***/
#include <fcntl.h>
+#include <fnmatch.h>
#include "path-util.h"
#include "special.h"
#include "cgroup-util.h"
#include "cgroup.h"
+#define CGROUP_CPU_QUOTA_PERIOD_USEC ((usec_t) 100 * USEC_PER_MSEC)
+
void cgroup_context_init(CGroupContext *c) {
assert(c);
/* Initialize everything to the kernel defaults, assuming the
* structure is preinitialized to 0 */
- c->cpu_shares = 1024;
+ c->cpu_shares = (unsigned long) -1;
+ c->startup_cpu_shares = (unsigned long) -1;
c->memory_limit = (uint64_t) -1;
- c->blockio_weight = 1000;
+ c->blockio_weight = (unsigned long) -1;
+ c->startup_blockio_weight = (unsigned long) -1;
+
+ c->cpu_quota_per_sec_usec = USEC_INFINITY;
}
void cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a) {
CGroupBlockIODeviceBandwidth *b;
CGroupBlockIODeviceWeight *w;
CGroupDeviceAllow *a;
+ char u[FORMAT_TIMESPAN_MAX];
assert(c);
assert(f);
"%sBlockIOAccounting=%s\n"
"%sMemoryAccounting=%s\n"
"%sCPUShares=%lu\n"
+ "%sStartupCPUShares=%lu\n"
+ "%sCPUQuotaPerSecSec=%s\n"
"%sBlockIOWeight=%lu\n"
+ "%sStartupBlockIOWeight=%lu\n"
"%sMemoryLimit=%" PRIu64 "\n"
- "%sDevicePolicy=%s\n",
+ "%sDevicePolicy=%s\n"
+ "%sDelegate=%s\n",
prefix, yes_no(c->cpu_accounting),
prefix, yes_no(c->blockio_accounting),
prefix, yes_no(c->memory_accounting),
prefix, c->cpu_shares,
+ prefix, c->startup_cpu_shares,
+ prefix, format_timespan(u, sizeof(u), c->cpu_quota_per_sec_usec, 1),
prefix, c->blockio_weight,
+ prefix, c->startup_blockio_weight,
prefix, c->memory_limit,
- prefix, cgroup_device_policy_to_string(c->device_policy));
+ prefix, cgroup_device_policy_to_string(c->device_policy),
+ prefix, yes_no(c->delegate));
LIST_FOREACH(device_allow, a, c->device_allow)
fprintf(f,
assert(dev);
r = stat(p, &st);
- if (r < 0) {
- log_warning("Couldn't stat device %s: %m", p);
- return -errno;
- }
+ if (r < 0)
+ return log_warning_errno(errno, "Couldn't stat device %s: %m", p);
if (S_ISBLK(st.st_mode))
*dev = st.st_rdev;
r = cg_set_attribute("devices", path, "devices.allow", buf);
if (r < 0)
- log_warning("Failed to set devices.allow on %s: %s", path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set devices.allow on %s: %m", path);
return r;
}
assert(type == 'b' || type == 'c');
f = fopen("/proc/devices", "re");
- if (!f) {
- log_warning("Cannot open /proc/devices to resolve %s (%c): %m", name, type);
- return -errno;
- }
+ if (!f)
+ return log_warning_errno(errno, "Cannot open /proc/devices to resolve %s (%c): %m", name, type);
FOREACH_LINE(line, f, goto fail) {
char buf[2+DECIMAL_STR_MAX(unsigned)+3+4], *p, *w;
w++;
w += strspn(w, WHITESPACE);
- if (!streq(w, name))
+
+ if (fnmatch(name, w, 0) != 0)
continue;
sprintf(buf,
r = cg_set_attribute("devices", path, "devices.allow", buf);
if (r < 0)
- log_warning("Failed to set devices.allow on %s: %s", path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set devices.allow on %s: %m", path);
}
return 0;
fail:
- log_warning("Failed to read /proc/devices: %m");
+ log_warning_errno(errno, "Failed to read /proc/devices: %m");
return -errno;
}
-void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const char *path) {
+void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const char *path, ManagerState state) {
+ bool is_root;
int r;
assert(c);
if (mask == 0)
return;
- if (mask & CGROUP_CPU) {
- char buf[DECIMAL_STR_MAX(unsigned long) + 1];
+ /* Some cgroup attributes are not supported on the root cgroup,
+ * hence silently ignore */
+ is_root = isempty(path) || path_equal(path, "/");
+ if (is_root)
+ /* Make sure we don't try to display messages with an empty path. */
+ path = "/";
+
+ /* We generally ignore errors caused by read-only mounted
+ * cgroup trees (assuming we are running in a container then),
+ * and missing cgroups, i.e. EROFS and ENOENT. */
- sprintf(buf, "%lu\n", c->cpu_shares);
+ if ((mask & CGROUP_CPU) && !is_root) {
+ char buf[MAX(DECIMAL_STR_MAX(unsigned long), DECIMAL_STR_MAX(usec_t)) + 1];
+
+ sprintf(buf, "%lu\n",
+ IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_cpu_shares != (unsigned long) -1 ? c->startup_cpu_shares :
+ c->cpu_shares != (unsigned long) -1 ? c->cpu_shares : 1024);
r = cg_set_attribute("cpu", path, "cpu.shares", buf);
if (r < 0)
- log_warning("Failed to set cpu.shares on %s: %s", path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set cpu.shares on %s: %m", path);
+
+ sprintf(buf, USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC);
+ r = cg_set_attribute("cpu", path, "cpu.cfs_period_us", buf);
+ if (r < 0)
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set cpu.cfs_period_us on %s: %m", path);
+
+ if (c->cpu_quota_per_sec_usec != USEC_INFINITY) {
+ sprintf(buf, USEC_FMT "\n", c->cpu_quota_per_sec_usec * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC);
+ r = cg_set_attribute("cpu", path, "cpu.cfs_quota_us", buf);
+ } else
+ r = cg_set_attribute("cpu", path, "cpu.cfs_quota_us", "-1");
+ if (r < 0)
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set cpu.cfs_quota_us on %s: %m", path);
}
if (mask & CGROUP_BLKIO) {
CGroupBlockIODeviceWeight *w;
CGroupBlockIODeviceBandwidth *b;
- sprintf(buf, "%lu\n", c->blockio_weight);
- r = cg_set_attribute("blkio", path, "blkio.weight", buf);
- if (r < 0)
- log_warning("Failed to set blkio.weight on %s: %s", path, strerror(-r));
-
- /* FIXME: no way to reset this list */
- LIST_FOREACH(device_weights, w, c->blockio_device_weights) {
- dev_t dev;
-
- r = lookup_blkio_device(w->path, &dev);
- if (r < 0)
- continue;
-
- sprintf(buf, "%u:%u %lu", major(dev), minor(dev), w->weight);
- r = cg_set_attribute("blkio", path, "blkio.weight_device", buf);
+ if (!is_root) {
+ sprintf(buf, "%lu\n", IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_blockio_weight != (unsigned long) -1 ? c->startup_blockio_weight :
+ c->blockio_weight != (unsigned long) -1 ? c->blockio_weight : 1000);
+ r = cg_set_attribute("blkio", path, "blkio.weight", buf);
if (r < 0)
- log_error("Failed to set blkio.weight_device on %s: %s", path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set blkio.weight on %s: %m", path);
+
+ /* FIXME: no way to reset this list */
+ LIST_FOREACH(device_weights, w, c->blockio_device_weights) {
+ dev_t dev;
+
+ r = lookup_blkio_device(w->path, &dev);
+ if (r < 0)
+ continue;
+
+ sprintf(buf, "%u:%u %lu", major(dev), minor(dev), w->weight);
+ r = cg_set_attribute("blkio", path, "blkio.weight_device", buf);
+ if (r < 0)
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set blkio.weight_device on %s: %m", path);
+ }
}
/* FIXME: no way to reset this list */
sprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), b->bandwidth);
r = cg_set_attribute("blkio", path, a, buf);
if (r < 0)
- log_error("Failed to set %s on %s: %s", a, path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set %s on %s: %m", a, path);
}
}
- if (mask & CGROUP_MEMORY) {
+ if ((mask & CGROUP_MEMORY) && !is_root) {
if (c->memory_limit != (uint64_t) -1) {
char buf[DECIMAL_STR_MAX(uint64_t) + 1];
r = cg_set_attribute("memory", path, "memory.limit_in_bytes", "-1");
if (r < 0)
- log_error("Failed to set memory.limit_in_bytes on %s: %s", path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to set memory.limit_in_bytes on %s: %m", path);
}
- if (mask & CGROUP_DEVICE) {
+ if ((mask & CGROUP_DEVICE) && !is_root) {
CGroupDeviceAllow *a;
+ /* Changing the devices list of a populated cgroup
+ * might result in EINVAL, hence ignore EINVAL
+ * here. */
+
if (c->device_allow || c->device_policy != CGROUP_AUTO)
r = cg_set_attribute("devices", path, "devices.deny", "a");
else
r = cg_set_attribute("devices", path, "devices.allow", "a");
if (r < 0)
- log_error("Failed to reset devices.list on %s: %s", path, strerror(-r));
+ log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to reset devices.list on %s: %m", path);
if (c->device_policy == CGROUP_CLOSED ||
(c->device_policy == CGROUP_AUTO && c->device_allow)) {
static const char auto_devices[] =
- "/dev/null\0" "rw\0"
- "/dev/zero\0" "rw\0"
- "/dev/full\0" "rw\0"
- "/dev/random\0" "rw\0"
- "/dev/urandom\0" "rw\0";
+ "/dev/null\0" "rwm\0"
+ "/dev/zero\0" "rwm\0"
+ "/dev/full\0" "rwm\0"
+ "/dev/random\0" "rwm\0"
+ "/dev/urandom\0" "rwm\0"
+ "/dev/tty\0" "rwm\0"
+ "/dev/pts/ptmx\0" "rw\0"; /* /dev/pts/ptmx may not be duplicated, but accessed */
const char *x, *y;
NULSTR_FOREACH_PAIR(x, y, auto_devices)
whitelist_device(path, x, y);
+
+ whitelist_major(path, "pts", 'c', "rw");
+ whitelist_major(path, "kdbus", 'c', "rw");
+ whitelist_major(path, "kdbus/*", 'c', "rw");
}
LIST_FOREACH(device_allow, a, c->device_allow) {
/* Figure out which controllers we need */
- if (c->cpu_accounting || c->cpu_shares != 1024)
+ if (c->cpu_accounting ||
+ c->cpu_shares != (unsigned long) -1 ||
+ c->startup_cpu_shares != (unsigned long) -1 ||
+ c->cpu_quota_per_sec_usec != USEC_INFINITY)
mask |= CGROUP_CPUACCT | CGROUP_CPU;
if (c->blockio_accounting ||
- c->blockio_weight != 1000 ||
+ c->blockio_weight != (unsigned long) -1 ||
+ c->startup_blockio_weight != (unsigned long) -1 ||
c->blockio_device_weights ||
c->blockio_device_bandwidths)
mask |= CGROUP_BLKIO;
c->memory_limit != (uint64_t) -1)
mask |= CGROUP_MEMORY;
- if (c->device_allow || c->device_policy != CGROUP_AUTO)
+ if (c->device_allow ||
+ c->device_policy != CGROUP_AUTO)
mask |= CGROUP_DEVICE;
return mask;
if (!c)
return 0;
+ /* If delegation is turned on, then turn on all cgroups,
+ * unless the process we fork into it is known to drop
+ * privileges anyway, and shouldn't get access to the
+ * controllers anyway. */
+
+ if (c->delegate) {
+ ExecContext *e;
+
+ e = unit_get_exec_context(u);
+ if (!e || exec_context_maintains_privileges(e))
+ return _CGROUP_CONTROLLER_MASK_ALL;
+ }
+
return cgroup_context_get_mask(c);
}
}
CGroupControllerMask unit_get_siblings_mask(Unit *u) {
- CGroupControllerMask m;
-
assert(u);
if (UNIT_ISSET(u->slice))
- m = unit_get_members_mask(UNIT_DEREF(u->slice));
- else
- m = unit_get_cgroup_mask(u) | unit_get_members_mask(u);
+ return unit_get_members_mask(UNIT_DEREF(u->slice));
- /* Sibling propagation is only relevant for weight-based
- * controllers, so let's mask out everything else */
- return m & (CGROUP_CPU|CGROUP_BLKIO|CGROUP_CPUACCT);
+ return unit_get_cgroup_mask(u) | unit_get_members_mask(u);
}
CGroupControllerMask unit_get_target_mask(Unit *u) {
}
static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) {
- _cleanup_free_ char *path = NULL;
+ CGroupContext *c;
int r;
assert(u);
- path = unit_default_cgroup_path(u);
- if (!path)
- return log_oom();
+ c = unit_get_cgroup_context(u);
+ if (!c)
+ return 0;
- r = hashmap_put(u->manager->cgroup_unit, path, u);
- if (r < 0) {
- log_error(r == -EEXIST ? "cgroup %s exists already: %s" : "hashmap_put failed for %s: %s", path, strerror(-r));
- return r;
- }
- if (r > 0) {
- u->cgroup_path = path;
- path = NULL;
+ if (!u->cgroup_path) {
+ _cleanup_free_ char *path = NULL;
+
+ path = unit_default_cgroup_path(u);
+ if (!path)
+ return log_oom();
+
+ r = hashmap_put(u->manager->cgroup_unit, path, u);
+ if (r < 0) {
+ log_error(r == -EEXIST ? "cgroup %s exists already: %s" : "hashmap_put failed for %s: %s", path, strerror(-r));
+ return r;
+ }
+ if (r > 0) {
+ u->cgroup_path = path;
+ path = NULL;
+ }
}
/* First, create our own group */
r = cg_create_everywhere(u->manager->cgroup_supported, mask, u->cgroup_path);
- if (r < 0) {
- log_error("Failed to create cgroup %s: %s", u->cgroup_path, strerror(-r));
- return r;
- }
+ if (r < 0)
+ return log_error_errno(r, "Failed to create cgroup %s: %m", u->cgroup_path);
/* Keep track that this is now realized */
u->cgroup_realized = true;
u->cgroup_realized_mask = mask;
- /* Then, possibly move things over */
- r = cg_migrate_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->cgroup_path, migrate_callback, u);
+ if (u->type != UNIT_SLICE && !c->delegate) {
+
+ /* Then, possibly move things over, but not if
+ * subgroups may contain processes, which is the case
+ * for slice and delegation units. */
+ r = cg_migrate_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->cgroup_path, migrate_callback, u);
+ if (r < 0)
+ log_warning_errno(r, "Failed to migrate cgroup from to %s: %m", u->cgroup_path);
+ }
+
+ return 0;
+}
+
+int unit_attach_pids_to_cgroup(Unit *u) {
+ int r;
+ assert(u);
+
+ r = unit_realize_cgroup(u);
if (r < 0)
- log_warning("Failed to migrate cgroup from to %s: %s", u->cgroup_path, strerror(-r));
+ return r;
+
+ r = cg_attach_many_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->pids, migrate_callback, u);
+ if (r < 0)
+ return r;
return 0;
}
* If not, create paths, move processes over, and set attributes.
*
* Returns 0 on success and < 0 on failure. */
-static int unit_realize_cgroup_now(Unit *u) {
+static int unit_realize_cgroup_now(Unit *u, ManagerState state) {
CGroupControllerMask mask;
int r;
/* First, realize parents */
if (UNIT_ISSET(u->slice)) {
- r = unit_realize_cgroup_now(UNIT_DEREF(u->slice));
+ r = unit_realize_cgroup_now(UNIT_DEREF(u->slice), state);
if (r < 0)
return r;
}
return r;
/* Finally, apply the necessary attributes. */
- cgroup_context_apply(unit_get_cgroup_context(u), mask, u->cgroup_path);
+ cgroup_context_apply(unit_get_cgroup_context(u), mask, u->cgroup_path, state);
return 0;
}
}
unsigned manager_dispatch_cgroup_queue(Manager *m) {
- Unit *i;
+ ManagerState state;
unsigned n = 0;
+ Unit *i;
int r;
+ state = manager_state(m);
+
while ((i = m->cgroup_queue)) {
assert(i->in_cgroup_queue);
- r = unit_realize_cgroup_now(i);
+ r = unit_realize_cgroup_now(i, state);
if (r < 0)
- log_warning("Failed to realize cgroups for queued unit %s: %s", i->id, strerror(-r));
+ log_warning_errno(r, "Failed to realize cgroups for queued unit %s: %m", i->id);
n++;
}
unit_queue_siblings(u);
/* And realize this one now (and apply the values) */
- return unit_realize_cgroup_now(u);
+ return unit_realize_cgroup_now(u, manager_state(u->manager));
}
-void unit_destroy_cgroup(Unit *u) {
+void unit_destroy_cgroup_if_empty(Unit *u) {
int r;
assert(u);
return;
r = cg_trim_everywhere(u->manager->cgroup_supported, u->cgroup_path, !unit_has_name(u, SPECIAL_ROOT_SLICE));
- if (r < 0)
- log_debug("Failed to destroy cgroup %s: %s", u->cgroup_path, strerror(-r));
+ if (r < 0) {
+ log_debug_errno(r, "Failed to destroy cgroup %s: %m", u->cgroup_path);
+ return;
+ }
hashmap_remove(u->manager->cgroup_unit, u->cgroup_path);
u->cgroup_path = NULL;
u->cgroup_realized = false;
u->cgroup_realized_mask = 0;
-
}
pid_t unit_search_main_pid(Unit *u) {
int manager_setup_cgroup(Manager *m) {
_cleanup_free_ char *path = NULL;
- char *e;
int r;
assert(m);
- /* 0. Be nice to Ingo Molnar #628004 */
- if (path_is_mount_point("/sys/fs/cgroup/systemd", false) <= 0) {
- log_warning("No control group support available, not creating root group.");
- return 0;
- }
-
/* 1. Determine hierarchy */
free(m->cgroup_root);
m->cgroup_root = NULL;
r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, &m->cgroup_root);
- if (r < 0) {
- log_error("Cannot determine cgroup we are running in: %s", strerror(-r));
- return r;
- }
+ if (r < 0)
+ return log_error_errno(r, "Cannot determine cgroup we are running in: %m");
/* LEGACY: Already in /system.slice? If so, let's cut this
* off. This is to support live upgrades from older systemd
* versions where PID 1 was moved there. */
if (m->running_as == SYSTEMD_SYSTEM) {
+ char *e;
+
e = endswith(m->cgroup_root, "/" SPECIAL_SYSTEM_SLICE);
if (!e)
e = endswith(m->cgroup_root, "/system");
/* 2. Show data */
r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, NULL, &path);
- if (r < 0) {
- log_error("Cannot find cgroup mount point: %s", strerror(-r));
- return r;
- }
+ if (r < 0)
+ return log_error_errno(r, "Cannot find cgroup mount point: %m");
log_debug("Using cgroup controller " SYSTEMD_CGROUP_CONTROLLER ". File system hierarchy is at %s.", path);
+ if (!m->test_run) {
- /* 3. Install agent */
- if (m->running_as == SYSTEMD_SYSTEM) {
- r = cg_install_release_agent(SYSTEMD_CGROUP_CONTROLLER, SYSTEMD_CGROUP_AGENT_PATH);
+ /* 3. Install agent */
+ if (m->running_as == SYSTEMD_SYSTEM) {
+ r = cg_install_release_agent(SYSTEMD_CGROUP_CONTROLLER, SYSTEMD_CGROUP_AGENT_PATH);
+ if (r < 0)
+ log_warning_errno(r, "Failed to install release agent, ignoring: %m");
+ else if (r > 0)
+ log_debug("Installed release agent.");
+ else
+ log_debug("Release agent already installed.");
+ }
+
+ /* 4. Make sure we are in the root cgroup */
+ r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, 0);
if (r < 0)
- log_warning("Failed to install release agent, ignoring: %s", strerror(-r));
- else if (r > 0)
- log_debug("Installed release agent.");
- else
- log_debug("Release agent already installed.");
- }
+ return log_error_errno(r, "Failed to create root cgroup hierarchy: %m");
- /* 4. Make sure we are in the root cgroup */
- r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, 0);
- if (r < 0) {
- log_error("Failed to create root cgroup hierarchy: %s", strerror(-r));
- return r;
- }
+ /* 5. And pin it, so that it cannot be unmounted */
+ safe_close(m->pin_cgroupfs_fd);
- /* 5. And pin it, so that it cannot be unmounted */
- if (m->pin_cgroupfs_fd >= 0)
- close_nointr_nofail(m->pin_cgroupfs_fd);
+ m->pin_cgroupfs_fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY|O_NOCTTY|O_NONBLOCK);
+ if (m->pin_cgroupfs_fd < 0)
+ return log_error_errno(errno, "Failed to open pin file: %m");
- m->pin_cgroupfs_fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY|O_NOCTTY|O_NONBLOCK);
- if (r < 0) {
- log_error("Failed to open pin file: %m");
- return -errno;
+ /* 6. Always enable hierarchical support if it exists... */
+ cg_set_attribute("memory", "/", "memory.use_hierarchy", "1");
}
- /* 6. Figure out which controllers are supported */
+ /* 7. Figure out which controllers are supported */
m->cgroup_supported = cg_mask_supported();
- /* 7. Always enable hierarchial support if it exists... */
- cg_set_attribute("memory", "/", "memory.use_hierarchy", "1");
-
return 0;
}
if (delete && m->cgroup_root)
cg_trim(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, false);
- if (m->pin_cgroupfs_fd >= 0) {
- close_nointr_nofail(m->pin_cgroupfs_fd);
- m->pin_cgroupfs_fd = -1;
- }
+ m->pin_cgroupfs_fd = safe_close(m->pin_cgroupfs_fd);
free(m->cgroup_root);
m->cgroup_root = NULL;