path: root/kernel/acct.c

// SPDX-License-Identifier: GPL-2.0
/*
 *  linux/kernel/acct.c
 *
 *  BSD Process Accounting for Linux
 *
 *  Author: Marco van Wieringen <mvw@planets.elm.net>
 *
 *  Some code based on ideas and code from:
 *  Thomas K. Dyas <tdyas@eden.rutgers.edu>
 *
 *  This file implements BSD-style process accounting. Whenever any
 *  process exits, an accounting record of type "struct acct" is
 *  written to the file specified with the acct() system call. It is
 *  up to user-level programs to do useful things with the accounting
 *  log. The kernel just provides the raw accounting information.
 *
 * (C) Copyright 1995 - 1997 Marco van Wieringen - ELM Consultancy B.V.
 *
 *  Plugged two leaks. 1) It didn't return acct_file into the free_filps if
 *  the file happened to be read-only. 2) If the accounting was suspended
 *  due to the lack of space it happily allowed to reopen it and completely
 *  lost the old acct_file. 3/10/98, Al Viro.
 *
 *  Now we silently close acct_file on attempt to reopen. Cleaned sys_acct().
 *  XTerms and EMACS are manifestations of pure evil. 21/10/98, AV.
 *
 *  Fixed a nasty interaction with sys_umount(). If the accounting
 *  was suspeneded we failed to stop it on umount(). Messy.
 *  Another one: remount to readonly didn't stop accounting.
 *	Question: what should we do if we have CAP_SYS_ADMIN but not
 *  CAP_SYS_PACCT? Current code does the following: umount returns -EBUSY
 *  unless we are messing with the root. In that case we are getting a
 *  real mess with do_remount_sb(). 9/11/98, AV.
 *
 *  Fixed a bunch of races (and pair of leaks). Probably not the best way,
 *  but this one obviously doesn't introduce deadlocks. Later. BTW, found
 *  one race (and leak) in BSD implementation.
 *  OK, that's better. ANOTHER race and leak in BSD variant. There always
 *  is one more bug... 10/11/98, AV.
 *
 *	Oh, fsck... Oopsable SMP race in do_process_acct() - we must hold
 * ->mmap_lock to walk the vma list of current->mm. Nasty, since it leaks
 * a struct file opened for write. Fixed. 2/6/2000, AV.
 */

#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/acct.h>
#include <linux/capability.h>
#include <linux/file.h>
#include <linux/tty.h>
#include <linux/security.h>
#include <linux/vfs.h>
#include <linux/jiffies.h>
#include <linux/times.h>
#include <linux/syscalls.h>
#include <linux/mount.h>
#include <linux/uaccess.h>
#include <linux/sched/cputime.h>

#include <asm/div64.h>
#include <linux/blkdev.h> /* sector_div */
#include <linux/pid_namespace.h>
#include <linux/fs_pin.h>

/*
 * These constants control the amount of freespace that suspend and
 * resume the process accounting system, and the time delay between
 * each check.
 * Turned into sysctl-controllable parameters. AV, 12/11/98
 */

int acct_parm[3] = {4, 2, 30};
#define RESUME		(acct_parm[0])	/* >foo% free space - resume */
#define SUSPEND		(acct_parm[1])	/* <foo% free space - suspend */
#define ACCT_TIMEOUT	(acct_parm[2])	/* foo second timeout between checks */

/*
 * External references and all of the globals.
 */

struct bsd_acct_struct {
	struct fs_pin		pin;
	atomic_long_t		count;
	struct rcu_head		rcu;
	struct mutex		lock;
	int			active;
	unsigned long		needcheck;
	struct file		*file;
	struct pid_namespace	*ns;
	struct work_struct	work;
	struct completion	done;
};

static void do_acct_process(struct bsd_acct_struct *acct);

/*
 * Check the amount of free space and suspend/resume accordingly.
 */
static int check_free_space(struct bsd_acct_struct *acct)
{
	struct kstatfs sbuf;

	if (time_is_after_jiffies(acct->needcheck))
		goto out;

	/* May block */
	if (vfs_statfs(&acct->file->f_path, &sbuf))
		goto out;

	if (acct->active) {
		u64 suspend = sbuf.f_blocks * SUSPEND;
		do_div(suspend, 100);
		if (sbuf.f_bavail <= suspend) {
			acct->active = 0;
			pr_info("Process accounting paused\n");
		}
	} else {
		u64 resume = sbuf.f_blocks * RESUME;
		do_div(resume, 100);
		if (sbuf.f_bavail >= resume) {
			acct->active = 1;
			pr_info("Process accounting resumed\n");
		}
	}

	acct->needcheck = jiffies + ACCT_TIMEOUT*HZ;
out:
	return acct->active;
}

static void acct_put(struct bsd_acct_struct *p)
{
	if (atomic_long_dec_and_test(&p->count))
		kfree_rcu(p, rcu);
}

static inline struct bsd_acct_struct *to_acct(struct fs_pin *p)
{
	return p ? container_of(p, struct bsd_acct_struct, pin) : NULL;
}

static struct bsd_acct_struct *acct_get(struct pid_namespace *ns)
{
	struct bsd_acct_struct *res;
again:
	smp_rmb();
	rcu_read_lock();
	res = to_acct(READ_ONCE(ns->bacct));
	if (!res) {
		rcu_read_unlock();
		return NULL;
	}
	if (!atomic_long_inc_not_zero(&res->count)) {
		rcu_read_unlock();
		cpu_relax();
		goto again;
	}
	rcu_read_unlock();
	mutex_lock(&res->lock);
	if (res != to_acct(READ_ONCE(ns->bacct))) {
		mutex_unlock(&res->lock);
		acct_put(res);
		goto again;
	}
	return res;
}

static void acct_pin_kill(struct fs_pin *pin)
{
	struct bsd_acct_struct *acct = to_acct(pin);
	mutex_lock(&acct->lock);
	do_acct_process(acct);
	schedule_work(&acct->work);
	wait_for_completion(&acct->done);
	cmpxchg(&acct->ns->bacct, pin, NULL);
	mutex_unlock(&acct->lock);
	pin_remove(pin);
	acct_put(acct);
}

static void close_work(struct work_struct *work)
{
	struct bsd_acct_struct *acct = container_of(work, struct bsd_acct_struct, work);
	struct file *file = acct->file;
	if (file->f_op->flush)
		file->f_op->flush(file, NULL);
	__fput_sync(file);
	complete(&acct->done);
}

static int acct_on(struct filename *pathname)
{
	struct file *file;
	struct vfsmount *mnt, *internal;
	struct pid_namespace *ns = task_active_pid_ns(current);
	struct bsd_acct_struct *acct;
	struct fs_pin *old;
	int err;

	acct = kzalloc(sizeof(struct bsd_acct_struct), GFP_KERNEL);
	if (!acct)
		return -ENOMEM;

	/* Difference from BSD - they don't do O_APPEND */
	file = file_open_name(pathname, O_WRONLY|O_APPEND|O_LARGEFILE, 0);
	if (IS_ERR(file)) {
		kfree(acct);
		return PTR_ERR(file);
	}

	if (!S_ISREG(file_inode(file)->i_mode)) {
		kfree(acct);
		filp_close(file, NULL);
		return -EACCES;
	}

	if (!(file->f_mode & FMODE_CAN_WRITE)) {
		kfree(acct);
		filp_close(file, NULL);
		return -EIO;
	}
	internal = mnt_clone_internal(&file->f_path);
	if (IS_ERR(internal)) {
		kfree(acct);
		filp_close(file, NULL);
		return PTR_ERR(internal);
	}
	err = __mnt_want_write(internal);
	if (err) {
		mntput(internal);
		kfree(acct);
		filp_close(file, NULL);
		return err;
	}
	mnt = file->f_path.mnt;
	file->f_path.mnt = internal;

	atomic_long_set(&acct->count, 1);
	init_fs_pin(&acct->pin, acct_pin_kill);
	acct->file = file;
	acct->needcheck = jiffies;
	acct->ns = ns;
	mutex_init(&acct->lock);
	INIT_WORK(&acct->work, close_work);
	init_completion(&acct->done);
	mutex_lock_nested(&acct->lock, 1);	/* nobody has seen it yet */
	pin_insert(&acct->pin, mnt);

	rcu_read_lock();
	old = xchg(&ns->bacct, &acct->pin);
	mutex_unlock(&acct->lock);
	pin_kill(old