Linux Audio

Check our new training course

Loading...
v4.10.11
 
   1/*
   2 *  linux/fs/proc/base.c
   3 *
   4 *  Copyright (C) 1991, 1992 Linus Torvalds
   5 *
   6 *  proc base directory handling functions
   7 *
   8 *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
   9 *  Instead of using magical inumbers to determine the kind of object
  10 *  we allocate and fill in-core inodes upon lookup. They don't even
  11 *  go into icache. We cache the reference to task_struct upon lookup too.
  12 *  Eventually it should become a filesystem in its own. We don't use the
  13 *  rest of procfs anymore.
  14 *
  15 *
  16 *  Changelog:
  17 *  17-Jan-2005
  18 *  Allan Bezerra
  19 *  Bruna Moreira <bruna.moreira@indt.org.br>
  20 *  Edjard Mota <edjard.mota@indt.org.br>
  21 *  Ilias Biris <ilias.biris@indt.org.br>
  22 *  Mauricio Lin <mauricio.lin@indt.org.br>
  23 *
  24 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  25 *
  26 *  A new process specific entry (smaps) included in /proc. It shows the
  27 *  size of rss for each memory area. The maps entry lacks information
  28 *  about physical memory size (rss) for each mapped file, i.e.,
  29 *  rss information for executables and library files.
  30 *  This additional information is useful for any tools that need to know
  31 *  about physical memory consumption for a process specific library.
  32 *
  33 *  Changelog:
  34 *  21-Feb-2005
  35 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
  36 *  Pud inclusion in the page table walking.
  37 *
  38 *  ChangeLog:
  39 *  10-Mar-2005
  40 *  10LE Instituto Nokia de Tecnologia - INdT:
  41 *  A better way to walks through the page table as suggested by Hugh Dickins.
  42 *
  43 *  Simo Piiroinen <simo.piiroinen@nokia.com>:
  44 *  Smaps information related to shared, private, clean and dirty pages.
  45 *
  46 *  Paul Mundt <paul.mundt@nokia.com>:
  47 *  Overall revision about smaps.
  48 */
  49
  50#include <linux/uaccess.h>
  51
  52#include <linux/errno.h>
  53#include <linux/time.h>
  54#include <linux/proc_fs.h>
  55#include <linux/stat.h>
  56#include <linux/task_io_accounting_ops.h>
  57#include <linux/init.h>
  58#include <linux/capability.h>
  59#include <linux/file.h>
  60#include <linux/fdtable.h>
 
  61#include <linux/string.h>
  62#include <linux/seq_file.h>
  63#include <linux/namei.h>
  64#include <linux/mnt_namespace.h>
  65#include <linux/mm.h>
  66#include <linux/swap.h>
  67#include <linux/rcupdate.h>
  68#include <linux/kallsyms.h>
  69#include <linux/stacktrace.h>
  70#include <linux/resource.h>
  71#include <linux/module.h>
  72#include <linux/mount.h>
  73#include <linux/security.h>
  74#include <linux/ptrace.h>
  75#include <linux/tracehook.h>
  76#include <linux/printk.h>
 
  77#include <linux/cgroup.h>
  78#include <linux/cpuset.h>
  79#include <linux/audit.h>
  80#include <linux/poll.h>
  81#include <linux/nsproxy.h>
  82#include <linux/oom.h>
  83#include <linux/elf.h>
  84#include <linux/pid_namespace.h>
  85#include <linux/user_namespace.h>
  86#include <linux/fs_struct.h>
  87#include <linux/slab.h>
  88#include <linux/flex_array.h>
 
 
 
 
  89#include <linux/posix-timers.h>
  90#ifdef CONFIG_HARDWALL
  91#include <asm/hardwall.h>
  92#endif
  93#include <trace/events/oom.h>
  94#include "internal.h"
  95#include "fd.h"
  96
 
 
  97/* NOTE:
  98 *	Implementing inode permission operations in /proc is almost
  99 *	certainly an error.  Permission checks need to happen during
 100 *	each system call not at open time.  The reason is that most of
 101 *	what we wish to check for permissions in /proc varies at runtime.
 102 *
 103 *	The classic example of a problem is opening file descriptors
 104 *	in /proc for a task before it execs a suid executable.
 105 */
 106
 107static u8 nlink_tid;
 108static u8 nlink_tgid;
 109
 110struct pid_entry {
 111	const char *name;
 112	unsigned int len;
 113	umode_t mode;
 114	const struct inode_operations *iop;
 115	const struct file_operations *fop;
 116	union proc_op op;
 117};
 118
 119#define NOD(NAME, MODE, IOP, FOP, OP) {			\
 120	.name = (NAME),					\
 121	.len  = sizeof(NAME) - 1,			\
 122	.mode = MODE,					\
 123	.iop  = IOP,					\
 124	.fop  = FOP,					\
 125	.op   = OP,					\
 126}
 127
 128#define DIR(NAME, MODE, iops, fops)	\
 129	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
 130#define LNK(NAME, get_link)					\
 131	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
 132		&proc_pid_link_inode_operations, NULL,		\
 133		{ .proc_get_link = get_link } )
 134#define REG(NAME, MODE, fops)				\
 135	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
 136#define ONE(NAME, MODE, show)				\
 137	NOD(NAME, (S_IFREG|(MODE)), 			\
 138		NULL, &proc_single_file_operations,	\
 139		{ .proc_show = show } )
 
 
 
 
 140
 141/*
 142 * Count the number of hardlinks for the pid_entry table, excluding the .
 143 * and .. links.
 144 */
 145static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
 146	unsigned int n)
 147{
 148	unsigned int i;
 149	unsigned int count;
 150
 151	count = 2;
 152	for (i = 0; i < n; ++i) {
 153		if (S_ISDIR(entries[i].mode))
 154			++count;
 155	}
 156
 157	return count;
 158}
 159
 160static int get_task_root(struct task_struct *task, struct path *root)
 161{
 162	int result = -ENOENT;
 163
 164	task_lock(task);
 165	if (task->fs) {
 166		get_fs_root(task->fs, root);
 167		result = 0;
 168	}
 169	task_unlock(task);
 170	return result;
 171}
 172
 173static int proc_cwd_link(struct dentry *dentry, struct path *path)
 174{
 175	struct task_struct *task = get_proc_task(d_inode(dentry));
 176	int result = -ENOENT;
 177
 178	if (task) {
 179		task_lock(task);
 180		if (task->fs) {
 181			get_fs_pwd(task->fs, path);
 182			result = 0;
 183		}
 184		task_unlock(task);
 185		put_task_struct(task);
 186	}
 187	return result;
 188}
 189
 190static int proc_root_link(struct dentry *dentry, struct path *path)
 191{
 192	struct task_struct *task = get_proc_task(d_inode(dentry));
 193	int result = -ENOENT;
 194
 195	if (task) {
 196		result = get_task_root(task, path);
 197		put_task_struct(task);
 198	}
 199	return result;
 200}
 201
 202static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 203				     size_t _count, loff_t *pos)
 
 
 
 
 
 204{
 205	struct task_struct *tsk;
 206	struct mm_struct *mm;
 207	char *page;
 208	unsigned long count = _count;
 209	unsigned long arg_start, arg_end, env_start, env_end;
 210	unsigned long len1, len2, len;
 211	unsigned long p;
 212	char c;
 213	ssize_t rv;
 214
 215	BUG_ON(*pos < 0);
 216
 217	tsk = get_proc_task(file_inode(file));
 218	if (!tsk)
 219		return -ESRCH;
 220	mm = get_task_mm(tsk);
 221	put_task_struct(tsk);
 222	if (!mm)
 223		return 0;
 224	/* Check if process spawned far enough to have cmdline. */
 225	if (!mm->env_end) {
 226		rv = 0;
 227		goto out_mmput;
 228	}
 229
 230	page = (char *)__get_free_page(GFP_TEMPORARY);
 231	if (!page) {
 232		rv = -ENOMEM;
 233		goto out_mmput;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 234	}
 
 
 
 235
 236	down_read(&mm->mmap_sem);
 
 
 
 
 
 
 
 
 
 
 
 237	arg_start = mm->arg_start;
 238	arg_end = mm->arg_end;
 239	env_start = mm->env_start;
 240	env_end = mm->env_end;
 241	up_read(&mm->mmap_sem);
 242
 243	BUG_ON(arg_start > arg_end);
 244	BUG_ON(env_start > env_end);
 245
 246	len1 = arg_end - arg_start;
 247	len2 = env_end - env_start;
 248
 249	/* Empty ARGV. */
 250	if (len1 == 0) {
 251		rv = 0;
 252		goto out_free_page;
 253	}
 254	/*
 255	 * Inherently racy -- command line shares address space
 256	 * with code and data.
 
 257	 */
 258	rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
 259	if (rv <= 0)
 260		goto out_free_page;
 261
 262	rv = 0;
 263
 264	if (c == '\0') {
 265		/* Command line (set of strings) occupies whole ARGV. */
 266		if (len1 <= *pos)
 267			goto out_free_page;
 268
 269		p = arg_start + *pos;
 270		len = len1 - *pos;
 271		while (count > 0 && len > 0) {
 272			unsigned int _count;
 273			int nr_read;
 274
 275			_count = min3(count, len, PAGE_SIZE);
 276			nr_read = access_remote_vm(mm, p, page, _count, 0);
 277			if (nr_read < 0)
 278				rv = nr_read;
 279			if (nr_read <= 0)
 280				goto out_free_page;
 281
 282			if (copy_to_user(buf, page, nr_read)) {
 283				rv = -EFAULT;
 284				goto out_free_page;
 285			}
 286
 287			p	+= nr_read;
 288			len	-= nr_read;
 289			buf	+= nr_read;
 290			count	-= nr_read;
 291			rv	+= nr_read;
 292		}
 293	} else {
 294		/*
 295		 * Command line (1 string) occupies ARGV and maybe
 296		 * extends into ENVP.
 297		 */
 298		if (len1 + len2 <= *pos)
 299			goto skip_argv_envp;
 300		if (len1 <= *pos)
 301			goto skip_argv;
 302
 303		p = arg_start + *pos;
 304		len = len1 - *pos;
 305		while (count > 0 && len > 0) {
 306			unsigned int _count, l;
 307			int nr_read;
 308			bool final;
 309
 310			_count = min3(count, len, PAGE_SIZE);
 311			nr_read = access_remote_vm(mm, p, page, _count, 0);
 312			if (nr_read < 0)
 313				rv = nr_read;
 314			if (nr_read <= 0)
 315				goto out_free_page;
 316
 317			/*
 318			 * Command line can be shorter than whole ARGV
 319			 * even if last "marker" byte says it is not.
 320			 */
 321			final = false;
 322			l = strnlen(page, nr_read);
 323			if (l < nr_read) {
 324				nr_read = l;
 325				final = true;
 326			}
 327
 328			if (copy_to_user(buf, page, nr_read)) {
 329				rv = -EFAULT;
 330				goto out_free_page;
 331			}
 332
 333			p	+= nr_read;
 334			len	-= nr_read;
 335			buf	+= nr_read;
 336			count	-= nr_read;
 337			rv	+= nr_read;
 338
 339			if (final)
 340				goto out_free_page;
 341		}
 342skip_argv:
 343		/*
 344		 * Command line (1 string) occupies ARGV and
 345		 * extends into ENVP.
 346		 */
 347		if (len1 <= *pos) {
 348			p = env_start + *pos - len1;
 349			len = len1 + len2 - *pos;
 350		} else {
 351			p = env_start;
 352			len = len2;
 353		}
 354		while (count > 0 && len > 0) {
 355			unsigned int _count, l;
 356			int nr_read;
 357			bool final;
 358
 359			_count = min3(count, len, PAGE_SIZE);
 360			nr_read = access_remote_vm(mm, p, page, _count, 0);
 361			if (nr_read < 0)
 362				rv = nr_read;
 363			if (nr_read <= 0)
 364				goto out_free_page;
 365
 366			/* Find EOS. */
 367			final = false;
 368			l = strnlen(page, nr_read);
 369			if (l < nr_read) {
 370				nr_read = l;
 371				final = true;
 372			}
 373
 374			if (copy_to_user(buf, page, nr_read)) {
 375				rv = -EFAULT;
 376				goto out_free_page;
 377			}
 378
 379			p	+= nr_read;
 380			len	-= nr_read;
 381			buf	+= nr_read;
 382			count	-= nr_read;
 383			rv	+= nr_read;
 384
 385			if (final)
 386				goto out_free_page;
 387		}
 388skip_argv_envp:
 389		;
 390	}
 391
 392out_free_page:
 393	free_page((unsigned long)page);
 394out_mmput:
 395	mmput(mm);
 396	if (rv > 0)
 397		*pos += rv;
 398	return rv;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 399}
 400
 401static const struct file_operations proc_pid_cmdline_ops = {
 402	.read	= proc_pid_cmdline_read,
 403	.llseek	= generic_file_llseek,
 404};
 405
 406#ifdef CONFIG_KALLSYMS
 407/*
 408 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
 409 * Returns the resolved symbol.  If that fails, simply return the address.
 410 */
 411static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
 412			  struct pid *pid, struct task_struct *task)
 413{
 414	unsigned long wchan;
 415	char symname[KSYM_NAME_LEN];
 416
 417	wchan = get_wchan(task);
 
 418
 419	if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
 420			&& !lookup_symbol_name(wchan, symname))
 421		seq_printf(m, "%s", symname);
 422	else
 423		seq_putc(m, '0');
 424
 
 
 425	return 0;
 426}
 427#endif /* CONFIG_KALLSYMS */
 428
 429static int lock_trace(struct task_struct *task)
 430{
 431	int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
 432	if (err)
 433		return err;
 434	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
 435		mutex_unlock(&task->signal->cred_guard_mutex);
 436		return -EPERM;
 437	}
 438	return 0;
 439}
 440
 441static void unlock_trace(struct task_struct *task)
 442{
 443	mutex_unlock(&task->signal->cred_guard_mutex);
 444}
 445
 446#ifdef CONFIG_STACKTRACE
 447
 448#define MAX_STACK_TRACE_DEPTH	64
 449
 450static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
 451			  struct pid *pid, struct task_struct *task)
 452{
 453	struct stack_trace trace;
 454	unsigned long *entries;
 455	int err;
 456	int i;
 457
 458	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 459	if (!entries)
 460		return -ENOMEM;
 461
 462	trace.nr_entries	= 0;
 463	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
 464	trace.entries		= entries;
 465	trace.skip		= 0;
 466
 467	err = lock_trace(task);
 468	if (!err) {
 469		save_stack_trace_tsk(task, &trace);
 470
 471		for (i = 0; i < trace.nr_entries; i++) {
 472			seq_printf(m, "[<%pK>] %pB\n",
 473				   (void *)entries[i], (void *)entries[i]);
 
 
 474		}
 
 475		unlock_trace(task);
 476	}
 477	kfree(entries);
 478
 479	return err;
 480}
 481#endif
 482
 483#ifdef CONFIG_SCHED_INFO
 484/*
 485 * Provides /proc/PID/schedstat
 486 */
 487static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
 488			      struct pid *pid, struct task_struct *task)
 489{
 490	if (unlikely(!sched_info_on()))
 491		seq_printf(m, "0 0 0\n");
 492	else
 493		seq_printf(m, "%llu %llu %lu\n",
 494		   (unsigned long long)task->se.sum_exec_runtime,
 495		   (unsigned long long)task->sched_info.run_delay,
 496		   task->sched_info.pcount);
 497
 498	return 0;
 499}
 500#endif
 501
 502#ifdef CONFIG_LATENCYTOP
 503static int lstats_show_proc(struct seq_file *m, void *v)
 504{
 505	int i;
 506	struct inode *inode = m->private;
 507	struct task_struct *task = get_proc_task(inode);
 508
 509	if (!task)
 510		return -ESRCH;
 511	seq_puts(m, "Latency Top version : v0.1\n");
 512	for (i = 0; i < 32; i++) {
 513		struct latency_record *lr = &task->latency_record[i];
 514		if (lr->backtrace[0]) {
 515			int q;
 516			seq_printf(m, "%i %li %li",
 517				   lr->count, lr->time, lr->max);
 518			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
 519				unsigned long bt = lr->backtrace[q];
 
 520				if (!bt)
 521					break;
 522				if (bt == ULONG_MAX)
 523					break;
 524				seq_printf(m, " %ps", (void *)bt);
 525			}
 526			seq_putc(m, '\n');
 527		}
 528
 529	}
 530	put_task_struct(task);
 531	return 0;
 532}
 533
 534static int lstats_open(struct inode *inode, struct file *file)
 535{
 536	return single_open(file, lstats_show_proc, inode);
 537}
 538
 539static ssize_t lstats_write(struct file *file, const char __user *buf,
 540			    size_t count, loff_t *offs)
 541{
 542	struct task_struct *task = get_proc_task(file_inode(file));
 543
 544	if (!task)
 545		return -ESRCH;
 546	clear_all_latency_tracing(task);
 547	put_task_struct(task);
 548
 549	return count;
 550}
 551
 552static const struct file_operations proc_lstats_operations = {
 553	.open		= lstats_open,
 554	.read		= seq_read,
 555	.write		= lstats_write,
 556	.llseek		= seq_lseek,
 557	.release	= single_release,
 558};
 559
 560#endif
 561
 562static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 563			  struct pid *pid, struct task_struct *task)
 564{
 565	unsigned long totalpages = totalram_pages + total_swap_pages;
 566	unsigned long points = 0;
 
 
 
 
 
 
 
 
 
 
 567
 568	points = oom_badness(task, NULL, NULL, totalpages) *
 569					1000 / totalpages;
 570	seq_printf(m, "%lu\n", points);
 571
 572	return 0;
 573}
 574
 575struct limit_names {
 576	const char *name;
 577	const char *unit;
 578};
 579
 580static const struct limit_names lnames[RLIM_NLIMITS] = {
 581	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
 582	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
 583	[RLIMIT_DATA] = {"Max data size", "bytes"},
 584	[RLIMIT_STACK] = {"Max stack size", "bytes"},
 585	[RLIMIT_CORE] = {"Max core file size", "bytes"},
 586	[RLIMIT_RSS] = {"Max resident set", "bytes"},
 587	[RLIMIT_NPROC] = {"Max processes", "processes"},
 588	[RLIMIT_NOFILE] = {"Max open files", "files"},
 589	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
 590	[RLIMIT_AS] = {"Max address space", "bytes"},
 591	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
 592	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
 593	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
 594	[RLIMIT_NICE] = {"Max nice priority", NULL},
 595	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
 596	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
 597};
 598
 599/* Display limits for a process */
 600static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
 601			   struct pid *pid, struct task_struct *task)
 602{
 603	unsigned int i;
 604	unsigned long flags;
 605
 606	struct rlimit rlim[RLIM_NLIMITS];
 607
 608	if (!lock_task_sighand(task, &flags))
 609		return 0;
 610	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
 611	unlock_task_sighand(task, &flags);
 612
 613	/*
 614	 * print the file header
 615	 */
 616       seq_printf(m, "%-25s %-20s %-20s %-10s\n",
 617		  "Limit", "Soft Limit", "Hard Limit", "Units");
 
 
 618
 619	for (i = 0; i < RLIM_NLIMITS; i++) {
 620		if (rlim[i].rlim_cur == RLIM_INFINITY)
 621			seq_printf(m, "%-25s %-20s ",
 622				   lnames[i].name, "unlimited");
 623		else
 624			seq_printf(m, "%-25s %-20lu ",
 625				   lnames[i].name, rlim[i].rlim_cur);
 626
 627		if (rlim[i].rlim_max == RLIM_INFINITY)
 628			seq_printf(m, "%-20s ", "unlimited");
 629		else
 630			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
 631
 632		if (lnames[i].unit)
 633			seq_printf(m, "%-10s\n", lnames[i].unit);
 634		else
 635			seq_putc(m, '\n');
 636	}
 637
 638	return 0;
 639}
 640
 641#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
 642static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
 643			    struct pid *pid, struct task_struct *task)
 644{
 645	long nr;
 646	unsigned long args[6], sp, pc;
 647	int res;
 648
 649	res = lock_trace(task);
 650	if (res)
 651		return res;
 652
 653	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
 654		seq_puts(m, "running\n");
 655	else if (nr < 0)
 656		seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
 
 657	else
 658		seq_printf(m,
 659		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
 660		       nr,
 661		       args[0], args[1], args[2], args[3], args[4], args[5],
 662		       sp, pc);
 663	unlock_trace(task);
 664
 665	return 0;
 666}
 667#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
 668
 669/************************************************************************/
 670/*                       Here the fs part begins                        */
 671/************************************************************************/
 672
 673/* permission checks */
 674static int proc_fd_access_allowed(struct inode *inode)
 675{
 676	struct task_struct *task;
 677	int allowed = 0;
 678	/* Allow access to a task's file descriptors if it is us or we
 679	 * may use ptrace attach to the process and find out that
 680	 * information.
 681	 */
 682	task = get_proc_task(inode);
 683	if (task) {
 684		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 685		put_task_struct(task);
 686	}
 687	return allowed;
 688}
 689
 690int proc_setattr(struct dentry *dentry, struct iattr *attr)
 691{
 692	int error;
 693	struct inode *inode = d_inode(dentry);
 694
 695	if (attr->ia_valid & ATTR_MODE)
 696		return -EPERM;
 697
 698	error = setattr_prepare(dentry, attr);
 699	if (error)
 700		return error;
 701
 702	setattr_copy(inode, attr);
 703	mark_inode_dirty(inode);
 704	return 0;
 705}
 706
 707/*
 708 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
 709 * or euid/egid (for hide_pid_min=2)?
 710 */
 711static bool has_pid_permissions(struct pid_namespace *pid,
 712				 struct task_struct *task,
 713				 int hide_pid_min)
 714{
 715	if (pid->hide_pid < hide_pid_min)
 
 
 
 
 
 
 
 
 716		return true;
 717	if (in_group_p(pid->pid_gid))
 718		return true;
 719	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
 720}
 721
 722
 723static int proc_pid_permission(struct inode *inode, int mask)
 724{
 725	struct pid_namespace *pid = inode->i_sb->s_fs_info;
 726	struct task_struct *task;
 727	bool has_perms;
 728
 729	task = get_proc_task(inode);
 730	if (!task)
 731		return -ESRCH;
 732	has_perms = has_pid_permissions(pid, task, 1);
 733	put_task_struct(task);
 734
 735	if (!has_perms) {
 736		if (pid->hide_pid == 2) {
 737			/*
 738			 * Let's make getdents(), stat(), and open()
 739			 * consistent with each other.  If a process
 740			 * may not stat() a file, it shouldn't be seen
 741			 * in procfs at all.
 742			 */
 743			return -ENOENT;
 744		}
 745
 746		return -EPERM;
 747	}
 748	return generic_permission(inode, mask);
 749}
 750
 751
 752
 753static const struct inode_operations proc_def_inode_operations = {
 754	.setattr	= proc_setattr,
 755};
 756
 757static int proc_single_show(struct seq_file *m, void *v)
 758{
 759	struct inode *inode = m->private;
 760	struct pid_namespace *ns;
 761	struct pid *pid;
 762	struct task_struct *task;
 763	int ret;
 764
 765	ns = inode->i_sb->s_fs_info;
 766	pid = proc_pid(inode);
 767	task = get_pid_task(pid, PIDTYPE_PID);
 768	if (!task)
 769		return -ESRCH;
 770
 771	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
 772
 773	put_task_struct(task);
 774	return ret;
 775}
 776
 777static int proc_single_open(struct inode *inode, struct file *filp)
 778{
 779	return single_open(filp, proc_single_show, inode);
 780}
 781
 782static const struct file_operations proc_single_file_operations = {
 783	.open		= proc_single_open,
 784	.read		= seq_read,
 785	.llseek		= seq_lseek,
 786	.release	= single_release,
 787};
 788
 789
 790struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
 791{
 792	struct task_struct *task = get_proc_task(inode);
 793	struct mm_struct *mm = ERR_PTR(-ESRCH);
 794
 795	if (task) {
 796		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
 797		put_task_struct(task);
 798
 799		if (!IS_ERR_OR_NULL(mm)) {
 800			/* ensure this mm_struct can't be freed */
 801			atomic_inc(&mm->mm_count);
 802			/* but do not pin its memory */
 803			mmput(mm);
 804		}
 805	}
 806
 807	return mm;
 808}
 809
 810static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
 811{
 812	struct mm_struct *mm = proc_mem_open(inode, mode);
 813
 814	if (IS_ERR(mm))
 815		return PTR_ERR(mm);
 816
 817	file->private_data = mm;
 818	return 0;
 819}
 820
 821static int mem_open(struct inode *inode, struct file *file)
 822{
 823	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
 824
 825	/* OK to pass negative loff_t, we can catch out-of-range */
 826	file->f_mode |= FMODE_UNSIGNED_OFFSET;
 827
 828	return ret;
 829}
 830
 831static ssize_t mem_rw(struct file *file, char __user *buf,
 832			size_t count, loff_t *ppos, int write)
 833{
 834	struct mm_struct *mm = file->private_data;
 835	unsigned long addr = *ppos;
 836	ssize_t copied;
 837	char *page;
 838	unsigned int flags;
 839
 840	if (!mm)
 841		return 0;
 842
 843	page = (char *)__get_free_page(GFP_TEMPORARY);
 844	if (!page)
 845		return -ENOMEM;
 846
 847	copied = 0;
 848	if (!atomic_inc_not_zero(&mm->mm_users))
 849		goto free;
 850
 851	/* Maybe we should limit FOLL_FORCE to actual ptrace users? */
 852	flags = FOLL_FORCE;
 853	if (write)
 854		flags |= FOLL_WRITE;
 855
 856	while (count > 0) {
 857		int this_len = min_t(int, 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_TEMPORARY);
 952	if (!page)
 953		return -ENOMEM;
 954
 955	ret = 0;
 956	if (!atomic_inc_not_zero(&mm->mm_users))
 957		goto free;
 958
 959	down_read(&mm->mmap_sem);
 960	env_start = mm->env_start;
 961	env_end = mm->env_end;
 962	up_read(&mm->mmap_sem);
 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, 0);
 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	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1052	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1053}
1054
1055static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1056{
1057	static DEFINE_MUTEX(oom_adj_mutex);
1058	struct mm_struct *mm = NULL;
1059	struct task_struct *task;
1060	int err = 0;
1061
1062	task = get_proc_task(file_inode(file));
1063	if (!task)
1064		return -ESRCH;
1065
1066	mutex_lock(&oom_adj_mutex);
1067	if (legacy) {
1068		if (oom_adj < task->signal->oom_score_adj &&
1069				!capable(CAP_SYS_RESOURCE)) {
1070			err = -EACCES;
1071			goto err_unlock;
1072		}
1073		/*
1074		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1075		 * /proc/pid/oom_score_adj instead.
1076		 */
1077		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1078			  current->comm, task_pid_nr(current), task_pid_nr(task),
1079			  task_pid_nr(task));
1080	} else {
1081		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1082				!capable(CAP_SYS_RESOURCE)) {
1083			err = -EACCES;
1084			goto err_unlock;
1085		}
1086	}
1087
1088	/*
1089	 * Make sure we will check other processes sharing the mm if this is
1090	 * not vfrok which wants its own oom_score_adj.
1091	 * pin the mm so it doesn't go away and get reused after task_unlock
1092	 */
1093	if (!task->vfork_done) {
1094		struct task_struct *p = find_lock_task_mm(task);
1095
1096		if (p) {
1097			if (atomic_read(&p->mm->mm_users) > 1) {
1098				mm = p->mm;
1099				atomic_inc(&mm->mm_count);
1100			}
1101			task_unlock(p);
1102		}
1103	}
1104
1105	task->signal->oom_score_adj = oom_adj;
1106	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1107		task->signal->oom_score_adj_min = (short)oom_adj;
1108	trace_oom_score_adj_update(task);
1109
1110	if (mm) {
1111		struct task_struct *p;
1112
1113		rcu_read_lock();
1114		for_each_process(p) {
1115			if (same_thread_group(task, p))
1116				continue;
1117
1118			/* do not touch kernel threads or the global init */
1119			if (p->flags & PF_KTHREAD || is_global_init(p))
1120				continue;
1121
1122			task_lock(p);
1123			if (!p->vfork_done && process_shares_mm(p, mm)) {
1124				pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1125						task_pid_nr(p), p->comm,
1126						p->signal->oom_score_adj, oom_adj,
1127						task_pid_nr(task), task->comm);
1128				p->signal->oom_score_adj = oom_adj;
1129				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1130					p->signal->oom_score_adj_min = (short)oom_adj;
1131			}
1132			task_unlock(p);
1133		}
1134		rcu_read_unlock();
1135		mmdrop(mm);
1136	}
1137err_unlock:
1138	mutex_unlock(&oom_adj_mutex);
1139	put_task_struct(task);
1140	return err;
1141}
1142
1143/*
1144 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1145 * kernels.  The effective policy is defined by oom_score_adj, which has a
1146 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1147 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1148 * Processes that become oom disabled via oom_adj will still be oom disabled
1149 * with this implementation.
1150 *
1151 * oom_adj cannot be removed since existing userspace binaries use it.
1152 */
1153static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1154			     size_t count, loff_t *ppos)
1155{
1156	char buffer[PROC_NUMBUF];
1157	int oom_adj;
1158	int err;
1159
1160	memset(buffer, 0, sizeof(buffer));
1161	if (count > sizeof(buffer) - 1)
1162		count = sizeof(buffer) - 1;
1163	if (copy_from_user(buffer, buf, count)) {
1164		err = -EFAULT;
1165		goto out;
1166	}
1167
1168	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1169	if (err)
1170		goto out;
1171	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1172	     oom_adj != OOM_DISABLE) {
1173		err = -EINVAL;
1174		goto out;
1175	}
1176
1177	/*
1178	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1179	 * value is always attainable.
1180	 */
1181	if (oom_adj == OOM_ADJUST_MAX)
1182		oom_adj = OOM_SCORE_ADJ_MAX;
1183	else
1184		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1185
1186	err = __set_oom_adj(file, oom_adj, true);
1187out:
1188	return err < 0 ? err : count;
1189}
1190
1191static const struct file_operations proc_oom_adj_operations = {
1192	.read		= oom_adj_read,
1193	.write		= oom_adj_write,
1194	.llseek		= generic_file_llseek,
1195};
1196
1197static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1198					size_t count, loff_t *ppos)
1199{
1200	struct task_struct *task = get_proc_task(file_inode(file));
1201	char buffer[PROC_NUMBUF];
1202	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1203	size_t len;
1204
1205	if (!task)
1206		return -ESRCH;
1207	oom_score_adj = task->signal->oom_score_adj;
1208	put_task_struct(task);
1209	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1210	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1211}
1212
1213static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1214					size_t count, loff_t *ppos)
1215{
1216	char buffer[PROC_NUMBUF];
1217	int oom_score_adj;
1218	int err;
1219
1220	memset(buffer, 0, sizeof(buffer));
1221	if (count > sizeof(buffer) - 1)
1222		count = sizeof(buffer) - 1;
1223	if (copy_from_user(buffer, buf, count)) {
1224		err = -EFAULT;
1225		goto out;
1226	}
1227
1228	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1229	if (err)
1230		goto out;
1231	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1232			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1233		err = -EINVAL;
1234		goto out;
1235	}
1236
1237	err = __set_oom_adj(file, oom_score_adj, false);
1238out:
1239	return err < 0 ? err : count;
1240}
1241
1242static const struct file_operations proc_oom_score_adj_operations = {
1243	.read		= oom_score_adj_read,
1244	.write		= oom_score_adj_write,
1245	.llseek		= default_llseek,
1246};
1247
1248#ifdef CONFIG_AUDITSYSCALL
1249#define TMPBUFLEN 11
1250static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1251				  size_t count, loff_t *ppos)
1252{
1253	struct inode * inode = file_inode(file);
1254	struct task_struct *task = get_proc_task(inode);
1255	ssize_t length;
1256	char tmpbuf[TMPBUFLEN];
1257
1258	if (!task)
1259		return -ESRCH;
1260	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1261			   from_kuid(file->f_cred->user_ns,
1262				     audit_get_loginuid(task)));
1263	put_task_struct(task);
1264	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1265}
1266
1267static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1268				   size_t count, loff_t *ppos)
1269{
1270	struct inode * inode = file_inode(file);
1271	uid_t loginuid;
1272	kuid_t kloginuid;
1273	int rv;
1274
1275	rcu_read_lock();
1276	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1277		rcu_read_unlock();
1278		return -EPERM;
1279	}
1280	rcu_read_unlock();
1281
1282	if (*ppos != 0) {
1283		/* No partial writes. */
1284		return -EINVAL;
1285	}
1286
1287	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1288	if (rv < 0)
1289		return rv;
1290
1291	/* is userspace tring to explicitly UNSET the loginuid? */
1292	if (loginuid == AUDIT_UID_UNSET) {
1293		kloginuid = INVALID_UID;
1294	} else {
1295		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1296		if (!uid_valid(kloginuid))
1297			return -EINVAL;
1298	}
1299
1300	rv = audit_set_loginuid(kloginuid);
1301	if (rv < 0)
1302		return rv;
1303	return count;
1304}
1305
1306static const struct file_operations proc_loginuid_operations = {
1307	.read		= proc_loginuid_read,
1308	.write		= proc_loginuid_write,
1309	.llseek		= generic_file_llseek,
1310};
1311
1312static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1313				  size_t count, loff_t *ppos)
1314{
1315	struct inode * inode = file_inode(file);
1316	struct task_struct *task = get_proc_task(inode);
1317	ssize_t length;
1318	char tmpbuf[TMPBUFLEN];
1319
1320	if (!task)
1321		return -ESRCH;
1322	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1323				audit_get_sessionid(task));
1324	put_task_struct(task);
1325	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1326}
1327
1328static const struct file_operations proc_sessionid_operations = {
1329	.read		= proc_sessionid_read,
1330	.llseek		= generic_file_llseek,
1331};
1332#endif
1333
1334#ifdef CONFIG_FAULT_INJECTION
1335static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1336				      size_t count, loff_t *ppos)
1337{
1338	struct task_struct *task = get_proc_task(file_inode(file));
1339	char buffer[PROC_NUMBUF];
1340	size_t len;
1341	int make_it_fail;
1342
1343	if (!task)
1344		return -ESRCH;
1345	make_it_fail = task->make_it_fail;
1346	put_task_struct(task);
1347
1348	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1349
1350	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1351}
1352
1353static ssize_t proc_fault_inject_write(struct file * file,
1354			const char __user * buf, size_t count, loff_t *ppos)
1355{
1356	struct task_struct *task;
1357	char buffer[PROC_NUMBUF];
1358	int make_it_fail;
1359	int rv;
1360
1361	if (!capable(CAP_SYS_RESOURCE))
1362		return -EPERM;
1363	memset(buffer, 0, sizeof(buffer));
1364	if (count > sizeof(buffer) - 1)
1365		count = sizeof(buffer) - 1;
1366	if (copy_from_user(buffer, buf, count))
1367		return -EFAULT;
1368	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1369	if (rv < 0)
1370		return rv;
1371	if (make_it_fail < 0 || make_it_fail > 1)
1372		return -EINVAL;
1373
1374	task = get_proc_task(file_inode(file));
1375	if (!task)
1376		return -ESRCH;
1377	task->make_it_fail = make_it_fail;
1378	put_task_struct(task);
1379
1380	return count;
1381}
1382
1383static const struct file_operations proc_fault_inject_operations = {
1384	.read		= proc_fault_inject_read,
1385	.write		= proc_fault_inject_write,
1386	.llseek		= generic_file_llseek,
1387};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1388#endif
1389
1390
1391#ifdef CONFIG_SCHED_DEBUG
1392/*
1393 * Print out various scheduling related per-task fields:
1394 */
1395static int sched_show(struct seq_file *m, void *v)
1396{
1397	struct inode *inode = m->private;
 
1398	struct task_struct *p;
1399
1400	p = get_proc_task(inode);
1401	if (!p)
1402		return -ESRCH;
1403	proc_sched_show_task(p, m);
1404
1405	put_task_struct(p);
1406
1407	return 0;
1408}
1409
1410static ssize_t
1411sched_write(struct file *file, const char __user *buf,
1412	    size_t count, loff_t *offset)
1413{
1414	struct inode *inode = file_inode(file);
1415	struct task_struct *p;
1416
1417	p = get_proc_task(inode);
1418	if (!p)
1419		return -ESRCH;
1420	proc_sched_set_task(p);
1421
1422	put_task_struct(p);
1423
1424	return count;
1425}
1426
1427static int sched_open(struct inode *inode, struct file *filp)
1428{
1429	return single_open(filp, sched_show, inode);
1430}
1431
1432static const struct file_operations proc_pid_sched_operations = {
1433	.open		= sched_open,
1434	.read		= seq_read,
1435	.write		= sched_write,
1436	.llseek		= seq_lseek,
1437	.release	= single_release,
1438};
1439
1440#endif
1441
1442#ifdef CONFIG_SCHED_AUTOGROUP
1443/*
1444 * Print out autogroup related information:
1445 */
1446static int sched_autogroup_show(struct seq_file *m, void *v)
1447{
1448	struct inode *inode = m->private;
1449	struct task_struct *p;
1450
1451	p = get_proc_task(inode);
1452	if (!p)
1453		return -ESRCH;
1454	proc_sched_autogroup_show_task(p, m);
1455
1456	put_task_struct(p);
1457
1458	return 0;
1459}
1460
1461static ssize_t
1462sched_autogroup_write(struct file *file, const char __user *buf,
1463	    size_t count, loff_t *offset)
1464{
1465	struct inode *inode = file_inode(file);
1466	struct task_struct *p;
1467	char buffer[PROC_NUMBUF];
1468	int nice;
1469	int err;
1470
1471	memset(buffer, 0, sizeof(buffer));
1472	if (count > sizeof(buffer) - 1)
1473		count = sizeof(buffer) - 1;
1474	if (copy_from_user(buffer, buf, count))
1475		return -EFAULT;
1476
1477	err = kstrtoint(strstrip(buffer), 0, &nice);
1478	if (err < 0)
1479		return err;
1480
1481	p = get_proc_task(inode);
1482	if (!p)
1483		return -ESRCH;
1484
1485	err = proc_sched_autogroup_set_nice(p, nice);
1486	if (err)
1487		count = err;
1488
1489	put_task_struct(p);
1490
1491	return count;
1492}
1493
1494static int sched_autogroup_open(struct inode *inode, struct file *filp)
1495{
1496	int ret;
1497
1498	ret = single_open(filp, sched_autogroup_show, NULL);
1499	if (!ret) {
1500		struct seq_file *m = filp->private_data;
1501
1502		m->private = inode;
1503	}
1504	return ret;
1505}
1506
1507static const struct file_operations proc_pid_sched_autogroup_operations = {
1508	.open		= sched_autogroup_open,
1509	.read		= seq_read,
1510	.write		= sched_autogroup_write,
1511	.llseek		= seq_lseek,
1512	.release	= single_release,
1513};
1514
1515#endif /* CONFIG_SCHED_AUTOGROUP */
1516
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1517static ssize_t comm_write(struct file *file, const char __user *buf,
1518				size_t count, loff_t *offset)
1519{
1520	struct inode *inode = file_inode(file);
1521	struct task_struct *p;
1522	char buffer[TASK_COMM_LEN];
1523	const size_t maxlen = sizeof(buffer) - 1;
1524
1525	memset(buffer, 0, sizeof(buffer));
1526	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1527		return -EFAULT;
1528
1529	p = get_proc_task(inode);
1530	if (!p)
1531		return -ESRCH;
1532
1533	if (same_thread_group(current, p))
1534		set_task_comm(p, buffer);
1535	else
1536		count = -EINVAL;
1537
1538	put_task_struct(p);
1539
1540	return count;
1541}
1542
1543static int comm_show(struct seq_file *m, void *v)
1544{
1545	struct inode *inode = m->private;
1546	struct task_struct *p;
1547
1548	p = get_proc_task(inode);
1549	if (!p)
1550		return -ESRCH;
1551
1552	task_lock(p);
1553	seq_printf(m, "%s\n", p->comm);
1554	task_unlock(p);
1555
1556	put_task_struct(p);
1557
1558	return 0;
1559}
1560
1561static int comm_open(struct inode *inode, struct file *filp)
1562{
1563	return single_open(filp, comm_show, inode);
1564}
1565
1566static const struct file_operations proc_pid_set_comm_operations = {
1567	.open		= comm_open,
1568	.read		= seq_read,
1569	.write		= comm_write,
1570	.llseek		= seq_lseek,
1571	.release	= single_release,
1572};
1573
1574static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1575{
1576	struct task_struct *task;
1577	struct file *exe_file;
1578
1579	task = get_proc_task(d_inode(dentry));
1580	if (!task)
1581		return -ENOENT;
1582	exe_file = get_task_exe_file(task);
1583	put_task_struct(task);
1584	if (exe_file) {
1585		*exe_path = exe_file->f_path;
1586		path_get(&exe_file->f_path);
1587		fput(exe_file);
1588		return 0;
1589	} else
1590		return -ENOENT;
1591}
1592
1593static const char *proc_pid_get_link(struct dentry *dentry,
1594				     struct inode *inode,
1595				     struct delayed_call *done)
1596{
1597	struct path path;
1598	int error = -EACCES;
1599
1600	if (!dentry)
1601		return ERR_PTR(-ECHILD);
1602
1603	/* Are we allowed to snoop on the tasks file descriptors? */
1604	if (!proc_fd_access_allowed(inode))
1605		goto out;
1606
1607	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1608	if (error)
1609		goto out;
1610
1611	nd_jump_link(&path);
1612	return NULL;
1613out:
1614	return ERR_PTR(error);
1615}
1616
1617static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1618{
1619	char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1620	char *pathname;
1621	int len;
1622
1623	if (!tmp)
1624		return -ENOMEM;
1625
1626	pathname = d_path(path, tmp, PAGE_SIZE);
1627	len = PTR_ERR(pathname);
1628	if (IS_ERR(pathname))
1629		goto out;
1630	len = tmp + PAGE_SIZE - 1 - pathname;
1631
1632	if (len > buflen)
1633		len = buflen;
1634	if (copy_to_user(buffer, pathname, len))
1635		len = -EFAULT;
1636 out:
1637	free_page((unsigned long)tmp);
1638	return len;
1639}
1640
1641static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1642{
1643	int error = -EACCES;
1644	struct inode *inode = d_inode(dentry);
1645	struct path path;
1646
1647	/* Are we allowed to snoop on the tasks file descriptors? */
1648	if (!proc_fd_access_allowed(inode))
1649		goto out;
1650
1651	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1652	if (error)
1653		goto out;
1654
1655	error = do_proc_readlink(&path, buffer, buflen);
1656	path_put(&path);
1657out:
1658	return error;
1659}
1660
1661const struct inode_operations proc_pid_link_inode_operations = {
1662	.readlink	= proc_pid_readlink,
1663	.get_link	= proc_pid_get_link,
1664	.setattr	= proc_setattr,
1665};
1666
1667
1668/* building an inode */
1669
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1670struct inode *proc_pid_make_inode(struct super_block * sb,
1671				  struct task_struct *task, umode_t mode)
1672{
1673	struct inode * inode;
1674	struct proc_inode *ei;
1675	const struct cred *cred;
1676
1677	/* We need a new inode */
1678
1679	inode = new_inode(sb);
1680	if (!inode)
1681		goto out;
1682
1683	/* Common stuff */
1684	ei = PROC_I(inode);
1685	inode->i_mode = mode;
1686	inode->i_ino = get_next_ino();
1687	inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1688	inode->i_op = &proc_def_inode_operations;
1689
1690	/*
1691	 * grab the reference to task.
1692	 */
1693	ei->pid = get_task_pid(task, PIDTYPE_PID);
1694	if (!ei->pid)
1695		goto out_unlock;
1696
1697	if (task_dumpable(task)) {
1698		rcu_read_lock();
1699		cred = __task_cred(task);
1700		inode->i_uid = cred->euid;
1701		inode->i_gid = cred->egid;
1702		rcu_read_unlock();
1703	}
 
 
1704	security_task_to_inode(task, inode);
1705
1706out:
1707	return inode;
1708
1709out_unlock:
1710	iput(inode);
1711	return NULL;
1712}
1713
1714int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 
1715{
1716	struct inode *inode = d_inode(dentry);
 
1717	struct task_struct *task;
1718	const struct cred *cred;
1719	struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1720
1721	generic_fillattr(inode, stat);
1722
1723	rcu_read_lock();
1724	stat->uid = GLOBAL_ROOT_UID;
1725	stat->gid = GLOBAL_ROOT_GID;
 
1726	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1727	if (task) {
1728		if (!has_pid_permissions(pid, task, 2)) {
1729			rcu_read_unlock();
1730			/*
1731			 * This doesn't prevent learning whether PID exists,
1732			 * it only makes getattr() consistent with readdir().
1733			 */
1734			return -ENOENT;
1735		}
1736		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1737		    task_dumpable(task)) {
1738			cred = __task_cred(task);
1739			stat->uid = cred->euid;
1740			stat->gid = cred->egid;
1741		}
1742	}
1743	rcu_read_unlock();
1744	return 0;
1745}
1746
1747/* dentry stuff */
1748
1749/*
1750 *	Exceptional case: normally we are not allowed to unhash a busy
1751 * directory. In this case, however, we can do it - no aliasing problems
1752 * due to the way we treat inodes.
1753 *
 
 
 
 
 
 
 
1754 * Rewrite the inode's ownerships here because the owning task may have
1755 * performed a setuid(), etc.
1756 *
1757 * Before the /proc/pid/status file was created the only way to read
1758 * the effective uid of a /process was to stat /proc/pid.  Reading
1759 * /proc/pid/status is slow enough that procps and other packages
1760 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1761 * made this apply to all per process world readable and executable
1762 * directories.
1763 */
1764int pid_revalidate(struct dentry *dentry, unsigned int flags)
1765{
1766	struct inode *inode;
1767	struct task_struct *task;
1768	const struct cred *cred;
1769
1770	if (flags & LOOKUP_RCU)
1771		return -ECHILD;
1772
1773	inode = d_inode(dentry);
1774	task = get_proc_task(inode);
1775
1776	if (task) {
1777		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1778		    task_dumpable(task)) {
1779			rcu_read_lock();
1780			cred = __task_cred(task);
1781			inode->i_uid = cred->euid;
1782			inode->i_gid = cred->egid;
1783			rcu_read_unlock();
1784		} else {
1785			inode->i_uid = GLOBAL_ROOT_UID;
1786			inode->i_gid = GLOBAL_ROOT_GID;
1787		}
1788		inode->i_mode &= ~(S_ISUID | S_ISGID);
1789		security_task_to_inode(task, inode);
1790		put_task_struct(task);
1791		return 1;
1792	}
1793	return 0;
1794}
1795
1796static inline bool proc_inode_is_dead(struct inode *inode)
1797{
1798	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1799}
1800
1801int pid_delete_dentry(const struct dentry *dentry)
1802{
1803	/* Is the task we represent dead?
1804	 * If so, then don't put the dentry on the lru list,
1805	 * kill it immediately.
1806	 */
1807	return proc_inode_is_dead(d_inode(dentry));
1808}
1809
1810const struct dentry_operations pid_dentry_operations =
1811{
1812	.d_revalidate	= pid_revalidate,
1813	.d_delete	= pid_delete_dentry,
1814};
1815
1816/* Lookups */
1817
1818/*
1819 * Fill a directory entry.
1820 *
1821 * If possible create the dcache entry and derive our inode number and
1822 * file type from dcache entry.
1823 *
1824 * Since all of the proc inode numbers are dynamically generated, the inode
1825 * numbers do not exist until the inode is cache.  This means creating the
1826 * the dcache entry in readdir is necessary to keep the inode numbers
1827 * reported by readdir in sync with the inode numbers reported
1828 * by stat.
1829 */
1830bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1831	const char *name, int len,
1832	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1833{
1834	struct dentry *child, *dir = file->f_path.dentry;
1835	struct qstr qname = QSTR_INIT(name, len);
1836	struct inode *inode;
1837	unsigned type;
1838	ino_t ino;
1839
1840	child = d_hash_and_lookup(dir, &qname);
1841	if (!child) {
1842		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1843		child = d_alloc_parallel(dir, &qname, &wq);
1844		if (IS_ERR(child))
1845			goto end_instantiate;
1846		if (d_in_lookup(child)) {
1847			int err = instantiate(d_inode(dir), child, task, ptr);
 
1848			d_lookup_done(child);
1849			if (err < 0) {
1850				dput(child);
1851				goto end_instantiate;
 
 
1852			}
1853		}
1854	}
1855	inode = d_inode(child);
1856	ino = inode->i_ino;
1857	type = inode->i_mode >> 12;
1858	dput(child);
1859	return dir_emit(ctx, name, len, ino, type);
1860
1861end_instantiate:
1862	return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1863}
1864
1865/*
1866 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1867 * which represent vma start and end addresses.
1868 */
1869static int dname_to_vma_addr(struct dentry *dentry,
1870			     unsigned long *start, unsigned long *end)
1871{
1872	if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
 
 
 
 
 
 
 
 
 
1873		return -EINVAL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1874
1875	return 0;
1876}
1877
1878static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1879{
1880	unsigned long vm_start, vm_end;
1881	bool exact_vma_exists = false;
1882	struct mm_struct *mm = NULL;
1883	struct task_struct *task;
1884	const struct cred *cred;
1885	struct inode *inode;
1886	int status = 0;
1887
1888	if (flags & LOOKUP_RCU)
1889		return -ECHILD;
1890
1891	inode = d_inode(dentry);
1892	task = get_proc_task(inode);
1893	if (!task)
1894		goto out_notask;
1895
1896	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
1897	if (IS_ERR_OR_NULL(mm))
1898		goto out;
1899
1900	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1901		down_read(&mm->mmap_sem);
1902		exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1903		up_read(&mm->mmap_sem);
 
 
 
1904	}
1905
1906	mmput(mm);
1907
1908	if (exact_vma_exists) {
1909		if (task_dumpable(task)) {
1910			rcu_read_lock();
1911			cred = __task_cred(task);
1912			inode->i_uid = cred->euid;
1913			inode->i_gid = cred->egid;
1914			rcu_read_unlock();
1915		} else {
1916			inode->i_uid = GLOBAL_ROOT_UID;
1917			inode->i_gid = GLOBAL_ROOT_GID;
1918		}
1919		security_task_to_inode(task, inode);
1920		status = 1;
1921	}
1922
1923out:
1924	put_task_struct(task);
1925
1926out_notask:
1927	return status;
1928}
1929
1930static const struct dentry_operations tid_map_files_dentry_operations = {
1931	.d_revalidate	= map_files_d_revalidate,
1932	.d_delete	= pid_delete_dentry,
1933};
1934
1935static int map_files_get_link(struct dentry *dentry, struct path *path)
1936{
1937	unsigned long vm_start, vm_end;
1938	struct vm_area_struct *vma;
1939	struct task_struct *task;
1940	struct mm_struct *mm;
1941	int rc;
1942
1943	rc = -ENOENT;
1944	task = get_proc_task(d_inode(dentry));
1945	if (!task)
1946		goto out;
1947
1948	mm = get_task_mm(task);
1949	put_task_struct(task);
1950	if (!mm)
1951		goto out;
1952
1953	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1954	if (rc)
1955		goto out_mmput;
1956
 
 
 
 
1957	rc = -ENOENT;
1958	down_read(&mm->mmap_sem);
1959	vma = find_exact_vma(mm, vm_start, vm_end);
1960	if (vma && vma->vm_file) {
1961		*path = vma->vm_file->f_path;
1962		path_get(path);
1963		rc = 0;
1964	}
1965	up_read(&mm->mmap_sem);
1966
1967out_mmput:
1968	mmput(mm);
1969out:
1970	return rc;
1971}
1972
1973struct map_files_info {
 
 
1974	fmode_t		mode;
1975	unsigned int	len;
1976	unsigned char	name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1977};
1978
1979/*
1980 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1981 * symlinks may be used to bypass permissions on ancestor directories in the
1982 * path to the file in question.
1983 */
1984static const char *
1985proc_map_files_get_link(struct dentry *dentry,
1986			struct inode *inode,
1987		        struct delayed_call *done)
1988{
1989	if (!capable(CAP_SYS_ADMIN))
1990		return ERR_PTR(-EPERM);
1991
1992	return proc_pid_get_link(dentry, inode, done);
1993}
1994
1995/*
1996 * Identical to proc_pid_link_inode_operations except for get_link()
1997 */
1998static const struct inode_operations proc_map_files_link_inode_operations = {
1999	.readlink	= proc_pid_readlink,
2000	.get_link	= proc_map_files_get_link,
2001	.setattr	= proc_setattr,
2002};
2003
2004static int
2005proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2006			   struct task_struct *task, const void *ptr)
2007{
2008	fmode_t mode = (fmode_t)(unsigned long)ptr;
2009	struct proc_inode *ei;
2010	struct inode *inode;
2011
2012	inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2013				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2014				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2015	if (!inode)
2016		return -ENOENT;
2017
2018	ei = PROC_I(inode);
2019	ei->op.proc_get_link = map_files_get_link;
2020
2021	inode->i_op = &proc_map_files_link_inode_operations;
2022	inode->i_size = 64;
2023
2024	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2025	d_add(dentry, inode);
2026
2027	return 0;
2028}
2029
2030static struct dentry *proc_map_files_lookup(struct inode *dir,
2031		struct dentry *dentry, unsigned int flags)
2032{
2033	unsigned long vm_start, vm_end;
2034	struct vm_area_struct *vma;
2035	struct task_struct *task;
2036	int result;
2037	struct mm_struct *mm;
2038
2039	result = -ENOENT;
2040	task = get_proc_task(dir);
2041	if (!task)
2042		goto out;
2043
2044	result = -EACCES;
2045	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2046		goto out_put_task;
2047
2048	result = -ENOENT;
2049	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2050		goto out_put_task;
2051
2052	mm = get_task_mm(task);
2053	if (!mm)
2054		goto out_put_task;
2055
2056	down_read(&mm->mmap_sem);
 
 
 
 
2057	vma = find_exact_vma(mm, vm_start, vm_end);
2058	if (!vma)
2059		goto out_no_vma;
2060
2061	if (vma->vm_file)
2062		result = proc_map_files_instantiate(dir, dentry, task,
2063				(void *)(unsigned long)vma->vm_file->f_mode);
2064
2065out_no_vma:
2066	up_read(&mm->mmap_sem);
 
2067	mmput(mm);
2068out_put_task:
2069	put_task_struct(task);
2070out:
2071	return ERR_PTR(result);
2072}
2073
2074static const struct inode_operations proc_map_files_inode_operations = {
2075	.lookup		= proc_map_files_lookup,
2076	.permission	= proc_fd_permission,
2077	.setattr	= proc_setattr,
2078};
2079
2080static int
2081proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2082{
2083	struct vm_area_struct *vma;
2084	struct task_struct *task;
2085	struct mm_struct *mm;
2086	unsigned long nr_files, pos, i;
2087	struct flex_array *fa = NULL;
2088	struct map_files_info info;
2089	struct map_files_info *p;
2090	int ret;
2091
 
 
2092	ret = -ENOENT;
2093	task = get_proc_task(file_inode(file));
2094	if (!task)
2095		goto out;
2096
2097	ret = -EACCES;
2098	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2099		goto out_put_task;
2100
2101	ret = 0;
2102	if (!dir_emit_dots(file, ctx))
2103		goto out_put_task;
2104
2105	mm = get_task_mm(task);
2106	if (!mm)
2107		goto out_put_task;
2108	down_read(&mm->mmap_sem);
 
 
 
 
 
2109
2110	nr_files = 0;
2111
2112	/*
2113	 * We need two passes here:
2114	 *
2115	 *  1) Collect vmas of mapped files with mmap_sem taken
2116	 *  2) Release mmap_sem and instantiate entries
2117	 *
2118	 * otherwise we get lockdep complained, since filldir()
2119	 * routine might require mmap_sem taken in might_fault().
2120	 */
2121
2122	for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2123		if (vma->vm_file && ++pos > ctx->pos)
2124			nr_files++;
2125	}
 
2126
2127	if (nr_files) {
2128		fa = flex_array_alloc(sizeof(info), nr_files,
2129					GFP_KERNEL);
2130		if (!fa || flex_array_prealloc(fa, 0, nr_files,
2131						GFP_KERNEL)) {
2132			ret = -ENOMEM;
2133			if (fa)
2134				flex_array_free(fa);
2135			up_read(&mm->mmap_sem);
2136			mmput(mm);
2137			goto out_put_task;
2138		}
2139		for (i = 0, vma = mm->mmap, pos = 2; vma;
2140				vma = vma->vm_next) {
2141			if (!vma->vm_file)
2142				continue;
2143			if (++pos <= ctx->pos)
2144				continue;
2145
2146			info.mode = vma->vm_file->f_mode;
2147			info.len = snprintf(info.name,
2148					sizeof(info.name), "%lx-%lx",
2149					vma->vm_start, vma->vm_end);
2150			if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2151				BUG();
2152		}
2153	}
2154	up_read(&mm->mmap_sem);
 
2155
2156	for (i = 0; i < nr_files; i++) {
2157		p = flex_array_get(fa, i);
 
 
 
 
2158		if (!proc_fill_cache(file, ctx,
2159				      p->name, p->len,
2160				      proc_map_files_instantiate,
2161				      task,
2162				      (void *)(unsigned long)p->mode))
2163			break;
2164		ctx->pos++;
2165	}
2166	if (fa)
2167		flex_array_free(fa);
2168	mmput(mm);
2169
2170out_put_task:
2171	put_task_struct(task);
2172out:
 
2173	return ret;
2174}
2175
2176static const struct file_operations proc_map_files_operations = {
2177	.read		= generic_read_dir,
2178	.iterate_shared	= proc_map_files_readdir,
2179	.llseek		= generic_file_llseek,
2180};
2181
2182#ifdef CONFIG_CHECKPOINT_RESTORE
2183struct timers_private {
2184	struct pid *pid;
2185	struct task_struct *task;
2186	struct sighand_struct *sighand;
2187	struct pid_namespace *ns;
2188	unsigned long flags;
2189};
2190
2191static void *timers_start(struct seq_file *m, loff_t *pos)
2192{
2193	struct timers_private *tp = m->private;
2194
2195	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2196	if (!tp->task)
2197		return ERR_PTR(-ESRCH);
2198
2199	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2200	if (!tp->sighand)
2201		return ERR_PTR(-ESRCH);
2202
2203	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2204}
2205
2206static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2207{
2208	struct timers_private *tp = m->private;
2209	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2210}
2211
2212static void timers_stop(struct seq_file *m, void *v)
2213{
2214	struct timers_private *tp = m->private;
2215
2216	if (tp->sighand) {
2217		unlock_task_sighand(tp->task, &tp->flags);
2218		tp->sighand = NULL;
2219	}
2220
2221	if (tp->task) {
2222		put_task_struct(tp->task);
2223		tp->task = NULL;
2224	}
2225}
2226
2227static int show_timer(struct seq_file *m, void *v)
2228{
2229	struct k_itimer *timer;
2230	struct timers_private *tp = m->private;
2231	int notify;
2232	static const char * const nstr[] = {
2233		[SIGEV_SIGNAL] = "signal",
2234		[SIGEV_NONE] = "none",
2235		[SIGEV_THREAD] = "thread",
2236	};
2237
2238	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2239	notify = timer->it_sigev_notify;
2240
2241	seq_printf(m, "ID: %d\n", timer->it_id);
2242	seq_printf(m, "signal: %d/%p\n",
2243		   timer->sigq->info.si_signo,
2244		   timer->sigq->info.si_value.sival_ptr);
2245	seq_printf(m, "notify: %s/%s.%d\n",
2246		   nstr[notify & ~SIGEV_THREAD_ID],
2247		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2248		   pid_nr_ns(timer->it_pid, tp->ns));
2249	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2250
2251	return 0;
2252}
2253
2254static const struct seq_operations proc_timers_seq_ops = {
2255	.start	= timers_start,
2256	.next	= timers_next,
2257	.stop	= timers_stop,
2258	.show	= show_timer,
2259};
2260
2261static int proc_timers_open(struct inode *inode, struct file *file)
2262{
2263	struct timers_private *tp;
2264
2265	tp = __seq_open_private(file, &proc_timers_seq_ops,
2266			sizeof(struct timers_private));
2267	if (!tp)
2268		return -ENOMEM;
2269
2270	tp->pid = proc_pid(inode);
2271	tp->ns = inode->i_sb->s_fs_info;
2272	return 0;
2273}
2274
2275static const struct file_operations proc_timers_operations = {
2276	.open		= proc_timers_open,
2277	.read		= seq_read,
2278	.llseek		= seq_lseek,
2279	.release	= seq_release_private,
2280};
2281#endif
2282
2283static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2284					size_t count, loff_t *offset)
2285{
2286	struct inode *inode = file_inode(file);
2287	struct task_struct *p;
2288	u64 slack_ns;
2289	int err;
2290
2291	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2292	if (err < 0)
2293		return err;
2294
2295	p = get_proc_task(inode);
2296	if (!p)
2297		return -ESRCH;
2298
2299	if (p != current) {
2300		if (!capable(CAP_SYS_NICE)) {
 
 
2301			count = -EPERM;
2302			goto out;
2303		}
 
2304
2305		err = security_task_setscheduler(p);
2306		if (err) {
2307			count = err;
2308			goto out;
2309		}
2310	}
2311
2312	task_lock(p);
2313	if (slack_ns == 0)
2314		p->timer_slack_ns = p->default_timer_slack_ns;
2315	else
2316		p->timer_slack_ns = slack_ns;
2317	task_unlock(p);
2318
2319out:
2320	put_task_struct(p);
2321
2322	return count;
2323}
2324
2325static int timerslack_ns_show(struct seq_file *m, void *v)
2326{
2327	struct inode *inode = m->private;
2328	struct task_struct *p;
2329	int err = 0;
2330
2331	p = get_proc_task(inode);
2332	if (!p)
2333		return -ESRCH;
2334
2335	if (p != current) {
2336
2337		if (!capable(CAP_SYS_NICE)) {
 
2338			err = -EPERM;
2339			goto out;
2340		}
 
 
2341		err = security_task_getscheduler(p);
2342		if (err)
2343			goto out;
2344	}
2345
2346	task_lock(p);
2347	seq_printf(m, "%llu\n", p->timer_slack_ns);
2348	task_unlock(p);
2349
2350out:
2351	put_task_struct(p);
2352
2353	return err;
2354}
2355
2356static int timerslack_ns_open(struct inode *inode, struct file *filp)
2357{
2358	return single_open(filp, timerslack_ns_show, inode);
2359}
2360
2361static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2362	.open		= timerslack_ns_open,
2363	.read		= seq_read,
2364	.write		= timerslack_ns_write,
2365	.llseek		= seq_lseek,
2366	.release	= single_release,
2367};
2368
2369static int proc_pident_instantiate(struct inode *dir,
2370	struct dentry *dentry, struct task_struct *task, const void *ptr)
2371{
2372	const struct pid_entry *p = ptr;
2373	struct inode *inode;
2374	struct proc_inode *ei;
2375
2376	inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
2377	if (!inode)
2378		goto out;
2379
2380	ei = PROC_I(inode);
2381	if (S_ISDIR(inode->i_mode))
2382		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2383	if (p->iop)
2384		inode->i_op = p->iop;
2385	if (p->fop)
2386		inode->i_fop = p->fop;
2387	ei->op = p->op;
 
2388	d_set_d_op(dentry, &pid_dentry_operations);
2389	d_add(dentry, inode);
2390	/* Close the race of the process dying before we return the dentry */
2391	if (pid_revalidate(dentry, 0))
2392		return 0;
2393out:
2394	return -ENOENT;
2395}
2396
2397static struct dentry *proc_pident_lookup(struct inode *dir, 
2398					 struct dentry *dentry,
2399					 const struct pid_entry *ents,
2400					 unsigned int nents)
2401{
2402	int error;
2403	struct task_struct *task = get_proc_task(dir);
2404	const struct pid_entry *p, *last;
2405
2406	error = -ENOENT;
2407
2408	if (!task)
2409		goto out_no_task;
2410
2411	/*
2412	 * Yes, it does not scale. And it should not. Don't add
2413	 * new entries into /proc/<tgid>/ without very good reasons.
2414	 */
2415	last = &ents[nents];
2416	for (p = ents; p < last; p++) {
2417		if (p->len != dentry->d_name.len)
2418			continue;
2419		if (!memcmp(dentry->d_name.name, p->name, p->len))
 
2420			break;
 
2421	}
2422	if (p >= last)
2423		goto out;
2424
2425	error = proc_pident_instantiate(dir, dentry, task, p);
2426out:
2427	put_task_struct(task);
2428out_no_task:
2429	return ERR_PTR(error);
2430}
2431
2432static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2433		const struct pid_entry *ents, unsigned int nents)
2434{
2435	struct task_struct *task = get_proc_task(file_inode(file));
2436	const struct pid_entry *p;
2437
2438	if (!task)
2439		return -ENOENT;
2440
2441	if (!dir_emit_dots(file, ctx))
2442		goto out;
2443
2444	if (ctx->pos >= nents + 2)
2445		goto out;
2446
2447	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2448		if (!proc_fill_cache(file, ctx, p->name, p->len,
2449				proc_pident_instantiate, task, p))
2450			break;
2451		ctx->pos++;
2452	}
2453out:
2454	put_task_struct(task);
2455	return 0;
2456}
2457
2458#ifdef CONFIG_SECURITY
2459static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2460				  size_t count, loff_t *ppos)
2461{
2462	struct inode * inode = file_inode(file);
2463	char *p = NULL;
2464	ssize_t length;
2465	struct task_struct *task = get_proc_task(inode);
2466
2467	if (!task)
2468		return -ESRCH;
2469
2470	length = security_getprocattr(task,
2471				      (char*)file->f_path.dentry->d_name.name,
2472				      &p);
2473	put_task_struct(task);
2474	if (length > 0)
2475		length = simple_read_from_buffer(buf, count, ppos, p, length);
2476	kfree(p);
2477	return length;
2478}
2479
2480static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2481				   size_t count, loff_t *ppos)
2482{
2483	struct inode * inode = file_inode(file);
 
2484	void *page;
2485	ssize_t length;
2486	struct task_struct *task = get_proc_task(inode);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2487
2488	length = -ESRCH;
2489	if (!task)
2490		goto out_no_task;
2491	if (count > PAGE_SIZE)
2492		count = PAGE_SIZE;
2493
2494	/* No partial writes. */
2495	length = -EINVAL;
2496	if (*ppos != 0)
2497		goto out;
2498
2499	page = memdup_user(buf, count);
2500	if (IS_ERR(page)) {
2501		length = PTR_ERR(page);
2502		goto out;
2503	}
2504
2505	/* Guard against adverse ptrace interaction */
2506	length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2507	if (length < 0)
2508		goto out_free;
2509
2510	length = security_setprocattr(task,
2511				      (char*)file->f_path.dentry->d_name.name,
2512				      page, count);
2513	mutex_unlock(&task->signal->cred_guard_mutex);
2514out_free:
2515	kfree(page);
2516out:
2517	put_task_struct(task);
2518out_no_task:
2519	return length;
2520}
2521
2522static const struct file_operations proc_pid_attr_operations = {
2523	.read		= proc_pid_attr_read,
2524	.write		= proc_pid_attr_write,
2525	.llseek		= generic_file_llseek,
2526};
2527
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2528static const struct pid_entry attr_dir_stuff[] = {
2529	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2530	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2531	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2532	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2533	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2534	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
 
 
 
 
 
 
 
 
2535};
2536
2537static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2538{
2539	return proc_pident_readdir(file, ctx, 
2540				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2541}
2542
2543static const struct file_operations proc_attr_dir_operations = {
2544	.read		= generic_read_dir,
2545	.iterate_shared	= proc_attr_dir_readdir,
2546	.llseek		= generic_file_llseek,
2547};
2548
2549static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2550				struct dentry *dentry, unsigned int flags)
2551{
2552	return proc_pident_lookup(dir, dentry,
2553				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
 
2554}
2555
2556static const struct inode_operations proc_attr_dir_inode_operations = {
2557	.lookup		= proc_attr_dir_lookup,
2558	.getattr	= pid_getattr,
2559	.setattr	= proc_setattr,
2560};
2561
2562#endif
2563
2564#ifdef CONFIG_ELF_CORE
2565static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2566					 size_t count, loff_t *ppos)
2567{
2568	struct task_struct *task = get_proc_task(file_inode(file));
2569	struct mm_struct *mm;
2570	char buffer[PROC_NUMBUF];
2571	size_t len;
2572	int ret;
2573
2574	if (!task)
2575		return -ESRCH;
2576
2577	ret = 0;
2578	mm = get_task_mm(task);
2579	if (mm) {
2580		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2581			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2582				MMF_DUMP_FILTER_SHIFT));
2583		mmput(mm);
2584		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2585	}
2586
2587	put_task_struct(task);
2588
2589	return ret;
2590}
2591
2592static ssize_t proc_coredump_filter_write(struct file *file,
2593					  const char __user *buf,
2594					  size_t count,
2595					  loff_t *ppos)
2596{
2597	struct task_struct *task;
2598	struct mm_struct *mm;
2599	unsigned int val;
2600	int ret;
2601	int i;
2602	unsigned long mask;
2603
2604	ret = kstrtouint_from_user(buf, count, 0, &val);
2605	if (ret < 0)
2606		return ret;
2607
2608	ret = -ESRCH;
2609	task = get_proc_task(file_inode(file));
2610	if (!task)
2611		goto out_no_task;
2612
2613	mm = get_task_mm(task);
2614	if (!mm)
2615		goto out_no_mm;
2616	ret = 0;
2617
2618	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2619		if (val & mask)
2620			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2621		else
2622			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2623	}
2624
2625	mmput(mm);
2626 out_no_mm:
2627	put_task_struct(task);
2628 out_no_task:
2629	if (ret < 0)
2630		return ret;
2631	return count;
2632}
2633
2634static const struct file_operations proc_coredump_filter_operations = {
2635	.read		= proc_coredump_filter_read,
2636	.write		= proc_coredump_filter_write,
2637	.llseek		= generic_file_llseek,
2638};
2639#endif
2640
2641#ifdef CONFIG_TASK_IO_ACCOUNTING
2642static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2643{
2644	struct task_io_accounting acct = task->ioac;
2645	unsigned long flags;
2646	int result;
2647
2648	result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2649	if (result)
2650		return result;
2651
2652	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2653		result = -EACCES;
2654		goto out_unlock;
2655	}
2656
2657	if (whole && lock_task_sighand(task, &flags)) {
2658		struct task_struct *t = task;
2659
2660		task_io_accounting_add(&acct, &task->signal->ioac);
2661		while_each_thread(task, t)
2662			task_io_accounting_add(&acct, &t->ioac);
2663
2664		unlock_task_sighand(task, &flags);
2665	}
2666	seq_printf(m,
2667		   "rchar: %llu\n"
2668		   "wchar: %llu\n"
2669		   "syscr: %llu\n"
2670		   "syscw: %llu\n"
2671		   "read_bytes: %llu\n"
2672		   "write_bytes: %llu\n"
2673		   "cancelled_write_bytes: %llu\n",
2674		   (unsigned long long)acct.rchar,
2675		   (unsigned long long)acct.wchar,
2676		   (unsigned long long)acct.syscr,
2677		   (unsigned long long)acct.syscw,
2678		   (unsigned long long)acct.read_bytes,
2679		   (unsigned long long)acct.write_bytes,
2680		   (unsigned long long)acct.cancelled_write_bytes);
2681	result = 0;
2682
2683out_unlock:
2684	mutex_unlock(&task->signal->cred_guard_mutex);
2685	return result;
2686}
2687
2688static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2689				  struct pid *pid, struct task_struct *task)
2690{
2691	return do_io_accounting(task, m, 0);
2692}
2693
2694static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2695				   struct pid *pid, struct task_struct *task)
2696{
2697	return do_io_accounting(task, m, 1);
2698}
2699#endif /* CONFIG_TASK_IO_ACCOUNTING */
2700
2701#ifdef CONFIG_USER_NS
2702static int proc_id_map_open(struct inode *inode, struct file *file,
2703	const struct seq_operations *seq_ops)
2704{
2705	struct user_namespace *ns = NULL;
2706	struct task_struct *task;
2707	struct seq_file *seq;
2708	int ret = -EINVAL;
2709
2710	task = get_proc_task(inode);
2711	if (task) {
2712		rcu_read_lock();
2713		ns = get_user_ns(task_cred_xxx(task, user_ns));
2714		rcu_read_unlock();
2715		put_task_struct(task);
2716	}
2717	if (!ns)
2718		goto err;
2719
2720	ret = seq_open(file, seq_ops);
2721	if (ret)
2722		goto err_put_ns;
2723
2724	seq = file->private_data;
2725	seq->private = ns;
2726
2727	return 0;
2728err_put_ns:
2729	put_user_ns(ns);
2730err:
2731	return ret;
2732}
2733
2734static int proc_id_map_release(struct inode *inode, struct file *file)
2735{
2736	struct seq_file *seq = file->private_data;
2737	struct user_namespace *ns = seq->private;
2738	put_user_ns(ns);
2739	return seq_release(inode, file);
2740}
2741
2742static int proc_uid_map_open(struct inode *inode, struct file *file)
2743{
2744	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2745}
2746
2747static int proc_gid_map_open(struct inode *inode, struct file *file)
2748{
2749	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2750}
2751
2752static int proc_projid_map_open(struct inode *inode, struct file *file)
2753{
2754	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2755}
2756
2757static const struct file_operations proc_uid_map_operations = {
2758	.open		= proc_uid_map_open,
2759	.write		= proc_uid_map_write,
2760	.read		= seq_read,
2761	.llseek		= seq_lseek,
2762	.release	= proc_id_map_release,
2763};
2764
2765static const struct file_operations proc_gid_map_operations = {
2766	.open		= proc_gid_map_open,
2767	.write		= proc_gid_map_write,
2768	.read		= seq_read,
2769	.llseek		= seq_lseek,
2770	.release	= proc_id_map_release,
2771};
2772
2773static const struct file_operations proc_projid_map_operations = {
2774	.open		= proc_projid_map_open,
2775	.write		= proc_projid_map_write,
2776	.read		= seq_read,
2777	.llseek		= seq_lseek,
2778	.release	= proc_id_map_release,
2779};
2780
2781static int proc_setgroups_open(struct inode *inode, struct file *file)
2782{
2783	struct user_namespace *ns = NULL;
2784	struct task_struct *task;
2785	int ret;
2786
2787	ret = -ESRCH;
2788	task = get_proc_task(inode);
2789	if (task) {
2790		rcu_read_lock();
2791		ns = get_user_ns(task_cred_xxx(task, user_ns));
2792		rcu_read_unlock();
2793		put_task_struct(task);
2794	}
2795	if (!ns)
2796		goto err;
2797
2798	if (file->f_mode & FMODE_WRITE) {
2799		ret = -EACCES;
2800		if (!ns_capable(ns, CAP_SYS_ADMIN))
2801			goto err_put_ns;
2802	}
2803
2804	ret = single_open(file, &proc_setgroups_show, ns);
2805	if (ret)
2806		goto err_put_ns;
2807
2808	return 0;
2809err_put_ns:
2810	put_user_ns(ns);
2811err:
2812	return ret;
2813}
2814
2815static int proc_setgroups_release(struct inode *inode, struct file *file)
2816{
2817	struct seq_file *seq = file->private_data;
2818	struct user_namespace *ns = seq->private;
2819	int ret = single_release(inode, file);
2820	put_user_ns(ns);
2821	return ret;
2822}
2823
2824static const struct file_operations proc_setgroups_operations = {
2825	.open		= proc_setgroups_open,
2826	.write		= proc_setgroups_write,
2827	.read		= seq_read,
2828	.llseek		= seq_lseek,
2829	.release	= proc_setgroups_release,
2830};
2831#endif /* CONFIG_USER_NS */
2832
2833static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2834				struct pid *pid, struct task_struct *task)
2835{
2836	int err = lock_trace(task);
2837	if (!err) {
2838		seq_printf(m, "%08x\n", task->personality);
2839		unlock_trace(task);
2840	}
2841	return err;
2842}
2843
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2844/*
2845 * Thread groups
2846 */
2847static const struct file_operations proc_task_operations;
2848static const struct inode_operations proc_task_inode_operations;
2849
2850static const struct pid_entry tgid_base_stuff[] = {
2851	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2852	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2853	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2854	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2855	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2856#ifdef CONFIG_NET
2857	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2858#endif
2859	REG("environ",    S_IRUSR, proc_environ_operations),
2860	REG("auxv",       S_IRUSR, proc_auxv_operations),
2861	ONE("status",     S_IRUGO, proc_pid_status),
2862	ONE("personality", S_IRUSR, proc_pid_personality),
2863	ONE("limits",	  S_IRUGO, proc_pid_limits),
2864#ifdef CONFIG_SCHED_DEBUG
2865	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2866#endif
2867#ifdef CONFIG_SCHED_AUTOGROUP
2868	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2869#endif
 
 
 
2870	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2871#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2872	ONE("syscall",    S_IRUSR, proc_pid_syscall),
2873#endif
2874	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
2875	ONE("stat",       S_IRUGO, proc_tgid_stat),
2876	ONE("statm",      S_IRUGO, proc_pid_statm),
2877	REG("maps",       S_IRUGO, proc_pid_maps_operations),
2878#ifdef CONFIG_NUMA
2879	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
2880#endif
2881	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2882	LNK("cwd",        proc_cwd_link),
2883	LNK("root",       proc_root_link),
2884	LNK("exe",        proc_exe_link),
2885	REG("mounts",     S_IRUGO, proc_mounts_operations),
2886	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2887	REG("mountstats", S_IRUSR, proc_mountstats_operations),
2888#ifdef CONFIG_PROC_PAGE_MONITOR
2889	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2890	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
 
2891	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
2892#endif
2893#ifdef CONFIG_SECURITY
2894	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2895#endif
2896#ifdef CONFIG_KALLSYMS
2897	ONE("wchan",      S_IRUGO, proc_pid_wchan),
2898#endif
2899#ifdef CONFIG_STACKTRACE
2900	ONE("stack",      S_IRUSR, proc_pid_stack),
2901#endif
2902#ifdef CONFIG_SCHED_INFO
2903	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
2904#endif
2905#ifdef CONFIG_LATENCYTOP
2906	REG("latency",  S_IRUGO, proc_lstats_operations),
2907#endif
2908#ifdef CONFIG_PROC_PID_CPUSET
2909	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
2910#endif
2911#ifdef CONFIG_CGROUPS
2912	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
2913#endif
 
 
 
2914	ONE("oom_score",  S_IRUGO, proc_oom_score),
2915	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2916	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2917#ifdef CONFIG_AUDITSYSCALL
2918	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2919	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2920#endif
2921#ifdef CONFIG_FAULT_INJECTION
2922	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
 
2923#endif
2924#ifdef CONFIG_ELF_CORE
2925	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2926#endif
2927#ifdef CONFIG_TASK_IO_ACCOUNTING
2928	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
2929#endif
2930#ifdef CONFIG_HARDWALL
2931	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
2932#endif
2933#ifdef CONFIG_USER_NS
2934	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
2935	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
2936	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2937	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
2938#endif
2939#ifdef CONFIG_CHECKPOINT_RESTORE
2940	REG("timers",	  S_IRUGO, proc_timers_operations),
2941#endif
2942	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
 
 
 
 
 
 
 
 
 
2943};
2944
2945static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2946{
2947	return proc_pident_readdir(file, ctx,
2948				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2949}
2950
2951static const struct file_operations proc_tgid_base_operations = {
2952	.read		= generic_read_dir,
2953	.iterate_shared	= proc_tgid_base_readdir,
2954	.llseek		= generic_file_llseek,
2955};
2956
 
 
 
 
 
 
 
 
2957static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2958{
2959	return proc_pident_lookup(dir, dentry,
2960				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
 
2961}
2962
2963static const struct inode_operations proc_tgid_base_inode_operations = {
2964	.lookup		= proc_tgid_base_lookup,
2965	.getattr	= pid_getattr,
2966	.setattr	= proc_setattr,
2967	.permission	= proc_pid_permission,
2968};
2969
2970static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2971{
2972	struct dentry *dentry, *leader, *dir;
2973	char buf[PROC_NUMBUF];
2974	struct qstr name;
2975
2976	name.name = buf;
2977	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2978	/* no ->d_hash() rejects on procfs */
2979	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2980	if (dentry) {
2981		d_invalidate(dentry);
2982		dput(dentry);
2983	}
2984
2985	if (pid == tgid)
2986		return;
2987
2988	name.name = buf;
2989	name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2990	leader = d_hash_and_lookup(mnt->mnt_root, &name);
2991	if (!leader)
2992		goto out;
2993
2994	name.name = "task";
2995	name.len = strlen(name.name);
2996	dir = d_hash_and_lookup(leader, &name);
2997	if (!dir)
2998		goto out_put_leader;
2999
3000	name.name = buf;
3001	name.len = snprintf(buf, sizeof(buf), "%d", pid);
3002	dentry = d_hash_and_lookup(dir, &name);
3003	if (dentry) {
3004		d_invalidate(dentry);
3005		dput(dentry);
3006	}
3007
3008	dput(dir);
3009out_put_leader:
3010	dput(leader);
3011out:
3012	return;
3013}
3014
3015/**
3016 * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
3017 * @task: task that should be flushed.
3018 *
3019 * When flushing dentries from proc, one needs to flush them from global
3020 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3021 * in. This call is supposed to do all of this job.
3022 *
3023 * Looks in the dcache for
3024 * /proc/@pid
3025 * /proc/@tgid/task/@pid
3026 * if either directory is present flushes it and all of it'ts children
3027 * from the dcache.
3028 *
3029 * It is safe and reasonable to cache /proc entries for a task until
3030 * that task exits.  After that they just clog up the dcache with
3031 * useless entries, possibly causing useful dcache entries to be
3032 * flushed instead.  This routine is proved to flush those useless
3033 * dcache entries at process exit time.
3034 *
3035 * NOTE: This routine is just an optimization so it does not guarantee
3036 *       that no dcache entries will exist at process exit time it
3037 *       just makes it very unlikely that any will persist.
3038 */
3039
3040void proc_flush_task(struct task_struct *task)
3041{
3042	int i;
3043	struct pid *pid, *tgid;
3044	struct upid *upid;
3045
3046	pid = task_pid(task);
3047	tgid = task_tgid(task);
3048
3049	for (i = 0; i <= pid->level; i++) {
3050		upid = &pid->numbers[i];
3051		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3052					tgid->numbers[i].nr);
3053	}
3054}
3055
3056static int proc_pid_instantiate(struct inode *dir,
3057				   struct dentry * dentry,
3058				   struct task_struct *task, const void *ptr)
3059{
3060	struct inode *inode;
3061
3062	inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3063	if (!inode)
3064		goto out;
3065
3066	inode->i_op = &proc_tgid_base_inode_operations;
3067	inode->i_fop = &proc_tgid_base_operations;
3068	inode->i_flags|=S_IMMUTABLE;
3069
3070	set_nlink(inode, nlink_tgid);
 
3071
3072	d_set_d_op(dentry, &pid_dentry_operations);
3073
3074	d_add(dentry, inode);
3075	/* Close the race of the process dying before we return the dentry */
3076	if (pid_revalidate(dentry, 0))
3077		return 0;
3078out:
3079	return -ENOENT;
3080}
3081
3082struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3083{
3084	int result = -ENOENT;
3085	struct task_struct *task;
3086	unsigned tgid;
 
3087	struct pid_namespace *ns;
 
3088
3089	tgid = name_to_int(&dentry->d_name);
3090	if (tgid == ~0U)
3091		goto out;
3092
3093	ns = dentry->d_sb->s_fs_info;
 
3094	rcu_read_lock();
3095	task = find_task_by_pid_ns(tgid, ns);
3096	if (task)
3097		get_task_struct(task);
3098	rcu_read_unlock();
3099	if (!task)
3100		goto out;
3101
3102	result = proc_pid_instantiate(dir, dentry, task, NULL);
 
 
 
 
 
 
 
3103	put_task_struct(task);
3104out:
3105	return ERR_PTR(result);
3106}
3107
3108/*
3109 * Find the first task with tgid >= tgid
3110 *
3111 */
3112struct tgid_iter {
3113	unsigned int tgid;
3114	struct task_struct *task;
3115};
3116static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3117{
3118	struct pid *pid;
3119
3120	if (iter.task)
3121		put_task_struct(iter.task);
3122	rcu_read_lock();
3123retry:
3124	iter.task = NULL;
3125	pid = find_ge_pid(iter.tgid, ns);
3126	if (pid) {
3127		iter.tgid = pid_nr_ns(pid, ns);
3128		iter.task = pid_task(pid, PIDTYPE_PID);
3129		/* What we to know is if the pid we have find is the
3130		 * pid of a thread_group_leader.  Testing for task
3131		 * being a thread_group_leader is the obvious thing
3132		 * todo but there is a window when it fails, due to
3133		 * the pid transfer logic in de_thread.
3134		 *
3135		 * So we perform the straight forward test of seeing
3136		 * if the pid we have found is the pid of a thread
3137		 * group leader, and don't worry if the task we have
3138		 * found doesn't happen to be a thread group leader.
3139		 * As we don't care in the case of readdir.
3140		 */
3141		if (!iter.task || !has_group_leader_pid(iter.task)) {
3142			iter.tgid += 1;
3143			goto retry;
3144		}
3145		get_task_struct(iter.task);
3146	}
3147	rcu_read_unlock();
3148	return iter;
3149}
3150
3151#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3152
3153/* for the /proc/ directory itself, after non-process stuff has been done */
3154int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3155{
3156	struct tgid_iter iter;
3157	struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
 
3158	loff_t pos = ctx->pos;
3159
3160	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3161		return 0;
3162
3163	if (pos == TGID_OFFSET - 2) {
3164		struct inode *inode = d_inode(ns->proc_self);
3165		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3166			return 0;
3167		ctx->pos = pos = pos + 1;
3168	}
3169	if (pos == TGID_OFFSET - 1) {
3170		struct inode *inode = d_inode(ns->proc_thread_self);
3171		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3172			return 0;
3173		ctx->pos = pos = pos + 1;
3174	}
3175	iter.tgid = pos - TGID_OFFSET;
3176	iter.task = NULL;
3177	for (iter = next_tgid(ns, iter);
3178	     iter.task;
3179	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3180		char name[PROC_NUMBUF];
3181		int len;
3182
3183		cond_resched();
3184		if (!has_pid_permissions(ns, iter.task, 2))
3185			continue;
3186
3187		len = snprintf(name, sizeof(name), "%d", iter.tgid);
3188		ctx->pos = iter.tgid + TGID_OFFSET;
3189		if (!proc_fill_cache(file, ctx, name, len,
3190				     proc_pid_instantiate, iter.task, NULL)) {
3191			put_task_struct(iter.task);
3192			return 0;
3193		}
3194	}
3195	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3196	return 0;
3197}
3198
3199/*
3200 * proc_tid_comm_permission is a special permission function exclusively
3201 * used for the node /proc/<pid>/task/<tid>/comm.
3202 * It bypasses generic permission checks in the case where a task of the same
3203 * task group attempts to access the node.
3204 * The rationale behind this is that glibc and bionic access this node for
3205 * cross thread naming (pthread_set/getname_np(!self)). However, if
3206 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3207 * which locks out the cross thread naming implementation.
3208 * This function makes sure that the node is always accessible for members of
3209 * same thread group.
3210 */
3211static int proc_tid_comm_permission(struct inode *inode, int mask)
3212{
3213	bool is_same_tgroup;
3214	struct task_struct *task;
3215
3216	task = get_proc_task(inode);
3217	if (!task)
3218		return -ESRCH;
3219	is_same_tgroup = same_thread_group(current, task);
3220	put_task_struct(task);
3221
3222	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3223		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3224		 * read or written by the members of the corresponding
3225		 * thread group.
3226		 */
3227		return 0;
3228	}
3229
3230	return generic_permission(inode, mask);
3231}
3232
3233static const struct inode_operations proc_tid_comm_inode_operations = {
3234		.permission = proc_tid_comm_permission,
3235};
3236
3237/*
3238 * Tasks
3239 */
3240static const struct pid_entry tid_base_stuff[] = {
3241	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3242	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3243	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3244#ifdef CONFIG_NET
3245	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3246#endif
3247	REG("environ",   S_IRUSR, proc_environ_operations),
3248	REG("auxv",      S_IRUSR, proc_auxv_operations),
3249	ONE("status",    S_IRUGO, proc_pid_status),
3250	ONE("personality", S_IRUSR, proc_pid_personality),
3251	ONE("limits",	 S_IRUGO, proc_pid_limits),
3252#ifdef CONFIG_SCHED_DEBUG
3253	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3254#endif
3255	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3256			 &proc_tid_comm_inode_operations,
3257			 &proc_pid_set_comm_operations, {}),
3258#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3259	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3260#endif
3261	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3262	ONE("stat",      S_IRUGO, proc_tid_stat),
3263	ONE("statm",     S_IRUGO, proc_pid_statm),
3264	REG("maps",      S_IRUGO, proc_tid_maps_operations),
3265#ifdef CONFIG_PROC_CHILDREN
3266	REG("children",  S_IRUGO, proc_tid_children_operations),
3267#endif
3268#ifdef CONFIG_NUMA
3269	REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3270#endif
3271	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3272	LNK("cwd",       proc_cwd_link),
3273	LNK("root",      proc_root_link),
3274	LNK("exe",       proc_exe_link),
3275	REG("mounts",    S_IRUGO, proc_mounts_operations),
3276	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3277#ifdef CONFIG_PROC_PAGE_MONITOR
3278	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3279	REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
 
3280	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3281#endif
3282#ifdef CONFIG_SECURITY
3283	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3284#endif
3285#ifdef CONFIG_KALLSYMS
3286	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3287#endif
3288#ifdef CONFIG_STACKTRACE
3289	ONE("stack",      S_IRUSR, proc_pid_stack),
3290#endif
3291#ifdef CONFIG_SCHED_INFO
3292	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3293#endif
3294#ifdef CONFIG_LATENCYTOP
3295	REG("latency",  S_IRUGO, proc_lstats_operations),
3296#endif
3297#ifdef CONFIG_PROC_PID_CPUSET
3298	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3299#endif
3300#ifdef CONFIG_CGROUPS
3301	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3302#endif
 
 
 
3303	ONE("oom_score", S_IRUGO, proc_oom_score),
3304	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3305	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3306#ifdef CONFIG_AUDITSYSCALL
3307	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3308	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3309#endif
3310#ifdef CONFIG_FAULT_INJECTION
3311	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
 
3312#endif
3313#ifdef CONFIG_TASK_IO_ACCOUNTING
3314	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3315#endif
3316#ifdef CONFIG_HARDWALL
3317	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
3318#endif
3319#ifdef CONFIG_USER_NS
3320	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3321	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3322	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3323	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3324#endif
 
 
 
 
 
 
3325};
3326
3327static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3328{
3329	return proc_pident_readdir(file, ctx,
3330				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3331}
3332
3333static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3334{
3335	return proc_pident_lookup(dir, dentry,
3336				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
 
3337}
3338
3339static const struct file_operations proc_tid_base_operations = {
3340	.read		= generic_read_dir,
3341	.iterate_shared	= proc_tid_base_readdir,
3342	.llseek		= generic_file_llseek,
3343};
3344
3345static const struct inode_operations proc_tid_base_inode_operations = {
3346	.lookup		= proc_tid_base_lookup,
3347	.getattr	= pid_getattr,
3348	.setattr	= proc_setattr,
3349};
3350
3351static int proc_task_instantiate(struct inode *dir,
3352	struct dentry *dentry, struct task_struct *task, const void *ptr)
3353{
3354	struct inode *inode;
3355	inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3356
3357	if (!inode)
3358		goto out;
 
3359	inode->i_op = &proc_tid_base_inode_operations;
3360	inode->i_fop = &proc_tid_base_operations;
3361	inode->i_flags|=S_IMMUTABLE;
3362
3363	set_nlink(inode, nlink_tid);
 
3364
3365	d_set_d_op(dentry, &pid_dentry_operations);
3366
3367	d_add(dentry, inode);
3368	/* Close the race of the process dying before we return the dentry */
3369	if (pid_revalidate(dentry, 0))
3370		return 0;
3371out:
3372	return -ENOENT;
3373}
3374
3375static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3376{
3377	int result = -ENOENT;
3378	struct task_struct *task;
3379	struct task_struct *leader = get_proc_task(dir);
3380	unsigned tid;
 
3381	struct pid_namespace *ns;
 
3382
3383	if (!leader)
3384		goto out_no_task;
3385
3386	tid = name_to_int(&dentry->d_name);
3387	if (tid == ~0U)
3388		goto out;
3389
3390	ns = dentry->d_sb->s_fs_info;
 
3391	rcu_read_lock();
3392	task = find_task_by_pid_ns(tid, ns);
3393	if (task)
3394		get_task_struct(task);
3395	rcu_read_unlock();
3396	if (!task)
3397		goto out;
3398	if (!same_thread_group(leader, task))
3399		goto out_drop_task;
3400
3401	result = proc_task_instantiate(dir, dentry, task, NULL);
3402out_drop_task:
3403	put_task_struct(task);
3404out:
3405	put_task_struct(leader);
3406out_no_task:
3407	return ERR_PTR(result);
3408}
3409
3410/*
3411 * Find the first tid of a thread group to return to user space.
3412 *
3413 * Usually this is just the thread group leader, but if the users
3414 * buffer was too small or there was a seek into the middle of the
3415 * directory we have more work todo.
3416 *
3417 * In the case of a short read we start with find_task_by_pid.
3418 *
3419 * In the case of a seek we start with the leader and walk nr
3420 * threads past it.
3421 */
3422static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3423					struct pid_namespace *ns)
3424{
3425	struct task_struct *pos, *task;
3426	unsigned long nr = f_pos;
3427
3428	if (nr != f_pos)	/* 32bit overflow? */
3429		return NULL;
3430
3431	rcu_read_lock();
3432	task = pid_task(pid, PIDTYPE_PID);
3433	if (!task)
3434		goto fail;
3435
3436	/* Attempt to start with the tid of a thread */
3437	if (tid && nr) {
3438		pos = find_task_by_pid_ns(tid, ns);
3439		if (pos && same_thread_group(pos, task))
3440			goto found;
3441	}
3442
3443	/* If nr exceeds the number of threads there is nothing todo */
3444	if (nr >= get_nr_threads(task))
3445		goto fail;
3446
3447	/* If we haven't found our starting place yet start
3448	 * with the leader and walk nr threads forward.
3449	 */
3450	pos = task = task->group_leader;
3451	do {
3452		if (!nr--)
3453			goto found;
3454	} while_each_thread(task, pos);
3455fail:
3456	pos = NULL;
3457	goto out;
3458found:
3459	get_task_struct(pos);
3460out:
3461	rcu_read_unlock();
3462	return pos;
3463}
3464
3465/*
3466 * Find the next thread in the thread list.
3467 * Return NULL if there is an error or no next thread.
3468 *
3469 * The reference to the input task_struct is released.
3470 */
3471static struct task_struct *next_tid(struct task_struct *start)
3472{
3473	struct task_struct *pos = NULL;
3474	rcu_read_lock();
3475	if (pid_alive(start)) {
3476		pos = next_thread(start);
3477		if (thread_group_leader(pos))
3478			pos = NULL;
3479		else
3480			get_task_struct(pos);
3481	}
3482	rcu_read_unlock();
3483	put_task_struct(start);
3484	return pos;
3485}
3486
3487/* for the /proc/TGID/task/ directories */
3488static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3489{
3490	struct inode *inode = file_inode(file);
3491	struct task_struct *task;
3492	struct pid_namespace *ns;
3493	int tid;
3494
3495	if (proc_inode_is_dead(inode))
3496		return -ENOENT;
3497
3498	if (!dir_emit_dots(file, ctx))
3499		return 0;
3500
3501	/* f_version caches the tgid value that the last readdir call couldn't
3502	 * return. lseek aka telldir automagically resets f_version to 0.
3503	 */
3504	ns = inode->i_sb->s_fs_info;
3505	tid = (int)file->f_version;
3506	file->f_version = 0;
3507	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3508	     task;
3509	     task = next_tid(task), ctx->pos++) {
3510		char name[PROC_NUMBUF];
3511		int len;
3512		tid = task_pid_nr_ns(task, ns);
3513		len = snprintf(name, sizeof(name), "%d", tid);
3514		if (!proc_fill_cache(file, ctx, name, len,
3515				proc_task_instantiate, task, NULL)) {
3516			/* returning this tgid failed, save it as the first
3517			 * pid for the next readir call */
3518			file->f_version = (u64)tid;
3519			put_task_struct(task);
3520			break;
3521		}
3522	}
3523
3524	return 0;
3525}
3526
3527static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 
3528{
3529	struct inode *inode = d_inode(dentry);
3530	struct task_struct *p = get_proc_task(inode);
3531	generic_fillattr(inode, stat);
3532
3533	if (p) {
3534		stat->nlink += get_nr_threads(p);
3535		put_task_struct(p);
3536	}
3537
3538	return 0;
3539}
3540
3541static const struct inode_operations proc_task_inode_operations = {
3542	.lookup		= proc_task_lookup,
3543	.getattr	= proc_task_getattr,
3544	.setattr	= proc_setattr,
3545	.permission	= proc_pid_permission,
3546};
3547
3548static const struct file_operations proc_task_operations = {
3549	.read		= generic_read_dir,
3550	.iterate_shared	= proc_task_readdir,
3551	.llseek		= generic_file_llseek,
3552};
3553
3554void __init set_proc_pid_nlink(void)
3555{
3556	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3557	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3558}
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}