1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
7 #ifndef FUSE_USE_VERSION
8 #define FUSE_USE_VERSION 26
11 #define _FILE_OFFSET_BITS 64
13 #define __STDC_FORMAT_MACROS
31 #include <linux/magic.h>
32 #include <linux/sched.h>
33 #include <sys/epoll.h>
35 #include <sys/mount.h>
36 #include <sys/param.h>
37 #include <sys/socket.h>
38 #include <sys/syscall.h>
39 #include <sys/sysinfo.h>
44 #include "cgroup_fuse.h"
45 #include "cgroups/cgroup.h"
46 #include "cgroups/cgroup_utils.h"
47 #include "cpuset_parse.h"
48 #include "memory_utils.h"
49 #include "proc_loadavg.h"
50 #include "proc_cpuview.h"
54 uint64_t hierarchical_memory_limit
;
55 uint64_t hierarchical_memsw_limit
;
58 uint64_t total_rss_huge
;
60 uint64_t total_mapped_file
;
62 uint64_t total_writeback
;
64 uint64_t total_pgpgin
;
65 uint64_t total_pgpgout
;
66 uint64_t total_pgfault
;
67 uint64_t total_pgmajfault
;
68 uint64_t total_inactive_anon
;
69 uint64_t total_active_anon
;
70 uint64_t total_inactive_file
;
71 uint64_t total_active_file
;
72 uint64_t total_unevictable
;
75 int proc_getattr(const char *path
, struct stat
*sb
)
79 memset(sb
, 0, sizeof(struct stat
));
80 if (clock_gettime(CLOCK_REALTIME
, &now
) < 0)
82 sb
->st_uid
= sb
->st_gid
= 0;
83 sb
->st_atim
= sb
->st_mtim
= sb
->st_ctim
= now
;
84 if (strcmp(path
, "/proc") == 0) {
85 sb
->st_mode
= S_IFDIR
| 00555;
89 if (strcmp(path
, "/proc/meminfo") == 0 ||
90 strcmp(path
, "/proc/cpuinfo") == 0 ||
91 strcmp(path
, "/proc/uptime") == 0 ||
92 strcmp(path
, "/proc/stat") == 0 ||
93 strcmp(path
, "/proc/diskstats") == 0 ||
94 strcmp(path
, "/proc/swaps") == 0 ||
95 strcmp(path
, "/proc/loadavg") == 0) {
97 sb
->st_mode
= S_IFREG
| 00444;
105 int proc_readdir(const char *path
, void *buf
, fuse_fill_dir_t filler
,
106 off_t offset
, struct fuse_file_info
*fi
)
108 if (filler(buf
, ".", NULL
, 0) != 0 ||
109 filler(buf
, "..", NULL
, 0) != 0 ||
110 filler(buf
, "cpuinfo", NULL
, 0) != 0 ||
111 filler(buf
, "meminfo", NULL
, 0) != 0 ||
112 filler(buf
, "stat", NULL
, 0) != 0 ||
113 filler(buf
, "uptime", NULL
, 0) != 0 ||
114 filler(buf
, "diskstats", NULL
, 0) != 0 ||
115 filler(buf
, "swaps", NULL
, 0) != 0 ||
116 filler(buf
, "loadavg", NULL
, 0) != 0)
122 static off_t
get_procfile_size(const char *which
)
124 FILE *f
= fopen(which
, "re");
127 ssize_t sz
, answer
= 0;
131 while ((sz
= getline(&line
, &len
, f
)) != -1)
139 int proc_open(const char *path
, struct fuse_file_info
*fi
)
141 __do_free
struct file_info
*info
= NULL
;
144 if (strcmp(path
, "/proc/meminfo") == 0)
145 type
= LXC_TYPE_PROC_MEMINFO
;
146 else if (strcmp(path
, "/proc/cpuinfo") == 0)
147 type
= LXC_TYPE_PROC_CPUINFO
;
148 else if (strcmp(path
, "/proc/uptime") == 0)
149 type
= LXC_TYPE_PROC_UPTIME
;
150 else if (strcmp(path
, "/proc/stat") == 0)
151 type
= LXC_TYPE_PROC_STAT
;
152 else if (strcmp(path
, "/proc/diskstats") == 0)
153 type
= LXC_TYPE_PROC_DISKSTATS
;
154 else if (strcmp(path
, "/proc/swaps") == 0)
155 type
= LXC_TYPE_PROC_SWAPS
;
156 else if (strcmp(path
, "/proc/loadavg") == 0)
157 type
= LXC_TYPE_PROC_LOADAVG
;
161 info
= malloc(sizeof(*info
));
165 memset(info
, 0, sizeof(*info
));
168 info
->buflen
= get_procfile_size(path
) + BUF_RESERVE_SIZE
;
170 info
->buf
= malloc(info
->buflen
);
174 memset(info
->buf
, 0, info
->buflen
);
175 /* set actual size to buffer size */
176 info
->size
= info
->buflen
;
178 fi
->fh
= PTR_TO_UINT64(move_ptr(info
));
182 int proc_access(const char *path
, int mask
)
184 if (strcmp(path
, "/proc") == 0 && access(path
, R_OK
) == 0)
187 /* these are all read-only */
188 if ((mask
& ~R_OK
) != 0)
193 int proc_release(const char *path
, struct fuse_file_info
*fi
)
195 do_release_file_info(fi
);
199 static unsigned long get_memlimit(const char *cgroup
, bool swap
)
202 __do_free
char *memlimit_str
= NULL
;
203 unsigned long memlimit
= -1;
206 ret
= cgroup_ops
->get_memory_swap_max(cgroup_ops
, cgroup
, &memlimit_str
);
208 ret
= cgroup_ops
->get_memory_max(cgroup_ops
, cgroup
, &memlimit_str
);
210 memlimit
= strtoul(memlimit_str
, NULL
, 10);
215 static unsigned long get_min_memlimit(const char *cgroup
, bool swap
)
217 __do_free
char *copy
= NULL
;
218 unsigned long memlimit
= 0;
219 unsigned long retlimit
;
221 copy
= strdup(cgroup
);
222 retlimit
= get_memlimit(copy
, swap
);
224 while (strcmp(copy
, "/") != 0) {
228 memlimit
= get_memlimit(it
, swap
);
229 if (memlimit
!= -1 && memlimit
< retlimit
)
236 static bool startswith(const char *line
, const char *pref
)
238 if (strncmp(line
, pref
, strlen(pref
)) == 0)
243 static int proc_swaps_read(char *buf
, size_t size
, off_t offset
,
244 struct fuse_file_info
*fi
)
246 __do_free
char *cg
= NULL
, *memswlimit_str
= NULL
, *memusage_str
= NULL
,
247 *memswusage_str
= NULL
;
248 struct fuse_context
*fc
= fuse_get_context();
249 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
250 unsigned long memswlimit
= 0, memlimit
= 0, memusage
= 0,
251 memswusage
= 0, swap_total
= 0, swap_free
= 0;
252 ssize_t total_len
= 0;
254 char *cache
= d
->buf
;
260 if (offset
> d
->size
)
266 left
= d
->size
- offset
;
267 total_len
= left
> size
? size
: left
;
268 memcpy(buf
, cache
+ offset
, total_len
);
273 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
274 if (initpid
<= 1 || is_shared_pidns(initpid
))
276 cg
= get_pid_cgroup(initpid
, "memory");
278 return read_file_fuse("/proc/swaps", buf
, size
, d
);
279 prune_init_slice(cg
);
281 memlimit
= get_min_memlimit(cg
, false);
283 ret
= cgroup_ops
->get_memory_current(cgroup_ops
, cg
, &memusage_str
);
287 memusage
= strtoul(memusage_str
, NULL
, 10);
289 ret
= cgroup_ops
->get_memory_swap_max(cgroup_ops
, cg
, &memswlimit_str
);
291 ret
= cgroup_ops
->get_memory_swap_current(cgroup_ops
, cg
, &memswusage_str
);
293 memswlimit
= get_min_memlimit(cg
, true);
294 memswusage
= strtoul(memswusage_str
, NULL
, 10);
295 swap_total
= (memswlimit
- memlimit
) / 1024;
296 swap_free
= (memswusage
- memusage
) / 1024;
299 total_len
= snprintf(d
->buf
, d
->size
, "Filename\t\t\t\tType\t\tSize\tUsed\tPriority\n");
301 /* When no mem + swap limit is specified or swapaccount=0*/
303 __do_free
char *line
= NULL
;
304 __do_free
void *fopen_cache
= NULL
;
305 __do_fclose
FILE *f
= NULL
;
308 f
= fopen_cached("/proc/meminfo", "re", &fopen_cache
);
312 while (getline(&line
, &linelen
, f
) != -1) {
313 if (startswith(line
, "SwapTotal:"))
314 sscanf(line
, "SwapTotal: %8lu kB", &swap_total
);
315 else if (startswith(line
, "SwapFree:"))
316 sscanf(line
, "SwapFree: %8lu kB", &swap_free
);
320 if (swap_total
> 0) {
321 l
= snprintf(d
->buf
+ total_len
, d
->size
- total_len
,
322 "none%*svirtual\t\t%lu\t%lu\t0\n", 36, " ",
323 swap_total
, swap_free
);
327 if (total_len
< 0 || l
< 0) {
328 perror("Error writing to cache");
333 d
->size
= (int)total_len
;
335 if (total_len
> size
) total_len
= size
;
336 memcpy(buf
, d
->buf
, total_len
);
340 static void get_blkio_io_value(char *str
, unsigned major
, unsigned minor
,
341 char *iotype
, unsigned long *v
)
347 snprintf(key
, 32, "%u:%u %s", major
, minor
, iotype
);
349 size_t len
= strlen(key
);
353 if (startswith(str
, key
)) {
354 sscanf(str
+ len
, "%lu", v
);
357 eol
= strchr(str
, '\n');
364 static int proc_diskstats_read(char *buf
, size_t size
, off_t offset
,
365 struct fuse_file_info
*fi
)
367 __do_free
char *cg
= NULL
, *io_serviced_str
= NULL
,
368 *io_merged_str
= NULL
, *io_service_bytes_str
= NULL
,
369 *io_wait_time_str
= NULL
, *io_service_time_str
= NULL
,
371 __do_free
void *fopen_cache
= NULL
;
372 __do_fclose
FILE *f
= NULL
;
373 struct fuse_context
*fc
= fuse_get_context();
374 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
375 unsigned long read
= 0, write
= 0;
376 unsigned long read_merged
= 0, write_merged
= 0;
377 unsigned long read_sectors
= 0, write_sectors
= 0;
378 unsigned long read_ticks
= 0, write_ticks
= 0;
379 unsigned long ios_pgr
= 0, tot_ticks
= 0, rq_ticks
= 0;
380 unsigned long rd_svctm
= 0, wr_svctm
= 0, rd_wait
= 0, wr_wait
= 0;
381 char *cache
= d
->buf
;
382 size_t cache_size
= d
->buflen
;
383 size_t linelen
= 0, total_len
= 0;
384 unsigned int major
= 0, minor
= 0;
392 if (offset
> d
->size
)
398 left
= d
->size
- offset
;
399 total_len
= left
> size
? size
: left
;
400 memcpy(buf
, cache
+ offset
, total_len
);
405 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
406 if (initpid
<= 1 || is_shared_pidns(initpid
))
408 cg
= get_pid_cgroup(initpid
, "blkio");
410 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
411 prune_init_slice(cg
);
413 ret
= cgroup_ops
->get_io_serviced(cgroup_ops
, cg
, &io_serviced_str
);
415 if (ret
== -EOPNOTSUPP
)
416 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
419 ret
= cgroup_ops
->get_io_merged(cgroup_ops
, cg
, &io_merged_str
);
421 if (ret
== -EOPNOTSUPP
)
422 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
425 ret
= cgroup_ops
->get_io_service_bytes(cgroup_ops
, cg
, &io_service_bytes_str
);
427 if (ret
== -EOPNOTSUPP
)
428 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
431 ret
= cgroup_ops
->get_io_wait_time(cgroup_ops
, cg
, &io_wait_time_str
);
433 if (ret
== -EOPNOTSUPP
)
434 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
437 ret
= cgroup_ops
->get_io_service_time(cgroup_ops
, cg
, &io_service_time_str
);
439 if (ret
== -EOPNOTSUPP
)
440 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
443 f
= fopen_cached("/proc/diskstats", "re", &fopen_cache
);
447 while (getline(&line
, &linelen
, f
) != -1) {
451 i
= sscanf(line
, "%u %u %71s", &major
, &minor
, dev_name
);
455 get_blkio_io_value(io_serviced_str
, major
, minor
, "Read", &read
);
456 get_blkio_io_value(io_serviced_str
, major
, minor
, "Write", &write
);
457 get_blkio_io_value(io_merged_str
, major
, minor
, "Read", &read_merged
);
458 get_blkio_io_value(io_merged_str
, major
, minor
, "Write", &write_merged
);
459 get_blkio_io_value(io_service_bytes_str
, major
, minor
, "Read", &read_sectors
);
460 read_sectors
= read_sectors
/512;
461 get_blkio_io_value(io_service_bytes_str
, major
, minor
, "Write", &write_sectors
);
462 write_sectors
= write_sectors
/512;
464 get_blkio_io_value(io_service_time_str
, major
, minor
, "Read", &rd_svctm
);
465 rd_svctm
= rd_svctm
/1000000;
466 get_blkio_io_value(io_wait_time_str
, major
, minor
, "Read", &rd_wait
);
467 rd_wait
= rd_wait
/1000000;
468 read_ticks
= rd_svctm
+ rd_wait
;
470 get_blkio_io_value(io_service_time_str
, major
, minor
, "Write", &wr_svctm
);
471 wr_svctm
= wr_svctm
/1000000;
472 get_blkio_io_value(io_wait_time_str
, major
, minor
, "Write", &wr_wait
);
473 wr_wait
= wr_wait
/1000000;
474 write_ticks
= wr_svctm
+ wr_wait
;
476 get_blkio_io_value(io_service_time_str
, major
, minor
, "Total", &tot_ticks
);
477 tot_ticks
= tot_ticks
/1000000;
479 memset(lbuf
, 0, 256);
480 if (read
|| write
|| read_merged
|| write_merged
|| read_sectors
|| write_sectors
|| read_ticks
|| write_ticks
)
481 snprintf(lbuf
, 256, "%u %u %s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
482 major
, minor
, dev_name
, read
, read_merged
, read_sectors
, read_ticks
,
483 write
, write_merged
, write_sectors
, write_ticks
, ios_pgr
, tot_ticks
, rq_ticks
);
487 l
= snprintf(cache
, cache_size
, "%s", lbuf
);
489 perror("Error writing to fuse buf");
492 if (l
>= cache_size
) {
493 lxcfs_error("%s\n", "Internal error: truncated write to cache.");
503 if (total_len
> size
) total_len
= size
;
504 memcpy(buf
, d
->buf
, total_len
);
510 static inline void iwashere(void)
512 mknod("/tmp/lxcfs-iwashere", S_IFREG
, 0644);
516 /* This function retrieves the busy time of a group of tasks by looking at
517 * cpuacct.usage. Unfortunately, this only makes sense when the container has
518 * been given it's own cpuacct cgroup. If not, this function will take the busy
519 * time of all other taks that do not actually belong to the container into
520 * account as well. If someone has a clever solution for this please send a
523 static double get_reaper_busy(pid_t task
)
525 __do_free
char *cgroup
= NULL
, *usage_str
= NULL
;
526 unsigned long usage
= 0;
529 initpid
= lookup_initpid_in_store(task
);
533 cgroup
= get_pid_cgroup(initpid
, "cpuacct");
536 prune_init_slice(cgroup
);
537 if (!cgroup_ops
->get(cgroup_ops
, "cpuacct", cgroup
, "cpuacct.usage",
541 usage
= strtoul(usage_str
, NULL
, 10);
542 return ((double)usage
/ 1000000000);
545 static uint64_t get_reaper_start_time(pid_t pid
)
547 __do_free
void *fopen_cache
= NULL
;
548 __do_fclose
FILE *f
= NULL
;
551 /* strlen("/proc/") = 6
555 * strlen("/stat") = 5
559 #define __PROC_PID_STAT_LEN (6 + LXCFS_NUMSTRLEN64 + 5 + 1)
560 char path
[__PROC_PID_STAT_LEN
];
563 qpid
= lookup_initpid_in_store(pid
);
565 /* Caller can check for EINVAL on 0. */
570 ret
= snprintf(path
, __PROC_PID_STAT_LEN
, "/proc/%d/stat", qpid
);
571 if (ret
< 0 || ret
>= __PROC_PID_STAT_LEN
) {
572 /* Caller can check for EINVAL on 0. */
577 f
= fopen_cached(path
, "re", &fopen_cache
);
579 /* Caller can check for EINVAL on 0. */
584 /* Note that the *scanf() argument supression requires that length
585 * modifiers such as "l" are omitted. Otherwise some compilers will yell
586 * at us. It's like telling someone you're not married and then asking
587 * if you can bring your wife to the party.
589 ret
= fscanf(f
, "%*d " /* (1) pid %d */
590 "%*s " /* (2) comm %s */
591 "%*c " /* (3) state %c */
592 "%*d " /* (4) ppid %d */
593 "%*d " /* (5) pgrp %d */
594 "%*d " /* (6) session %d */
595 "%*d " /* (7) tty_nr %d */
596 "%*d " /* (8) tpgid %d */
597 "%*u " /* (9) flags %u */
598 "%*u " /* (10) minflt %lu */
599 "%*u " /* (11) cminflt %lu */
600 "%*u " /* (12) majflt %lu */
601 "%*u " /* (13) cmajflt %lu */
602 "%*u " /* (14) utime %lu */
603 "%*u " /* (15) stime %lu */
604 "%*d " /* (16) cutime %ld */
605 "%*d " /* (17) cstime %ld */
606 "%*d " /* (18) priority %ld */
607 "%*d " /* (19) nice %ld */
608 "%*d " /* (20) num_threads %ld */
609 "%*d " /* (21) itrealvalue %ld */
610 "%" PRIu64
, /* (22) starttime %llu */
613 return ret_set_errno(0, EINVAL
);
615 return ret_set_errno(starttime
, 0);
618 static double get_reaper_start_time_in_sec(pid_t pid
)
620 uint64_t clockticks
, ticks_per_sec
;
624 clockticks
= get_reaper_start_time(pid
);
625 if (clockticks
== 0 && errno
== EINVAL
) {
626 lxcfs_debug("failed to retrieve start time of pid %d\n", pid
);
630 ret
= sysconf(_SC_CLK_TCK
);
631 if (ret
< 0 && errno
== EINVAL
) {
634 "failed to determine number of clock ticks in a second");
638 ticks_per_sec
= (uint64_t)ret
;
639 res
= (double)clockticks
/ ticks_per_sec
;
643 static double get_reaper_age(pid_t pid
)
646 double procstart
, procage
;
648 /* We need to substract the time the process has started since system
649 * boot minus the time when the system has started to get the actual
652 procstart
= get_reaper_start_time_in_sec(pid
);
656 struct timespec spec
;
658 ret
= clock_gettime(CLOCK_BOOTTIME
, &spec
);
662 /* We could make this more precise here by using the tv_nsec
663 * field in the timespec struct and convert it to milliseconds
664 * and then create a double for the seconds and milliseconds but
665 * that seems more work than it is worth.
667 uptime_ms
= (spec
.tv_sec
* 1000) + (spec
.tv_nsec
* 1e-6);
668 procage
= (uptime_ms
- (procstart
* 1000)) / 1000;
675 * We read /proc/uptime and reuse its second field.
676 * For the first field, we use the mtime for the reaper for
677 * the calling pid as returned by getreaperage
679 static int proc_uptime_read(char *buf
, size_t size
, off_t offset
,
680 struct fuse_file_info
*fi
)
682 struct fuse_context
*fc
= fuse_get_context();
683 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
684 double busytime
= get_reaper_busy(fc
->pid
);
685 char *cache
= d
->buf
;
686 ssize_t total_len
= 0;
687 double idletime
, reaperage
;
696 if (offset
> d
->size
)
698 int left
= d
->size
- offset
;
699 total_len
= left
> size
? size
: left
;
700 memcpy(buf
, cache
+ offset
, total_len
);
704 reaperage
= get_reaper_age(fc
->pid
);
705 /* To understand why this is done, please read the comment to the
706 * get_reaper_busy() function.
708 idletime
= reaperage
;
709 if (reaperage
>= busytime
)
710 idletime
= reaperage
- busytime
;
712 total_len
= snprintf(d
->buf
, d
->buflen
, "%.2lf %.2lf\n", reaperage
, idletime
);
713 if (total_len
< 0 || total_len
>= d
->buflen
){
714 lxcfs_error("%s\n", "failed to write to cache");
718 d
->size
= (int)total_len
;
721 if (total_len
> size
) total_len
= size
;
723 memcpy(buf
, d
->buf
, total_len
);
727 #define CPUALL_MAX_SIZE (BUF_RESERVE_SIZE / 2)
728 static int proc_stat_read(char *buf
, size_t size
, off_t offset
,
729 struct fuse_file_info
*fi
)
731 __do_free
char *cg
= NULL
, *cpuset
= NULL
, *line
= NULL
;
732 __do_free
void *fopen_cache
= NULL
;
733 __do_free
struct cpuacct_usage
*cg_cpu_usage
= NULL
;
734 __do_fclose
FILE *f
= NULL
;
735 struct fuse_context
*fc
= fuse_get_context();
736 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
737 size_t linelen
= 0, total_len
= 0;
738 int curcpu
= -1; /* cpu numbering starts at 0 */
740 unsigned long user
= 0, nice
= 0, system
= 0, idle
= 0, iowait
= 0,
741 irq
= 0, softirq
= 0, steal
= 0, guest
= 0, guest_nice
= 0;
742 unsigned long user_sum
= 0, nice_sum
= 0, system_sum
= 0, idle_sum
= 0,
743 iowait_sum
= 0, irq_sum
= 0, softirq_sum
= 0,
744 steal_sum
= 0, guest_sum
= 0, guest_nice_sum
= 0;
745 char cpuall
[CPUALL_MAX_SIZE
];
746 /* reserve for cpu all */
747 char *cache
= d
->buf
+ CPUALL_MAX_SIZE
;
748 size_t cache_size
= d
->buflen
- CPUALL_MAX_SIZE
;
749 int cg_cpu_usage_size
= 0;
752 if (offset
> d
->size
)
756 int left
= d
->size
- offset
;
757 total_len
= left
> size
? size
: left
;
758 memcpy(buf
, d
->buf
+ offset
, total_len
);
762 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
763 lxcfs_v("initpid: %d\n", initpid
);
768 * when container run with host pid namespace initpid == 1, cgroup will "/"
769 * we should return host os's /proc contents.
770 * in some case cpuacct_usage.all in "/" will larger then /proc/stat
773 return read_file_fuse("/proc/stat", buf
, size
, d
);
776 cg
= get_pid_cgroup(initpid
, "cpuset");
777 lxcfs_v("cg: %s\n", cg
);
779 return read_file_fuse("/proc/stat", buf
, size
, d
);
780 prune_init_slice(cg
);
782 cpuset
= get_cpuset(cg
);
787 * Read cpuacct.usage_all for all CPUs.
788 * If the cpuacct cgroup is present, it is used to calculate the container's
789 * CPU usage. If not, values from the host's /proc/stat are used.
791 if (read_cpuacct_usage_all(cg
, cpuset
, &cg_cpu_usage
, &cg_cpu_usage_size
) != 0)
792 lxcfs_v("%s\n", "proc_stat_read failed to read from cpuacct, falling back to the host's /proc/stat");
794 f
= fopen_cached("/proc/stat", "re", &fopen_cache
);
799 if (getline(&line
, &linelen
, f
) < 0) {
800 lxcfs_error("%s\n", "proc_stat_read read first line failed.");
804 if (cgroup_ops
->can_use_cpuview(cgroup_ops
) && cg_cpu_usage
) {
805 total_len
= cpuview_proc_stat(cg
, cpuset
, cg_cpu_usage
, cg_cpu_usage_size
,
806 f
, d
->buf
, d
->buflen
);
810 while (getline(&line
, &linelen
, f
) != -1) {
812 char cpu_char
[10]; /* That's a lot of cores */
814 uint64_t all_used
, cg_used
, new_idle
;
817 if (strlen(line
) == 0)
819 if (sscanf(line
, "cpu%9[^ ]", cpu_char
) != 1) {
820 /* not a ^cpuN line containing a number N, just print it */
821 l
= snprintf(cache
, cache_size
, "%s", line
);
823 perror("Error writing to cache");
826 if (l
>= cache_size
) {
827 lxcfs_error("%s\n", "Internal error: truncated write to cache.");
836 if (sscanf(cpu_char
, "%d", &physcpu
) != 1)
838 if (!cpu_in_cpuset(physcpu
, cpuset
))
842 ret
= sscanf(line
, "%*s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu",
854 if (ret
!= 10 || !cg_cpu_usage
) {
855 c
= strchr(line
, ' ');
858 l
= snprintf(cache
, cache_size
, "cpu%d%s", curcpu
, c
);
860 perror("Error writing to cache");
864 if (l
>= cache_size
) {
865 lxcfs_error("%s\n", "Internal error: truncated write to cache.");
878 if (physcpu
>= cg_cpu_usage_size
)
881 all_used
= user
+ nice
+ system
+ iowait
+ irq
+ softirq
+ steal
+ guest
+ guest_nice
;
882 cg_used
= cg_cpu_usage
[physcpu
].user
+ cg_cpu_usage
[physcpu
].system
;
884 if (all_used
>= cg_used
) {
885 new_idle
= idle
+ (all_used
- cg_used
);
888 lxcfs_error("cpu%d from %s has unexpected cpu time: %" PRIu64
" in /proc/stat, %" PRIu64
" in cpuacct.usage_all; unable to determine idle time",
889 curcpu
, cg
, all_used
, cg_used
);
893 l
= snprintf(cache
, cache_size
,
894 "cpu%d %" PRIu64
" 0 %" PRIu64
" %" PRIu64
" 0 0 0 0 0 0\n",
895 curcpu
, cg_cpu_usage
[physcpu
].user
,
896 cg_cpu_usage
[physcpu
].system
, new_idle
);
899 perror("Error writing to cache");
903 if (l
>= cache_size
) {
904 lxcfs_error("%s\n", "Internal error: truncated write to cache.");
912 user_sum
+= cg_cpu_usage
[physcpu
].user
;
913 system_sum
+= cg_cpu_usage
[physcpu
].system
;
914 idle_sum
+= new_idle
;
919 system_sum
+= system
;
921 iowait_sum
+= iowait
;
923 softirq_sum
+= softirq
;
926 guest_nice_sum
+= guest_nice
;
932 int cpuall_len
= snprintf(cpuall
, CPUALL_MAX_SIZE
, "cpu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
943 if (cpuall_len
> 0 && cpuall_len
< CPUALL_MAX_SIZE
) {
944 memcpy(cache
, cpuall
, cpuall_len
);
947 /* shouldn't happen */
948 lxcfs_error("proc_stat_read copy cpuall failed, cpuall_len=%d.", cpuall_len
);
952 memmove(cache
, d
->buf
+ CPUALL_MAX_SIZE
, total_len
);
953 total_len
+= cpuall_len
;
958 if (total_len
> size
)
961 memcpy(buf
, d
->buf
, total_len
);
965 /* Note that "memory.stat" in cgroup2 is hierarchical by default. */
966 static bool cgroup_parse_memory_stat(const char *cgroup
, struct memory_stat
*mstat
)
968 __do_close_prot_errno
int fd
= -EBADF
;
969 __do_fclose
FILE *f
= NULL
;
970 __do_free
char *line
= NULL
;
975 fd
= cgroup_ops
->get_memory_stats_fd(cgroup_ops
, cgroup
);
979 f
= fdopen(fd
, "re");
982 /* Transferring ownership to fdopen(). */
985 unified
= pure_unified_layout(cgroup_ops
);
986 while ((linelen
= getline(&line
, &len
, f
)) != -1) {
987 if (!unified
&& startswith(line
, "hierarchical_memory_limit")) {
988 sscanf(line
, "hierarchical_memory_limit %" PRIu64
, &(mstat
->hierarchical_memory_limit
));
989 } else if (!unified
&& startswith(line
, "hierarchical_memsw_limit")) {
990 sscanf(line
, "hierarchical_memsw_limit %" PRIu64
, &(mstat
->hierarchical_memsw_limit
));
991 } else if (startswith(line
, unified
? "file" :"total_cache")) {
992 sscanf(line
, unified
? "file %" PRIu64
: "total_cache %" PRIu64
, &(mstat
->total_cache
));
993 } else if (!unified
&& startswith(line
, "total_rss")) {
994 sscanf(line
, "total_rss %" PRIu64
, &(mstat
->total_rss
));
995 } else if (!unified
&& startswith(line
, "total_rss_huge")) {
996 sscanf(line
, "total_rss_huge %" PRIu64
, &(mstat
->total_rss_huge
));
997 } else if (startswith(line
, unified
? "shmem" : "total_shmem")) {
998 sscanf(line
, unified
? "shmem %" PRIu64
: "total_shmem %" PRIu64
, &(mstat
->total_shmem
));
999 } else if (startswith(line
, unified
? "file_mapped" : "total_mapped_file")) {
1000 sscanf(line
, unified
? "file_mapped %" PRIu64
: "total_mapped_file %" PRIu64
, &(mstat
->total_mapped_file
));
1001 } else if (!unified
&& startswith(line
, "total_dirty")) {
1002 sscanf(line
, "total_dirty %" PRIu64
, &(mstat
->total_dirty
));
1003 } else if (!unified
&& startswith(line
, "total_writeback")) {
1004 sscanf(line
, "total_writeback %" PRIu64
, &(mstat
->total_writeback
));
1005 } else if (!unified
&& startswith(line
, "total_swap")) {
1006 sscanf(line
, "total_swap %" PRIu64
, &(mstat
->total_swap
));
1007 } else if (!unified
&& startswith(line
, "total_pgpgin")) {
1008 sscanf(line
, "total_pgpgin %" PRIu64
, &(mstat
->total_pgpgin
));
1009 } else if (!unified
&& startswith(line
, "total_pgpgout")) {
1010 sscanf(line
, "total_pgpgout %" PRIu64
, &(mstat
->total_pgpgout
));
1011 } else if (startswith(line
, unified
? "pgfault" : "total_pgfault")) {
1012 sscanf(line
, unified
? "pgfault %" PRIu64
: "total_pgfault %" PRIu64
, &(mstat
->total_pgfault
));
1013 } else if (startswith(line
, unified
? "pgmajfault" : "total_pgmajfault")) {
1014 sscanf(line
, unified
? "pgmajfault %" PRIu64
: "total_pgmajfault %" PRIu64
, &(mstat
->total_pgmajfault
));
1015 } else if (startswith(line
, unified
? "inactive_anon" : "total_inactive_anon")) {
1016 sscanf(line
, unified
? "inactive_anon %" PRIu64
: "total_inactive_anon %" PRIu64
, &(mstat
->total_inactive_anon
));
1017 } else if (startswith(line
, unified
? "active_anon" : "total_active_anon")) {
1018 sscanf(line
, unified
? "active_anon %" PRIu64
: "total_active_anon %" PRIu64
, &(mstat
->total_active_anon
));
1019 } else if (startswith(line
, unified
? "inactive_file" : "total_inactive_file")) {
1020 sscanf(line
, unified
? "inactive_file %" PRIu64
: "total_inactive_file %" PRIu64
, &(mstat
->total_inactive_file
));
1021 } else if (startswith(line
, unified
? "active_file" : "total_active_file")) {
1022 sscanf(line
, unified
? "active_file %" PRIu64
: "total_active_file %" PRIu64
, &(mstat
->total_active_file
));
1023 } else if (startswith(line
, unified
? "unevictable" : "total_unevictable")) {
1024 sscanf(line
, unified
? "unevictable %" PRIu64
: "total_unevictable %" PRIu64
, &(mstat
->total_unevictable
));
1031 static int proc_meminfo_read(char *buf
, size_t size
, off_t offset
,
1032 struct fuse_file_info
*fi
)
1034 __do_free
char *cgroup
= NULL
, *line
= NULL
,
1035 *memusage_str
= NULL
, *memstat_str
= NULL
,
1036 *memswlimit_str
= NULL
, *memswusage_str
= NULL
;
1037 __do_free
void *fopen_cache
= NULL
;
1038 __do_fclose
FILE *f
= NULL
;
1039 struct fuse_context
*fc
= fuse_get_context();
1040 struct lxcfs_opts
*opts
= (struct lxcfs_opts
*) fuse_get_context()->private_data
;
1041 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
1042 uint64_t memlimit
= 0, memusage
= 0, memswlimit
= 0, memswusage
= 0,
1044 struct memory_stat mstat
= {};
1045 size_t linelen
= 0, total_len
= 0;
1046 char *cache
= d
->buf
;
1047 size_t cache_size
= d
->buflen
;
1053 if (offset
> d
->size
)
1059 left
= d
->size
- offset
;
1060 total_len
= left
> size
? size
: left
;
1061 memcpy(buf
, cache
+ offset
, total_len
);
1066 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
1067 if (initpid
<= 1 || is_shared_pidns(initpid
))
1070 cgroup
= get_pid_cgroup(initpid
, "memory");
1072 return read_file_fuse("/proc/meminfo", buf
, size
, d
);
1074 prune_init_slice(cgroup
);
1076 memlimit
= get_min_memlimit(cgroup
, false);
1078 ret
= cgroup_ops
->get_memory_current(cgroup_ops
, cgroup
, &memusage_str
);
1082 if (!cgroup_parse_memory_stat(cgroup
, &mstat
))
1086 * Following values are allowed to fail, because swapaccount might be
1087 * turned off for current kernel.
1089 ret
= cgroup_ops
->get_memory_swap_max(cgroup_ops
, cgroup
, &memswlimit_str
);
1091 ret
= cgroup_ops
->get_memory_swap_current(cgroup_ops
, cgroup
, &memswusage_str
);
1093 memswlimit
= get_min_memlimit(cgroup
, true);
1094 memswusage
= strtoul(memswusage_str
, NULL
, 10);
1095 memswlimit
= memswlimit
/ 1024;
1096 memswusage
= memswusage
/ 1024;
1099 memusage
= strtoul(memusage_str
, NULL
, 10);
1103 f
= fopen_cached("/proc/meminfo", "re", &fopen_cache
);
1107 while (getline(&line
, &linelen
, f
) != -1) {
1109 char *printme
, lbuf
[100];
1111 memset(lbuf
, 0, 100);
1112 if (startswith(line
, "MemTotal:")) {
1113 sscanf(line
+sizeof("MemTotal:")-1, "%" PRIu64
, &hosttotal
);
1114 if (hosttotal
< memlimit
)
1115 memlimit
= hosttotal
;
1116 snprintf(lbuf
, 100, "MemTotal: %8" PRIu64
" kB\n", memlimit
);
1118 } else if (startswith(line
, "MemFree:")) {
1119 snprintf(lbuf
, 100, "MemFree: %8" PRIu64
" kB\n", memlimit
- memusage
);
1121 } else if (startswith(line
, "MemAvailable:")) {
1122 snprintf(lbuf
, 100, "MemAvailable: %8" PRIu64
" kB\n", memlimit
- memusage
+ mstat
.total_cache
/ 1024);
1124 } else if (startswith(line
, "SwapTotal:") && memswlimit
> 0 &&
1125 opts
&& opts
->swap_off
== false) {
1126 memswlimit
-= memlimit
;
1127 snprintf(lbuf
, 100, "SwapTotal: %8" PRIu64
" kB\n", memswlimit
);
1129 } else if (startswith(line
, "SwapTotal:") && opts
&& opts
->swap_off
== true) {
1130 snprintf(lbuf
, 100, "SwapTotal: %8" PRIu64
" kB\n", (uint64_t)0);
1132 } else if (startswith(line
, "SwapFree:") && memswlimit
> 0 &&
1133 memswusage
> 0 && opts
&& opts
->swap_off
== false) {
1134 uint64_t swaptotal
= memswlimit
,
1135 swapusage
= memusage
> memswusage
1137 : memswusage
- memusage
,
1138 swapfree
= swapusage
< swaptotal
1139 ? swaptotal
- swapusage
1141 snprintf(lbuf
, 100, "SwapFree: %8" PRIu64
" kB\n", swapfree
);
1143 } else if (startswith(line
, "SwapFree:") && opts
&& opts
->swap_off
== true) {
1144 snprintf(lbuf
, 100, "SwapFree: %8" PRIu64
" kB\n", (uint64_t)0);
1146 } else if (startswith(line
, "Slab:")) {
1147 snprintf(lbuf
, 100, "Slab: %8" PRIu64
" kB\n", (uint64_t)0);
1149 } else if (startswith(line
, "Buffers:")) {
1150 snprintf(lbuf
, 100, "Buffers: %8" PRIu64
" kB\n", (uint64_t)0);
1152 } else if (startswith(line
, "Cached:")) {
1153 snprintf(lbuf
, 100, "Cached: %8" PRIu64
" kB\n",
1154 mstat
.total_cache
/ 1024);
1156 } else if (startswith(line
, "SwapCached:")) {
1157 snprintf(lbuf
, 100, "SwapCached: %8" PRIu64
" kB\n", (uint64_t)0);
1159 } else if (startswith(line
, "Active:")) {
1160 snprintf(lbuf
, 100, "Active: %8" PRIu64
" kB\n",
1161 (mstat
.total_active_anon
+
1162 mstat
.total_active_file
) /
1165 } else if (startswith(line
, "Inactive:")) {
1166 snprintf(lbuf
, 100, "Inactive: %8" PRIu64
" kB\n",
1167 (mstat
.total_inactive_anon
+
1168 mstat
.total_inactive_file
) /
1171 } else if (startswith(line
, "Active(anon)")) {
1172 snprintf(lbuf
, 100, "Active(anon): %8" PRIu64
" kB\n",
1173 mstat
.total_active_anon
/ 1024);
1175 } else if (startswith(line
, "Inactive(anon)")) {
1176 snprintf(lbuf
, 100, "Inactive(anon): %8" PRIu64
" kB\n",
1177 mstat
.total_inactive_anon
/ 1024);
1179 } else if (startswith(line
, "Active(file)")) {
1180 snprintf(lbuf
, 100, "Active(file): %8" PRIu64
" kB\n",
1181 mstat
.total_active_file
/ 1024);
1183 } else if (startswith(line
, "Inactive(file)")) {
1184 snprintf(lbuf
, 100, "Inactive(file): %8" PRIu64
" kB\n",
1185 mstat
.total_inactive_file
/ 1024);
1187 } else if (startswith(line
, "Unevictable")) {
1188 snprintf(lbuf
, 100, "Unevictable: %8" PRIu64
" kB\n",
1189 mstat
.total_unevictable
/ 1024);
1191 } else if (startswith(line
, "Dirty")) {
1192 snprintf(lbuf
, 100, "Dirty: %8" PRIu64
" kB\n",
1193 mstat
.total_dirty
/ 1024);
1195 } else if (startswith(line
, "Writeback")) {
1196 snprintf(lbuf
, 100, "Writeback: %8" PRIu64
" kB\n",
1197 mstat
.total_writeback
/ 1024);
1199 } else if (startswith(line
, "AnonPages")) {
1200 snprintf(lbuf
, 100, "AnonPages: %8" PRIu64
" kB\n",
1201 (mstat
.total_active_anon
+
1202 mstat
.total_inactive_anon
- mstat
.total_shmem
) /
1205 } else if (startswith(line
, "Mapped")) {
1206 snprintf(lbuf
, 100, "Mapped: %8" PRIu64
" kB\n",
1207 mstat
.total_mapped_file
/ 1024);
1209 } else if (startswith(line
, "SReclaimable")) {
1210 snprintf(lbuf
, 100, "SReclaimable: %8" PRIu64
" kB\n", (uint64_t)0);
1212 } else if (startswith(line
, "SUnreclaim")) {
1213 snprintf(lbuf
, 100, "SUnreclaim: %8" PRIu64
" kB\n", (uint64_t)0);
1215 } else if (startswith(line
, "Shmem:")) {
1216 snprintf(lbuf
, 100, "Shmem: %8" PRIu64
" kB\n",
1217 mstat
.total_shmem
/ 1024);
1219 } else if (startswith(line
, "ShmemHugePages")) {
1220 snprintf(lbuf
, 100, "ShmemHugePages: %8" PRIu64
" kB\n", (uint64_t)0);
1222 } else if (startswith(line
, "ShmemPmdMapped")) {
1223 snprintf(lbuf
, 100, "ShmemPmdMapped: %8" PRIu64
" kB\n", (uint64_t)0);
1225 } else if (startswith(line
, "AnonHugePages")) {
1226 snprintf(lbuf
, 100, "AnonHugePages: %8" PRIu64
" kB\n",
1227 mstat
.total_rss_huge
/ 1024);
1233 l
= snprintf(cache
, cache_size
, "%s", printme
);
1235 perror("Error writing to cache");
1239 if (l
>= cache_size
) {
1240 lxcfs_error("%s\n", "Internal error: truncated write to cache.");
1250 d
->size
= total_len
;
1251 if (total_len
> size
) total_len
= size
;
1252 memcpy(buf
, d
->buf
, total_len
);
1257 int proc_read(const char *path
, char *buf
, size_t size
, off_t offset
,
1258 struct fuse_file_info
*fi
)
1260 struct file_info
*f
= INTTYPE_TO_PTR(fi
->fh
);
1263 case LXC_TYPE_PROC_MEMINFO
:
1264 return proc_meminfo_read(buf
, size
, offset
, fi
);
1265 case LXC_TYPE_PROC_CPUINFO
:
1266 return proc_cpuinfo_read(buf
, size
, offset
, fi
);
1267 case LXC_TYPE_PROC_UPTIME
:
1268 return proc_uptime_read(buf
, size
, offset
, fi
);
1269 case LXC_TYPE_PROC_STAT
:
1270 return proc_stat_read(buf
, size
, offset
, fi
);
1271 case LXC_TYPE_PROC_DISKSTATS
:
1272 return proc_diskstats_read(buf
, size
, offset
, fi
);
1273 case LXC_TYPE_PROC_SWAPS
:
1274 return proc_swaps_read(buf
, size
, offset
, fi
);
1275 case LXC_TYPE_PROC_LOADAVG
:
1276 return proc_loadavg_read(buf
, size
, offset
, fi
);