Linux Audio

Check our new training course

Loading...
v5.14.15
   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/fdtable.h>
  62#include <linux/generic-radix-tree.h>
  63#include <linux/string.h>
  64#include <linux/seq_file.h>
  65#include <linux/namei.h>
  66#include <linux/mnt_namespace.h>
  67#include <linux/mm.h>
  68#include <linux/swap.h>
  69#include <linux/rcupdate.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/tracehook.h>
  77#include <linux/printk.h>
  78#include <linux/cache.h>
  79#include <linux/cgroup.h>
  80#include <linux/cpuset.h>
  81#include <linux/audit.h>
  82#include <linux/poll.h>
  83#include <linux/nsproxy.h>
  84#include <linux/oom.h>
  85#include <linux/elf.h>
  86#include <linux/pid_namespace.h>
  87#include <linux/user_namespace.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 <trace/events/oom.h>
  99#include "internal.h"
 100#include "fd.h"
 101
 102#include "../../lib/kstrtox.h"
 103
 104/* NOTE:
 105 *	Implementing inode permission operations in /proc is almost
 106 *	certainly an error.  Permission checks need to happen during
 107 *	each system call not at open time.  The reason is that most of
 108 *	what we wish to check for permissions in /proc varies at runtime.
 109 *
 110 *	The classic example of a problem is opening file descriptors
 111 *	in /proc for a task before it execs a suid executable.
 112 */
 113
 114static u8 nlink_tid __ro_after_init;
 115static u8 nlink_tgid __ro_after_init;
 116
 117struct pid_entry {
 118	const char *name;
 119	unsigned int len;
 120	umode_t mode;
 121	const struct inode_operations *iop;
 122	const struct file_operations *fop;
 123	union proc_op op;
 124};
 125
 126#define NOD(NAME, MODE, IOP, FOP, OP) {			\
 127	.name = (NAME),					\
 128	.len  = sizeof(NAME) - 1,			\
 129	.mode = MODE,					\
 130	.iop  = IOP,					\
 131	.fop  = FOP,					\
 132	.op   = OP,					\
 133}
 134
 135#define DIR(NAME, MODE, iops, fops)	\
 136	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
 137#define LNK(NAME, get_link)					\
 138	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
 139		&proc_pid_link_inode_operations, NULL,		\
 140		{ .proc_get_link = get_link } )
 141#define REG(NAME, MODE, fops)				\
 142	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
 143#define ONE(NAME, MODE, show)				\
 144	NOD(NAME, (S_IFREG|(MODE)),			\
 145		NULL, &proc_single_file_operations,	\
 146		{ .proc_show = show } )
 147#define ATTR(LSM, NAME, MODE)				\
 148	NOD(NAME, (S_IFREG|(MODE)),			\
 149		NULL, &proc_pid_attr_operations,	\
 150		{ .lsm = LSM })
 151
 152/*
 153 * Count the number of hardlinks for the pid_entry table, excluding the .
 154 * and .. links.
 155 */
 156static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
 157	unsigned int n)
 158{
 159	unsigned int i;
 160	unsigned int count;
 161
 162	count = 2;
 163	for (i = 0; i < n; ++i) {
 164		if (S_ISDIR(entries[i].mode))
 165			++count;
 166	}
 167
 168	return count;
 169}
 170
 171static int get_task_root(struct task_struct *task, struct path *root)
 172{
 173	int result = -ENOENT;
 174
 175	task_lock(task);
 176	if (task->fs) {
 177		get_fs_root(task->fs, root);
 178		result = 0;
 179	}
 180	task_unlock(task);
 181	return result;
 182}
 183
 184static int proc_cwd_link(struct dentry *dentry, struct path *path)
 185{
 186	struct task_struct *task = get_proc_task(d_inode(dentry));
 187	int result = -ENOENT;
 188
 189	if (task) {
 190		task_lock(task);
 191		if (task->fs) {
 192			get_fs_pwd(task->fs, path);
 193			result = 0;
 194		}
 195		task_unlock(task);
 196		put_task_struct(task);
 197	}
 198	return result;
 199}
 200
 201static int proc_root_link(struct dentry *dentry, struct path *path)
 202{
 203	struct task_struct *task = get_proc_task(d_inode(dentry));
 204	int result = -ENOENT;
 205
 206	if (task) {
 207		result = get_task_root(task, path);
 208		put_task_struct(task);
 209	}
 210	return result;
 211}
 212
 213/*
 214 * If the user used setproctitle(), we just get the string from
 215 * user space at arg_start, and limit it to a maximum of one page.
 216 */
 217static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
 218				size_t count, unsigned long pos,
 219				unsigned long arg_start)
 220{
 221	char *page;
 222	int ret, got;
 223
 224	if (pos >= PAGE_SIZE)
 225		return 0;
 226
 227	page = (char *)__get_free_page(GFP_KERNEL);
 228	if (!page)
 229		return -ENOMEM;
 230
 231	ret = 0;
 232	got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
 233	if (got > 0) {
 234		int len = strnlen(page, got);
 235
 236		/* Include the NUL character if it was found */
 237		if (len < got)
 238			len++;
 239
 240		if (len > pos) {
 241			len -= pos;
 242			if (len > count)
 243				len = count;
 244			len -= copy_to_user(buf, page+pos, len);
 245			if (!len)
 246				len = -EFAULT;
 247			ret = len;
 248		}
 249	}
 250	free_page((unsigned long)page);
 251	return ret;
 252}
 253
 254static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
 255			      size_t count, loff_t *ppos)
 256{
 257	unsigned long arg_start, arg_end, env_start, env_end;
 258	unsigned long pos, len;
 259	char *page, c;
 260
 261	/* Check if process spawned far enough to have cmdline. */
 262	if (!mm->env_end)
 263		return 0;
 264
 265	spin_lock(&mm->arg_lock);
 266	arg_start = mm->arg_start;
 267	arg_end = mm->arg_end;
 268	env_start = mm->env_start;
 269	env_end = mm->env_end;
 270	spin_unlock(&mm->arg_lock);
 271
 272	if (arg_start >= arg_end)
 273		return 0;
 274
 275	/*
 276	 * We allow setproctitle() to overwrite the argument
 277	 * strings, and overflow past the original end. But
 278	 * only when it overflows into the environment area.
 279	 */
 280	if (env_start != arg_end || env_end < env_start)
 281		env_start = env_end = arg_end;
 282	len = env_end - arg_start;
 283
 284	/* We're not going to care if "*ppos" has high bits set */
 285	pos = *ppos;
 286	if (pos >= len)
 287		return 0;
 288	if (count > len - pos)
 289		count = len - pos;
 290	if (!count)
 291		return 0;
 292
 293	/*
 294	 * Magical special case: if the argv[] end byte is not
 295	 * zero, the user has overwritten it with setproctitle(3).
 296	 *
 297	 * Possible future enhancement: do this only once when
 298	 * pos is 0, and set a flag in the 'struct file'.
 299	 */
 300	if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
 301		return get_mm_proctitle(mm, buf, count, pos, arg_start);
 302
 303	/*
 304	 * For the non-setproctitle() case we limit things strictly
 305	 * to the [arg_start, arg_end[ range.
 306	 */
 307	pos += arg_start;
 308	if (pos < arg_start || pos >= arg_end)
 309		return 0;
 310	if (count > arg_end - pos)
 311		count = arg_end - pos;
 312
 313	page = (char *)__get_free_page(GFP_KERNEL);
 314	if (!page)
 315		return -ENOMEM;
 316
 317	len = 0;
 318	while (count) {
 319		int got;
 320		size_t size = min_t(size_t, PAGE_SIZE, count);
 321
 322		got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
 323		if (got <= 0)
 324			break;
 325		got -= copy_to_user(buf, page, got);
 326		if (unlikely(!got)) {
 327			if (!len)
 328				len = -EFAULT;
 329			break;
 330		}
 331		pos += got;
 332		buf += got;
 333		len += got;
 334		count -= got;
 335	}
 336
 337	free_page((unsigned long)page);
 338	return len;
 339}
 340
 341static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
 342				size_t count, loff_t *pos)
 343{
 344	struct mm_struct *mm;
 345	ssize_t ret;
 346
 347	mm = get_task_mm(tsk);
 348	if (!mm)
 349		return 0;
 350
 351	ret = get_mm_cmdline(mm, buf, count, pos);
 352	mmput(mm);
 353	return ret;
 354}
 355
 356static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 357				     size_t count, loff_t *pos)
 358{
 359	struct task_struct *tsk;
 360	ssize_t ret;
 361
 362	BUG_ON(*pos < 0);
 363
 364	tsk = get_proc_task(file_inode(file));
 365	if (!tsk)
 366		return -ESRCH;
 367	ret = get_task_cmdline(tsk, buf, count, pos);
 368	put_task_struct(tsk);
 369	if (ret > 0)
 370		*pos += ret;
 371	return ret;
 372}
 373
 374static const struct file_operations proc_pid_cmdline_ops = {
 375	.read	= proc_pid_cmdline_read,
 376	.llseek	= generic_file_llseek,
 377};
 378
 379#ifdef CONFIG_KALLSYMS
 380/*
 381 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 382 * Returns the resolved symbol.  If that fails, simply return the address.
 383 */
 384static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
 385			  struct pid *pid, struct task_struct *task)
 386{
 387	unsigned long wchan;
 
 388
 389	if (ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
 390		wchan = get_wchan(task);
 391	else
 392		wchan = 0;
 393
 394	if (wchan)
 395		seq_printf(m, "%ps", (void *) wchan);
 396	else
 397		seq_putc(m, '0');
 
 398
 
 
 399	return 0;
 400}
 401#endif /* CONFIG_KALLSYMS */
 402
 403static int lock_trace(struct task_struct *task)
 404{
 405	int err = down_read_killable(&task->signal->exec_update_lock);
 406	if (err)
 407		return err;
 408	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
 409		up_read(&task->signal->exec_update_lock);
 410		return -EPERM;
 411	}
 412	return 0;
 413}
 414
 415static void unlock_trace(struct task_struct *task)
 416{
 417	up_read(&task->signal->exec_update_lock);
 418}
 419
 420#ifdef CONFIG_STACKTRACE
 421
 422#define MAX_STACK_TRACE_DEPTH	64
 423
 424static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
 425			  struct pid *pid, struct task_struct *task)
 426{
 427	unsigned long *entries;
 428	int err;
 429
 430	/*
 431	 * The ability to racily run the kernel stack unwinder on a running task
 432	 * and then observe the unwinder output is scary; while it is useful for
 433	 * debugging kernel issues, it can also allow an attacker to leak kernel
 434	 * stack contents.
 435	 * Doing this in a manner that is at least safe from races would require
 436	 * some work to ensure that the remote task can not be scheduled; and
 437	 * even then, this would still expose the unwinder as local attack
 438	 * surface.
 439	 * Therefore, this interface is restricted to root.
 440	 */
 441	if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
 442		return -EACCES;
 443
 444	entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
 445				GFP_KERNEL);
 446	if (!entries)
 447		return -ENOMEM;
 448
 449	err = lock_trace(task);
 450	if (!err) {
 451		unsigned int i, nr_entries;
 452
 453		nr_entries = stack_trace_save_tsk(task, entries,
 454						  MAX_STACK_TRACE_DEPTH, 0);
 455
 456		for (i = 0; i < nr_entries; i++) {
 457			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
 458		}
 459
 460		unlock_trace(task);
 461	}
 462	kfree(entries);
 463
 464	return err;
 465}
 466#endif
 467
 468#ifdef CONFIG_SCHED_INFO
 469/*
 470 * Provides /proc/PID/schedstat
 471 */
 472static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
 473			      struct pid *pid, struct task_struct *task)
 474{
 475	if (unlikely(!sched_info_on()))
 476		seq_puts(m, "0 0 0\n");
 477	else
 478		seq_printf(m, "%llu %llu %lu\n",
 479		   (unsigned long long)task->se.sum_exec_runtime,
 480		   (unsigned long long)task->sched_info.run_delay,
 481		   task->sched_info.pcount);
 482
 483	return 0;
 484}
 485#endif
 486
 487#ifdef CONFIG_LATENCYTOP
 488static int lstats_show_proc(struct seq_file *m, void *v)
 489{
 490	int i;
 491	struct inode *inode = m->private;
 492	struct task_struct *task = get_proc_task(inode);
 493
 494	if (!task)
 495		return -ESRCH;
 496	seq_puts(m, "Latency Top version : v0.1\n");
 497	for (i = 0; i < LT_SAVECOUNT; i++) {
 498		struct latency_record *lr = &task->latency_record[i];
 499		if (lr->backtrace[0]) {
 500			int q;
 501			seq_printf(m, "%i %li %li",
 502				   lr->count, lr->time, lr->max);
 503			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
 504				unsigned long bt = lr->backtrace[q];
 505
 506				if (!bt)
 507					break;
 508				seq_printf(m, " %ps", (void *)bt);
 509			}
 510			seq_putc(m, '\n');
 511		}
 512
 513	}
 514	put_task_struct(task);
 515	return 0;
 516}
 517
 518static int lstats_open(struct inode *inode, struct file *file)
 519{
 520	return single_open(file, lstats_show_proc, inode);
 521}
 522
 523static ssize_t lstats_write(struct file *file, const char __user *buf,
 524			    size_t count, loff_t *offs)
 525{
 526	struct task_struct *task = get_proc_task(file_inode(file));
 527
 528	if (!task)
 529		return -ESRCH;
 530	clear_tsk_latency_tracing(task);
 531	put_task_struct(task);
 532
 533	return count;
 534}
 535
 536static const struct file_operations proc_lstats_operations = {
 537	.open		= lstats_open,
 538	.read		= seq_read,
 539	.write		= lstats_write,
 540	.llseek		= seq_lseek,
 541	.release	= single_release,
 542};
 543
 544#endif
 545
 546static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 547			  struct pid *pid, struct task_struct *task)
 548{
 549	unsigned long totalpages = totalram_pages() + total_swap_pages;
 550	unsigned long points = 0;
 551	long badness;
 552
 553	badness = oom_badness(task, totalpages);
 554	/*
 555	 * Special case OOM_SCORE_ADJ_MIN for all others scale the
 556	 * badness value into [0, 2000] range which we have been
 557	 * exporting for a long time so userspace might depend on it.
 558	 */
 559	if (badness != LONG_MIN)
 560		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
 561
 562	seq_printf(m, "%lu\n", points);
 563
 564	return 0;
 565}
 566
 567struct limit_names {
 568	const char *name;
 569	const char *unit;
 570};
 571
 572static const struct limit_names lnames[RLIM_NLIMITS] = {
 573	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
 574	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
 575	[RLIMIT_DATA] = {"Max data size", "bytes"},
 576	[RLIMIT_STACK] = {"Max stack size", "bytes"},
 577	[RLIMIT_CORE] = {"Max core file size", "bytes"},
 578	[RLIMIT_RSS] = {"Max resident set", "bytes"},
 579	[RLIMIT_NPROC] = {"Max processes", "processes"},
 580	[RLIMIT_NOFILE] = {"Max open files", "files"},
 581	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
 582	[RLIMIT_AS] = {"Max address space", "bytes"},
 583	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
 584	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
 585	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
 586	[RLIMIT_NICE] = {"Max nice priority", NULL},
 587	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
 588	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
 589};
 590
 591/* Display limits for a process */
 592static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
 593			   struct pid *pid, struct task_struct *task)
 594{
 595	unsigned int i;
 596	unsigned long flags;
 597
 598	struct rlimit rlim[RLIM_NLIMITS];
 599
 600	if (!lock_task_sighand(task, &flags))
 601		return 0;
 602	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
 603	unlock_task_sighand(task, &flags);
 604
 605	/*
 606	 * print the file header
 607	 */
 608	seq_puts(m, "Limit                     "
 609		"Soft Limit           "
 610		"Hard Limit           "
 611		"Units     \n");
 612
 613	for (i = 0; i < RLIM_NLIMITS; i++) {
 614		if (rlim[i].rlim_cur == RLIM_INFINITY)
 615			seq_printf(m, "%-25s %-20s ",
 616				   lnames[i].name, "unlimited");
 617		else
 618			seq_printf(m, "%-25s %-20lu ",
 619				   lnames[i].name, rlim[i].rlim_cur);
 620
 621		if (rlim[i].rlim_max == RLIM_INFINITY)
 622			seq_printf(m, "%-20s ", "unlimited");
 623		else
 624			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
 625
 626		if (lnames[i].unit)
 627			seq_printf(m, "%-10s\n", lnames[i].unit);
 628		else
 629			seq_putc(m, '\n');
 630	}
 631
 632	return 0;
 633}
 634
 635#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 636static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
 637			    struct pid *pid, struct task_struct *task)
 638{
 639	struct syscall_info info;
 640	u64 *args = &info.data.args[0];
 641	int res;
 642
 643	res = lock_trace(task);
 644	if (res)
 645		return res;
 646
 647	if (task_current_syscall(task, &info))
 648		seq_puts(m, "running\n");
 649	else if (info.data.nr < 0)
 650		seq_printf(m, "%d 0x%llx 0x%llx\n",
 651			   info.data.nr, info.sp, info.data.instruction_pointer);
 652	else
 653		seq_printf(m,
 654		       "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
 655		       info.data.nr,
 656		       args[0], args[1], args[2], args[3], args[4], args[5],
 657		       info.sp, info.data.instruction_pointer);
 658	unlock_trace(task);
 659
 660	return 0;
 661}
 662#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
 663
 664/************************************************************************/
 665/*                       Here the fs part begins                        */
 666/************************************************************************/
 667
 668/* permission checks */
 669static int proc_fd_access_allowed(struct inode *inode)
 670{
 671	struct task_struct *task;
 672	int allowed = 0;
 673	/* Allow access to a task's file descriptors if it is us or we
 674	 * may use ptrace attach to the process and find out that
 675	 * information.
 676	 */
 677	task = get_proc_task(inode);
 678	if (task) {
 679		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 680		put_task_struct(task);
 681	}
 682	return allowed;
 683}
 684
 685int proc_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 686		 struct iattr *attr)
 687{
 688	int error;
 689	struct inode *inode = d_inode(dentry);
 690
 691	if (attr->ia_valid & ATTR_MODE)
 692		return -EPERM;
 693
 694	error = setattr_prepare(&init_user_ns, dentry, attr);
 695	if (error)
 696		return error;
 697
 698	setattr_copy(&init_user_ns, inode, attr);
 699	mark_inode_dirty(inode);
 700	return 0;
 701}
 702
 703/*
 704 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
 705 * or euid/egid (for hide_pid_min=2)?
 706 */
 707static bool has_pid_permissions(struct proc_fs_info *fs_info,
 708				 struct task_struct *task,
 709				 enum proc_hidepid hide_pid_min)
 710{
 711	/*
 712	 * If 'hidpid' mount option is set force a ptrace check,
 713	 * we indicate that we are using a filesystem syscall
 714	 * by passing PTRACE_MODE_READ_FSCREDS
 715	 */
 716	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
 717		return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 718
 719	if (fs_info->hide_pid < hide_pid_min)
 720		return true;
 721	if (in_group_p(fs_info->pid_gid))
 722		return true;
 723	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 724}
 725
 726
 727static int proc_pid_permission(struct user_namespace *mnt_userns,
 728			       struct inode *inode, int mask)
 729{
 730	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
 731	struct task_struct *task;
 732	bool has_perms;
 733
 734	task = get_proc_task(inode);
 735	if (!task)
 736		return -ESRCH;
 737	has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
 738	put_task_struct(task);
 739
 740	if (!has_perms) {
 741		if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
 742			/*
 743			 * Let's make getdents(), stat(), and open()
 744			 * consistent with each other.  If a process
 745			 * may not stat() a file, it shouldn't be seen
 746			 * in procfs at all.
 747			 */
 748			return -ENOENT;
 749		}
 750
 751		return -EPERM;
 752	}
 753	return generic_permission(&init_user_ns, inode, mask);
 754}
 755
 756
 757
 758static const struct inode_operations proc_def_inode_operations = {
 759	.setattr	= proc_setattr,
 760};
 761
 762static int proc_single_show(struct seq_file *m, void *v)
 763{
 764	struct inode *inode = m->private;
 765	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
 766	struct pid *pid = proc_pid(inode);
 767	struct task_struct *task;
 768	int ret;
 769
 770	task = get_pid_task(pid, PIDTYPE_PID);
 771	if (!task)
 772		return -ESRCH;
 773
 774	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
 775
 776	put_task_struct(task);
 777	return ret;
 778}
 779
 780static int proc_single_open(struct inode *inode, struct file *filp)
 781{
 782	return single_open(filp, proc_single_show, inode);
 783}
 784
 785static const struct file_operations proc_single_file_operations = {
 786	.open		= proc_single_open,
 787	.read		= seq_read,
 788	.llseek		= seq_lseek,
 789	.release	= single_release,
 790};
 791
 792
 793struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
 794{
 795	struct task_struct *task = get_proc_task(inode);
 796	struct mm_struct *mm = ERR_PTR(-ESRCH);
 797
 798	if (task) {
 799		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
 800		put_task_struct(task);
 801
 802		if (!IS_ERR_OR_NULL(mm)) {
 803			/* ensure this mm_struct can't be freed */
 804			mmgrab(mm);
 805			/* but do not pin its memory */
 806			mmput(mm);
 807		}
 808	}
 809
 810	return mm;
 811}
 812
 813static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
 814{
 815	struct mm_struct *mm = proc_mem_open(inode, mode);
 816
 817	if (IS_ERR(mm))
 818		return PTR_ERR(mm);
 819
 820	file->private_data = mm;
 821	return 0;
 822}
 823
 824static int mem_open(struct inode *inode, struct file *file)
 825{
 826	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
 827
 828	/* OK to pass negative loff_t, we can catch out-of-range */
 829	file->f_mode |= FMODE_UNSIGNED_OFFSET;
 830
 831	return ret;
 832}
 833
 834static ssize_t mem_rw(struct file *file, char __user *buf,
 835			size_t count, loff_t *ppos, int write)
 836{
 837	struct mm_struct *mm = file->private_data;
 838	unsigned long addr = *ppos;
 839	ssize_t copied;
 840	char *page;
 841	unsigned int flags;
 842
 843	if (!mm)
 844		return 0;
 845
 846	page = (char *)__get_free_page(GFP_KERNEL);
 847	if (!page)
 848		return -ENOMEM;
 849
 850	copied = 0;
 851	if (!mmget_not_zero(mm))
 852		goto free;
 853
 854	flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
 855
 856	while (count > 0) {
 857		size_t this_len = min_t(size_t, count, PAGE_SIZE);
 858
 859		if (write && copy_from_user(page, buf, this_len)) {
 860			copied = -EFAULT;
 861			break;
 862		}
 863
 864		this_len = access_remote_vm(mm, addr, page, this_len, flags);
 865		if (!this_len) {
 866			if (!copied)
 867				copied = -EIO;
 868			break;
 869		}
 870
 871		if (!write && copy_to_user(buf, page, this_len)) {
 872			copied = -EFAULT;
 873			break;
 874		}
 875
 876		buf += this_len;
 877		addr += this_len;
 878		copied += this_len;
 879		count -= this_len;
 880	}
 881	*ppos = addr;
 882
 883	mmput(mm);
 884free:
 885	free_page((unsigned long) page);
 886	return copied;
 887}
 888
 889static ssize_t mem_read(struct file *file, char __user *buf,
 890			size_t count, loff_t *ppos)
 891{
 892	return mem_rw(file, buf, count, ppos, 0);
 893}
 894
 895static ssize_t mem_write(struct file *file, const char __user *buf,
 896			 size_t count, loff_t *ppos)
 897{
 898	return mem_rw(file, (char __user*)buf, count, ppos, 1);
 899}
 900
 901loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 902{
 903	switch (orig) {
 904	case 0:
 905		file->f_pos = offset;
 906		break;
 907	case 1:
 908		file->f_pos += offset;
 909		break;
 910	default:
 911		return -EINVAL;
 912	}
 913	force_successful_syscall_return();
 914	return file->f_pos;
 915}
 916
 917static int mem_release(struct inode *inode, struct file *file)
 918{
 919	struct mm_struct *mm = file->private_data;
 920	if (mm)
 921		mmdrop(mm);
 922	return 0;
 923}
 924
 925static const struct file_operations proc_mem_operations = {
 926	.llseek		= mem_lseek,
 927	.read		= mem_read,
 928	.write		= mem_write,
 929	.open		= mem_open,
 930	.release	= mem_release,
 931};
 932
 933static int environ_open(struct inode *inode, struct file *file)
 934{
 935	return __mem_open(inode, file, PTRACE_MODE_READ);
 936}
 937
 938static ssize_t environ_read(struct file *file, char __user *buf,
 939			size_t count, loff_t *ppos)
 940{
 941	char *page;
 942	unsigned long src = *ppos;
 943	int ret = 0;
 944	struct mm_struct *mm = file->private_data;
 945	unsigned long env_start, env_end;
 946
 947	/* Ensure the process spawned far enough to have an environment. */
 948	if (!mm || !mm->env_end)
 949		return 0;
 950
 951	page = (char *)__get_free_page(GFP_KERNEL);
 952	if (!page)
 953		return -ENOMEM;
 954
 955	ret = 0;
 956	if (!mmget_not_zero(mm))
 957		goto free;
 958
 959	spin_lock(&mm->arg_lock);
 960	env_start = mm->env_start;
 961	env_end = mm->env_end;
 962	spin_unlock(&mm->arg_lock);
 963
 964	while (count > 0) {
 965		size_t this_len, max_len;
 966		int retval;
 967
 968		if (src >= (env_end - env_start))
 969			break;
 970
 971		this_len = env_end - (env_start + src);
 972
 973		max_len = min_t(size_t, PAGE_SIZE, count);
 974		this_len = min(max_len, this_len);
 975
 976		retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
 977
 978		if (retval <= 0) {
 979			ret = retval;
 980			break;
 981		}
 982
 983		if (copy_to_user(buf, page, retval)) {
 984			ret = -EFAULT;
 985			break;
 986		}
 987
 988		ret += retval;
 989		src += retval;
 990		buf += retval;
 991		count -= retval;
 992	}
 993	*ppos = src;
 994	mmput(mm);
 995
 996free:
 997	free_page((unsigned long) page);
 998	return ret;
 999}
1000
1001static const struct file_operations proc_environ_operations = {
1002	.open		= environ_open,
1003	.read		= environ_read,
1004	.llseek		= generic_file_llseek,
1005	.release	= mem_release,
1006};
1007
1008static int auxv_open(struct inode *inode, struct file *file)
1009{
1010	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1011}
1012
1013static ssize_t auxv_read(struct file *file, char __user *buf,
1014			size_t count, loff_t *ppos)
1015{
1016	struct mm_struct *mm = file->private_data;
1017	unsigned int nwords = 0;
1018
1019	if (!mm)
1020		return 0;
1021	do {
1022		nwords += 2;
1023	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1024	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1025				       nwords * sizeof(mm->saved_auxv[0]));
1026}
1027
1028static const struct file_operations proc_auxv_operations = {
1029	.open		= auxv_open,
1030	.read		= auxv_read,
1031	.llseek		= generic_file_llseek,
1032	.release	= mem_release,
1033};
1034
1035static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1036			    loff_t *ppos)
1037{
1038	struct task_struct *task = get_proc_task(file_inode(file));
1039	char buffer[PROC_NUMBUF];
1040	int oom_adj = OOM_ADJUST_MIN;
1041	size_t len;
1042
1043	if (!task)
1044		return -ESRCH;
1045	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1046		oom_adj = OOM_ADJUST_MAX;
1047	else
1048		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1049			  OOM_SCORE_ADJ_MAX;
1050	put_task_struct(task);
1051	if (oom_adj > OOM_ADJUST_MAX)
1052		oom_adj = OOM_ADJUST_MAX;
1053	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1054	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1055}
1056
1057static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1058{
 
1059	struct mm_struct *mm = NULL;
1060	struct task_struct *task;
1061	int err = 0;
1062
1063	task = get_proc_task(file_inode(file));
1064	if (!task)
1065		return -ESRCH;
1066
1067	mutex_lock(&oom_adj_mutex);
1068	if (legacy) {
1069		if (oom_adj < task->signal->oom_score_adj &&
1070				!capable(CAP_SYS_RESOURCE)) {
1071			err = -EACCES;
1072			goto err_unlock;
1073		}
1074		/*
1075		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1076		 * /proc/pid/oom_score_adj instead.
1077		 */
1078		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1079			  current->comm, task_pid_nr(current), task_pid_nr(task),
1080			  task_pid_nr(task));
1081	} else {
1082		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1083				!capable(CAP_SYS_RESOURCE)) {
1084			err = -EACCES;
1085			goto err_unlock;
1086		}
1087	}
1088
1089	/*
1090	 * Make sure we will check other processes sharing the mm if this is
1091	 * not vfrok which wants its own oom_score_adj.
1092	 * pin the mm so it doesn't go away and get reused after task_unlock
1093	 */
1094	if (!task->vfork_done) {
1095		struct task_struct *p = find_lock_task_mm(task);
1096
1097		if (p) {
1098			if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1099				mm = p->mm;
1100				mmgrab(mm);
1101			}
1102			task_unlock(p);
1103		}
1104	}
1105
1106	task->signal->oom_score_adj = oom_adj;
1107	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1108		task->signal->oom_score_adj_min = (short)oom_adj;
1109	trace_oom_score_adj_update(task);
1110
1111	if (mm) {
1112		struct task_struct *p;
1113
1114		rcu_read_lock();
1115		for_each_process(p) {
1116			if (same_thread_group(task, p))
1117				continue;
1118
1119			/* do not touch kernel threads or the global init */
1120			if (p->flags & PF_KTHREAD || is_global_init(p))
1121				continue;
1122
1123			task_lock(p);
1124			if (!p->vfork_done && process_shares_mm(p, mm)) {
1125				p->signal->oom_score_adj = oom_adj;
1126				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1127					p->signal->oom_score_adj_min = (short)oom_adj;
1128			}
1129			task_unlock(p);
1130		}
1131		rcu_read_unlock();
1132		mmdrop(mm);
1133	}
1134err_unlock:
1135	mutex_unlock(&oom_adj_mutex);
1136	put_task_struct(task);
1137	return err;
1138}
1139
1140/*
1141 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1142 * kernels.  The effective policy is defined by oom_score_adj, which has a
1143 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1144 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1145 * Processes that become oom disabled via oom_adj will still be oom disabled
1146 * with this implementation.
1147 *
1148 * oom_adj cannot be removed since existing userspace binaries use it.
1149 */
1150static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1151			     size_t count, loff_t *ppos)
1152{
1153	char buffer[PROC_NUMBUF];
1154	int oom_adj;
1155	int err;
1156
1157	memset(buffer, 0, sizeof(buffer));
1158	if (count > sizeof(buffer) - 1)
1159		count = sizeof(buffer) - 1;
1160	if (copy_from_user(buffer, buf, count)) {
1161		err = -EFAULT;
1162		goto out;
1163	}
1164
1165	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1166	if (err)
1167		goto out;
1168	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1169	     oom_adj != OOM_DISABLE) {
1170		err = -EINVAL;
1171		goto out;
1172	}
1173
1174	/*
1175	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1176	 * value is always attainable.
1177	 */
1178	if (oom_adj == OOM_ADJUST_MAX)
1179		oom_adj = OOM_SCORE_ADJ_MAX;
1180	else
1181		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1182
1183	err = __set_oom_adj(file, oom_adj, true);
1184out:
1185	return err < 0 ? err : count;
1186}
1187
1188static const struct file_operations proc_oom_adj_operations = {
1189	.read		= oom_adj_read,
1190	.write		= oom_adj_write,
1191	.llseek		= generic_file_llseek,
1192};
1193
1194static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1195					size_t count, loff_t *ppos)
1196{
1197	struct task_struct *task = get_proc_task(file_inode(file));
1198	char buffer[PROC_NUMBUF];
1199	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1200	size_t len;
1201
1202	if (!task)
1203		return -ESRCH;
1204	oom_score_adj = task->signal->oom_score_adj;
1205	put_task_struct(task);
1206	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1207	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1208}
1209
1210static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1211					size_t count, loff_t *ppos)
1212{
1213	char buffer[PROC_NUMBUF];
1214	int oom_score_adj;
1215	int err;
1216
1217	memset(buffer, 0, sizeof(buffer));
1218	if (count > sizeof(buffer) - 1)
1219		count = sizeof(buffer) - 1;
1220	if (copy_from_user(buffer, buf, count)) {
1221		err = -EFAULT;
1222		goto out;
1223	}
1224
1225	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1226	if (err)
1227		goto out;
1228	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1229			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1230		err = -EINVAL;
1231		goto out;
1232	}
1233
1234	err = __set_oom_adj(file, oom_score_adj, false);
1235out:
1236	return err < 0 ? err : count;
1237}
1238
1239static const struct file_operations proc_oom_score_adj_operations = {
1240	.read		= oom_score_adj_read,
1241	.write		= oom_score_adj_write,
1242	.llseek		= default_llseek,
1243};
1244
1245#ifdef CONFIG_AUDIT
1246#define TMPBUFLEN 11
1247static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1248				  size_t count, loff_t *ppos)
1249{
1250	struct inode * inode = file_inode(file);
1251	struct task_struct *task = get_proc_task(inode);
1252	ssize_t length;
1253	char tmpbuf[TMPBUFLEN];
1254
1255	if (!task)
1256		return -ESRCH;
1257	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1258			   from_kuid(file->f_cred->user_ns,
1259				     audit_get_loginuid(task)));
1260	put_task_struct(task);
1261	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1262}
1263
1264static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1265				   size_t count, loff_t *ppos)
1266{
1267	struct inode * inode = file_inode(file);
1268	uid_t loginuid;
1269	kuid_t kloginuid;
1270	int rv;
1271
1272	/* Don't let kthreads write their own loginuid */
1273	if (current->flags & PF_KTHREAD)
1274		return -EPERM;
1275
1276	rcu_read_lock();
1277	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1278		rcu_read_unlock();
1279		return -EPERM;
1280	}
1281	rcu_read_unlock();
1282
1283	if (*ppos != 0) {
1284		/* No partial writes. */
1285		return -EINVAL;
1286	}
1287
1288	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1289	if (rv < 0)
1290		return rv;
1291
1292	/* is userspace tring to explicitly UNSET the loginuid? */
1293	if (loginuid == AUDIT_UID_UNSET) {
1294		kloginuid = INVALID_UID;
1295	} else {
1296		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1297		if (!uid_valid(kloginuid))
1298			return -EINVAL;
1299	}
1300
1301	rv = audit_set_loginuid(kloginuid);
1302	if (rv < 0)
1303		return rv;
1304	return count;
1305}
1306
1307static const struct file_operations proc_loginuid_operations = {
1308	.read		= proc_loginuid_read,
1309	.write		= proc_loginuid_write,
1310	.llseek		= generic_file_llseek,
1311};
1312
1313static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1314				  size_t count, loff_t *ppos)
1315{
1316	struct inode * inode = file_inode(file);
1317	struct task_struct *task = get_proc_task(inode);
1318	ssize_t length;
1319	char tmpbuf[TMPBUFLEN];
1320
1321	if (!task)
1322		return -ESRCH;
1323	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1324				audit_get_sessionid(task));
1325	put_task_struct(task);
1326	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1327}
1328
1329static const struct file_operations proc_sessionid_operations = {
1330	.read		= proc_sessionid_read,
1331	.llseek		= generic_file_llseek,
1332};
1333#endif
1334
1335#ifdef CONFIG_FAULT_INJECTION
1336static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1337				      size_t count, loff_t *ppos)
1338{
1339	struct task_struct *task = get_proc_task(file_inode(file));
1340	char buffer[PROC_NUMBUF];
1341	size_t len;
1342	int make_it_fail;
1343
1344	if (!task)
1345		return -ESRCH;
1346	make_it_fail = task->make_it_fail;
1347	put_task_struct(task);
1348
1349	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1350
1351	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1352}
1353
1354static ssize_t proc_fault_inject_write(struct file * file,
1355			const char __user * buf, size_t count, loff_t *ppos)
1356{
1357	struct task_struct *task;
1358	char buffer[PROC_NUMBUF];
1359	int make_it_fail;
1360	int rv;
1361
1362	if (!capable(CAP_SYS_RESOURCE))
1363		return -EPERM;
1364	memset(buffer, 0, sizeof(buffer));
1365	if (count > sizeof(buffer) - 1)
1366		count = sizeof(buffer) - 1;
1367	if (copy_from_user(buffer, buf, count))
1368		return -EFAULT;
1369	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1370	if (rv < 0)
1371		return rv;
1372	if (make_it_fail < 0 || make_it_fail > 1)
1373		return -EINVAL;
1374
1375	task = get_proc_task(file_inode(file));
1376	if (!task)
1377		return -ESRCH;
1378	task->make_it_fail = make_it_fail;
1379	put_task_struct(task);
1380
1381	return count;
1382}
1383
1384static const struct file_operations proc_fault_inject_operations = {
1385	.read		= proc_fault_inject_read,
1386	.write		= proc_fault_inject_write,
1387	.llseek		= generic_file_llseek,
1388};
1389
1390static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1391				   size_t count, loff_t *ppos)
1392{
1393	struct task_struct *task;
1394	int err;
1395	unsigned int n;
1396
1397	err = kstrtouint_from_user(buf, count, 0, &n);
1398	if (err)
1399		return err;
1400
1401	task = get_proc_task(file_inode(file));
1402	if (!task)
1403		return -ESRCH;
1404	task->fail_nth = n;
1405	put_task_struct(task);
1406
1407	return count;
1408}
1409
1410static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1411				  size_t count, loff_t *ppos)
1412{
1413	struct task_struct *task;
1414	char numbuf[PROC_NUMBUF];
1415	ssize_t len;
1416
1417	task = get_proc_task(file_inode(file));
1418	if (!task)
1419		return -ESRCH;
1420	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1421	put_task_struct(task);
1422	return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1423}
1424
1425static const struct file_operations proc_fail_nth_operations = {
1426	.read		= proc_fail_nth_read,
1427	.write		= proc_fail_nth_write,
1428};
1429#endif
1430
1431
1432#ifdef CONFIG_SCHED_DEBUG
1433/*
1434 * Print out various scheduling related per-task fields:
1435 */
1436static int sched_show(struct seq_file *m, void *v)
1437{
1438	struct inode *inode = m->private;
1439	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1440	struct task_struct *p;
1441
1442	p = get_proc_task(inode);
1443	if (!p)
1444		return -ESRCH;
1445	proc_sched_show_task(p, ns, m);
1446
1447	put_task_struct(p);
1448
1449	return 0;
1450}
1451
1452static ssize_t
1453sched_write(struct file *file, const char __user *buf,
1454	    size_t count, loff_t *offset)
1455{
1456	struct inode *inode = file_inode(file);
1457	struct task_struct *p;
1458
1459	p = get_proc_task(inode);
1460	if (!p)
1461		return -ESRCH;
1462	proc_sched_set_task(p);
1463
1464	put_task_struct(p);
1465
1466	return count;
1467}
1468
1469static int sched_open(struct inode *inode, struct file *filp)
1470{
1471	return single_open(filp, sched_show, inode);
1472}
1473
1474static const struct file_operations proc_pid_sched_operations = {
1475	.open		= sched_open,
1476	.read		= seq_read,
1477	.write		= sched_write,
1478	.llseek		= seq_lseek,
1479	.release	= single_release,
1480};
1481
1482#endif
1483
1484#ifdef CONFIG_SCHED_AUTOGROUP
1485/*
1486 * Print out autogroup related information:
1487 */
1488static int sched_autogroup_show(struct seq_file *m, void *v)
1489{
1490	struct inode *inode = m->private;
1491	struct task_struct *p;
1492
1493	p = get_proc_task(inode);
1494	if (!p)
1495		return -ESRCH;
1496	proc_sched_autogroup_show_task(p, m);
1497
1498	put_task_struct(p);
1499
1500	return 0;
1501}
1502
1503static ssize_t
1504sched_autogroup_write(struct file *file, const char __user *buf,
1505	    size_t count, loff_t *offset)
1506{
1507	struct inode *inode = file_inode(file);
1508	struct task_struct *p;
1509	char buffer[PROC_NUMBUF];
1510	int nice;
1511	int err;
1512
1513	memset(buffer, 0, sizeof(buffer));
1514	if (count > sizeof(buffer) - 1)
1515		count = sizeof(buffer) - 1;
1516	if (copy_from_user(buffer, buf, count))
1517		return -EFAULT;
1518
1519	err = kstrtoint(strstrip(buffer), 0, &nice);
1520	if (err < 0)
1521		return err;
1522
1523	p = get_proc_task(inode);
1524	if (!p)
1525		return -ESRCH;
1526
1527	err = proc_sched_autogroup_set_nice(p, nice);
1528	if (err)
1529		count = err;
1530
1531	put_task_struct(p);
1532
1533	return count;
1534}
1535
1536static int sched_autogroup_open(struct inode *inode, struct file *filp)
1537{
1538	int ret;
1539
1540	ret = single_open(filp, sched_autogroup_show, NULL);
1541	if (!ret) {
1542		struct seq_file *m = filp->private_data;
1543
1544		m->private = inode;
1545	}
1546	return ret;
1547}
1548
1549static const struct file_operations proc_pid_sched_autogroup_operations = {
1550	.open		= sched_autogroup_open,
1551	.read		= seq_read,
1552	.write		= sched_autogroup_write,
1553	.llseek		= seq_lseek,
1554	.release	= single_release,
1555};
1556
1557#endif /* CONFIG_SCHED_AUTOGROUP */
1558
1559#ifdef CONFIG_TIME_NS
1560static int timens_offsets_show(struct seq_file *m, void *v)
1561{
1562	struct task_struct *p;
1563
1564	p = get_proc_task(file_inode(m->file));
1565	if (!p)
1566		return -ESRCH;
1567	proc_timens_show_offsets(p, m);
1568
1569	put_task_struct(p);
1570
1571	return 0;
1572}
1573
1574static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1575				    size_t count, loff_t *ppos)
1576{
1577	struct inode *inode = file_inode(file);
1578	struct proc_timens_offset offsets[2];
1579	char *kbuf = NULL, *pos, *next_line;
1580	struct task_struct *p;
1581	int ret, noffsets;
1582
1583	/* Only allow < page size writes at the beginning of the file */
1584	if ((*ppos != 0) || (count >= PAGE_SIZE))
1585		return -EINVAL;
1586
1587	/* Slurp in the user data */
1588	kbuf = memdup_user_nul(buf, count);
1589	if (IS_ERR(kbuf))
1590		return PTR_ERR(kbuf);
1591
1592	/* Parse the user data */
1593	ret = -EINVAL;
1594	noffsets = 0;
1595	for (pos = kbuf; pos; pos = next_line) {
1596		struct proc_timens_offset *off = &offsets[noffsets];
1597		char clock[10];
1598		int err;
1599
1600		/* Find the end of line and ensure we don't look past it */
1601		next_line = strchr(pos, '\n');
1602		if (next_line) {
1603			*next_line = '\0';
1604			next_line++;
1605			if (*next_line == '\0')
1606				next_line = NULL;
1607		}
1608
1609		err = sscanf(pos, "%9s %lld %lu", clock,
1610				&off->val.tv_sec, &off->val.tv_nsec);
1611		if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1612			goto out;
1613
1614		clock[sizeof(clock) - 1] = 0;
1615		if (strcmp(clock, "monotonic") == 0 ||
1616		    strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1617			off->clockid = CLOCK_MONOTONIC;
1618		else if (strcmp(clock, "boottime") == 0 ||
1619			 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1620			off->clockid = CLOCK_BOOTTIME;
1621		else
1622			goto out;
1623
1624		noffsets++;
1625		if (noffsets == ARRAY_SIZE(offsets)) {
1626			if (next_line)
1627				count = next_line - kbuf;
1628			break;
1629		}
1630	}
1631
1632	ret = -ESRCH;
1633	p = get_proc_task(inode);
1634	if (!p)
1635		goto out;
1636	ret = proc_timens_set_offset(file, p, offsets, noffsets);
1637	put_task_struct(p);
1638	if (ret)
1639		goto out;
1640
1641	ret = count;
1642out:
1643	kfree(kbuf);
1644	return ret;
1645}
1646
1647static int timens_offsets_open(struct inode *inode, struct file *filp)
1648{
1649	return single_open(filp, timens_offsets_show, inode);
1650}
1651
1652static const struct file_operations proc_timens_offsets_operations = {
1653	.open		= timens_offsets_open,
1654	.read		= seq_read,
1655	.write		= timens_offsets_write,
1656	.llseek		= seq_lseek,
1657	.release	= single_release,
1658};
1659#endif /* CONFIG_TIME_NS */
1660
1661static ssize_t comm_write(struct file *file, const char __user *buf,
1662				size_t count, loff_t *offset)
1663{
1664	struct inode *inode = file_inode(file);
1665	struct task_struct *p;
1666	char buffer[TASK_COMM_LEN];
1667	const size_t maxlen = sizeof(buffer) - 1;
1668
1669	memset(buffer, 0, sizeof(buffer));
1670	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1671		return -EFAULT;
1672
1673	p = get_proc_task(inode);
1674	if (!p)
1675		return -ESRCH;
1676
1677	if (same_thread_group(current, p))
1678		set_task_comm(p, buffer);
1679	else
1680		count = -EINVAL;
1681
1682	put_task_struct(p);
1683
1684	return count;
1685}
1686
1687static int comm_show(struct seq_file *m, void *v)
1688{
1689	struct inode *inode = m->private;
1690	struct task_struct *p;
1691
1692	p = get_proc_task(inode);
1693	if (!p)
1694		return -ESRCH;
1695
1696	proc_task_name(m, p, false);
1697	seq_putc(m, '\n');
1698
1699	put_task_struct(p);
1700
1701	return 0;
1702}
1703
1704static int comm_open(struct inode *inode, struct file *filp)
1705{
1706	return single_open(filp, comm_show, inode);
1707}
1708
1709static const struct file_operations proc_pid_set_comm_operations = {
1710	.open		= comm_open,
1711	.read		= seq_read,
1712	.write		= comm_write,
1713	.llseek		= seq_lseek,
1714	.release	= single_release,
1715};
1716
1717static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1718{
1719	struct task_struct *task;
1720	struct file *exe_file;
1721
1722	task = get_proc_task(d_inode(dentry));
1723	if (!task)
1724		return -ENOENT;
1725	exe_file = get_task_exe_file(task);
1726	put_task_struct(task);
1727	if (exe_file) {
1728		*exe_path = exe_file->f_path;
1729		path_get(&exe_file->f_path);
1730		fput(exe_file);
1731		return 0;
1732	} else
1733		return -ENOENT;
1734}
1735
1736static const char *proc_pid_get_link(struct dentry *dentry,
1737				     struct inode *inode,
1738				     struct delayed_call *done)
1739{
1740	struct path path;
1741	int error = -EACCES;
1742
1743	if (!dentry)
1744		return ERR_PTR(-ECHILD);
1745
1746	/* Are we allowed to snoop on the tasks file descriptors? */
1747	if (!proc_fd_access_allowed(inode))
1748		goto out;
1749
1750	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1751	if (error)
1752		goto out;
1753
1754	error = nd_jump_link(&path);
1755out:
1756	return ERR_PTR(error);
1757}
1758
1759static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1760{
1761	char *tmp = (char *)__get_free_page(GFP_KERNEL);
1762	char *pathname;
1763	int len;
1764
1765	if (!tmp)
1766		return -ENOMEM;
1767
1768	pathname = d_path(path, tmp, PAGE_SIZE);
1769	len = PTR_ERR(pathname);
1770	if (IS_ERR(pathname))
1771		goto out;
1772	len = tmp + PAGE_SIZE - 1 - pathname;
1773
1774	if (len > buflen)
1775		len = buflen;
1776	if (copy_to_user(buffer, pathname, len))
1777		len = -EFAULT;
1778 out:
1779	free_page((unsigned long)tmp);
1780	return len;
1781}
1782
1783static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1784{
1785	int error = -EACCES;
1786	struct inode *inode = d_inode(dentry);
1787	struct path path;
1788
1789	/* Are we allowed to snoop on the tasks file descriptors? */
1790	if (!proc_fd_access_allowed(inode))
1791		goto out;
1792
1793	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1794	if (error)
1795		goto out;
1796
1797	error = do_proc_readlink(&path, buffer, buflen);
1798	path_put(&path);
1799out:
1800	return error;
1801}
1802
1803const struct inode_operations proc_pid_link_inode_operations = {
1804	.readlink	= proc_pid_readlink,
1805	.get_link	= proc_pid_get_link,
1806	.setattr	= proc_setattr,
1807};
1808
1809
1810/* building an inode */
1811
1812void task_dump_owner(struct task_struct *task, umode_t mode,
1813		     kuid_t *ruid, kgid_t *rgid)
1814{
1815	/* Depending on the state of dumpable compute who should own a
1816	 * proc file for a task.
1817	 */
1818	const struct cred *cred;
1819	kuid_t uid;
1820	kgid_t gid;
1821
1822	if (unlikely(task->flags & PF_KTHREAD)) {
1823		*ruid = GLOBAL_ROOT_UID;
1824		*rgid = GLOBAL_ROOT_GID;
1825		return;
1826	}
1827
1828	/* Default to the tasks effective ownership */
1829	rcu_read_lock();
1830	cred = __task_cred(task);
1831	uid = cred->euid;
1832	gid = cred->egid;
1833	rcu_read_unlock();
1834
1835	/*
1836	 * Before the /proc/pid/status file was created the only way to read
1837	 * the effective uid of a /process was to stat /proc/pid.  Reading
1838	 * /proc/pid/status is slow enough that procps and other packages
1839	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1840	 * made this apply to all per process world readable and executable
1841	 * directories.
1842	 */
1843	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1844		struct mm_struct *mm;
1845		task_lock(task);
1846		mm = task->mm;
1847		/* Make non-dumpable tasks owned by some root */
1848		if (mm) {
1849			if (get_dumpable(mm) != SUID_DUMP_USER) {
1850				struct user_namespace *user_ns = mm->user_ns;
1851
1852				uid = make_kuid(user_ns, 0);
1853				if (!uid_valid(uid))
1854					uid = GLOBAL_ROOT_UID;
1855
1856				gid = make_kgid(user_ns, 0);
1857				if (!gid_valid(gid))
1858					gid = GLOBAL_ROOT_GID;
1859			}
1860		} else {
1861			uid = GLOBAL_ROOT_UID;
1862			gid = GLOBAL_ROOT_GID;
1863		}
1864		task_unlock(task);
1865	}
1866	*ruid = uid;
1867	*rgid = gid;
1868}
1869
1870void proc_pid_evict_inode(struct proc_inode *ei)
1871{
1872	struct pid *pid = ei->pid;
1873
1874	if (S_ISDIR(ei->vfs_inode.i_mode)) {
1875		spin_lock(&pid->lock);
1876		hlist_del_init_rcu(&ei->sibling_inodes);
1877		spin_unlock(&pid->lock);
1878	}
1879
1880	put_pid(pid);
1881}
1882
1883struct inode *proc_pid_make_inode(struct super_block * sb,
1884				  struct task_struct *task, umode_t mode)
1885{
1886	struct inode * inode;
1887	struct proc_inode *ei;
1888	struct pid *pid;
1889
1890	/* We need a new inode */
1891
1892	inode = new_inode(sb);
1893	if (!inode)
1894		goto out;
1895
1896	/* Common stuff */
1897	ei = PROC_I(inode);
1898	inode->i_mode = mode;
1899	inode->i_ino = get_next_ino();
1900	inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1901	inode->i_op = &proc_def_inode_operations;
1902
1903	/*
1904	 * grab the reference to task.
1905	 */
1906	pid = get_task_pid(task, PIDTYPE_PID);
1907	if (!pid)
1908		goto out_unlock;
1909
1910	/* Let the pid remember us for quick removal */
1911	ei->pid = pid;
1912	if (S_ISDIR(mode)) {
1913		spin_lock(&pid->lock);
1914		hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
1915		spin_unlock(&pid->lock);
1916	}
1917
1918	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1919	security_task_to_inode(task, inode);
1920
1921out:
1922	return inode;
1923
1924out_unlock:
1925	iput(inode);
1926	return NULL;
1927}
1928
1929int pid_getattr(struct user_namespace *mnt_userns, const struct path *path,
1930		struct kstat *stat, u32 request_mask, unsigned int query_flags)
1931{
1932	struct inode *inode = d_inode(path->dentry);
1933	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
1934	struct task_struct *task;
1935
1936	generic_fillattr(&init_user_ns, inode, stat);
1937
1938	stat->uid = GLOBAL_ROOT_UID;
1939	stat->gid = GLOBAL_ROOT_GID;
1940	rcu_read_lock();
1941	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1942	if (task) {
1943		if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
1944			rcu_read_unlock();
1945			/*
1946			 * This doesn't prevent learning whether PID exists,
1947			 * it only makes getattr() consistent with readdir().
1948			 */
1949			return -ENOENT;
1950		}
1951		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1952	}
1953	rcu_read_unlock();
1954	return 0;
1955}
1956
1957/* dentry stuff */
1958
1959/*
1960 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1961 */
1962void pid_update_inode(struct task_struct *task, struct inode *inode)
1963{
1964	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1965
1966	inode->i_mode &= ~(S_ISUID | S_ISGID);
1967	security_task_to_inode(task, inode);
1968}
1969
1970/*
1971 * Rewrite the inode's ownerships here because the owning task may have
1972 * performed a setuid(), etc.
1973 *
1974 */
1975static int pid_revalidate(struct dentry *dentry, unsigned int flags)
1976{
1977	struct inode *inode;
1978	struct task_struct *task;
1979
1980	if (flags & LOOKUP_RCU)
1981		return -ECHILD;
1982
1983	inode = d_inode(dentry);
1984	task = get_proc_task(inode);
1985
1986	if (task) {
1987		pid_update_inode(task, inode);
1988		put_task_struct(task);
1989		return 1;
1990	}
1991	return 0;
1992}
1993
1994static inline bool proc_inode_is_dead(struct inode *inode)
1995{
1996	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1997}
1998
1999int pid_delete_dentry(const struct dentry *dentry)
2000{
2001	/* Is the task we represent dead?
2002	 * If so, then don't put the dentry on the lru list,
2003	 * kill it immediately.
2004	 */
2005	return proc_inode_is_dead(d_inode(dentry));
2006}
2007
2008const struct dentry_operations pid_dentry_operations =
2009{
2010	.d_revalidate	= pid_revalidate,
2011	.d_delete	= pid_delete_dentry,
2012};
2013
2014/* Lookups */
2015
2016/*
2017 * Fill a directory entry.
2018 *
2019 * If possible create the dcache entry and derive our inode number and
2020 * file type from dcache entry.
2021 *
2022 * Since all of the proc inode numbers are dynamically generated, the inode
2023 * numbers do not exist until the inode is cache.  This means creating
2024 * the dcache entry in readdir is necessary to keep the inode numbers
2025 * reported by readdir in sync with the inode numbers reported
2026 * by stat.
2027 */
2028bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2029	const char *name, unsigned int len,
2030	instantiate_t instantiate, struct task_struct *task, const void *ptr)
2031{
2032	struct dentry *child, *dir = file->f_path.dentry;
2033	struct qstr qname = QSTR_INIT(name, len);
2034	struct inode *inode;
2035	unsigned type = DT_UNKNOWN;
2036	ino_t ino = 1;
2037
2038	child = d_hash_and_lookup(dir, &qname);
2039	if (!child) {
2040		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2041		child = d_alloc_parallel(dir, &qname, &wq);
2042		if (IS_ERR(child))
2043			goto end_instantiate;
2044		if (d_in_lookup(child)) {
2045			struct dentry *res;
2046			res = instantiate(child, task, ptr);
2047			d_lookup_done(child);
2048			if (unlikely(res)) {
2049				dput(child);
2050				child = res;
2051				if (IS_ERR(child))
2052					goto end_instantiate;
2053			}
2054		}
2055	}
2056	inode = d_inode(child);
2057	ino = inode->i_ino;
2058	type = inode->i_mode >> 12;
2059	dput(child);
2060end_instantiate:
2061	return dir_emit(ctx, name, len, ino, type);
2062}
2063
2064/*
2065 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2066 * which represent vma start and end addresses.
2067 */
2068static int dname_to_vma_addr(struct dentry *dentry,
2069			     unsigned long *start, unsigned long *end)
2070{
2071	const char *str = dentry->d_name.name;
2072	unsigned long long sval, eval;
2073	unsigned int len;
2074
2075	if (str[0] == '0' && str[1] != '-')
2076		return -EINVAL;
2077	len = _parse_integer(str, 16, &sval);
2078	if (len & KSTRTOX_OVERFLOW)
2079		return -EINVAL;
2080	if (sval != (unsigned long)sval)
2081		return -EINVAL;
2082	str += len;
2083
2084	if (*str != '-')
2085		return -EINVAL;
2086	str++;
2087
2088	if (str[0] == '0' && str[1])
2089		return -EINVAL;
2090	len = _parse_integer(str, 16, &eval);
2091	if (len & KSTRTOX_OVERFLOW)
2092		return -EINVAL;
2093	if (eval != (unsigned long)eval)
2094		return -EINVAL;
2095	str += len;
2096
2097	if (*str != '\0')
2098		return -EINVAL;
2099
2100	*start = sval;
2101	*end = eval;
2102
2103	return 0;
2104}
2105
2106static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2107{
2108	unsigned long vm_start, vm_end;
2109	bool exact_vma_exists = false;
2110	struct mm_struct *mm = NULL;
2111	struct task_struct *task;
2112	struct inode *inode;
2113	int status = 0;
2114
2115	if (flags & LOOKUP_RCU)
2116		return -ECHILD;
2117
2118	inode = d_inode(dentry);
2119	task = get_proc_task(inode);
2120	if (!task)
2121		goto out_notask;
2122
2123	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2124	if (IS_ERR_OR_NULL(mm))
2125		goto out;
2126
2127	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2128		status = mmap_read_lock_killable(mm);
2129		if (!status) {
2130			exact_vma_exists = !!find_exact_vma(mm, vm_start,
2131							    vm_end);
2132			mmap_read_unlock(mm);
2133		}
2134	}
2135
2136	mmput(mm);
2137
2138	if (exact_vma_exists) {
2139		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2140
2141		security_task_to_inode(task, inode);
2142		status = 1;
2143	}
2144
2145out:
2146	put_task_struct(task);
2147
2148out_notask:
2149	return status;
2150}
2151
2152static const struct dentry_operations tid_map_files_dentry_operations = {
2153	.d_revalidate	= map_files_d_revalidate,
2154	.d_delete	= pid_delete_dentry,
2155};
2156
2157static int map_files_get_link(struct dentry *dentry, struct path *path)
2158{
2159	unsigned long vm_start, vm_end;
2160	struct vm_area_struct *vma;
2161	struct task_struct *task;
2162	struct mm_struct *mm;
2163	int rc;
2164
2165	rc = -ENOENT;
2166	task = get_proc_task(d_inode(dentry));
2167	if (!task)
2168		goto out;
2169
2170	mm = get_task_mm(task);
2171	put_task_struct(task);
2172	if (!mm)
2173		goto out;
2174
2175	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2176	if (rc)
2177		goto out_mmput;
2178
2179	rc = mmap_read_lock_killable(mm);
2180	if (rc)
2181		goto out_mmput;
2182
2183	rc = -ENOENT;
2184	vma = find_exact_vma(mm, vm_start, vm_end);
2185	if (vma && vma->vm_file) {
2186		*path = vma->vm_file->f_path;
2187		path_get(path);
2188		rc = 0;
2189	}
2190	mmap_read_unlock(mm);
2191
2192out_mmput:
2193	mmput(mm);
2194out:
2195	return rc;
2196}
2197
2198struct map_files_info {
2199	unsigned long	start;
2200	unsigned long	end;
2201	fmode_t		mode;
2202};
2203
2204/*
2205 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2206 * to concerns about how the symlinks may be used to bypass permissions on
2207 * ancestor directories in the path to the file in question.
2208 */
2209static const char *
2210proc_map_files_get_link(struct dentry *dentry,
2211			struct inode *inode,
2212		        struct delayed_call *done)
2213{
2214	if (!checkpoint_restore_ns_capable(&init_user_ns))
2215		return ERR_PTR(-EPERM);
2216
2217	return proc_pid_get_link(dentry, inode, done);
2218}
2219
2220/*
2221 * Identical to proc_pid_link_inode_operations except for get_link()
2222 */
2223static const struct inode_operations proc_map_files_link_inode_operations = {
2224	.readlink	= proc_pid_readlink,
2225	.get_link	= proc_map_files_get_link,
2226	.setattr	= proc_setattr,
2227};
2228
2229static struct dentry *
2230proc_map_files_instantiate(struct dentry *dentry,
2231			   struct task_struct *task, const void *ptr)
2232{
2233	fmode_t mode = (fmode_t)(unsigned long)ptr;
2234	struct proc_inode *ei;
2235	struct inode *inode;
2236
2237	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2238				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2239				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2240	if (!inode)
2241		return ERR_PTR(-ENOENT);
2242
2243	ei = PROC_I(inode);
2244	ei->op.proc_get_link = map_files_get_link;
2245
2246	inode->i_op = &proc_map_files_link_inode_operations;
2247	inode->i_size = 64;
2248
2249	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2250	return d_splice_alias(inode, dentry);
2251}
2252
2253static struct dentry *proc_map_files_lookup(struct inode *dir,
2254		struct dentry *dentry, unsigned int flags)
2255{
2256	unsigned long vm_start, vm_end;
2257	struct vm_area_struct *vma;
2258	struct task_struct *task;
2259	struct dentry *result;
2260	struct mm_struct *mm;
2261
2262	result = ERR_PTR(-ENOENT);
2263	task = get_proc_task(dir);
2264	if (!task)
2265		goto out;
2266
2267	result = ERR_PTR(-EACCES);
2268	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2269		goto out_put_task;
2270
2271	result = ERR_PTR(-ENOENT);
2272	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2273		goto out_put_task;
2274
2275	mm = get_task_mm(task);
2276	if (!mm)
2277		goto out_put_task;
2278
2279	result = ERR_PTR(-EINTR);
2280	if (mmap_read_lock_killable(mm))
2281		goto out_put_mm;
2282
2283	result = ERR_PTR(-ENOENT);
2284	vma = find_exact_vma(mm, vm_start, vm_end);
2285	if (!vma)
2286		goto out_no_vma;
2287
2288	if (vma->vm_file)
2289		result = proc_map_files_instantiate(dentry, task,
2290				(void *)(unsigned long)vma->vm_file->f_mode);
2291
2292out_no_vma:
2293	mmap_read_unlock(mm);
2294out_put_mm:
2295	mmput(mm);
2296out_put_task:
2297	put_task_struct(task);
2298out:
2299	return result;
2300}
2301
2302static const struct inode_operations proc_map_files_inode_operations = {
2303	.lookup		= proc_map_files_lookup,
2304	.permission	= proc_fd_permission,
2305	.setattr	= proc_setattr,
2306};
2307
2308static int
2309proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2310{
2311	struct vm_area_struct *vma;
2312	struct task_struct *task;
2313	struct mm_struct *mm;
2314	unsigned long nr_files, pos, i;
2315	GENRADIX(struct map_files_info) fa;
2316	struct map_files_info *p;
2317	int ret;
2318
2319	genradix_init(&fa);
2320
2321	ret = -ENOENT;
2322	task = get_proc_task(file_inode(file));
2323	if (!task)
2324		goto out;
2325
2326	ret = -EACCES;
2327	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2328		goto out_put_task;
2329
2330	ret = 0;
2331	if (!dir_emit_dots(file, ctx))
2332		goto out_put_task;
2333
2334	mm = get_task_mm(task);
2335	if (!mm)
2336		goto out_put_task;
2337
2338	ret = mmap_read_lock_killable(mm);
2339	if (ret) {
2340		mmput(mm);
2341		goto out_put_task;
2342	}
2343
2344	nr_files = 0;
2345
2346	/*
2347	 * We need two passes here:
2348	 *
2349	 *  1) Collect vmas of mapped files with mmap_lock taken
2350	 *  2) Release mmap_lock and instantiate entries
2351	 *
2352	 * otherwise we get lockdep complained, since filldir()
2353	 * routine might require mmap_lock taken in might_fault().
2354	 */
2355
2356	for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2357		if (!vma->vm_file)
2358			continue;
2359		if (++pos <= ctx->pos)
2360			continue;
2361
2362		p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2363		if (!p) {
2364			ret = -ENOMEM;
2365			mmap_read_unlock(mm);
2366			mmput(mm);
2367			goto out_put_task;
2368		}
2369
2370		p->start = vma->vm_start;
2371		p->end = vma->vm_end;
2372		p->mode = vma->vm_file->f_mode;
2373	}
2374	mmap_read_unlock(mm);
2375	mmput(mm);
2376
2377	for (i = 0; i < nr_files; i++) {
2378		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2379		unsigned int len;
2380
2381		p = genradix_ptr(&fa, i);
2382		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2383		if (!proc_fill_cache(file, ctx,
2384				      buf, len,
2385				      proc_map_files_instantiate,
2386				      task,
2387				      (void *)(unsigned long)p->mode))
2388			break;
2389		ctx->pos++;
2390	}
2391
2392out_put_task:
2393	put_task_struct(task);
2394out:
2395	genradix_free(&fa);
2396	return ret;
2397}
2398
2399static const struct file_operations proc_map_files_operations = {
2400	.read		= generic_read_dir,
2401	.iterate_shared	= proc_map_files_readdir,
2402	.llseek		= generic_file_llseek,
2403};
2404
2405#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2406struct timers_private {
2407	struct pid *pid;
2408	struct task_struct *task;
2409	struct sighand_struct *sighand;
2410	struct pid_namespace *ns;
2411	unsigned long flags;
2412};
2413
2414static void *timers_start(struct seq_file *m, loff_t *pos)
2415{
2416	struct timers_private *tp = m->private;
2417
2418	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2419	if (!tp->task)
2420		return ERR_PTR(-ESRCH);
2421
2422	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2423	if (!tp->sighand)
2424		return ERR_PTR(-ESRCH);
2425
2426	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2427}
2428
2429static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2430{
2431	struct timers_private *tp = m->private;
2432	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2433}
2434
2435static void timers_stop(struct seq_file *m, void *v)
2436{
2437	struct timers_private *tp = m->private;
2438
2439	if (tp->sighand) {
2440		unlock_task_sighand(tp->task, &tp->flags);
2441		tp->sighand = NULL;
2442	}
2443
2444	if (tp->task) {
2445		put_task_struct(tp->task);
2446		tp->task = NULL;
2447	}
2448}
2449
2450static int show_timer(struct seq_file *m, void *v)
2451{
2452	struct k_itimer *timer;
2453	struct timers_private *tp = m->private;
2454	int notify;
2455	static const char * const nstr[] = {
2456		[SIGEV_SIGNAL] = "signal",
2457		[SIGEV_NONE] = "none",
2458		[SIGEV_THREAD] = "thread",
2459	};
2460
2461	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2462	notify = timer->it_sigev_notify;
2463
2464	seq_printf(m, "ID: %d\n", timer->it_id);
2465	seq_printf(m, "signal: %d/%px\n",
2466		   timer->sigq->info.si_signo,
2467		   timer->sigq->info.si_value.sival_ptr);
2468	seq_printf(m, "notify: %s/%s.%d\n",
2469		   nstr[notify & ~SIGEV_THREAD_ID],
2470		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2471		   pid_nr_ns(timer->it_pid, tp->ns));
2472	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2473
2474	return 0;
2475}
2476
2477static const struct seq_operations proc_timers_seq_ops = {
2478	.start	= timers_start,
2479	.next	= timers_next,
2480	.stop	= timers_stop,
2481	.show	= show_timer,
2482};
2483
2484static int proc_timers_open(struct inode *inode, struct file *file)
2485{
2486	struct timers_private *tp;
2487
2488	tp = __seq_open_private(file, &proc_timers_seq_ops,
2489			sizeof(struct timers_private));
2490	if (!tp)
2491		return -ENOMEM;
2492
2493	tp->pid = proc_pid(inode);
2494	tp->ns = proc_pid_ns(inode->i_sb);
2495	return 0;
2496}
2497
2498static const struct file_operations proc_timers_operations = {
2499	.open		= proc_timers_open,
2500	.read		= seq_read,
2501	.llseek		= seq_lseek,
2502	.release	= seq_release_private,
2503};
2504#endif
2505
2506static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2507					size_t count, loff_t *offset)
2508{
2509	struct inode *inode = file_inode(file);
2510	struct task_struct *p;
2511	u64 slack_ns;
2512	int err;
2513
2514	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2515	if (err < 0)
2516		return err;
2517
2518	p = get_proc_task(inode);
2519	if (!p)
2520		return -ESRCH;
2521
2522	if (p != current) {
2523		rcu_read_lock();
2524		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2525			rcu_read_unlock();
2526			count = -EPERM;
2527			goto out;
2528		}
2529		rcu_read_unlock();
2530
2531		err = security_task_setscheduler(p);
2532		if (err) {
2533			count = err;
2534			goto out;
2535		}
2536	}
2537
2538	task_lock(p);
2539	if (slack_ns == 0)
2540		p->timer_slack_ns = p->default_timer_slack_ns;
2541	else
2542		p->timer_slack_ns = slack_ns;
2543	task_unlock(p);
2544
2545out:
2546	put_task_struct(p);
2547
2548	return count;
2549}
2550
2551static int timerslack_ns_show(struct seq_file *m, void *v)
2552{
2553	struct inode *inode = m->private;
2554	struct task_struct *p;
2555	int err = 0;
2556
2557	p = get_proc_task(inode);
2558	if (!p)
2559		return -ESRCH;
2560
2561	if (p != current) {
2562		rcu_read_lock();
2563		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2564			rcu_read_unlock();
2565			err = -EPERM;
2566			goto out;
2567		}
2568		rcu_read_unlock();
2569
2570		err = security_task_getscheduler(p);
2571		if (err)
2572			goto out;
2573	}
2574
2575	task_lock(p);
2576	seq_printf(m, "%llu\n", p->timer_slack_ns);
2577	task_unlock(p);
2578
2579out:
2580	put_task_struct(p);
2581
2582	return err;
2583}
2584
2585static int timerslack_ns_open(struct inode *inode, struct file *filp)
2586{
2587	return single_open(filp, timerslack_ns_show, inode);
2588}
2589
2590static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2591	.open		= timerslack_ns_open,
2592	.read		= seq_read,
2593	.write		= timerslack_ns_write,
2594	.llseek		= seq_lseek,
2595	.release	= single_release,
2596};
2597
2598static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2599	struct task_struct *task, const void *ptr)
2600{
2601	const struct pid_entry *p = ptr;
2602	struct inode *inode;
2603	struct proc_inode *ei;
2604
2605	inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2606	if (!inode)
2607		return ERR_PTR(-ENOENT);
2608
2609	ei = PROC_I(inode);
2610	if (S_ISDIR(inode->i_mode))
2611		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2612	if (p->iop)
2613		inode->i_op = p->iop;
2614	if (p->fop)
2615		inode->i_fop = p->fop;
2616	ei->op = p->op;
2617	pid_update_inode(task, inode);
2618	d_set_d_op(dentry, &pid_dentry_operations);
2619	return d_splice_alias(inode, dentry);
2620}
2621
2622static struct dentry *proc_pident_lookup(struct inode *dir, 
2623					 struct dentry *dentry,
2624					 const struct pid_entry *p,
2625					 const struct pid_entry *end)
2626{
2627	struct task_struct *task = get_proc_task(dir);
2628	struct dentry *res = ERR_PTR(-ENOENT);
2629
2630	if (!task)
2631		goto out_no_task;
2632
2633	/*
2634	 * Yes, it does not scale. And it should not. Don't add
2635	 * new entries into /proc/<tgid>/ without very good reasons.
2636	 */
2637	for (; p < end; p++) {
2638		if (p->len != dentry->d_name.len)
2639			continue;
2640		if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2641			res = proc_pident_instantiate(dentry, task, p);
2642			break;
2643		}
2644	}
2645	put_task_struct(task);
2646out_no_task:
2647	return res;
2648}
2649
2650static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2651		const struct pid_entry *ents, unsigned int nents)
2652{
2653	struct task_struct *task = get_proc_task(file_inode(file));
2654	const struct pid_entry *p;
2655
2656	if (!task)
2657		return -ENOENT;
2658
2659	if (!dir_emit_dots(file, ctx))
2660		goto out;
2661
2662	if (ctx->pos >= nents + 2)
2663		goto out;
2664
2665	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2666		if (!proc_fill_cache(file, ctx, p->name, p->len,
2667				proc_pident_instantiate, task, p))
2668			break;
2669		ctx->pos++;
2670	}
2671out:
2672	put_task_struct(task);
2673	return 0;
2674}
2675
2676#ifdef CONFIG_SECURITY
2677static int proc_pid_attr_open(struct inode *inode, struct file *file)
2678{
2679	file->private_data = NULL;
2680	__mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2681	return 0;
2682}
2683
2684static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2685				  size_t count, loff_t *ppos)
2686{
2687	struct inode * inode = file_inode(file);
2688	char *p = NULL;
2689	ssize_t length;
2690	struct task_struct *task = get_proc_task(inode);
2691
2692	if (!task)
2693		return -ESRCH;
2694
2695	length = security_getprocattr(task, PROC_I(inode)->op.lsm,
2696				      (char*)file->f_path.dentry->d_name.name,
2697				      &p);
2698	put_task_struct(task);
2699	if (length > 0)
2700		length = simple_read_from_buffer(buf, count, ppos, p, length);
2701	kfree(p);
2702	return length;
2703}
2704
2705static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2706				   size_t count, loff_t *ppos)
2707{
2708	struct inode * inode = file_inode(file);
2709	struct task_struct *task;
2710	void *page;
2711	int rv;
2712
2713	/* A task may only write when it was the opener. */
2714	if (file->private_data != current->mm)
2715		return -EPERM;
2716
2717	rcu_read_lock();
2718	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2719	if (!task) {
2720		rcu_read_unlock();
2721		return -ESRCH;
2722	}
2723	/* A task may only write its own attributes. */
2724	if (current != task) {
2725		rcu_read_unlock();
2726		return -EACCES;
2727	}
2728	/* Prevent changes to overridden credentials. */
2729	if (current_cred() != current_real_cred()) {
2730		rcu_read_unlock();
2731		return -EBUSY;
2732	}
2733	rcu_read_unlock();
2734
2735	if (count > PAGE_SIZE)
2736		count = PAGE_SIZE;
2737
2738	/* No partial writes. */
2739	if (*ppos != 0)
2740		return -EINVAL;
2741
2742	page = memdup_user(buf, count);
2743	if (IS_ERR(page)) {
2744		rv = PTR_ERR(page);
2745		goto out;
2746	}
2747
2748	/* Guard against adverse ptrace interaction */
2749	rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2750	if (rv < 0)
2751		goto out_free;
2752
2753	rv = security_setprocattr(PROC_I(inode)->op.lsm,
2754				  file->f_path.dentry->d_name.name, page,
2755				  count);
2756	mutex_unlock(&current->signal->cred_guard_mutex);
2757out_free:
2758	kfree(page);
2759out:
2760	return rv;
2761}
2762
2763static const struct file_operations proc_pid_attr_operations = {
2764	.open		= proc_pid_attr_open,
2765	.read		= proc_pid_attr_read,
2766	.write		= proc_pid_attr_write,
2767	.llseek		= generic_file_llseek,
2768	.release	= mem_release,
2769};
2770
2771#define LSM_DIR_OPS(LSM) \
2772static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2773			     struct dir_context *ctx) \
2774{ \
2775	return proc_pident_readdir(filp, ctx, \
2776				   LSM##_attr_dir_stuff, \
2777				   ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2778} \
2779\
2780static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2781	.read		= generic_read_dir, \
2782	.iterate	= proc_##LSM##_attr_dir_iterate, \
2783	.llseek		= default_llseek, \
2784}; \
2785\
2786static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2787				struct dentry *dentry, unsigned int flags) \
2788{ \
2789	return proc_pident_lookup(dir, dentry, \
2790				  LSM##_attr_dir_stuff, \
2791				  LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2792} \
2793\
2794static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2795	.lookup		= proc_##LSM##_attr_dir_lookup, \
2796	.getattr	= pid_getattr, \
2797	.setattr	= proc_setattr, \
2798}
2799
2800#ifdef CONFIG_SECURITY_SMACK
2801static const struct pid_entry smack_attr_dir_stuff[] = {
2802	ATTR("smack", "current",	0666),
2803};
2804LSM_DIR_OPS(smack);
2805#endif
2806
2807#ifdef CONFIG_SECURITY_APPARMOR
2808static const struct pid_entry apparmor_attr_dir_stuff[] = {
2809	ATTR("apparmor", "current",	0666),
2810	ATTR("apparmor", "prev",	0444),
2811	ATTR("apparmor", "exec",	0666),
2812};
2813LSM_DIR_OPS(apparmor);
2814#endif
2815
2816static const struct pid_entry attr_dir_stuff[] = {
2817	ATTR(NULL, "current",		0666),
2818	ATTR(NULL, "prev",		0444),
2819	ATTR(NULL, "exec",		0666),
2820	ATTR(NULL, "fscreate",		0666),
2821	ATTR(NULL, "keycreate",		0666),
2822	ATTR(NULL, "sockcreate",	0666),
2823#ifdef CONFIG_SECURITY_SMACK
2824	DIR("smack",			0555,
2825	    proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2826#endif
2827#ifdef CONFIG_SECURITY_APPARMOR
2828	DIR("apparmor",			0555,
2829	    proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2830#endif
2831};
2832
2833static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2834{
2835	return proc_pident_readdir(file, ctx, 
2836				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2837}
2838
2839static const struct file_operations proc_attr_dir_operations = {
2840	.read		= generic_read_dir,
2841	.iterate_shared	= proc_attr_dir_readdir,
2842	.llseek		= generic_file_llseek,
2843};
2844
2845static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2846				struct dentry *dentry, unsigned int flags)
2847{
2848	return proc_pident_lookup(dir, dentry,
2849				  attr_dir_stuff,
2850				  attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2851}
2852
2853static const struct inode_operations proc_attr_dir_inode_operations = {
2854	.lookup		= proc_attr_dir_lookup,
2855	.getattr	= pid_getattr,
2856	.setattr	= proc_setattr,
2857};
2858
2859#endif
2860
2861#ifdef CONFIG_ELF_CORE
2862static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2863					 size_t count, loff_t *ppos)
2864{
2865	struct task_struct *task = get_proc_task(file_inode(file));
2866	struct mm_struct *mm;
2867	char buffer[PROC_NUMBUF];
2868	size_t len;
2869	int ret;
2870
2871	if (!task)
2872		return -ESRCH;
2873
2874	ret = 0;
2875	mm = get_task_mm(task);
2876	if (mm) {
2877		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2878			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2879				MMF_DUMP_FILTER_SHIFT));
2880		mmput(mm);
2881		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2882	}
2883
2884	put_task_struct(task);
2885
2886	return ret;
2887}
2888
2889static ssize_t proc_coredump_filter_write(struct file *file,
2890					  const char __user *buf,
2891					  size_t count,
2892					  loff_t *ppos)
2893{
2894	struct task_struct *task;
2895	struct mm_struct *mm;
2896	unsigned int val;
2897	int ret;
2898	int i;
2899	unsigned long mask;
2900
2901	ret = kstrtouint_from_user(buf, count, 0, &val);
2902	if (ret < 0)
2903		return ret;
2904
2905	ret = -ESRCH;
2906	task = get_proc_task(file_inode(file));
2907	if (!task)
2908		goto out_no_task;
2909
2910	mm = get_task_mm(task);
2911	if (!mm)
2912		goto out_no_mm;
2913	ret = 0;
2914
2915	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2916		if (val & mask)
2917			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2918		else
2919			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2920	}
2921
2922	mmput(mm);
2923 out_no_mm:
2924	put_task_struct(task);
2925 out_no_task:
2926	if (ret < 0)
2927		return ret;
2928	return count;
2929}
2930
2931static const struct file_operations proc_coredump_filter_operations = {
2932	.read		= proc_coredump_filter_read,
2933	.write		= proc_coredump_filter_write,
2934	.llseek		= generic_file_llseek,
2935};
2936#endif
2937
2938#ifdef CONFIG_TASK_IO_ACCOUNTING
2939static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2940{
2941	struct task_io_accounting acct = task->ioac;
2942	unsigned long flags;
2943	int result;
2944
2945	result = down_read_killable(&task->signal->exec_update_lock);
2946	if (result)
2947		return result;
2948
2949	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2950		result = -EACCES;
2951		goto out_unlock;
2952	}
2953
2954	if (whole && lock_task_sighand(task, &flags)) {
2955		struct task_struct *t = task;
2956
2957		task_io_accounting_add(&acct, &task->signal->ioac);
2958		while_each_thread(task, t)
2959			task_io_accounting_add(&acct, &t->ioac);
2960
2961		unlock_task_sighand(task, &flags);
2962	}
2963	seq_printf(m,
2964		   "rchar: %llu\n"
2965		   "wchar: %llu\n"
2966		   "syscr: %llu\n"
2967		   "syscw: %llu\n"
2968		   "read_bytes: %llu\n"
2969		   "write_bytes: %llu\n"
2970		   "cancelled_write_bytes: %llu\n",
2971		   (unsigned long long)acct.rchar,
2972		   (unsigned long long)acct.wchar,
2973		   (unsigned long long)acct.syscr,
2974		   (unsigned long long)acct.syscw,
2975		   (unsigned long long)acct.read_bytes,
2976		   (unsigned long long)acct.write_bytes,
2977		   (unsigned long long)acct.cancelled_write_bytes);
2978	result = 0;
2979
2980out_unlock:
2981	up_read(&task->signal->exec_update_lock);
2982	return result;
2983}
2984
2985static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2986				  struct pid *pid, struct task_struct *task)
2987{
2988	return do_io_accounting(task, m, 0);
2989}
2990
2991static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2992				   struct pid *pid, struct task_struct *task)
2993{
2994	return do_io_accounting(task, m, 1);
2995}
2996#endif /* CONFIG_TASK_IO_ACCOUNTING */
2997
2998#ifdef CONFIG_USER_NS
2999static int proc_id_map_open(struct inode *inode, struct file *file,
3000	const struct seq_operations *seq_ops)
3001{
3002	struct user_namespace *ns = NULL;
3003	struct task_struct *task;
3004	struct seq_file *seq;
3005	int ret = -EINVAL;
3006
3007	task = get_proc_task(inode);
3008	if (task) {
3009		rcu_read_lock();
3010		ns = get_user_ns(task_cred_xxx(task, user_ns));
3011		rcu_read_unlock();
3012		put_task_struct(task);
3013	}
3014	if (!ns)
3015		goto err;
3016
3017	ret = seq_open(file, seq_ops);
3018	if (ret)
3019		goto err_put_ns;
3020
3021	seq = file->private_data;
3022	seq->private = ns;
3023
3024	return 0;
3025err_put_ns:
3026	put_user_ns(ns);
3027err:
3028	return ret;
3029}
3030
3031static int proc_id_map_release(struct inode *inode, struct file *file)
3032{
3033	struct seq_file *seq = file->private_data;
3034	struct user_namespace *ns = seq->private;
3035	put_user_ns(ns);
3036	return seq_release(inode, file);
3037}
3038
3039static int proc_uid_map_open(struct inode *inode, struct file *file)
3040{
3041	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3042}
3043
3044static int proc_gid_map_open(struct inode *inode, struct file *file)
3045{
3046	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3047}
3048
3049static int proc_projid_map_open(struct inode *inode, struct file *file)
3050{
3051	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3052}
3053
3054static const struct file_operations proc_uid_map_operations = {
3055	.open		= proc_uid_map_open,
3056	.write		= proc_uid_map_write,
3057	.read		= seq_read,
3058	.llseek		= seq_lseek,
3059	.release	= proc_id_map_release,
3060};
3061
3062static const struct file_operations proc_gid_map_operations = {
3063	.open		= proc_gid_map_open,
3064	.write		= proc_gid_map_write,
3065	.read		= seq_read,
3066	.llseek		= seq_lseek,
3067	.release	= proc_id_map_release,
3068};
3069
3070static const struct file_operations proc_projid_map_operations = {
3071	.open		= proc_projid_map_open,
3072	.write		= proc_projid_map_write,
3073	.read		= seq_read,
3074	.llseek		= seq_lseek,
3075	.release	= proc_id_map_release,
3076};
3077
3078static int proc_setgroups_open(struct inode *inode, struct file *file)
3079{
3080	struct user_namespace *ns = NULL;
3081	struct task_struct *task;
3082	int ret;
3083
3084	ret = -ESRCH;
3085	task = get_proc_task(inode);
3086	if (task) {
3087		rcu_read_lock();
3088		ns = get_user_ns(task_cred_xxx(task, user_ns));
3089		rcu_read_unlock();
3090		put_task_struct(task);
3091	}
3092	if (!ns)
3093		goto err;
3094
3095	if (file->f_mode & FMODE_WRITE) {
3096		ret = -EACCES;
3097		if (!ns_capable(ns, CAP_SYS_ADMIN))
3098			goto err_put_ns;
3099	}
3100
3101	ret = single_open(file, &proc_setgroups_show, ns);
3102	if (ret)
3103		goto err_put_ns;
3104
3105	return 0;
3106err_put_ns:
3107	put_user_ns(ns);
3108err:
3109	return ret;
3110}
3111
3112static int proc_setgroups_release(struct inode *inode, struct file *file)
3113{
3114	struct seq_file *seq = file->private_data;
3115	struct user_namespace *ns = seq->private;
3116	int ret = single_release(inode, file);
3117	put_user_ns(ns);
3118	return ret;
3119}
3120
3121static const struct file_operations proc_setgroups_operations = {
3122	.open		= proc_setgroups_open,
3123	.write		= proc_setgroups_write,
3124	.read		= seq_read,
3125	.llseek		= seq_lseek,
3126	.release	= proc_setgroups_release,
3127};
3128#endif /* CONFIG_USER_NS */
3129
3130static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3131				struct pid *pid, struct task_struct *task)
3132{
3133	int err = lock_trace(task);
3134	if (!err) {
3135		seq_printf(m, "%08x\n", task->personality);
3136		unlock_trace(task);
3137	}
3138	return err;
3139}
3140
3141#ifdef CONFIG_LIVEPATCH
3142static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3143				struct pid *pid, struct task_struct *task)
3144{
3145	seq_printf(m, "%d\n", task->patch_state);
3146	return 0;
3147}
3148#endif /* CONFIG_LIVEPATCH */
3149
3150#ifdef CONFIG_STACKLEAK_METRICS
3151static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3152				struct pid *pid, struct task_struct *task)
3153{
3154	unsigned long prev_depth = THREAD_SIZE -
3155				(task->prev_lowest_stack & (THREAD_SIZE - 1));
3156	unsigned long depth = THREAD_SIZE -
3157				(task->lowest_stack & (THREAD_SIZE - 1));
3158
3159	seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3160							prev_depth, depth);
3161	return 0;
3162}
3163#endif /* CONFIG_STACKLEAK_METRICS */
3164
3165/*
3166 * Thread groups
3167 */
3168static const struct file_operations proc_task_operations;
3169static const struct inode_operations proc_task_inode_operations;
3170
3171static const struct pid_entry tgid_base_stuff[] = {
3172	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3173	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3174	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3175	DIR("fdinfo",     S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3176	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3177#ifdef CONFIG_NET
3178	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3179#endif
3180	REG("environ",    S_IRUSR, proc_environ_operations),
3181	REG("auxv",       S_IRUSR, proc_auxv_operations),
3182	ONE("status",     S_IRUGO, proc_pid_status),
3183	ONE("personality", S_IRUSR, proc_pid_personality),
3184	ONE("limits",	  S_IRUGO, proc_pid_limits),
3185#ifdef CONFIG_SCHED_DEBUG
3186	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3187#endif
3188#ifdef CONFIG_SCHED_AUTOGROUP
3189	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3190#endif
3191#ifdef CONFIG_TIME_NS
3192	REG("timens_offsets",  S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3193#endif
3194	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3195#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3196	ONE("syscall",    S_IRUSR, proc_pid_syscall),
3197#endif
3198	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
3199	ONE("stat",       S_IRUGO, proc_tgid_stat),
3200	ONE("statm",      S_IRUGO, proc_pid_statm),
3201	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3202#ifdef CONFIG_NUMA
3203	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3204#endif
3205	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3206	LNK("cwd",        proc_cwd_link),
3207	LNK("root",       proc_root_link),
3208	LNK("exe",        proc_exe_link),
3209	REG("mounts",     S_IRUGO, proc_mounts_operations),
3210	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3211	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3212#ifdef CONFIG_PROC_PAGE_MONITOR
3213	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3214	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3215	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3216	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3217#endif
3218#ifdef CONFIG_SECURITY
3219	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3220#endif
3221#ifdef CONFIG_KALLSYMS
3222	ONE("wchan",      S_IRUGO, proc_pid_wchan),
3223#endif
3224#ifdef CONFIG_STACKTRACE
3225	ONE("stack",      S_IRUSR, proc_pid_stack),
3226#endif
3227#ifdef CONFIG_SCHED_INFO
3228	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
3229#endif
3230#ifdef CONFIG_LATENCYTOP
3231	REG("latency",  S_IRUGO, proc_lstats_operations),
3232#endif
3233#ifdef CONFIG_PROC_PID_CPUSET
3234	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
3235#endif
3236#ifdef CONFIG_CGROUPS
3237	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3238#endif
3239#ifdef CONFIG_PROC_CPU_RESCTRL
3240	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3241#endif
3242	ONE("oom_score",  S_IRUGO, proc_oom_score),
3243	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3244	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3245#ifdef CONFIG_AUDIT
3246	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3247	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3248#endif
3249#ifdef CONFIG_FAULT_INJECTION
3250	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3251	REG("fail-nth", 0644, proc_fail_nth_operations),
3252#endif
3253#ifdef CONFIG_ELF_CORE
3254	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3255#endif
3256#ifdef CONFIG_TASK_IO_ACCOUNTING
3257	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3258#endif
3259#ifdef CONFIG_USER_NS
3260	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3261	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3262	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3263	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3264#endif
3265#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3266	REG("timers",	  S_IRUGO, proc_timers_operations),
3267#endif
3268	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3269#ifdef CONFIG_LIVEPATCH
3270	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3271#endif
3272#ifdef CONFIG_STACKLEAK_METRICS
3273	ONE("stack_depth", S_IRUGO, proc_stack_depth),
3274#endif
3275#ifdef CONFIG_PROC_PID_ARCH_STATUS
3276	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3277#endif
3278#ifdef CONFIG_SECCOMP_CACHE_DEBUG
3279	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3280#endif
3281};
3282
3283static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3284{
3285	return proc_pident_readdir(file, ctx,
3286				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3287}
3288
3289static const struct file_operations proc_tgid_base_operations = {
3290	.read		= generic_read_dir,
3291	.iterate_shared	= proc_tgid_base_readdir,
3292	.llseek		= generic_file_llseek,
3293};
3294
3295struct pid *tgid_pidfd_to_pid(const struct file *file)
3296{
3297	if (file->f_op != &proc_tgid_base_operations)
3298		return ERR_PTR(-EBADF);
3299
3300	return proc_pid(file_inode(file));
3301}
3302
3303static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3304{
3305	return proc_pident_lookup(dir, dentry,
3306				  tgid_base_stuff,
3307				  tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3308}
3309
3310static const struct inode_operations proc_tgid_base_inode_operations = {
3311	.lookup		= proc_tgid_base_lookup,
3312	.getattr	= pid_getattr,
3313	.setattr	= proc_setattr,
3314	.permission	= proc_pid_permission,
3315};
3316
3317/**
3318 * proc_flush_pid -  Remove dcache entries for @pid from the /proc dcache.
3319 * @pid: pid that should be flushed.
3320 *
3321 * This function walks a list of inodes (that belong to any proc
3322 * filesystem) that are attached to the pid and flushes them from
3323 * the dentry cache.
3324 *
3325 * It is safe and reasonable to cache /proc entries for a task until
3326 * that task exits.  After that they just clog up the dcache with
3327 * useless entries, possibly causing useful dcache entries to be
3328 * flushed instead.  This routine is provided to flush those useless
3329 * dcache entries when a process is reaped.
3330 *
3331 * NOTE: This routine is just an optimization so it does not guarantee
3332 *       that no dcache entries will exist after a process is reaped
3333 *       it just makes it very unlikely that any will persist.
3334 */
3335
3336void proc_flush_pid(struct pid *pid)
3337{
3338	proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3339}
3340
3341static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3342				   struct task_struct *task, const void *ptr)
3343{
3344	struct inode *inode;
3345
3346	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3347	if (!inode)
3348		return ERR_PTR(-ENOENT);
3349
3350	inode->i_op = &proc_tgid_base_inode_operations;
3351	inode->i_fop = &proc_tgid_base_operations;
3352	inode->i_flags|=S_IMMUTABLE;
3353
3354	set_nlink(inode, nlink_tgid);
3355	pid_update_inode(task, inode);
3356
3357	d_set_d_op(dentry, &pid_dentry_operations);
3358	return d_splice_alias(inode, dentry);
3359}
3360
3361struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3362{
3363	struct task_struct *task;
3364	unsigned tgid;
3365	struct proc_fs_info *fs_info;
3366	struct pid_namespace *ns;
3367	struct dentry *result = ERR_PTR(-ENOENT);
3368
3369	tgid = name_to_int(&dentry->d_name);
3370	if (tgid == ~0U)
3371		goto out;
3372
3373	fs_info = proc_sb_info(dentry->d_sb);
3374	ns = fs_info->pid_ns;
3375	rcu_read_lock();
3376	task = find_task_by_pid_ns(tgid, ns);
3377	if (task)
3378		get_task_struct(task);
3379	rcu_read_unlock();
3380	if (!task)
3381		goto out;
3382
3383	/* Limit procfs to only ptraceable tasks */
3384	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3385		if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3386			goto out_put_task;
3387	}
3388
3389	result = proc_pid_instantiate(dentry, task, NULL);
3390out_put_task:
3391	put_task_struct(task);
3392out:
3393	return result;
3394}
3395
3396/*
3397 * Find the first task with tgid >= tgid
3398 *
3399 */
3400struct tgid_iter {
3401	unsigned int tgid;
3402	struct task_struct *task;
3403};
3404static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3405{
3406	struct pid *pid;
3407
3408	if (iter.task)
3409		put_task_struct(iter.task);
3410	rcu_read_lock();
3411retry:
3412	iter.task = NULL;
3413	pid = find_ge_pid(iter.tgid, ns);
3414	if (pid) {
3415		iter.tgid = pid_nr_ns(pid, ns);
3416		iter.task = pid_task(pid, PIDTYPE_TGID);
3417		if (!iter.task) {
3418			iter.tgid += 1;
3419			goto retry;
3420		}
3421		get_task_struct(iter.task);
3422	}
3423	rcu_read_unlock();
3424	return iter;
3425}
3426
3427#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3428
3429/* for the /proc/ directory itself, after non-process stuff has been done */
3430int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3431{
3432	struct tgid_iter iter;
3433	struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3434	struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3435	loff_t pos = ctx->pos;
3436
3437	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3438		return 0;
3439
3440	if (pos == TGID_OFFSET - 2) {
3441		struct inode *inode = d_inode(fs_info->proc_self);
3442		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3443			return 0;
3444		ctx->pos = pos = pos + 1;
3445	}
3446	if (pos == TGID_OFFSET - 1) {
3447		struct inode *inode = d_inode(fs_info->proc_thread_self);
3448		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3449			return 0;
3450		ctx->pos = pos = pos + 1;
3451	}
3452	iter.tgid = pos - TGID_OFFSET;
3453	iter.task = NULL;
3454	for (iter = next_tgid(ns, iter);
3455	     iter.task;
3456	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3457		char name[10 + 1];
3458		unsigned int len;
3459
3460		cond_resched();
3461		if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3462			continue;
3463
3464		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3465		ctx->pos = iter.tgid + TGID_OFFSET;
3466		if (!proc_fill_cache(file, ctx, name, len,
3467				     proc_pid_instantiate, iter.task, NULL)) {
3468			put_task_struct(iter.task);
3469			return 0;
3470		}
3471	}
3472	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3473	return 0;
3474}
3475
3476/*
3477 * proc_tid_comm_permission is a special permission function exclusively
3478 * used for the node /proc/<pid>/task/<tid>/comm.
3479 * It bypasses generic permission checks in the case where a task of the same
3480 * task group attempts to access the node.
3481 * The rationale behind this is that glibc and bionic access this node for
3482 * cross thread naming (pthread_set/getname_np(!self)). However, if
3483 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3484 * which locks out the cross thread naming implementation.
3485 * This function makes sure that the node is always accessible for members of
3486 * same thread group.
3487 */
3488static int proc_tid_comm_permission(struct user_namespace *mnt_userns,
3489				    struct inode *inode, int mask)
3490{
3491	bool is_same_tgroup;
3492	struct task_struct *task;
3493
3494	task = get_proc_task(inode);
3495	if (!task)
3496		return -ESRCH;
3497	is_same_tgroup = same_thread_group(current, task);
3498	put_task_struct(task);
3499
3500	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3501		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3502		 * read or written by the members of the corresponding
3503		 * thread group.
3504		 */
3505		return 0;
3506	}
3507
3508	return generic_permission(&init_user_ns, inode, mask);
3509}
3510
3511static const struct inode_operations proc_tid_comm_inode_operations = {
3512		.permission = proc_tid_comm_permission,
3513};
3514
3515/*
3516 * Tasks
3517 */
3518static const struct pid_entry tid_base_stuff[] = {
3519	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3520	DIR("fdinfo",    S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3521	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3522#ifdef CONFIG_NET
3523	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3524#endif
3525	REG("environ",   S_IRUSR, proc_environ_operations),
3526	REG("auxv",      S_IRUSR, proc_auxv_operations),
3527	ONE("status",    S_IRUGO, proc_pid_status),
3528	ONE("personality", S_IRUSR, proc_pid_personality),
3529	ONE("limits",	 S_IRUGO, proc_pid_limits),
3530#ifdef CONFIG_SCHED_DEBUG
3531	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3532#endif
3533	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3534			 &proc_tid_comm_inode_operations,
3535			 &proc_pid_set_comm_operations, {}),
3536#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3537	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3538#endif
3539	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3540	ONE("stat",      S_IRUGO, proc_tid_stat),
3541	ONE("statm",     S_IRUGO, proc_pid_statm),
3542	REG("maps",      S_IRUGO, proc_pid_maps_operations),
3543#ifdef CONFIG_PROC_CHILDREN
3544	REG("children",  S_IRUGO, proc_tid_children_operations),
3545#endif
3546#ifdef CONFIG_NUMA
3547	REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3548#endif
3549	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3550	LNK("cwd",       proc_cwd_link),
3551	LNK("root",      proc_root_link),
3552	LNK("exe",       proc_exe_link),
3553	REG("mounts",    S_IRUGO, proc_mounts_operations),
3554	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3555#ifdef CONFIG_PROC_PAGE_MONITOR
3556	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3557	REG("smaps",     S_IRUGO, proc_pid_smaps_operations),
3558	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3559	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3560#endif
3561#ifdef CONFIG_SECURITY
3562	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3563#endif
3564#ifdef CONFIG_KALLSYMS
3565	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3566#endif
3567#ifdef CONFIG_STACKTRACE
3568	ONE("stack",      S_IRUSR, proc_pid_stack),
3569#endif
3570#ifdef CONFIG_SCHED_INFO
3571	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3572#endif
3573#ifdef CONFIG_LATENCYTOP
3574	REG("latency",  S_IRUGO, proc_lstats_operations),
3575#endif
3576#ifdef CONFIG_PROC_PID_CPUSET
3577	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3578#endif
3579#ifdef CONFIG_CGROUPS
3580	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3581#endif
3582#ifdef CONFIG_PROC_CPU_RESCTRL
3583	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3584#endif
3585	ONE("oom_score", S_IRUGO, proc_oom_score),
3586	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3587	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3588#ifdef CONFIG_AUDIT
3589	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3590	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3591#endif
3592#ifdef CONFIG_FAULT_INJECTION
3593	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3594	REG("fail-nth", 0644, proc_fail_nth_operations),
3595#endif
3596#ifdef CONFIG_TASK_IO_ACCOUNTING
3597	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3598#endif
3599#ifdef CONFIG_USER_NS
3600	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3601	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3602	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3603	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3604#endif
3605#ifdef CONFIG_LIVEPATCH
3606	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3607#endif
3608#ifdef CONFIG_PROC_PID_ARCH_STATUS
3609	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3610#endif
3611#ifdef CONFIG_SECCOMP_CACHE_DEBUG
3612	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3613#endif
3614};
3615
3616static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3617{
3618	return proc_pident_readdir(file, ctx,
3619				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3620}
3621
3622static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3623{
3624	return proc_pident_lookup(dir, dentry,
3625				  tid_base_stuff,
3626				  tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3627}
3628
3629static const struct file_operations proc_tid_base_operations = {
3630	.read		= generic_read_dir,
3631	.iterate_shared	= proc_tid_base_readdir,
3632	.llseek		= generic_file_llseek,
3633};
3634
3635static const struct inode_operations proc_tid_base_inode_operations = {
3636	.lookup		= proc_tid_base_lookup,
3637	.getattr	= pid_getattr,
3638	.setattr	= proc_setattr,
3639};
3640
3641static struct dentry *proc_task_instantiate(struct dentry *dentry,
3642	struct task_struct *task, const void *ptr)
3643{
3644	struct inode *inode;
3645	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3646	if (!inode)
3647		return ERR_PTR(-ENOENT);
3648
3649	inode->i_op = &proc_tid_base_inode_operations;
3650	inode->i_fop = &proc_tid_base_operations;
3651	inode->i_flags |= S_IMMUTABLE;
3652
3653	set_nlink(inode, nlink_tid);
3654	pid_update_inode(task, inode);
3655
3656	d_set_d_op(dentry, &pid_dentry_operations);
3657	return d_splice_alias(inode, dentry);
3658}
3659
3660static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3661{
3662	struct task_struct *task;
3663	struct task_struct *leader = get_proc_task(dir);
3664	unsigned tid;
3665	struct proc_fs_info *fs_info;
3666	struct pid_namespace *ns;
3667	struct dentry *result = ERR_PTR(-ENOENT);
3668
3669	if (!leader)
3670		goto out_no_task;
3671
3672	tid = name_to_int(&dentry->d_name);
3673	if (tid == ~0U)
3674		goto out;
3675
3676	fs_info = proc_sb_info(dentry->d_sb);
3677	ns = fs_info->pid_ns;
3678	rcu_read_lock();
3679	task = find_task_by_pid_ns(tid, ns);
3680	if (task)
3681		get_task_struct(task);
3682	rcu_read_unlock();
3683	if (!task)
3684		goto out;
3685	if (!same_thread_group(leader, task))
3686		goto out_drop_task;
3687
3688	result = proc_task_instantiate(dentry, task, NULL);
3689out_drop_task:
3690	put_task_struct(task);
3691out:
3692	put_task_struct(leader);
3693out_no_task:
3694	return result;
3695}
3696
3697/*
3698 * Find the first tid of a thread group to return to user space.
3699 *
3700 * Usually this is just the thread group leader, but if the users
3701 * buffer was too small or there was a seek into the middle of the
3702 * directory we have more work todo.
3703 *
3704 * In the case of a short read we start with find_task_by_pid.
3705 *
3706 * In the case of a seek we start with the leader and walk nr
3707 * threads past it.
3708 */
3709static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3710					struct pid_namespace *ns)
3711{
3712	struct task_struct *pos, *task;
3713	unsigned long nr = f_pos;
3714
3715	if (nr != f_pos)	/* 32bit overflow? */
3716		return NULL;
3717
3718	rcu_read_lock();
3719	task = pid_task(pid, PIDTYPE_PID);
3720	if (!task)
3721		goto fail;
3722
3723	/* Attempt to start with the tid of a thread */
3724	if (tid && nr) {
3725		pos = find_task_by_pid_ns(tid, ns);
3726		if (pos && same_thread_group(pos, task))
3727			goto found;
3728	}
3729
3730	/* If nr exceeds the number of threads there is nothing todo */
3731	if (nr >= get_nr_threads(task))
3732		goto fail;
3733
3734	/* If we haven't found our starting place yet start
3735	 * with the leader and walk nr threads forward.
3736	 */
3737	pos = task = task->group_leader;
3738	do {
3739		if (!nr--)
3740			goto found;
3741	} while_each_thread(task, pos);
3742fail:
3743	pos = NULL;
3744	goto out;
3745found:
3746	get_task_struct(pos);
3747out:
3748	rcu_read_unlock();
3749	return pos;
3750}
3751
3752/*
3753 * Find the next thread in the thread list.
3754 * Return NULL if there is an error or no next thread.
3755 *
3756 * The reference to the input task_struct is released.
3757 */
3758static struct task_struct *next_tid(struct task_struct *start)
3759{
3760	struct task_struct *pos = NULL;
3761	rcu_read_lock();
3762	if (pid_alive(start)) {
3763		pos = next_thread(start);
3764		if (thread_group_leader(pos))
3765			pos = NULL;
3766		else
3767			get_task_struct(pos);
3768	}
3769	rcu_read_unlock();
3770	put_task_struct(start);
3771	return pos;
3772}
3773
3774/* for the /proc/TGID/task/ directories */
3775static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3776{
3777	struct inode *inode = file_inode(file);
3778	struct task_struct *task;
3779	struct pid_namespace *ns;
3780	int tid;
3781
3782	if (proc_inode_is_dead(inode))
3783		return -ENOENT;
3784
3785	if (!dir_emit_dots(file, ctx))
3786		return 0;
3787
3788	/* f_version caches the tgid value that the last readdir call couldn't
3789	 * return. lseek aka telldir automagically resets f_version to 0.
3790	 */
3791	ns = proc_pid_ns(inode->i_sb);
3792	tid = (int)file->f_version;
3793	file->f_version = 0;
3794	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3795	     task;
3796	     task = next_tid(task), ctx->pos++) {
3797		char name[10 + 1];
3798		unsigned int len;
3799		tid = task_pid_nr_ns(task, ns);
3800		len = snprintf(name, sizeof(name), "%u", tid);
3801		if (!proc_fill_cache(file, ctx, name, len,
3802				proc_task_instantiate, task, NULL)) {
3803			/* returning this tgid failed, save it as the first
3804			 * pid for the next readir call */
3805			file->f_version = (u64)tid;
3806			put_task_struct(task);
3807			break;
3808		}
3809	}
3810
3811	return 0;
3812}
3813
3814static int proc_task_getattr(struct user_namespace *mnt_userns,
3815			     const struct path *path, struct kstat *stat,
3816			     u32 request_mask, unsigned int query_flags)
3817{
3818	struct inode *inode = d_inode(path->dentry);
3819	struct task_struct *p = get_proc_task(inode);
3820	generic_fillattr(&init_user_ns, inode, stat);
3821
3822	if (p) {
3823		stat->nlink += get_nr_threads(p);
3824		put_task_struct(p);
3825	}
3826
3827	return 0;
3828}
3829
3830static const struct inode_operations proc_task_inode_operations = {
3831	.lookup		= proc_task_lookup,
3832	.getattr	= proc_task_getattr,
3833	.setattr	= proc_setattr,
3834	.permission	= proc_pid_permission,
3835};
3836
3837static const struct file_operations proc_task_operations = {
3838	.read		= generic_read_dir,
3839	.iterate_shared	= proc_task_readdir,
3840	.llseek		= generic_file_llseek,
3841};
3842
3843void __init set_proc_pid_nlink(void)
3844{
3845	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3846	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3847}
v5.9
   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/fdtable.h>
  62#include <linux/generic-radix-tree.h>
  63#include <linux/string.h>
  64#include <linux/seq_file.h>
  65#include <linux/namei.h>
  66#include <linux/mnt_namespace.h>
  67#include <linux/mm.h>
  68#include <linux/swap.h>
  69#include <linux/rcupdate.h>
  70#include <linux/kallsyms.h>
  71#include <linux/stacktrace.h>
  72#include <linux/resource.h>
  73#include <linux/module.h>
  74#include <linux/mount.h>
  75#include <linux/security.h>
  76#include <linux/ptrace.h>
  77#include <linux/tracehook.h>
  78#include <linux/printk.h>
  79#include <linux/cache.h>
  80#include <linux/cgroup.h>
  81#include <linux/cpuset.h>
  82#include <linux/audit.h>
  83#include <linux/poll.h>
  84#include <linux/nsproxy.h>
  85#include <linux/oom.h>
  86#include <linux/elf.h>
  87#include <linux/pid_namespace.h>
  88#include <linux/user_namespace.h>
  89#include <linux/fs_struct.h>
  90#include <linux/slab.h>
  91#include <linux/sched/autogroup.h>
  92#include <linux/sched/mm.h>
  93#include <linux/sched/coredump.h>
  94#include <linux/sched/debug.h>
  95#include <linux/sched/stat.h>
  96#include <linux/posix-timers.h>
  97#include <linux/time_namespace.h>
  98#include <linux/resctrl.h>
  99#include <trace/events/oom.h>
 100#include "internal.h"
 101#include "fd.h"
 102
 103#include "../../lib/kstrtox.h"
 104
 105/* NOTE:
 106 *	Implementing inode permission operations in /proc is almost
 107 *	certainly an error.  Permission checks need to happen during
 108 *	each system call not at open time.  The reason is that most of
 109 *	what we wish to check for permissions in /proc varies at runtime.
 110 *
 111 *	The classic example of a problem is opening file descriptors
 112 *	in /proc for a task before it execs a suid executable.
 113 */
 114
 115static u8 nlink_tid __ro_after_init;
 116static u8 nlink_tgid __ro_after_init;
 117
 118struct pid_entry {
 119	const char *name;
 120	unsigned int len;
 121	umode_t mode;
 122	const struct inode_operations *iop;
 123	const struct file_operations *fop;
 124	union proc_op op;
 125};
 126
 127#define NOD(NAME, MODE, IOP, FOP, OP) {			\
 128	.name = (NAME),					\
 129	.len  = sizeof(NAME) - 1,			\
 130	.mode = MODE,					\
 131	.iop  = IOP,					\
 132	.fop  = FOP,					\
 133	.op   = OP,					\
 134}
 135
 136#define DIR(NAME, MODE, iops, fops)	\
 137	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
 138#define LNK(NAME, get_link)					\
 139	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
 140		&proc_pid_link_inode_operations, NULL,		\
 141		{ .proc_get_link = get_link } )
 142#define REG(NAME, MODE, fops)				\
 143	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
 144#define ONE(NAME, MODE, show)				\
 145	NOD(NAME, (S_IFREG|(MODE)),			\
 146		NULL, &proc_single_file_operations,	\
 147		{ .proc_show = show } )
 148#define ATTR(LSM, NAME, MODE)				\
 149	NOD(NAME, (S_IFREG|(MODE)),			\
 150		NULL, &proc_pid_attr_operations,	\
 151		{ .lsm = LSM })
 152
 153/*
 154 * Count the number of hardlinks for the pid_entry table, excluding the .
 155 * and .. links.
 156 */
 157static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
 158	unsigned int n)
 159{
 160	unsigned int i;
 161	unsigned int count;
 162
 163	count = 2;
 164	for (i = 0; i < n; ++i) {
 165		if (S_ISDIR(entries[i].mode))
 166			++count;
 167	}
 168
 169	return count;
 170}
 171
 172static int get_task_root(struct task_struct *task, struct path *root)
 173{
 174	int result = -ENOENT;
 175
 176	task_lock(task);
 177	if (task->fs) {
 178		get_fs_root(task->fs, root);
 179		result = 0;
 180	}
 181	task_unlock(task);
 182	return result;
 183}
 184
 185static int proc_cwd_link(struct dentry *dentry, struct path *path)
 186{
 187	struct task_struct *task = get_proc_task(d_inode(dentry));
 188	int result = -ENOENT;
 189
 190	if (task) {
 191		task_lock(task);
 192		if (task->fs) {
 193			get_fs_pwd(task->fs, path);
 194			result = 0;
 195		}
 196		task_unlock(task);
 197		put_task_struct(task);
 198	}
 199	return result;
 200}
 201
 202static int proc_root_link(struct dentry *dentry, struct path *path)
 203{
 204	struct task_struct *task = get_proc_task(d_inode(dentry));
 205	int result = -ENOENT;
 206
 207	if (task) {
 208		result = get_task_root(task, path);
 209		put_task_struct(task);
 210	}
 211	return result;
 212}
 213
 214/*
 215 * If the user used setproctitle(), we just get the string from
 216 * user space at arg_start, and limit it to a maximum of one page.
 217 */
 218static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
 219				size_t count, unsigned long pos,
 220				unsigned long arg_start)
 221{
 222	char *page;
 223	int ret, got;
 224
 225	if (pos >= PAGE_SIZE)
 226		return 0;
 227
 228	page = (char *)__get_free_page(GFP_KERNEL);
 229	if (!page)
 230		return -ENOMEM;
 231
 232	ret = 0;
 233	got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
 234	if (got > 0) {
 235		int len = strnlen(page, got);
 236
 237		/* Include the NUL character if it was found */
 238		if (len < got)
 239			len++;
 240
 241		if (len > pos) {
 242			len -= pos;
 243			if (len > count)
 244				len = count;
 245			len -= copy_to_user(buf, page+pos, len);
 246			if (!len)
 247				len = -EFAULT;
 248			ret = len;
 249		}
 250	}
 251	free_page((unsigned long)page);
 252	return ret;
 253}
 254
 255static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
 256			      size_t count, loff_t *ppos)
 257{
 258	unsigned long arg_start, arg_end, env_start, env_end;
 259	unsigned long pos, len;
 260	char *page, c;
 261
 262	/* Check if process spawned far enough to have cmdline. */
 263	if (!mm->env_end)
 264		return 0;
 265
 266	spin_lock(&mm->arg_lock);
 267	arg_start = mm->arg_start;
 268	arg_end = mm->arg_end;
 269	env_start = mm->env_start;
 270	env_end = mm->env_end;
 271	spin_unlock(&mm->arg_lock);
 272
 273	if (arg_start >= arg_end)
 274		return 0;
 275
 276	/*
 277	 * We allow setproctitle() to overwrite the argument
 278	 * strings, and overflow past the original end. But
 279	 * only when it overflows into the environment area.
 280	 */
 281	if (env_start != arg_end || env_end < env_start)
 282		env_start = env_end = arg_end;
 283	len = env_end - arg_start;
 284
 285	/* We're not going to care if "*ppos" has high bits set */
 286	pos = *ppos;
 287	if (pos >= len)
 288		return 0;
 289	if (count > len - pos)
 290		count = len - pos;
 291	if (!count)
 292		return 0;
 293
 294	/*
 295	 * Magical special case: if the argv[] end byte is not
 296	 * zero, the user has overwritten it with setproctitle(3).
 297	 *
 298	 * Possible future enhancement: do this only once when
 299	 * pos is 0, and set a flag in the 'struct file'.
 300	 */
 301	if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
 302		return get_mm_proctitle(mm, buf, count, pos, arg_start);
 303
 304	/*
 305	 * For the non-setproctitle() case we limit things strictly
 306	 * to the [arg_start, arg_end[ range.
 307	 */
 308	pos += arg_start;
 309	if (pos < arg_start || pos >= arg_end)
 310		return 0;
 311	if (count > arg_end - pos)
 312		count = arg_end - pos;
 313
 314	page = (char *)__get_free_page(GFP_KERNEL);
 315	if (!page)
 316		return -ENOMEM;
 317
 318	len = 0;
 319	while (count) {
 320		int got;
 321		size_t size = min_t(size_t, PAGE_SIZE, count);
 322
 323		got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
 324		if (got <= 0)
 325			break;
 326		got -= copy_to_user(buf, page, got);
 327		if (unlikely(!got)) {
 328			if (!len)
 329				len = -EFAULT;
 330			break;
 331		}
 332		pos += got;
 333		buf += got;
 334		len += got;
 335		count -= got;
 336	}
 337
 338	free_page((unsigned long)page);
 339	return len;
 340}
 341
 342static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
 343				size_t count, loff_t *pos)
 344{
 345	struct mm_struct *mm;
 346	ssize_t ret;
 347
 348	mm = get_task_mm(tsk);
 349	if (!mm)
 350		return 0;
 351
 352	ret = get_mm_cmdline(mm, buf, count, pos);
 353	mmput(mm);
 354	return ret;
 355}
 356
 357static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 358				     size_t count, loff_t *pos)
 359{
 360	struct task_struct *tsk;
 361	ssize_t ret;
 362
 363	BUG_ON(*pos < 0);
 364
 365	tsk = get_proc_task(file_inode(file));
 366	if (!tsk)
 367		return -ESRCH;
 368	ret = get_task_cmdline(tsk, buf, count, pos);
 369	put_task_struct(tsk);
 370	if (ret > 0)
 371		*pos += ret;
 372	return ret;
 373}
 374
 375static const struct file_operations proc_pid_cmdline_ops = {
 376	.read	= proc_pid_cmdline_read,
 377	.llseek	= generic_file_llseek,
 378};
 379
 380#ifdef CONFIG_KALLSYMS
 381/*
 382 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 383 * Returns the resolved symbol.  If that fails, simply return the address.
 384 */
 385static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
 386			  struct pid *pid, struct task_struct *task)
 387{
 388	unsigned long wchan;
 389	char symname[KSYM_NAME_LEN];
 390
 391	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
 392		goto print0;
 
 
 393
 394	wchan = get_wchan(task);
 395	if (wchan && !lookup_symbol_name(wchan, symname)) {
 396		seq_puts(m, symname);
 397		return 0;
 398	}
 399
 400print0:
 401	seq_putc(m, '0');
 402	return 0;
 403}
 404#endif /* CONFIG_KALLSYMS */
 405
 406static int lock_trace(struct task_struct *task)
 407{
 408	int err = mutex_lock_killable(&task->signal->exec_update_mutex);
 409	if (err)
 410		return err;
 411	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
 412		mutex_unlock(&task->signal->exec_update_mutex);
 413		return -EPERM;
 414	}
 415	return 0;
 416}
 417
 418static void unlock_trace(struct task_struct *task)
 419{
 420	mutex_unlock(&task->signal->exec_update_mutex);
 421}
 422
 423#ifdef CONFIG_STACKTRACE
 424
 425#define MAX_STACK_TRACE_DEPTH	64
 426
 427static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
 428			  struct pid *pid, struct task_struct *task)
 429{
 430	unsigned long *entries;
 431	int err;
 432
 433	/*
 434	 * The ability to racily run the kernel stack unwinder on a running task
 435	 * and then observe the unwinder output is scary; while it is useful for
 436	 * debugging kernel issues, it can also allow an attacker to leak kernel
 437	 * stack contents.
 438	 * Doing this in a manner that is at least safe from races would require
 439	 * some work to ensure that the remote task can not be scheduled; and
 440	 * even then, this would still expose the unwinder as local attack
 441	 * surface.
 442	 * Therefore, this interface is restricted to root.
 443	 */
 444	if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
 445		return -EACCES;
 446
 447	entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
 448				GFP_KERNEL);
 449	if (!entries)
 450		return -ENOMEM;
 451
 452	err = lock_trace(task);
 453	if (!err) {
 454		unsigned int i, nr_entries;
 455
 456		nr_entries = stack_trace_save_tsk(task, entries,
 457						  MAX_STACK_TRACE_DEPTH, 0);
 458
 459		for (i = 0; i < nr_entries; i++) {
 460			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
 461		}
 462
 463		unlock_trace(task);
 464	}
 465	kfree(entries);
 466
 467	return err;
 468}
 469#endif
 470
 471#ifdef CONFIG_SCHED_INFO
 472/*
 473 * Provides /proc/PID/schedstat
 474 */
 475static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
 476			      struct pid *pid, struct task_struct *task)
 477{
 478	if (unlikely(!sched_info_on()))
 479		seq_puts(m, "0 0 0\n");
 480	else
 481		seq_printf(m, "%llu %llu %lu\n",
 482		   (unsigned long long)task->se.sum_exec_runtime,
 483		   (unsigned long long)task->sched_info.run_delay,
 484		   task->sched_info.pcount);
 485
 486	return 0;
 487}
 488#endif
 489
 490#ifdef CONFIG_LATENCYTOP
 491static int lstats_show_proc(struct seq_file *m, void *v)
 492{
 493	int i;
 494	struct inode *inode = m->private;
 495	struct task_struct *task = get_proc_task(inode);
 496
 497	if (!task)
 498		return -ESRCH;
 499	seq_puts(m, "Latency Top version : v0.1\n");
 500	for (i = 0; i < LT_SAVECOUNT; i++) {
 501		struct latency_record *lr = &task->latency_record[i];
 502		if (lr->backtrace[0]) {
 503			int q;
 504			seq_printf(m, "%i %li %li",
 505				   lr->count, lr->time, lr->max);
 506			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
 507				unsigned long bt = lr->backtrace[q];
 508
 509				if (!bt)
 510					break;
 511				seq_printf(m, " %ps", (void *)bt);
 512			}
 513			seq_putc(m, '\n');
 514		}
 515
 516	}
 517	put_task_struct(task);
 518	return 0;
 519}
 520
 521static int lstats_open(struct inode *inode, struct file *file)
 522{
 523	return single_open(file, lstats_show_proc, inode);
 524}
 525
 526static ssize_t lstats_write(struct file *file, const char __user *buf,
 527			    size_t count, loff_t *offs)
 528{
 529	struct task_struct *task = get_proc_task(file_inode(file));
 530
 531	if (!task)
 532		return -ESRCH;
 533	clear_tsk_latency_tracing(task);
 534	put_task_struct(task);
 535
 536	return count;
 537}
 538
 539static const struct file_operations proc_lstats_operations = {
 540	.open		= lstats_open,
 541	.read		= seq_read,
 542	.write		= lstats_write,
 543	.llseek		= seq_lseek,
 544	.release	= single_release,
 545};
 546
 547#endif
 548
 549static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 550			  struct pid *pid, struct task_struct *task)
 551{
 552	unsigned long totalpages = totalram_pages() + total_swap_pages;
 553	unsigned long points = 0;
 554	long badness;
 555
 556	badness = oom_badness(task, totalpages);
 557	/*
 558	 * Special case OOM_SCORE_ADJ_MIN for all others scale the
 559	 * badness value into [0, 2000] range which we have been
 560	 * exporting for a long time so userspace might depend on it.
 561	 */
 562	if (badness != LONG_MIN)
 563		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
 564
 565	seq_printf(m, "%lu\n", points);
 566
 567	return 0;
 568}
 569
 570struct limit_names {
 571	const char *name;
 572	const char *unit;
 573};
 574
 575static const struct limit_names lnames[RLIM_NLIMITS] = {
 576	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
 577	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
 578	[RLIMIT_DATA] = {"Max data size", "bytes"},
 579	[RLIMIT_STACK] = {"Max stack size", "bytes"},
 580	[RLIMIT_CORE] = {"Max core file size", "bytes"},
 581	[RLIMIT_RSS] = {"Max resident set", "bytes"},
 582	[RLIMIT_NPROC] = {"Max processes", "processes"},
 583	[RLIMIT_NOFILE] = {"Max open files", "files"},
 584	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
 585	[RLIMIT_AS] = {"Max address space", "bytes"},
 586	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
 587	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
 588	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
 589	[RLIMIT_NICE] = {"Max nice priority", NULL},
 590	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
 591	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
 592};
 593
 594/* Display limits for a process */
 595static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
 596			   struct pid *pid, struct task_struct *task)
 597{
 598	unsigned int i;
 599	unsigned long flags;
 600
 601	struct rlimit rlim[RLIM_NLIMITS];
 602
 603	if (!lock_task_sighand(task, &flags))
 604		return 0;
 605	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
 606	unlock_task_sighand(task, &flags);
 607
 608	/*
 609	 * print the file header
 610	 */
 611	seq_puts(m, "Limit                     "
 612		"Soft Limit           "
 613		"Hard Limit           "
 614		"Units     \n");
 615
 616	for (i = 0; i < RLIM_NLIMITS; i++) {
 617		if (rlim[i].rlim_cur == RLIM_INFINITY)
 618			seq_printf(m, "%-25s %-20s ",
 619				   lnames[i].name, "unlimited");
 620		else
 621			seq_printf(m, "%-25s %-20lu ",
 622				   lnames[i].name, rlim[i].rlim_cur);
 623
 624		if (rlim[i].rlim_max == RLIM_INFINITY)
 625			seq_printf(m, "%-20s ", "unlimited");
 626		else
 627			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
 628
 629		if (lnames[i].unit)
 630			seq_printf(m, "%-10s\n", lnames[i].unit);
 631		else
 632			seq_putc(m, '\n');
 633	}
 634
 635	return 0;
 636}
 637
 638#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 639static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
 640			    struct pid *pid, struct task_struct *task)
 641{
 642	struct syscall_info info;
 643	u64 *args = &info.data.args[0];
 644	int res;
 645
 646	res = lock_trace(task);
 647	if (res)
 648		return res;
 649
 650	if (task_current_syscall(task, &info))
 651		seq_puts(m, "running\n");
 652	else if (info.data.nr < 0)
 653		seq_printf(m, "%d 0x%llx 0x%llx\n",
 654			   info.data.nr, info.sp, info.data.instruction_pointer);
 655	else
 656		seq_printf(m,
 657		       "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
 658		       info.data.nr,
 659		       args[0], args[1], args[2], args[3], args[4], args[5],
 660		       info.sp, info.data.instruction_pointer);
 661	unlock_trace(task);
 662
 663	return 0;
 664}
 665#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
 666
 667/************************************************************************/
 668/*                       Here the fs part begins                        */
 669/************************************************************************/
 670
 671/* permission checks */
 672static int proc_fd_access_allowed(struct inode *inode)
 673{
 674	struct task_struct *task;
 675	int allowed = 0;
 676	/* Allow access to a task's file descriptors if it is us or we
 677	 * may use ptrace attach to the process and find out that
 678	 * information.
 679	 */
 680	task = get_proc_task(inode);
 681	if (task) {
 682		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 683		put_task_struct(task);
 684	}
 685	return allowed;
 686}
 687
 688int proc_setattr(struct dentry *dentry, struct iattr *attr)
 
 689{
 690	int error;
 691	struct inode *inode = d_inode(dentry);
 692
 693	if (attr->ia_valid & ATTR_MODE)
 694		return -EPERM;
 695
 696	error = setattr_prepare(dentry, attr);
 697	if (error)
 698		return error;
 699
 700	setattr_copy(inode, attr);
 701	mark_inode_dirty(inode);
 702	return 0;
 703}
 704
 705/*
 706 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
 707 * or euid/egid (for hide_pid_min=2)?
 708 */
 709static bool has_pid_permissions(struct proc_fs_info *fs_info,
 710				 struct task_struct *task,
 711				 enum proc_hidepid hide_pid_min)
 712{
 713	/*
 714	 * If 'hidpid' mount option is set force a ptrace check,
 715	 * we indicate that we are using a filesystem syscall
 716	 * by passing PTRACE_MODE_READ_FSCREDS
 717	 */
 718	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
 719		return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 720
 721	if (fs_info->hide_pid < hide_pid_min)
 722		return true;
 723	if (in_group_p(fs_info->pid_gid))
 724		return true;
 725	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 726}
 727
 728
 729static int proc_pid_permission(struct inode *inode, int mask)
 
 730{
 731	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
 732	struct task_struct *task;
 733	bool has_perms;
 734
 735	task = get_proc_task(inode);
 736	if (!task)
 737		return -ESRCH;
 738	has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
 739	put_task_struct(task);
 740
 741	if (!has_perms) {
 742		if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
 743			/*
 744			 * Let's make getdents(), stat(), and open()
 745			 * consistent with each other.  If a process
 746			 * may not stat() a file, it shouldn't be seen
 747			 * in procfs at all.
 748			 */
 749			return -ENOENT;
 750		}
 751
 752		return -EPERM;
 753	}
 754	return generic_permission(inode, mask);
 755}
 756
 757
 758
 759static const struct inode_operations proc_def_inode_operations = {
 760	.setattr	= proc_setattr,
 761};
 762
 763static int proc_single_show(struct seq_file *m, void *v)
 764{
 765	struct inode *inode = m->private;
 766	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
 767	struct pid *pid = proc_pid(inode);
 768	struct task_struct *task;
 769	int ret;
 770
 771	task = get_pid_task(pid, PIDTYPE_PID);
 772	if (!task)
 773		return -ESRCH;
 774
 775	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
 776
 777	put_task_struct(task);
 778	return ret;
 779}
 780
 781static int proc_single_open(struct inode *inode, struct file *filp)
 782{
 783	return single_open(filp, proc_single_show, inode);
 784}
 785
 786static const struct file_operations proc_single_file_operations = {
 787	.open		= proc_single_open,
 788	.read		= seq_read,
 789	.llseek		= seq_lseek,
 790	.release	= single_release,
 791};
 792
 793
 794struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
 795{
 796	struct task_struct *task = get_proc_task(inode);
 797	struct mm_struct *mm = ERR_PTR(-ESRCH);
 798
 799	if (task) {
 800		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
 801		put_task_struct(task);
 802
 803		if (!IS_ERR_OR_NULL(mm)) {
 804			/* ensure this mm_struct can't be freed */
 805			mmgrab(mm);
 806			/* but do not pin its memory */
 807			mmput(mm);
 808		}
 809	}
 810
 811	return mm;
 812}
 813
 814static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
 815{
 816	struct mm_struct *mm = proc_mem_open(inode, mode);
 817
 818	if (IS_ERR(mm))
 819		return PTR_ERR(mm);
 820
 821	file->private_data = mm;
 822	return 0;
 823}
 824
 825static int mem_open(struct inode *inode, struct file *file)
 826{
 827	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
 828
 829	/* OK to pass negative loff_t, we can catch out-of-range */
 830	file->f_mode |= FMODE_UNSIGNED_OFFSET;
 831
 832	return ret;
 833}
 834
 835static ssize_t mem_rw(struct file *file, char __user *buf,
 836			size_t count, loff_t *ppos, int write)
 837{
 838	struct mm_struct *mm = file->private_data;
 839	unsigned long addr = *ppos;
 840	ssize_t copied;
 841	char *page;
 842	unsigned int flags;
 843
 844	if (!mm)
 845		return 0;
 846
 847	page = (char *)__get_free_page(GFP_KERNEL);
 848	if (!page)
 849		return -ENOMEM;
 850
 851	copied = 0;
 852	if (!mmget_not_zero(mm))
 853		goto free;
 854
 855	flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
 856
 857	while (count > 0) {
 858		int this_len = min_t(int, count, PAGE_SIZE);
 859
 860		if (write && copy_from_user(page, buf, this_len)) {
 861			copied = -EFAULT;
 862			break;
 863		}
 864
 865		this_len = access_remote_vm(mm, addr, page, this_len, flags);
 866		if (!this_len) {
 867			if (!copied)
 868				copied = -EIO;
 869			break;
 870		}
 871
 872		if (!write && copy_to_user(buf, page, this_len)) {
 873			copied = -EFAULT;
 874			break;
 875		}
 876
 877		buf += this_len;
 878		addr += this_len;
 879		copied += this_len;
 880		count -= this_len;
 881	}
 882	*ppos = addr;
 883
 884	mmput(mm);
 885free:
 886	free_page((unsigned long) page);
 887	return copied;
 888}
 889
 890static ssize_t mem_read(struct file *file, char __user *buf,
 891			size_t count, loff_t *ppos)
 892{
 893	return mem_rw(file, buf, count, ppos, 0);
 894}
 895
 896static ssize_t mem_write(struct file *file, const char __user *buf,
 897			 size_t count, loff_t *ppos)
 898{
 899	return mem_rw(file, (char __user*)buf, count, ppos, 1);
 900}
 901
 902loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 903{
 904	switch (orig) {
 905	case 0:
 906		file->f_pos = offset;
 907		break;
 908	case 1:
 909		file->f_pos += offset;
 910		break;
 911	default:
 912		return -EINVAL;
 913	}
 914	force_successful_syscall_return();
 915	return file->f_pos;
 916}
 917
 918static int mem_release(struct inode *inode, struct file *file)
 919{
 920	struct mm_struct *mm = file->private_data;
 921	if (mm)
 922		mmdrop(mm);
 923	return 0;
 924}
 925
 926static const struct file_operations proc_mem_operations = {
 927	.llseek		= mem_lseek,
 928	.read		= mem_read,
 929	.write		= mem_write,
 930	.open		= mem_open,
 931	.release	= mem_release,
 932};
 933
 934static int environ_open(struct inode *inode, struct file *file)
 935{
 936	return __mem_open(inode, file, PTRACE_MODE_READ);
 937}
 938
 939static ssize_t environ_read(struct file *file, char __user *buf,
 940			size_t count, loff_t *ppos)
 941{
 942	char *page;
 943	unsigned long src = *ppos;
 944	int ret = 0;
 945	struct mm_struct *mm = file->private_data;
 946	unsigned long env_start, env_end;
 947
 948	/* Ensure the process spawned far enough to have an environment. */
 949	if (!mm || !mm->env_end)
 950		return 0;
 951
 952	page = (char *)__get_free_page(GFP_KERNEL);
 953	if (!page)
 954		return -ENOMEM;
 955
 956	ret = 0;
 957	if (!mmget_not_zero(mm))
 958		goto free;
 959
 960	spin_lock(&mm->arg_lock);
 961	env_start = mm->env_start;
 962	env_end = mm->env_end;
 963	spin_unlock(&mm->arg_lock);
 964
 965	while (count > 0) {
 966		size_t this_len, max_len;
 967		int retval;
 968
 969		if (src >= (env_end - env_start))
 970			break;
 971
 972		this_len = env_end - (env_start + src);
 973
 974		max_len = min_t(size_t, PAGE_SIZE, count);
 975		this_len = min(max_len, this_len);
 976
 977		retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
 978
 979		if (retval <= 0) {
 980			ret = retval;
 981			break;
 982		}
 983
 984		if (copy_to_user(buf, page, retval)) {
 985			ret = -EFAULT;
 986			break;
 987		}
 988
 989		ret += retval;
 990		src += retval;
 991		buf += retval;
 992		count -= retval;
 993	}
 994	*ppos = src;
 995	mmput(mm);
 996
 997free:
 998	free_page((unsigned long) page);
 999	return ret;
1000}
1001
1002static const struct file_operations proc_environ_operations = {
1003	.open		= environ_open,
1004	.read		= environ_read,
1005	.llseek		= generic_file_llseek,
1006	.release	= mem_release,
1007};
1008
1009static int auxv_open(struct inode *inode, struct file *file)
1010{
1011	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1012}
1013
1014static ssize_t auxv_read(struct file *file, char __user *buf,
1015			size_t count, loff_t *ppos)
1016{
1017	struct mm_struct *mm = file->private_data;
1018	unsigned int nwords = 0;
1019
1020	if (!mm)
1021		return 0;
1022	do {
1023		nwords += 2;
1024	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1025	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1026				       nwords * sizeof(mm->saved_auxv[0]));
1027}
1028
1029static const struct file_operations proc_auxv_operations = {
1030	.open		= auxv_open,
1031	.read		= auxv_read,
1032	.llseek		= generic_file_llseek,
1033	.release	= mem_release,
1034};
1035
1036static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1037			    loff_t *ppos)
1038{
1039	struct task_struct *task = get_proc_task(file_inode(file));
1040	char buffer[PROC_NUMBUF];
1041	int oom_adj = OOM_ADJUST_MIN;
1042	size_t len;
1043
1044	if (!task)
1045		return -ESRCH;
1046	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1047		oom_adj = OOM_ADJUST_MAX;
1048	else
1049		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1050			  OOM_SCORE_ADJ_MAX;
1051	put_task_struct(task);
 
 
1052	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1053	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1054}
1055
1056static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1057{
1058	static DEFINE_MUTEX(oom_adj_mutex);
1059	struct mm_struct *mm = NULL;
1060	struct task_struct *task;
1061	int err = 0;
1062
1063	task = get_proc_task(file_inode(file));
1064	if (!task)
1065		return -ESRCH;
1066
1067	mutex_lock(&oom_adj_mutex);
1068	if (legacy) {
1069		if (oom_adj < task->signal->oom_score_adj &&
1070				!capable(CAP_SYS_RESOURCE)) {
1071			err = -EACCES;
1072			goto err_unlock;
1073		}
1074		/*
1075		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1076		 * /proc/pid/oom_score_adj instead.
1077		 */
1078		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1079			  current->comm, task_pid_nr(current), task_pid_nr(task),
1080			  task_pid_nr(task));
1081	} else {
1082		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1083				!capable(CAP_SYS_RESOURCE)) {
1084			err = -EACCES;
1085			goto err_unlock;
1086		}
1087	}
1088
1089	/*
1090	 * Make sure we will check other processes sharing the mm if this is
1091	 * not vfrok which wants its own oom_score_adj.
1092	 * pin the mm so it doesn't go away and get reused after task_unlock
1093	 */
1094	if (!task->vfork_done) {
1095		struct task_struct *p = find_lock_task_mm(task);
1096
1097		if (p) {
1098			if (atomic_read(&p->mm->mm_users) > 1) {
1099				mm = p->mm;
1100				mmgrab(mm);
1101			}
1102			task_unlock(p);
1103		}
1104	}
1105
1106	task->signal->oom_score_adj = oom_adj;
1107	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1108		task->signal->oom_score_adj_min = (short)oom_adj;
1109	trace_oom_score_adj_update(task);
1110
1111	if (mm) {
1112		struct task_struct *p;
1113
1114		rcu_read_lock();
1115		for_each_process(p) {
1116			if (same_thread_group(task, p))
1117				continue;
1118
1119			/* do not touch kernel threads or the global init */
1120			if (p->flags & PF_KTHREAD || is_global_init(p))
1121				continue;
1122
1123			task_lock(p);
1124			if (!p->vfork_done && process_shares_mm(p, mm)) {
1125				p->signal->oom_score_adj = oom_adj;
1126				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1127					p->signal->oom_score_adj_min = (short)oom_adj;
1128			}
1129			task_unlock(p);
1130		}
1131		rcu_read_unlock();
1132		mmdrop(mm);
1133	}
1134err_unlock:
1135	mutex_unlock(&oom_adj_mutex);
1136	put_task_struct(task);
1137	return err;
1138}
1139
1140/*
1141 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1142 * kernels.  The effective policy is defined by oom_score_adj, which has a
1143 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1144 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1145 * Processes that become oom disabled via oom_adj will still be oom disabled
1146 * with this implementation.
1147 *
1148 * oom_adj cannot be removed since existing userspace binaries use it.
1149 */
1150static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1151			     size_t count, loff_t *ppos)
1152{
1153	char buffer[PROC_NUMBUF];
1154	int oom_adj;
1155	int err;
1156
1157	memset(buffer, 0, sizeof(buffer));
1158	if (count > sizeof(buffer) - 1)
1159		count = sizeof(buffer) - 1;
1160	if (copy_from_user(buffer, buf, count)) {
1161		err = -EFAULT;
1162		goto out;
1163	}
1164
1165	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1166	if (err)
1167		goto out;
1168	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1169	     oom_adj != OOM_DISABLE) {
1170		err = -EINVAL;
1171		goto out;
1172	}
1173
1174	/*
1175	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1176	 * value is always attainable.
1177	 */
1178	if (oom_adj == OOM_ADJUST_MAX)
1179		oom_adj = OOM_SCORE_ADJ_MAX;
1180	else
1181		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1182
1183	err = __set_oom_adj(file, oom_adj, true);
1184out:
1185	return err < 0 ? err : count;
1186}
1187
1188static const struct file_operations proc_oom_adj_operations = {
1189	.read		= oom_adj_read,
1190	.write		= oom_adj_write,
1191	.llseek		= generic_file_llseek,
1192};
1193
1194static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1195					size_t count, loff_t *ppos)
1196{
1197	struct task_struct *task = get_proc_task(file_inode(file));
1198	char buffer[PROC_NUMBUF];
1199	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1200	size_t len;
1201
1202	if (!task)
1203		return -ESRCH;
1204	oom_score_adj = task->signal->oom_score_adj;
1205	put_task_struct(task);
1206	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1207	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1208}
1209
1210static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1211					size_t count, loff_t *ppos)
1212{
1213	char buffer[PROC_NUMBUF];
1214	int oom_score_adj;
1215	int err;
1216
1217	memset(buffer, 0, sizeof(buffer));
1218	if (count > sizeof(buffer) - 1)
1219		count = sizeof(buffer) - 1;
1220	if (copy_from_user(buffer, buf, count)) {
1221		err = -EFAULT;
1222		goto out;
1223	}
1224
1225	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1226	if (err)
1227		goto out;
1228	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1229			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1230		err = -EINVAL;
1231		goto out;
1232	}
1233
1234	err = __set_oom_adj(file, oom_score_adj, false);
1235out:
1236	return err < 0 ? err : count;
1237}
1238
1239static const struct file_operations proc_oom_score_adj_operations = {
1240	.read		= oom_score_adj_read,
1241	.write		= oom_score_adj_write,
1242	.llseek		= default_llseek,
1243};
1244
1245#ifdef CONFIG_AUDIT
1246#define TMPBUFLEN 11
1247static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1248				  size_t count, loff_t *ppos)
1249{
1250	struct inode * inode = file_inode(file);
1251	struct task_struct *task = get_proc_task(inode);
1252	ssize_t length;
1253	char tmpbuf[TMPBUFLEN];
1254
1255	if (!task)
1256		return -ESRCH;
1257	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1258			   from_kuid(file->f_cred->user_ns,
1259				     audit_get_loginuid(task)));
1260	put_task_struct(task);
1261	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1262}
1263
1264static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1265				   size_t count, loff_t *ppos)
1266{
1267	struct inode * inode = file_inode(file);
1268	uid_t loginuid;
1269	kuid_t kloginuid;
1270	int rv;
1271
 
 
 
 
1272	rcu_read_lock();
1273	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1274		rcu_read_unlock();
1275		return -EPERM;
1276	}
1277	rcu_read_unlock();
1278
1279	if (*ppos != 0) {
1280		/* No partial writes. */
1281		return -EINVAL;
1282	}
1283
1284	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1285	if (rv < 0)
1286		return rv;
1287
1288	/* is userspace tring to explicitly UNSET the loginuid? */
1289	if (loginuid == AUDIT_UID_UNSET) {
1290		kloginuid = INVALID_UID;
1291	} else {
1292		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1293		if (!uid_valid(kloginuid))
1294			return -EINVAL;
1295	}
1296
1297	rv = audit_set_loginuid(kloginuid);
1298	if (rv < 0)
1299		return rv;
1300	return count;
1301}
1302
1303static const struct file_operations proc_loginuid_operations = {
1304	.read		= proc_loginuid_read,
1305	.write		= proc_loginuid_write,
1306	.llseek		= generic_file_llseek,
1307};
1308
1309static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1310				  size_t count, loff_t *ppos)
1311{
1312	struct inode * inode = file_inode(file);
1313	struct task_struct *task = get_proc_task(inode);
1314	ssize_t length;
1315	char tmpbuf[TMPBUFLEN];
1316
1317	if (!task)
1318		return -ESRCH;
1319	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1320				audit_get_sessionid(task));
1321	put_task_struct(task);
1322	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1323}
1324
1325static const struct file_operations proc_sessionid_operations = {
1326	.read		= proc_sessionid_read,
1327	.llseek		= generic_file_llseek,
1328};
1329#endif
1330
1331#ifdef CONFIG_FAULT_INJECTION
1332static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1333				      size_t count, loff_t *ppos)
1334{
1335	struct task_struct *task = get_proc_task(file_inode(file));
1336	char buffer[PROC_NUMBUF];
1337	size_t len;
1338	int make_it_fail;
1339
1340	if (!task)
1341		return -ESRCH;
1342	make_it_fail = task->make_it_fail;
1343	put_task_struct(task);
1344
1345	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1346
1347	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1348}
1349
1350static ssize_t proc_fault_inject_write(struct file * file,
1351			const char __user * buf, size_t count, loff_t *ppos)
1352{
1353	struct task_struct *task;
1354	char buffer[PROC_NUMBUF];
1355	int make_it_fail;
1356	int rv;
1357
1358	if (!capable(CAP_SYS_RESOURCE))
1359		return -EPERM;
1360	memset(buffer, 0, sizeof(buffer));
1361	if (count > sizeof(buffer) - 1)
1362		count = sizeof(buffer) - 1;
1363	if (copy_from_user(buffer, buf, count))
1364		return -EFAULT;
1365	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1366	if (rv < 0)
1367		return rv;
1368	if (make_it_fail < 0 || make_it_fail > 1)
1369		return -EINVAL;
1370
1371	task = get_proc_task(file_inode(file));
1372	if (!task)
1373		return -ESRCH;
1374	task->make_it_fail = make_it_fail;
1375	put_task_struct(task);
1376
1377	return count;
1378}
1379
1380static const struct file_operations proc_fault_inject_operations = {
1381	.read		= proc_fault_inject_read,
1382	.write		= proc_fault_inject_write,
1383	.llseek		= generic_file_llseek,
1384};
1385
1386static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1387				   size_t count, loff_t *ppos)
1388{
1389	struct task_struct *task;
1390	int err;
1391	unsigned int n;
1392
1393	err = kstrtouint_from_user(buf, count, 0, &n);
1394	if (err)
1395		return err;
1396
1397	task = get_proc_task(file_inode(file));
1398	if (!task)
1399		return -ESRCH;
1400	task->fail_nth = n;
1401	put_task_struct(task);
1402
1403	return count;
1404}
1405
1406static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1407				  size_t count, loff_t *ppos)
1408{
1409	struct task_struct *task;
1410	char numbuf[PROC_NUMBUF];
1411	ssize_t len;
1412
1413	task = get_proc_task(file_inode(file));
1414	if (!task)
1415		return -ESRCH;
1416	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1417	put_task_struct(task);
1418	return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1419}
1420
1421static const struct file_operations proc_fail_nth_operations = {
1422	.read		= proc_fail_nth_read,
1423	.write		= proc_fail_nth_write,
1424};
1425#endif
1426
1427
1428#ifdef CONFIG_SCHED_DEBUG
1429/*
1430 * Print out various scheduling related per-task fields:
1431 */
1432static int sched_show(struct seq_file *m, void *v)
1433{
1434	struct inode *inode = m->private;
1435	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1436	struct task_struct *p;
1437
1438	p = get_proc_task(inode);
1439	if (!p)
1440		return -ESRCH;
1441	proc_sched_show_task(p, ns, m);
1442
1443	put_task_struct(p);
1444
1445	return 0;
1446}
1447
1448static ssize_t
1449sched_write(struct file *file, const char __user *buf,
1450	    size_t count, loff_t *offset)
1451{
1452	struct inode *inode = file_inode(file);
1453	struct task_struct *p;
1454
1455	p = get_proc_task(inode);
1456	if (!p)
1457		return -ESRCH;
1458	proc_sched_set_task(p);
1459
1460	put_task_struct(p);
1461
1462	return count;
1463}
1464
1465static int sched_open(struct inode *inode, struct file *filp)
1466{
1467	return single_open(filp, sched_show, inode);
1468}
1469
1470static const struct file_operations proc_pid_sched_operations = {
1471	.open		= sched_open,
1472	.read		= seq_read,
1473	.write		= sched_write,
1474	.llseek		= seq_lseek,
1475	.release	= single_release,
1476};
1477
1478#endif
1479
1480#ifdef CONFIG_SCHED_AUTOGROUP
1481/*
1482 * Print out autogroup related information:
1483 */
1484static int sched_autogroup_show(struct seq_file *m, void *v)
1485{
1486	struct inode *inode = m->private;
1487	struct task_struct *p;
1488
1489	p = get_proc_task(inode);
1490	if (!p)
1491		return -ESRCH;
1492	proc_sched_autogroup_show_task(p, m);
1493
1494	put_task_struct(p);
1495
1496	return 0;
1497}
1498
1499static ssize_t
1500sched_autogroup_write(struct file *file, const char __user *buf,
1501	    size_t count, loff_t *offset)
1502{
1503	struct inode *inode = file_inode(file);
1504	struct task_struct *p;
1505	char buffer[PROC_NUMBUF];
1506	int nice;
1507	int err;
1508
1509	memset(buffer, 0, sizeof(buffer));
1510	if (count > sizeof(buffer) - 1)
1511		count = sizeof(buffer) - 1;
1512	if (copy_from_user(buffer, buf, count))
1513		return -EFAULT;
1514
1515	err = kstrtoint(strstrip(buffer), 0, &nice);
1516	if (err < 0)
1517		return err;
1518
1519	p = get_proc_task(inode);
1520	if (!p)
1521		return -ESRCH;
1522
1523	err = proc_sched_autogroup_set_nice(p, nice);
1524	if (err)
1525		count = err;
1526
1527	put_task_struct(p);
1528
1529	return count;
1530}
1531
1532static int sched_autogroup_open(struct inode *inode, struct file *filp)
1533{
1534	int ret;
1535
1536	ret = single_open(filp, sched_autogroup_show, NULL);
1537	if (!ret) {
1538		struct seq_file *m = filp->private_data;
1539
1540		m->private = inode;
1541	}
1542	return ret;
1543}
1544
1545static const struct file_operations proc_pid_sched_autogroup_operations = {
1546	.open		= sched_autogroup_open,
1547	.read		= seq_read,
1548	.write		= sched_autogroup_write,
1549	.llseek		= seq_lseek,
1550	.release	= single_release,
1551};
1552
1553#endif /* CONFIG_SCHED_AUTOGROUP */
1554
1555#ifdef CONFIG_TIME_NS
1556static int timens_offsets_show(struct seq_file *m, void *v)
1557{
1558	struct task_struct *p;
1559
1560	p = get_proc_task(file_inode(m->file));
1561	if (!p)
1562		return -ESRCH;
1563	proc_timens_show_offsets(p, m);
1564
1565	put_task_struct(p);
1566
1567	return 0;
1568}
1569
1570static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1571				    size_t count, loff_t *ppos)
1572{
1573	struct inode *inode = file_inode(file);
1574	struct proc_timens_offset offsets[2];
1575	char *kbuf = NULL, *pos, *next_line;
1576	struct task_struct *p;
1577	int ret, noffsets;
1578
1579	/* Only allow < page size writes at the beginning of the file */
1580	if ((*ppos != 0) || (count >= PAGE_SIZE))
1581		return -EINVAL;
1582
1583	/* Slurp in the user data */
1584	kbuf = memdup_user_nul(buf, count);
1585	if (IS_ERR(kbuf))
1586		return PTR_ERR(kbuf);
1587
1588	/* Parse the user data */
1589	ret = -EINVAL;
1590	noffsets = 0;
1591	for (pos = kbuf; pos; pos = next_line) {
1592		struct proc_timens_offset *off = &offsets[noffsets];
1593		char clock[10];
1594		int err;
1595
1596		/* Find the end of line and ensure we don't look past it */
1597		next_line = strchr(pos, '\n');
1598		if (next_line) {
1599			*next_line = '\0';
1600			next_line++;
1601			if (*next_line == '\0')
1602				next_line = NULL;
1603		}
1604
1605		err = sscanf(pos, "%9s %lld %lu", clock,
1606				&off->val.tv_sec, &off->val.tv_nsec);
1607		if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1608			goto out;
1609
1610		clock[sizeof(clock) - 1] = 0;
1611		if (strcmp(clock, "monotonic") == 0 ||
1612		    strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1613			off->clockid = CLOCK_MONOTONIC;
1614		else if (strcmp(clock, "boottime") == 0 ||
1615			 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1616			off->clockid = CLOCK_BOOTTIME;
1617		else
1618			goto out;
1619
1620		noffsets++;
1621		if (noffsets == ARRAY_SIZE(offsets)) {
1622			if (next_line)
1623				count = next_line - kbuf;
1624			break;
1625		}
1626	}
1627
1628	ret = -ESRCH;
1629	p = get_proc_task(inode);
1630	if (!p)
1631		goto out;
1632	ret = proc_timens_set_offset(file, p, offsets, noffsets);
1633	put_task_struct(p);
1634	if (ret)
1635		goto out;
1636
1637	ret = count;
1638out:
1639	kfree(kbuf);
1640	return ret;
1641}
1642
1643static int timens_offsets_open(struct inode *inode, struct file *filp)
1644{
1645	return single_open(filp, timens_offsets_show, inode);
1646}
1647
1648static const struct file_operations proc_timens_offsets_operations = {
1649	.open		= timens_offsets_open,
1650	.read		= seq_read,
1651	.write		= timens_offsets_write,
1652	.llseek		= seq_lseek,
1653	.release	= single_release,
1654};
1655#endif /* CONFIG_TIME_NS */
1656
1657static ssize_t comm_write(struct file *file, const char __user *buf,
1658				size_t count, loff_t *offset)
1659{
1660	struct inode *inode = file_inode(file);
1661	struct task_struct *p;
1662	char buffer[TASK_COMM_LEN];
1663	const size_t maxlen = sizeof(buffer) - 1;
1664
1665	memset(buffer, 0, sizeof(buffer));
1666	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1667		return -EFAULT;
1668
1669	p = get_proc_task(inode);
1670	if (!p)
1671		return -ESRCH;
1672
1673	if (same_thread_group(current, p))
1674		set_task_comm(p, buffer);
1675	else
1676		count = -EINVAL;
1677
1678	put_task_struct(p);
1679
1680	return count;
1681}
1682
1683static int comm_show(struct seq_file *m, void *v)
1684{
1685	struct inode *inode = m->private;
1686	struct task_struct *p;
1687
1688	p = get_proc_task(inode);
1689	if (!p)
1690		return -ESRCH;
1691
1692	proc_task_name(m, p, false);
1693	seq_putc(m, '\n');
1694
1695	put_task_struct(p);
1696
1697	return 0;
1698}
1699
1700static int comm_open(struct inode *inode, struct file *filp)
1701{
1702	return single_open(filp, comm_show, inode);
1703}
1704
1705static const struct file_operations proc_pid_set_comm_operations = {
1706	.open		= comm_open,
1707	.read		= seq_read,
1708	.write		= comm_write,
1709	.llseek		= seq_lseek,
1710	.release	= single_release,
1711};
1712
1713static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1714{
1715	struct task_struct *task;
1716	struct file *exe_file;
1717
1718	task = get_proc_task(d_inode(dentry));
1719	if (!task)
1720		return -ENOENT;
1721	exe_file = get_task_exe_file(task);
1722	put_task_struct(task);
1723	if (exe_file) {
1724		*exe_path = exe_file->f_path;
1725		path_get(&exe_file->f_path);
1726		fput(exe_file);
1727		return 0;
1728	} else
1729		return -ENOENT;
1730}
1731
1732static const char *proc_pid_get_link(struct dentry *dentry,
1733				     struct inode *inode,
1734				     struct delayed_call *done)
1735{
1736	struct path path;
1737	int error = -EACCES;
1738
1739	if (!dentry)
1740		return ERR_PTR(-ECHILD);
1741
1742	/* Are we allowed to snoop on the tasks file descriptors? */
1743	if (!proc_fd_access_allowed(inode))
1744		goto out;
1745
1746	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1747	if (error)
1748		goto out;
1749
1750	error = nd_jump_link(&path);
1751out:
1752	return ERR_PTR(error);
1753}
1754
1755static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1756{
1757	char *tmp = (char *)__get_free_page(GFP_KERNEL);
1758	char *pathname;
1759	int len;
1760
1761	if (!tmp)
1762		return -ENOMEM;
1763
1764	pathname = d_path(path, tmp, PAGE_SIZE);
1765	len = PTR_ERR(pathname);
1766	if (IS_ERR(pathname))
1767		goto out;
1768	len = tmp + PAGE_SIZE - 1 - pathname;
1769
1770	if (len > buflen)
1771		len = buflen;
1772	if (copy_to_user(buffer, pathname, len))
1773		len = -EFAULT;
1774 out:
1775	free_page((unsigned long)tmp);
1776	return len;
1777}
1778
1779static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1780{
1781	int error = -EACCES;
1782	struct inode *inode = d_inode(dentry);
1783	struct path path;
1784
1785	/* Are we allowed to snoop on the tasks file descriptors? */
1786	if (!proc_fd_access_allowed(inode))
1787		goto out;
1788
1789	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1790	if (error)
1791		goto out;
1792
1793	error = do_proc_readlink(&path, buffer, buflen);
1794	path_put(&path);
1795out:
1796	return error;
1797}
1798
1799const struct inode_operations proc_pid_link_inode_operations = {
1800	.readlink	= proc_pid_readlink,
1801	.get_link	= proc_pid_get_link,
1802	.setattr	= proc_setattr,
1803};
1804
1805
1806/* building an inode */
1807
1808void task_dump_owner(struct task_struct *task, umode_t mode,
1809		     kuid_t *ruid, kgid_t *rgid)
1810{
1811	/* Depending on the state of dumpable compute who should own a
1812	 * proc file for a task.
1813	 */
1814	const struct cred *cred;
1815	kuid_t uid;
1816	kgid_t gid;
1817
1818	if (unlikely(task->flags & PF_KTHREAD)) {
1819		*ruid = GLOBAL_ROOT_UID;
1820		*rgid = GLOBAL_ROOT_GID;
1821		return;
1822	}
1823
1824	/* Default to the tasks effective ownership */
1825	rcu_read_lock();
1826	cred = __task_cred(task);
1827	uid = cred->euid;
1828	gid = cred->egid;
1829	rcu_read_unlock();
1830
1831	/*
1832	 * Before the /proc/pid/status file was created the only way to read
1833	 * the effective uid of a /process was to stat /proc/pid.  Reading
1834	 * /proc/pid/status is slow enough that procps and other packages
1835	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1836	 * made this apply to all per process world readable and executable
1837	 * directories.
1838	 */
1839	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1840		struct mm_struct *mm;
1841		task_lock(task);
1842		mm = task->mm;
1843		/* Make non-dumpable tasks owned by some root */
1844		if (mm) {
1845			if (get_dumpable(mm) != SUID_DUMP_USER) {
1846				struct user_namespace *user_ns = mm->user_ns;
1847
1848				uid = make_kuid(user_ns, 0);
1849				if (!uid_valid(uid))
1850					uid = GLOBAL_ROOT_UID;
1851
1852				gid = make_kgid(user_ns, 0);
1853				if (!gid_valid(gid))
1854					gid = GLOBAL_ROOT_GID;
1855			}
1856		} else {
1857			uid = GLOBAL_ROOT_UID;
1858			gid = GLOBAL_ROOT_GID;
1859		}
1860		task_unlock(task);
1861	}
1862	*ruid = uid;
1863	*rgid = gid;
1864}
1865
1866void proc_pid_evict_inode(struct proc_inode *ei)
1867{
1868	struct pid *pid = ei->pid;
1869
1870	if (S_ISDIR(ei->vfs_inode.i_mode)) {
1871		spin_lock(&pid->lock);
1872		hlist_del_init_rcu(&ei->sibling_inodes);
1873		spin_unlock(&pid->lock);
1874	}
1875
1876	put_pid(pid);
1877}
1878
1879struct inode *proc_pid_make_inode(struct super_block * sb,
1880				  struct task_struct *task, umode_t mode)
1881{
1882	struct inode * inode;
1883	struct proc_inode *ei;
1884	struct pid *pid;
1885
1886	/* We need a new inode */
1887
1888	inode = new_inode(sb);
1889	if (!inode)
1890		goto out;
1891
1892	/* Common stuff */
1893	ei = PROC_I(inode);
1894	inode->i_mode = mode;
1895	inode->i_ino = get_next_ino();
1896	inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1897	inode->i_op = &proc_def_inode_operations;
1898
1899	/*
1900	 * grab the reference to task.
1901	 */
1902	pid = get_task_pid(task, PIDTYPE_PID);
1903	if (!pid)
1904		goto out_unlock;
1905
1906	/* Let the pid remember us for quick removal */
1907	ei->pid = pid;
1908	if (S_ISDIR(mode)) {
1909		spin_lock(&pid->lock);
1910		hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
1911		spin_unlock(&pid->lock);
1912	}
1913
1914	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1915	security_task_to_inode(task, inode);
1916
1917out:
1918	return inode;
1919
1920out_unlock:
1921	iput(inode);
1922	return NULL;
1923}
1924
1925int pid_getattr(const struct path *path, struct kstat *stat,
1926		u32 request_mask, unsigned int query_flags)
1927{
1928	struct inode *inode = d_inode(path->dentry);
1929	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
1930	struct task_struct *task;
1931
1932	generic_fillattr(inode, stat);
1933
1934	stat->uid = GLOBAL_ROOT_UID;
1935	stat->gid = GLOBAL_ROOT_GID;
1936	rcu_read_lock();
1937	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1938	if (task) {
1939		if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
1940			rcu_read_unlock();
1941			/*
1942			 * This doesn't prevent learning whether PID exists,
1943			 * it only makes getattr() consistent with readdir().
1944			 */
1945			return -ENOENT;
1946		}
1947		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1948	}
1949	rcu_read_unlock();
1950	return 0;
1951}
1952
1953/* dentry stuff */
1954
1955/*
1956 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1957 */
1958void pid_update_inode(struct task_struct *task, struct inode *inode)
1959{
1960	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1961
1962	inode->i_mode &= ~(S_ISUID | S_ISGID);
1963	security_task_to_inode(task, inode);
1964}
1965
1966/*
1967 * Rewrite the inode's ownerships here because the owning task may have
1968 * performed a setuid(), etc.
1969 *
1970 */
1971static int pid_revalidate(struct dentry *dentry, unsigned int flags)
1972{
1973	struct inode *inode;
1974	struct task_struct *task;
1975
1976	if (flags & LOOKUP_RCU)
1977		return -ECHILD;
1978
1979	inode = d_inode(dentry);
1980	task = get_proc_task(inode);
1981
1982	if (task) {
1983		pid_update_inode(task, inode);
1984		put_task_struct(task);
1985		return 1;
1986	}
1987	return 0;
1988}
1989
1990static inline bool proc_inode_is_dead(struct inode *inode)
1991{
1992	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1993}
1994
1995int pid_delete_dentry(const struct dentry *dentry)
1996{
1997	/* Is the task we represent dead?
1998	 * If so, then don't put the dentry on the lru list,
1999	 * kill it immediately.
2000	 */
2001	return proc_inode_is_dead(d_inode(dentry));
2002}
2003
2004const struct dentry_operations pid_dentry_operations =
2005{
2006	.d_revalidate	= pid_revalidate,
2007	.d_delete	= pid_delete_dentry,
2008};
2009
2010/* Lookups */
2011
2012/*
2013 * Fill a directory entry.
2014 *
2015 * If possible create the dcache entry and derive our inode number and
2016 * file type from dcache entry.
2017 *
2018 * Since all of the proc inode numbers are dynamically generated, the inode
2019 * numbers do not exist until the inode is cache.  This means creating the
2020 * the dcache entry in readdir is necessary to keep the inode numbers
2021 * reported by readdir in sync with the inode numbers reported
2022 * by stat.
2023 */
2024bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2025	const char *name, unsigned int len,
2026	instantiate_t instantiate, struct task_struct *task, const void *ptr)
2027{
2028	struct dentry *child, *dir = file->f_path.dentry;
2029	struct qstr qname = QSTR_INIT(name, len);
2030	struct inode *inode;
2031	unsigned type = DT_UNKNOWN;
2032	ino_t ino = 1;
2033
2034	child = d_hash_and_lookup(dir, &qname);
2035	if (!child) {
2036		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2037		child = d_alloc_parallel(dir, &qname, &wq);
2038		if (IS_ERR(child))
2039			goto end_instantiate;
2040		if (d_in_lookup(child)) {
2041			struct dentry *res;
2042			res = instantiate(child, task, ptr);
2043			d_lookup_done(child);
2044			if (unlikely(res)) {
2045				dput(child);
2046				child = res;
2047				if (IS_ERR(child))
2048					goto end_instantiate;
2049			}
2050		}
2051	}
2052	inode = d_inode(child);
2053	ino = inode->i_ino;
2054	type = inode->i_mode >> 12;
2055	dput(child);
2056end_instantiate:
2057	return dir_emit(ctx, name, len, ino, type);
2058}
2059
2060/*
2061 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2062 * which represent vma start and end addresses.
2063 */
2064static int dname_to_vma_addr(struct dentry *dentry,
2065			     unsigned long *start, unsigned long *end)
2066{
2067	const char *str = dentry->d_name.name;
2068	unsigned long long sval, eval;
2069	unsigned int len;
2070
2071	if (str[0] == '0' && str[1] != '-')
2072		return -EINVAL;
2073	len = _parse_integer(str, 16, &sval);
2074	if (len & KSTRTOX_OVERFLOW)
2075		return -EINVAL;
2076	if (sval != (unsigned long)sval)
2077		return -EINVAL;
2078	str += len;
2079
2080	if (*str != '-')
2081		return -EINVAL;
2082	str++;
2083
2084	if (str[0] == '0' && str[1])
2085		return -EINVAL;
2086	len = _parse_integer(str, 16, &eval);
2087	if (len & KSTRTOX_OVERFLOW)
2088		return -EINVAL;
2089	if (eval != (unsigned long)eval)
2090		return -EINVAL;
2091	str += len;
2092
2093	if (*str != '\0')
2094		return -EINVAL;
2095
2096	*start = sval;
2097	*end = eval;
2098
2099	return 0;
2100}
2101
2102static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2103{
2104	unsigned long vm_start, vm_end;
2105	bool exact_vma_exists = false;
2106	struct mm_struct *mm = NULL;
2107	struct task_struct *task;
2108	struct inode *inode;
2109	int status = 0;
2110
2111	if (flags & LOOKUP_RCU)
2112		return -ECHILD;
2113
2114	inode = d_inode(dentry);
2115	task = get_proc_task(inode);
2116	if (!task)
2117		goto out_notask;
2118
2119	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2120	if (IS_ERR_OR_NULL(mm))
2121		goto out;
2122
2123	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2124		status = mmap_read_lock_killable(mm);
2125		if (!status) {
2126			exact_vma_exists = !!find_exact_vma(mm, vm_start,
2127							    vm_end);
2128			mmap_read_unlock(mm);
2129		}
2130	}
2131
2132	mmput(mm);
2133
2134	if (exact_vma_exists) {
2135		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2136
2137		security_task_to_inode(task, inode);
2138		status = 1;
2139	}
2140
2141out:
2142	put_task_struct(task);
2143
2144out_notask:
2145	return status;
2146}
2147
2148static const struct dentry_operations tid_map_files_dentry_operations = {
2149	.d_revalidate	= map_files_d_revalidate,
2150	.d_delete	= pid_delete_dentry,
2151};
2152
2153static int map_files_get_link(struct dentry *dentry, struct path *path)
2154{
2155	unsigned long vm_start, vm_end;
2156	struct vm_area_struct *vma;
2157	struct task_struct *task;
2158	struct mm_struct *mm;
2159	int rc;
2160
2161	rc = -ENOENT;
2162	task = get_proc_task(d_inode(dentry));
2163	if (!task)
2164		goto out;
2165
2166	mm = get_task_mm(task);
2167	put_task_struct(task);
2168	if (!mm)
2169		goto out;
2170
2171	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2172	if (rc)
2173		goto out_mmput;
2174
2175	rc = mmap_read_lock_killable(mm);
2176	if (rc)
2177		goto out_mmput;
2178
2179	rc = -ENOENT;
2180	vma = find_exact_vma(mm, vm_start, vm_end);
2181	if (vma && vma->vm_file) {
2182		*path = vma->vm_file->f_path;
2183		path_get(path);
2184		rc = 0;
2185	}
2186	mmap_read_unlock(mm);
2187
2188out_mmput:
2189	mmput(mm);
2190out:
2191	return rc;
2192}
2193
2194struct map_files_info {
2195	unsigned long	start;
2196	unsigned long	end;
2197	fmode_t		mode;
2198};
2199
2200/*
2201 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2202 * to concerns about how the symlinks may be used to bypass permissions on
2203 * ancestor directories in the path to the file in question.
2204 */
2205static const char *
2206proc_map_files_get_link(struct dentry *dentry,
2207			struct inode *inode,
2208		        struct delayed_call *done)
2209{
2210	if (!checkpoint_restore_ns_capable(&init_user_ns))
2211		return ERR_PTR(-EPERM);
2212
2213	return proc_pid_get_link(dentry, inode, done);
2214}
2215
2216/*
2217 * Identical to proc_pid_link_inode_operations except for get_link()
2218 */
2219static const struct inode_operations proc_map_files_link_inode_operations = {
2220	.readlink	= proc_pid_readlink,
2221	.get_link	= proc_map_files_get_link,
2222	.setattr	= proc_setattr,
2223};
2224
2225static struct dentry *
2226proc_map_files_instantiate(struct dentry *dentry,
2227			   struct task_struct *task, const void *ptr)
2228{
2229	fmode_t mode = (fmode_t)(unsigned long)ptr;
2230	struct proc_inode *ei;
2231	struct inode *inode;
2232
2233	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2234				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2235				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2236	if (!inode)
2237		return ERR_PTR(-ENOENT);
2238
2239	ei = PROC_I(inode);
2240	ei->op.proc_get_link = map_files_get_link;
2241
2242	inode->i_op = &proc_map_files_link_inode_operations;
2243	inode->i_size = 64;
2244
2245	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2246	return d_splice_alias(inode, dentry);
2247}
2248
2249static struct dentry *proc_map_files_lookup(struct inode *dir,
2250		struct dentry *dentry, unsigned int flags)
2251{
2252	unsigned long vm_start, vm_end;
2253	struct vm_area_struct *vma;
2254	struct task_struct *task;
2255	struct dentry *result;
2256	struct mm_struct *mm;
2257
2258	result = ERR_PTR(-ENOENT);
2259	task = get_proc_task(dir);
2260	if (!task)
2261		goto out;
2262
2263	result = ERR_PTR(-EACCES);
2264	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2265		goto out_put_task;
2266
2267	result = ERR_PTR(-ENOENT);
2268	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2269		goto out_put_task;
2270
2271	mm = get_task_mm(task);
2272	if (!mm)
2273		goto out_put_task;
2274
2275	result = ERR_PTR(-EINTR);
2276	if (mmap_read_lock_killable(mm))
2277		goto out_put_mm;
2278
2279	result = ERR_PTR(-ENOENT);
2280	vma = find_exact_vma(mm, vm_start, vm_end);
2281	if (!vma)
2282		goto out_no_vma;
2283
2284	if (vma->vm_file)
2285		result = proc_map_files_instantiate(dentry, task,
2286				(void *)(unsigned long)vma->vm_file->f_mode);
2287
2288out_no_vma:
2289	mmap_read_unlock(mm);
2290out_put_mm:
2291	mmput(mm);
2292out_put_task:
2293	put_task_struct(task);
2294out:
2295	return result;
2296}
2297
2298static const struct inode_operations proc_map_files_inode_operations = {
2299	.lookup		= proc_map_files_lookup,
2300	.permission	= proc_fd_permission,
2301	.setattr	= proc_setattr,
2302};
2303
2304static int
2305proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2306{
2307	struct vm_area_struct *vma;
2308	struct task_struct *task;
2309	struct mm_struct *mm;
2310	unsigned long nr_files, pos, i;
2311	GENRADIX(struct map_files_info) fa;
2312	struct map_files_info *p;
2313	int ret;
2314
2315	genradix_init(&fa);
2316
2317	ret = -ENOENT;
2318	task = get_proc_task(file_inode(file));
2319	if (!task)
2320		goto out;
2321
2322	ret = -EACCES;
2323	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2324		goto out_put_task;
2325
2326	ret = 0;
2327	if (!dir_emit_dots(file, ctx))
2328		goto out_put_task;
2329
2330	mm = get_task_mm(task);
2331	if (!mm)
2332		goto out_put_task;
2333
2334	ret = mmap_read_lock_killable(mm);
2335	if (ret) {
2336		mmput(mm);
2337		goto out_put_task;
2338	}
2339
2340	nr_files = 0;
2341
2342	/*
2343	 * We need two passes here:
2344	 *
2345	 *  1) Collect vmas of mapped files with mmap_lock taken
2346	 *  2) Release mmap_lock and instantiate entries
2347	 *
2348	 * otherwise we get lockdep complained, since filldir()
2349	 * routine might require mmap_lock taken in might_fault().
2350	 */
2351
2352	for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2353		if (!vma->vm_file)
2354			continue;
2355		if (++pos <= ctx->pos)
2356			continue;
2357
2358		p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2359		if (!p) {
2360			ret = -ENOMEM;
2361			mmap_read_unlock(mm);
2362			mmput(mm);
2363			goto out_put_task;
2364		}
2365
2366		p->start = vma->vm_start;
2367		p->end = vma->vm_end;
2368		p->mode = vma->vm_file->f_mode;
2369	}
2370	mmap_read_unlock(mm);
2371	mmput(mm);
2372
2373	for (i = 0; i < nr_files; i++) {
2374		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2375		unsigned int len;
2376
2377		p = genradix_ptr(&fa, i);
2378		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2379		if (!proc_fill_cache(file, ctx,
2380				      buf, len,
2381				      proc_map_files_instantiate,
2382				      task,
2383				      (void *)(unsigned long)p->mode))
2384			break;
2385		ctx->pos++;
2386	}
2387
2388out_put_task:
2389	put_task_struct(task);
2390out:
2391	genradix_free(&fa);
2392	return ret;
2393}
2394
2395static const struct file_operations proc_map_files_operations = {
2396	.read		= generic_read_dir,
2397	.iterate_shared	= proc_map_files_readdir,
2398	.llseek		= generic_file_llseek,
2399};
2400
2401#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2402struct timers_private {
2403	struct pid *pid;
2404	struct task_struct *task;
2405	struct sighand_struct *sighand;
2406	struct pid_namespace *ns;
2407	unsigned long flags;
2408};
2409
2410static void *timers_start(struct seq_file *m, loff_t *pos)
2411{
2412	struct timers_private *tp = m->private;
2413
2414	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2415	if (!tp->task)
2416		return ERR_PTR(-ESRCH);
2417
2418	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2419	if (!tp->sighand)
2420		return ERR_PTR(-ESRCH);
2421
2422	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2423}
2424
2425static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2426{
2427	struct timers_private *tp = m->private;
2428	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2429}
2430
2431static void timers_stop(struct seq_file *m, void *v)
2432{
2433	struct timers_private *tp = m->private;
2434
2435	if (tp->sighand) {
2436		unlock_task_sighand(tp->task, &tp->flags);
2437		tp->sighand = NULL;
2438	}
2439
2440	if (tp->task) {
2441		put_task_struct(tp->task);
2442		tp->task = NULL;
2443	}
2444}
2445
2446static int show_timer(struct seq_file *m, void *v)
2447{
2448	struct k_itimer *timer;
2449	struct timers_private *tp = m->private;
2450	int notify;
2451	static const char * const nstr[] = {
2452		[SIGEV_SIGNAL] = "signal",
2453		[SIGEV_NONE] = "none",
2454		[SIGEV_THREAD] = "thread",
2455	};
2456
2457	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2458	notify = timer->it_sigev_notify;
2459
2460	seq_printf(m, "ID: %d\n", timer->it_id);
2461	seq_printf(m, "signal: %d/%px\n",
2462		   timer->sigq->info.si_signo,
2463		   timer->sigq->info.si_value.sival_ptr);
2464	seq_printf(m, "notify: %s/%s.%d\n",
2465		   nstr[notify & ~SIGEV_THREAD_ID],
2466		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2467		   pid_nr_ns(timer->it_pid, tp->ns));
2468	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2469
2470	return 0;
2471}
2472
2473static const struct seq_operations proc_timers_seq_ops = {
2474	.start	= timers_start,
2475	.next	= timers_next,
2476	.stop	= timers_stop,
2477	.show	= show_timer,
2478};
2479
2480static int proc_timers_open(struct inode *inode, struct file *file)
2481{
2482	struct timers_private *tp;
2483
2484	tp = __seq_open_private(file, &proc_timers_seq_ops,
2485			sizeof(struct timers_private));
2486	if (!tp)
2487		return -ENOMEM;
2488
2489	tp->pid = proc_pid(inode);
2490	tp->ns = proc_pid_ns(inode->i_sb);
2491	return 0;
2492}
2493
2494static const struct file_operations proc_timers_operations = {
2495	.open		= proc_timers_open,
2496	.read		= seq_read,
2497	.llseek		= seq_lseek,
2498	.release	= seq_release_private,
2499};
2500#endif
2501
2502static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2503					size_t count, loff_t *offset)
2504{
2505	struct inode *inode = file_inode(file);
2506	struct task_struct *p;
2507	u64 slack_ns;
2508	int err;
2509
2510	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2511	if (err < 0)
2512		return err;
2513
2514	p = get_proc_task(inode);
2515	if (!p)
2516		return -ESRCH;
2517
2518	if (p != current) {
2519		rcu_read_lock();
2520		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2521			rcu_read_unlock();
2522			count = -EPERM;
2523			goto out;
2524		}
2525		rcu_read_unlock();
2526
2527		err = security_task_setscheduler(p);
2528		if (err) {
2529			count = err;
2530			goto out;
2531		}
2532	}
2533
2534	task_lock(p);
2535	if (slack_ns == 0)
2536		p->timer_slack_ns = p->default_timer_slack_ns;
2537	else
2538		p->timer_slack_ns = slack_ns;
2539	task_unlock(p);
2540
2541out:
2542	put_task_struct(p);
2543
2544	return count;
2545}
2546
2547static int timerslack_ns_show(struct seq_file *m, void *v)
2548{
2549	struct inode *inode = m->private;
2550	struct task_struct *p;
2551	int err = 0;
2552
2553	p = get_proc_task(inode);
2554	if (!p)
2555		return -ESRCH;
2556
2557	if (p != current) {
2558		rcu_read_lock();
2559		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2560			rcu_read_unlock();
2561			err = -EPERM;
2562			goto out;
2563		}
2564		rcu_read_unlock();
2565
2566		err = security_task_getscheduler(p);
2567		if (err)
2568			goto out;
2569	}
2570
2571	task_lock(p);
2572	seq_printf(m, "%llu\n", p->timer_slack_ns);
2573	task_unlock(p);
2574
2575out:
2576	put_task_struct(p);
2577
2578	return err;
2579}
2580
2581static int timerslack_ns_open(struct inode *inode, struct file *filp)
2582{
2583	return single_open(filp, timerslack_ns_show, inode);
2584}
2585
2586static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2587	.open		= timerslack_ns_open,
2588	.read		= seq_read,
2589	.write		= timerslack_ns_write,
2590	.llseek		= seq_lseek,
2591	.release	= single_release,
2592};
2593
2594static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2595	struct task_struct *task, const void *ptr)
2596{
2597	const struct pid_entry *p = ptr;
2598	struct inode *inode;
2599	struct proc_inode *ei;
2600
2601	inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2602	if (!inode)
2603		return ERR_PTR(-ENOENT);
2604
2605	ei = PROC_I(inode);
2606	if (S_ISDIR(inode->i_mode))
2607		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2608	if (p->iop)
2609		inode->i_op = p->iop;
2610	if (p->fop)
2611		inode->i_fop = p->fop;
2612	ei->op = p->op;
2613	pid_update_inode(task, inode);
2614	d_set_d_op(dentry, &pid_dentry_operations);
2615	return d_splice_alias(inode, dentry);
2616}
2617
2618static struct dentry *proc_pident_lookup(struct inode *dir, 
2619					 struct dentry *dentry,
2620					 const struct pid_entry *p,
2621					 const struct pid_entry *end)
2622{
2623	struct task_struct *task = get_proc_task(dir);
2624	struct dentry *res = ERR_PTR(-ENOENT);
2625
2626	if (!task)
2627		goto out_no_task;
2628
2629	/*
2630	 * Yes, it does not scale. And it should not. Don't add
2631	 * new entries into /proc/<tgid>/ without very good reasons.
2632	 */
2633	for (; p < end; p++) {
2634		if (p->len != dentry->d_name.len)
2635			continue;
2636		if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2637			res = proc_pident_instantiate(dentry, task, p);
2638			break;
2639		}
2640	}
2641	put_task_struct(task);
2642out_no_task:
2643	return res;
2644}
2645
2646static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2647		const struct pid_entry *ents, unsigned int nents)
2648{
2649	struct task_struct *task = get_proc_task(file_inode(file));
2650	const struct pid_entry *p;
2651
2652	if (!task)
2653		return -ENOENT;
2654
2655	if (!dir_emit_dots(file, ctx))
2656		goto out;
2657
2658	if (ctx->pos >= nents + 2)
2659		goto out;
2660
2661	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2662		if (!proc_fill_cache(file, ctx, p->name, p->len,
2663				proc_pident_instantiate, task, p))
2664			break;
2665		ctx->pos++;
2666	}
2667out:
2668	put_task_struct(task);
2669	return 0;
2670}
2671
2672#ifdef CONFIG_SECURITY
 
 
 
 
 
 
 
2673static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2674				  size_t count, loff_t *ppos)
2675{
2676	struct inode * inode = file_inode(file);
2677	char *p = NULL;
2678	ssize_t length;
2679	struct task_struct *task = get_proc_task(inode);
2680
2681	if (!task)
2682		return -ESRCH;
2683
2684	length = security_getprocattr(task, PROC_I(inode)->op.lsm,
2685				      (char*)file->f_path.dentry->d_name.name,
2686				      &p);
2687	put_task_struct(task);
2688	if (length > 0)
2689		length = simple_read_from_buffer(buf, count, ppos, p, length);
2690	kfree(p);
2691	return length;
2692}
2693
2694static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2695				   size_t count, loff_t *ppos)
2696{
2697	struct inode * inode = file_inode(file);
2698	struct task_struct *task;
2699	void *page;
2700	int rv;
2701
 
 
 
 
2702	rcu_read_lock();
2703	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2704	if (!task) {
2705		rcu_read_unlock();
2706		return -ESRCH;
2707	}
2708	/* A task may only write its own attributes. */
2709	if (current != task) {
2710		rcu_read_unlock();
2711		return -EACCES;
2712	}
2713	/* Prevent changes to overridden credentials. */
2714	if (current_cred() != current_real_cred()) {
2715		rcu_read_unlock();
2716		return -EBUSY;
2717	}
2718	rcu_read_unlock();
2719
2720	if (count > PAGE_SIZE)
2721		count = PAGE_SIZE;
2722
2723	/* No partial writes. */
2724	if (*ppos != 0)
2725		return -EINVAL;
2726
2727	page = memdup_user(buf, count);
2728	if (IS_ERR(page)) {
2729		rv = PTR_ERR(page);
2730		goto out;
2731	}
2732
2733	/* Guard against adverse ptrace interaction */
2734	rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2735	if (rv < 0)
2736		goto out_free;
2737
2738	rv = security_setprocattr(PROC_I(inode)->op.lsm,
2739				  file->f_path.dentry->d_name.name, page,
2740				  count);
2741	mutex_unlock(&current->signal->cred_guard_mutex);
2742out_free:
2743	kfree(page);
2744out:
2745	return rv;
2746}
2747
2748static const struct file_operations proc_pid_attr_operations = {
 
2749	.read		= proc_pid_attr_read,
2750	.write		= proc_pid_attr_write,
2751	.llseek		= generic_file_llseek,
 
2752};
2753
2754#define LSM_DIR_OPS(LSM) \
2755static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2756			     struct dir_context *ctx) \
2757{ \
2758	return proc_pident_readdir(filp, ctx, \
2759				   LSM##_attr_dir_stuff, \
2760				   ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2761} \
2762\
2763static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2764	.read		= generic_read_dir, \
2765	.iterate	= proc_##LSM##_attr_dir_iterate, \
2766	.llseek		= default_llseek, \
2767}; \
2768\
2769static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2770				struct dentry *dentry, unsigned int flags) \
2771{ \
2772	return proc_pident_lookup(dir, dentry, \
2773				  LSM##_attr_dir_stuff, \
2774				  LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2775} \
2776\
2777static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2778	.lookup		= proc_##LSM##_attr_dir_lookup, \
2779	.getattr	= pid_getattr, \
2780	.setattr	= proc_setattr, \
2781}
2782
2783#ifdef CONFIG_SECURITY_SMACK
2784static const struct pid_entry smack_attr_dir_stuff[] = {
2785	ATTR("smack", "current",	0666),
2786};
2787LSM_DIR_OPS(smack);
2788#endif
2789
2790#ifdef CONFIG_SECURITY_APPARMOR
2791static const struct pid_entry apparmor_attr_dir_stuff[] = {
2792	ATTR("apparmor", "current",	0666),
2793	ATTR("apparmor", "prev",	0444),
2794	ATTR("apparmor", "exec",	0666),
2795};
2796LSM_DIR_OPS(apparmor);
2797#endif
2798
2799static const struct pid_entry attr_dir_stuff[] = {
2800	ATTR(NULL, "current",		0666),
2801	ATTR(NULL, "prev",		0444),
2802	ATTR(NULL, "exec",		0666),
2803	ATTR(NULL, "fscreate",		0666),
2804	ATTR(NULL, "keycreate",		0666),
2805	ATTR(NULL, "sockcreate",	0666),
2806#ifdef CONFIG_SECURITY_SMACK
2807	DIR("smack",			0555,
2808	    proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2809#endif
2810#ifdef CONFIG_SECURITY_APPARMOR
2811	DIR("apparmor",			0555,
2812	    proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2813#endif
2814};
2815
2816static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2817{
2818	return proc_pident_readdir(file, ctx, 
2819				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2820}
2821
2822static const struct file_operations proc_attr_dir_operations = {
2823	.read		= generic_read_dir,
2824	.iterate_shared	= proc_attr_dir_readdir,
2825	.llseek		= generic_file_llseek,
2826};
2827
2828static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2829				struct dentry *dentry, unsigned int flags)
2830{
2831	return proc_pident_lookup(dir, dentry,
2832				  attr_dir_stuff,
2833				  attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2834}
2835
2836static const struct inode_operations proc_attr_dir_inode_operations = {
2837	.lookup		= proc_attr_dir_lookup,
2838	.getattr	= pid_getattr,
2839	.setattr	= proc_setattr,
2840};
2841
2842#endif
2843
2844#ifdef CONFIG_ELF_CORE
2845static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2846					 size_t count, loff_t *ppos)
2847{
2848	struct task_struct *task = get_proc_task(file_inode(file));
2849	struct mm_struct *mm;
2850	char buffer[PROC_NUMBUF];
2851	size_t len;
2852	int ret;
2853
2854	if (!task)
2855		return -ESRCH;
2856
2857	ret = 0;
2858	mm = get_task_mm(task);
2859	if (mm) {
2860		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2861			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2862				MMF_DUMP_FILTER_SHIFT));
2863		mmput(mm);
2864		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2865	}
2866
2867	put_task_struct(task);
2868
2869	return ret;
2870}
2871
2872static ssize_t proc_coredump_filter_write(struct file *file,
2873					  const char __user *buf,
2874					  size_t count,
2875					  loff_t *ppos)
2876{
2877	struct task_struct *task;
2878	struct mm_struct *mm;
2879	unsigned int val;
2880	int ret;
2881	int i;
2882	unsigned long mask;
2883
2884	ret = kstrtouint_from_user(buf, count, 0, &val);
2885	if (ret < 0)
2886		return ret;
2887
2888	ret = -ESRCH;
2889	task = get_proc_task(file_inode(file));
2890	if (!task)
2891		goto out_no_task;
2892
2893	mm = get_task_mm(task);
2894	if (!mm)
2895		goto out_no_mm;
2896	ret = 0;
2897
2898	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2899		if (val & mask)
2900			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2901		else
2902			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2903	}
2904
2905	mmput(mm);
2906 out_no_mm:
2907	put_task_struct(task);
2908 out_no_task:
2909	if (ret < 0)
2910		return ret;
2911	return count;
2912}
2913
2914static const struct file_operations proc_coredump_filter_operations = {
2915	.read		= proc_coredump_filter_read,
2916	.write		= proc_coredump_filter_write,
2917	.llseek		= generic_file_llseek,
2918};
2919#endif
2920
2921#ifdef CONFIG_TASK_IO_ACCOUNTING
2922static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2923{
2924	struct task_io_accounting acct = task->ioac;
2925	unsigned long flags;
2926	int result;
2927
2928	result = mutex_lock_killable(&task->signal->exec_update_mutex);
2929	if (result)
2930		return result;
2931
2932	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2933		result = -EACCES;
2934		goto out_unlock;
2935	}
2936
2937	if (whole && lock_task_sighand(task, &flags)) {
2938		struct task_struct *t = task;
2939
2940		task_io_accounting_add(&acct, &task->signal->ioac);
2941		while_each_thread(task, t)
2942			task_io_accounting_add(&acct, &t->ioac);
2943
2944		unlock_task_sighand(task, &flags);
2945	}
2946	seq_printf(m,
2947		   "rchar: %llu\n"
2948		   "wchar: %llu\n"
2949		   "syscr: %llu\n"
2950		   "syscw: %llu\n"
2951		   "read_bytes: %llu\n"
2952		   "write_bytes: %llu\n"
2953		   "cancelled_write_bytes: %llu\n",
2954		   (unsigned long long)acct.rchar,
2955		   (unsigned long long)acct.wchar,
2956		   (unsigned long long)acct.syscr,
2957		   (unsigned long long)acct.syscw,
2958		   (unsigned long long)acct.read_bytes,
2959		   (unsigned long long)acct.write_bytes,
2960		   (unsigned long long)acct.cancelled_write_bytes);
2961	result = 0;
2962
2963out_unlock:
2964	mutex_unlock(&task->signal->exec_update_mutex);
2965	return result;
2966}
2967
2968static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2969				  struct pid *pid, struct task_struct *task)
2970{
2971	return do_io_accounting(task, m, 0);
2972}
2973
2974static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2975				   struct pid *pid, struct task_struct *task)
2976{
2977	return do_io_accounting(task, m, 1);
2978}
2979#endif /* CONFIG_TASK_IO_ACCOUNTING */
2980
2981#ifdef CONFIG_USER_NS
2982static int proc_id_map_open(struct inode *inode, struct file *file,
2983	const struct seq_operations *seq_ops)
2984{
2985	struct user_namespace *ns = NULL;
2986	struct task_struct *task;
2987	struct seq_file *seq;
2988	int ret = -EINVAL;
2989
2990	task = get_proc_task(inode);
2991	if (task) {
2992		rcu_read_lock();
2993		ns = get_user_ns(task_cred_xxx(task, user_ns));
2994		rcu_read_unlock();
2995		put_task_struct(task);
2996	}
2997	if (!ns)
2998		goto err;
2999
3000	ret = seq_open(file, seq_ops);
3001	if (ret)
3002		goto err_put_ns;
3003
3004	seq = file->private_data;
3005	seq->private = ns;
3006
3007	return 0;
3008err_put_ns:
3009	put_user_ns(ns);
3010err:
3011	return ret;
3012}
3013
3014static int proc_id_map_release(struct inode *inode, struct file *file)
3015{
3016	struct seq_file *seq = file->private_data;
3017	struct user_namespace *ns = seq->private;
3018	put_user_ns(ns);
3019	return seq_release(inode, file);
3020}
3021
3022static int proc_uid_map_open(struct inode *inode, struct file *file)
3023{
3024	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3025}
3026
3027static int proc_gid_map_open(struct inode *inode, struct file *file)
3028{
3029	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3030}
3031
3032static int proc_projid_map_open(struct inode *inode, struct file *file)
3033{
3034	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3035}
3036
3037static const struct file_operations proc_uid_map_operations = {
3038	.open		= proc_uid_map_open,
3039	.write		= proc_uid_map_write,
3040	.read		= seq_read,
3041	.llseek		= seq_lseek,
3042	.release	= proc_id_map_release,
3043};
3044
3045static const struct file_operations proc_gid_map_operations = {
3046	.open		= proc_gid_map_open,
3047	.write		= proc_gid_map_write,
3048	.read		= seq_read,
3049	.llseek		= seq_lseek,
3050	.release	= proc_id_map_release,
3051};
3052
3053static const struct file_operations proc_projid_map_operations = {
3054	.open		= proc_projid_map_open,
3055	.write		= proc_projid_map_write,
3056	.read		= seq_read,
3057	.llseek		= seq_lseek,
3058	.release	= proc_id_map_release,
3059};
3060
3061static int proc_setgroups_open(struct inode *inode, struct file *file)
3062{
3063	struct user_namespace *ns = NULL;
3064	struct task_struct *task;
3065	int ret;
3066
3067	ret = -ESRCH;
3068	task = get_proc_task(inode);
3069	if (task) {
3070		rcu_read_lock();
3071		ns = get_user_ns(task_cred_xxx(task, user_ns));
3072		rcu_read_unlock();
3073		put_task_struct(task);
3074	}
3075	if (!ns)
3076		goto err;
3077
3078	if (file->f_mode & FMODE_WRITE) {
3079		ret = -EACCES;
3080		if (!ns_capable(ns, CAP_SYS_ADMIN))
3081			goto err_put_ns;
3082	}
3083
3084	ret = single_open(file, &proc_setgroups_show, ns);
3085	if (ret)
3086		goto err_put_ns;
3087
3088	return 0;
3089err_put_ns:
3090	put_user_ns(ns);
3091err:
3092	return ret;
3093}
3094
3095static int proc_setgroups_release(struct inode *inode, struct file *file)
3096{
3097	struct seq_file *seq = file->private_data;
3098	struct user_namespace *ns = seq->private;
3099	int ret = single_release(inode, file);
3100	put_user_ns(ns);
3101	return ret;
3102}
3103
3104static const struct file_operations proc_setgroups_operations = {
3105	.open		= proc_setgroups_open,
3106	.write		= proc_setgroups_write,
3107	.read		= seq_read,
3108	.llseek		= seq_lseek,
3109	.release	= proc_setgroups_release,
3110};
3111#endif /* CONFIG_USER_NS */
3112
3113static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3114				struct pid *pid, struct task_struct *task)
3115{
3116	int err = lock_trace(task);
3117	if (!err) {
3118		seq_printf(m, "%08x\n", task->personality);
3119		unlock_trace(task);
3120	}
3121	return err;
3122}
3123
3124#ifdef CONFIG_LIVEPATCH
3125static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3126				struct pid *pid, struct task_struct *task)
3127{
3128	seq_printf(m, "%d\n", task->patch_state);
3129	return 0;
3130}
3131#endif /* CONFIG_LIVEPATCH */
3132
3133#ifdef CONFIG_STACKLEAK_METRICS
3134static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3135				struct pid *pid, struct task_struct *task)
3136{
3137	unsigned long prev_depth = THREAD_SIZE -
3138				(task->prev_lowest_stack & (THREAD_SIZE - 1));
3139	unsigned long depth = THREAD_SIZE -
3140				(task->lowest_stack & (THREAD_SIZE - 1));
3141
3142	seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3143							prev_depth, depth);
3144	return 0;
3145}
3146#endif /* CONFIG_STACKLEAK_METRICS */
3147
3148/*
3149 * Thread groups
3150 */
3151static const struct file_operations proc_task_operations;
3152static const struct inode_operations proc_task_inode_operations;
3153
3154static const struct pid_entry tgid_base_stuff[] = {
3155	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3156	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3157	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3158	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3159	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3160#ifdef CONFIG_NET
3161	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3162#endif
3163	REG("environ",    S_IRUSR, proc_environ_operations),
3164	REG("auxv",       S_IRUSR, proc_auxv_operations),
3165	ONE("status",     S_IRUGO, proc_pid_status),
3166	ONE("personality", S_IRUSR, proc_pid_personality),
3167	ONE("limits",	  S_IRUGO, proc_pid_limits),
3168#ifdef CONFIG_SCHED_DEBUG
3169	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3170#endif
3171#ifdef CONFIG_SCHED_AUTOGROUP
3172	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3173#endif
3174#ifdef CONFIG_TIME_NS
3175	REG("timens_offsets",  S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3176#endif
3177	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3178#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3179	ONE("syscall",    S_IRUSR, proc_pid_syscall),
3180#endif
3181	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
3182	ONE("stat",       S_IRUGO, proc_tgid_stat),
3183	ONE("statm",      S_IRUGO, proc_pid_statm),
3184	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3185#ifdef CONFIG_NUMA
3186	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3187#endif
3188	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3189	LNK("cwd",        proc_cwd_link),
3190	LNK("root",       proc_root_link),
3191	LNK("exe",        proc_exe_link),
3192	REG("mounts",     S_IRUGO, proc_mounts_operations),
3193	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3194	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3195#ifdef CONFIG_PROC_PAGE_MONITOR
3196	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3197	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3198	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3199	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3200#endif
3201#ifdef CONFIG_SECURITY
3202	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3203#endif
3204#ifdef CONFIG_KALLSYMS
3205	ONE("wchan",      S_IRUGO, proc_pid_wchan),
3206#endif
3207#ifdef CONFIG_STACKTRACE
3208	ONE("stack",      S_IRUSR, proc_pid_stack),
3209#endif
3210#ifdef CONFIG_SCHED_INFO
3211	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
3212#endif
3213#ifdef CONFIG_LATENCYTOP
3214	REG("latency",  S_IRUGO, proc_lstats_operations),
3215#endif
3216#ifdef CONFIG_PROC_PID_CPUSET
3217	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
3218#endif
3219#ifdef CONFIG_CGROUPS
3220	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3221#endif
3222#ifdef CONFIG_PROC_CPU_RESCTRL
3223	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3224#endif
3225	ONE("oom_score",  S_IRUGO, proc_oom_score),
3226	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3227	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3228#ifdef CONFIG_AUDIT
3229	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3230	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3231#endif
3232#ifdef CONFIG_FAULT_INJECTION
3233	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3234	REG("fail-nth", 0644, proc_fail_nth_operations),
3235#endif
3236#ifdef CONFIG_ELF_CORE
3237	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3238#endif
3239#ifdef CONFIG_TASK_IO_ACCOUNTING
3240	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3241#endif
3242#ifdef CONFIG_USER_NS
3243	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3244	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3245	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3246	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3247#endif
3248#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3249	REG("timers",	  S_IRUGO, proc_timers_operations),
3250#endif
3251	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3252#ifdef CONFIG_LIVEPATCH
3253	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3254#endif
3255#ifdef CONFIG_STACKLEAK_METRICS
3256	ONE("stack_depth", S_IRUGO, proc_stack_depth),
3257#endif
3258#ifdef CONFIG_PROC_PID_ARCH_STATUS
3259	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3260#endif
 
 
 
3261};
3262
3263static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3264{
3265	return proc_pident_readdir(file, ctx,
3266				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3267}
3268
3269static const struct file_operations proc_tgid_base_operations = {
3270	.read		= generic_read_dir,
3271	.iterate_shared	= proc_tgid_base_readdir,
3272	.llseek		= generic_file_llseek,
3273};
3274
3275struct pid *tgid_pidfd_to_pid(const struct file *file)
3276{
3277	if (file->f_op != &proc_tgid_base_operations)
3278		return ERR_PTR(-EBADF);
3279
3280	return proc_pid(file_inode(file));
3281}
3282
3283static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3284{
3285	return proc_pident_lookup(dir, dentry,
3286				  tgid_base_stuff,
3287				  tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3288}
3289
3290static const struct inode_operations proc_tgid_base_inode_operations = {
3291	.lookup		= proc_tgid_base_lookup,
3292	.getattr	= pid_getattr,
3293	.setattr	= proc_setattr,
3294	.permission	= proc_pid_permission,
3295};
3296
3297/**
3298 * proc_flush_pid -  Remove dcache entries for @pid from the /proc dcache.
3299 * @pid: pid that should be flushed.
3300 *
3301 * This function walks a list of inodes (that belong to any proc
3302 * filesystem) that are attached to the pid and flushes them from
3303 * the dentry cache.
3304 *
3305 * It is safe and reasonable to cache /proc entries for a task until
3306 * that task exits.  After that they just clog up the dcache with
3307 * useless entries, possibly causing useful dcache entries to be
3308 * flushed instead.  This routine is provided to flush those useless
3309 * dcache entries when a process is reaped.
3310 *
3311 * NOTE: This routine is just an optimization so it does not guarantee
3312 *       that no dcache entries will exist after a process is reaped
3313 *       it just makes it very unlikely that any will persist.
3314 */
3315
3316void proc_flush_pid(struct pid *pid)
3317{
3318	proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3319}
3320
3321static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3322				   struct task_struct *task, const void *ptr)
3323{
3324	struct inode *inode;
3325
3326	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3327	if (!inode)
3328		return ERR_PTR(-ENOENT);
3329
3330	inode->i_op = &proc_tgid_base_inode_operations;
3331	inode->i_fop = &proc_tgid_base_operations;
3332	inode->i_flags|=S_IMMUTABLE;
3333
3334	set_nlink(inode, nlink_tgid);
3335	pid_update_inode(task, inode);
3336
3337	d_set_d_op(dentry, &pid_dentry_operations);
3338	return d_splice_alias(inode, dentry);
3339}
3340
3341struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3342{
3343	struct task_struct *task;
3344	unsigned tgid;
3345	struct proc_fs_info *fs_info;
3346	struct pid_namespace *ns;
3347	struct dentry *result = ERR_PTR(-ENOENT);
3348
3349	tgid = name_to_int(&dentry->d_name);
3350	if (tgid == ~0U)
3351		goto out;
3352
3353	fs_info = proc_sb_info(dentry->d_sb);
3354	ns = fs_info->pid_ns;
3355	rcu_read_lock();
3356	task = find_task_by_pid_ns(tgid, ns);
3357	if (task)
3358		get_task_struct(task);
3359	rcu_read_unlock();
3360	if (!task)
3361		goto out;
3362
3363	/* Limit procfs to only ptraceable tasks */
3364	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3365		if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3366			goto out_put_task;
3367	}
3368
3369	result = proc_pid_instantiate(dentry, task, NULL);
3370out_put_task:
3371	put_task_struct(task);
3372out:
3373	return result;
3374}
3375
3376/*
3377 * Find the first task with tgid >= tgid
3378 *
3379 */
3380struct tgid_iter {
3381	unsigned int tgid;
3382	struct task_struct *task;
3383};
3384static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3385{
3386	struct pid *pid;
3387
3388	if (iter.task)
3389		put_task_struct(iter.task);
3390	rcu_read_lock();
3391retry:
3392	iter.task = NULL;
3393	pid = find_ge_pid(iter.tgid, ns);
3394	if (pid) {
3395		iter.tgid = pid_nr_ns(pid, ns);
3396		iter.task = pid_task(pid, PIDTYPE_TGID);
3397		if (!iter.task) {
3398			iter.tgid += 1;
3399			goto retry;
3400		}
3401		get_task_struct(iter.task);
3402	}
3403	rcu_read_unlock();
3404	return iter;
3405}
3406
3407#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3408
3409/* for the /proc/ directory itself, after non-process stuff has been done */
3410int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3411{
3412	struct tgid_iter iter;
3413	struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3414	struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3415	loff_t pos = ctx->pos;
3416
3417	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3418		return 0;
3419
3420	if (pos == TGID_OFFSET - 2) {
3421		struct inode *inode = d_inode(fs_info->proc_self);
3422		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3423			return 0;
3424		ctx->pos = pos = pos + 1;
3425	}
3426	if (pos == TGID_OFFSET - 1) {
3427		struct inode *inode = d_inode(fs_info->proc_thread_self);
3428		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3429			return 0;
3430		ctx->pos = pos = pos + 1;
3431	}
3432	iter.tgid = pos - TGID_OFFSET;
3433	iter.task = NULL;
3434	for (iter = next_tgid(ns, iter);
3435	     iter.task;
3436	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3437		char name[10 + 1];
3438		unsigned int len;
3439
3440		cond_resched();
3441		if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3442			continue;
3443
3444		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3445		ctx->pos = iter.tgid + TGID_OFFSET;
3446		if (!proc_fill_cache(file, ctx, name, len,
3447				     proc_pid_instantiate, iter.task, NULL)) {
3448			put_task_struct(iter.task);
3449			return 0;
3450		}
3451	}
3452	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3453	return 0;
3454}
3455
3456/*
3457 * proc_tid_comm_permission is a special permission function exclusively
3458 * used for the node /proc/<pid>/task/<tid>/comm.
3459 * It bypasses generic permission checks in the case where a task of the same
3460 * task group attempts to access the node.
3461 * The rationale behind this is that glibc and bionic access this node for
3462 * cross thread naming (pthread_set/getname_np(!self)). However, if
3463 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3464 * which locks out the cross thread naming implementation.
3465 * This function makes sure that the node is always accessible for members of
3466 * same thread group.
3467 */
3468static int proc_tid_comm_permission(struct inode *inode, int mask)
 
3469{
3470	bool is_same_tgroup;
3471	struct task_struct *task;
3472
3473	task = get_proc_task(inode);
3474	if (!task)
3475		return -ESRCH;
3476	is_same_tgroup = same_thread_group(current, task);
3477	put_task_struct(task);
3478
3479	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3480		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3481		 * read or written by the members of the corresponding
3482		 * thread group.
3483		 */
3484		return 0;
3485	}
3486
3487	return generic_permission(inode, mask);
3488}
3489
3490static const struct inode_operations proc_tid_comm_inode_operations = {
3491		.permission = proc_tid_comm_permission,
3492};
3493
3494/*
3495 * Tasks
3496 */
3497static const struct pid_entry tid_base_stuff[] = {
3498	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3499	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3500	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3501#ifdef CONFIG_NET
3502	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3503#endif
3504	REG("environ",   S_IRUSR, proc_environ_operations),
3505	REG("auxv",      S_IRUSR, proc_auxv_operations),
3506	ONE("status",    S_IRUGO, proc_pid_status),
3507	ONE("personality", S_IRUSR, proc_pid_personality),
3508	ONE("limits",	 S_IRUGO, proc_pid_limits),
3509#ifdef CONFIG_SCHED_DEBUG
3510	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3511#endif
3512	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3513			 &proc_tid_comm_inode_operations,
3514			 &proc_pid_set_comm_operations, {}),
3515#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3516	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3517#endif
3518	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3519	ONE("stat",      S_IRUGO, proc_tid_stat),
3520	ONE("statm",     S_IRUGO, proc_pid_statm),
3521	REG("maps",      S_IRUGO, proc_pid_maps_operations),
3522#ifdef CONFIG_PROC_CHILDREN
3523	REG("children",  S_IRUGO, proc_tid_children_operations),
3524#endif
3525#ifdef CONFIG_NUMA
3526	REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3527#endif
3528	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3529	LNK("cwd",       proc_cwd_link),
3530	LNK("root",      proc_root_link),
3531	LNK("exe",       proc_exe_link),
3532	REG("mounts",    S_IRUGO, proc_mounts_operations),
3533	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3534#ifdef CONFIG_PROC_PAGE_MONITOR
3535	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3536	REG("smaps",     S_IRUGO, proc_pid_smaps_operations),
3537	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3538	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3539#endif
3540#ifdef CONFIG_SECURITY
3541	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3542#endif
3543#ifdef CONFIG_KALLSYMS
3544	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3545#endif
3546#ifdef CONFIG_STACKTRACE
3547	ONE("stack",      S_IRUSR, proc_pid_stack),
3548#endif
3549#ifdef CONFIG_SCHED_INFO
3550	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3551#endif
3552#ifdef CONFIG_LATENCYTOP
3553	REG("latency",  S_IRUGO, proc_lstats_operations),
3554#endif
3555#ifdef CONFIG_PROC_PID_CPUSET
3556	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3557#endif
3558#ifdef CONFIG_CGROUPS
3559	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3560#endif
3561#ifdef CONFIG_PROC_CPU_RESCTRL
3562	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3563#endif
3564	ONE("oom_score", S_IRUGO, proc_oom_score),
3565	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3566	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3567#ifdef CONFIG_AUDIT
3568	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3569	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3570#endif
3571#ifdef CONFIG_FAULT_INJECTION
3572	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3573	REG("fail-nth", 0644, proc_fail_nth_operations),
3574#endif
3575#ifdef CONFIG_TASK_IO_ACCOUNTING
3576	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3577#endif
3578#ifdef CONFIG_USER_NS
3579	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3580	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3581	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3582	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3583#endif
3584#ifdef CONFIG_LIVEPATCH
3585	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3586#endif
3587#ifdef CONFIG_PROC_PID_ARCH_STATUS
3588	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3589#endif
 
 
 
3590};
3591
3592static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3593{
3594	return proc_pident_readdir(file, ctx,
3595				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3596}
3597
3598static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3599{
3600	return proc_pident_lookup(dir, dentry,
3601				  tid_base_stuff,
3602				  tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3603}
3604
3605static const struct file_operations proc_tid_base_operations = {
3606	.read		= generic_read_dir,
3607	.iterate_shared	= proc_tid_base_readdir,
3608	.llseek		= generic_file_llseek,
3609};
3610
3611static const struct inode_operations proc_tid_base_inode_operations = {
3612	.lookup		= proc_tid_base_lookup,
3613	.getattr	= pid_getattr,
3614	.setattr	= proc_setattr,
3615};
3616
3617static struct dentry *proc_task_instantiate(struct dentry *dentry,
3618	struct task_struct *task, const void *ptr)
3619{
3620	struct inode *inode;
3621	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3622	if (!inode)
3623		return ERR_PTR(-ENOENT);
3624
3625	inode->i_op = &proc_tid_base_inode_operations;
3626	inode->i_fop = &proc_tid_base_operations;
3627	inode->i_flags |= S_IMMUTABLE;
3628
3629	set_nlink(inode, nlink_tid);
3630	pid_update_inode(task, inode);
3631
3632	d_set_d_op(dentry, &pid_dentry_operations);
3633	return d_splice_alias(inode, dentry);
3634}
3635
3636static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3637{
3638	struct task_struct *task;
3639	struct task_struct *leader = get_proc_task(dir);
3640	unsigned tid;
3641	struct proc_fs_info *fs_info;
3642	struct pid_namespace *ns;
3643	struct dentry *result = ERR_PTR(-ENOENT);
3644
3645	if (!leader)
3646		goto out_no_task;
3647
3648	tid = name_to_int(&dentry->d_name);
3649	if (tid == ~0U)
3650		goto out;
3651
3652	fs_info = proc_sb_info(dentry->d_sb);
3653	ns = fs_info->pid_ns;
3654	rcu_read_lock();
3655	task = find_task_by_pid_ns(tid, ns);
3656	if (task)
3657		get_task_struct(task);
3658	rcu_read_unlock();
3659	if (!task)
3660		goto out;
3661	if (!same_thread_group(leader, task))
3662		goto out_drop_task;
3663
3664	result = proc_task_instantiate(dentry, task, NULL);
3665out_drop_task:
3666	put_task_struct(task);
3667out:
3668	put_task_struct(leader);
3669out_no_task:
3670	return result;
3671}
3672
3673/*
3674 * Find the first tid of a thread group to return to user space.
3675 *
3676 * Usually this is just the thread group leader, but if the users
3677 * buffer was too small or there was a seek into the middle of the
3678 * directory we have more work todo.
3679 *
3680 * In the case of a short read we start with find_task_by_pid.
3681 *
3682 * In the case of a seek we start with the leader and walk nr
3683 * threads past it.
3684 */
3685static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3686					struct pid_namespace *ns)
3687{
3688	struct task_struct *pos, *task;
3689	unsigned long nr = f_pos;
3690
3691	if (nr != f_pos)	/* 32bit overflow? */
3692		return NULL;
3693
3694	rcu_read_lock();
3695	task = pid_task(pid, PIDTYPE_PID);
3696	if (!task)
3697		goto fail;
3698
3699	/* Attempt to start with the tid of a thread */
3700	if (tid && nr) {
3701		pos = find_task_by_pid_ns(tid, ns);
3702		if (pos && same_thread_group(pos, task))
3703			goto found;
3704	}
3705
3706	/* If nr exceeds the number of threads there is nothing todo */
3707	if (nr >= get_nr_threads(task))
3708		goto fail;
3709
3710	/* If we haven't found our starting place yet start
3711	 * with the leader and walk nr threads forward.
3712	 */
3713	pos = task = task->group_leader;
3714	do {
3715		if (!nr--)
3716			goto found;
3717	} while_each_thread(task, pos);
3718fail:
3719	pos = NULL;
3720	goto out;
3721found:
3722	get_task_struct(pos);
3723out:
3724	rcu_read_unlock();
3725	return pos;
3726}
3727
3728/*
3729 * Find the next thread in the thread list.
3730 * Return NULL if there is an error or no next thread.
3731 *
3732 * The reference to the input task_struct is released.
3733 */
3734static struct task_struct *next_tid(struct task_struct *start)
3735{
3736	struct task_struct *pos = NULL;
3737	rcu_read_lock();
3738	if (pid_alive(start)) {
3739		pos = next_thread(start);
3740		if (thread_group_leader(pos))
3741			pos = NULL;
3742		else
3743			get_task_struct(pos);
3744	}
3745	rcu_read_unlock();
3746	put_task_struct(start);
3747	return pos;
3748}
3749
3750/* for the /proc/TGID/task/ directories */
3751static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3752{
3753	struct inode *inode = file_inode(file);
3754	struct task_struct *task;
3755	struct pid_namespace *ns;
3756	int tid;
3757
3758	if (proc_inode_is_dead(inode))
3759		return -ENOENT;
3760
3761	if (!dir_emit_dots(file, ctx))
3762		return 0;
3763
3764	/* f_version caches the tgid value that the last readdir call couldn't
3765	 * return. lseek aka telldir automagically resets f_version to 0.
3766	 */
3767	ns = proc_pid_ns(inode->i_sb);
3768	tid = (int)file->f_version;
3769	file->f_version = 0;
3770	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3771	     task;
3772	     task = next_tid(task), ctx->pos++) {
3773		char name[10 + 1];
3774		unsigned int len;
3775		tid = task_pid_nr_ns(task, ns);
3776		len = snprintf(name, sizeof(name), "%u", tid);
3777		if (!proc_fill_cache(file, ctx, name, len,
3778				proc_task_instantiate, task, NULL)) {
3779			/* returning this tgid failed, save it as the first
3780			 * pid for the next readir call */
3781			file->f_version = (u64)tid;
3782			put_task_struct(task);
3783			break;
3784		}
3785	}
3786
3787	return 0;
3788}
3789
3790static int proc_task_getattr(const struct path *path, struct kstat *stat,
 
3791			     u32 request_mask, unsigned int query_flags)
3792{
3793	struct inode *inode = d_inode(path->dentry);
3794	struct task_struct *p = get_proc_task(inode);
3795	generic_fillattr(inode, stat);
3796
3797	if (p) {
3798		stat->nlink += get_nr_threads(p);
3799		put_task_struct(p);
3800	}
3801
3802	return 0;
3803}
3804
3805static const struct inode_operations proc_task_inode_operations = {
3806	.lookup		= proc_task_lookup,
3807	.getattr	= proc_task_getattr,
3808	.setattr	= proc_setattr,
3809	.permission	= proc_pid_permission,
3810};
3811
3812static const struct file_operations proc_task_operations = {
3813	.read		= generic_read_dir,
3814	.iterate_shared	= proc_task_readdir,
3815	.llseek		= generic_file_llseek,
3816};
3817
3818void __init set_proc_pid_nlink(void)
3819{
3820	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3821	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3822}