#include "alloc-util.h"
#include "blockdev-util.h"
+#include "bpf-devices.h"
#include "bpf-firewall.h"
#include "btrfs-util.h"
-#include "bpf-devices.h"
#include "bus-error.h"
#include "cgroup-util.h"
#include "cgroup.h"
#include "fd-util.h"
#include "fileio.h"
#include "fs-util.h"
+#include "nulstr-util.h"
#include "parse-util.h"
#include "path-util.h"
#include "process-util.h"
#include "string-util.h"
#include "virt.h"
-#define CGROUP_CPU_QUOTA_PERIOD_USEC ((usec_t) 100 * USEC_PER_MSEC)
+#define CGROUP_CPU_QUOTA_DEFAULT_PERIOD_USEC ((usec_t) 100 * USEC_PER_MSEC)
/* Returns the log level to use when cgroup attribute writes fail. When an attribute is missing or we have access
* problems we downgrade to LOG_DEBUG. This is supposed to be nice to container managers and kernels which want to mask
.cpu_weight = CGROUP_WEIGHT_INVALID,
.startup_cpu_weight = CGROUP_WEIGHT_INVALID,
.cpu_quota_per_sec_usec = USEC_INFINITY,
+ .cpu_quota_period_usec = USEC_INFINITY,
.cpu_shares = CGROUP_CPU_SHARES_INVALID,
.startup_cpu_shares = CGROUP_CPU_SHARES_INVALID,
CGroupDeviceAllow *a;
IPAddressAccessItem *iaai;
char u[FORMAT_TIMESPAN_MAX];
+ char v[FORMAT_TIMESPAN_MAX];
assert(c);
assert(f);
"%sCPUShares=%" PRIu64 "\n"
"%sStartupCPUShares=%" PRIu64 "\n"
"%sCPUQuotaPerSecSec=%s\n"
+ "%sCPUQuotaPeriodSec=%s\n"
"%sIOWeight=%" PRIu64 "\n"
"%sStartupIOWeight=%" PRIu64 "\n"
"%sBlockIOWeight=%" PRIu64 "\n"
prefix, c->cpu_shares,
prefix, c->startup_cpu_shares,
prefix, format_timespan(u, sizeof(u), c->cpu_quota_per_sec_usec, 1),
+ prefix, format_timespan(v, sizeof(v), c->cpu_quota_period_usec, 1),
prefix, c->io_weight,
prefix, c->startup_io_weight,
prefix, c->blockio_weight,
return CGROUP_CPU_SHARES_DEFAULT;
}
+usec_t cgroup_cpu_adjust_period(usec_t period, usec_t quota, usec_t resolution, usec_t max_period) {
+ /* kernel uses a minimum resolution of 1ms, so both period and (quota * period)
+ * need to be higher than that boundary. quota is specified in USecPerSec.
+ * Additionally, period must be at most max_period. */
+ assert(quota > 0);
+
+ return MIN(MAX3(period, resolution, resolution * USEC_PER_SEC / quota), max_period);
+}
+
+static usec_t cgroup_cpu_adjust_period_and_log(Unit *u, usec_t period, usec_t quota) {
+ usec_t new_period;
+
+ if (quota == USEC_INFINITY)
+ /* Always use default period for infinity quota. */
+ return CGROUP_CPU_QUOTA_DEFAULT_PERIOD_USEC;
+
+ if (period == USEC_INFINITY)
+ /* Default period was requested. */
+ period = CGROUP_CPU_QUOTA_DEFAULT_PERIOD_USEC;
+
+ /* Clamp to interval [1ms, 1s] */
+ new_period = cgroup_cpu_adjust_period(period, quota, USEC_PER_MSEC, USEC_PER_SEC);
+
+ if (new_period != period) {
+ char v[FORMAT_TIMESPAN_MAX];
+ log_unit_full(u, u->warned_clamping_cpu_quota_period ? LOG_DEBUG : LOG_WARNING, 0,
+ "Clamping CPU interval for cpu.max: period is now %s",
+ format_timespan(v, sizeof(v), new_period, 1));
+ u->warned_clamping_cpu_quota_period = true;
+ }
+
+ return new_period;
+}
+
static void cgroup_apply_unified_cpu_weight(Unit *u, uint64_t weight) {
char buf[DECIMAL_STR_MAX(uint64_t) + 2];
(void) set_attribute_and_warn(u, "cpu", "cpu.weight", buf);
}
-static void cgroup_apply_unified_cpu_quota(Unit *u, usec_t quota) {
+static void cgroup_apply_unified_cpu_quota(Unit *u, usec_t quota, usec_t period) {
char buf[(DECIMAL_STR_MAX(usec_t) + 1) * 2 + 1];
+ period = cgroup_cpu_adjust_period_and_log(u, period, quota);
if (quota != USEC_INFINITY)
xsprintf(buf, USEC_FMT " " USEC_FMT "\n",
- quota * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC, CGROUP_CPU_QUOTA_PERIOD_USEC);
+ MAX(quota * period / USEC_PER_SEC, USEC_PER_MSEC), period);
else
- xsprintf(buf, "max " USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC);
+ xsprintf(buf, "max " USEC_FMT "\n", period);
(void) set_attribute_and_warn(u, "cpu", "cpu.max", buf);
}
(void) set_attribute_and_warn(u, "cpu", "cpu.shares", buf);
}
-static void cgroup_apply_legacy_cpu_quota(Unit *u, usec_t quota) {
+static void cgroup_apply_legacy_cpu_quota(Unit *u, usec_t quota, usec_t period) {
char buf[DECIMAL_STR_MAX(usec_t) + 2];
- xsprintf(buf, USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC);
+ period = cgroup_cpu_adjust_period_and_log(u, period, quota);
+
+ xsprintf(buf, USEC_FMT "\n", period);
(void) set_attribute_and_warn(u, "cpu", "cpu.cfs_period_us", buf);
if (quota != USEC_INFINITY) {
- xsprintf(buf, USEC_FMT "\n", quota * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC);
+ xsprintf(buf, USEC_FMT "\n", MAX(quota * period / USEC_PER_SEC, USEC_PER_MSEC));
(void) set_attribute_and_warn(u, "cpu", "cpu.cfs_quota_us", buf);
} else
(void) set_attribute_and_warn(u, "cpu", "cpu.cfs_quota_us", "-1\n");
weight = CGROUP_WEIGHT_DEFAULT;
cgroup_apply_unified_cpu_weight(u, weight);
- cgroup_apply_unified_cpu_quota(u, c->cpu_quota_per_sec_usec);
+ cgroup_apply_unified_cpu_quota(u, c->cpu_quota_per_sec_usec, c->cpu_quota_period_usec);
} else {
uint64_t shares;
shares = CGROUP_CPU_SHARES_DEFAULT;
cgroup_apply_legacy_cpu_shares(u, shares);
- cgroup_apply_legacy_cpu_quota(u, c->cpu_quota_per_sec_usec);
+ cgroup_apply_legacy_cpu_quota(u, c->cpu_quota_per_sec_usec, c->cpu_quota_period_usec);
}
}
assert(u);
+ if (streq_ptr(u->cgroup_path, path))
+ return 0;
+
if (path) {
p = strdup(path);
if (!p)
return -ENOMEM;
- } else
- p = NULL;
-
- if (streq_ptr(u->cgroup_path, p))
- return 0;
+ }
if (p) {
r = hashmap_put(u->manager->cgroup_unit, p, u);
}
unit_release_cgroup(u);
-
u->cgroup_path = TAKE_PTR(p);
return 1;
int unit_search_main_pid(Unit *u, pid_t *ret) {
_cleanup_fclose_ FILE *f = NULL;
- pid_t pid = 0, npid, mypid;
+ pid_t pid = 0, npid;
int r;
assert(u);
if (r < 0)
return r;
- mypid = getpid_cached();
while (cg_read_pid(f, &npid) > 0) {
- pid_t ppid;
if (npid == pid)
continue;
- /* Ignore processes that aren't our kids */
- if (get_process_ppid(npid, &ppid) >= 0 && ppid != mypid)
+ if (pid_is_my_child(npid) == 0)
continue;
if (pid != 0)
pid_t pid;
while ((r = cg_read_pid(f, &pid)) > 0) {
- r = unit_watch_pid(u, pid);
+ r = unit_watch_pid(u, pid, false);
if (r < 0 && ret >= 0)
ret = r;
}
if (r < 0)
return log_error_errno(r, "Failed to create cgroup empty event source: %m");
+ /* Schedule cgroup empty checks early, but after having processed service notification messages or
+ * SIGCHLD signals, so that a cgroup running empty is always just the last safety net of
+ * notification, and we collected the metadata the notification and SIGCHLD stuff offers first. */
r = sd_event_source_set_priority(m->cgroup_empty_event_source, SD_EVENT_PRIORITY_NORMAL-5);
if (r < 0)
return log_error_errno(r, "Failed to set priority of cgroup empty event source: %m");
if (r < 0)
return log_error_errno(r, "Failed to watch control group inotify object: %m");
- /* Process cgroup empty notifications early, but after service notifications and SIGCHLD. Also
- * see handling of cgroup agent notifications, for the classic cgroup hierarchy support. */
- r = sd_event_source_set_priority(m->cgroup_inotify_event_source, SD_EVENT_PRIORITY_NORMAL-4);
+ /* Process cgroup empty notifications early. Note that when this event is dispatched it'll
+ * just add the unit to a cgroup empty queue, hence let's run earlier than that. Also see
+ * handling of cgroup agent notifications, for the classic cgroup hierarchy support. */
+ r = sd_event_source_set_priority(m->cgroup_inotify_event_source, SD_EVENT_PRIORITY_NORMAL-9);
if (r < 0)
return log_error_errno(r, "Failed to set priority of inotify event source: %m");