#define FUSE_USE_VERSION 26
-#include <stdio.h>
#include <dirent.h>
+#include <errno.h>
#include <fcntl.h>
#include <fuse.h>
-#include <unistd.h>
-#include <errno.h>
-#include <stdbool.h>
-#include <time.h>
-#include <string.h>
-#include <stdlib.h>
#include <libgen.h>
-#include <sched.h>
#include <pthread.h>
+#include <sched.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+#include <unistd.h>
+#include <wait.h>
#include <linux/sched.h>
+#include <sys/epoll.h>
+#include <sys/mman.h>
+#include <sys/mount.h>
#include <sys/param.h>
#include <sys/socket.h>
-#include <sys/mount.h>
-#include <sys/epoll.h>
-#include <wait.h>
+#include <sys/syscall.h>
#include "bindings.h"
-
#include "config.h" // for VERSION
+/* Define pivot_root() if missing from the C library */
+#ifndef HAVE_PIVOT_ROOT
+static int pivot_root(const char * new_root, const char * put_old)
+{
+#ifdef __NR_pivot_root
+return syscall(__NR_pivot_root, new_root, put_old);
+#else
+errno = ENOSYS;
+return -1;
+#endif
+}
+#else
+extern int pivot_root(const char * new_root, const char * put_old);
+#endif
+
+#ifdef DEBUG
+#define lxcfs_debug(format, ...) \
+ do { \
+ fprintf(stderr, "%s: %d: %s: " format, __FILE__, __LINE__, \
+ __func__, __VA_ARGS__); \
+ } while (false)
+#else
+#define lxcfs_debug(format, ...)
+#endif /* DEBUG */
+
enum {
LXC_TYPE_CGDIR,
LXC_TYPE_CGFILE,
}
}
+/* READ-ONLY after __constructor__ collect_and_mount_subsystems() has run.
+ * Number of hierarchies mounted. */
+static int num_hierarchies;
+
+/* READ-ONLY after __constructor__ collect_and_mount_subsystems() has run.
+ * Hierachies mounted {cpuset, blkio, ...}:
+ * Initialized via __constructor__ collect_and_mount_subsystems(). */
+static char **hierarchies;
+
+/* READ-ONLY after __constructor__ collect_and_mount_subsystems() has run.
+ * Open file descriptors:
+ * @fd_hierarchies[i] refers to cgroup @hierarchies[i]. They are mounted in a
+ * private mount namespace.
+ * Initialized via __constructor__ collect_and_mount_subsystems().
+ * @fd_hierarchies[i] can be used to perform file operations on the cgroup
+ * mounts and respective files in the private namespace even when located in
+ * another namespace using the *at() family of functions
+ * {openat(), fchownat(), ...}. */
+static int *fd_hierarchies;
+
static void unlock_mutex(pthread_mutex_t *l)
{
int ret;
snprintf(fnam, 100, "/proc/%d", e->initpid);
if (stat(fnam, &initsb) < 0)
return false;
-#if DEBUG
- fprintf(stderr, "comparing ctime %ld %ld for pid %d\n",
- e->ctime, initsb.st_ctime, e->initpid);
-#endif
+
+ lxcfs_debug("Comparing ctime %ld == %ld for pid %d.\n", e->ctime,
+ initsb.st_ctime, e->initpid);
+
if (e->ctime != initsb.st_ctime)
return false;
return true;
struct pidns_init_store *tmp;
int h;
-#if DEBUG
- fprintf(stderr, "remove_initpid: removing entry for %d\n", e->initpid);
-#endif
+ lxcfs_debug("Remove_initpid: removing entry for %d.\n", e->initpid);
+
h = HASH(e->ino);
if (pidns_hash_table[h] == e) {
pidns_hash_table[h] = e->next;
now = time(NULL);
if (now < last_prune + PURGE_SECS)
return;
-#if DEBUG
- fprintf(stderr, "pruning\n");
-#endif
+
+ lxcfs_debug("%s\n", "Pruning.");
+
last_prune = now;
threshold = now - 2 * PURGE_SECS;
for (i = 0; i < PIDNS_HASH_SIZE; i++) {
for (prev = NULL, e = pidns_hash_table[i]; e; ) {
if (e->lastcheck < threshold) {
-#if DEBUG
- fprintf(stderr, "Removing cached entry for %d\n", e->initpid);
-#endif
+
+ lxcfs_debug("Removing cached entry for %d.\n", e->initpid);
+
delme = e;
if (prev)
prev->next = e->next;
struct stat procsb;
int h;
-#if DEBUG
- fprintf(stderr, "save_initpid: adding entry for %d\n", pid);
-#endif
+ lxcfs_debug("Save_initpid: adding entry for %d.\n", pid);
+
snprintf(fpath, 100, "/proc/%d", pid);
if (stat(fpath, &procsb) < 0)
return;
return NULL;
}
-static int is_dir(const char *path)
+static int is_dir(const char *path, int fd)
{
struct stat statbuf;
- int ret = stat(path, &statbuf);
+ int ret = fstatat(fd, path, &statbuf, fd);
if (ret == 0 && S_ISDIR(statbuf.st_mode))
return 1;
return 0;
const char *s = haystack, *e;
size_t nlen = strlen(needle);
- while (*s && (e = index(s, ','))) {
+ while (*s && (e = strchr(s, ','))) {
if (nlen != e - s) {
s = e + 1;
continue;
{
int ret, fd, cfd;
size_t len;
- char *fnam, *tmpc = find_mounted_controller(controller, &cfd);
+ char *fnam, *tmpc;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return false;
- /* . + /cgroup + / + file + \0 */
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cgroup + / + file + \0
+ */
len = strlen(cgroup) + strlen(file) + 3;
fnam = alloca(len);
ret = snprintf(fnam, len, "%s%s/%s", *cgroup == '/' ? "." : "", cgroup, file);
{
int cfd;
size_t len;
- char *dirnam, *tmpc = find_mounted_controller(controller, &cfd);
+ char *dirnam, *tmpc;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return -EINVAL;
- /* . + /cg + \0 */
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cg + \0
+ */
len = strlen(cg) + 2;
dirnam = alloca(len);
snprintf(dirnam, len, "%s%s", *cg == '/' ? "." : "", cg);
return 0;
}
-static bool recursive_rmdir(const char *dirname, int fd)
+static bool recursive_rmdir(const char *dirname, int fd, const int cfd)
{
struct dirent *direntp;
DIR *dir;
dir = fdopendir(dupfd);
if (!dir) {
-#if DEBUG
- fprintf(stderr, "%s: failed to open %s: %s\n", __func__, dirname, strerror(errno));
-#endif
+ lxcfs_debug("Failed to open %s: %s.\n", dirname, strerror(errno));
+ close(dupfd);
return false;
}
struct stat mystat;
int rc;
- if (!direntp)
- break;
-
if (!strcmp(direntp->d_name, ".") ||
!strcmp(direntp->d_name, ".."))
continue;
continue;
}
- ret = fstatat(fd, pathname, &mystat, AT_SYMLINK_NOFOLLOW);
- if (ret) {
-#if DEBUG
- fprintf(stderr, "%s: failed to stat %s: %s\n", __func__, pathname, strerror(errno));
-#endif
+ rc = fstatat(cfd, pathname, &mystat, AT_SYMLINK_NOFOLLOW);
+ if (rc) {
+ lxcfs_debug("Failed to stat %s: %s.\n", pathname, strerror(errno));
continue;
}
- if (S_ISDIR(mystat.st_mode)) {
- if (!recursive_rmdir(pathname, fd)) {
-#if DEBUG
- fprintf(stderr, "Error removing %s\n", pathname);
-#endif
- }
- }
+ if (S_ISDIR(mystat.st_mode))
+ if (!recursive_rmdir(pathname, fd, cfd))
+ lxcfs_debug("Error removing %s.\n", pathname);
}
ret = true;
ret = false;
}
- if (unlinkat(fd, dirname, AT_REMOVEDIR) < 0) {
-#if DEBUG
- fprintf(stderr, "%s: failed to delete %s: %s\n", __func__, dirname, strerror(errno));
-#endif
+ if (unlinkat(cfd, dirname, AT_REMOVEDIR) < 0) {
+ lxcfs_debug("Failed to delete %s: %s.\n", dirname, strerror(errno));
ret = false;
}
- close(fd);
+
+ close(dupfd);
return ret;
}
{
int fd, cfd;
size_t len;
- char *dirnam, *tmpc = find_mounted_controller(controller, &cfd);
+ char *dirnam, *tmpc;
+ bool bret;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return false;
- /* . + /cg + \0 */
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cg + \0
+ */
len = strlen(cg) + 2;
dirnam = alloca(len);
snprintf(dirnam, len, "%s%s", *cg == '/' ? "." : "", cg);
if (fd < 0)
return false;
- return recursive_rmdir(dirnam, fd);
+ bret = recursive_rmdir(dirnam, fd, cfd);
+ close(fd);
+ return bret;
}
bool cgfs_chmod_file(const char *controller, const char *file, mode_t mode)
{
int cfd;
size_t len;
- char *pathname, *tmpc = find_mounted_controller(controller, &cfd);
+ char *pathname, *tmpc;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return false;
- /* . + /file + \0 */
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /file + \0
+ */
len = strlen(file) + 2;
pathname = alloca(len);
snprintf(pathname, len, "%s%s", *file == '/' ? "." : "", file);
return true;
}
-static int chown_tasks_files(const char *dirname, uid_t uid, gid_t gid)
+static int chown_tasks_files(const char *dirname, uid_t uid, gid_t gid, int fd)
{
size_t len;
char *fname;
len = strlen(dirname) + strlen("/cgroup.procs") + 1;
fname = alloca(len);
snprintf(fname, len, "%s/tasks", dirname);
- if (chown(fname, uid, gid) != 0)
+ if (fchownat(fd, fname, uid, gid, 0) != 0)
return -errno;
snprintf(fname, len, "%s/cgroup.procs", dirname);
- if (chown(fname, uid, gid) != 0)
+ if (fchownat(fd, fname, uid, gid, 0) != 0)
return -errno;
return 0;
}
{
int cfd;
size_t len;
- char *pathname, *tmpc = find_mounted_controller(controller, &cfd);
+ char *pathname, *tmpc;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return -EINVAL;
- /* BASEDIR / tmpc / file \0 */
- len = strlen(BASEDIR) + strlen(tmpc) + strlen(file) + 3;
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /file + \0
+ */
+ len = strlen(file) + 2;
pathname = alloca(len);
- snprintf(pathname, len, "%s/%s/%s", BASEDIR, tmpc, file);
- if (chown(pathname, uid, gid) < 0)
+ snprintf(pathname, len, "%s%s", *file == '/' ? "." : "", file);
+ if (fchownat(cfd, pathname, uid, gid, 0) < 0)
return -errno;
- if (is_dir(pathname))
+ if (is_dir(pathname, cfd))
// like cgmanager did, we want to chown the tasks file as well
- return chown_tasks_files(pathname, uid, gid);
+ return chown_tasks_files(pathname, uid, gid, cfd);
return 0;
}
FILE *open_pids_file(const char *controller, const char *cgroup)
{
- int cfd;
+ int fd, cfd;
size_t len;
- char *pathname, *tmpc = find_mounted_controller(controller, &cfd);
+ char *pathname, *tmpc;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return NULL;
- /* BASEDIR / tmpc / cgroup / "cgroup.procs" \0 */
- len = strlen(BASEDIR) + strlen(tmpc) + strlen(cgroup) + 4 + strlen("cgroup.procs");
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cgroup + / "cgroup.procs" + \0
+ */
+ len = strlen(cgroup) + strlen("cgroup.procs") + 3;
pathname = alloca(len);
- snprintf(pathname, len, "%s/%s/%s/cgroup.procs", BASEDIR, tmpc, cgroup);
- return fopen(pathname, "w");
+ snprintf(pathname, len, "%s%s/cgroup.procs", *cgroup == '/' ? "." : "", cgroup);
+
+ fd = openat(cfd, pathname, O_WRONLY);
+ if (fd < 0)
+ return NULL;
+
+ return fdopen(fd, "w");
}
static bool cgfs_iterate_cgroup(const char *controller, const char *cgroup, bool directories,
if (!tmpc)
return false;
- /* Make sure we pass a relative path to openat(). */
+ /* Make sure we pass a relative path to *at() family of functions. */
len = strlen(cgroup) + 1 /* . */ + 1 /* \0 */;
cg = alloca(len);
ret = snprintf(cg, len, "%s%s", *cgroup == '/' ? "." : "", cgroup);
{
int ret, fd, cfd;
size_t len;
- char *fnam, *tmpc = find_mounted_controller(controller, &cfd);
+ char *fnam, *tmpc;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return false;
- /* . + /cgroup + / + file + \0 */
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cgroup + / + file + \0
+ */
len = strlen(cgroup) + strlen(file) + 3;
fnam = alloca(len);
ret = snprintf(fnam, len, "%s%s/%s", *cgroup == '/' ? "." : "", cgroup, file);
{
int ret, cfd;
size_t len;
- char *fnam, *tmpc = find_mounted_controller(controller, &cfd);
+ char *fnam, *tmpc;
struct stat sb;
struct cgfs_files *newkey;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return false;
if (file && *file == '/')
file++;
- if (file && index(file, '/'))
+ if (file && strchr(file, '/'))
return NULL;
- /* . + /cgroup + / + file + \0 */
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cgroup + / + file + \0
+ */
len = strlen(cgroup) + 3;
if (file)
len += strlen(file) + 1;
} while (!newkey);
if (file)
newkey->name = must_copy_string(file);
- else if (rindex(cgroup, '/'))
- newkey->name = must_copy_string(rindex(cgroup, '/'));
+ else if (strrchr(cgroup, '/'))
+ newkey->name = must_copy_string(strrchr(cgroup, '/'));
else
newkey->name = must_copy_string(cgroup);
newkey->uid = sb.st_uid;
{
int cfd;
size_t len;
- char *fnam, *tmpc = find_mounted_controller(controller, &cfd);
+ char *fnam, *tmpc;
int ret;
struct stat sb;
+ tmpc = find_mounted_controller(controller, &cfd);
if (!tmpc)
return false;
- /* . + /cgroup + / + f + \0 */
+
+ /* Make sure we pass a relative path to *at() family of functions.
+ * . + /cgroup + / + f + \0
+ */
len = strlen(cgroup) + strlen(f) + 3;
fnam = alloca(len);
ret = snprintf(fnam, len, "%s%s/%s", *cgroup == '/' ? "." : "", cgroup, f);
ret = fstatat(cfd, fnam, &sb, 0);
if (ret < 0 || !S_ISDIR(sb.st_mode))
return false;
+
return true;
}
return NULL;
}
- if (strcmp(querycg, "/") == 0)
+ if ((strcmp(querycg, "/") == 0) || (strcmp(querycg, "./") == 0))
start = strdup(taskcg + 1);
else
start = strdup(taskcg + strlen(querycg) + 1);
const char *p1;
char *contr, *slash;
- if (strlen(path) < 9)
+ if (strlen(path) < 9) {
+ errno = EACCES;
return NULL;
- if (*(path+7) != '/')
+ }
+ if (*(path + 7) != '/') {
+ errno = EINVAL;
return NULL;
- p1 = path+8;
+ }
+ p1 = path + 8;
contr = strdupa(p1);
- if (!contr)
+ if (!contr) {
+ errno = ENOMEM;
return NULL;
+ }
slash = strstr(contr, "/");
if (slash)
*slash = '\0';
int i;
- for (i = 0; i < num_hierarchies; i++) {
+ for (i = 0; i < num_hierarchies; i++) {
if (hierarchies[i] && strcmp(hierarchies[i], contr) == 0)
return hierarchies[i];
}
+ errno = ENOENT;
return NULL;
}
{
const char *p1;
- if (strlen(path) < 9)
+ if (strlen(path) < 9) {
+ errno = EACCES;
return NULL;
- p1 = strstr(path+8, "/");
- if (!p1)
+ }
+ p1 = strstr(path + 8, "/");
+ if (!p1) {
+ errno = EINVAL;
return NULL;
- return p1+1;
+ }
+ errno = 0;
+ return p1 + 1;
}
/*
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EIO;
+ return -errno;
cgroup = find_cgroup_in_path(path);
if (!cgroup) {
/* this is just /cgroup/controller, return it as a dir */
// return list of keys for the controller, and list of child cgroups
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EIO;
+ return -errno;
cgroup = find_cgroup_in_path(path);
if (!cgroup) {
struct fuse_context *fc = fuse_get_context();
char **clist = NULL;
+ if (filler(buf, ".", NULL, 0) != 0 || filler(buf, "..", NULL, 0) != 0)
+ return -EIO;
+
if (d->type != LXC_TYPE_CGDIR) {
fprintf(stderr, "Internal error: file cache info used in readdir\n");
return -EIO;
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EIO;
+ return -errno;
cgroup = find_cgroup_in_path(path);
if (!cgroup)
- return -EINVAL;
+ return -errno;
get_cgdir_and_path(cgroup, &cgdir, &last);
if (!last) {
int cg_access(const char *path, int mode)
{
+ int ret;
const char *cgroup;
- char *last = NULL, *path1, *path2, * cgdir = NULL, *controller;
+ char *path1, *path2, *controller;
+ char *last = NULL, *cgdir = NULL;
struct cgfs_files *k = NULL;
struct fuse_context *fc = fuse_get_context();
- int ret;
+
+ if (strcmp(path, "/cgroup") == 0)
+ return 0;
if (!fc)
return -EIO;
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EIO;
+ return -errno;
cgroup = find_cgroup_in_path(path);
if (!cgroup) {
// access("/sys/fs/cgroup/systemd", mode) - rx allowed, w not
return -EIO;
if (strcmp(path, "/cgroup") == 0)
- return -EINVAL;
+ return -EPERM;
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EINVAL;
+ return errno == ENOENT ? -EPERM : -errno;
+
cgroup = find_cgroup_in_path(path);
if (!cgroup)
/* this is just /cgroup/controller */
- return -EINVAL;
+ return -EPERM;
get_cgdir_and_path(cgroup, &cgdir, &last);
return -EIO;
if (strcmp(path, "/cgroup") == 0)
- return -EINVAL;
+ return -EPERM;
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EINVAL;
+ return errno == ENOENT ? -EPERM : -errno;
+
cgroup = find_cgroup_in_path(path);
if (!cgroup)
/* this is just /cgroup/controller */
- return -EINVAL;
+ return -EPERM;
get_cgdir_and_path(cgroup, &cgdir, &last);
if (!fc)
return -EIO;
-
controller = pick_controller_from_path(fc, path);
if (!controller)
- return -EINVAL;
+ return errno == ENOENT ? -EPERM : -errno;
cgroup = find_cgroup_in_path(path);
if (!cgroup)
- return -EINVAL;
+ return -errno;
get_cgdir_and_path(cgroup, &cgdir, &last);
if (!last)
else if (last && strcmp(next, last) == 0)
ret = -EEXIST;
else
- ret = -ENOENT;
+ ret = -EPERM;
goto out;
}
return -EIO;
controller = pick_controller_from_path(fc, path);
- if (!controller)
- return -EINVAL;
+ if (!controller) /* Someone's trying to delete "/cgroup". */
+ return -EPERM;
cgroup = find_cgroup_in_path(path);
- if (!cgroup)
- return -EINVAL;
+ if (!cgroup) /* Someone's trying to delete a controller e.g. "/blkio". */
+ return -EPERM;
get_cgdir_and_path(cgroup, &cgdir, &last);
if (!last) {
- ret = -EINVAL;
+ /* Someone's trying to delete a cgroup on the same level as the
+ * "/lxc" cgroup e.g. rmdir "/cgroup/blkio/lxc" or
+ * rmdir "/cgroup/blkio/init.slice".
+ */
+ ret = -EPERM;
goto out;
}
return false;
}
-static void get_mem_cached(char *memstat, unsigned long *v)
+static void parse_memstat(char *memstat, unsigned long *cached,
+ unsigned long *active_anon, unsigned long *inactive_anon,
+ unsigned long *active_file, unsigned long *inactive_file,
+ unsigned long *unevictable)
{
char *eol;
- *v = 0;
while (*memstat) {
- if (startswith(memstat, "total_cache")) {
- sscanf(memstat + 11, "%lu", v);
- *v /= 1024;
- return;
+ if (startswith(memstat, "cache")) {
+ sscanf(memstat + 11, "%lu", cached);
+ *cached /= 1024;
+ } else if (startswith(memstat, "active_anon")) {
+ sscanf(memstat + 11, "%lu", active_anon);
+ *active_anon /= 1024;
+ } else if (startswith(memstat, "inactive_anon")) {
+ sscanf(memstat + 11, "%lu", inactive_anon);
+ *inactive_anon /= 1024;
+ } else if (startswith(memstat, "active_file")) {
+ sscanf(memstat + 11, "%lu", active_file);
+ *active_file /= 1024;
+ } else if (startswith(memstat, "inactive_file")) {
+ sscanf(memstat + 11, "%lu", inactive_file);
+ *inactive_file /= 1024;
+ } else if (startswith(memstat, "unevictable")) {
+ sscanf(memstat + 11, "%lu", unevictable);
+ *unevictable /= 1024;
}
eol = strchr(memstat, '\n');
if (!eol)
*memswlimit_str = NULL, *memswusage_str = NULL,
*memswlimit_default_str = NULL, *memswusage_default_str = NULL;
unsigned long memlimit = 0, memusage = 0, memswlimit = 0, memswusage = 0,
- cached = 0, hosttotal = 0;
+ cached = 0, hosttotal = 0, active_anon = 0, inactive_anon = 0,
+ active_file = 0, inactive_file = 0, unevictable = 0;
char *line = NULL;
size_t linelen = 0, total_len = 0, rv = 0;
char *cache = d->buf;
memlimit /= 1024;
memusage /= 1024;
- get_mem_cached(memstat_str, &cached);
+ parse_memstat(memstat_str, &cached, &active_anon,
+ &inactive_anon, &active_file, &inactive_file,
+ &unevictable);
f = fopen("/proc/meminfo", "r");
if (!f)
snprintf(lbuf, 100, "SwapTotal: %8lu kB\n", memswlimit - memlimit);
printme = lbuf;
} else if (startswith(line, "SwapFree:") && memswlimit > 0 && memswusage > 0) {
- snprintf(lbuf, 100, "SwapFree: %8lu kB\n",
- (memswlimit - memlimit) - (memswusage - memusage));
+ unsigned long swaptotal = memswlimit - memlimit,
+ swapusage = memswusage - memusage,
+ swapfree = swapusage < swaptotal ? swaptotal - swapusage : 0;
+ snprintf(lbuf, 100, "SwapFree: %8lu kB\n", swapfree);
printme = lbuf;
} else if (startswith(line, "Slab:")) {
snprintf(lbuf, 100, "Slab: %8lu kB\n", 0UL);
} else if (startswith(line, "SwapCached:")) {
snprintf(lbuf, 100, "SwapCached: %8lu kB\n", 0UL);
printme = lbuf;
+ } else if (startswith(line, "Active")) {
+ snprintf(lbuf, 100, "Active: %8lu kB\n",
+ active_anon + active_file);
+ printme = lbuf;
+ } else if (startswith(line, "Inactive")) {
+ snprintf(lbuf, 100, "Inactive: %8lu kB\n",
+ inactive_anon + inactive_file);
+ printme = lbuf;
+ } else if (startswith(line, "Active(anon)")) {
+ snprintf(lbuf, 100, "Active(anon): %8lu kB\n", active_anon);
+ printme = lbuf;
+ } else if (startswith(line, "Inactive(anon)")) {
+ snprintf(lbuf, 100, "Inactive(anon): %8lu kB\n", inactive_anon);
+ printme = lbuf;
+ } else if (startswith(line, "Active(file)")) {
+ snprintf(lbuf, 100, "Active(file): %8lu kB\n", active_file);
+ printme = lbuf;
+ } else if (startswith(line, "Inactive(file)")) {
+ snprintf(lbuf, 100, "Inactive(file): %8lu kB\n", inactive_file);
+ printme = lbuf;
+ } else if (startswith(line, "Unevictable")) {
+ snprintf(lbuf, 100, "Unevictable: %8lu kB\n", unevictable);
+ printme = lbuf;
+ } else if (startswith(line, "SReclaimable")) {
+ snprintf(lbuf, 100, "SReclaimable: %8lu kB\n", 0UL);
+ printme = lbuf;
+ } else if (startswith(line, "SUnreclaim")) {
+ snprintf(lbuf, 100, "SUnreclaim: %8lu kB\n", 0UL);
+ printme = lbuf;
} else
printme = line;
char cpu_char[10]; /* That's a lot of cores */
char *c;
+ if (strlen(line) == 0)
+ continue;
if (sscanf(line, "cpu%9[^ ]", cpu_char) != 1) {
/* not a ^cpuN line containing a number N, just print it */
l = snprintf(cache, cache_size, "%s", line);
#if RELOADTEST
void iwashere(void)
{
- char *name, *cwd = get_current_dir_name();
- size_t len;
int fd;
- if (!cwd)
- exit(1);
- len = strlen(cwd) + strlen("/iwashere") + 1;
- name = alloca(len);
- snprintf(name, len, "%s/iwashere", cwd);
- free(cwd);
- fd = creat(name, 0755);
+ fd = creat("/tmp/lxcfs-iwashere", 0644);
if (fd >= 0)
close(fd);
}
int proc_readdir(const char *path, void *buf, fuse_fill_dir_t filler, off_t offset,
struct fuse_file_info *fi)
{
- if (filler(buf, "cpuinfo", NULL, 0) != 0 ||
- filler(buf, "meminfo", NULL, 0) != 0 ||
- filler(buf, "stat", NULL, 0) != 0 ||
- filler(buf, "uptime", NULL, 0) != 0 ||
- filler(buf, "diskstats", NULL, 0) != 0 ||
- filler(buf, "swaps", NULL, 0) != 0)
+ if (filler(buf, ".", NULL, 0) != 0 ||
+ filler(buf, "..", NULL, 0) != 0 ||
+ filler(buf, "cpuinfo", NULL, 0) != 0 ||
+ filler(buf, "meminfo", NULL, 0) != 0 ||
+ filler(buf, "stat", NULL, 0) != 0 ||
+ filler(buf, "uptime", NULL, 0) != 0 ||
+ filler(buf, "diskstats", NULL, 0) != 0 ||
+ filler(buf, "swaps", NULL, 0) != 0)
return -EINVAL;
return 0;
}
int proc_access(const char *path, int mask)
{
+ if (strcmp(path, "/proc") == 0 && access(path, R_OK) == 0)
+ return 0;
+
/* these are all read-only */
if ((mask & ~R_OK) != 0)
return -EACCES;
}
}
-static void __attribute__((constructor)) collect_subsystems(void)
+/*
+ * Functions needed to setup cgroups in the __constructor__.
+ */
+
+static bool mkdir_p(const char *dir, mode_t mode)
+{
+ const char *tmp = dir;
+ const char *orig = dir;
+ char *makeme;
+
+ do {
+ dir = tmp + strspn(tmp, "/");
+ tmp = dir + strcspn(dir, "/");
+ makeme = strndup(orig, dir - orig);
+ if (!makeme)
+ return false;
+ if (mkdir(makeme, mode) && errno != EEXIST) {
+ fprintf(stderr, "failed to create directory '%s': %s",
+ makeme, strerror(errno));
+ free(makeme);
+ return false;
+ }
+ free(makeme);
+ } while(tmp != dir);
+
+ return true;
+}
+
+static bool umount_if_mounted(void)
+{
+ if (umount2(BASEDIR, MNT_DETACH) < 0 && errno != EINVAL) {
+ fprintf(stderr, "failed to unmount %s: %s.\n", BASEDIR, strerror(errno));
+ return false;
+ }
+ return true;
+}
+
+static int pivot_enter(void)
+{
+ int ret = -1, oldroot = -1, newroot = -1;
+
+ oldroot = open("/", O_DIRECTORY | O_RDONLY);
+ if (oldroot < 0) {
+ fprintf(stderr, "%s: Failed to open old root for fchdir.\n", __func__);
+ return ret;
+ }
+
+ newroot = open(ROOTDIR, O_DIRECTORY | O_RDONLY);
+ if (newroot < 0) {
+ fprintf(stderr, "%s: Failed to open new root for fchdir.\n", __func__);
+ goto err;
+ }
+
+ /* change into new root fs */
+ if (fchdir(newroot) < 0) {
+ fprintf(stderr, "%s: Failed to change directory to new rootfs: %s.\n", __func__, ROOTDIR);
+ goto err;
+ }
+
+ /* pivot_root into our new root fs */
+ if (pivot_root(".", ".") < 0) {
+ fprintf(stderr, "%s: pivot_root() syscall failed: %s.\n", __func__, strerror(errno));
+ goto err;
+ }
+
+ /*
+ * At this point the old-root is mounted on top of our new-root.
+ * To unmounted it we must not be chdir'd into it, so escape back
+ * to the old-root.
+ */
+ if (fchdir(oldroot) < 0) {
+ fprintf(stderr, "%s: Failed to enter old root.\n", __func__);
+ goto err;
+ }
+ if (umount2(".", MNT_DETACH) < 0) {
+ fprintf(stderr, "%s: Failed to detach old root.\n", __func__);
+ goto err;
+ }
+
+ if (fchdir(newroot) < 0) {
+ fprintf(stderr, "%s: Failed to re-enter new root.\n", __func__);
+ goto err;
+ }
+
+ ret = 0;
+
+err:
+ if (oldroot > 0)
+ close(oldroot);
+ if (newroot > 0)
+ close(newroot);
+ return ret;
+}
+
+/* Prepare our new clean root. */
+static int pivot_prepare(void)
+{
+ if (mkdir(ROOTDIR, 0700) < 0 && errno != EEXIST) {
+ fprintf(stderr, "%s: Failed to create directory for new root.\n", __func__);
+ return -1;
+ }
+
+ if (mount("/", ROOTDIR, NULL, MS_BIND, 0) < 0) {
+ fprintf(stderr, "%s: Failed to bind-mount / for new root: %s.\n", __func__, strerror(errno));
+ return -1;
+ }
+
+ if (mount(RUNTIME_PATH, ROOTDIR RUNTIME_PATH, NULL, MS_BIND, 0) < 0) {
+ fprintf(stderr, "%s: Failed to bind-mount /run into new root: %s.\n", __func__, strerror(errno));
+ return -1;
+ }
+
+ if (mount(BASEDIR, ROOTDIR BASEDIR, NULL, MS_REC | MS_MOVE, 0) < 0) {
+ printf("%s: failed to move " BASEDIR " into new root: %s.\n", __func__, strerror(errno));
+ return -1;
+ }
+
+ return 0;
+}
+
+static bool pivot_new_root(void)
+{
+ /* Prepare new root. */
+ if (pivot_prepare() < 0)
+ return false;
+
+ /* Pivot into new root. */
+ if (pivot_enter() < 0)
+ return false;
+
+ return true;
+}
+
+static bool setup_cgfs_dir(void)
+{
+ if (!mkdir_p(BASEDIR, 0700)) {
+ fprintf(stderr, "Failed to create lxcfs cgroup mountpoint.\n");
+ return false;
+ }
+
+ if (!umount_if_mounted()) {
+ fprintf(stderr, "Failed to clean up old lxcfs cgroup mountpoint.\n");
+ return false;
+ }
+
+ if (unshare(CLONE_NEWNS) < 0) {
+ fprintf(stderr, "%s: Failed to unshare mount namespace: %s.\n", __func__, strerror(errno));
+ return false;
+ }
+
+ if (mount(NULL, "/", NULL, MS_REC | MS_PRIVATE, 0) < 0) {
+ fprintf(stderr, "%s: Failed to remount / private: %s.\n", __func__, strerror(errno));
+ return false;
+ }
+
+ if (mount("tmpfs", BASEDIR, "tmpfs", 0, "size=100000,mode=700") < 0) {
+ fprintf(stderr, "Failed to mount tmpfs over lxcfs cgroup mountpoint.\n");
+ return false;
+ }
+
+ return true;
+}
+
+static bool do_mount_cgroups(void)
+{
+ char *target;
+ size_t clen, len;
+ int i, ret;
+
+ for (i = 0; i < num_hierarchies; i++) {
+ char *controller = hierarchies[i];
+ clen = strlen(controller);
+ len = strlen(BASEDIR) + clen + 2;
+ target = malloc(len);
+ if (!target)
+ return false;
+ ret = snprintf(target, len, "%s/%s", BASEDIR, controller);
+ if (ret < 0 || ret >= len) {
+ free(target);
+ return false;
+ }
+ if (mkdir(target, 0755) < 0 && errno != EEXIST) {
+ free(target);
+ return false;
+ }
+ if (mount(controller, target, "cgroup", 0, controller) < 0) {
+ fprintf(stderr, "Failed mounting cgroup %s\n", controller);
+ free(target);
+ return false;
+ }
+
+ fd_hierarchies[i] = open(target, O_DIRECTORY);
+ if (fd_hierarchies[i] < 0) {
+ free(target);
+ return false;
+ }
+ free(target);
+ }
+ return true;
+}
+
+static bool cgfs_setup_controllers(void)
+{
+ if (!setup_cgfs_dir())
+ return false;
+
+ if (!do_mount_cgroups()) {
+ fprintf(stderr, "Failed to set up private lxcfs cgroup mounts.\n");
+ return false;
+ }
+
+ if (!pivot_new_root())
+ return false;
+
+ return true;
+}
+
+static int preserve_ns(int pid)
+{
+ int ret;
+ size_t len = 5 /* /proc */ + 21 /* /int_as_str */ + 7 /* /ns/mnt */ + 1 /* \0 */;
+ char path[len];
+
+ ret = snprintf(path, len, "/proc/%d/ns/mnt", pid);
+ if (ret < 0 || (size_t)ret >= len)
+ return -1;
+
+ return open(path, O_RDONLY | O_CLOEXEC);
+}
+
+static void __attribute__((constructor)) collect_and_mount_subsystems(void)
{
FILE *f;
char *line = NULL;
size_t len = 0;
+ int i, init_ns = -1;
if ((f = fopen("/proc/self/cgroup", "r")) == NULL) {
fprintf(stderr, "Error opening /proc/self/cgroup: %s\n", strerror(errno));
goto out;
}
+ /* Preserve initial namespace. */
+ init_ns = preserve_ns(getpid());
+ if (init_ns < 0)
+ goto out;
+
+ fd_hierarchies = malloc(sizeof(int *) * num_hierarchies);
+ if (!fd_hierarchies)
+ goto out;
+
+ for (i = 0; i < num_hierarchies; i++)
+ fd_hierarchies[i] = -1;
+
+ /* This function calls unshare(CLONE_NEWNS) our initial mount namespace
+ * to privately mount lxcfs cgroups. */
+ if (!cgfs_setup_controllers())
+ goto out;
+
+ if (setns(init_ns, 0) < 0)
+ goto out;
+
print_subsystems();
out:
free(line);
fclose(f);
+ if (init_ns >= 0)
+ close(init_ns);
}
static void __attribute__((destructor)) free_subsystems(void)
{
int i;
- for (i = 0; i < num_hierarchies; i++)
+ for (i = 0; i < num_hierarchies; i++) {
if (hierarchies[i])
free(hierarchies[i]);
+ if (fd_hierarchies && fd_hierarchies[i] >= 0)
+ close(fd_hierarchies[i]);
+ }
free(hierarchies);
+ free(fd_hierarchies);
}