964		return -ESRCH;
2965
2966	ret = 0;
2967	mm = get_task_mm(task);
2968	if (mm) {
2969		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2970			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2971				MMF_DUMP_FILTER_SHIFT));
2972		mmput(mm);
2973		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2974	}
2975
2976	put_task_struct(task);
2977
2978	return ret;
2979}
2980
2981static ssize_t proc_coredump_filter_write(struct file *file,
2982					  const char __user *buf,
2983					  size_t count,
2984					  loff_t *ppos)
2985{
2986	struct task_struct *task;
2987	struct mm_struct *mm;
 
2988	unsigned int val;
2989	int ret;
2990	int i;
2991	unsigned long mask;
2992
2993	ret = kstrtouint_from_user(buf, count, 0, &val);
2994	if (ret < 0)
2995		return ret;
 
 
 
 
 
 
 
 
 
 
2996
2997	ret = -ESRCH;
2998	task = get_proc_task(file_inode(file));
2999	if (!task)
3000		goto out_no_task;
3001
 
3002	mm = get_task_mm(task);
3003	if (!mm)
3004		goto out_no_mm;
3005	ret = 0;
3006
3007	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
3008		if (val & mask)
3009			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3010		else
3011			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3012	}
3013
3014	mmput(mm);
3015 out_no_mm:
3016	put_task_struct(task);
3017 out_no_task:
3018	if (ret < 0)
3019		return ret;
3020	return count;
3021}
3022
3023static const struct file_operations proc_coredump_filter_operations = {
3024	.read		= proc_coredump_filter_read,
3025	.write		= proc_coredump_filter_write,
3026	.llseek		= generic_file_llseek,
3027};
3028#endif
3029
3030#ifdef CONFIG_TASK_IO_ACCOUNTING
3031static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
 
 
 
3032{
3033	struct task_io_accounting acct;
3034	int result;
 
 
 
 
 
 
3035
3036	result = down_read_killable(&task->signal->exec_update_lock);
3037	if (result)
3038		return result;
3039
3040	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
3041		result = -EACCES;
3042		goto out_unlock;
 
 
 
 
3043	}
 
 
 
3044
3045	if (whole) {
3046		struct signal_struct *sig = task->signal;
3047		struct task_struct *t;
3048		unsigned int seq = 1;
3049		unsigned long flags;
 
 
3050
3051		rcu_read_lock();
3052		do {
3053			seq++; /* 2 on the 1st/lockless path, otherwise odd */
3054			flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
 
3055
3056			acct = sig->ioac;
3057			__for_each_thread(sig, t)
3058				task_io_accounting_add(&acct, &t->ioac);
 
 
 
 
 
 
 
 
3059
3060		} while (need_seqretry(&sig->stats_lock, seq));
3061		done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
3062		rcu_read_unlock();
3063	} else {
3064		acct = task->ioac;
3065	}
 
3066
3067	seq_printf(m,
3068		   "rchar: %llu\n"
3069		   "wchar: %llu\n"
3070		   "syscr: %llu\n"
3071		   "syscw: %llu\n"
3072		   "read_bytes: %llu\n"
3073		   "write_bytes: %llu\n"
3074		   "cancelled_write_bytes: %llu\n",
3075		   (unsigned long long)acct.rchar,
3076		   (unsigned long long)acct.wchar,
3077		   (unsigned long long)acct.syscr,
3078		   (unsigned long long)acct.syscw,
3079		   (unsigned long long)acct.read_bytes,
3080		   (unsigned long long)acct.write_bytes,
3081		   (unsigned long long)acct.cancelled_write_bytes);
3082	result = 0;
3083
3084out_unlock:
3085	up_read(&task->signal->exec_update_lock);
3086	return result;
3087}
3088
3089static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3090				  struct pid *pid, struct task_struct *task)
3091{
3092	return do_io_accounting(task, m, 0);
3093}
 
3094
3095static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3096				   struct pid *pid, struct task_struct *task)
3097{
3098	return do_io_accounting(task, m, 1);
 
 
 
 
 
 
 
 
 
 
 
 
 
3099}
3100#endif /* CONFIG_TASK_IO_ACCOUNTING */
3101
3102#ifdef CONFIG_USER_NS
3103static int proc_id_map_open(struct inode *inode, struct file *file,
3104	const struct seq_operations *seq_ops)
3105{
3106	struct user_namespace *ns = NULL;
3107	struct task_struct *task;
3108	struct seq_file *seq;
3109	int ret = -EINVAL;
3110
3111	task = get_proc_task(inode);
3112	if (task) {
3113		rcu_read_lock();
3114		ns = get_user_ns(task_cred_xxx(task, user_ns));
3115		rcu_read_unlock();
3116		put_task_struct(task);
3117	}
3118	if (!ns)
3119		goto err;
3120
3121	ret = seq_open(file, seq_ops);
3122	if (ret)
3123		goto err_put_ns;
3124
3125	seq = file->private_data;
3126	seq->private = ns;
3127
3128	return 0;
3129err_put_ns:
3130	put_user_ns(ns);
3131err:
3132	return ret;
3133}
 
3134
3135static int proc_id_map_release(struct inode *inode, struct file *file)
3136{
3137	struct seq_file *seq = file->private_data;
3138	struct user_namespace *ns = seq->private;
3139	put_user_ns(ns);
3140	return seq_release(inode, file);
3141}
3142
3143static int proc_uid_map_open(struct inode *inode, struct file *file)
3144{
3145	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
 
3146}
3147
3148static int proc_gid_map_open(struct inode *inode, struct file *file)
 
3149{
3150	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
 
3151}
3152
3153static int proc_projid_map_open(struct inode *inode, struct file *file)
 
3154{
3155	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3156}
 
3157
3158static const struct file_operations proc_uid_map_operations = {
3159	.open		= proc_uid_map_open,
3160	.write		= proc_uid_map_write,
3161	.read		= seq_read,
3162	.llseek		= seq_lseek,
3163	.release	= proc_id_map_release,
3164};
3165
3166static const struct file_operations proc_gid_map_operations = {
3167	.open		= proc_gid_map_open,
3168	.write		= proc_gid_map_write,
3169	.read		= seq_read,
3170	.llseek		= seq_lseek,
3171	.release	= proc_id_map_release,
3172};
3173
3174static const struct file_operations proc_projid_map_operations = {
3175	.open		= proc_projid_map_open,
3176	.write		= proc_projid_map_write,
3177	.read		= seq_read,
3178	.llseek		= seq_lseek,
3179	.release	= proc_id_map_release,
3180};
3181
3182static int proc_setgroups_open(struct inode *inode, struct file *file)
3183{
3184	struct user_namespace *ns = NULL;
3185	struct task_struct *task;
3186	int ret;
3187
3188	ret = -ESRCH;
3189	task = get_proc_task(inode);
3190	if (task) {
3191		rcu_read_lock();
3192		ns = get_user_ns(task_cred_xxx(task, user_ns));
3193		rcu_read_unlock();
3194		put_task_struct(task);
3195	}
3196	if (!ns)
3197		goto err;
3198
3199	if (file->f_mode & FMODE_WRITE) {
3200		ret = -EACCES;
3201		if (!ns_capable(ns, CAP_SYS_ADMIN))
3202			goto err_put_ns;
3203	}
3204
3205	ret = single_open(file, &proc_setgroups_show, ns);
3206	if (ret)
3207		goto err_put_ns;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3208
3209	return 0;
3210err_put_ns:
3211	put_user_ns(ns);
3212err:
3213	return ret;
3214}
3215
3216static int proc_setgroups_release(struct inode *inode, struct file *file)
3217{
3218	struct seq_file *seq = file->private_data;
3219	struct user_namespace *ns = seq->private;
3220	int ret = single_release(inode, file);
3221	put_user_ns(ns);
3222	return ret;
3223}
3224
3225static const struct file_operations proc_setgroups_operations = {
3226	.open		= proc_setgroups_open,
3227	.write		= proc_setgroups_write,
3228	.read		= seq_read,
3229	.llseek		= seq_lseek,
3230	.release	= proc_setgroups_release,
3231};
3232#endif /* CONFIG_USER_NS */
3233
3234static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3235				struct pid *pid, struct task_struct *task)
3236{
3237	int err = lock_trace(task);
3238	if (!err) {
3239		seq_printf(m, "%08x\n", task->personality);
3240		unlock_trace(task);
3241	}
3242	return err;
3243}
3244
3245#ifdef CONFIG_LIVEPATCH
3246static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3247				struct pid *pid, struct task_struct *task)
3248{
3249	seq_printf(m, "%d\n", task->patch_state);
3250	return 0;
3251}
3252#endif /* CONFIG_LIVEPATCH */
3253
3254#ifdef CONFIG_KSM
3255static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
3256				struct pid *pid, struct task_struct *task)
3257{
3258	struct mm_struct *mm;
3259
3260	mm = get_task_mm(task);
3261	if (mm) {
3262		seq_printf(m, "%lu\n", mm->ksm_merging_pages);
3263		mmput(mm);
3264	}
3265
3266	return 0;
3267}
3268static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
3269				struct pid *pid, struct task_struct *task)
3270{
3271	struct mm_struct *mm;
3272
3273	mm = get_task_mm(task);
3274	if (mm) {
3275		seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
3276		seq_printf(m, "ksm_zero_pages %ld\n", mm_ksm_zero_pages(mm));
3277		seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
3278		seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
3279		mmput(mm);
3280	}
3281
3282	return 0;
3283}
3284#endif /* CONFIG_KSM */
3285
3286#ifdef CONFIG_STACKLEAK_METRICS
3287static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3288				struct pid *pid, struct task_struct *task)
3289{
3290	unsigned long prev_depth = THREAD_SIZE -
3291				(task->prev_lowest_stack & (THREAD_SIZE - 1));
3292	unsigned long depth = THREAD_SIZE -
3293				(task->lowest_stack & (THREAD_SIZE - 1));
3294
3295	seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3296							prev_depth, depth);
3297	return 0;
3298}
3299#endif /* CONFIG_STACKLEAK_METRICS */
3300
3301/*
3302 * Thread groups
3303 */
3304static const struct file_operations proc_task_operations;
3305static const struct inode_operations proc_task_inode_operations;
3306
3307static const struct pid_entry tgid_base_stuff[] = {
3308	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3309	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3310	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3311	DIR("fdinfo",     S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3312	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3313#ifdef CONFIG_NET
3314	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3315#endif
3316	REG("environ",    S_IRUSR, proc_environ_operations),
3317	REG("auxv",       S_IRUSR, proc_auxv_operations),
3318	ONE("status",     S_IRUGO, proc_pid_status),
3319	ONE("personality", S_IRUSR, proc_pid_personality),
3320	ONE("limits",	  S_IRUGO, proc_pid_limits),
3321#ifdef CONFIG_SCHED_DEBUG
3322	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3323#endif
3324#ifdef CONFIG_SCHED_AUTOGROUP
3325	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3326#endif
3327#ifdef CONFIG_TIME_NS
3328	REG("timens_offsets",  S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3329#endif
3330	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3331#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3332	ONE("syscall",    S_IRUSR, proc_pid_syscall),
3333#endif
3334	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
3335	ONE("stat",       S_IRUGO, proc_tgid_stat),
3336	ONE("statm",      S_IRUGO, proc_pid_statm),
3337	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3338#ifdef CONFIG_NUMA
3339	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3340#endif
3341	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3342	LNK("cwd",        proc_cwd_link),
3343	LNK("root",       proc_root_link),
3344	LNK("exe",        proc_exe_link),
3345	REG("mounts",     S_IRUGO, proc_mounts_operations),
3346	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3347	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3348#ifdef CONFIG_PROC_PAGE_MONITOR
3349	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3350	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3351	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3352	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3353#endif
3354#ifdef CONFIG_SECURITY
3355	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3356#endif
3357#ifdef CONFIG_KALLSYMS
3358	ONE("wchan",      S_IRUGO, proc_pid_wchan),
3359#endif
3360#ifdef CONFIG_STACKTRACE
3361	ONE("stack",      S_IRUSR, proc_pid_stack),
3362#endif
3363#ifdef CONFIG_SCHED_INFO
3364	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
3365#endif
3366#ifdef CONFIG_LATENCYTOP
3367	REG("latency",  S_IRUGO, proc_lstats_operations),
3368#endif
3369#ifdef CONFIG_PROC_PID_CPUSET
3370	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
3371#endif
3372#ifdef CONFIG_CGROUPS
3373	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3374#endif
3375#ifdef CONFIG_PROC_CPU_RESCTRL
3376	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3377#endif
3378	ONE("oom_score",  S_IRUGO, proc_oom_score),
3379	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3380	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3381#ifdef CONFIG_AUDIT
3382	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3383	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3384#endif
3385#ifdef CONFIG_FAULT_INJECTION
3386	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3387	REG("fail-nth", 0644, proc_fail_nth_operations),
3388#endif
3389#ifdef CONFIG_ELF_CORE
3390	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3391#endif
3392#ifdef CONFIG_TASK_IO_ACCOUNTING
3393	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3394#endif
3395#ifdef CONFIG_USER_NS
3396	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3397	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3398	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3399	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3400#endif
3401#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3402	REG("timers",	  S_IRUGO, proc_timers_operations),
3403#endif
3404	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3405#ifdef CONFIG_LIVEPATCH
3406	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3407#endif
3408#ifdef CONFIG_STACKLEAK_METRICS
3409	ONE("stack_depth", S_IRUGO, proc_stack_depth),
3410#endif
3411#ifdef CONFIG_PROC_PID_ARCH_STATUS
3412	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3413#endif
3414#ifdef CONFIG_SECCOMP_CACHE_DEBUG
3415	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3416#endif
3417#ifdef CONFIG_KSM
3418	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
3419	ONE("ksm_stat",  S_IRUSR, proc_pid_ksm_stat),
3420#endif
3421};
3422
3423static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
 
3424{
3425	return proc_pident_readdir(file, ctx,
3426				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3427}
3428
3429static const struct file_operations proc_tgid_base_operations = {
3430	.read		= generic_read_dir,
3431	.iterate_shared	= proc_tgid_base_readdir,
3432	.llseek		= generic_file_llseek,
3433};
3434
3435struct pid *tgid_pidfd_to_pid(const struct file *file)
3436{
3437	if (file->f_op != &proc_tgid_base_operations)
3438		return ERR_PTR(-EBADF);
3439
3440	return proc_pid(file_inode(file));
3441}
3442
3443static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3444{
3445	return proc_pident_lookup(dir, dentry,
3446				  tgid_base_stuff,
3447				  tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3448}
3449
3450static const struct inode_operations proc_tgid_base_inode_operations = {
3451	.lookup		= proc_tgid_base_lookup,
3452	.getattr	= pid_getattr,
3453	.setattr	= proc_setattr,
3454	.permission	= proc_pid_permission,
3455};
3456
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3457/**
3458 * proc_flush_pid -  Remove dcache entries for @pid from the /proc dcache.
3459 * @pid: pid that should be flushed.
3460 *
3461 * This function walks a list of inodes (that belong to any proc
3462 * filesystem) that are attached to the pid and flushes them from
3463 * the dentry cache.
 
 
 
 
 
 
3464 *
3465 * It is safe and reasonable to cache /proc entries for a task until
3466 * that task exits.  After that they just clog up the dcache with
3467 * useless entries, possibly causing useful dcache entries to be
3468 * flushed instead.  This routine is provided to flush those useless
3469 * dcache entries when a process is reaped.
3470 *
3471 * NOTE: This routine is just an optimization so it does not guarantee
3472 *       that no dcache entries will exist after a process is reaped
3473 *       it just makes it very unlikely that any will persist.
3474 */
3475
3476void proc_flush_pid(struct pid *pid)
3477{
3478	proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3479}
3480
3481static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3482				   struct task_struct *task, const void *ptr)
 
3483{
 
3484	struct inode *inode;
3485
3486	inode = proc_pid_make_base_inode(dentry->d_sb, task,
3487					 S_IFDIR | S_IRUGO | S_IXUGO);
3488	if (!inode)
3489		return ERR_PTR(-ENOENT);
3490
 
3491	inode->i_op = &proc_tgid_base_inode_operations;
3492	inode->i_fop = &proc_tgid_base_operations;
3493	inode->i_flags|=S_IMMUTABLE;
3494
3495	set_nlink(inode, nlink_tgid);
3496	pid_update_inode(task, inode);
3497
3498	d_set_d_op(dentry, &pid_dentry_operations);
3499	return d_splice_alias(inode, dentry);
 
 
 
 
 
 
3500}
3501
3502struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3503{
 
3504	struct task_struct *task;
3505	unsigned tgid;
3506	struct proc_fs_info *fs_info;
3507	struct pid_namespace *ns;
3508	struct dentry *result = ERR_PTR(-ENOENT);
3509
3510	tgid = name_to_int(&dentry->d_name);
 
 
 
 
3511	if (tgid == ~0U)
3512		goto out;
3513
3514	fs_info = proc_sb_info(dentry->d_sb);
3515	ns = fs_info->pid_ns;
3516	rcu_read_lock();
3517	task = find_task_by_pid_ns(tgid, ns);
3518	if (task)
3519		get_task_struct(task);
3520	rcu_read_unlock();
3521	if (!task)
3522		goto out;
3523
3524	/* Limit procfs to only ptraceable tasks */
3525	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3526		if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3527			goto out_put_task;
3528	}
3529
3530	result = proc_pid_instantiate(dentry, task, NULL);
3531out_put_task:
3532	put_task_struct(task);
3533out:
3534	return result;
3535}
3536
3537/*
3538 * Find the first task with tgid >= tgid
3539 *
3540 */
3541struct tgid_iter {
3542	unsigned int tgid;
3543	struct task_struct *task;
3544};
3545static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3546{
3547	struct pid *pid;
3548
3549	if (iter.task)
3550		put_task_struct(iter.task);
3551	rcu_read_lock();
3552retry:
3553	iter.task = NULL;
3554	pid = find_ge_pid(iter.tgid, ns);
3555	if (pid) {
3556		iter.tgid = pid_nr_ns(pid, ns);
3557		iter.task = pid_task(pid, PIDTYPE_TGID);
3558		if (!iter.task) {
 
 
 
 
 
 
 
 
 
 
 
 
3559			iter.tgid += 1;
3560			goto retry;
3561		}
3562		get_task_struct(iter.task);
3563	}
3564	rcu_read_unlock();
3565	return iter;
3566}
3567
3568#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
 
 
 
 
 
 
 
 
 
3569
3570/* for the /proc/ directory itself, after non-process stuff has been done */
3571int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3572{
 
 
3573	struct tgid_iter iter;
3574	struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3575	struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3576	loff_t pos = ctx->pos;
 
 
3577
3578	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3579		return 0;
 
3580
3581	if (pos == TGID_OFFSET - 2) {
3582		struct inode *inode = d_inode(fs_info->proc_self);
3583		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3584			return 0;
3585		ctx->pos = pos = pos + 1;
3586	}
3587	if (pos == TGID_OFFSET - 1) {
3588		struct inode *inode = d_inode(fs_info->proc_thread_self);
3589		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3590			return 0;
3591		ctx->pos = pos = pos + 1;
3592	}
3593	iter.tgid = pos - TGID_OFFSET;
3594	iter.task = NULL;
 
3595	for (iter = next_tgid(ns, iter);
3596	     iter.task;
3597	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3598		char name[10 + 1];
3599		unsigned int len;
3600
3601		cond_resched();
3602		if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3603			continue;
3604
3605		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3606		ctx->pos = iter.tgid + TGID_OFFSET;
3607		if (!proc_fill_cache(file, ctx, name, len,
3608				     proc_pid_instantiate, iter.task, NULL)) {
3609			put_task_struct(iter.task);
3610			return 0;
3611		}
3612	}
3613	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
 
 
 
3614	return 0;
3615}
3616
3617/*
3618 * proc_tid_comm_permission is a special permission function exclusively
3619 * used for the node /proc/<pid>/task/<tid>/comm.
3620 * It bypasses generic permission checks in the case where a task of the same
3621 * task group attempts to access the node.
3622 * The rationale behind this is that glibc and bionic access this node for
3623 * cross thread naming (pthread_set/getname_np(!self)). However, if
3624 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3625 * which locks out the cross thread naming implementation.
3626 * This function makes sure that the node is always accessible for members of
3627 * same thread group.
3628 */
3629static int proc_tid_comm_permission(struct mnt_idmap *idmap,
3630				    struct inode *inode, int mask)
3631{
3632	bool is_same_tgroup;
3633	struct task_struct *task;
3634
3635	task = get_proc_task(inode);
3636	if (!task)
3637		return -ESRCH;
3638	is_same_tgroup = same_thread_group(current, task);
3639	put_task_struct(task);
3640
3641	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3642		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3643		 * read or written by the members of the corresponding
3644		 * thread group.
3645		 */
3646		return 0;
3647	}
3648
3649	return generic_permission(&nop_mnt_idmap, inode, mask);
3650}
3651
3652static const struct inode_operations proc_tid_comm_inode_operations = {
3653		.setattr	= proc_setattr,
3654		.permission	= proc_tid_comm_permission,
3655};
3656
3657/*
3658 * Tasks
3659 */
3660static const struct pid_entry tid_base_stuff[] = {
3661	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3662	DIR("fdinfo",    S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3663	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3664#ifdef CONFIG_NET
3665	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3666#endif
3667	REG("environ",   S_IRUSR, proc_environ_operations),
3668	REG("auxv",      S_IRUSR, proc_auxv_operations),
3669	ONE("status",    S_IRUGO, proc_pid_status),
3670	ONE("personality", S_IRUSR, proc_pid_personality),
3671	ONE("limits",	 S_IRUGO, proc_pid_limits),
3672#ifdef CONFIG_SCHED_DEBUG
3673	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3674#endif
3675	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3676			 &proc_tid_comm_inode_operations,
3677			 &proc_pid_set_comm_operations, {}),
3678#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3679	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3680#endif
3681	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3682	ONE("stat",      S_IRUGO, proc_tid_stat),
3683	ONE("statm",     S_IRUGO, proc_pid_statm),
3684	REG("maps",      S_IRUGO, proc_pid_maps_operations),
3685#ifdef CONFIG_PROC_CHILDREN
3686	REG("children",  S_IRUGO, proc_tid_children_operations),
3687#endif
3688#ifdef CONFIG_NUMA
3689	REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3690#endif
3691	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3692	LNK("cwd",       proc_cwd_link),
3693	LNK("root",      proc_root_link),
3694	LNK("exe",       proc_exe_link),
3695	REG("mounts",    S_IRUGO, proc_mounts_operations),
3696	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3697#ifdef CONFIG_PROC_PAGE_MONITOR
3698	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3699	REG("smaps",     S_IRUGO, proc_pid_smaps_operations),
3700	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3701	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3702#endif
3703#ifdef CONFIG_SECURITY
3704	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3705#endif
3706#ifdef CONFIG_KALLSYMS
3707	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3708#endif
3709#ifdef CONFIG_STACKTRACE
3710	ONE("stack",      S_IRUSR, proc_pid_stack),
3711#endif
3712#ifdef CONFIG_SCHED_INFO
3713	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3714#endif
3715#ifdef CONFIG_LATENCYTOP
3716	REG("latency",  S_IRUGO, proc_lstats_operations),
3717#endif
3718#ifdef CONFIG_PROC_PID_CPUSET
3719	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3720#endif
3721#ifdef CONFIG_CGROUPS
3722	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3723#endif
3724#ifdef CONFIG_PROC_CPU_RESCTRL
3725	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3726#endif
3727	ONE("oom_score", S_IRUGO, proc_oom_score),
3728	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3729	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3730#ifdef CONFIG_AUDIT
3731	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3732	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3733#endif
3734#ifdef CONFIG_FAULT_INJECTION
3735	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3736	REG("fail-nth", 0644, proc_fail_nth_operations),
3737#endif
3738#ifdef CONFIG_TASK_IO_ACCOUNTING
3739	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3740#endif
3741#ifdef CONFIG_USER_NS
3742	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3743	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3744	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3745	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3746#endif
3747#ifdef CONFIG_LIVEPATCH
3748	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3749#endif
3750#ifdef CONFIG_PROC_PID_ARCH_STATUS
3751	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3752#endif
3753#ifdef CONFIG_SECCOMP_CACHE_DEBUG
3754	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3755#endif
3756#ifdef CONFIG_KSM
3757	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
3758	ONE("ksm_stat",  S_IRUSR, proc_pid_ksm_stat),
3759#endif
3760};
3761
3762static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
 
3763{
3764	return proc_pident_readdir(file, ctx,
3765				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3766}
3767
3768static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3769{
3770	return proc_pident_lookup(dir, dentry,
3771				  tid_base_stuff,
3772				  tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3773}
3774
3775static const struct file_operations proc_tid_base_operations = {
3776	.read		= generic_read_dir,
3777	.iterate_shared	= proc_tid_base_readdir,
3778	.llseek		= generic_file_llseek,
3779};
3780
3781static const struct inode_operations proc_tid_base_inode_operations = {
3782	.lookup		= proc_tid_base_lookup,
3783	.getattr	= pid_getattr,
3784	.setattr	= proc_setattr,
3785};
3786
3787static struct dentry *proc_task_instantiate(struct dentry *dentry,
3788	struct task_struct *task, const void *ptr)
3789{
 
3790	struct inode *inode;
3791	inode = proc_pid_make_base_inode(dentry->d_sb, task,
3792					 S_IFDIR | S_IRUGO | S_IXUGO);
3793	if (!inode)
3794		return ERR_PTR(-ENOENT);
3795
3796	inode->i_op = &proc_tid_base_inode_operations;
3797	inode->i_fop = &proc_tid_base_operations;
3798	inode->i_flags |= S_IMMUTABLE;
3799
3800	set_nlink(inode, nlink_tid);
3801	pid_update_inode(task, inode);
3802
3803	d_set_d_op(dentry, &pid_dentry_operations);
3804	return d_splice_alias(inode, dentry);
 
 
 
 
 
 
3805}
3806
3807static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3808{
 
3809	struct task_struct *task;
3810	struct task_struct *leader = get_proc_task(dir);
3811	unsigned tid;
3812	struct proc_fs_info *fs_info;
3813	struct pid_namespace *ns;
3814	struct dentry *result = ERR_PTR(-ENOENT);
3815
3816	if (!leader)
3817		goto out_no_task;
3818
3819	tid = name_to_int(&dentry->d_name);
3820	if (tid == ~0U)
3821		goto out;
3822
3823	fs_info = proc_sb_info(dentry->d_sb);
3824	ns = fs_info->pid_ns;
3825	rcu_read_lock();
3826	task = find_task_by_pid_ns(tid, ns);
3827	if (task)
3828		get_task_struct(task);
3829	rcu_read_unlock();
3830	if (!task)
3831		goto out;
3832	if (!same_thread_group(leader, task))
3833		goto out_drop_task;
3834
3835	result = proc_task_instantiate(dentry, task, NULL);
3836out_drop_task:
3837	put_task_struct(task);
3838out:
3839	put_task_struct(leader);
3840out_no_task:
3841	return result;
3842}
3843
3844/*
3845 * Find the first tid of a thread group to return to user space.
3846 *
3847 * Usually this is just the thread group leader, but if the users
3848 * buffer was too small or there was a seek into the middle of the
3849 * directory we have more work todo.
3850 *
3851 * In the case of a short read we start with find_task_by_pid.
3852 *
3853 * In the case of a seek we start with the leader and walk nr
3854 * threads past it.
3855 */
3856static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3857					struct pid_namespace *ns)
3858{
3859	struct task_struct *pos, *task;
3860	unsigned long nr = f_pos;
3861
3862	if (nr != f_pos)	/* 32bit overflow? */
3863		return NULL;
3864
3865	rcu_read_lock();
3866	task = pid_task(pid, PIDTYPE_PID);
3867	if (!task)
3868		goto fail;
3869
3870	/* Attempt to start with the tid of a thread */
3871	if (tid && nr) {
3872		pos = find_task_by_pid_ns(tid, ns);
3873		if (pos && same_thread_group(pos, task))
3874			goto found;
3875	}
3876
3877	/* If nr exceeds the number of threads there is nothing todo */
3878	if (nr >= get_nr_threads(task))
3879		goto fail;
 
3880
3881	/* If we haven't found our starting place yet start
3882	 * with the leader and walk nr threads forward.
3883	 */
3884	for_each_thread(task, pos) {
3885		if (!nr--)
3886			goto found;
 
 
 
3887	}
3888fail:
3889	pos = NULL;
3890	goto out;
3891found:
3892	get_task_struct(pos);
3893out:
3894	rcu_read_unlock();
3895	return pos;
3896}
3897
3898/*
3899 * Find the next thread in the thread list.
3900 * Return NULL if there is an error or no next thread.
3901 *
3902 * The reference to the input task_struct is released.
3903 */
3904static struct task_struct *next_tid(struct task_struct *start)
3905{
3906	struct task_struct *pos = NULL;
3907	rcu_read_lock();
3908	if (pid_alive(start)) {
3909		pos = __next_thread(start);
3910		if (pos)
 
 
3911			get_task_struct(pos);
3912	}
3913	rcu_read_unlock();
3914	put_task_struct(start);
3915	return pos;
3916}
3917
 
 
 
 
 
 
 
 
 
3918/* for the /proc/TGID/task/ directories */
3919static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3920{
3921	struct inode *inode = file_inode(file);
 
 
3922	struct task_struct *task;
 
 
 
3923	struct pid_namespace *ns;
3924	int tid;
3925
3926	if (proc_inode_is_dead(inode))
3927		return -ENOENT;
 
 
 
 
 
 
 
 
 
 
 
3928
3929	if (!dir_emit_dots(file, ctx))
3930		return 0;
 
 
 
 
 
 
 
 
 
 
 
 
3931
3932	/* We cache the tgid value that the last readdir call couldn't
3933	 * return and lseek resets it to 0.
3934	 */
3935	ns = proc_pid_ns(inode->i_sb);
3936	tid = (int)(intptr_t)file->private_data;
3937	file->private_data = NULL;
3938	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3939	     task;
3940	     task = next_tid(task), ctx->pos++) {
3941		char name[10 + 1];
3942		unsigned int len;
3943
3944		tid = task_pid_nr_ns(task, ns);
3945		if (!tid)
3946			continue;	/* The task has just exited. */
3947		len = snprintf(name, sizeof(name), "%u", tid);
3948		if (!proc_fill_cache(file, ctx, name, len,
3949				proc_task_instantiate, task, NULL)) {
3950			/* returning this tgid failed, save it as the first
3951			 * pid for the next readir call */
3952			file->private_data = (void *)(intptr_t)tid;
3953			put_task_struct(task);
3954			break;
3955		}
3956	}
3957
3958	return 0;
 
 
3959}
3960
3961static int proc_task_getattr(struct mnt_idmap *idmap,
3962			     const struct path *path, struct kstat *stat,
3963			     u32 request_mask, unsigned int query_flags)
3964{
3965	struct inode *inode = d_inode(path->dentry);
3966	struct task_struct *p = get_proc_task(inode);
3967	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
3968
3969	if (p) {
3970		stat->nlink += get_nr_threads(p);
3971		put_task_struct(p);
3972	}
3973
3974	return 0;
3975}
3976
3977/*
3978 * proc_task_readdir() set @file->private_data to a positive integer
3979 * value, so casting that to u64 is safe. generic_llseek_cookie() will
3980 * set @cookie to 0, so casting to an int is safe. The WARN_ON_ONCE() is
3981 * here to catch any unexpected change in behavior either in
3982 * proc_task_readdir() or generic_llseek_cookie().
3983 */
3984static loff_t proc_dir_llseek(struct file *file, loff_t offset, int whence)
3985{
3986	u64 cookie = (u64)(intptr_t)file->private_data;
3987	loff_t off;
3988
3989	off = generic_llseek_cookie(file, offset, whence, &cookie);
3990	WARN_ON_ONCE(cookie > INT_MAX);
3991	file->private_data = (void *)(intptr_t)cookie; /* serialized by f_pos_lock */
3992	return off;
3993}
3994
3995static const struct inode_operations proc_task_inode_operations = {
3996	.lookup		= proc_task_lookup,
3997	.getattr	= proc_task_getattr,
3998	.setattr	= proc_setattr,
3999	.permission	= proc_pid_permission,
4000};
4001
4002static const struct file_operations proc_task_operations = {
4003	.read		= generic_read_dir,
4004	.iterate_shared	= proc_task_readdir,
4005	.llseek		= proc_dir_llseek,
4006};
4007
4008void __init set_proc_pid_nlink(void)
4009{
4010	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
4011	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
4012}