X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fcore%2Fcgroup.c;h=8d7d57e7a04499569f855065847d5744487731c0;hb=58172968ad2f9a350172f062cc459d070569916d;hp=54da47182f29ad4f8736b5019b35d668e11a1065;hpb=3b22396a4b2767a98172f6915929c47738cb0a1e;p=elogind.git diff --git a/src/core/cgroup.c b/src/core/cgroup.c index 54da47182..8d7d57e7a 100644 --- a/src/core/cgroup.c +++ b/src/core/cgroup.c @@ -22,29 +22,38 @@ #include #include -#include "process-util.h" -#include "path-util.h" -// #include "special.h" +#include "alloc-util.h" #include "cgroup-util.h" #include "cgroup.h" +#include "fd-util.h" +#include "fileio.h" +#include "fs-util.h" +#include "parse-util.h" +#include "path-util.h" +#include "process-util.h" +//#include "special.h" +#include "string-table.h" +#include "string-util.h" #define CGROUP_CPU_QUOTA_PERIOD_USEC ((usec_t) 100 * USEC_PER_MSEC) -// UNNEEDED by elogind -#if 0 +#if 0 /// UNNEEDED by elogind void cgroup_context_init(CGroupContext *c) { assert(c); /* Initialize everything to the kernel defaults, assuming the * structure is preinitialized to 0 */ - c->cpu_shares = (unsigned long) -1; - c->startup_cpu_shares = (unsigned long) -1; + c->cpu_shares = CGROUP_CPU_SHARES_INVALID; + c->startup_cpu_shares = CGROUP_CPU_SHARES_INVALID; + c->cpu_quota_per_sec_usec = USEC_INFINITY; + c->memory_limit = (uint64_t) -1; - c->blockio_weight = (unsigned long) -1; - c->startup_blockio_weight = (unsigned long) -1; - c->cpu_quota_per_sec_usec = USEC_INFINITY; + c->blockio_weight = CGROUP_BLKIO_WEIGHT_INVALID; + c->startup_blockio_weight = CGROUP_BLKIO_WEIGHT_INVALID; + + c->tasks_max = (uint64_t) -1; } void cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a) { @@ -102,23 +111,27 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { "%sCPUAccounting=%s\n" "%sBlockIOAccounting=%s\n" "%sMemoryAccounting=%s\n" - "%sCPUShares=%lu\n" - "%sStartupCPUShares=%lu\n" + "%sTasksAccounting=%s\n" + "%sCPUShares=%" PRIu64 "\n" + "%sStartupCPUShares=%" PRIu64 "\n" "%sCPUQuotaPerSecSec=%s\n" - "%sBlockIOWeight=%lu\n" - "%sStartupBlockIOWeight=%lu\n" + "%sBlockIOWeight=%" PRIu64 "\n" + "%sStartupBlockIOWeight=%" PRIu64 "\n" "%sMemoryLimit=%" PRIu64 "\n" + "%sTasksMax=%" PRIu64 "\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, yes_no(c->tasks_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, c->tasks_max, prefix, cgroup_device_policy_to_string(c->device_policy), prefix, yes_no(c->delegate)); @@ -131,7 +144,7 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { LIST_FOREACH(device_weights, w, c->blockio_device_weights) fprintf(f, - "%sBlockIODeviceWeight=%s %lu", + "%sBlockIODeviceWeight=%s %" PRIu64, prefix, w->path, w->weight); @@ -203,7 +216,7 @@ 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_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set devices.allow on %s: %m", path); return r; @@ -274,7 +287,7 @@ 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_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set devices.allow on %s: %m", path); } @@ -307,20 +320,20 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M * and missing cgroups, i.e. EROFS and ENOENT. */ if ((mask & CGROUP_MASK_CPU) && !is_root) { - char buf[MAX(DECIMAL_STR_MAX(unsigned long), DECIMAL_STR_MAX(usec_t)) + 1]; + char buf[MAX(DECIMAL_STR_MAX(uint64_t), 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); + sprintf(buf, "%" PRIu64 "\n", + IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID ? c->startup_cpu_shares : + c->cpu_shares != CGROUP_CPU_SHARES_INVALID ? c->cpu_shares : CGROUP_CPU_SHARES_DEFAULT); r = cg_set_attribute("cpu", path, "cpu.shares", buf); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? 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, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? 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) { @@ -329,23 +342,23 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M } 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, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set cpu.cfs_quota_us on %s: %m", path); } if (mask & CGROUP_MASK_BLKIO) { - char buf[MAX3(DECIMAL_STR_MAX(unsigned long)+1, - DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(unsigned long)*1, + char buf[MAX(DECIMAL_STR_MAX(uint64_t)+1, DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(uint64_t)+1)]; CGroupBlockIODeviceWeight *w; CGroupBlockIODeviceBandwidth *b; 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); + sprintf(buf, "%" PRIu64 "\n", + IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID ? c->startup_blockio_weight : + c->blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID ? c->blockio_weight : CGROUP_BLKIO_WEIGHT_DEFAULT); r = cg_set_attribute("blkio", path, "blkio.weight", buf); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set blkio.weight on %s: %m", path); /* FIXME: no way to reset this list */ @@ -356,10 +369,10 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M if (r < 0) continue; - sprintf(buf, "%u:%u %lu", major(dev), minor(dev), w->weight); + sprintf(buf, "%u:%u %" PRIu64 "\n", 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, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set blkio.weight_device on %s: %m", path); } } @@ -378,7 +391,7 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M sprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), b->bandwidth); r = cg_set_attribute("blkio", path, a, buf); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set %s on %s: %m", a, path); } } @@ -402,11 +415,11 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M } if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set memory.limit_in_bytes/memory.max on %s: %m", path); } - if ((mask & CGROUP_MASK_DEVICE) && !is_root) { + if ((mask & CGROUP_MASK_DEVICES) && !is_root) { CGroupDeviceAllow *a; /* Changing the devices list of a populated cgroup @@ -418,7 +431,7 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M else r = cg_set_attribute("devices", path, "devices.allow", "a"); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to reset devices.list on %s: %m", path); if (c->device_policy == CGROUP_CLOSED || @@ -468,6 +481,21 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M log_debug("Ignoring device %s while writing cgroup attribute.", a->path); } } + + if ((mask & CGROUP_MASK_PIDS) && !is_root) { + + if (c->tasks_max != (uint64_t) -1) { + char buf[DECIMAL_STR_MAX(uint64_t) + 2]; + + sprintf(buf, "%" PRIu64 "\n", c->tasks_max); + r = cg_set_attribute("pids", path, "pids.max", buf); + } else + r = cg_set_attribute("pids", path, "pids.max", "max"); + + if (r < 0) + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set pids.max on %s: %m", path); + } } CGroupMask cgroup_context_get_mask(CGroupContext *c) { @@ -476,14 +504,14 @@ CGroupMask cgroup_context_get_mask(CGroupContext *c) { /* Figure out which controllers we need */ if (c->cpu_accounting || - c->cpu_shares != (unsigned long) -1 || - c->startup_cpu_shares != (unsigned long) -1 || + c->cpu_shares != CGROUP_CPU_SHARES_INVALID || + c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID || c->cpu_quota_per_sec_usec != USEC_INFINITY) mask |= CGROUP_MASK_CPUACCT | CGROUP_MASK_CPU; if (c->blockio_accounting || - c->blockio_weight != (unsigned long) -1 || - c->startup_blockio_weight != (unsigned long) -1 || + c->blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID || + c->startup_blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID || c->blockio_device_weights || c->blockio_device_bandwidths) mask |= CGROUP_MASK_BLKIO; @@ -494,7 +522,11 @@ CGroupMask cgroup_context_get_mask(CGroupContext *c) { if (c->device_allow || c->device_policy != CGROUP_AUTO) - mask |= CGROUP_MASK_DEVICE; + mask |= CGROUP_MASK_DEVICES; + + if (c->tasks_accounting || + c->tasks_max != (uint64_t) -1) + mask |= CGROUP_MASK_PIDS; return mask; } @@ -1065,7 +1097,7 @@ int unit_search_main_pid(Unit *u, pid_t *ret) { continue; /* Ignore processes that aren't our kids */ - if (get_parent_of_pid(npid, &ppid) >= 0 && ppid != mypid) + if (get_process_ppid(npid, &ppid) >= 0 && ppid != mypid) continue; if (pid != 0) @@ -1230,8 +1262,7 @@ int manager_setup_cgroup(Manager *m) { if (r < 0) return log_error_errno(r, "Cannot determine cgroup we are running in: %m"); -/// elogind does not support systemd scopes and slices -#if 0 +#if 0 /// elogind does not support systemd scopes and slices /* Chop off the init scope, if we are already located in it */ e = endswith(m->cgroup_root, "/" SPECIAL_INIT_SCOPE); @@ -1239,7 +1270,7 @@ int manager_setup_cgroup(Manager *m) { * it. This is to support live upgrades from older systemd * versions where PID 1 was moved there. Also see * cg_get_root_path(). */ - if (!e) { + if (!e && m->running_as == MANAGER_SYSTEM) { e = endswith(m->cgroup_root, "/" SPECIAL_SYSTEM_SLICE); if (!e) e = endswith(m->cgroup_root, "/system"); /* even more legacy */ @@ -1277,8 +1308,8 @@ int manager_setup_cgroup(Manager *m) { /* In the unified hierarchy we can can get * cgroup empty notifications via inotify. */ -/// elogind does not support the unified hierarchy, yet. -#if 0 + +#if 0 /// elogind does not support the unified hierarchy, yet. m->cgroup_inotify_event_source = sd_event_source_unref(m->cgroup_inotify_event_source); safe_close(m->cgroup_inotify_fd); @@ -1300,6 +1331,7 @@ int manager_setup_cgroup(Manager *m) { return log_error_errno(EOPNOTSUPP, "Unified cgroup hierarchy not supported: %m"); #endif // 0 } else if (m->running_as == MANAGER_SYSTEM) { + /* On the legacy hierarchy we only get * notifications via cgroup agents. (Which * isn't really reliable, since it does not @@ -1315,8 +1347,7 @@ int manager_setup_cgroup(Manager *m) { log_debug("Release agent already installed."); } -/// elogind is not meant to run in systemd init scope -#if 0 +#if 0 /// elogind is not meant to run in systemd init scope /* 4. Make sure we are in the special "init.scope" unit in the root slice. */ scope_path = strjoina(m->cgroup_root, "/" SPECIAL_INIT_SCOPE); r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, scope_path, 0); @@ -1374,8 +1405,7 @@ void manager_shutdown_cgroup(Manager *m, bool delete) { if (delete && m->cgroup_root) (void) cg_trim(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, false); -/// elogind does not support the unified hierarchy, yet. -#if 0 +#if 0 /// elogind does not support the unified hierarchy, yet. m->cgroup_inotify_wd_unit = hashmap_free(m->cgroup_inotify_wd_unit); m->cgroup_inotify_event_source = sd_event_source_unref(m->cgroup_inotify_event_source); @@ -1387,8 +1417,7 @@ void manager_shutdown_cgroup(Manager *m, bool delete) { m->cgroup_root = mfree(m->cgroup_root); } -/// UNNEEDED by elogind -#if 0 +#if 0 /// UNNEEDED by elogind Unit* manager_get_unit_by_cgroup(Manager *m, const char *cgroup) { char *p; Unit *u; @@ -1492,6 +1521,28 @@ int unit_get_memory_current(Unit *u, uint64_t *ret) { return safe_atou64(v, ret); } +int unit_get_tasks_current(Unit *u, uint64_t *ret) { + _cleanup_free_ char *v = NULL; + int r; + + assert(u); + assert(ret); + + if (!u->cgroup_path) + return -ENODATA; + + if ((u->cgroup_realized_mask & CGROUP_MASK_PIDS) == 0) + return -ENODATA; + + r = cg_get_attribute("pids", u->cgroup_path, "pids.current", &v); + if (r == -ENOENT) + return -ENODATA; + if (r < 0) + return r; + + return safe_atou64(v, ret); +} + static int unit_get_cpu_usage_raw(Unit *u, nsec_t *ret) { _cleanup_free_ char *v = NULL; uint64_t ns; @@ -1565,6 +1616,32 @@ bool unit_cgroup_delegate(Unit *u) { return c->delegate; } +void unit_invalidate_cgroup(Unit *u, CGroupMask m) { + assert(u); + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return; + + if (m == 0) + return; + + if ((u->cgroup_realized_mask & m) == 0) + return; + + u->cgroup_realized_mask &= ~m; + unit_add_to_cgroup_queue(u); +} + +void manager_invalidate_startup_units(Manager *m) { + Iterator i; + Unit *u; + + assert(m); + + SET_FOREACH(u, m->startup_units, i) + unit_invalidate_cgroup(u, CGROUP_MASK_CPU|CGROUP_MASK_BLKIO); +} + static const char* const cgroup_device_policy_table[_CGROUP_DEVICE_POLICY_MAX] = { [CGROUP_AUTO] = "auto", [CGROUP_CLOSED] = "closed",