X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fcore%2Fcgroup.c;h=6815ca907ac5a41f331771c9a39d603aaa158882;hb=b1491eba40715438b0ac874f2be79d5622b971e1;hp=1327486509a6eaaf54af54119087f077f8429f42;hpb=01efdf13a6ee9a14fd6d8b41a5d522d5917e1fbc;p=elogind.git diff --git a/src/core/cgroup.c b/src/core/cgroup.c index 132748650..6815ca907 100644 --- a/src/core/cgroup.c +++ b/src/core/cgroup.c @@ -20,21 +20,28 @@ ***/ #include +#include #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) { @@ -81,6 +88,7 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { CGroupBlockIODeviceBandwidth *b; CGroupBlockIODeviceWeight *w; CGroupDeviceAllow *a; + char u[FORMAT_TIMESPAN_MAX]; assert(c); assert(f); @@ -92,16 +100,24 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { "%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, @@ -137,10 +153,8 @@ static int lookup_blkio_device(const char *p, dev_t *dev) { 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; @@ -186,7 +200,7 @@ static int whitelist_device(const char *path, const char *node, const char *acc) 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set devices.allow on %s: %s", path, strerror(-r)); return r; } @@ -202,10 +216,8 @@ static int whitelist_major(const char *path, const char *name, char type, const 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; @@ -246,7 +258,8 @@ static int whitelist_major(const char *path, const char *name, char type, const w++; w += strspn(w, WHITESPACE); - if (!streq(w, name)) + + if (fnmatch(name, w, 0) != 0) continue; sprintf(buf, @@ -257,17 +270,17 @@ static int whitelist_major(const char *path, const char *name, char type, const 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set devices.allow on %s: %s", path, strerror(-r)); } 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; @@ -282,12 +295,27 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha is_root = isempty(path) || path_equal(path, "/"); if ((mask & CGROUP_CPU) && !is_root) { - char buf[DECIMAL_STR_MAX(unsigned long) + 1]; + char buf[MAX(DECIMAL_STR_MAX(unsigned long), DECIMAL_STR_MAX(usec_t)) + 1]; - sprintf(buf, "%lu\n", c->cpu_shares); + 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set cpu.shares on %s: %s", path, strerror(-r)); + + 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set cpu.cfs_period_us on %s: %s", path, strerror(-r)); + + 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set cpu.cfs_quota_us on %s: %s", path, strerror(-r)); } if (mask & CGROUP_BLKIO) { @@ -298,10 +326,11 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha CGroupBlockIODeviceBandwidth *b; if (!is_root) { - sprintf(buf, "%lu\n", c->blockio_weight); + 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_warning("Failed to set blkio.weight on %s: %s", path, strerror(-r)); + log_full(r == -ENOENT ? LOG_DEBUG : 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) { @@ -314,7 +343,7 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha 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_error("Failed to set blkio.weight_device on %s: %s", path, strerror(-r)); + log_full(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set blkio.weight_device on %s: %s", path, strerror(-r)); } } @@ -332,7 +361,7 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set %s on %s: %s", a, path, strerror(-r)); } } @@ -346,7 +375,7 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha 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(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to set memory.limit_in_bytes on %s: %s", path, strerror(-r)); } if ((mask & CGROUP_DEVICE) && !is_root) { @@ -357,21 +386,27 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha else r = cg_set_attribute("devices", path, "devices.allow", "a"); if (r < 0) - log_warning("Failed to reset devices.list on %s: %s", path, strerror(-r)); + log_full(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, "Failed to reset devices.list on %s: %s", path, strerror(-r)); 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) { @@ -407,11 +442,15 @@ CGroupControllerMask cgroup_context_get_mask(CGroupContext *c) { /* 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; @@ -420,7 +459,8 @@ CGroupControllerMask cgroup_context_get_mask(CGroupContext *c) { 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; @@ -433,6 +473,19 @@ CGroupControllerMask unit_get_cgroup_mask(Unit *u) { 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); } @@ -467,18 +520,12 @@ CGroupControllerMask unit_get_members_mask(Unit *u) { } 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) { @@ -579,10 +626,8 @@ static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) { /* 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; @@ -591,7 +636,7 @@ static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) { /* Then, possibly move things over */ r = cg_migrate_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->cgroup_path, migrate_callback, u); if (r < 0) - log_warning("Failed to migrate cgroup from to %s: %s", u->cgroup_path, strerror(-r)); + log_warning_errno(r, "Failed to migrate cgroup from to %s: %m", u->cgroup_path); return 0; } @@ -608,7 +653,7 @@ static bool unit_has_mask_realized(Unit *u, CGroupControllerMask mask) { * 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; @@ -626,7 +671,7 @@ static int unit_realize_cgroup_now(Unit *u) { /* 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; } @@ -637,7 +682,7 @@ static int unit_realize_cgroup_now(Unit *u) { 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; } @@ -652,16 +697,19 @@ static void unit_add_to_cgroup_queue(Unit *u) { } 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++; } @@ -731,10 +779,10 @@ int unit_realize_cgroup(Unit *u) { 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); @@ -743,8 +791,10 @@ void unit_destroy_cgroup(Unit *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); @@ -752,7 +802,6 @@ void unit_destroy_cgroup(Unit *u) { u->cgroup_path = NULL; u->cgroup_realized = false; u->cgroup_realized_mask = 0; - } pid_t unit_search_main_pid(Unit *u) { @@ -794,31 +843,24 @@ 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"); @@ -834,47 +876,42 @@ int manager_setup_cgroup(Manager *m) { /* 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 hierarchial 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; } @@ -886,10 +923,7 @@ void manager_shutdown_cgroup(Manager *m, bool delete) { 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;