LXC_TYPE_PROC_STAT,
LXC_TYPE_PROC_DISKSTATS,
LXC_TYPE_PROC_SWAPS,
+ LXC_TYPE_PROC_LOADAVG,
};
struct file_info {
int cached;
};
+struct cpuacct_usage {
+ uint64_t user;
+ uint64_t system;
+};
+
+/* The function of hash table.*/
+#define LOAD_SIZE 100 /*the size of hash_table */
+#define FLUSH_TIME 5 /*the flush rate */
+#define DEPTH_DIR 3 /*the depth of per cgroup */
+/* The function of calculate loadavg .*/
+#define FSHIFT 11 /* nr of bits of precision */
+#define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
+#define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
+#define EXP_5 2014 /* 1/exp(5sec/5min) */
+#define EXP_15 2037 /* 1/exp(5sec/15min) */
+#define LOAD_INT(x) ((x) >> FSHIFT)
+#define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100)
+/*
+ * This parameter is used for proc_loadavg_read().
+ * 1 means use loadavg, 0 means not use.
+ */
+static int loadavg = 0;
+static volatile sig_atomic_t loadavg_stop = 0;
+static int calc_hash(char *name)
+{
+ unsigned int hash = 0;
+ unsigned int x = 0;
+ /* ELFHash algorithm. */
+ while (*name) {
+ hash = (hash << 4) + *name++;
+ x = hash & 0xf0000000;
+ if (x != 0)
+ hash ^= (x >> 24);
+ hash &= ~x;
+ }
+ return ((hash & 0x7fffffff) % LOAD_SIZE);
+}
+
+struct load_node {
+ char *cg; /*cg */
+ unsigned long avenrun[3]; /* Load averages */
+ unsigned int run_pid;
+ unsigned int total_pid;
+ unsigned int last_pid;
+ int cfd; /* The file descriptor of the mounted cgroup */
+ struct load_node *next;
+ struct load_node **pre;
+};
+
+struct load_head {
+ /*
+ * The lock is about insert load_node and refresh load_node.To the first
+ * load_node of each hash bucket, insert and refresh in this hash bucket is
+ * mutually exclusive.
+ */
+ pthread_mutex_t lock;
+ /*
+ * The rdlock is about read loadavg and delete load_node.To each hash
+ * bucket, read and delete is mutually exclusive. But at the same time, we
+ * allow paratactic read operation. This rdlock is at list level.
+ */
+ pthread_rwlock_t rdlock;
+ /*
+ * The rilock is about read loadavg and insert load_node.To the first
+ * load_node of each hash bucket, read and insert is mutually exclusive.
+ * But at the same time, we allow paratactic read operation.
+ */
+ pthread_rwlock_t rilock;
+ struct load_node *next;
+};
+
+static struct load_head load_hash[LOAD_SIZE]; /* hash table */
+/*
+ * init_load initialize the hash table.
+ * Return 0 on success, return -1 on failure.
+ */
+static int init_load(void)
+{
+ int i;
+ int ret;
+
+ for (i = 0; i < LOAD_SIZE; i++) {
+ load_hash[i].next = NULL;
+ ret = pthread_mutex_init(&load_hash[i].lock, NULL);
+ if (ret != 0) {
+ lxcfs_error("%s\n", "Failed to initialize lock");
+ goto out3;
+ }
+ ret = pthread_rwlock_init(&load_hash[i].rdlock, NULL);
+ if (ret != 0) {
+ lxcfs_error("%s\n", "Failed to initialize rdlock");
+ goto out2;
+ }
+ ret = pthread_rwlock_init(&load_hash[i].rilock, NULL);
+ if (ret != 0) {
+ lxcfs_error("%s\n", "Failed to initialize rilock");
+ goto out1;
+ }
+ }
+ return 0;
+out1:
+ pthread_rwlock_destroy(&load_hash[i].rdlock);
+out2:
+ pthread_mutex_destroy(&load_hash[i].lock);
+out3:
+ while (i > 0) {
+ i--;
+ pthread_mutex_destroy(&load_hash[i].lock);
+ pthread_rwlock_destroy(&load_hash[i].rdlock);
+ pthread_rwlock_destroy(&load_hash[i].rilock);
+ }
+ return -1;
+}
+
+static void insert_node(struct load_node **n, int locate)
+{
+ struct load_node *f;
+
+ pthread_mutex_lock(&load_hash[locate].lock);
+ pthread_rwlock_wrlock(&load_hash[locate].rilock);
+ f = load_hash[locate].next;
+ load_hash[locate].next = *n;
+
+ (*n)->pre = &(load_hash[locate].next);
+ if (f)
+ f->pre = &((*n)->next);
+ (*n)->next = f;
+ pthread_mutex_unlock(&load_hash[locate].lock);
+ pthread_rwlock_unlock(&load_hash[locate].rilock);
+}
+/*
+ * locate_node() finds special node. Not return NULL means success.
+ * It should be noted that rdlock isn't unlocked at the end of code
+ * because this function is used to read special node. Delete is not
+ * allowed before read has ended.
+ * unlock rdlock only in proc_loadavg_read().
+ */
+static struct load_node *locate_node(char *cg, int locate)
+{
+ struct load_node *f = NULL;
+ int i = 0;
+
+ pthread_rwlock_rdlock(&load_hash[locate].rilock);
+ pthread_rwlock_rdlock(&load_hash[locate].rdlock);
+ if (load_hash[locate].next == NULL) {
+ pthread_rwlock_unlock(&load_hash[locate].rilock);
+ return f;
+ }
+ f = load_hash[locate].next;
+ pthread_rwlock_unlock(&load_hash[locate].rilock);
+ while (f && ((i = strcmp(f->cg, cg)) != 0))
+ f = f->next;
+ return f;
+}
+/* Delete the load_node n and return the next node of it. */
+static struct load_node *del_node(struct load_node *n, int locate)
+{
+ struct load_node *g;
+
+ pthread_rwlock_wrlock(&load_hash[locate].rdlock);
+ if (n->next == NULL) {
+ *(n->pre) = NULL;
+ } else {
+ *(n->pre) = n->next;
+ n->next->pre = n->pre;
+ }
+ g = n->next;
+ free(n->cg);
+ free(n);
+ pthread_rwlock_unlock(&load_hash[locate].rdlock);
+ return g;
+}
+
+static void load_free(void)
+{
+ int i;
+ struct load_node *f, *p;
+
+ for (i = 0; i < LOAD_SIZE; i++) {
+ pthread_mutex_lock(&load_hash[i].lock);
+ pthread_rwlock_wrlock(&load_hash[i].rilock);
+ pthread_rwlock_wrlock(&load_hash[i].rdlock);
+ if (load_hash[i].next == NULL) {
+ pthread_mutex_unlock(&load_hash[i].lock);
+ pthread_mutex_destroy(&load_hash[i].lock);
+ pthread_rwlock_unlock(&load_hash[i].rilock);
+ pthread_rwlock_destroy(&load_hash[i].rilock);
+ pthread_rwlock_unlock(&load_hash[i].rdlock);
+ pthread_rwlock_destroy(&load_hash[i].rdlock);
+ continue;
+ }
+ for (f = load_hash[i].next; f; ) {
+ free(f->cg);
+ p = f->next;
+ free(f);
+ f = p;
+ }
+ pthread_mutex_unlock(&load_hash[i].lock);
+ pthread_mutex_destroy(&load_hash[i].lock);
+ pthread_rwlock_unlock(&load_hash[i].rilock);
+ pthread_rwlock_destroy(&load_hash[i].rilock);
+ pthread_rwlock_unlock(&load_hash[i].rdlock);
+ pthread_rwlock_destroy(&load_hash[i].rdlock);
+ }
+}
/* Reserve buffer size to account for file size changes. */
#define BUF_RESERVE_SIZE 512
FILE *f;
size_t len, ret;
- if (!(f = fdopen(fd, "w")))
+ f = fdopen(fd, "w");
+ if (!f)
return false;
+
len = strlen(string);
ret = fwrite(string, 1, len, f);
if (ret != len) {
- lxcfs_error("Error writing to file: %s\n", strerror(errno));
+ lxcfs_error("%s - Error writing \"%s\" to \"%s\"\n",
+ strerror(errno), string, fnam);
fclose(f);
return false;
}
+
if (fclose(f) < 0) {
- lxcfs_error("Error writing to file: %s\n", strerror(errno));
+ lxcfs_error("%s - Failed to close \"%s\"\n", strerror(errno), fnam);
return false;
}
+
return true;
}
free(f->buf);
f->buf = NULL;
free(f);
+ f = NULL;
}
int cg_releasedir(const char *path, struct fuse_file_info *fi)
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)
+ unsigned long *unevictable, unsigned long *shmem)
{
char *eol;
} else if (startswith(memstat, "total_unevictable")) {
sscanf(memstat + 17, "%lu", unevictable);
*unevictable /= 1024;
+ } else if (startswith(memstat, "total_shmem")) {
+ sscanf(memstat + 11, "%lu", shmem);
+ *shmem /= 1024;
}
eol = strchr(memstat, '\n');
if (!eol)
*memswlimit_str = NULL, *memswusage_str = NULL;
unsigned long memlimit = 0, memusage = 0, memswlimit = 0, memswusage = 0,
cached = 0, hosttotal = 0, active_anon = 0, inactive_anon = 0,
- active_file = 0, inactive_file = 0, unevictable = 0,
+ active_file = 0, inactive_file = 0, unevictable = 0, shmem = 0,
hostswtotal = 0;
char *line = NULL;
size_t linelen = 0, total_len = 0, rv = 0;
parse_memstat(memstat_str, &cached, &active_anon,
&inactive_anon, &active_file, &inactive_file,
- &unevictable);
+ &unevictable, &shmem);
f = fopen("/proc/meminfo", "r");
if (!f)
} else if (startswith(line, "SUnreclaim")) {
snprintf(lbuf, 100, "SUnreclaim: %8lu kB\n", 0UL);
printme = lbuf;
+ } else if (startswith(line, "Shmem:")) {
+ snprintf(lbuf, 100, "Shmem: %8lu kB\n", shmem);
+ printme = lbuf;
+ } else if (startswith(line, "ShmemHugePages")) {
+ snprintf(lbuf, 100, "ShmemHugePages: %8lu kB\n", 0UL);
+ printme = lbuf;
+ } else if (startswith(line, "ShmemPmdMapped")) {
+ snprintf(lbuf, 100, "ShmemPmdMapped: %8lu kB\n", 0UL);
+ printme = lbuf;
} else
printme = line;
return procage;
}
+/*
+ * Returns 0 on success.
+ * It is the caller's responsibility to free `return_usage`, unless this
+ * function returns an error.
+ */
+static int read_cpuacct_usage_all(char *cg, char *cpuset, struct cpuacct_usage **return_usage)
+{
+ int cpucount = get_nprocs();
+ struct cpuacct_usage *cpu_usage;
+ int rv = 0, i, j, ret, read_pos = 0, read_cnt;
+ int cg_cpu;
+ uint64_t cg_user, cg_system;
+ int64_t ticks_per_sec;
+ char *usage_str = NULL;
+
+ ticks_per_sec = sysconf(_SC_CLK_TCK);
+
+ if (ticks_per_sec < 0 && errno == EINVAL) {
+ lxcfs_debug(
+ "%s\n",
+ "read_cpuacct_usage_all failed to determine number of clock ticks "
+ "in a second");
+ return -1;
+ }
+
+ cpu_usage = malloc(sizeof(struct cpuacct_usage) * cpucount);
+ if (!cpu_usage)
+ return -ENOMEM;
+
+ if (!cgfs_get_value("cpuacct", cg, "cpuacct.usage_all", &usage_str)) {
+ rv = -1;
+ goto err;
+ }
+
+ if (sscanf(usage_str, "cpu user system\n%n", &read_cnt) != 0) {
+ lxcfs_error("read_cpuacct_usage_all reading first line from "
+ "%s/cpuacct.usage_all failed.\n", cg);
+ rv = -1;
+ goto err;
+ }
+
+ read_pos += read_cnt;
+
+ for (i = 0, j = 0; i < cpucount; i++) {
+ ret = sscanf(usage_str + read_pos, "%d %lu %lu\n%n", &cg_cpu, &cg_user,
+ &cg_system, &read_cnt);
+
+ if (ret == EOF)
+ break;
+
+ if (ret != 3) {
+ lxcfs_error("read_cpuacct_usage_all reading from %s/cpuacct.usage_all "
+ "failed.\n", cg);
+ rv = -1;
+ goto err;
+ }
+
+ read_pos += read_cnt;
+
+ if (!cpu_in_cpuset(i, cpuset))
+ continue;
+
+ /* Convert the time from nanoseconds to USER_HZ */
+ cpu_usage[j].user = cg_user / 1000.0 / 1000 / 1000 * ticks_per_sec;
+ cpu_usage[j].system = cg_system / 1000.0 / 1000 / 1000 * ticks_per_sec;
+ j++;
+ }
+
+ rv = 0;
+ *return_usage = cpu_usage;
+
+err:
+ if (usage_str)
+ free(usage_str);
+
+ if (rv != 0) {
+ free(cpu_usage);
+ *return_usage = NULL;
+ }
+
+ return rv;
+}
+
#define CPUALL_MAX_SIZE (BUF_RESERVE_SIZE / 2)
static int proc_stat_read(char *buf, size_t size, off_t offset,
struct fuse_file_info *fi)
char *cache = d->buf + CPUALL_MAX_SIZE;
size_t cache_size = d->buflen - CPUALL_MAX_SIZE;
FILE *f = NULL;
+ struct cpuacct_usage *cg_cpu_usage = NULL;
if (offset){
if (offset > d->size)
if (!cpuset)
goto err;
+ /*
+ * Read cpuacct.usage_all for all CPUs.
+ * If the cpuacct cgroup is present, it is used to calculate the container's
+ * CPU usage. If not, values from the host's /proc/stat are used.
+ */
+ if (read_cpuacct_usage_all(cg, cpuset, &cg_cpu_usage) != 0) {
+ lxcfs_debug("%s\n", "proc_stat_read failed to read from cpuacct, "
+ "falling back to the host's /proc/stat");
+ }
+
f = fopen("/proc/stat", "r");
if (!f)
goto err;
int cpu;
char cpu_char[10]; /* That's a lot of cores */
char *c;
+ uint64_t all_used, cg_used, new_idle;
+ int ret;
if (strlen(line) == 0)
continue;
continue;
curcpu ++;
- c = strchr(line, ' ');
- if (!c)
- continue;
- l = snprintf(cache, cache_size, "cpu%d%s", curcpu, c);
- if (l < 0) {
- perror("Error writing to cache");
- rv = 0;
- goto err;
-
- }
- if (l >= cache_size) {
- lxcfs_error("%s\n", "Internal error: truncated write to cache.");
- rv = 0;
- goto err;
- }
-
- cache += l;
- cache_size -= l;
- total_len += l;
-
- if (sscanf(line, "%*s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu",
+ ret = sscanf(line, "%*s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu",
&user,
&nice,
&system,
&softirq,
&steal,
&guest,
- &guest_nice) != 10)
- continue;
- user_sum += user;
- nice_sum += nice;
- system_sum += system;
- idle_sum += idle;
- iowait_sum += iowait;
- irq_sum += irq;
- softirq_sum += softirq;
- steal_sum += steal;
- guest_sum += guest;
- guest_nice_sum += guest_nice;
+ &guest_nice);
+
+ if (ret != 10 || !cg_cpu_usage) {
+ c = strchr(line, ' ');
+ if (!c)
+ continue;
+ l = snprintf(cache, cache_size, "cpu%d%s", curcpu, c);
+ if (l < 0) {
+ perror("Error writing to cache");
+ rv = 0;
+ goto err;
+
+ }
+ if (l >= cache_size) {
+ lxcfs_error("%s\n", "Internal error: truncated write to cache.");
+ rv = 0;
+ goto err;
+ }
+
+ cache += l;
+ cache_size -= l;
+ total_len += l;
+
+ if (ret != 10)
+ continue;
+ }
+
+ if (cg_cpu_usage) {
+ all_used = user + nice + system + iowait + irq + softirq + steal + guest + guest_nice;
+ cg_used = cg_cpu_usage[curcpu].user + cg_cpu_usage[curcpu].system;
+
+ if (all_used >= cg_used) {
+ new_idle = idle + (all_used - cg_used);
+
+ } else {
+ lxcfs_error("cpu%d from %s has unexpected cpu time: %lu in /proc/stat, "
+ "%lu in cpuacct.usage_all; unable to determine idle time\n",
+ curcpu, cg, all_used, cg_used);
+ new_idle = idle;
+ }
+
+ l = snprintf(cache, cache_size, "cpu%d %lu 0 %lu %lu 0 0 0 0 0 0\n",
+ curcpu, cg_cpu_usage[curcpu].user, cg_cpu_usage[curcpu].system,
+ new_idle);
+
+ if (l < 0) {
+ perror("Error writing to cache");
+ rv = 0;
+ goto err;
+
+ }
+ if (l >= cache_size) {
+ lxcfs_error("%s\n", "Internal error: truncated write to cache.");
+ rv = 0;
+ goto err;
+ }
+
+ cache += l;
+ cache_size -= l;
+ total_len += l;
+
+ user_sum += cg_cpu_usage[curcpu].user;
+ system_sum += cg_cpu_usage[curcpu].system;
+ idle_sum += new_idle;
+
+ } else {
+ user_sum += user;
+ nice_sum += nice;
+ system_sum += system;
+ idle_sum += idle;
+ iowait_sum += iowait;
+ irq_sum += irq;
+ softirq_sum += softirq;
+ steal_sum += steal;
+ guest_sum += guest;
+ guest_nice_sum += guest_nice;
+ }
}
cache = d->buf;
err:
if (f)
fclose(f);
+ if (cg_cpu_usage)
+ free(cg_cpu_usage);
free(line);
free(cpuset);
free(cg);
free(memswusage_str);
return rv;
}
+/*
+ * Find the process pid from cgroup path.
+ * eg:from /sys/fs/cgroup/cpu/docker/containerid/cgroup.procs to find the process pid.
+ * @pid_buf : put pid to pid_buf.
+ * @dpath : the path of cgroup. eg: /docker/containerid or /docker/containerid/child-cgroup ...
+ * @depth : the depth of cgroup in container.
+ * @sum : return the number of pid.
+ * @cfd : the file descriptor of the mounted cgroup. eg: /sys/fs/cgroup/cpu
+ */
+static int calc_pid(char ***pid_buf, char *dpath, int depth, int sum, int cfd)
+{
+ DIR *dir;
+ int fd;
+ struct dirent *file;
+ FILE *f = NULL;
+ size_t linelen = 0;
+ char *line = NULL;
+ int pd;
+ char *path_dir, *path;
+ char **pid;
+
+ /* path = dpath + "/cgroup.procs" + /0 */
+ do {
+ path = malloc(strlen(dpath) + 20);
+ } while (!path);
+
+ strcpy(path, dpath);
+ fd = openat(cfd, path, O_RDONLY);
+ if (fd < 0)
+ goto out;
+
+ dir = fdopendir(fd);
+ if (dir == NULL) {
+ close(fd);
+ goto out;
+ }
+
+ while (((file = readdir(dir)) != NULL) && depth > 0) {
+ if (strncmp(file->d_name, ".", 1) == 0)
+ continue;
+ if (strncmp(file->d_name, "..", 1) == 0)
+ continue;
+ if (file->d_type == DT_DIR) {
+ /* path + '/' + d_name +/0 */
+ do {
+ path_dir = malloc(strlen(path) + 2 + sizeof(file->d_name));
+ } while (!path_dir);
+ strcpy(path_dir, path);
+ strcat(path_dir, "/");
+ strcat(path_dir, file->d_name);
+ pd = depth - 1;
+ sum = calc_pid(pid_buf, path_dir, pd, sum, cfd);
+ free(path_dir);
+ }
+ }
+ closedir(dir);
+
+ strcat(path, "/cgroup.procs");
+ fd = openat(cfd, path, O_RDONLY);
+ if (fd < 0)
+ goto out;
+
+ f = fdopen(fd, "r");
+ if (!f) {
+ close(fd);
+ goto out;
+ }
+
+ while (getline(&line, &linelen, f) != -1) {
+ do {
+ pid = realloc(*pid_buf, sizeof(char *) * (sum + 1));
+ } while (!pid);
+ *pid_buf = pid;
+ do {
+ *(*pid_buf + sum) = malloc(strlen(line) + 1);
+ } while (*(*pid_buf + sum) == NULL);
+ strcpy(*(*pid_buf + sum), line);
+ sum++;
+ }
+ fclose(f);
+out:
+ if (line)
+ free(line);
+ free(path);
+ return sum;
+}
+/*
+ * calc_load calculates the load according to the following formula:
+ * load1 = load0 * exp + active * (1 - exp)
+ *
+ * @load1: the new loadavg.
+ * @load0: the former loadavg.
+ * @active: the total number of running pid at this moment.
+ * @exp: the fixed-point defined in the beginning.
+ */
+static unsigned long
+calc_load(unsigned long load, unsigned long exp, unsigned long active)
+{
+ unsigned long newload;
+
+ active = active > 0 ? active * FIXED_1 : 0;
+ newload = load * exp + active * (FIXED_1 - exp);
+ if (active >= load)
+ newload += FIXED_1 - 1;
+
+ return newload / FIXED_1;
+}
+
+/*
+ * Return 0 means that container p->cg is closed.
+ * Return -1 means that error occurred in refresh.
+ * Positive num equals the total number of pid.
+ */
+static int refresh_load(struct load_node *p, char *path)
+{
+ FILE *f = NULL;
+ char **idbuf;
+ char proc_path[256];
+ int i, ret, run_pid = 0, total_pid = 0, last_pid = 0;
+ char *line = NULL;
+ size_t linelen = 0;
+ int sum, length;
+ DIR *dp;
+ struct dirent *file;
+
+ do {
+ idbuf = malloc(sizeof(char *));
+ } while (!idbuf);
+ sum = calc_pid(&idbuf, path, DEPTH_DIR, 0, p->cfd);
+ /* normal exit */
+ if (sum == 0)
+ goto out;
+
+ for (i = 0; i < sum; i++) {
+ /*clean up '\n' */
+ length = strlen(idbuf[i])-1;
+ idbuf[i][length] = '\0';
+ ret = snprintf(proc_path, 256, "/proc/%s/task", idbuf[i]);
+ if (ret < 0 || ret > 255) {
+ lxcfs_error("%s\n", "snprintf() failed in refresh_load.");
+ i = sum;
+ sum = -1;
+ goto err_out;
+ }
+
+ dp = opendir(proc_path);
+ if (!dp) {
+ lxcfs_error("%s\n", "Open proc_path failed in refresh_load.");
+ continue;
+ }
+ while ((file = readdir(dp)) != NULL) {
+ if (strncmp(file->d_name, ".", 1) == 0)
+ continue;
+ if (strncmp(file->d_name, "..", 1) == 0)
+ continue;
+ total_pid++;
+ /* We make the biggest pid become last_pid.*/
+ ret = atof(file->d_name);
+ last_pid = (ret > last_pid) ? ret : last_pid;
+
+ ret = snprintf(proc_path, 256, "/proc/%s/task/%s/status", idbuf[i], file->d_name);
+ if (ret < 0 || ret > 255) {
+ lxcfs_error("%s\n", "snprintf() failed in refresh_load.");
+ i = sum;
+ sum = -1;
+ closedir(dp);
+ goto err_out;
+ }
+ f = fopen(proc_path, "r");
+ if (f != NULL) {
+ while (getline(&line, &linelen, f) != -1) {
+ /* Find State */
+ if ((line[0] == 'S') && (line[1] == 't'))
+ break;
+ }
+ if ((line[7] == 'R') || (line[7] == 'D'))
+ run_pid++;
+ fclose(f);
+ }
+ }
+ closedir(dp);
+ }
+ /*Calculate the loadavg.*/
+ p->avenrun[0] = calc_load(p->avenrun[0], EXP_1, run_pid);
+ p->avenrun[1] = calc_load(p->avenrun[1], EXP_5, run_pid);
+ p->avenrun[2] = calc_load(p->avenrun[2], EXP_15, run_pid);
+ p->run_pid = run_pid;
+ p->total_pid = total_pid;
+ p->last_pid = last_pid;
+
+ free(line);
+err_out:
+ for (; i > 0; i--)
+ free(idbuf[i-1]);
+out:
+ free(idbuf);
+ return sum;
+}
+/*
+ * Traverse the hash table and update it.
+ */
+void *load_begin(void *arg)
+{
+
+ char *path = NULL;
+ int i, sum, length, ret;
+ struct load_node *f;
+ int first_node;
+ clock_t time1, time2;
+
+ while (1) {
+ if (loadavg_stop == 1)
+ return NULL;
+
+ time1 = clock();
+ for (i = 0; i < LOAD_SIZE; i++) {
+ pthread_mutex_lock(&load_hash[i].lock);
+ if (load_hash[i].next == NULL) {
+ pthread_mutex_unlock(&load_hash[i].lock);
+ continue;
+ }
+ f = load_hash[i].next;
+ first_node = 1;
+ while (f) {
+ length = strlen(f->cg) + 2;
+ do {
+ /* strlen(f->cg) + '.' or '' + \0 */
+ path = malloc(length);
+ } while (!path);
+
+ ret = snprintf(path, length, "%s%s", *(f->cg) == '/' ? "." : "", f->cg);
+ if (ret < 0 || ret > length - 1) {
+ /* snprintf failed, ignore the node.*/
+ lxcfs_error("Refresh node %s failed for snprintf().\n", f->cg);
+ goto out;
+ }
+ sum = refresh_load(f, path);
+ if (sum == 0) {
+ f = del_node(f, i);
+ } else {
+out: f = f->next;
+ }
+ free(path);
+ /* load_hash[i].lock locks only on the first node.*/
+ if (first_node == 1) {
+ first_node = 0;
+ pthread_mutex_unlock(&load_hash[i].lock);
+ }
+ }
+ }
+
+ if (loadavg_stop == 1)
+ return NULL;
+
+ time2 = clock();
+ usleep(FLUSH_TIME * 1000000 - (int)((time2 - time1) * 1000000 / CLOCKS_PER_SEC));
+ }
+}
+
+static int proc_loadavg_read(char *buf, size_t size, off_t offset,
+ struct fuse_file_info *fi)
+{
+ struct fuse_context *fc = fuse_get_context();
+ struct file_info *d = (struct file_info *)fi->fh;
+ pid_t initpid;
+ char *cg;
+ size_t total_len = 0;
+ char *cache = d->buf;
+ struct load_node *n;
+ int hash;
+ int cfd, rv = 0;
+ unsigned long a, b, c;
+
+ if (offset) {
+ if (offset > d->size)
+ return -EINVAL;
+ if (!d->cached)
+ return 0;
+ int left = d->size - offset;
+ total_len = left > size ? size : left;
+ memcpy(buf, cache + offset, total_len);
+ return total_len;
+ }
+ if (!loadavg)
+ return read_file("/proc/loadavg", buf, size, d);
+
+ initpid = lookup_initpid_in_store(fc->pid);
+ if (initpid <= 0)
+ initpid = fc->pid;
+ cg = get_pid_cgroup(initpid, "cpu");
+ if (!cg)
+ return read_file("/proc/loadavg", buf, size, d);
+
+ prune_init_slice(cg);
+ hash = calc_hash(cg);
+ n = locate_node(cg, hash);
+
+ /* First time */
+ if (n == NULL) {
+ if (!find_mounted_controller("cpu", &cfd)) {
+ /*
+ * In locate_node() above, pthread_rwlock_unlock() isn't used
+ * because delete is not allowed before read has ended.
+ */
+ pthread_rwlock_unlock(&load_hash[hash].rdlock);
+ rv = 0;
+ goto err;
+ }
+ do {
+ n = malloc(sizeof(struct load_node));
+ } while (!n);
+
+ do {
+ n->cg = malloc(strlen(cg)+1);
+ } while (!n->cg);
+ strcpy(n->cg, cg);
+ n->avenrun[0] = 0;
+ n->avenrun[1] = 0;
+ n->avenrun[2] = 0;
+ n->run_pid = 0;
+ n->total_pid = 1;
+ n->last_pid = initpid;
+ n->cfd = cfd;
+ insert_node(&n, hash);
+ }
+ a = n->avenrun[0] + (FIXED_1/200);
+ b = n->avenrun[1] + (FIXED_1/200);
+ c = n->avenrun[2] + (FIXED_1/200);
+ total_len = snprintf(d->buf, d->buflen, "%lu.%02lu %lu.%02lu %lu.%02lu %d/%d %d\n",
+ LOAD_INT(a), LOAD_FRAC(a),
+ LOAD_INT(b), LOAD_FRAC(b),
+ LOAD_INT(c), LOAD_FRAC(c),
+ n->run_pid, n->total_pid, n->last_pid);
+ pthread_rwlock_unlock(&load_hash[hash].rdlock);
+ if (total_len < 0 || total_len >= d->buflen) {
+ lxcfs_error("%s\n", "Failed to write to cache");
+ rv = 0;
+ goto err;
+ }
+ d->size = (int)total_len;
+ d->cached = 1;
+
+ if (total_len > size)
+ total_len = size;
+ memcpy(buf, d->buf, total_len);
+ rv = total_len;
+
+err:
+ free(cg);
+ return rv;
+}
+/* Return a positive number on success, return 0 on failure.*/
+pthread_t load_daemon(int load_use)
+{
+ int ret;
+ pthread_t pid;
+
+ ret = init_load();
+ if (ret == -1) {
+ lxcfs_error("%s\n", "Initialize hash_table fails in load_daemon!");
+ return 0;
+ }
+ ret = pthread_create(&pid, NULL, load_begin, NULL);
+ if (ret != 0) {
+ lxcfs_error("%s\n", "Create pthread fails in load_daemon!");
+ load_free();
+ return 0;
+ }
+ /* use loadavg, here loadavg = 1*/
+ loadavg = load_use;
+ return pid;
+}
+
+/* Returns 0 on success. */
+int stop_load_daemon(pthread_t pid)
+{
+ int s;
+
+ /* Signal the thread to gracefully stop */
+ loadavg_stop = 1;
+
+ s = pthread_join(pid, NULL); /* Make sure sub thread has been canceled. */
+ if (s != 0) {
+ lxcfs_error("%s\n", "stop_load_daemon error: failed to join");
+ return -1;
+ }
+
+ load_free();
+ loadavg_stop = 0;
+
+ return 0;
+}
static off_t get_procfile_size(const char *which)
{
strcmp(path, "/proc/uptime") == 0 ||
strcmp(path, "/proc/stat") == 0 ||
strcmp(path, "/proc/diskstats") == 0 ||
- strcmp(path, "/proc/swaps") == 0) {
+ strcmp(path, "/proc/swaps") == 0 ||
+ strcmp(path, "/proc/loadavg") == 0) {
sb->st_size = 0;
sb->st_mode = S_IFREG | 00444;
sb->st_nlink = 1;
filler(buf, "stat", NULL, 0) != 0 ||
filler(buf, "uptime", NULL, 0) != 0 ||
filler(buf, "diskstats", NULL, 0) != 0 ||
- filler(buf, "swaps", NULL, 0) != 0)
+ filler(buf, "swaps", NULL, 0) != 0 ||
+ filler(buf, "loadavg", NULL, 0) != 0)
return -EINVAL;
return 0;
}
type = LXC_TYPE_PROC_DISKSTATS;
else if (strcmp(path, "/proc/swaps") == 0)
type = LXC_TYPE_PROC_SWAPS;
+ else if (strcmp(path, "/proc/loadavg") == 0)
+ type = LXC_TYPE_PROC_LOADAVG;
if (type == -1)
return -ENOENT;
return proc_diskstats_read(buf, size, offset, fi);
case LXC_TYPE_PROC_SWAPS:
return proc_swaps_read(buf, size, offset, fi);
+ case LXC_TYPE_PROC_LOADAVG:
+ return proc_loadavg_read(buf, size, offset, fi);
default:
return -EINVAL;
}
projects / mirror_lxcfs.git / blobdiff