diff options
Diffstat (limited to 'src/core')
-rw-r--r-- | src/core/cgroup.c | 195 | ||||
-rw-r--r-- | src/core/cgroup.h | 40 | ||||
-rw-r--r-- | src/core/dbus-cgroup.c | 155 | ||||
-rw-r--r-- | src/core/dbus-execute.c | 54 | ||||
-rw-r--r-- | src/core/dbus-manager.c | 32 | ||||
-rw-r--r-- | src/core/dbus-unit.c | 1 | ||||
-rw-r--r-- | src/core/execute.c | 34 | ||||
-rw-r--r-- | src/core/kill.c | 5 | ||||
-rw-r--r-- | src/core/kill.h | 3 | ||||
-rw-r--r-- | src/core/load-fragment-gperf.gperf.m4 | 3 | ||||
-rw-r--r-- | src/core/load-fragment.c | 110 | ||||
-rw-r--r-- | src/core/load-fragment.h | 1 | ||||
-rw-r--r-- | src/core/main.c | 21 | ||||
-rw-r--r-- | src/core/manager.c | 41 | ||||
-rw-r--r-- | src/core/manager.h | 11 | ||||
-rw-r--r-- | src/core/service.c | 2 | ||||
-rw-r--r-- | src/core/shutdown.c | 22 | ||||
-rw-r--r-- | src/core/transaction.c | 4 | ||||
-rw-r--r-- | src/core/unit.c | 66 | ||||
-rw-r--r-- | src/core/unit.h | 5 |
20 files changed, 547 insertions, 258 deletions
diff --git a/src/core/cgroup.c b/src/core/cgroup.c index 48000d4e6e..0c790c33da 100644 --- a/src/core/cgroup.c +++ b/src/core/cgroup.c @@ -37,14 +37,18 @@ void cgroup_context_init(CGroupContext *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->blockio_weight = CGROUP_BLKIO_WEIGHT_INVALID; + c->startup_blockio_weight = CGROUP_BLKIO_WEIGHT_INVALID; + c->tasks_max = (uint64_t) -1; - c->cpu_quota_per_sec_usec = USEC_INFINITY; + c->netclass_type = CGROUP_NETCLASS_TYPE_NONE; } void cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a) { @@ -102,11 +106,12 @@ 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" @@ -114,6 +119,7 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { 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), @@ -133,7 +139,7 @@ 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); @@ -287,7 +293,7 @@ fail: return -errno; } -void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, ManagerState state) { +void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, uint32_t netclass, ManagerState state) { bool is_root; int r; @@ -309,11 +315,11 @@ 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, @@ -336,15 +342,15 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M } 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, - DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(uint64_t)+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, @@ -358,7 +364,7 @@ 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, @@ -485,6 +491,17 @@ void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, M log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set pids.max on %s: %m", path); } + + if (mask & CGROUP_MASK_NET_CLS) { + char buf[DECIMAL_STR_MAX(uint32_t)]; + + sprintf(buf, "%" PRIu32, netclass); + + r = cg_set_attribute("net_cls", path, "net_cls.classid", buf); + if (r < 0) + log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set net_cls.classid on %s: %m", path); + } } CGroupMask cgroup_context_get_mask(CGroupContext *c) { @@ -493,14 +510,14 @@ 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; @@ -517,6 +534,9 @@ CGroupMask cgroup_context_get_mask(CGroupContext *c) { c->tasks_max != (uint64_t) -1) mask |= CGROUP_MASK_PIDS; + if (c->netclass_type != CGROUP_NETCLASS_TYPE_NONE) + mask |= CGROUP_MASK_NET_CLS; + return mask; } @@ -884,6 +904,103 @@ static bool unit_has_mask_realized(Unit *u, CGroupMask target_mask) { return u->cgroup_realized && u->cgroup_realized_mask == target_mask; } +static int unit_find_free_netclass_cgroup(Unit *u, uint32_t *ret) { + + uint32_t start, i; + Manager *m; + + assert(u); + + m = u->manager; + + i = start = m->cgroup_netclass_registry_last; + + do { + i++; + + if (!hashmap_get(m->cgroup_netclass_registry, UINT_TO_PTR(i))) { + m->cgroup_netclass_registry_last = i; + *ret = i; + return 0; + } + + if (i == UINT32_MAX) + i = CGROUP_NETCLASS_FIXED_MAX; + + } while (i != start); + + return -ENOBUFS; +} + +int unit_add_to_netclass_cgroup(Unit *u) { + + CGroupContext *cc; + Unit *first; + void *key; + int r; + + assert(u); + + cc = unit_get_cgroup_context(u); + if (!cc) + return 0; + + switch (cc->netclass_type) { + case CGROUP_NETCLASS_TYPE_NONE: + return 0; + + case CGROUP_NETCLASS_TYPE_FIXED: + u->cgroup_netclass_id = cc->netclass_id; + break; + + case CGROUP_NETCLASS_TYPE_AUTO: + /* Allocate a new ID in case it was requested and not done yet */ + if (u->cgroup_netclass_id == 0) { + r = unit_find_free_netclass_cgroup(u, &u->cgroup_netclass_id); + if (r < 0) + return r; + + log_debug("Dynamically assigned netclass cgroup id %" PRIu32 " to %s", u->cgroup_netclass_id, u->id); + } + + break; + } + + r = hashmap_ensure_allocated(&u->manager->cgroup_netclass_registry, &trivial_hash_ops); + if (r < 0) + return r; + + key = UINT32_TO_PTR(u->cgroup_netclass_id); + first = hashmap_get(u->manager->cgroup_netclass_registry, key); + + if (first) { + LIST_PREPEND(cgroup_netclass, first, u); + return hashmap_replace(u->manager->cgroup_netclass_registry, key, u); + } + + return hashmap_put(u->manager->cgroup_netclass_registry, key, u); +} + +int unit_remove_from_netclass_cgroup(Unit *u) { + + Unit *head; + void *key; + + assert(u); + + key = UINT32_TO_PTR(u->cgroup_netclass_id); + + LIST_FIND_HEAD(cgroup_netclass, u, head); + LIST_REMOVE(cgroup_netclass, head, u); + + if (head) + return hashmap_replace(u->manager->cgroup_netclass_registry, key, head); + + hashmap_remove(u->manager->cgroup_netclass_registry, key); + + return 0; +} + /* Check if necessary controllers and attributes for a unit are in place. * * If so, do nothing. @@ -919,7 +1036,7 @@ static int unit_realize_cgroup_now(Unit *u, ManagerState state) { return r; /* Finally, apply the necessary attributes. */ - cgroup_context_apply(unit_get_cgroup_context(u), target_mask, u->cgroup_path, state); + cgroup_context_apply(unit_get_cgroup_context(u), target_mask, u->cgroup_path, u->cgroup_netclass_id, state); return 0; } @@ -1577,6 +1694,32 @@ 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", diff --git a/src/core/cgroup.h b/src/core/cgroup.h index 3ba09d56a4..457544b49f 100644 --- a/src/core/cgroup.h +++ b/src/core/cgroup.h @@ -26,6 +26,11 @@ #include "list.h" #include "time-util.h" +/* Maximum value for fixed (manual) net class ID assignment, + * and also the value at which the range of automatic assignments starts + */ +#define CGROUP_NETCLASS_FIXED_MAX UINT32_C(65535) + typedef struct CGroupContext CGroupContext; typedef struct CGroupDeviceAllow CGroupDeviceAllow; typedef struct CGroupBlockIODeviceWeight CGroupBlockIODeviceWeight; @@ -47,6 +52,17 @@ typedef enum CGroupDevicePolicy { _CGROUP_DEVICE_POLICY_INVALID = -1 } CGroupDevicePolicy; +typedef enum CGroupNetClassType { + /* Default - do not assign a net class */ + CGROUP_NETCLASS_TYPE_NONE, + + /* Automatically assign a net class */ + CGROUP_NETCLASS_TYPE_AUTO, + + /* Assign the net class that was provided by the user */ + CGROUP_NETCLASS_TYPE_FIXED, +} CGroupNetClassType; + struct CGroupDeviceAllow { LIST_FIELDS(CGroupDeviceAllow, device_allow); char *path; @@ -58,7 +74,7 @@ struct CGroupDeviceAllow { struct CGroupBlockIODeviceWeight { LIST_FIELDS(CGroupBlockIODeviceWeight, device_weights); char *path; - unsigned long weight; + uint64_t weight; }; struct CGroupBlockIODeviceBandwidth { @@ -74,12 +90,12 @@ struct CGroupContext { bool memory_accounting; bool tasks_accounting; - unsigned long cpu_shares; - unsigned long startup_cpu_shares; + uint64_t cpu_shares; + uint64_t startup_cpu_shares; usec_t cpu_quota_per_sec_usec; - unsigned long blockio_weight; - unsigned long startup_blockio_weight; + uint64_t blockio_weight; + uint64_t startup_blockio_weight; LIST_HEAD(CGroupBlockIODeviceWeight, blockio_device_weights); LIST_HEAD(CGroupBlockIODeviceBandwidth, blockio_device_bandwidths); @@ -88,9 +104,12 @@ struct CGroupContext { CGroupDevicePolicy device_policy; LIST_HEAD(CGroupDeviceAllow, device_allow); - bool delegate; + CGroupNetClassType netclass_type; + uint32_t netclass_id; uint64_t tasks_max; + + bool delegate; }; #include "unit.h" @@ -99,7 +118,7 @@ struct CGroupContext { void cgroup_context_init(CGroupContext *c); void cgroup_context_done(CGroupContext *c); void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix); -void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, ManagerState state); +void cgroup_context_apply(CGroupContext *c, CGroupMask mask, const char *path, uint32_t netclass_id, ManagerState state); CGroupMask cgroup_context_get_mask(CGroupContext *c); @@ -127,6 +146,9 @@ int unit_watch_cgroup(Unit *u); int unit_attach_pids_to_cgroup(Unit *u); +int unit_add_to_netclass_cgroup(Unit *u); +int unit_remove_from_netclass_cgroup(Unit *u); + int manager_setup_cgroup(Manager *m); void manager_shutdown_cgroup(Manager *m, bool delete); @@ -149,5 +171,9 @@ bool unit_cgroup_delegate(Unit *u); int unit_notify_cgroup_empty(Unit *u); int manager_notify_cgroup_empty(Manager *m, const char *group); +void unit_invalidate_cgroup(Unit *u, CGroupMask m); + +void manager_invalidate_startup_units(Manager *m); + const char* cgroup_device_policy_to_string(CGroupDevicePolicy i) _const_; CGroupDevicePolicy cgroup_device_policy_from_string(const char *s) _pure_; diff --git a/src/core/dbus-cgroup.c b/src/core/dbus-cgroup.c index 94de92c3dd..f334dc928d 100644 --- a/src/core/dbus-cgroup.c +++ b/src/core/dbus-cgroup.c @@ -133,34 +133,16 @@ static int property_get_device_allow( return sd_bus_message_close_container(reply); } -static int property_get_ulong_as_u64( - sd_bus *bus, - const char *path, - const char *interface, - const char *property, - sd_bus_message *reply, - void *userdata, - sd_bus_error *error) { - - unsigned long *ul = userdata; - - assert(bus); - assert(reply); - assert(ul); - - return sd_bus_message_append(reply, "t", *ul == (unsigned long) -1 ? (uint64_t) -1 : (uint64_t) *ul); -} - const sd_bus_vtable bus_cgroup_vtable[] = { SD_BUS_VTABLE_START(0), SD_BUS_PROPERTY("Delegate", "b", bus_property_get_bool, offsetof(CGroupContext, delegate), 0), SD_BUS_PROPERTY("CPUAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, cpu_accounting), 0), - SD_BUS_PROPERTY("CPUShares", "t", property_get_ulong_as_u64, offsetof(CGroupContext, cpu_shares), 0), - SD_BUS_PROPERTY("StartupCPUShares", "t", property_get_ulong_as_u64, offsetof(CGroupContext, startup_cpu_shares), 0), + SD_BUS_PROPERTY("CPUShares", "t", NULL, offsetof(CGroupContext, cpu_shares), 0), + SD_BUS_PROPERTY("StartupCPUShares", "t", NULL, offsetof(CGroupContext, startup_cpu_shares), 0), SD_BUS_PROPERTY("CPUQuotaPerSecUSec", "t", bus_property_get_usec, offsetof(CGroupContext, cpu_quota_per_sec_usec), 0), SD_BUS_PROPERTY("BlockIOAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, blockio_accounting), 0), - SD_BUS_PROPERTY("BlockIOWeight", "t", property_get_ulong_as_u64, offsetof(CGroupContext, blockio_weight), 0), - SD_BUS_PROPERTY("StartupBlockIOWeight", "t", property_get_ulong_as_u64, offsetof(CGroupContext, startup_blockio_weight), 0), + SD_BUS_PROPERTY("BlockIOWeight", "t", NULL, offsetof(CGroupContext, blockio_weight), 0), + SD_BUS_PROPERTY("StartupBlockIOWeight", "t", NULL, offsetof(CGroupContext, startup_blockio_weight), 0), SD_BUS_PROPERTY("BlockIODeviceWeight", "a(st)", property_get_blockio_device_weight, 0, 0), SD_BUS_PROPERTY("BlockIOReadBandwidth", "a(st)", property_get_blockio_device_bandwidths, 0, 0), SD_BUS_PROPERTY("BlockIOWriteBandwidth", "a(st)", property_get_blockio_device_bandwidths, 0, 0), @@ -230,56 +212,52 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->cpu_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_MASK_CPUACCT; + unit_invalidate_cgroup(u, CGROUP_MASK_CPUACCT|CGROUP_MASK_CPU); unit_write_drop_in_private(u, mode, name, b ? "CPUAccounting=yes" : "CPUAccounting=no"); } return 1; } else if (streq(name, "CPUShares")) { - uint64_t u64; - unsigned long ul; + uint64_t shares; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &shares); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul <= 0 || (uint64_t) ul != u64) - return sd_bus_error_set_errnof(error, EINVAL, "CPUShares value out of range"); - } + if (!CGROUP_CPU_SHARES_IS_OK(shares)) + return sd_bus_error_set_errnof(error, EINVAL, "CPUShares value out of range"); if (mode != UNIT_CHECK) { - c->cpu_shares = ul; - u->cgroup_realized_mask &= ~CGROUP_MASK_CPU; - unit_write_drop_in_private_format(u, mode, name, "CPUShares=%lu", ul); + c->cpu_shares = shares; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (shares == CGROUP_CPU_SHARES_INVALID) + unit_write_drop_in_private(u, mode, name, "CPUShares="); + else + unit_write_drop_in_private_format(u, mode, name, "CPUShares=%" PRIu64, shares); } return 1; } else if (streq(name, "StartupCPUShares")) { - uint64_t u64; - unsigned long ul; + uint64_t shares; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &shares); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul <= 0 || (uint64_t) ul != u64) - return sd_bus_error_set_errnof(error, EINVAL, "StartupCPUShares value out of range"); - } + if (!CGROUP_CPU_SHARES_IS_OK(shares)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupCPUShares value out of range"); if (mode != UNIT_CHECK) { - c->startup_cpu_shares = ul; - u->cgroup_realized_mask &= ~CGROUP_MASK_CPU; - unit_write_drop_in_private_format(u, mode, name, "StartupCPUShares=%lu", ul); + c->startup_cpu_shares = shares; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (shares == CGROUP_CPU_SHARES_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupCPUShares="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupCPUShares=%" PRIu64, shares); } return 1; @@ -296,7 +274,7 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->cpu_quota_per_sec_usec = u64; - u->cgroup_realized_mask &= ~CGROUP_MASK_CPU; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); unit_write_drop_in_private_format(u, mode, "CPUQuota", "CPUQuota=%0.f%%", (double) (c->cpu_quota_per_sec_usec / 10000)); } @@ -311,56 +289,52 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->blockio_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_MASK_BLKIO; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); unit_write_drop_in_private(u, mode, name, b ? "BlockIOAccounting=yes" : "BlockIOAccounting=no"); } return 1; } else if (streq(name, "BlockIOWeight")) { - uint64_t u64; - unsigned long ul; + uint64_t weight; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &weight); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul < 10 || ul > 1000) - return sd_bus_error_set_errnof(error, EINVAL, "BlockIOWeight value out of range"); - } + if (!CGROUP_BLKIO_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "BlockIOWeight value out of range"); if (mode != UNIT_CHECK) { - c->blockio_weight = ul; - u->cgroup_realized_mask &= ~CGROUP_MASK_BLKIO; - unit_write_drop_in_private_format(u, mode, name, "BlockIOWeight=%lu", ul); + c->blockio_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + + if (weight == CGROUP_BLKIO_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "BlockIOWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "BlockIOWeight=%" PRIu64, weight); } return 1; } else if (streq(name, "StartupBlockIOWeight")) { - uint64_t u64; - unsigned long ul; + uint64_t weight; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &weight); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul < 10 || ul > 1000) - return sd_bus_error_set_errnof(error, EINVAL, "StartupBlockIOWeight value out of range"); - } + if (CGROUP_BLKIO_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupBlockIOWeight value out of range"); if (mode != UNIT_CHECK) { - c->startup_blockio_weight = ul; - u->cgroup_realized_mask &= ~CGROUP_MASK_BLKIO; - unit_write_drop_in_private_format(u, mode, name, "StartupBlockIOWeight=%lu", ul); + c->startup_blockio_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + + if (weight == CGROUP_BLKIO_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupBlockIOWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupBlockIOWeight=%" PRIu64, weight); } return 1; @@ -429,7 +403,7 @@ int bus_cgroup_set_property( cgroup_context_free_blockio_device_bandwidth(c, a); } - u->cgroup_realized_mask &= ~CGROUP_MASK_BLKIO; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); f = open_memstream(&buf, &size); if (!f) @@ -455,17 +429,16 @@ int bus_cgroup_set_property( } else if (streq(name, "BlockIODeviceWeight")) { const char *path; - uint64_t u64; + uint64_t weight; unsigned n = 0; r = sd_bus_message_enter_container(message, 'a', "(st)"); if (r < 0) return r; - while ((r = sd_bus_message_read(message, "(st)", &path, &u64)) > 0) { - unsigned long ul = u64; + while ((r = sd_bus_message_read(message, "(st)", &path, &weight)) > 0) { - if (ul < 10 || ul > 1000) + if (!CGROUP_BLKIO_WEIGHT_IS_OK(weight) || weight == CGROUP_BLKIO_WEIGHT_INVALID) return sd_bus_error_set_errnof(error, EINVAL, "BlockIODeviceWeight out of range"); if (mode != UNIT_CHECK) { @@ -491,7 +464,7 @@ int bus_cgroup_set_property( LIST_PREPEND(device_weights,c->blockio_device_weights, a); } - a->weight = ul; + a->weight = weight; } n++; @@ -512,7 +485,7 @@ int bus_cgroup_set_property( cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights); } - u->cgroup_realized_mask &= ~CGROUP_MASK_BLKIO; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); f = open_memstream(&buf, &size); if (!f) @@ -520,7 +493,7 @@ int bus_cgroup_set_property( fputs("BlockIODeviceWeight=\n", f); LIST_FOREACH(device_weights, a, c->blockio_device_weights) - fprintf(f, "BlockIODeviceWeight=%s %lu\n", a->path, a->weight); + fprintf(f, "BlockIODeviceWeight=%s %" PRIu64 "\n", a->path, a->weight); fflush(f); unit_write_drop_in_private(u, mode, name, buf); @@ -537,7 +510,7 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->memory_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_MASK_MEMORY; + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); unit_write_drop_in_private(u, mode, name, b ? "MemoryAccounting=yes" : "MemoryAccounting=no"); } @@ -552,7 +525,7 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->memory_limit = limit; - u->cgroup_realized_mask &= ~CGROUP_MASK_MEMORY; + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); if (limit == (uint64_t) -1) unit_write_drop_in_private(u, mode, name, "MemoryLimit=infinity"); @@ -578,7 +551,7 @@ int bus_cgroup_set_property( char *buf; c->device_policy = p; - u->cgroup_realized_mask &= ~CGROUP_MASK_DEVICES; + unit_invalidate_cgroup(u, CGROUP_MASK_DEVICES); buf = strjoina("DevicePolicy=", policy); unit_write_drop_in_private(u, mode, name, buf); @@ -657,7 +630,7 @@ int bus_cgroup_set_property( cgroup_context_free_device_allow(c, c->device_allow); } - u->cgroup_realized_mask &= ~CGROUP_MASK_DEVICES; + unit_invalidate_cgroup(u, CGROUP_MASK_DEVICES); f = open_memstream(&buf, &size); if (!f) @@ -682,7 +655,7 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->tasks_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_MASK_PIDS; + unit_invalidate_cgroup(u, CGROUP_MASK_PIDS); unit_write_drop_in_private(u, mode, name, b ? "TasksAccounting=yes" : "TasksAccounting=no"); } @@ -697,7 +670,7 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->tasks_max = limit; - u->cgroup_realized_mask &= ~CGROUP_MASK_PIDS; + unit_invalidate_cgroup(u, CGROUP_MASK_PIDS); if (limit == (uint64_t) -1) unit_write_drop_in_private(u, mode, name, "TasksMax=infinity"); diff --git a/src/core/dbus-execute.c b/src/core/dbus-execute.c index ed55fcfca2..fd13c6d019 100644 --- a/src/core/dbus-execute.c +++ b/src/core/dbus-execute.c @@ -933,7 +933,10 @@ int bus_exec_context_set_transient_property( return 1; - } else if (streq(name, "IgnoreSIGPIPE")) { + } else if (STR_IN_SET(name, + "IgnoreSIGPIPE", "TTYVHangup", "TTYReset", + "PrivateTmp", "PrivateDevices", "PrivateNetwork", + "NoNewPrivileges")) { int b; r = sd_bus_message_read(message, "b", &b); @@ -941,39 +944,22 @@ int bus_exec_context_set_transient_property( return r; if (mode != UNIT_CHECK) { - c->ignore_sigpipe = b; - - unit_write_drop_in_private_format(u, mode, name, "IgnoreSIGPIPE=%s\n", yes_no(b)); - } - - return 1; - - } else if (streq(name, "TTYVHangup")) { - int b; - - r = sd_bus_message_read(message, "b", &b); - if (r < 0) - return r; - - if (mode != UNIT_CHECK) { - c->tty_vhangup = b; - - unit_write_drop_in_private_format(u, mode, name, "TTYVHangup=%s\n", yes_no(b)); - } - - return 1; - - } else if (streq(name, "TTYReset")) { - int b; - - r = sd_bus_message_read(message, "b", &b); - if (r < 0) - return r; - - if (mode != UNIT_CHECK) { - c->tty_reset = b; - - unit_write_drop_in_private_format(u, mode, name, "TTYReset=%s\n", yes_no(b)); + if (streq(name, "IgnoreSIGPIPE")) + c->ignore_sigpipe = b; + else if (streq(name, "TTYVHangup")) + c->tty_vhangup = b; + else if (streq(name, "TTYReset")) + c->tty_reset = b; + else if (streq(name, "PrivateTmp")) + c->private_tmp = b; + else if (streq(name, "PrivateDevices")) + c->private_devices = b; + else if (streq(name, "PrivateNetwork")) + c->private_network = b; + else if (streq(name, "NoNewPrivileges")) + c->no_new_privileges = b; + + unit_write_drop_in_private_format(u, mode, name, "%s=%s\n", name, yes_no(b)); } return 1; diff --git a/src/core/dbus-manager.c b/src/core/dbus-manager.c index 4e5d67fc19..561b6f8bfa 100644 --- a/src/core/dbus-manager.c +++ b/src/core/dbus-manager.c @@ -1201,8 +1201,10 @@ static int method_exit(sd_bus_message *message, void *userdata, sd_bus_error *er if (r < 0) return r; - if (m->running_as == MANAGER_SYSTEM) - return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Exit is only supported for user service managers."); + /* Exit() (in contrast to SetExitCode()) is actually allowed even if + * we are running on the host. It will fall back on reboot() in + * systemd-shutdown if it cannot do the exit() because it isn't a + * container. */ m->exit_code = MANAGER_EXIT; @@ -1450,6 +1452,30 @@ static int method_unset_and_set_environment(sd_bus_message *message, void *userd return sd_bus_reply_method_return(message, NULL); } +static int method_set_exit_code(sd_bus_message *message, void *userdata, sd_bus_error *error) { + uint8_t code; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "exit", error); + if (r < 0) + return r; + + r = sd_bus_message_read_basic(message, 'y', &code); + if (r < 0) + return r; + + if (m->running_as == MANAGER_SYSTEM && detect_container() <= 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "ExitCode can only be set for user service managers or in containers."); + + m->return_value = code; + + return sd_bus_reply_method_return(message, NULL); +} + static int method_list_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; Manager *m = userdata; @@ -1933,6 +1959,7 @@ const sd_bus_vtable bus_manager_vtable[] = { SD_BUS_WRITABLE_PROPERTY("ShutdownWatchdogUSec", "t", bus_property_get_usec, bus_property_set_usec, offsetof(Manager, shutdown_watchdog), 0), SD_BUS_PROPERTY("ControlGroup", "s", NULL, offsetof(Manager, cgroup_root), 0), SD_BUS_PROPERTY("SystemState", "s", property_get_system_state, 0, 0), + SD_BUS_PROPERTY("ExitCode", "y", bus_property_get_unsigned, offsetof(Manager, return_value), 0), SD_BUS_METHOD("GetUnit", "s", "o", method_get_unit, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("GetUnitByPID", "u", "o", method_get_unit_by_pid, SD_BUS_VTABLE_UNPRIVILEGED), @@ -1986,6 +2013,7 @@ const sd_bus_vtable bus_manager_vtable[] = { SD_BUS_METHOD("GetDefaultTarget", NULL, "s", method_get_default_target, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("PresetAllUnitFiles", "sbb", "a(sss)", method_preset_all_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("AddDependencyUnitFiles", "asssbb", "a(sss)", method_add_dependency_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("SetExitCode", "y", NULL, method_set_exit_code, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_SIGNAL("UnitNew", "so", 0), SD_BUS_SIGNAL("UnitRemoved", "so", 0), diff --git a/src/core/dbus-unit.c b/src/core/dbus-unit.c index f7e9795928..cd88a87340 100644 --- a/src/core/dbus-unit.c +++ b/src/core/dbus-unit.c @@ -679,6 +679,7 @@ const sd_bus_vtable bus_unit_vtable[] = { SD_BUS_PROPERTY("Asserts", "a(sbbsi)", property_get_conditions, offsetof(Unit, asserts), 0), SD_BUS_PROPERTY("LoadError", "(ss)", property_get_load_error, 0, SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("Transient", "b", bus_property_get_bool, offsetof(Unit, transient), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NetClass", "u", bus_property_get_unsigned, offsetof(Unit, cgroup_netclass_id), 0), SD_BUS_METHOD("Start", "s", "o", method_start, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("Stop", "s", "o", method_stop, SD_BUS_VTABLE_UNPRIVILEGED), diff --git a/src/core/execute.c b/src/core/execute.c index d1acda6682..6e14848cd4 100644 --- a/src/core/execute.c +++ b/src/core/execute.c @@ -629,15 +629,6 @@ static int enforce_groups(const ExecContext *context, const char *username, gid_ * we avoid NSS lookups for gid=0. */ if (context->group || username) { - - if (context->group) { - const char *g = context->group; - - r = get_group_creds(&g, &gid); - if (r < 0) - return r; - } - /* First step, initialize groups from /etc/groups */ if (username && gid != 0) { if (initgroups(username, gid) < 0) @@ -1160,8 +1151,8 @@ static void do_idle_pipe_dance(int idle_pipe[4]) { assert(idle_pipe); - safe_close(idle_pipe[1]); - safe_close(idle_pipe[2]); + idle_pipe[1] = safe_close(idle_pipe[1]); + idle_pipe[2] = safe_close(idle_pipe[2]); if (idle_pipe[0] >= 0) { int r; @@ -1169,18 +1160,20 @@ static void do_idle_pipe_dance(int idle_pipe[4]) { r = fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT_USEC); if (idle_pipe[3] >= 0 && r == 0 /* timeout */) { + ssize_t n; + /* Signal systemd that we are bored and want to continue. */ - r = write(idle_pipe[3], "x", 1); - if (r > 0) + n = write(idle_pipe[3], "x", 1); + if (n > 0) /* Wait for systemd to react to the signal above. */ fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT2_USEC); } - safe_close(idle_pipe[0]); + idle_pipe[0] = safe_close(idle_pipe[0]); } - safe_close(idle_pipe[3]); + idle_pipe[3] = safe_close(idle_pipe[3]); } static int build_environment( @@ -1412,6 +1405,17 @@ static int exec_child( } } + if (context->group) { + const char *g = context->group; + + r = get_group_creds(&g, &gid); + if (r < 0) { + *exit_status = EXIT_GROUP; + return r; + } + } + + /* If a socket is connected to STDIN/STDOUT/STDERR, we * must sure to drop O_NONBLOCK */ if (socket_fd >= 0) diff --git a/src/core/kill.c b/src/core/kill.c index 2de71c6bf9..bddfa4460f 100644 --- a/src/core/kill.c +++ b/src/core/kill.c @@ -60,7 +60,10 @@ DEFINE_STRING_TABLE_LOOKUP(kill_mode, KillMode); static const char* const kill_who_table[_KILL_WHO_MAX] = { [KILL_MAIN] = "main", [KILL_CONTROL] = "control", - [KILL_ALL] = "all" + [KILL_ALL] = "all", + [KILL_MAIN_FAIL] = "main-fail", + [KILL_CONTROL_FAIL] = "control-fail", + [KILL_ALL_FAIL] = "all-fail" }; DEFINE_STRING_TABLE_LOOKUP(kill_who, KillWho); diff --git a/src/core/kill.h b/src/core/kill.h index d5f125fa41..5d97abb104 100644 --- a/src/core/kill.h +++ b/src/core/kill.h @@ -50,6 +50,9 @@ typedef enum KillWho { KILL_MAIN, KILL_CONTROL, KILL_ALL, + KILL_MAIN_FAIL, + KILL_CONTROL_FAIL, + KILL_ALL_FAIL, _KILL_WHO_MAX, _KILL_WHO_INVALID = -1 } KillWho; diff --git a/src/core/load-fragment-gperf.gperf.m4 b/src/core/load-fragment-gperf.gperf.m4 index e056fd863c..ec744214c1 100644 --- a/src/core/load-fragment-gperf.gperf.m4 +++ b/src/core/load-fragment-gperf.gperf.m4 @@ -126,7 +126,8 @@ $1.BlockIOReadBandwidth, config_parse_blockio_bandwidth, 0, $1.BlockIOWriteBandwidth, config_parse_blockio_bandwidth, 0, offsetof($1, cgroup_context) $1.TasksAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.tasks_accounting) $1.TasksMax, config_parse_tasks_max, 0, offsetof($1, cgroup_context) -$1.Delegate, config_parse_bool, 0, offsetof($1, cgroup_context.delegate)' +$1.Delegate, config_parse_bool, 0, offsetof($1, cgroup_context.delegate) +$1.NetClass, config_parse_netclass, 0, offsetof($1, cgroup_context)' )m4_dnl Unit.Description, config_parse_unit_string_printf, 0, offsetof(Unit, description) Unit.Documentation, config_parse_documentation, 0, offsetof(Unit, documentation) diff --git a/src/core/load-fragment.c b/src/core/load-fragment.c index f7a8539910..fcf863c5c7 100644 --- a/src/core/load-fragment.c +++ b/src/core/load-fragment.c @@ -2605,26 +2605,19 @@ int config_parse_cpu_shares( void *data, void *userdata) { - unsigned long *shares = data, lu; + uint64_t *shares = data; int r; assert(filename); assert(lvalue); assert(rvalue); - if (isempty(rvalue)) { - *shares = (unsigned long) -1; - return 0; - } - - r = safe_atolu(rvalue, &lu); - if (r < 0 || lu <= 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "CPU shares '%s' invalid. Ignoring.", rvalue); + r = cg_cpu_shares_parse(rvalue, shares); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "CPU shares '%s' invalid. Ignoring.", rvalue); return 0; } - *shares = lu; return 0; } @@ -2805,26 +2798,19 @@ int config_parse_blockio_weight( void *data, void *userdata) { - unsigned long *weight = data, lu; + uint64_t *weight = data; int r; assert(filename); assert(lvalue); assert(rvalue); - if (isempty(rvalue)) { - *weight = (unsigned long) -1; - return 0; - } - - r = safe_atolu(rvalue, &lu); - if (r < 0 || lu < 10 || lu > 1000) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Block IO weight '%s' invalid. Ignoring.", rvalue); + r = cg_blkio_weight_parse(rvalue, weight); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Block IO weight '%s' invalid. Ignoring.", rvalue); return 0; } - *weight = lu; return 0; } @@ -2843,8 +2829,8 @@ int config_parse_blockio_device_weight( _cleanup_free_ char *path = NULL; CGroupBlockIODeviceWeight *w; CGroupContext *c = data; - unsigned long lu; const char *weight; + uint64_t u; size_t n; int r; @@ -2861,9 +2847,10 @@ int config_parse_blockio_device_weight( n = strcspn(rvalue, WHITESPACE); weight = rvalue + n; - if (!*weight) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Expected block device and device weight. Ignoring."); + weight += strspn(weight, WHITESPACE); + + if (isempty(weight)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Expected block device and device weight. Ignoring."); return 0; } @@ -2872,19 +2859,18 @@ int config_parse_blockio_device_weight( return log_oom(); if (!path_startswith(path, "/dev")) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid device node path '%s'. Ignoring.", path); + log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Invalid device node path '%s'. Ignoring.", path); return 0; } - weight += strspn(weight, WHITESPACE); - r = safe_atolu(weight, &lu); - if (r < 0 || lu < 10 || lu > 1000) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Block IO weight '%s' invalid. Ignoring.", rvalue); + r = cg_blkio_weight_parse(weight, &u); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Block IO weight '%s' invalid. Ignoring.", weight); return 0; } + assert(u != CGROUP_BLKIO_WEIGHT_INVALID); + w = new0(CGroupBlockIODeviceWeight, 1); if (!w) return log_oom(); @@ -2892,7 +2878,7 @@ int config_parse_blockio_device_weight( w->path = path; path = NULL; - w->weight = lu; + w->weight = u; LIST_PREPEND(device_weights, c->blockio_device_weights, w); return 0; @@ -2976,6 +2962,48 @@ int config_parse_blockio_bandwidth( return 0; } +int config_parse_netclass( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + CGroupContext *c = data; + unsigned v; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (streq(rvalue, "auto")) { + c->netclass_type = CGROUP_NETCLASS_TYPE_AUTO; + return 0; + } + + r = safe_atou32(rvalue, &v); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Netclass '%s' invalid. Ignoring.", rvalue); + return 0; + } + + if (v > CGROUP_NETCLASS_FIXED_MAX) + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Fixed netclass %" PRIu32 " out of allowed range (0-%d). Applying anyway.", v, (uint32_t) CGROUP_NETCLASS_FIXED_MAX); + + c->netclass_id = v; + c->netclass_type = CGROUP_NETCLASS_TYPE_FIXED; + + return 0; +} + DEFINE_CONFIG_PARSE_ENUM(config_parse_job_mode, job_mode, JobMode, "Failed to parse job mode"); int config_parse_job_mode_isolate( @@ -3021,6 +3049,7 @@ int config_parse_runtime_directory( void *userdata) { char***rt = data; + Unit *u = userdata; const char *word, *state; size_t l; int r; @@ -3037,12 +3066,19 @@ int config_parse_runtime_directory( } FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *n; + _cleanup_free_ char *t = NULL, *n = NULL; - n = strndup(word, l); - if (!n) + t = strndup(word, l); + if (!t) return log_oom(); + r = unit_name_printf(u, t, &n); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, -r, + "Failed to resolve specifiers, ignoring: %s", strerror(-r)); + continue; + } + if (!filename_is_valid(n)) { log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Runtime directory is not valid, ignoring assignment: %s", rvalue); diff --git a/src/core/load-fragment.h b/src/core/load-fragment.h index 638b343a6e..5d0a09249f 100644 --- a/src/core/load-fragment.h +++ b/src/core/load-fragment.h @@ -90,6 +90,7 @@ int config_parse_device_allow(const char *unit, const char *filename, unsigned l int config_parse_blockio_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_blockio_device_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_blockio_bandwidth(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_netclass(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_job_mode(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_job_mode_isolate(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_exec_selinux_context(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); diff --git a/src/core/main.c b/src/core/main.c index 200fe740da..9b59648279 100644 --- a/src/core/main.c +++ b/src/core/main.c @@ -1254,6 +1254,7 @@ int main(int argc, char *argv[]) { char *switch_root_dir = NULL, *switch_root_init = NULL; struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0); const char *error_message = NULL; + uint8_t shutdown_exit_code = 0; #ifdef HAVE_SYSV_COMPAT if (getpid() != 1 && strstr(program_invocation_short_name, "init")) { @@ -1764,11 +1765,6 @@ int main(int argc, char *argv[]) { switch (m->exit_code) { - case MANAGER_EXIT: - retval = EXIT_SUCCESS; - log_debug("Exit."); - goto finish; - case MANAGER_RELOAD: log_info("Reloading."); @@ -1810,11 +1806,13 @@ int main(int argc, char *argv[]) { log_notice("Switching root."); goto finish; + case MANAGER_EXIT: case MANAGER_REBOOT: case MANAGER_POWEROFF: case MANAGER_HALT: case MANAGER_KEXEC: { static const char * const table[_MANAGER_EXIT_CODE_MAX] = { + [MANAGER_EXIT] = "exit", [MANAGER_REBOOT] = "reboot", [MANAGER_POWEROFF] = "poweroff", [MANAGER_HALT] = "halt", @@ -1836,8 +1834,10 @@ int main(int argc, char *argv[]) { finish: pager_close(); - if (m) + if (m) { arg_shutdown_watchdog = m->shutdown_watchdog; + shutdown_exit_code = m->return_value; + } m = manager_free(m); for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) @@ -1978,7 +1978,8 @@ finish: if (shutdown_verb) { char log_level[DECIMAL_STR_MAX(int) + 1]; - const char* command_line[9] = { + char exit_code[DECIMAL_STR_MAX(uint8_t) + 1]; + const char* command_line[11] = { SYSTEMD_SHUTDOWN_BINARY_PATH, shutdown_verb, "--log-level", log_level, @@ -2015,6 +2016,12 @@ finish: if (log_get_show_location()) command_line[pos++] = "--log-location"; + if (streq(shutdown_verb, "exit")) { + command_line[pos++] = "--exit-code"; + command_line[pos++] = exit_code; + xsprintf(exit_code, "%d", shutdown_exit_code); + } + assert(pos < ELEMENTSOF(command_line)); if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) { diff --git a/src/core/manager.c b/src/core/manager.c index d918007bb8..4e672a8c48 100644 --- a/src/core/manager.c +++ b/src/core/manager.c @@ -317,6 +317,8 @@ static int manager_watch_idle_pipe(Manager *m) { static void manager_close_idle_pipe(Manager *m) { assert(m); + m->idle_pipe_event_source = sd_event_source_unref(m->idle_pipe_event_source); + safe_close_pair(m->idle_pipe); safe_close_pair(m->idle_pipe + 2); } @@ -577,6 +579,8 @@ int manager_new(ManagerRunningAs running_as, bool test_run, Manager **_m) { m->have_ask_password = -EINVAL; /* we don't know */ m->first_boot = -1; + m->cgroup_netclass_registry_last = CGROUP_NETCLASS_FIXED_MAX; + m->test_run = test_run; /* Reboot immediately if the user hits C-A-D more often than 7x per 2s */ @@ -602,14 +606,6 @@ int manager_new(ManagerRunningAs running_as, bool test_run, Manager **_m) { if (r < 0) goto fail; - r = set_ensure_allocated(&m->startup_units, NULL); - if (r < 0) - goto fail; - - r = set_ensure_allocated(&m->failed_units, NULL); - if (r < 0) - goto fail; - r = sd_event_default(&m->event); if (r < 0) goto fail; @@ -944,7 +940,6 @@ Manager* manager_free(Manager *m) { sd_event_source_unref(m->notify_event_source); sd_event_source_unref(m->time_change_event_source); sd_event_source_unref(m->jobs_in_progress_event_source); - sd_event_source_unref(m->idle_pipe_event_source); sd_event_source_unref(m->run_queue_event_source); safe_close(m->signal_fd); @@ -967,6 +962,8 @@ Manager* manager_free(Manager *m) { hashmap_free(m->cgroup_unit); set_free_free(m->unit_path_cache); + hashmap_free(m->cgroup_netclass_registry); + free(m->switch_root); free(m->switch_root_init); @@ -1962,7 +1959,6 @@ static int manager_dispatch_idle_pipe_fd(sd_event_source *source, int fd, uint32 m->no_console_output = m->n_on_console > 0; - m->idle_pipe_event_source = sd_event_source_unref(m->idle_pipe_event_source); manager_close_idle_pipe(m); return 0; @@ -2675,9 +2671,6 @@ static void manager_notify_finished(Manager *m) { } void manager_check_finished(Manager *m) { - Unit *u = NULL; - Iterator i; - assert(m); if (m->n_reloading > 0) @@ -2690,11 +2683,9 @@ void manager_check_finished(Manager *m) { return; if (hashmap_size(m->jobs) > 0) { - if (m->jobs_in_progress_event_source) /* Ignore any failure, this is only for feedback */ - (void) sd_event_source_set_time(m->jobs_in_progress_event_source, - now(CLOCK_MONOTONIC) + JOBS_IN_PROGRESS_WAIT_USEC); + (void) sd_event_source_set_time(m->jobs_in_progress_event_source, now(CLOCK_MONOTONIC) + JOBS_IN_PROGRESS_WAIT_USEC); return; } @@ -2702,7 +2693,6 @@ void manager_check_finished(Manager *m) { manager_flip_auto_status(m, false); /* Notify Type=idle units that we are done now */ - m->idle_pipe_event_source = sd_event_source_unref(m->idle_pipe_event_source); manager_close_idle_pipe(m); /* Turn off confirm spawn now */ @@ -2721,9 +2711,7 @@ void manager_check_finished(Manager *m) { manager_notify_finished(m); - SET_FOREACH(u, m->startup_units, i) - if (u->cgroup_path) - cgroup_context_apply(unit_get_cgroup_context(u), unit_get_own_mask(u), u->cgroup_path, manager_state(m)); + manager_invalidate_startup_units(m); } static int create_generator_dir(Manager *m, char **generator, const char *name) { @@ -3069,8 +3057,9 @@ const char *manager_get_runtime_prefix(Manager *m) { getenv("XDG_RUNTIME_DIR"); } -void manager_update_failed_units(Manager *m, Unit *u, bool failed) { +int manager_update_failed_units(Manager *m, Unit *u, bool failed) { unsigned size; + int r; assert(m); assert(u->manager == m); @@ -3078,13 +3067,19 @@ void manager_update_failed_units(Manager *m, Unit *u, bool failed) { size = set_size(m->failed_units); if (failed) { + r = set_ensure_allocated(&m->failed_units, NULL); + if (r < 0) + return log_oom(); + if (set_put(m->failed_units, u) < 0) - log_oom(); + return log_oom(); } else - set_remove(m->failed_units, u); + (void) set_remove(m->failed_units, u); if (set_size(m->failed_units) != size) bus_manager_send_change_signal(m); + + return 0; } ManagerState manager_state(Manager *m) { diff --git a/src/core/manager.h b/src/core/manager.h index 78a0e50a33..1384eb33a4 100644 --- a/src/core/manager.h +++ b/src/core/manager.h @@ -242,6 +242,11 @@ struct Manager { bool test_run:1; + /* If non-zero, exit with the following value when the systemd + * process terminate. Useful for containers: systemd-nspawn could get + * the return value. */ + uint8_t return_value; + ShowStatus show_status; bool confirm_spawn; bool no_console_output; @@ -303,6 +308,10 @@ struct Manager { const char *unit_log_format_string; int first_boot; + + /* Used for NetClass=auto units */ + Hashmap *cgroup_netclass_registry; + uint32_t cgroup_netclass_registry_last; }; int manager_new(ManagerRunningAs running_as, bool test_run, Manager **m); @@ -369,7 +378,7 @@ const char *manager_get_runtime_prefix(Manager *m); ManagerState manager_state(Manager *m); -void manager_update_failed_units(Manager *m, Unit *u, bool failed); +int manager_update_failed_units(Manager *m, Unit *u, bool failed); const char *manager_state_to_string(ManagerState m) _const_; ManagerState manager_state_from_string(const char *s) _pure_; diff --git a/src/core/service.c b/src/core/service.c index 248a9e8c62..fc28ba4d07 100644 --- a/src/core/service.c +++ b/src/core/service.c @@ -2737,6 +2737,8 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us case SERVICE_RELOAD: log_unit_warning(UNIT(s), "Reload operation timed out. Stopping."); + service_unwatch_control_pid(s); + service_kill_control_processes(s); s->reload_result = SERVICE_FAILURE_TIMEOUT; service_enter_running(s, SERVICE_SUCCESS); break; diff --git a/src/core/shutdown.c b/src/core/shutdown.c index 8cc6efc5b8..5296efce1d 100644 --- a/src/core/shutdown.c +++ b/src/core/shutdown.c @@ -48,6 +48,7 @@ #define FINALIZE_ATTEMPTS 50 static char* arg_verb; +static uint8_t arg_exit_code; static int parse_argv(int argc, char *argv[]) { enum { @@ -55,6 +56,7 @@ static int parse_argv(int argc, char *argv[]) { ARG_LOG_TARGET, ARG_LOG_COLOR, ARG_LOG_LOCATION, + ARG_EXIT_CODE, }; static const struct option options[] = { @@ -62,6 +64,7 @@ static int parse_argv(int argc, char *argv[]) { { "log-target", required_argument, NULL, ARG_LOG_TARGET }, { "log-color", optional_argument, NULL, ARG_LOG_COLOR }, { "log-location", optional_argument, NULL, ARG_LOG_LOCATION }, + { "exit-code", required_argument, NULL, ARG_EXIT_CODE }, {} }; @@ -110,6 +113,13 @@ static int parse_argv(int argc, char *argv[]) { break; + case ARG_EXIT_CODE: + r = safe_atou8(optarg, &arg_exit_code); + if (r < 0) + log_error("Failed to parse exit code %s, ignoring", optarg); + + break; + case '\001': if (!arg_verb) arg_verb = optarg; @@ -183,6 +193,8 @@ int main(int argc, char *argv[]) { cmd = RB_HALT_SYSTEM; else if (streq(arg_verb, "kexec")) cmd = LINUX_REBOOT_CMD_KEXEC; + else if (streq(arg_verb, "exit")) + cmd = 0; /* ignored, just checking that arg_verb is valid */ else { r = -EINVAL; log_error("Unknown action '%s'.", arg_verb); @@ -339,6 +351,16 @@ int main(int argc, char *argv[]) { if (!in_container) sync(); + if (streq(arg_verb, "exit")) { + if (in_container) + exit(arg_exit_code); + else { + /* We cannot exit() on the host, fallback on another + * method. */ + cmd = RB_POWER_OFF; + } + } + switch (cmd) { case LINUX_REBOOT_CMD_KEXEC: diff --git a/src/core/transaction.c b/src/core/transaction.c index b8f69ec6f3..2d120af4b5 100644 --- a/src/core/transaction.c +++ b/src/core/transaction.c @@ -736,8 +736,8 @@ int transaction_activate(Transaction *tr, Manager *m, JobMode mode, sd_bus_error if (m->idle_pipe[0] < 0 && m->idle_pipe[1] < 0 && m->idle_pipe[2] < 0 && m->idle_pipe[3] < 0) { - pipe2(m->idle_pipe, O_NONBLOCK|O_CLOEXEC); - pipe2(m->idle_pipe + 2, O_NONBLOCK|O_CLOEXEC); + (void) pipe2(m->idle_pipe, O_NONBLOCK|O_CLOEXEC); + (void) pipe2(m->idle_pipe + 2, O_NONBLOCK|O_CLOEXEC); } } diff --git a/src/core/unit.c b/src/core/unit.c index 2ebfb09a7a..3a6313e4a2 100644 --- a/src/core/unit.c +++ b/src/core/unit.c @@ -452,6 +452,7 @@ static void unit_free_requires_mounts_for(Unit *u) { static void unit_done(Unit *u) { ExecContext *ec; CGroupContext *cc; + int r; assert(u); @@ -468,6 +469,10 @@ static void unit_done(Unit *u) { cc = unit_get_cgroup_context(u); if (cc) cgroup_context_done(cc); + + r = unit_remove_from_netclass_cgroup(u); + if (r < 0) + log_warning_errno(r, "Unable to remove unit from netclass group: %m"); } void unit_free(Unit *u) { @@ -528,7 +533,7 @@ void unit_free(Unit *u) { unit_release_cgroup(u); - manager_update_failed_units(u->manager, u, false); + (void) manager_update_failed_units(u->manager, u, false); set_remove(u->manager->startup_units, u); free(u->description); @@ -1172,15 +1177,20 @@ static int unit_add_mount_dependencies(Unit *u) { static int unit_add_startup_units(Unit *u) { CGroupContext *c; + int r; c = unit_get_cgroup_context(u); if (!c) return 0; - if (c->startup_cpu_shares == (unsigned long) -1 && - c->startup_blockio_weight == (unsigned long) -1) + if (c->startup_cpu_shares == CGROUP_CPU_SHARES_INVALID && + c->startup_blockio_weight == CGROUP_BLKIO_WEIGHT_INVALID) return 0; + r = set_ensure_allocated(&u->manager->startup_units, NULL); + if (r < 0) + return r; + return set_put(u->manager->startup_units, u); } @@ -1236,6 +1246,14 @@ int unit_load(Unit *u) { } unit_update_cgroup_members_masks(u); + + /* If we are reloading, we need to wait for the deserializer + * to restore the net_cls ids that have been set previously */ + if (u->manager->n_reloading <= 0) { + r = unit_add_to_netclass_cgroup(u); + if (r < 0) + return r; + } } assert((u->load_state != UNIT_MERGED) == !u->merged_into); @@ -1807,7 +1825,7 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su } /* Keep track of failed units */ - manager_update_failed_units(u->manager, u, ns == UNIT_FAILED); + (void) manager_update_failed_units(u->manager, u, ns == UNIT_FAILED); /* Make sure the cgroup is always removed when we become inactive */ if (UNIT_IS_INACTIVE_OR_FAILED(ns)) @@ -2586,6 +2604,9 @@ int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { unit_serialize_item(u, f, "cgroup", u->cgroup_path); unit_serialize_item(u, f, "cgroup-realized", yes_no(u->cgroup_realized)); + if (u->cgroup_netclass_id) + unit_serialize_item_format(u, f, "netclass-id", "%" PRIu32, u->cgroup_netclass_id); + if (serialize_jobs) { if (u->job) { fprintf(f, "job\n"); @@ -2773,6 +2794,17 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { u->cgroup_realized = b; continue; + } else if (streq(l, "netclass-id")) { + r = safe_atou32(v, &u->cgroup_netclass_id); + if (r < 0) + log_unit_debug(u, "Failed to parse netclass ID %s, ignoring.", v); + else { + r = unit_add_to_netclass_cgroup(u); + if (r < 0) + log_unit_debug_errno(u, r, "Failed to add unit to netclass cgroup, ignoring: %m"); + } + + continue; } if (unit_can_serialize(u)) { @@ -3032,32 +3064,39 @@ int unit_kill_common( sd_bus_error *error) { int r = 0; + bool killed = false; - if (who == KILL_MAIN) { + if (IN_SET(who, KILL_MAIN, KILL_MAIN_FAIL)) { if (main_pid < 0) return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_PROCESS, "%s units have no main processes", unit_type_to_string(u->type)); else if (main_pid == 0) return sd_bus_error_set_const(error, BUS_ERROR_NO_SUCH_PROCESS, "No main process to kill"); } - if (who == KILL_CONTROL) { + if (IN_SET(who, KILL_CONTROL, KILL_CONTROL_FAIL)) { if (control_pid < 0) return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_PROCESS, "%s units have no control processes", unit_type_to_string(u->type)); else if (control_pid == 0) return sd_bus_error_set_const(error, BUS_ERROR_NO_SUCH_PROCESS, "No control process to kill"); } - if (who == KILL_CONTROL || who == KILL_ALL) - if (control_pid > 0) + if (IN_SET(who, KILL_CONTROL, KILL_CONTROL_FAIL, KILL_ALL, KILL_ALL_FAIL)) + if (control_pid > 0) { if (kill(control_pid, signo) < 0) r = -errno; + else + killed = true; + } - if (who == KILL_MAIN || who == KILL_ALL) - if (main_pid > 0) + if (IN_SET(who, KILL_MAIN, KILL_MAIN_FAIL, KILL_ALL, KILL_ALL_FAIL)) + if (main_pid > 0) { if (kill(main_pid, signo) < 0) r = -errno; + else + killed = true; + } - if (who == KILL_ALL && u->cgroup_path) { + if (IN_SET(who, KILL_ALL, KILL_ALL_FAIL) && u->cgroup_path) { _cleanup_set_free_ Set *pid_set = NULL; int q; @@ -3069,8 +3108,13 @@ int unit_kill_common( q = cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, signo, false, false, false, pid_set); if (q < 0 && q != -EAGAIN && q != -ESRCH && q != -ENOENT) r = q; + else + killed = true; } + if (r == 0 && !killed && IN_SET(who, KILL_ALL_FAIL, KILL_CONTROL_FAIL, KILL_ALL_FAIL)) + return -ESRCH; + return r; } diff --git a/src/core/unit.h b/src/core/unit.h index 3c7684411b..c868d75c79 100644 --- a/src/core/unit.h +++ b/src/core/unit.h @@ -161,6 +161,9 @@ struct Unit { /* CGroup realize members queue */ LIST_FIELDS(Unit, cgroup_queue); + /* Units with the same CGroup netclass */ + LIST_FIELDS(Unit, cgroup_netclass); + /* PIDs we keep an eye on. Note that a unit might have many * more, but these are the ones we care enough about to * process SIGCHLD for */ @@ -189,6 +192,8 @@ struct Unit { CGroupMask cgroup_members_mask; int cgroup_inotify_wd; + uint32_t cgroup_netclass_id; + /* How to start OnFailure units */ JobMode on_failure_job_mode; |