1 /* SPDX-License-Identifier: LGPL-2.1+ */
21 #include <linux/magic.h>
22 #include <linux/sched.h>
23 #include <sys/epoll.h>
25 #include <sys/mount.h>
26 #include <sys/param.h>
27 #include <sys/socket.h>
28 #include <sys/syscall.h>
29 #include <sys/sysinfo.h>
32 #include "proc_fuse.h"
35 #include "cgroup_fuse.h"
36 #include "cgroups/cgroup.h"
37 #include "cgroups/cgroup_utils.h"
38 #include "cpuset_parse.h"
39 #include "lxcfs_fuse_compat.h"
40 #include "memory_utils.h"
41 #include "proc_loadavg.h"
42 #include "proc_cpuview.h"
46 uint64_t hierarchical_memory_limit
;
47 uint64_t hierarchical_memsw_limit
;
50 uint64_t total_rss_huge
;
52 uint64_t total_mapped_file
;
54 uint64_t total_writeback
;
56 uint64_t total_pgpgin
;
57 uint64_t total_pgpgout
;
58 uint64_t total_pgfault
;
59 uint64_t total_pgmajfault
;
60 uint64_t total_inactive_anon
;
61 uint64_t total_active_anon
;
62 uint64_t total_inactive_file
;
63 uint64_t total_active_file
;
64 uint64_t total_unevictable
;
67 static off_t
get_procfile_size(const char *path
)
69 __do_fclose
FILE *f
= NULL
;
70 __do_free
char *line
= NULL
;
72 ssize_t sz
, answer
= 0;
74 f
= fopen(path
, "re");
78 while ((sz
= getline(&line
, &len
, f
)) != -1)
84 __lxcfs_fuse_ops
int proc_getattr(const char *path
, struct stat
*sb
)
88 memset(sb
, 0, sizeof(struct stat
));
89 if (clock_gettime(CLOCK_REALTIME
, &now
) < 0)
92 sb
->st_uid
= sb
->st_gid
= 0;
93 sb
->st_atim
= sb
->st_mtim
= sb
->st_ctim
= now
;
94 if (strcmp(path
, "/proc") == 0) {
95 sb
->st_mode
= S_IFDIR
| 00555;
100 if (strcmp(path
, "/proc/meminfo") == 0 ||
101 strcmp(path
, "/proc/cpuinfo") == 0 ||
102 strcmp(path
, "/proc/uptime") == 0 ||
103 strcmp(path
, "/proc/stat") == 0 ||
104 strcmp(path
, "/proc/diskstats") == 0 ||
105 strcmp(path
, "/proc/swaps") == 0 ||
106 strcmp(path
, "/proc/loadavg") == 0 ||
107 strcmp(path
, "/proc/slabinfo") == 0) {
108 sb
->st_size
= get_procfile_size(path
);
109 sb
->st_mode
= S_IFREG
| 00444;
117 __lxcfs_fuse_ops
int proc_readdir(const char *path
, void *buf
,
118 fuse_fill_dir_t filler
, off_t offset
,
119 struct fuse_file_info
*fi
)
121 if (DIR_FILLER(filler
, buf
, ".", NULL
, 0) != 0 ||
122 DIR_FILLER(filler
, buf
, "..", NULL
, 0) != 0 ||
123 DIR_FILLER(filler
, buf
, "cpuinfo", NULL
, 0) != 0 ||
124 DIR_FILLER(filler
, buf
, "meminfo", NULL
, 0) != 0 ||
125 DIR_FILLER(filler
, buf
, "stat", NULL
, 0) != 0 ||
126 DIR_FILLER(filler
, buf
, "uptime", NULL
, 0) != 0 ||
127 DIR_FILLER(filler
, buf
, "diskstats", NULL
, 0) != 0 ||
128 DIR_FILLER(filler
, buf
, "swaps", NULL
, 0) != 0 ||
129 DIR_FILLER(filler
, buf
, "loadavg", NULL
, 0) != 0 ||
130 DIR_FILLER(filler
, buf
, "slabinfo", NULL
, 0) != 0)
136 __lxcfs_fuse_ops
int proc_open(const char *path
, struct fuse_file_info
*fi
)
138 __do_free
struct file_info
*info
= NULL
;
141 if (strcmp(path
, "/proc/meminfo") == 0)
142 type
= LXC_TYPE_PROC_MEMINFO
;
143 else if (strcmp(path
, "/proc/cpuinfo") == 0)
144 type
= LXC_TYPE_PROC_CPUINFO
;
145 else if (strcmp(path
, "/proc/uptime") == 0)
146 type
= LXC_TYPE_PROC_UPTIME
;
147 else if (strcmp(path
, "/proc/stat") == 0)
148 type
= LXC_TYPE_PROC_STAT
;
149 else if (strcmp(path
, "/proc/diskstats") == 0)
150 type
= LXC_TYPE_PROC_DISKSTATS
;
151 else if (strcmp(path
, "/proc/swaps") == 0)
152 type
= LXC_TYPE_PROC_SWAPS
;
153 else if (strcmp(path
, "/proc/loadavg") == 0)
154 type
= LXC_TYPE_PROC_LOADAVG
;
155 else if (strcmp(path
, "/proc/slabinfo") == 0)
156 type
= LXC_TYPE_PROC_SLABINFO
;
160 info
= zalloc(sizeof(*info
));
166 info
->buflen
= get_procfile_size(path
) + BUF_RESERVE_SIZE
;
168 info
->buf
= zalloc(info
->buflen
);
171 /* set actual size to buffer size */
172 info
->size
= info
->buflen
;
174 fi
->fh
= PTR_TO_UINT64(move_ptr(info
));
178 __lxcfs_fuse_ops
int proc_access(const char *path
, int mask
)
180 if (strcmp(path
, "/proc") == 0 && access(path
, R_OK
) == 0)
183 /* these are all read-only */
184 if ((mask
& ~R_OK
) != 0)
190 __lxcfs_fuse_ops
int proc_release(const char *path
, struct fuse_file_info
*fi
)
192 do_release_file_info(fi
);
196 static uint64_t get_memlimit(const char *cgroup
, bool swap
)
198 __do_free
char *memlimit_str
= NULL
;
199 uint64_t memlimit
= 0;
203 ret
= cgroup_ops
->get_memory_swap_max(cgroup_ops
, cgroup
, &memlimit_str
);
205 ret
= cgroup_ops
->get_memory_max(cgroup_ops
, cgroup
, &memlimit_str
);
206 if (ret
> 0 && memlimit_str
[0] && safe_uint64(memlimit_str
, &memlimit
, 10) < 0)
207 lxcfs_error("Failed to convert memlimit %s", memlimit_str
);
213 * This function taken from glibc-2.32, as POSIX dirname("/some-dir") will
214 * return "/some-dir" as opposed to "/", which breaks `get_min_memlimit()`
216 static char *gnu_dirname(char *path
)
218 static const char dot
[] = ".";
222 last_slash
= path
!= NULL
? strrchr(path
, '/') : NULL
;
224 if (last_slash
!= NULL
&& last_slash
!= path
&& last_slash
[1] == '\0') {
225 /* Determine whether all remaining characters are slashes. */
228 for (runp
= last_slash
; runp
!= path
; --runp
)
232 /* The '/' is the last character, we have to look further. */
234 last_slash
= memrchr(path
, '/', runp
- path
);
237 if (last_slash
!= NULL
) {
238 /* Determine whether all remaining characters are slashes. */
241 for (runp
= last_slash
; runp
!= path
; --runp
)
245 /* Terminate the path. */
248 * The last slash is the first character in the string.
249 * We have to return "/". As a special case we have to
250 * return "//" if there are exactly two slashes at the
251 * beginning of the string. See XBD 4.10 Path Name
252 * Resolution for more information
254 if (last_slash
== path
+ 1)
257 last_slash
= path
+ 1;
261 last_slash
[0] = '\0';
264 * This assignment is ill-designed but the XPG specs require to
265 * return a string containing "." in any case no directory part
266 * is found and so a static and constant string is required.
274 static uint64_t get_min_memlimit(const char *cgroup
, bool swap
)
276 __do_free
char *copy
= NULL
;
277 uint64_t memlimit
= 0, retlimit
= 0;
279 copy
= strdup(cgroup
);
281 return log_error_errno(0, ENOMEM
, "Failed to allocate memory");
283 retlimit
= get_memlimit(copy
, swap
);
286 * If the cgroup doesn't start with / (probably won't happen), dirname()
287 * will terminate with "" instead of "/"
289 while (*copy
&& strcmp(copy
, "/") != 0) {
292 it
= gnu_dirname(it
);
293 memlimit
= get_memlimit(it
, swap
);
294 if (memlimit
> 0 && memlimit
< retlimit
)
301 static inline bool startswith(const char *line
, const char *pref
)
303 return strncmp(line
, pref
, strlen(pref
)) == 0;
306 static int proc_swaps_read(char *buf
, size_t size
, off_t offset
,
307 struct fuse_file_info
*fi
)
309 __do_free
char *cgroup
= NULL
, *memusage_str
= NULL
,
310 *memswusage_str
= NULL
, *memswpriority_str
= NULL
;
311 struct fuse_context
*fc
= fuse_get_context();
312 bool wants_swap
= lxcfs_has_opt(fuse_get_context()->private_data
, LXCFS_SWAP_ON
);
313 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
314 uint64_t memswlimit
= 0, memlimit
= 0, memusage
= 0, memswusage
= 0,
315 swtotal
= 0, swusage
= 0, memswpriority
= 1,
316 hostswtotal
= 0, hostswfree
= 0;
317 ssize_t total_len
= 0;
319 char *cache
= d
->buf
;
321 __do_free
char *line
= NULL
;
322 __do_free
void *fopen_cache
= NULL
;
323 __do_fclose
FILE *f
= NULL
;
329 if (offset
> d
->size
)
335 left
= d
->size
- offset
;
336 total_len
= left
> size
? size
: left
;
337 memcpy(buf
, cache
+ offset
, total_len
);
342 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
343 if (initpid
<= 1 || is_shared_pidns(initpid
))
346 cgroup
= get_pid_cgroup(initpid
, "memory");
348 return read_file_fuse("/proc/swaps", buf
, size
, d
);
349 prune_init_slice(cgroup
);
351 memlimit
= get_min_memlimit(cgroup
, false);
353 ret
= cgroup_ops
->get_memory_current(cgroup_ops
, cgroup
, &memusage_str
);
357 if (safe_uint64(memusage_str
, &memusage
, 10) < 0)
358 lxcfs_error("Failed to convert memusage %s", memusage_str
);
361 memswlimit
= get_min_memlimit(cgroup
, true);
362 if (memswlimit
> 0) {
363 ret
= cgroup_ops
->get_memory_swap_current(cgroup_ops
, cgroup
, &memswusage_str
);
364 if (ret
>= 0 && safe_uint64(memswusage_str
, &memswusage
, 10) == 0) {
365 if (memlimit
> memswlimit
)
368 swtotal
= (memswlimit
- memlimit
) / 1024;
369 if (memusage
> memswusage
|| swtotal
== 0)
372 swusage
= (memswusage
- memusage
) / 1024;
375 ret
= cgroup_ops
->get_memory_swappiness(cgroup_ops
, cgroup
, &memswpriority_str
);
377 safe_uint64(memswpriority_str
, &memswpriority
, 10);
381 total_len
= snprintf(d
->buf
, d
->size
, "Filename\t\t\t\tType\t\tSize\tUsed\tPriority\n");
383 /* Read host total and free values */
384 f
= fopen_cached("/proc/meminfo", "re", &fopen_cache
);
388 while (getline(&line
, &linelen
, f
) != -1) {
389 if (startswith(line
, "SwapTotal:"))
390 sscanf(line
, "SwapTotal: %8" PRIu64
" kB", &hostswtotal
);
391 else if (startswith(line
, "SwapFree:"))
392 sscanf(line
, "SwapFree: %8" PRIu64
" kB", &hostswfree
);
396 /* The total amount of swap is always reported to be the
397 lesser of the RAM+SWAP limit or the SWAP device size.
398 This is because the kernel can swap as much as it
399 wants and not only up to swtotal. */
400 swtotal
= memlimit
/ 1024 + swtotal
;
401 if (hostswtotal
< swtotal
) {
402 swtotal
= hostswtotal
;
405 /* When swappiness is 0, pretend we can't swap. */
406 if (memswpriority
== 0) {
412 l
= snprintf(d
->buf
+ total_len
, d
->size
- total_len
,
413 "none%*svirtual\t\t%" PRIu64
"\t%" PRIu64
"\t0\n",
414 36, " ", swtotal
, swusage
);
418 if (total_len
< 0 || l
< 0)
419 return log_error(0, "Failed writing to cache");
422 d
->size
= (int)total_len
;
424 if ((size_t)total_len
> size
)
426 memcpy(buf
, d
->buf
, total_len
);
431 static void get_blkio_io_value(char *str
, unsigned major
, unsigned minor
,
432 char *iotype
, uint64_t *v
)
439 snprintf(key
, 32, "%u:%u %s", major
, minor
, iotype
);
444 if (startswith(str
, key
)) {
445 sscanf(str
+ len
, "%lu", v
);
448 eol
= strchr(str
, '\n');
455 struct lxcfs_diskstats
{
456 unsigned int major
; /* 1 - major number */
457 unsigned int minor
; /* 2 - minor mumber */
458 char dev_name
[72]; /* 3 - device name */
459 uint64_t read
; /* 4 - reads completed successfully */
460 uint64_t read_merged
; /* 5 - reads merged */
461 uint64_t read_sectors
; /* 6 - sectors read */
462 uint64_t read_ticks
; /* 7 - time spent reading (ms) */
463 uint64_t write
; /* 8 - writes completed */
464 uint64_t write_merged
; /* 9 - writes merged */
465 uint64_t write_sectors
; /* 10 - sectors written */
466 uint64_t write_ticks
; /* 11 - time spent writing (ms) */
467 uint64_t ios_pgr
; /* 12 - I/Os currently in progress */
468 uint64_t total_ticks
; /* 13 - time spent doing I/Os (ms) */
469 uint64_t rq_ticks
; /* 14 - weighted time spent doing I/Os (ms) */
470 uint64_t discard
; /* 15 - discards completed successfully (4.18+) */
471 uint64_t discard_merged
; /* 16 - discards merged (4.18+) */
472 uint64_t discard_sectors
; /* 17 - sectors discarded (4.18+) */
473 uint64_t discard_ticks
; /* 18 - time spent discarding (4.18+) */
476 static int proc_diskstats_read(char *buf
, size_t size
, off_t offset
,
477 struct fuse_file_info
*fi
)
479 __do_free
char *cg
= NULL
, *io_serviced_str
= NULL
,
480 *io_merged_str
= NULL
, *io_service_bytes_str
= NULL
,
481 *io_wait_time_str
= NULL
, *io_service_time_str
= NULL
,
483 __do_free
void *fopen_cache
= NULL
;
484 __do_fclose
FILE *f
= NULL
;
485 struct fuse_context
*fc
= fuse_get_context();
486 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
487 struct lxcfs_diskstats stats
= {};
489 uint64_t read_service_time
, write_service_time
, discard_service_time
, read_wait_time
,
490 write_wait_time
, discard_wait_time
;
491 char *cache
= d
->buf
;
492 size_t cache_size
= d
->buflen
;
493 size_t linelen
= 0, total_len
= 0;
500 if (offset
> d
->size
)
506 left
= d
->size
- offset
;
507 total_len
= left
> size
? size
: left
;
508 memcpy(buf
, cache
+ offset
, total_len
);
513 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
514 if (initpid
<= 1 || is_shared_pidns(initpid
))
517 cg
= get_pid_cgroup(initpid
, "blkio");
519 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
520 prune_init_slice(cg
);
522 ret
= cgroup_ops
->get_io_serviced(cgroup_ops
, cg
, &io_serviced_str
);
524 if (ret
== -EOPNOTSUPP
)
525 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
528 ret
= cgroup_ops
->get_io_merged(cgroup_ops
, cg
, &io_merged_str
);
530 if (ret
== -EOPNOTSUPP
)
531 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
534 ret
= cgroup_ops
->get_io_service_bytes(cgroup_ops
, cg
, &io_service_bytes_str
);
536 if (ret
== -EOPNOTSUPP
)
537 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
540 ret
= cgroup_ops
->get_io_wait_time(cgroup_ops
, cg
, &io_wait_time_str
);
542 if (ret
== -EOPNOTSUPP
)
543 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
546 ret
= cgroup_ops
->get_io_service_time(cgroup_ops
, cg
, &io_service_time_str
);
548 if (ret
== -EOPNOTSUPP
)
549 return read_file_fuse("/proc/diskstats", buf
, size
, d
);
552 f
= fopen_cached("/proc/diskstats", "re", &fopen_cache
);
556 while (getline(&line
, &linelen
, f
) != -1) {
560 i
= sscanf(line
, "%u %u %71s", &stats
.major
, &stats
.minor
, stats
.dev_name
);
564 get_blkio_io_value(io_serviced_str
, stats
.major
, stats
.minor
, "Read", &stats
.read
);
565 get_blkio_io_value(io_serviced_str
, stats
.major
, stats
.minor
, "Write", &stats
.write
);
566 get_blkio_io_value(io_serviced_str
, stats
.major
, stats
.minor
, "Discard", &stats
.discard
);
568 get_blkio_io_value(io_merged_str
, stats
.major
, stats
.minor
, "Read", &stats
.read_merged
);
569 get_blkio_io_value(io_merged_str
, stats
.major
, stats
.minor
, "Write", &stats
.write_merged
);
570 get_blkio_io_value(io_merged_str
, stats
.major
, stats
.minor
, "Discard", &stats
.discard_merged
);
572 get_blkio_io_value(io_service_bytes_str
, stats
.major
, stats
.minor
, "Read", &stats
.read_sectors
);
573 stats
.read_sectors
= stats
.read_sectors
/ 512;
574 get_blkio_io_value(io_service_bytes_str
, stats
.major
, stats
.minor
, "Write", &stats
.write_sectors
);
575 stats
.write_sectors
= stats
.write_sectors
/ 512;
576 get_blkio_io_value(io_service_bytes_str
, stats
.major
, stats
.minor
, "Discard", &stats
.discard_sectors
);
577 stats
.discard_sectors
= stats
.discard_sectors
/ 512;
579 get_blkio_io_value(io_service_time_str
, stats
.major
, stats
.minor
, "Read", &read_service_time
);
580 read_service_time
= read_service_time
/ 1000000;
581 get_blkio_io_value(io_wait_time_str
, stats
.major
, stats
.minor
, "Read", &read_wait_time
);
582 read_wait_time
= read_wait_time
/ 1000000;
583 stats
.read_ticks
= read_service_time
+ read_wait_time
;
585 get_blkio_io_value(io_service_time_str
, stats
.major
, stats
.minor
, "Write", &write_service_time
);
586 write_service_time
= write_service_time
/ 1000000;
587 get_blkio_io_value(io_wait_time_str
, stats
.major
, stats
.minor
, "Write", &write_wait_time
);
588 write_wait_time
= write_wait_time
/ 1000000;
589 stats
.write_ticks
= write_service_time
+ write_wait_time
;
591 get_blkio_io_value(io_service_time_str
, stats
.major
, stats
.minor
, "Discard", &discard_service_time
);
592 discard_service_time
= discard_service_time
/ 1000000;
593 get_blkio_io_value(io_wait_time_str
, stats
.major
, stats
.minor
, "Discard", &discard_wait_time
);
594 discard_wait_time
= discard_wait_time
/ 1000000;
595 stats
.discard_ticks
= discard_service_time
+ discard_wait_time
;
597 get_blkio_io_value(io_service_time_str
, stats
.major
, stats
.minor
, "Total", &stats
.total_ticks
);
598 stats
.total_ticks
= stats
.total_ticks
/ 1000000;
600 memset(lbuf
, 0, 256);
601 if (stats
.read
|| stats
.write
|| stats
.read_merged
|| stats
.write_merged
||
602 stats
.read_sectors
|| stats
.write_sectors
|| stats
.read_ticks
||
603 stats
.write_ticks
|| stats
.ios_pgr
|| stats
.total_ticks
|| stats
.rq_ticks
||
604 stats
.discard_merged
|| stats
.discard_sectors
|| stats
.discard_ticks
)
605 snprintf(lbuf
, 256, "%u %u %s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
620 stats
.discard_merged
,
621 stats
.discard_sectors
,
622 stats
.discard_ticks
);
626 l
= snprintf(cache
, cache_size
, "%s", lbuf
);
628 return log_error(0, "Failed to write cache");
629 if ((size_t)l
>= cache_size
)
630 return log_error(0, "Write to cache was truncated");
639 if (total_len
> size
)
641 memcpy(buf
, d
->buf
, total_len
);
647 static inline void iwashere(void)
649 mknod("/tmp/lxcfs-iwashere", S_IFREG
, 0644);
654 * This function retrieves the busy time of a group of tasks by looking at
655 * cpuacct.usage. Unfortunately, this only makes sense when the container has
656 * been given it's own cpuacct cgroup. If not, this function will take the busy
657 * time of all other taks that do not actually belong to the container into
658 * account as well. If someone has a clever solution for this please send a
661 static double get_reaper_busy(pid_t task
)
663 __do_free
char *cgroup
= NULL
, *usage_str
= NULL
;
667 initpid
= lookup_initpid_in_store(task
);
671 cgroup
= get_pid_cgroup(initpid
, "cpuacct");
674 prune_init_slice(cgroup
);
676 if (!cgroup_ops
->get(cgroup_ops
, "cpuacct", cgroup
, "cpuacct.usage", &usage_str
))
679 if (safe_uint64(usage_str
, &usage
, 10) < 0)
680 lxcfs_error("Failed to convert usage %s", usage_str
);
682 return ((double)usage
/ 1000000000);
685 static uint64_t get_reaper_start_time(pid_t pid
)
687 __do_free
void *fopen_cache
= NULL
;
688 __do_fclose
FILE *f
= NULL
;
691 char path
[STRLITERALLEN("/proc/") + LXCFS_NUMSTRLEN64
+
692 STRLITERALLEN("/stat") + 1];
695 qpid
= lookup_initpid_in_store(pid
);
697 return ret_errno(EINVAL
);
699 ret
= snprintf(path
, sizeof(path
), "/proc/%d/stat", qpid
);
700 if (ret
< 0 || (size_t)ret
>= sizeof(path
))
701 return ret_errno(EINVAL
);
703 f
= fopen_cached(path
, "re", &fopen_cache
);
705 return ret_errno(EINVAL
);
707 /* Note that the *scanf() argument supression requires that length
708 * modifiers such as "l" are omitted. Otherwise some compilers will yell
709 * at us. It's like telling someone you're not married and then asking
710 * if you can bring your wife to the party.
712 ret
= fscanf(f
, "%*d " /* (1) pid %d */
713 "%*s " /* (2) comm %s */
714 "%*c " /* (3) state %c */
715 "%*d " /* (4) ppid %d */
716 "%*d " /* (5) pgrp %d */
717 "%*d " /* (6) session %d */
718 "%*d " /* (7) tty_nr %d */
719 "%*d " /* (8) tpgid %d */
720 "%*u " /* (9) flags %u */
721 "%*u " /* (10) minflt %lu */
722 "%*u " /* (11) cminflt %lu */
723 "%*u " /* (12) majflt %lu */
724 "%*u " /* (13) cmajflt %lu */
725 "%*u " /* (14) utime %lu */
726 "%*u " /* (15) stime %lu */
727 "%*d " /* (16) cutime %ld */
728 "%*d " /* (17) cstime %ld */
729 "%*d " /* (18) priority %ld */
730 "%*d " /* (19) nice %ld */
731 "%*d " /* (20) num_threads %ld */
732 "%*d " /* (21) itrealvalue %ld */
733 "%" PRIu64
, /* (22) starttime %llu */
736 return ret_errno(EINVAL
);
738 return ret_set_errno(starttime
, 0);
741 static double get_reaper_start_time_in_sec(pid_t pid
)
743 uint64_t clockticks
, ticks_per_sec
;
747 clockticks
= get_reaper_start_time(pid
);
749 return log_debug(0, "Failed to retrieve start time of pid %d", pid
);
751 ret
= sysconf(_SC_CLK_TCK
);
753 return log_debug(0, "Failed to determine number of clock ticks in a second");
755 ticks_per_sec
= (uint64_t)ret
;
756 res
= (double)clockticks
/ ticks_per_sec
;
760 static double get_reaper_age(pid_t pid
)
763 double procstart
, procage
;
766 * We need to substract the time the process has started since system
767 * boot minus the time when the system has started to get the actual
770 procstart
= get_reaper_start_time_in_sec(pid
);
774 struct timespec spec
;
776 ret
= clock_gettime(CLOCK_BOOTTIME
, &spec
);
780 uptime_ms
= (spec
.tv_sec
* 1000) + (spec
.tv_nsec
* 1e-6);
781 procage
= (uptime_ms
- (procstart
* 1000)) / 1000;
788 * We read /proc/uptime and reuse its second field.
789 * For the first field, we use the mtime for the reaper for
790 * the calling pid as returned by getreaperage
792 static int proc_uptime_read(char *buf
, size_t size
, off_t offset
,
793 struct fuse_file_info
*fi
)
795 struct fuse_context
*fc
= fuse_get_context();
796 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
797 char *cache
= d
->buf
;
798 ssize_t total_len
= 0, ret
= 0;
799 double busytime
, idletime
, reaperage
;
808 if (offset
> d
->size
)
814 left
= d
->size
- offset
;
815 total_len
= left
> size
? size
: left
;
816 memcpy(buf
, cache
+ offset
, total_len
);
821 reaperage
= get_reaper_age(fc
->pid
);
823 * To understand why this is done, please read the comment to the
824 * get_reaper_busy() function.
826 idletime
= reaperage
;
827 busytime
= get_reaper_busy(fc
->pid
);
828 if (reaperage
>= busytime
)
829 idletime
= reaperage
- busytime
;
831 ret
= snprintf(d
->buf
, d
->buflen
, "%.2lf %.2lf\n", reaperage
, idletime
);
832 if (ret
< 0 || ret
>= d
->buflen
)
833 return read_file_fuse("/proc/uptime", buf
, size
, d
);
838 if ((size_t)total_len
> size
)
840 memcpy(buf
, d
->buf
, total_len
);
845 #define CPUALL_MAX_SIZE (BUF_RESERVE_SIZE / 2)
846 static int proc_stat_read(char *buf
, size_t size
, off_t offset
,
847 struct fuse_file_info
*fi
)
849 __do_free
char *cg
= NULL
, *cpuset
= NULL
, *line
= NULL
;
850 __do_free
void *fopen_cache
= NULL
;
851 __do_free
struct cpuacct_usage
*cg_cpu_usage
= NULL
;
852 __do_fclose
FILE *f
= NULL
;
853 struct fuse_context
*fc
= fuse_get_context();
854 struct lxcfs_opts
*opts
= (struct lxcfs_opts
*)fc
->private_data
;
855 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
856 size_t linelen
= 0, total_len
= 0;
857 int curcpu
= -1; /* cpu numbering starts at 0 */
859 uint64_t user
= 0, nice
= 0, system
= 0, idle
= 0, iowait
= 0, irq
= 0,
860 softirq
= 0, steal
= 0, guest
= 0, guest_nice
= 0;
861 uint64_t user_sum
= 0, nice_sum
= 0, system_sum
= 0, idle_sum
= 0,
862 iowait_sum
= 0, irq_sum
= 0, softirq_sum
= 0, steal_sum
= 0,
863 guest_sum
= 0, guest_nice_sum
= 0;
864 char cpuall
[CPUALL_MAX_SIZE
];
865 /* reserve for cpu all */
866 char *cache
= d
->buf
+ CPUALL_MAX_SIZE
;
867 size_t cache_size
= d
->buflen
- CPUALL_MAX_SIZE
;
868 int cg_cpu_usage_size
= 0;
873 if (offset
> d
->size
)
879 left
= d
->size
- offset
;
880 total_len
= left
> size
? size
: left
;
881 memcpy(buf
, d
->buf
+ offset
, total_len
);
886 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
887 if (initpid
<= 1 || is_shared_pidns(initpid
))
891 * when container run with host pid namespace initpid == 1, cgroup will "/"
892 * we should return host os's /proc contents.
893 * in some case cpuacct_usage.all in "/" will larger then /proc/stat
896 return read_file_fuse("/proc/stat", buf
, size
, d
);
898 cg
= get_pid_cgroup(initpid
, "cpuset");
900 return read_file_fuse("/proc/stat", buf
, size
, d
);
901 prune_init_slice(cg
);
903 cpuset
= get_cpuset(cg
);
907 f
= fopen_cached("/proc/stat", "re", &fopen_cache
);
911 /* Skip first system cpu line. */
912 if (getline(&line
, &linelen
, f
) < 0)
913 return log_error(0, "proc_stat_read read first line failed");
916 * Read cpuacct.usage_all for all CPUs.
917 * If the cpuacct cgroup is present, it is used to calculate the container's
918 * CPU usage. If not, values from the host's /proc/stat are used.
920 if (read_cpuacct_usage_all(cg
, cpuset
, &cg_cpu_usage
, &cg_cpu_usage_size
) == 0) {
921 if (cgroup_ops
->can_use_cpuview(cgroup_ops
) && opts
&& opts
->use_cfs
) {
922 total_len
= cpuview_proc_stat(cg
, cpuset
, cg_cpu_usage
,
923 cg_cpu_usage_size
, f
,
928 lxcfs_v("proc_stat_read failed to read from cpuacct, falling back to the host's /proc/stat");
931 while (getline(&line
, &linelen
, f
) != -1) {
933 char cpu_char
[10]; /* That's a lot of cores */
935 uint64_t all_used
, cg_used
, new_idle
;
938 if (strlen(line
) == 0)
940 if (sscanf(line
, "cpu%9[^ ]", cpu_char
) != 1) {
941 /* not a ^cpuN line containing a number N, just print it */
942 l
= snprintf(cache
, cache_size
, "%s", line
);
944 return log_error(0, "Failed to write cache");
945 if ((size_t)l
>= cache_size
)
946 return log_error(0, "Write to cache was truncated");
955 if (sscanf(cpu_char
, "%d", &physcpu
) != 1)
958 if (!cpu_in_cpuset(physcpu
, cpuset
))
963 ret
= sscanf(line
, "%*s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu",
974 if (ret
!= 10 || !cg_cpu_usage
) {
975 c
= strchr(line
, ' ');
979 l
= snprintf(cache
, cache_size
, "cpu%d%s", curcpu
, c
);
981 return log_error(0, "Failed to write cache");
982 if ((size_t)l
>= cache_size
)
983 return log_error(0, "Write to cache was truncated");
994 if (physcpu
>= cg_cpu_usage_size
)
997 all_used
= user
+ nice
+ system
+ iowait
+ irq
+ softirq
+ steal
+ guest
+ guest_nice
;
998 cg_used
= cg_cpu_usage
[physcpu
].user
+ cg_cpu_usage
[physcpu
].system
;
1000 if (all_used
>= cg_used
) {
1001 new_idle
= idle
+ (all_used
- cg_used
);
1003 lxcfs_debug("cpu%d from %s has unexpected cpu time: %" PRIu64
" in /proc/stat, %" PRIu64
" in cpuacct.usage_all; unable to determine idle time",
1004 curcpu
, cg
, all_used
, cg_used
);
1008 l
= snprintf(cache
, cache_size
,
1009 "cpu%d %" PRIu64
" 0 %" PRIu64
" %" PRIu64
" 0 0 0 0 0 0\n",
1010 curcpu
, cg_cpu_usage
[physcpu
].user
,
1011 cg_cpu_usage
[physcpu
].system
, new_idle
);
1013 return log_error(0, "Failed to write cache");
1014 if ((size_t)l
>= cache_size
)
1015 return log_error(0, "Write to cache was truncated");
1021 user_sum
+= cg_cpu_usage
[physcpu
].user
;
1022 system_sum
+= cg_cpu_usage
[physcpu
].system
;
1023 idle_sum
+= new_idle
;
1027 system_sum
+= system
;
1029 iowait_sum
+= iowait
;
1031 softirq_sum
+= softirq
;
1034 guest_nice_sum
+= guest_nice
;
1040 int cpuall_len
= snprintf(cpuall
, CPUALL_MAX_SIZE
, "cpu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
1051 if (cpuall_len
> 0 && cpuall_len
< CPUALL_MAX_SIZE
) {
1052 memcpy(cache
, cpuall
, cpuall_len
);
1053 cache
+= cpuall_len
;
1055 /* shouldn't happen */
1056 lxcfs_error("proc_stat_read copy cpuall failed, cpuall_len=%d", cpuall_len
);
1060 memmove(cache
, d
->buf
+ CPUALL_MAX_SIZE
, total_len
);
1061 total_len
+= cpuall_len
;
1065 d
->size
= total_len
;
1066 if (total_len
> size
)
1069 memcpy(buf
, d
->buf
, total_len
);
1073 /* Note that "memory.stat" in cgroup2 is hierarchical by default. */
1074 static bool cgroup_parse_memory_stat(const char *cgroup
, struct memory_stat
*mstat
)
1076 __do_close
int fd
= -EBADF
;
1077 __do_fclose
FILE *f
= NULL
;
1078 __do_free
char *line
= NULL
;
1079 __do_free
void *fdopen_cache
= NULL
;
1084 fd
= cgroup_ops
->get_memory_stats_fd(cgroup_ops
, cgroup
);
1088 f
= fdopen_cached(fd
, "re", &fdopen_cache
);
1092 unified
= pure_unified_layout(cgroup_ops
);
1093 while ((linelen
= getline(&line
, &len
, f
)) != -1) {
1094 if (!unified
&& startswith(line
, "hierarchical_memory_limit")) {
1095 sscanf(line
, "hierarchical_memory_limit %" PRIu64
, &(mstat
->hierarchical_memory_limit
));
1096 } else if (!unified
&& startswith(line
, "hierarchical_memsw_limit")) {
1097 sscanf(line
, "hierarchical_memsw_limit %" PRIu64
, &(mstat
->hierarchical_memsw_limit
));
1098 } else if (startswith(line
, unified
? "file" :"total_cache")) {
1099 sscanf(line
, unified
? "file %" PRIu64
: "total_cache %" PRIu64
, &(mstat
->total_cache
));
1100 } else if (!unified
&& startswith(line
, "total_rss")) {
1101 sscanf(line
, "total_rss %" PRIu64
, &(mstat
->total_rss
));
1102 } else if (!unified
&& startswith(line
, "total_rss_huge")) {
1103 sscanf(line
, "total_rss_huge %" PRIu64
, &(mstat
->total_rss_huge
));
1104 } else if (startswith(line
, unified
? "shmem" : "total_shmem")) {
1105 sscanf(line
, unified
? "shmem %" PRIu64
: "total_shmem %" PRIu64
, &(mstat
->total_shmem
));
1106 } else if (startswith(line
, unified
? "file_mapped" : "total_mapped_file")) {
1107 sscanf(line
, unified
? "file_mapped %" PRIu64
: "total_mapped_file %" PRIu64
, &(mstat
->total_mapped_file
));
1108 } else if (!unified
&& startswith(line
, "total_dirty")) {
1109 sscanf(line
, "total_dirty %" PRIu64
, &(mstat
->total_dirty
));
1110 } else if (!unified
&& startswith(line
, "total_writeback")) {
1111 sscanf(line
, "total_writeback %" PRIu64
, &(mstat
->total_writeback
));
1112 } else if (!unified
&& startswith(line
, "total_swap")) {
1113 sscanf(line
, "total_swap %" PRIu64
, &(mstat
->total_swap
));
1114 } else if (!unified
&& startswith(line
, "total_pgpgin")) {
1115 sscanf(line
, "total_pgpgin %" PRIu64
, &(mstat
->total_pgpgin
));
1116 } else if (!unified
&& startswith(line
, "total_pgpgout")) {
1117 sscanf(line
, "total_pgpgout %" PRIu64
, &(mstat
->total_pgpgout
));
1118 } else if (startswith(line
, unified
? "pgfault" : "total_pgfault")) {
1119 sscanf(line
, unified
? "pgfault %" PRIu64
: "total_pgfault %" PRIu64
, &(mstat
->total_pgfault
));
1120 } else if (startswith(line
, unified
? "pgmajfault" : "total_pgmajfault")) {
1121 sscanf(line
, unified
? "pgmajfault %" PRIu64
: "total_pgmajfault %" PRIu64
, &(mstat
->total_pgmajfault
));
1122 } else if (startswith(line
, unified
? "inactive_anon" : "total_inactive_anon")) {
1123 sscanf(line
, unified
? "inactive_anon %" PRIu64
: "total_inactive_anon %" PRIu64
, &(mstat
->total_inactive_anon
));
1124 } else if (startswith(line
, unified
? "active_anon" : "total_active_anon")) {
1125 sscanf(line
, unified
? "active_anon %" PRIu64
: "total_active_anon %" PRIu64
, &(mstat
->total_active_anon
));
1126 } else if (startswith(line
, unified
? "inactive_file" : "total_inactive_file")) {
1127 sscanf(line
, unified
? "inactive_file %" PRIu64
: "total_inactive_file %" PRIu64
, &(mstat
->total_inactive_file
));
1128 } else if (startswith(line
, unified
? "active_file" : "total_active_file")) {
1129 sscanf(line
, unified
? "active_file %" PRIu64
: "total_active_file %" PRIu64
, &(mstat
->total_active_file
));
1130 } else if (startswith(line
, unified
? "unevictable" : "total_unevictable")) {
1131 sscanf(line
, unified
? "unevictable %" PRIu64
: "total_unevictable %" PRIu64
, &(mstat
->total_unevictable
));
1138 static int proc_meminfo_read(char *buf
, size_t size
, off_t offset
,
1139 struct fuse_file_info
*fi
)
1141 __do_free
char *cgroup
= NULL
, *line
= NULL
, *memusage_str
= NULL
,
1142 *memswusage_str
= NULL
, *memswpriority_str
= NULL
;
1143 __do_free
void *fopen_cache
= NULL
;
1144 __do_fclose
FILE *f
= NULL
;
1145 struct fuse_context
*fc
= fuse_get_context();
1146 bool wants_swap
= lxcfs_has_opt(fuse_get_context()->private_data
, LXCFS_SWAP_ON
);
1147 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
1148 uint64_t memlimit
= 0, memusage
= 0, memswlimit
= 0, memswusage
= 0,
1149 hosttotal
= 0, swfree
= 0, swusage
= 0, swtotal
= 0,
1151 struct memory_stat mstat
= {};
1152 size_t linelen
= 0, total_len
= 0;
1153 char *cache
= d
->buf
;
1154 size_t cache_size
= d
->buflen
;
1160 if (offset
> d
->size
)
1166 left
= d
->size
- offset
;
1167 total_len
= left
> size
? size
: left
;
1168 memcpy(buf
, cache
+ offset
, total_len
);
1173 pid_t initpid
= lookup_initpid_in_store(fc
->pid
);
1174 if (initpid
<= 1 || is_shared_pidns(initpid
))
1177 cgroup
= get_pid_cgroup(initpid
, "memory");
1179 return read_file_fuse("/proc/meminfo", buf
, size
, d
);
1181 prune_init_slice(cgroup
);
1184 ret
= cgroup_ops
->get_memory_current(cgroup_ops
, cgroup
, &memusage_str
);
1186 return read_file_fuse("/proc/meminfo", buf
, size
, d
);
1188 if (safe_uint64(memusage_str
, &memusage
, 10) < 0)
1189 lxcfs_error("Failed to convert memusage %s", memusage_str
);
1191 if (!cgroup_parse_memory_stat(cgroup
, &mstat
))
1192 return read_file_fuse("/proc/meminfo", buf
, size
, d
);
1194 memlimit
= get_min_memlimit(cgroup
, false);
1197 * Following values are allowed to fail, because swapaccount might be
1198 * turned off for current kernel.
1201 memswlimit
= get_min_memlimit(cgroup
, true);
1202 if (memswlimit
> 0) {
1203 ret
= cgroup_ops
->get_memory_swap_current(cgroup_ops
, cgroup
, &memswusage_str
);
1204 if (ret
>= 0 && safe_uint64(memswusage_str
, &memswusage
, 10) == 0) {
1205 if (memlimit
> memswlimit
)
1208 swtotal
= (memswlimit
- memlimit
) / 1024;
1209 if (memusage
> memswusage
|| swtotal
== 0)
1212 swusage
= (memswusage
- memusage
) / 1024;
1216 ret
= cgroup_ops
->get_memory_swappiness(cgroup_ops
, cgroup
, &memswpriority_str
);
1218 safe_uint64(memswpriority_str
, &memswpriority
, 10);
1221 f
= fopen_cached("/proc/meminfo", "re", &fopen_cache
);
1223 return read_file_fuse("/proc/meminfo", buf
, size
, d
);
1227 while (getline(&line
, &linelen
, f
) != -1) {
1229 char *printme
, lbuf
[100];
1231 memset(lbuf
, 0, 100);
1232 if (startswith(line
, "MemTotal:")) {
1233 sscanf(line
+sizeof("MemTotal:")-1, "%" PRIu64
, &hosttotal
);
1235 memlimit
= hosttotal
;
1237 if (hosttotal
< memlimit
)
1238 memlimit
= hosttotal
;
1239 snprintf(lbuf
, 100, "MemTotal: %8" PRIu64
" kB\n", memlimit
);
1241 } else if (startswith(line
, "MemFree:")) {
1242 snprintf(lbuf
, 100, "MemFree: %8" PRIu64
" kB\n", memlimit
- memusage
);
1244 } else if (startswith(line
, "MemAvailable:")) {
1245 snprintf(lbuf
, 100, "MemAvailable: %8" PRIu64
" kB\n", memlimit
- memusage
+ mstat
.total_cache
/ 1024);
1247 } else if (startswith(line
, "SwapTotal:")) {
1249 uint64_t hostswtotal
= 0;
1251 sscanf(line
+ STRLITERALLEN("SwapTotal:"), "%" PRIu64
, &hostswtotal
);
1253 /* The total amount of swap is always reported to be the
1254 lesser of the RAM+SWAP limit or the SWAP device size.
1255 This is because the kernel can swap as much as it
1256 wants and not only up to swtotal. */
1258 swtotal
= memlimit
+ swtotal
;
1259 if (hostswtotal
< swtotal
) {
1260 swtotal
= hostswtotal
;
1263 /* When swappiness is 0, pretend we can't swap. */
1264 if (memswpriority
== 0) {
1269 snprintf(lbuf
, 100, "SwapTotal: %8" PRIu64
" kB\n", swtotal
);
1271 } else if (startswith(line
, "SwapFree:")) {
1273 swfree
= swtotal
- swusage
;
1276 snprintf(lbuf
, 100, "SwapFree: %8" PRIu64
" kB\n", swfree
);
1278 } else if (startswith(line
, "Slab:")) {
1279 snprintf(lbuf
, 100, "Slab: %8" PRIu64
" kB\n", (uint64_t)0);
1281 } else if (startswith(line
, "Buffers:")) {
1282 snprintf(lbuf
, 100, "Buffers: %8" PRIu64
" kB\n", (uint64_t)0);
1284 } else if (startswith(line
, "Cached:")) {
1285 snprintf(lbuf
, 100, "Cached: %8" PRIu64
" kB\n",
1286 mstat
.total_cache
/ 1024);
1288 } else if (startswith(line
, "SwapCached:")) {
1289 snprintf(lbuf
, 100, "SwapCached: %8" PRIu64
" kB\n", (uint64_t)0);
1291 } else if (startswith(line
, "Active:")) {
1292 snprintf(lbuf
, 100, "Active: %8" PRIu64
" kB\n",
1293 (mstat
.total_active_anon
+
1294 mstat
.total_active_file
) /
1297 } else if (startswith(line
, "Inactive:")) {
1298 snprintf(lbuf
, 100, "Inactive: %8" PRIu64
" kB\n",
1299 (mstat
.total_inactive_anon
+
1300 mstat
.total_inactive_file
) /
1303 } else if (startswith(line
, "Active(anon):")) {
1304 snprintf(lbuf
, 100, "Active(anon): %8" PRIu64
" kB\n",
1305 mstat
.total_active_anon
/ 1024);
1307 } else if (startswith(line
, "Inactive(anon):")) {
1308 snprintf(lbuf
, 100, "Inactive(anon): %8" PRIu64
" kB\n",
1309 mstat
.total_inactive_anon
/ 1024);
1311 } else if (startswith(line
, "Active(file):")) {
1312 snprintf(lbuf
, 100, "Active(file): %8" PRIu64
" kB\n",
1313 mstat
.total_active_file
/ 1024);
1315 } else if (startswith(line
, "Inactive(file):")) {
1316 snprintf(lbuf
, 100, "Inactive(file): %8" PRIu64
" kB\n",
1317 mstat
.total_inactive_file
/ 1024);
1319 } else if (startswith(line
, "Unevictable:")) {
1320 snprintf(lbuf
, 100, "Unevictable: %8" PRIu64
" kB\n",
1321 mstat
.total_unevictable
/ 1024);
1323 } else if (startswith(line
, "Dirty:")) {
1324 snprintf(lbuf
, 100, "Dirty: %8" PRIu64
" kB\n",
1325 mstat
.total_dirty
/ 1024);
1327 } else if (startswith(line
, "Writeback:")) {
1328 snprintf(lbuf
, 100, "Writeback: %8" PRIu64
" kB\n",
1329 mstat
.total_writeback
/ 1024);
1331 } else if (startswith(line
, "AnonPages:")) {
1332 snprintf(lbuf
, 100, "AnonPages: %8" PRIu64
" kB\n",
1333 (mstat
.total_active_anon
+
1334 mstat
.total_inactive_anon
- mstat
.total_shmem
) /
1337 } else if (startswith(line
, "Mapped:")) {
1338 snprintf(lbuf
, 100, "Mapped: %8" PRIu64
" kB\n",
1339 mstat
.total_mapped_file
/ 1024);
1341 } else if (startswith(line
, "SReclaimable:")) {
1342 snprintf(lbuf
, 100, "SReclaimable: %8" PRIu64
" kB\n", (uint64_t)0);
1344 } else if (startswith(line
, "SUnreclaim:")) {
1345 snprintf(lbuf
, 100, "SUnreclaim: %8" PRIu64
" kB\n", (uint64_t)0);
1347 } else if (startswith(line
, "Shmem:")) {
1348 snprintf(lbuf
, 100, "Shmem: %8" PRIu64
" kB\n",
1349 mstat
.total_shmem
/ 1024);
1351 } else if (startswith(line
, "ShmemHugePages:")) {
1352 snprintf(lbuf
, 100, "ShmemHugePages: %8" PRIu64
" kB\n", (uint64_t)0);
1354 } else if (startswith(line
, "ShmemPmdMapped:")) {
1355 snprintf(lbuf
, 100, "ShmemPmdMapped: %8" PRIu64
" kB\n", (uint64_t)0);
1357 } else if (startswith(line
, "AnonHugePages:")) {
1358 snprintf(lbuf
, 100, "AnonHugePages: %8" PRIu64
" kB\n",
1359 mstat
.total_rss_huge
/ 1024);
1365 l
= snprintf(cache
, cache_size
, "%s", printme
);
1367 return log_error(0, "Failed to write cache");
1368 if ((size_t)l
>= cache_size
)
1369 return log_error(0, "Write to cache was truncated");
1377 d
->size
= total_len
;
1378 if (total_len
> size
)
1380 memcpy(buf
, d
->buf
, total_len
);
1385 static int proc_slabinfo_read(char *buf
, size_t size
, off_t offset
,
1386 struct fuse_file_info
*fi
)
1388 __do_free
char *cgroup
= NULL
, *line
= NULL
;
1389 __do_free
void *fopen_cache
= NULL
;
1390 __do_fclose
FILE *f
= NULL
;
1391 __do_close
int fd
= -EBADF
;
1392 struct fuse_context
*fc
= fuse_get_context();
1393 struct file_info
*d
= INTTYPE_TO_PTR(fi
->fh
);
1394 size_t linelen
= 0, total_len
= 0;
1395 char *cache
= d
->buf
;
1396 size_t cache_size
= d
->buflen
;
1402 if (offset
> d
->size
)
1408 left
= d
->size
- offset
;
1409 total_len
= left
> size
? size
: left
;
1410 memcpy(buf
, cache
+ offset
, total_len
);
1415 initpid
= lookup_initpid_in_store(fc
->pid
);
1416 if (initpid
<= 1 || is_shared_pidns(initpid
))
1419 cgroup
= get_pid_cgroup(initpid
, "memory");
1421 return read_file_fuse("/proc/slabinfo", buf
, size
, d
);
1423 prune_init_slice(cgroup
);
1425 fd
= cgroup_ops
->get_memory_slabinfo_fd(cgroup_ops
, cgroup
);
1427 return read_file_fuse("/proc/slabinfo", buf
, size
, d
);
1429 f
= fdopen_cached(fd
, "re", &fopen_cache
);
1431 return read_file_fuse("/proc/slabinfo", buf
, size
, d
);
1433 while (getline(&line
, &linelen
, f
) != -1) {
1434 ssize_t l
= snprintf(cache
, cache_size
, "%s", line
);
1436 return log_error(0, "Failed to write cache");
1437 if ((size_t)l
>= cache_size
)
1438 return log_error(0, "Write to cache was truncated");
1446 d
->size
= total_len
;
1447 if (total_len
> size
)
1449 memcpy(buf
, d
->buf
, total_len
);
1454 __lxcfs_fuse_ops
int proc_read(const char *path
, char *buf
, size_t size
,
1455 off_t offset
, struct fuse_file_info
*fi
)
1457 struct file_info
*f
= INTTYPE_TO_PTR(fi
->fh
);
1460 case LXC_TYPE_PROC_MEMINFO
:
1461 if (liblxcfs_functional())
1462 return proc_meminfo_read(buf
, size
, offset
, fi
);
1464 return read_file_fuse_with_offset(LXC_TYPE_PROC_MEMINFO_PATH
,
1465 buf
, size
, offset
, f
);
1466 case LXC_TYPE_PROC_CPUINFO
:
1467 if (liblxcfs_functional())
1468 return proc_cpuinfo_read(buf
, size
, offset
, fi
);
1470 return read_file_fuse_with_offset(LXC_TYPE_PROC_CPUINFO_PATH
,
1471 buf
, size
, offset
, f
);
1472 case LXC_TYPE_PROC_UPTIME
:
1473 if (liblxcfs_functional())
1474 return proc_uptime_read(buf
, size
, offset
, fi
);
1476 return read_file_fuse_with_offset(LXC_TYPE_PROC_UPTIME_PATH
,
1477 buf
, size
, offset
, f
);
1478 case LXC_TYPE_PROC_STAT
:
1479 if (liblxcfs_functional())
1480 return proc_stat_read(buf
, size
, offset
, fi
);
1482 return read_file_fuse_with_offset(LXC_TYPE_PROC_STAT_PATH
, buf
,
1484 case LXC_TYPE_PROC_DISKSTATS
:
1485 if (liblxcfs_functional())
1486 return proc_diskstats_read(buf
, size
, offset
, fi
);
1488 return read_file_fuse_with_offset(LXC_TYPE_PROC_DISKSTATS_PATH
,
1489 buf
, size
, offset
, f
);
1490 case LXC_TYPE_PROC_SWAPS
:
1491 if (liblxcfs_functional())
1492 return proc_swaps_read(buf
, size
, offset
, fi
);
1494 return read_file_fuse_with_offset(LXC_TYPE_PROC_SWAPS_PATH
, buf
,
1496 case LXC_TYPE_PROC_LOADAVG
:
1497 if (liblxcfs_functional())
1498 return proc_loadavg_read(buf
, size
, offset
, fi
);
1500 return read_file_fuse_with_offset(LXC_TYPE_PROC_LOADAVG_PATH
,
1501 buf
, size
, offset
, f
);
1502 case LXC_TYPE_PROC_SLABINFO
:
1503 if (liblxcfs_functional())
1504 return proc_slabinfo_read(buf
, size
, offset
, fi
);
1506 return read_file_fuse_with_offset(LXC_TYPE_PROC_SLABINFO_PATH
,
1507 buf
, size
, offset
, f
);