1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46#include <linux/mm.h>
47#include <linux/slab.h>
48#include <linux/acct.h>
49#include <linux/capability.h>
50#include <linux/file.h>
51#include <linux/tty.h>
52#include <linux/security.h>
53#include <linux/vfs.h>
54#include <linux/jiffies.h>
55#include <linux/times.h>
56#include <linux/syscalls.h>
57#include <linux/mount.h>
58#include <asm/uaccess.h>
59#include <asm/div64.h>
60#include <linux/blkdev.h>
61#include <linux/pid_namespace.h>
62
63
64
65
66
67
68
69
70int acct_parm[3] = {4, 2, 30};
71#define RESUME (acct_parm[0])
72#define SUSPEND (acct_parm[1])
73#define ACCT_TIMEOUT (acct_parm[2])
74
75
76
77
78static void do_acct_process(struct bsd_acct_struct *acct,
79 struct pid_namespace *ns, struct file *);
80
81
82
83
84
85
86struct bsd_acct_struct {
87 volatile int active;
88 volatile int needcheck;
89 struct file *file;
90 struct pid_namespace *ns;
91 struct timer_list timer;
92 struct list_head list;
93};
94
95static DEFINE_SPINLOCK(acct_lock);
96static LIST_HEAD(acct_list);
97
98
99
100
101static void acct_timeout(unsigned long x)
102{
103 struct bsd_acct_struct *acct = (struct bsd_acct_struct *)x;
104 acct->needcheck = 1;
105}
106
107
108
109
110static int check_free_space(struct bsd_acct_struct *acct, struct file *file)
111{
112 struct kstatfs sbuf;
113 int res;
114 int act;
115 sector_t resume;
116 sector_t suspend;
117
118 spin_lock(&acct_lock);
119 res = acct->active;
120 if (!file || !acct->needcheck)
121 goto out;
122 spin_unlock(&acct_lock);
123
124
125 if (vfs_statfs(&file->f_path, &sbuf))
126 return res;
127 suspend = sbuf.f_blocks * SUSPEND;
128 resume = sbuf.f_blocks * RESUME;
129
130 sector_div(suspend, 100);
131 sector_div(resume, 100);
132
133 if (sbuf.f_bavail <= suspend)
134 act = -1;
135 else if (sbuf.f_bavail >= resume)
136 act = 1;
137 else
138 act = 0;
139
140
141
142
143
144 spin_lock(&acct_lock);
145 if (file != acct->file) {
146 if (act)
147 res = act>0;
148 goto out;
149 }
150
151 if (acct->active) {
152 if (act < 0) {
153 acct->active = 0;
154 printk(KERN_INFO "Process accounting paused\n");
155 }
156 } else {
157 if (act > 0) {
158 acct->active = 1;
159 printk(KERN_INFO "Process accounting resumed\n");
160 }
161 }
162
163 del_timer(&acct->timer);
164 acct->needcheck = 0;
165 acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
166 add_timer(&acct->timer);
167 res = acct->active;
168out:
169 spin_unlock(&acct_lock);
170 return res;
171}
172
173
174
175
176
177
178
179static void acct_file_reopen(struct bsd_acct_struct *acct, struct file *file,
180 struct pid_namespace *ns)
181{
182 struct file *old_acct = NULL;
183 struct pid_namespace *old_ns = NULL;
184
185 if (acct->file) {
186 old_acct = acct->file;
187 old_ns = acct->ns;
188 del_timer(&acct->timer);
189 acct->active = 0;
190 acct->needcheck = 0;
191 acct->file = NULL;
192 acct->ns = NULL;
193 list_del(&acct->list);
194 }
195 if (file) {
196 acct->file = file;
197 acct->ns = ns;
198 acct->needcheck = 0;
199 acct->active = 1;
200 list_add(&acct->list, &acct_list);
201
202 setup_timer(&acct->timer, acct_timeout, (unsigned long)acct);
203 acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
204 add_timer(&acct->timer);
205 }
206 if (old_acct) {
207 mnt_unpin(old_acct->f_path.mnt);
208 spin_unlock(&acct_lock);
209 do_acct_process(acct, old_ns, old_acct);
210 filp_close(old_acct, NULL);
211 spin_lock(&acct_lock);
212 }
213}
214
215static int acct_on(char *name)
216{
217 struct file *file;
218 struct vfsmount *mnt;
219 struct pid_namespace *ns;
220 struct bsd_acct_struct *acct = NULL;
221
222
223 file = filp_open(name, O_WRONLY|O_APPEND|O_LARGEFILE, 0);
224 if (IS_ERR(file))
225 return PTR_ERR(file);
226
227 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode)) {
228 filp_close(file, NULL);
229 return -EACCES;
230 }
231
232 if (!file->f_op->write) {
233 filp_close(file, NULL);
234 return -EIO;
235 }
236
237 ns = task_active_pid_ns(current);
238 if (ns->bacct == NULL) {
239 acct = kzalloc(sizeof(struct bsd_acct_struct), GFP_KERNEL);
240 if (acct == NULL) {
241 filp_close(file, NULL);
242 return -ENOMEM;
243 }
244 }
245
246 spin_lock(&acct_lock);
247 if (ns->bacct == NULL) {
248 ns->bacct = acct;
249 acct = NULL;
250 }
251
252 mnt = file->f_path.mnt;
253 mnt_pin(mnt);
254 acct_file_reopen(ns->bacct, file, ns);
255 spin_unlock(&acct_lock);
256
257 mntput(mnt);
258 kfree(acct);
259
260 return 0;
261}
262
263
264
265
266
267
268
269
270
271
272
273
274SYSCALL_DEFINE1(acct, const char __user *, name)
275{
276 int error = 0;
277
278 if (!capable(CAP_SYS_PACCT))
279 return -EPERM;
280
281 if (name) {
282 char *tmp = getname(name);
283 if (IS_ERR(tmp))
284 return (PTR_ERR(tmp));
285 error = acct_on(tmp);
286 putname(tmp);
287 } else {
288 struct bsd_acct_struct *acct;
289
290 acct = task_active_pid_ns(current)->bacct;
291 if (acct == NULL)
292 return 0;
293
294 spin_lock(&acct_lock);
295 acct_file_reopen(acct, NULL, NULL);
296 spin_unlock(&acct_lock);
297 }
298
299 return error;
300}
301
302
303
304
305
306
307
308
309void acct_auto_close_mnt(struct vfsmount *m)
310{
311 struct bsd_acct_struct *acct;
312
313 spin_lock(&acct_lock);
314restart:
315 list_for_each_entry(acct, &acct_list, list)
316 if (acct->file && acct->file->f_path.mnt == m) {
317 acct_file_reopen(acct, NULL, NULL);
318 goto restart;
319 }
320 spin_unlock(&acct_lock);
321}
322
323
324
325
326
327
328
329
330void acct_auto_close(struct super_block *sb)
331{
332 struct bsd_acct_struct *acct;
333
334 spin_lock(&acct_lock);
335restart:
336 list_for_each_entry(acct, &acct_list, list)
337 if (acct->file && acct->file->f_path.mnt->mnt_sb == sb) {
338 acct_file_reopen(acct, NULL, NULL);
339 goto restart;
340 }
341 spin_unlock(&acct_lock);
342}
343
344void acct_exit_ns(struct pid_namespace *ns)
345{
346 struct bsd_acct_struct *acct = ns->bacct;
347
348 if (acct == NULL)
349 return;
350
351 del_timer_sync(&acct->timer);
352 spin_lock(&acct_lock);
353 if (acct->file != NULL)
354 acct_file_reopen(acct, NULL, NULL);
355 spin_unlock(&acct_lock);
356
357 kfree(acct);
358}
359
360
361
362
363
364
365
366
367
368#define MANTSIZE 13
369#define EXPSIZE 3
370#define MAXFRACT ((1 << MANTSIZE) - 1)
371
372static comp_t encode_comp_t(unsigned long value)
373{
374 int exp, rnd;
375
376 exp = rnd = 0;
377 while (value > MAXFRACT) {
378 rnd = value & (1 << (EXPSIZE - 1));
379 value >>= EXPSIZE;
380 exp++;
381 }
382
383
384
385
386 if (rnd && (++value > MAXFRACT)) {
387 value >>= EXPSIZE;
388 exp++;
389 }
390
391
392
393
394 exp <<= MANTSIZE;
395 exp += value;
396 return exp;
397}
398
399#if ACCT_VERSION==1 || ACCT_VERSION==2
400
401
402
403
404
405
406
407
408
409#define MANTSIZE2 20
410#define EXPSIZE2 5
411#define MAXFRACT2 ((1ul << MANTSIZE2) - 1)
412#define MAXEXP2 ((1 <<EXPSIZE2) - 1)
413
414static comp2_t encode_comp2_t(u64 value)
415{
416 int exp, rnd;
417
418 exp = (value > (MAXFRACT2>>1));
419 rnd = 0;
420 while (value > MAXFRACT2) {
421 rnd = value & 1;
422 value >>= 1;
423 exp++;
424 }
425
426
427
428
429 if (rnd && (++value > MAXFRACT2)) {
430 value >>= 1;
431 exp++;
432 }
433
434 if (exp > MAXEXP2) {
435
436 return (1ul << (MANTSIZE2+EXPSIZE2-1)) - 1;
437 } else {
438 return (value & (MAXFRACT2>>1)) | (exp << (MANTSIZE2-1));
439 }
440}
441#endif
442
443#if ACCT_VERSION==3
444
445
446
447static u32 encode_float(u64 value)
448{
449 unsigned exp = 190;
450 unsigned u;
451
452 if (value==0) return 0;
453 while ((s64)value > 0){
454 value <<= 1;
455 exp--;
456 }
457 u = (u32)(value >> 40) & 0x7fffffu;
458 return u | (exp << 23);
459}
460#endif
461
462
463
464
465
466
467
468
469
470
471
472
473
474static void do_acct_process(struct bsd_acct_struct *acct,
475 struct pid_namespace *ns, struct file *file)
476{
477 struct pacct_struct *pacct = ¤t->signal->pacct;
478 acct_t ac;
479 mm_segment_t fs;
480 unsigned long flim;
481 u64 elapsed;
482 u64 run_time;
483 struct timespec uptime;
484 struct tty_struct *tty;
485 const struct cred *orig_cred;
486
487
488 orig_cred = override_creds(file->f_cred);
489
490
491
492
493
494 if (!check_free_space(acct, file))
495 goto out;
496
497
498
499
500
501 memset((caddr_t)&ac, 0, sizeof(acct_t));
502
503 ac.ac_version = ACCT_VERSION | ACCT_BYTEORDER;
504 strlcpy(ac.ac_comm, current->comm, sizeof(ac.ac_comm));
505
506
507 do_posix_clock_monotonic_gettime(&uptime);
508 run_time = (u64)uptime.tv_sec*NSEC_PER_SEC + uptime.tv_nsec;
509 run_time -= (u64)current->group_leader->start_time.tv_sec * NSEC_PER_SEC
510 + current->group_leader->start_time.tv_nsec;
511
512 elapsed = nsec_to_AHZ(run_time);
513#if ACCT_VERSION==3
514 ac.ac_etime = encode_float(elapsed);
515#else
516 ac.ac_etime = encode_comp_t(elapsed < (unsigned long) -1l ?
517 (unsigned long) elapsed : (unsigned long) -1l);
518#endif
519#if ACCT_VERSION==1 || ACCT_VERSION==2
520 {
521
522 comp2_t etime = encode_comp2_t(elapsed);
523 ac.ac_etime_hi = etime >> 16;
524 ac.ac_etime_lo = (u16) etime;
525 }
526#endif
527 do_div(elapsed, AHZ);
528 ac.ac_btime = get_seconds() - elapsed;
529
530 ac.ac_uid = orig_cred->uid;
531 ac.ac_gid = orig_cred->gid;
532#if ACCT_VERSION==2
533 ac.ac_ahz = AHZ;
534#endif
535#if ACCT_VERSION==1 || ACCT_VERSION==2
536
537 ac.ac_uid16 = ac.ac_uid;
538 ac.ac_gid16 = ac.ac_gid;
539#endif
540#if ACCT_VERSION==3
541 ac.ac_pid = task_tgid_nr_ns(current, ns);
542 rcu_read_lock();
543 ac.ac_ppid = task_tgid_nr_ns(rcu_dereference(current->real_parent), ns);
544 rcu_read_unlock();
545#endif
546
547 spin_lock_irq(¤t->sighand->siglock);
548 tty = current->signal->tty;
549 ac.ac_tty = tty ? old_encode_dev(tty_devnum(tty)) : 0;
550 ac.ac_utime = encode_comp_t(jiffies_to_AHZ(cputime_to_jiffies(pacct->ac_utime)));
551 ac.ac_stime = encode_comp_t(jiffies_to_AHZ(cputime_to_jiffies(pacct->ac_stime)));
552 ac.ac_flag = pacct->ac_flag;
553 ac.ac_mem = encode_comp_t(pacct->ac_mem);
554 ac.ac_minflt = encode_comp_t(pacct->ac_minflt);
555 ac.ac_majflt = encode_comp_t(pacct->ac_majflt);
556 ac.ac_exitcode = pacct->ac_exitcode;
557 spin_unlock_irq(¤t->sighand->siglock);
558 ac.ac_io = encode_comp_t(0 );
559 ac.ac_rw = encode_comp_t(ac.ac_io / 1024);
560 ac.ac_swaps = encode_comp_t(0);
561
562
563
564
565
566 fs = get_fs();
567 set_fs(KERNEL_DS);
568
569
570
571 flim = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
572 current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
573 file->f_op->write(file, (char *)&ac,
574 sizeof(acct_t), &file->f_pos);
575 current->signal->rlim[RLIMIT_FSIZE].rlim_cur = flim;
576 set_fs(fs);
577out:
578 revert_creds(orig_cred);
579}
580
581
582
583
584
585
586void acct_collect(long exitcode, int group_dead)
587{
588 struct pacct_struct *pacct = ¤t->signal->pacct;
589 unsigned long vsize = 0;
590
591 if (group_dead && current->mm) {
592 struct vm_area_struct *vma;
593 down_read(¤t->mm->mmap_sem);
594 vma = current->mm->mmap;
595 while (vma) {
596 vsize += vma->vm_end - vma->vm_start;
597 vma = vma->vm_next;
598 }
599 up_read(¤t->mm->mmap_sem);
600 }
601
602 spin_lock_irq(¤t->sighand->siglock);
603 if (group_dead)
604 pacct->ac_mem = vsize / 1024;
605 if (thread_group_leader(current)) {
606 pacct->ac_exitcode = exitcode;
607 if (current->flags & PF_FORKNOEXEC)
608 pacct->ac_flag |= AFORK;
609 }
610 if (current->flags & PF_SUPERPRIV)
611 pacct->ac_flag |= ASU;
612 if (current->flags & PF_DUMPCORE)
613 pacct->ac_flag |= ACORE;
614 if (current->flags & PF_SIGNALED)
615 pacct->ac_flag |= AXSIG;
616 pacct->ac_utime = cputime_add(pacct->ac_utime, current->utime);
617 pacct->ac_stime = cputime_add(pacct->ac_stime, current->stime);
618 pacct->ac_minflt += current->min_flt;
619 pacct->ac_majflt += current->maj_flt;
620 spin_unlock_irq(¤t->sighand->siglock);
621}
622
623static void acct_process_in_ns(struct pid_namespace *ns)
624{
625 struct file *file = NULL;
626 struct bsd_acct_struct *acct;
627
628 acct = ns->bacct;
629
630
631
632 if (!acct || !acct->file)
633 return;
634
635 spin_lock(&acct_lock);
636 file = acct->file;
637 if (unlikely(!file)) {
638 spin_unlock(&acct_lock);
639 return;
640 }
641 get_file(file);
642 spin_unlock(&acct_lock);
643
644 do_acct_process(acct, ns, file);
645 fput(file);
646}
647
648
649
650
651
652
653
654void acct_process(void)
655{
656 struct pid_namespace *ns;
657
658
659
660
661
662
663 for (ns = task_active_pid_ns(current); ns != NULL; ns = ns->parent)
664 acct_process_in_ns(ns);
665}
666