/*
* 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 with sys_umount(). If the accointing
* 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_sem 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 <asm/uaccess.h>
#include <asm/div64.h>
#include <linux/blkdev.h> /* sector_div */
#include <linux/pid_namespace.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.
*/
static void do_acct_process(struct bsd_acct_struct *acct,
struct pid_namespace *ns, struct file *);
/*
* This structure is used so that all the data protected by lock
* can be placed in the same cache line as the lock. This primes
* the cache line to have the data after getting the lock.
*/
struct bsd_acct_struct {
volatile int active;
volatile int needcheck;
struct file *file;
struct pid_namespace *ns;
struct timer_list timer;
struct list_head list;
};
static DEFINE_SPINLOCK(acct_lock);
static LIST_HEAD(acct_list);
/*
* Called whenever the timer says to check the free space.
*/
static void acct_timeout(unsigned long x)
{
struct bsd_acct_struct *acct = (struct bsd_acct_struct *)x;
acct->needcheck = 1;
}
/*
* Check the amount of free space and suspend/resume accordingly.
*/
static int check_free_space(struct bsd_acct_struct *acct, struct file *file)
{
struct kstatfs sbuf;
int res;
int act;
sector_t resume;
sector_t suspend;
spin_lock(&acct_lock);
res = acct->active;
if (!file || !acct->needcheck)
goto out;
spin_unlock(&acct_lock);
/* May block */
if (vfs_statfs(&file->f_path, &sbuf))
return res;
suspend = sbuf.f_blocks * SUSPEND;
resume = sbuf.f_blocks * RESUME;
sector_div(suspend, 100);
sector_div(resume, 100);
if (sbuf.f_bavail <= suspend)
act = -1;
else if (sbuf.f_bavail >= resume)
act = 1;
else
act = 0;
/*
* If some joker switched acct->file under us we'ld better be
* silent and _not_ touch anything.
*/
spin_lock(&acct_lock);
if (file != acct->file) {
if (act)
res = act>0;
goto out;
}
if (acct->active) {
if (act < 0) {
acct->active = 0;
printk(KERN_INFO "Process accounting paused\n");
}
} else {
if (act > 0) {
acct->active = 1;
printk(KERN_INFO "Process accounting resumed\n");
}
}
del_timer(&acct->timer);
acct->needcheck = 0;
acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
add_timer(&acct->timer);
res = acct->active;
out:
spin_unlock(&acct_lock);
return res;
}
/*
* Close the old accounting file (if currently open) and then replace
* it with file (if non-NULL).
*
* NOTE: acct_lock MUST be held on entry and exit.
*/
static void acct_file_reopen(struct bsd_acct_struct *acct, struct file *file,
struct pid_namespace *ns)
{
struct file *old_acct = NULL;
struct pid_namespace *old_ns = NULL;
if (acct->file) {
old_acct = acct->file;
old_ns = acct->ns;
del_timer(&acct->timer);
acct->active = 0;
acct->needcheck = 0;
acct->file = NULL;
acct->ns = NULL;
list_del(&acct->list);
}
if (file) {
acct->file = file;
acct->ns = ns;
acct->needcheck = 0;
acct->active = 1;
list_add(&acct->list, &acct_list);
/* It's been deleted if it was used before so this is safe */
setup_timer(&acct->timer, acct_timeout, (unsigned long)acct);
acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
add_timer(&acct->timer);
}
if (old_acct) {
mnt_unpin(old_acct->f_path.mnt);
spin_unlock(&acct_lock);
do_acct_process(acct, old_ns, old_acct);
filp_close(old_acct, NULL);
spin_lock(&acct_lock);
}
}
static int acct_on(char *name)
{
struct file *file;
struct vfsmount *mnt;
struct pid_namespace *ns;
struct bsd_acct_struct *acct = NULL;
/* Difference from BSD - they don't do O_APPEND */
file = filp_open(name, O_WRONLY|O_APPEND|O_LARGEFILE, 0);
if (IS_ERR(file))
return PTR_ERR(file);
if (!S_ISREG(file->f_path.dentry->d_inode->i_mode)) {
filp_close(file, NULL);
return -EACCES;
}
if (!file->f_op->write) {
filp_close(file, NULL);
return -EIO;
}
ns = task_active_pid_ns(current);
if (ns->bacct == NULL) {
acct = kzalloc(sizeof(struct bsd_acct_struct), GFP_KERNEL);
if (acct == NULL) {
filp_close(file, NULL);
return -ENOMEM;
}
}
spin_lock(&acct_lock);
if (ns->bacct == NULL) {
ns->bacct = acct;
acct = NULL;
}
mnt = file->f_path.mnt;
mnt_pin(mnt);
acct_file_reopen(ns->bacct, file, ns);
spin_unlock(&acct_lock);
mntput(mnt); /* it's pinned, now give up active reference */
kfree(acct);
return 0;
}
/**
* sys_acct - enable/disable process accounting
* @name: file name for accounting records or NULL to shutdown accounting
*
* Returns 0 for success or negative errno values for failure.
*
* sys_acct() is the only system call needed to implement process
* accounting. It takes the name of the file where accounting records
* should be written. If the filename is NULL, accounting will be
* shutdown.
*/
SYSCALL_DEFINE1(acct, const char __user *, name)
{