diff --git a/man/org.freedesktop.systemd1.xml b/man/org.freedesktop.systemd1.xml index cbeb25efcd..027a8deeb4 100644 --- a/man/org.freedesktop.systemd1.xml +++ b/man/org.freedesktop.systemd1.xml @@ -552,6 +552,10 @@ node /org/freedesktop/systemd1 { readonly t DefaultMemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly s DefaultMemoryPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t DefaultCPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s DefaultCPUPressureWatch = '...'; @org.freedesktop.DBus.Property.EmitsChangedSignal("const") readonly t TimerSlackNSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("const") @@ -793,6 +797,10 @@ node /org/freedesktop/systemd1 { + + + + @@ -1243,6 +1251,10 @@ node /org/freedesktop/systemd1 { + + + + @@ -3066,6 +3078,10 @@ node /org/freedesktop/systemd1/unit/avahi_2ddaemon_2eservice { @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly t MemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s CPUPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t CPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly a(iiss) NFTSet = [...]; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly b CoredumpReceive = ...; @@ -3735,6 +3751,10 @@ node /org/freedesktop/systemd1/unit/avahi_2ddaemon_2eservice { + + + + @@ -4427,6 +4447,10 @@ node /org/freedesktop/systemd1/unit/avahi_2ddaemon_2eservice { + + + + @@ -5326,6 +5350,10 @@ node /org/freedesktop/systemd1/unit/avahi_2ddaemon_2esocket { @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly t MemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s CPUPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t CPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly a(iiss) NFTSet = [...]; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly b CoredumpReceive = ...; @@ -6011,6 +6039,10 @@ node /org/freedesktop/systemd1/unit/avahi_2ddaemon_2esocket { + + + + @@ -6677,6 +6709,10 @@ node /org/freedesktop/systemd1/unit/avahi_2ddaemon_2esocket { + + + + @@ -7399,6 +7435,10 @@ node /org/freedesktop/systemd1/unit/home_2emount { @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly t MemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s CPUPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t CPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly a(iiss) NFTSet = [...]; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly b CoredumpReceive = ...; @@ -8008,6 +8048,10 @@ node /org/freedesktop/systemd1/unit/home_2emount { + + + + @@ -8582,6 +8626,10 @@ node /org/freedesktop/systemd1/unit/home_2emount { + + + + @@ -9437,6 +9485,10 @@ node /org/freedesktop/systemd1/unit/dev_2dsda3_2eswap { @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly t MemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s CPUPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t CPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly a(iiss) NFTSet = [...]; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly b CoredumpReceive = ...; @@ -10028,6 +10080,10 @@ node /org/freedesktop/systemd1/unit/dev_2dsda3_2eswap { + + + + @@ -10584,6 +10640,10 @@ node /org/freedesktop/systemd1/unit/dev_2dsda3_2eswap { + + + + @@ -11292,6 +11352,10 @@ node /org/freedesktop/systemd1/unit/system_2eslice { @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly t MemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s CPUPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t CPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly a(iiss) NFTSet = [...]; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly b CoredumpReceive = ...; @@ -11465,6 +11529,10 @@ node /org/freedesktop/systemd1/unit/system_2eslice { + + + + @@ -11653,6 +11721,10 @@ node /org/freedesktop/systemd1/unit/system_2eslice { + + + + @@ -11864,6 +11936,10 @@ node /org/freedesktop/systemd1/unit/session_2d1_2escope { @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly t MemoryPressureThresholdUSec = ...; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly s CPUPressureWatch = '...'; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") + readonly t CPUPressureThresholdUSec = ...; + @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly a(iiss) NFTSet = [...]; @org.freedesktop.DBus.Property.EmitsChangedSignal("false") readonly b CoredumpReceive = ...; @@ -12051,6 +12127,10 @@ node /org/freedesktop/systemd1/unit/session_2d1_2escope { + + + + @@ -12263,6 +12343,10 @@ node /org/freedesktop/systemd1/unit/session_2d1_2escope { + + + + @@ -12475,7 +12559,9 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ RemoveSubgroupFromUnit(), and KillUnitSubgroup() were added in version 258. TransactionsWithOrderingCycle was added in version 259. - DefaultMemoryZSwapWriteback was added in version 261. + DefaultMemoryZSwapWriteback, + DefaultCPUPressureThresholdUSec and + DefaultCPUPressureWatch were added in version 261. Unit Objects @@ -12567,6 +12653,8 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ ExecReloadPostEx were added in version 259. BindNetworkInterface, MemoryTHP, RefreshOnReload, and RootMStack were added in version 260. + CPUPressureThresholdUSec and + CPUPressureWatch were added in version 261. Socket Unit Objects @@ -12637,6 +12725,8 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ ManagedOOMKills were added in 259. BindNetworkInterface MemoryTHP, and RootMStack were added in version 260. + CPUPressureThresholdUSec and + CPUPressureWatch were added in version 261. Mount Unit Objects @@ -12702,6 +12792,8 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ ManagedOOMKills were added in 259. BindNetworkInterface MemoryTHP, and RootMStack were added in version 260. + CPUPressureThresholdUSec and + CPUPressureWatch were added in version 261. Swap Unit Objects @@ -12765,6 +12857,8 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ ManagedOOMKills were added in 259. BindNetworkInterface, MemoryTHP, and RootMStack were added in version 260. + CPUPressureThresholdUSec and + CPUPressureWatch were added in version 261. Slice Unit Objects @@ -12798,6 +12892,8 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ OOMKills, and ManagedOOMKills were added in 259. BindNetworkInterface was added in version 260. + CPUPressureThresholdUSec and + CPUPressureWatch were added in version 261. Scope Unit Objects @@ -12829,6 +12925,8 @@ $ gdbus introspect --system --dest org.freedesktop.systemd1 \ OOMKills, and ManagedOOMKills were added in 259. BindNetworkInterface was added in version 260. + CPUPressureThresholdUSec and + CPUPressureWatch were added in version 261. Job Objects diff --git a/man/systemd-system.conf.xml b/man/systemd-system.conf.xml index e9e7d3d78d..79133dc15e 100644 --- a/man/systemd-system.conf.xml +++ b/man/systemd-system.conf.xml @@ -326,6 +326,20 @@ + + + DefaultCPUPressureWatch= + DefaultCPUPressureThresholdSec= + + Configures the default settings for the per-unit + CPUPressureWatch= and CPUPressureThresholdSec= + settings. See + systemd.resource-control5 + for details. Defaults to auto and 200ms, respectively. This + also sets the CPU pressure monitoring threshold for the service manager itself. + + + diff --git a/man/systemd.exec.xml b/man/systemd.exec.xml index 48bec7361b..1048fcadfc 100644 --- a/man/systemd.exec.xml +++ b/man/systemd.exec.xml @@ -4705,6 +4705,18 @@ StandardInputData=V2XigLJyZSBubyBzdHJhbmdlcnMgdG8gbG92ZQpZb3Uga25vdyB0aGUgcnVsZX + + $CPU_PRESSURE_WATCH + $CPU_PRESSURE_WRITE + + If CPU pressure monitoring is enabled for this service unit, the path to watch + and the data to write into it. See Resource Pressure + Handling for details about these variables and the service protocol data they + convey. + + + + $FDSTORE diff --git a/man/systemd.resource-control.xml b/man/systemd.resource-control.xml index ac31971e54..8d9e27f3d3 100644 --- a/man/systemd.resource-control.xml +++ b/man/systemd.resource-control.xml @@ -1657,6 +1657,55 @@ DeviceAllow=/dev/loop-control + + + CPUPressureWatch= + + Controls CPU pressure monitoring for invoked processes. Takes a boolean or one of + auto and skip. If no, tells the service not + to watch for CPU pressure events, by setting the $CPU_PRESSURE_WATCH + environment variable to the literal string /dev/null. If yes, + tells the service to watch for CPU pressure events. This ensures the + cpu.pressure cgroup attribute file is accessible for + reading and writing by the service's user. It then sets the $CPU_PRESSURE_WATCH + environment variable for processes invoked by the unit to the file system path to this file. The + threshold information configured with CPUPressureThresholdSec= is encoded in + the $CPU_PRESSURE_WRITE environment variable. If the auto + value is set the protocol is enabled if CPU resource controls are configured for the unit (e.g. because + CPUWeight= or CPUQuota= is set), and + disabled otherwise. If set to skip the logic is neither enabled, nor disabled and + the two environment variables are not set. + + Note that services are free to use the two environment variables, but it is unproblematic if + they ignore them. CPU pressure handling must be implemented individually in each service, and + usually means different things for different software. + + Services implemented using + sd-event3 may use + sd_event_add_cpu_pressure3 + to watch for and handle CPU pressure events. + + If not explicitly set, defaults to the DefaultCPUPressureWatch= setting in + systemd-system.conf5. + + + + + + CPUPressureThresholdSec= + + Sets the CPU pressure threshold time for CPU pressure monitor as configured via + CPUPressureWatch=. Specifies the maximum CPU stall time before a CPU + pressure event is signalled to the service, per 2s window. If not specified, defaults to the + DefaultCPUPressureThresholdSec= setting in + systemd-system.conf5 + (which in turn defaults to 200ms). The specified value expects a time unit such as + ms or μs, see + systemd.time7 for + details on the permitted syntax. + + + Coredump Control diff --git a/src/basic/psi-util.h b/src/basic/psi-util.h index aed74ef742..b8737b8976 100644 --- a/src/basic/psi-util.h +++ b/src/basic/psi-util.h @@ -43,6 +43,11 @@ typedef struct PressureResourceInfo { extern const PressureResourceInfo pressure_resource_info[_PRESSURE_RESOURCE_MAX]; +static inline const PressureResourceInfo* pressure_resource_get_info(PressureResource resource) { + assert(resource >= 0 && resource < _PRESSURE_RESOURCE_MAX); + return &pressure_resource_info[resource]; +} + DECLARE_STRING_TABLE_LOOKUP(pressure_resource, PressureResource); /* Default parameters for pressure watch logic in sd-event and PID 1 */ diff --git a/src/core/cgroup.c b/src/core/cgroup.c index 7bcd6777df..a9982de659 100644 --- a/src/core/cgroup.c +++ b/src/core/cgroup.c @@ -185,8 +185,10 @@ void cgroup_context_init(CGroupContext *c) { * moom_mem_pressure_duration_usec is set to infinity. */ .moom_mem_pressure_duration_usec = USEC_INFINITY, - .memory_pressure_watch = _CGROUP_PRESSURE_WATCH_INVALID, - .memory_pressure_threshold_usec = USEC_INFINITY, + .pressure = { + [PRESSURE_MEMORY] = { .watch = _CGROUP_PRESSURE_WATCH_INVALID, .threshold_usec = USEC_INFINITY }, + [PRESSURE_CPU] = { .watch = _CGROUP_PRESSURE_WATCH_INVALID, .threshold_usec = USEC_INFINITY }, + }, }; } @@ -528,6 +530,7 @@ void cgroup_context_dump(Unit *u, FILE* f, const char *prefix) { "%sManagedOOMMemoryPressureLimit: " PERMYRIAD_AS_PERCENT_FORMAT_STR "\n" "%sManagedOOMPreference: %s\n" "%sMemoryPressureWatch: %s\n" + "%sCPUPressureWatch: %s\n" "%sCoredumpReceive: %s\n", prefix, yes_no(c->io_accounting), prefix, yes_no(c->memory_accounting), @@ -563,7 +566,8 @@ void cgroup_context_dump(Unit *u, FILE* f, const char *prefix) { prefix, managed_oom_mode_to_string(c->moom_mem_pressure), prefix, PERMYRIAD_AS_PERCENT_FORMAT_VAL(UINT32_SCALE_TO_PERMYRIAD(c->moom_mem_pressure_limit)), prefix, managed_oom_preference_to_string(c->moom_preference), - prefix, cgroup_pressure_watch_to_string(c->memory_pressure_watch), + prefix, cgroup_pressure_watch_to_string(c->pressure[PRESSURE_MEMORY].watch), + prefix, cgroup_pressure_watch_to_string(c->pressure[PRESSURE_CPU].watch), prefix, yes_no(c->coredump_receive)); if (c->delegate_subgroup) @@ -574,9 +578,13 @@ void cgroup_context_dump(Unit *u, FILE* f, const char *prefix) { fprintf(f, "%sBindNetworkInterface: %s\n", prefix, c->bind_network_interface); - if (c->memory_pressure_threshold_usec != USEC_INFINITY) + if (c->pressure[PRESSURE_MEMORY].threshold_usec != USEC_INFINITY) fprintf(f, "%sMemoryPressureThresholdSec: %s\n", - prefix, FORMAT_TIMESPAN(c->memory_pressure_threshold_usec, 1)); + prefix, FORMAT_TIMESPAN(c->pressure[PRESSURE_MEMORY].threshold_usec, 1)); + + if (c->pressure[PRESSURE_CPU].threshold_usec != USEC_INFINITY) + fprintf(f, "%sCPUPressureThresholdSec: %s\n", + prefix, FORMAT_TIMESPAN(c->pressure[PRESSURE_CPU].threshold_usec, 1)); if (c->moom_mem_pressure_duration_usec != USEC_INFINITY) fprintf(f, "%sManagedOOMMemoryPressureDurationSec: %s\n", @@ -2107,12 +2115,13 @@ static int unit_update_cgroup( cgroup_context_apply(u, target_mask, state); cgroup_xattr_apply(u); - /* For most units we expect that memory monitoring is set up before the unit is started and we won't - * touch it after. For PID 1 this is different though, because we couldn't possibly do that given - * that PID 1 runs before init.scope is even set up. Hence, whenever init.scope is realized, let's - * try to open the memory pressure interface anew. */ + /* For most units we expect that pressure monitoring is set up before the unit is started and we + * won't touch it after. For PID 1 this is different though, because we couldn't possibly do that + * given that PID 1 runs before init.scope is even set up. Hence, whenever init.scope is realized, + * let's try to open the pressure interfaces anew. */ if (unit_has_name(u, SPECIAL_INIT_SCOPE)) - (void) manager_setup_memory_pressure_event_source(u->manager); + for (PressureResource t = 0; t < _PRESSURE_RESOURCE_MAX; t++) + (void) manager_setup_pressure_event_source(u->manager, t); return 0; } diff --git a/src/core/cgroup.h b/src/core/cgroup.h index 0cd290e92f..c4a2276567 100644 --- a/src/core/cgroup.h +++ b/src/core/cgroup.h @@ -6,6 +6,7 @@ #include "cpu-set-util.h" #include "firewall-util.h" #include "list.h" +#include "psi-util.h" typedef struct CGroupTasksMax { /* If scale == 0, just use value; otherwise, value / scale. @@ -95,14 +96,19 @@ typedef struct CGroupSocketBindItem { } CGroupSocketBindItem; typedef enum CGroupPressureWatch { - CGROUP_PRESSURE_WATCH_NO, /* → tells the service payload explicitly not to watch for memory pressure */ + CGROUP_PRESSURE_WATCH_NO, /* → tells the service payload explicitly not to watch for pressure */ CGROUP_PRESSURE_WATCH_YES, - CGROUP_PRESSURE_WATCH_AUTO, /* → on if memory account is on anyway for the unit, otherwise off */ - CGROUP_PRESSURE_WATCH_SKIP, /* → doesn't set up memory pressure watch, but also doesn't explicitly tell payload to avoid it */ + CGROUP_PRESSURE_WATCH_AUTO, /* → on if relevant accounting is on anyway for the unit, otherwise off */ + CGROUP_PRESSURE_WATCH_SKIP, /* → doesn't set up pressure watch, but also doesn't explicitly tell payload to avoid it */ _CGROUP_PRESSURE_WATCH_MAX, _CGROUP_PRESSURE_WATCH_INVALID = -EINVAL, } CGroupPressureWatch; +typedef struct CGroupPressure { + CGroupPressureWatch watch; + usec_t threshold_usec; +} CGroupPressure; + /* The user-supplied cgroup-related configuration options. This remains mostly immutable while the service * manager is running (except for an occasional SetProperties() configuration change), outside of reload * cycles. */ @@ -189,11 +195,8 @@ typedef struct CGroupContext { usec_t moom_mem_pressure_duration_usec; ManagedOOMPreference moom_preference; - /* Memory pressure logic */ - CGroupPressureWatch memory_pressure_watch; - usec_t memory_pressure_threshold_usec; - /* NB: For now we don't make the period configurable, not the type, nor do we allow multiple - * triggers, nor triggers for non-memory pressure. We might add that later. */ + /* Pressure logic */ + CGroupPressure pressure[_PRESSURE_RESOURCE_MAX]; NFTSetContext nft_set_context; @@ -353,11 +356,29 @@ void cgroup_context_free_io_device_latency(CGroupContext *c, CGroupIODeviceLaten void cgroup_context_remove_bpf_foreign_program(CGroupContext *c, CGroupBPFForeignProgram *p); void cgroup_context_remove_socket_bind(CGroupSocketBindItem **head); -static inline bool cgroup_context_want_memory_pressure(const CGroupContext *c) { +static inline bool cgroup_context_want_pressure(const CGroupContext *c, PressureResource t) { assert(c); + assert(t >= 0 && t < _PRESSURE_RESOURCE_MAX); - return c->memory_pressure_watch == CGROUP_PRESSURE_WATCH_YES || - (c->memory_pressure_watch == CGROUP_PRESSURE_WATCH_AUTO && c->memory_accounting); + if (c->pressure[t].watch == CGROUP_PRESSURE_WATCH_YES) + return true; + + if (c->pressure[t].watch != CGROUP_PRESSURE_WATCH_AUTO) + return false; + + switch (t) { + + case PRESSURE_MEMORY: + return c->memory_accounting; + + case PRESSURE_CPU: + return c->cpu_weight != CGROUP_WEIGHT_INVALID || + c->startup_cpu_weight != CGROUP_WEIGHT_INVALID || + c->cpu_quota_per_sec_usec != USEC_INFINITY; + + default: + assert_not_reached(); + } } static inline bool cgroup_context_has_device_policy(const CGroupContext *c) { diff --git a/src/core/dbus-cgroup.c b/src/core/dbus-cgroup.c index a7508c96da..c5a3302e08 100644 --- a/src/core/dbus-cgroup.c +++ b/src/core/dbus-cgroup.c @@ -427,8 +427,10 @@ const sd_bus_vtable bus_cgroup_vtable[] = { SD_BUS_PROPERTY("SocketBindDeny", "a(iiqq)", property_get_socket_bind, offsetof(CGroupContext, socket_bind_deny), 0), SD_BUS_PROPERTY("RestrictNetworkInterfaces", "(bas)", property_get_restrict_network_interfaces, 0, 0), SD_BUS_PROPERTY("BindNetworkInterface", "s", NULL, offsetof(CGroupContext, bind_network_interface), 0), - SD_BUS_PROPERTY("MemoryPressureWatch", "s", bus_property_get_cgroup_pressure_watch, offsetof(CGroupContext, memory_pressure_watch), 0), - SD_BUS_PROPERTY("MemoryPressureThresholdUSec", "t", bus_property_get_usec, offsetof(CGroupContext, memory_pressure_threshold_usec), 0), + SD_BUS_PROPERTY("MemoryPressureWatch", "s", bus_property_get_cgroup_pressure_watch, offsetof(CGroupContext, pressure[PRESSURE_MEMORY].watch), 0), + SD_BUS_PROPERTY("MemoryPressureThresholdUSec", "t", bus_property_get_usec, offsetof(CGroupContext, pressure[PRESSURE_MEMORY].threshold_usec), 0), + SD_BUS_PROPERTY("CPUPressureWatch", "s", bus_property_get_cgroup_pressure_watch, offsetof(CGroupContext, pressure[PRESSURE_CPU].watch), 0), + SD_BUS_PROPERTY("CPUPressureThresholdUSec", "t", bus_property_get_usec, offsetof(CGroupContext, pressure[PRESSURE_CPU].threshold_usec), 0), SD_BUS_PROPERTY("NFTSet", "a(iiss)", property_get_cgroup_nft_set, 0, 0), SD_BUS_PROPERTY("CoredumpReceive", "b", bus_property_get_bool, offsetof(CGroupContext, coredump_receive), 0), @@ -712,10 +714,12 @@ static int bus_cgroup_set_transient_property( return 1; - } else if (streq(name, "MemoryPressureWatch")) { + } else if (STR_IN_SET(name, "MemoryPressureWatch", "CPUPressureWatch")) { CGroupPressureWatch p; const char *t; + PressureResource pt = streq(name, "MemoryPressureWatch") ? PRESSURE_MEMORY : PRESSURE_CPU; + r = sd_bus_message_read(message, "s", &t); if (r < 0) return r; @@ -729,26 +733,28 @@ static int bus_cgroup_set_transient_property( } if (!UNIT_WRITE_FLAGS_NOOP(flags)) { - c->memory_pressure_watch = p; - unit_write_settingf(u, flags, name, "MemoryPressureWatch=%s", strempty(cgroup_pressure_watch_to_string(p))); + c->pressure[pt].watch = p; + unit_write_settingf(u, flags, name, "%s=%s", name, strempty(cgroup_pressure_watch_to_string(p))); } return 1; - } else if (streq(name, "MemoryPressureThresholdUSec")) { + } else if (STR_IN_SET(name, "MemoryPressureThresholdUSec", "CPUPressureThresholdUSec")) { uint64_t t; + PressureResource pt = streq(name, "MemoryPressureThresholdUSec") ? PRESSURE_MEMORY : PRESSURE_CPU; + r = sd_bus_message_read(message, "t", &t); if (r < 0) return r; if (!UNIT_WRITE_FLAGS_NOOP(flags)) { - c->memory_pressure_threshold_usec = t; + c->pressure[pt].threshold_usec = t; if (t == UINT64_MAX) - unit_write_setting(u, flags, name, "MemoryPressureThresholdUSec="); + unit_write_settingf(u, flags, name, "%s=", name); else - unit_write_settingf(u, flags, name, "MemoryPressureThresholdUSec=%" PRIu64, t); + unit_write_settingf(u, flags, name, "%s=%" PRIu64, name, t); } return 1; diff --git a/src/core/dbus-manager.c b/src/core/dbus-manager.c index 749e2261af..23f4c4c3de 100644 --- a/src/core/dbus-manager.c +++ b/src/core/dbus-manager.c @@ -2980,8 +2980,10 @@ const sd_bus_vtable bus_manager_vtable[] = { SD_BUS_PROPERTY("DefaultLimitRTTIME", "t", bus_property_get_rlimit, offsetof(Manager, defaults.rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("DefaultLimitRTTIMESoft", "t", bus_property_get_rlimit, offsetof(Manager, defaults.rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("DefaultTasksMax", "t", bus_property_get_tasks_max, offsetof(Manager, defaults.tasks_max), 0), - SD_BUS_PROPERTY("DefaultMemoryPressureThresholdUSec", "t", bus_property_get_usec, offsetof(Manager, defaults.memory_pressure_threshold_usec), 0), - SD_BUS_PROPERTY("DefaultMemoryPressureWatch", "s", bus_property_get_cgroup_pressure_watch, offsetof(Manager, defaults.memory_pressure_watch), 0), + SD_BUS_PROPERTY("DefaultMemoryPressureThresholdUSec", "t", bus_property_get_usec, offsetof(Manager, defaults.pressure[PRESSURE_MEMORY].threshold_usec), 0), + SD_BUS_PROPERTY("DefaultMemoryPressureWatch", "s", bus_property_get_cgroup_pressure_watch, offsetof(Manager, defaults.pressure[PRESSURE_MEMORY].watch), 0), + SD_BUS_PROPERTY("DefaultCPUPressureThresholdUSec", "t", bus_property_get_usec, offsetof(Manager, defaults.pressure[PRESSURE_CPU].threshold_usec), 0), + SD_BUS_PROPERTY("DefaultCPUPressureWatch", "s", bus_property_get_cgroup_pressure_watch, offsetof(Manager, defaults.pressure[PRESSURE_CPU].watch), 0), SD_BUS_PROPERTY("TimerSlackNSec", "t", property_get_timer_slack_nsec, 0, SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("DefaultOOMPolicy", "s", bus_property_get_oom_policy, offsetof(Manager, defaults.oom_policy), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("DefaultOOMScoreAdjust", "i", property_get_oom_score_adjust, 0, SD_BUS_VTABLE_PROPERTY_CONST), diff --git a/src/core/exec-invoke.c b/src/core/exec-invoke.c index 7500888c41..1361038af8 100644 --- a/src/core/exec-invoke.c +++ b/src/core/exec-invoke.c @@ -2034,7 +2034,7 @@ static int build_environment( const char *shell, dev_t journal_stream_dev, ino_t journal_stream_ino, - const char *memory_pressure_path, + char *const *pressure_path, bool needs_sandboxing, char ***ret) { @@ -2215,25 +2215,38 @@ static int build_environment( if (r < 0) return r; - if (memory_pressure_path) { - r = strv_extend_joined_with_size(&e, &n, "MEMORY_PRESSURE_WATCH=", memory_pressure_path); + for (PressureResource t = 0; t < _PRESSURE_RESOURCE_MAX; t++) { + if (!pressure_path[t]) + continue; + + const PressureResourceInfo *info = pressure_resource_get_info(t); + + _cleanup_free_ char *env_watch = strjoin(info->env_watch, "="); + if (!env_watch) + return -ENOMEM; + + r = strv_extend_joined_with_size(&e, &n, env_watch, pressure_path[t]); if (r < 0) return r; - if (!path_equal(memory_pressure_path, "/dev/null")) { + if (!path_equal(pressure_path[t], "/dev/null")) { _cleanup_free_ char *b = NULL, *x = NULL; if (asprintf(&b, "%s " USEC_FMT " " USEC_FMT, PRESSURE_DEFAULT_TYPE, - cgroup_context->memory_pressure_threshold_usec == USEC_INFINITY ? PRESSURE_DEFAULT_THRESHOLD_USEC : - CLAMP(cgroup_context->memory_pressure_threshold_usec, 1U, PRESSURE_DEFAULT_WINDOW_USEC), + cgroup_context->pressure[t].threshold_usec == USEC_INFINITY ? PRESSURE_DEFAULT_THRESHOLD_USEC : + CLAMP(cgroup_context->pressure[t].threshold_usec, 1U, PRESSURE_DEFAULT_WINDOW_USEC), PRESSURE_DEFAULT_WINDOW_USEC) < 0) return -ENOMEM; if (base64mem(b, strlen(b) + 1, &x) < 0) return -ENOMEM; - r = strv_extend_joined_with_size(&e, &n, "MEMORY_PRESSURE_WRITE=", x); + _cleanup_free_ char *env_write = strjoin(info->env_write, "="); + if (!env_write) + return -ENOMEM; + + r = strv_extend_joined_with_size(&e, &n, env_write, x); if (r < 0) return r; } @@ -3855,7 +3868,7 @@ static int apply_mount_namespace( const ExecParameters *params, const ExecRuntime *runtime, const PinnedResource *rootfs, - const char *memory_pressure_path, + char *const *pressure_path, bool needs_sandboxing, uid_t exec_directory_uid, gid_t exec_directory_gid, @@ -3887,16 +3900,28 @@ static int apply_mount_namespace( if (r < 0) return r; - /* We need to make the pressure path writable even if /sys/fs/cgroups is made read-only, as the - * service will need to write to it in order to start the notifications. */ - if (exec_is_cgroup_mount_read_only(context) && memory_pressure_path && !streq(memory_pressure_path, "/dev/null")) { + /* We need to make the pressure paths writable even if /sys/fs/cgroups is made read-only, as the + * service will need to write to them in order to start the notifications. */ + bool need_pressure_rw = false; + for (PressureResource t = 0; t < _PRESSURE_RESOURCE_MAX; t++) + if (pressure_path[t] && !streq(pressure_path[t], "/dev/null")) { + need_pressure_rw = true; + break; + } + + if (exec_is_cgroup_mount_read_only(context) && need_pressure_rw) { read_write_paths_cleanup = strv_copy(context->read_write_paths); if (!read_write_paths_cleanup) return -ENOMEM; - r = strv_extend(&read_write_paths_cleanup, memory_pressure_path); - if (r < 0) - return r; + for (PressureResource t = 0; t < _PRESSURE_RESOURCE_MAX; t++) { + if (!pressure_path[t] || streq(pressure_path[t], "/dev/null")) + continue; + + r = strv_extend(&read_write_paths_cleanup, pressure_path[t]); + if (r < 0) + return r; + } read_write_paths = read_write_paths_cleanup; } else @@ -4689,7 +4714,7 @@ static int setup_delegated_namespaces( const ExecRuntime *runtime, const PinnedResource *rootfs, bool delegate, - const char *memory_pressure_path, + char *const *pressure_path, uid_t uid, gid_t gid, const ExecCommand *command, @@ -4820,7 +4845,7 @@ static int setup_delegated_namespaces( params, runtime, rootfs, - memory_pressure_path, + pressure_path, needs_sandboxing, uid, gid, @@ -5146,6 +5171,10 @@ static int setup_term_environment(const ExecContext *context, char ***env) { return strv_env_replace_strdup(env, "TERM=" FALLBACK_TERM); } +static inline void free_pressure_paths(char *(*p)[_PRESSURE_RESOURCE_MAX]) { + free_many_charp(*p, _PRESSURE_RESOURCE_MAX); +} + int exec_invoke( const ExecCommand *command, const ExecContext *context, @@ -5157,7 +5186,8 @@ int exec_invoke( _cleanup_strv_free_ char **our_env = NULL, **pass_env = NULL, **joined_exec_search_path = NULL, **accum_env = NULL; int r; const char *username = NULL, *groupname = NULL; - _cleanup_free_ char *home_buffer = NULL, *memory_pressure_path = NULL, *own_user = NULL; + _cleanup_free_ char *home_buffer = NULL, *own_user = NULL; + _cleanup_(free_pressure_paths) char *pressure_path[_PRESSURE_RESOURCE_MAX] = {}; const char *pwent_home = NULL, *shell = NULL; dev_t journal_stream_dev = 0; ino_t journal_stream_ino = 0; @@ -5753,36 +5783,44 @@ int exec_invoke( } if (is_pressure_supported() > 0) { - if (cgroup_context_want_memory_pressure(cgroup_context)) { - r = cg_get_path(params->cgroup_path, "memory.pressure", &memory_pressure_path); - if (r < 0) { - *exit_status = EXIT_MEMORY; - return log_oom(); - } + for (PressureResource t = 0; t < _PRESSURE_RESOURCE_MAX; t++) { + if (cgroup_context_want_pressure(cgroup_context, t)) { + _cleanup_free_ char *pressure_file = strjoin(pressure_resource_to_string(t), ".pressure"); + if (!pressure_file) { + *exit_status = EXIT_MEMORY; + return log_oom(); + } - r = chmod_and_chown(memory_pressure_path, 0644, uid, gid); - if (r < 0) { - log_full_errno(r == -ENOENT || ERRNO_IS_PRIVILEGE(r) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to adjust ownership of '%s', ignoring: %m", memory_pressure_path); - memory_pressure_path = mfree(memory_pressure_path); - } - /* First we use the current cgroup path to chmod and chown the memory pressure path, then pass the path relative - * to the cgroup namespace to environment variables and mounts. If chown/chmod fails, we should not pass memory - * pressure path environment variable or read-write mount to the unit. This is why we check if - * memory_pressure_path != NULL in the conditional below. */ - if (memory_pressure_path && needs_sandboxing && exec_needs_cgroup_namespace(context)) { - memory_pressure_path = mfree(memory_pressure_path); - r = cg_get_path("/", "memory.pressure", &memory_pressure_path); + r = cg_get_path(params->cgroup_path, pressure_file, &pressure_path[t]); if (r < 0) { *exit_status = EXIT_MEMORY; return log_oom(); } - } - } else if (cgroup_context->memory_pressure_watch == CGROUP_PRESSURE_WATCH_NO) { - memory_pressure_path = strdup("/dev/null"); /* /dev/null is explicit indicator for turning of memory pressure watch */ - if (!memory_pressure_path) { - *exit_status = EXIT_MEMORY; - return log_oom(); + + r = chmod_and_chown(pressure_path[t], 0644, uid, gid); + if (r < 0) { + log_full_errno(r == -ENOENT || ERRNO_IS_PRIVILEGE(r) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to adjust ownership of '%s', ignoring: %m", pressure_path[t]); + pressure_path[t] = mfree(pressure_path[t]); + } + /* First we use the current cgroup path to chmod and chown the pressure path, then pass the + * path relative to the cgroup namespace to environment variables and mounts. If chown/chmod + * fails, we should not pass pressure path environment variable or read-write mount to the + * unit. This is why we check if pressure_path[t] != NULL in the conditional below. */ + if (pressure_path[t] && needs_sandboxing && exec_needs_cgroup_namespace(context)) { + pressure_path[t] = mfree(pressure_path[t]); + r = cg_get_path("/", pressure_file, &pressure_path[t]); + if (r < 0) { + *exit_status = EXIT_MEMORY; + return log_oom(); + } + } + } else if (cgroup_context->pressure[t].watch == CGROUP_PRESSURE_WATCH_NO) { + pressure_path[t] = strdup("/dev/null"); /* /dev/null is explicit indicator for turning off pressure watch */ + if (!pressure_path[t]) { + *exit_status = EXIT_MEMORY; + return log_oom(); + } } } } @@ -5829,7 +5867,7 @@ int exec_invoke( shell, journal_stream_dev, journal_stream_ino, - memory_pressure_path, + pressure_path, needs_sandboxing, &our_env); if (r < 0) { @@ -6047,7 +6085,7 @@ int exec_invoke( runtime, &rootfs, /* delegate= */ false, - memory_pressure_path, + pressure_path, uid, gid, command, @@ -6144,7 +6182,7 @@ int exec_invoke( runtime, &rootfs, /* delegate= */ true, - memory_pressure_path, + pressure_path, uid, gid, command, diff --git a/src/core/execute-serialize.c b/src/core/execute-serialize.c index c8f802687e..d3d23500a9 100644 --- a/src/core/execute-serialize.c +++ b/src/core/execute-serialize.c @@ -279,7 +279,11 @@ static int exec_cgroup_context_serialize(const CGroupContext *c, FILE *f) { if (r < 0) return r; - r = serialize_item(f, "exec-cgroup-context-memory-pressure-watch", cgroup_pressure_watch_to_string(c->memory_pressure_watch)); + r = serialize_item(f, "exec-cgroup-context-memory-pressure-watch", cgroup_pressure_watch_to_string(c->pressure[PRESSURE_MEMORY].watch)); + if (r < 0) + return r; + + r = serialize_item(f, "exec-cgroup-context-cpu-pressure-watch", cgroup_pressure_watch_to_string(c->pressure[PRESSURE_CPU].watch)); if (r < 0) return r; @@ -287,8 +291,14 @@ static int exec_cgroup_context_serialize(const CGroupContext *c, FILE *f) { if (r < 0) return r; - if (c->memory_pressure_threshold_usec != USEC_INFINITY) { - r = serialize_usec(f, "exec-cgroup-context-memory-pressure-threshold-usec", c->memory_pressure_threshold_usec); + if (c->pressure[PRESSURE_MEMORY].threshold_usec != USEC_INFINITY) { + r = serialize_usec(f, "exec-cgroup-context-memory-pressure-threshold-usec", c->pressure[PRESSURE_MEMORY].threshold_usec); + if (r < 0) + return r; + } + + if (c->pressure[PRESSURE_CPU].threshold_usec != USEC_INFINITY) { + r = serialize_usec(f, "exec-cgroup-context-cpu-pressure-threshold-usec", c->pressure[PRESSURE_CPU].threshold_usec); if (r < 0) return r; } @@ -621,15 +631,23 @@ static int exec_cgroup_context_deserialize(CGroupContext *c, FILE *f) { if (r < 0) return r; } else if ((val = startswith(l, "exec-cgroup-context-memory-pressure-watch="))) { - c->memory_pressure_watch = cgroup_pressure_watch_from_string(val); - if (c->memory_pressure_watch < 0) + c->pressure[PRESSURE_MEMORY].watch = cgroup_pressure_watch_from_string(val); + if (c->pressure[PRESSURE_MEMORY].watch < 0) + return -EINVAL; + } else if ((val = startswith(l, "exec-cgroup-context-cpu-pressure-watch="))) { + c->pressure[PRESSURE_CPU].watch = cgroup_pressure_watch_from_string(val); + if (c->pressure[PRESSURE_CPU].watch < 0) return -EINVAL; } else if ((val = startswith(l, "exec-cgroup-context-delegate-subgroup="))) { r = free_and_strdup(&c->delegate_subgroup, val); if (r < 0) return r; } else if ((val = startswith(l, "exec-cgroup-context-memory-pressure-threshold-usec="))) { - r = deserialize_usec(val, &c->memory_pressure_threshold_usec); + r = deserialize_usec(val, &c->pressure[PRESSURE_MEMORY].threshold_usec); + if (r < 0) + return r; + } else if ((val = startswith(l, "exec-cgroup-context-cpu-pressure-threshold-usec="))) { + r = deserialize_usec(val, &c->pressure[PRESSURE_CPU].threshold_usec); if (r < 0) return r; } else if ((val = startswith(l, "exec-cgroup-context-device-allow="))) { diff --git a/src/core/load-fragment-gperf.gperf.in b/src/core/load-fragment-gperf.gperf.in index bf808d220b..297836def1 100644 --- a/src/core/load-fragment-gperf.gperf.in +++ b/src/core/load-fragment-gperf.gperf.in @@ -276,8 +276,10 @@ {{type}}.SocketBindAllow, config_parse_cgroup_socket_bind, 0, offsetof({{type}}, cgroup_context.socket_bind_allow) {{type}}.SocketBindDeny, config_parse_cgroup_socket_bind, 0, offsetof({{type}}, cgroup_context.socket_bind_deny) {{type}}.RestrictNetworkInterfaces, config_parse_restrict_network_interfaces, 0, offsetof({{type}}, cgroup_context) -{{type}}.MemoryPressureThresholdSec, config_parse_sec, 0, offsetof({{type}}, cgroup_context.memory_pressure_threshold_usec) -{{type}}.MemoryPressureWatch, config_parse_memory_pressure_watch, 0, offsetof({{type}}, cgroup_context.memory_pressure_watch) +{{type}}.MemoryPressureThresholdSec, config_parse_sec, 0, offsetof({{type}}, cgroup_context.pressure[PRESSURE_MEMORY].threshold_usec) +{{type}}.MemoryPressureWatch, config_parse_pressure_watch, 0, offsetof({{type}}, cgroup_context.pressure[PRESSURE_MEMORY].watch) +{{type}}.CPUPressureThresholdSec, config_parse_sec, 0, offsetof({{type}}, cgroup_context.pressure[PRESSURE_CPU].threshold_usec) +{{type}}.CPUPressureWatch, config_parse_pressure_watch, 0, offsetof({{type}}, cgroup_context.pressure[PRESSURE_CPU].watch) {{type}}.NFTSet, config_parse_cgroup_nft_set, NFT_SET_PARSE_CGROUP, offsetof({{type}}, cgroup_context) {{type}}.CoredumpReceive, config_parse_bool, 0, offsetof({{type}}, cgroup_context.coredump_receive) {{type}}.BindNetworkInterface, config_parse_bind_network_interface, 0, offsetof({{type}}, cgroup_context) diff --git a/src/core/load-fragment.c b/src/core/load-fragment.c index 840804fcf8..16f3afafe9 100644 --- a/src/core/load-fragment.c +++ b/src/core/load-fragment.c @@ -154,7 +154,7 @@ DEFINE_CONFIG_PARSE_ENUM(config_parse_service_timeout_failure_mode, service_time DEFINE_CONFIG_PARSE_ENUM(config_parse_socket_bind, socket_address_bind_ipv6_only_or_bool, SocketAddressBindIPv6Only); DEFINE_CONFIG_PARSE_ENUM(config_parse_oom_policy, oom_policy, OOMPolicy); DEFINE_CONFIG_PARSE_ENUM(config_parse_managed_oom_preference, managed_oom_preference, ManagedOOMPreference); -DEFINE_CONFIG_PARSE_ENUM(config_parse_memory_pressure_watch, cgroup_pressure_watch, CGroupPressureWatch); +DEFINE_CONFIG_PARSE_ENUM(config_parse_pressure_watch, cgroup_pressure_watch, CGroupPressureWatch); DEFINE_CONFIG_PARSE_ENUM_WITH_DEFAULT(config_parse_ip_tos, ip_tos, int, -1); DEFINE_CONFIG_PARSE_PTR(config_parse_cg_weight, cg_weight_parse, uint64_t); DEFINE_CONFIG_PARSE_PTR(config_parse_cg_cpu_weight, cg_cpu_weight_parse, uint64_t); diff --git a/src/core/load-fragment.h b/src/core/load-fragment.h index 4677564904..a5b7595dbf 100644 --- a/src/core/load-fragment.h +++ b/src/core/load-fragment.h @@ -164,7 +164,7 @@ CONFIG_PARSER_PROTOTYPE(config_parse_watchdog_sec); CONFIG_PARSER_PROTOTYPE(config_parse_tty_size); CONFIG_PARSER_PROTOTYPE(config_parse_log_filter_patterns); CONFIG_PARSER_PROTOTYPE(config_parse_open_file); -CONFIG_PARSER_PROTOTYPE(config_parse_memory_pressure_watch); +CONFIG_PARSER_PROTOTYPE(config_parse_pressure_watch); CONFIG_PARSER_PROTOTYPE(config_parse_cgroup_nft_set); CONFIG_PARSER_PROTOTYPE(config_parse_mount_node); CONFIG_PARSER_PROTOTYPE(config_parse_concurrency_max); diff --git a/src/core/main.c b/src/core/main.c index d0c0023f89..7fcd0fa672 100644 --- a/src/core/main.c +++ b/src/core/main.c @@ -747,89 +747,91 @@ static int config_parse_crash_reboot( static int parse_config_file(void) { const ConfigTableItem items[] = { - { "Manager", "LogLevel", config_parse_level2, 0, NULL }, - { "Manager", "LogTarget", config_parse_target, 0, NULL }, - { "Manager", "LogColor", config_parse_color, 0, NULL }, - { "Manager", "LogLocation", config_parse_location, 0, NULL }, - { "Manager", "LogTime", config_parse_time, 0, NULL }, - { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, - { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, &arg_crash_chvt }, - { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, &arg_crash_chvt }, - { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, - { "Manager", "CrashReboot", config_parse_crash_reboot, 0, &arg_crash_action }, - { "Manager", "CrashAction", config_parse_crash_action, 0, &arg_crash_action }, - { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status }, - { "Manager", "StatusUnitFormat", config_parse_status_unit_format, 0, &arg_status_unit_format }, - { "Manager", "CPUAffinity", config_parse_cpu_set, 0, &arg_cpu_affinity }, - { "Manager", "NUMAPolicy", config_parse_numa_policy, 0, &arg_numa_policy.type }, - { "Manager", "NUMAMask", config_parse_numa_mask, 0, &arg_numa_policy.nodes }, - { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_LEGACY, NULL }, - { "Manager", "RuntimeWatchdogSec", config_parse_watchdog_sec, 0, &arg_runtime_watchdog }, - { "Manager", "RuntimeWatchdogPreSec", config_parse_watchdog_sec, 0, &arg_pretimeout_watchdog }, - { "Manager", "RebootWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, - { "Manager", "ShutdownWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, /* obsolete alias */ - { "Manager", "KExecWatchdogSec", config_parse_watchdog_sec, 0, &arg_kexec_watchdog }, - { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device }, - { "Manager", "RuntimeWatchdogPreGovernor", config_parse_string, CONFIG_PARSE_STRING_SAFE, &arg_watchdog_pretimeout_governor }, - { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set }, - { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs }, - { "Manager", "ProtectSystem", config_parse_protect_system_pid1, 0, &arg_protect_system }, + { "Manager", "LogLevel", config_parse_level2, 0, NULL }, + { "Manager", "LogTarget", config_parse_target, 0, NULL }, + { "Manager", "LogColor", config_parse_color, 0, NULL }, + { "Manager", "LogLocation", config_parse_location, 0, NULL }, + { "Manager", "LogTime", config_parse_time, 0, NULL }, + { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, + { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, &arg_crash_chvt }, + { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, &arg_crash_chvt }, + { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, + { "Manager", "CrashReboot", config_parse_crash_reboot, 0, &arg_crash_action }, + { "Manager", "CrashAction", config_parse_crash_action, 0, &arg_crash_action }, + { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status }, + { "Manager", "StatusUnitFormat", config_parse_status_unit_format, 0, &arg_status_unit_format }, + { "Manager", "CPUAffinity", config_parse_cpu_set, 0, &arg_cpu_affinity }, + { "Manager", "NUMAPolicy", config_parse_numa_policy, 0, &arg_numa_policy.type }, + { "Manager", "NUMAMask", config_parse_numa_mask, 0, &arg_numa_policy.nodes }, + { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_LEGACY, NULL }, + { "Manager", "RuntimeWatchdogSec", config_parse_watchdog_sec, 0, &arg_runtime_watchdog }, + { "Manager", "RuntimeWatchdogPreSec", config_parse_watchdog_sec, 0, &arg_pretimeout_watchdog }, + { "Manager", "RebootWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, + { "Manager", "ShutdownWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, /* obsolete alias */ + { "Manager", "KExecWatchdogSec", config_parse_watchdog_sec, 0, &arg_kexec_watchdog }, + { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device }, + { "Manager", "RuntimeWatchdogPreGovernor", config_parse_string, CONFIG_PARSE_STRING_SAFE, &arg_watchdog_pretimeout_governor }, + { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set }, + { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs }, + { "Manager", "ProtectSystem", config_parse_protect_system_pid1, 0, &arg_protect_system }, #if HAVE_SECCOMP - { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs }, + { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs }, #else - { "Manager", "SystemCallArchitectures", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, + { "Manager", "SystemCallArchitectures", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, #endif - { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec }, - { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_defaults.timer_accuracy_usec }, - { "Manager", "DefaultStandardOutput", config_parse_output_restricted, 0, &arg_defaults.std_output }, - { "Manager", "DefaultStandardError", config_parse_output_restricted, 0, &arg_defaults.std_error }, - { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_defaults.timeout_start_usec }, - { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_defaults.timeout_stop_usec }, - { "Manager", "DefaultTimeoutAbortSec", config_parse_default_timeout_abort, 0, NULL }, - { "Manager", "DefaultDeviceTimeoutSec", config_parse_sec, 0, &arg_defaults.device_timeout_usec }, - { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_defaults.restart_usec }, - { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_defaults.start_limit.interval }, /* obsolete alias */ - { "Manager", "DefaultStartLimitIntervalSec", config_parse_sec, 0, &arg_defaults.start_limit.interval }, - { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_defaults.start_limit.burst }, - { "Manager", "DefaultRestrictSUIDSGID", config_parse_bool, 0, &arg_defaults.restrict_suid_sgid }, - { "Manager", "DefaultEnvironment", config_parse_environ, arg_runtime_scope, &arg_default_environment }, - { "Manager", "ManagerEnvironment", config_parse_environ, arg_runtime_scope, &arg_manager_environment }, - { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_defaults.rlimit }, - { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_defaults.rlimit }, - { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_defaults.rlimit }, - { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_defaults.rlimit }, - { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_defaults.rlimit }, - { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_defaults.rlimit }, - { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_defaults.rlimit }, - { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_defaults.rlimit }, - { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_defaults.rlimit }, - { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_defaults.rlimit }, - { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_defaults.rlimit }, - { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_defaults.rlimit }, - { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_defaults.rlimit }, - { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_defaults.rlimit }, - { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_defaults.rlimit }, - { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_defaults.rlimit }, - { "Manager", "DefaultCPUAccounting", config_parse_warn_compat, DISABLED_LEGACY, NULL }, - { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_defaults.io_accounting }, - { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_defaults.ip_accounting }, - { "Manager", "DefaultBlockIOAccounting", config_parse_warn_compat, DISABLED_LEGACY, NULL }, - { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_defaults.memory_accounting }, - { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_defaults.tasks_accounting }, - { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_defaults.tasks_max }, - { "Manager", "DefaultMemoryPressureThresholdSec", config_parse_sec, 0, &arg_defaults.memory_pressure_threshold_usec }, - { "Manager", "DefaultMemoryPressureWatch", config_parse_memory_pressure_watch, 0, &arg_defaults.memory_pressure_watch }, - { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, arg_runtime_scope, &arg_cad_burst_action }, - { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_defaults.oom_policy }, - { "Manager", "DefaultOOMScoreAdjust", config_parse_oom_score_adjust, 0, NULL }, - { "Manager", "ReloadLimitIntervalSec", config_parse_sec, 0, &arg_reload_limit_interval_sec }, - { "Manager", "ReloadLimitBurst", config_parse_unsigned, 0, &arg_reload_limit_burst }, - { "Manager", "DefaultMemoryZSwapWriteback", config_parse_bool, 0, &arg_defaults.memory_zswap_writeback }, - { "Manager", "MinimumUptimeSec", config_parse_sec, 0, &arg_minimum_uptime_usec }, + { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec }, + { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_defaults.timer_accuracy_usec }, + { "Manager", "DefaultStandardOutput", config_parse_output_restricted, 0, &arg_defaults.std_output }, + { "Manager", "DefaultStandardError", config_parse_output_restricted, 0, &arg_defaults.std_error }, + { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_defaults.timeout_start_usec }, + { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_defaults.timeout_stop_usec }, + { "Manager", "DefaultTimeoutAbortSec", config_parse_default_timeout_abort, 0, NULL }, + { "Manager", "DefaultDeviceTimeoutSec", config_parse_sec, 0, &arg_defaults.device_timeout_usec }, + { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_defaults.restart_usec }, + { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_defaults.start_limit.interval }, /* obsolete alias */ + { "Manager", "DefaultStartLimitIntervalSec", config_parse_sec, 0, &arg_defaults.start_limit.interval }, + { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_defaults.start_limit.burst }, + { "Manager", "DefaultRestrictSUIDSGID", config_parse_bool, 0, &arg_defaults.restrict_suid_sgid }, + { "Manager", "DefaultEnvironment", config_parse_environ, arg_runtime_scope, &arg_default_environment }, + { "Manager", "ManagerEnvironment", config_parse_environ, arg_runtime_scope, &arg_manager_environment }, + { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_defaults.rlimit }, + { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_defaults.rlimit }, + { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_defaults.rlimit }, + { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_defaults.rlimit }, + { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_defaults.rlimit }, + { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_defaults.rlimit }, + { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_defaults.rlimit }, + { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_defaults.rlimit }, + { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_defaults.rlimit }, + { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_defaults.rlimit }, + { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_defaults.rlimit }, + { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_defaults.rlimit }, + { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_defaults.rlimit }, + { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_defaults.rlimit }, + { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_defaults.rlimit }, + { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_defaults.rlimit }, + { "Manager", "DefaultCPUAccounting", config_parse_warn_compat, DISABLED_LEGACY, NULL }, + { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_defaults.io_accounting }, + { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_defaults.ip_accounting }, + { "Manager", "DefaultBlockIOAccounting", config_parse_warn_compat, DISABLED_LEGACY, NULL }, + { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_defaults.memory_accounting }, + { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_defaults.tasks_accounting }, + { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_defaults.tasks_max }, + { "Manager", "DefaultMemoryPressureThresholdSec", config_parse_sec, 0, &arg_defaults.pressure[PRESSURE_MEMORY].threshold_usec }, + { "Manager", "DefaultMemoryPressureWatch", config_parse_pressure_watch, 0, &arg_defaults.pressure[PRESSURE_MEMORY].watch }, + { "Manager", "DefaultCPUPressureThresholdSec", config_parse_sec, 0, &arg_defaults.pressure[PRESSURE_CPU].threshold_usec }, + { "Manager", "DefaultCPUPressureWatch", config_parse_pressure_watch, 0, &arg_defaults.pressure[PRESSURE_CPU].watch }, + { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, arg_runtime_scope, &arg_cad_burst_action }, + { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_defaults.oom_policy }, + { "Manager", "DefaultOOMScoreAdjust", config_parse_oom_score_adjust, 0, NULL }, + { "Manager", "ReloadLimitIntervalSec", config_parse_sec, 0, &arg_reload_limit_interval_sec }, + { "Manager", "ReloadLimitBurst", config_parse_unsigned, 0, &arg_reload_limit_burst }, + { "Manager", "DefaultMemoryZSwapWriteback", config_parse_bool, 0, &arg_defaults.memory_zswap_writeback }, + { "Manager", "MinimumUptimeSec", config_parse_sec, 0, &arg_minimum_uptime_usec }, #if ENABLE_SMACK - { "Manager", "DefaultSmackProcessLabel", config_parse_string, 0, &arg_defaults.smack_process_label }, + { "Manager", "DefaultSmackProcessLabel", config_parse_string, 0, &arg_defaults.smack_process_label }, #else - { "Manager", "DefaultSmackProcessLabel", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, + { "Manager", "DefaultSmackProcessLabel", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, #endif {} }; diff --git a/src/core/manager.c b/src/core/manager.c index 85b68b86d2..c71d1a5d69 100644 --- a/src/core/manager.c +++ b/src/core/manager.c @@ -616,6 +616,8 @@ static char** sanitize_environment(char **l) { l, "CACHE_DIRECTORY", "CONFIGURATION_DIRECTORY", + "CPU_PRESSURE_WATCH", + "CPU_PRESSURE_WRITE", "CREDENTIALS_DIRECTORY", "EXIT_CODE", "EXIT_STATUS", @@ -796,26 +798,37 @@ static int manager_setup_sigchld_event_source(Manager *m) { return 0; } -int manager_setup_memory_pressure_event_source(Manager *m) { +typedef int (*pressure_add_t)(sd_event *, sd_event_source **, sd_event_handler_t, void *); +typedef int (*pressure_set_period_t)(sd_event_source *, usec_t, usec_t); + +static const struct { + pressure_add_t add; + pressure_set_period_t set_period; +} pressure_dispatch_table[_PRESSURE_RESOURCE_MAX] = { + [PRESSURE_MEMORY] = { sd_event_add_memory_pressure, sd_event_source_set_memory_pressure_period }, + [PRESSURE_CPU] = { sd_event_add_cpu_pressure, sd_event_source_set_cpu_pressure_period }, +}; + +int manager_setup_pressure_event_source(Manager *m, PressureResource t) { int r; assert(m); + assert(t >= 0 && t < _PRESSURE_RESOURCE_MAX); - m->memory_pressure_event_source = sd_event_source_disable_unref(m->memory_pressure_event_source); + m->pressure_event_source[t] = sd_event_source_disable_unref(m->pressure_event_source[t]); - r = sd_event_add_memory_pressure(m->event, &m->memory_pressure_event_source, NULL, NULL); + r = pressure_dispatch_table[t].add(m->event, &m->pressure_event_source[t], NULL, NULL); if (r < 0) log_full_errno(ERRNO_IS_NOT_SUPPORTED(r) || ERRNO_IS_PRIVILEGE(r) || (r == -EHOSTDOWN) ? LOG_DEBUG : LOG_NOTICE, r, - "Failed to establish memory pressure event source, ignoring: %m"); - else if (m->defaults.memory_pressure_threshold_usec != USEC_INFINITY) { + "Failed to establish %s pressure event source, ignoring: %m", pressure_resource_to_string(t)); + else if (m->defaults.pressure[t].threshold_usec != USEC_INFINITY) { - /* If there's a default memory pressure threshold set, also apply it to the service manager itself */ - r = sd_event_source_set_memory_pressure_period( - m->memory_pressure_event_source, - m->defaults.memory_pressure_threshold_usec, - MEMORY_PRESSURE_DEFAULT_WINDOW_USEC); + r = pressure_dispatch_table[t].set_period( + m->pressure_event_source[t], + m->defaults.pressure[t].threshold_usec, + PRESSURE_DEFAULT_WINDOW_USEC); if (r < 0) - log_warning_errno(r, "Failed to adjust memory pressure threshold, ignoring: %m"); + log_warning_errno(r, "Failed to adjust %s pressure threshold, ignoring: %m", pressure_resource_to_string(t)); } return 0; @@ -1001,9 +1014,11 @@ int manager_new(RuntimeScope runtime_scope, ManagerTestRunFlags test_run_flags, if (r < 0) return r; - r = manager_setup_memory_pressure_event_source(m); - if (r < 0) - return r; + for (PressureResource t = 0; t < _PRESSURE_RESOURCE_MAX; t++) { + r = manager_setup_pressure_event_source(m, t); + if (r < 0) + return r; + } #if HAVE_LIBBPF if (MANAGER_IS_SYSTEM(m) && bpf_restrict_fs_supported(/* initialize= */ true)) { @@ -1711,7 +1726,8 @@ Manager* manager_free(Manager *m) { sd_event_source_unref(m->user_lookup_event_source); sd_event_source_unref(m->handoff_timestamp_event_source); sd_event_source_unref(m->pidref_event_source); - sd_event_source_unref(m->memory_pressure_event_source); + FOREACH_ARRAY(pressure_event_source, m->pressure_event_source, _PRESSURE_RESOURCE_MAX) + sd_event_source_unref(*pressure_event_source); safe_close(m->signal_fd); safe_close(m->notify_fd); @@ -4300,8 +4316,7 @@ int manager_set_unit_defaults(Manager *m, const UnitDefaults *defaults) { m->defaults.oom_score_adjust = defaults->oom_score_adjust; m->defaults.oom_score_adjust_set = defaults->oom_score_adjust_set; - m->defaults.memory_pressure_watch = defaults->memory_pressure_watch; - m->defaults.memory_pressure_threshold_usec = defaults->memory_pressure_threshold_usec; + memcpy(m->defaults.pressure, defaults->pressure, sizeof(m->defaults.pressure)); m->defaults.memory_zswap_writeback = defaults->memory_zswap_writeback; @@ -5195,8 +5210,10 @@ void unit_defaults_init(UnitDefaults *defaults, RuntimeScope scope) { .tasks_max = DEFAULT_TASKS_MAX, .timer_accuracy_usec = 1 * USEC_PER_MINUTE, - .memory_pressure_watch = CGROUP_PRESSURE_WATCH_AUTO, - .memory_pressure_threshold_usec = MEMORY_PRESSURE_DEFAULT_THRESHOLD_USEC, + .pressure = { + [PRESSURE_MEMORY] = { .watch = CGROUP_PRESSURE_WATCH_AUTO, .threshold_usec = PRESSURE_DEFAULT_THRESHOLD_USEC }, + [PRESSURE_CPU] = { .watch = CGROUP_PRESSURE_WATCH_AUTO, .threshold_usec = PRESSURE_DEFAULT_THRESHOLD_USEC }, + }, .oom_policy = OOM_STOP, .oom_score_adjust_set = false, diff --git a/src/core/manager.h b/src/core/manager.h index 1c04deabee..7d58c330a1 100644 --- a/src/core/manager.h +++ b/src/core/manager.h @@ -149,8 +149,7 @@ typedef struct UnitDefaults { bool memory_zswap_writeback; - CGroupPressureWatch memory_pressure_watch; - usec_t memory_pressure_threshold_usec; + CGroupPressure pressure[_PRESSURE_RESOURCE_MAX]; char *smack_process_label; @@ -481,7 +480,7 @@ typedef struct Manager { /* Dump*() are slow, so always rate limit them to 10 per 10 minutes */ RateLimit dump_ratelimit; - sd_event_source *memory_pressure_event_source; + sd_event_source *pressure_event_source[_PRESSURE_RESOURCE_MAX]; /* For NFTSet= */ sd_netlink *nfnl; @@ -562,7 +561,7 @@ void manager_unwatch_pidref(Manager *m, const PidRef *pid); unsigned manager_dispatch_load_queue(Manager *m); -int manager_setup_memory_pressure_event_source(Manager *m); +int manager_setup_pressure_event_source(Manager *m, PressureResource t); int manager_default_environment(Manager *m); int manager_transient_environment_add(Manager *m, char **plus); diff --git a/src/core/system.conf.in b/src/core/system.conf.in index ef2c59bd79..d3cb0160a0 100644 --- a/src/core/system.conf.in +++ b/src/core/system.conf.in @@ -78,6 +78,8 @@ #DefaultLimitRTTIME= #DefaultMemoryPressureThresholdSec=200ms #DefaultMemoryPressureWatch=auto +#DefaultCPUPressureThresholdSec=200ms +#DefaultCPUPressureWatch=auto #DefaultOOMPolicy=stop #DefaultSmackProcessLabel= #DefaultRestrictSUIDSGID= diff --git a/src/core/unit.c b/src/core/unit.c index 6dd5599f0a..dc205097e0 100644 --- a/src/core/unit.c +++ b/src/core/unit.c @@ -178,10 +178,9 @@ static void unit_init(Unit *u) { if (u->type != UNIT_SLICE) cc->tasks_max = u->manager->defaults.tasks_max; - cc->memory_pressure_watch = u->manager->defaults.memory_pressure_watch; - cc->memory_pressure_threshold_usec = u->manager->defaults.memory_pressure_threshold_usec; - cc->memory_zswap_writeback = u->manager->defaults.memory_zswap_writeback; + + memcpy(cc->pressure, u->manager->defaults.pressure, sizeof(cc->pressure)); } ec = unit_get_exec_context(u); diff --git a/src/core/user.conf.in b/src/core/user.conf.in index 9c37f4b54e..fe45c00b74 100644 --- a/src/core/user.conf.in +++ b/src/core/user.conf.in @@ -54,6 +54,8 @@ #DefaultLimitRTTIME= #DefaultMemoryPressureThresholdSec=200ms #DefaultMemoryPressureWatch=auto +#DefaultCPUPressureThresholdSec=200ms +#DefaultCPUPressureWatch=auto #DefaultSmackProcessLabel= #DefaultRestrictSUIDSGID= #ReloadLimitIntervalSec= diff --git a/src/core/varlink-cgroup.c b/src/core/varlink-cgroup.c index 65c07ecfad..d4ec6049e6 100644 --- a/src/core/varlink-cgroup.c +++ b/src/core/varlink-cgroup.c @@ -323,8 +323,10 @@ int unit_cgroup_context_build_json(sd_json_variant **ret, const char *name, void JSON_BUILD_PAIR_UNSIGNED_NON_ZERO("ManagedOOMMemoryPressureLimit", c->moom_mem_pressure_limit), JSON_BUILD_PAIR_FINITE_USEC("ManagedOOMMemoryPressureDurationUSec", c->moom_mem_pressure_duration_usec), SD_JSON_BUILD_PAIR_STRING("ManagedOOMPreference", managed_oom_preference_to_string(c->moom_preference)), - SD_JSON_BUILD_PAIR_STRING("MemoryPressureWatch", cgroup_pressure_watch_to_string(c->memory_pressure_watch)), - JSON_BUILD_PAIR_FINITE_USEC("MemoryPressureThresholdUSec", c->memory_pressure_threshold_usec), + SD_JSON_BUILD_PAIR_STRING("MemoryPressureWatch", cgroup_pressure_watch_to_string(c->pressure[PRESSURE_MEMORY].watch)), + JSON_BUILD_PAIR_FINITE_USEC("MemoryPressureThresholdUSec", c->pressure[PRESSURE_MEMORY].threshold_usec), + SD_JSON_BUILD_PAIR_STRING("CPUPressureWatch", cgroup_pressure_watch_to_string(c->pressure[PRESSURE_CPU].watch)), + JSON_BUILD_PAIR_FINITE_USEC("CPUPressureThresholdUSec", c->pressure[PRESSURE_CPU].threshold_usec), /* Others */ SD_JSON_BUILD_PAIR_BOOLEAN("CoredumpReceive", c->coredump_receive)); diff --git a/src/core/varlink-manager.c b/src/core/varlink-manager.c index c039ea8e53..3953b8619f 100644 --- a/src/core/varlink-manager.c +++ b/src/core/varlink-manager.c @@ -106,8 +106,10 @@ static int manager_context_build_json(sd_json_variant **ret, const char *name, v SD_JSON_BUILD_PAIR_BOOLEAN("DefaultTasksAccounting", m->defaults.tasks_accounting), SD_JSON_BUILD_PAIR_CALLBACK("DefaultLimits", rlimit_table_build_json, m->defaults.rlimit), SD_JSON_BUILD_PAIR_UNSIGNED("DefaultTasksMax", cgroup_tasks_max_resolve(&m->defaults.tasks_max)), - JSON_BUILD_PAIR_FINITE_USEC("DefaultMemoryPressureThresholdUSec", m->defaults.memory_pressure_threshold_usec), - SD_JSON_BUILD_PAIR_STRING("DefaultMemoryPressureWatch", cgroup_pressure_watch_to_string(m->defaults.memory_pressure_watch)), + JSON_BUILD_PAIR_FINITE_USEC("DefaultMemoryPressureThresholdUSec", m->defaults.pressure[PRESSURE_MEMORY].threshold_usec), + SD_JSON_BUILD_PAIR_STRING("DefaultMemoryPressureWatch", cgroup_pressure_watch_to_string(m->defaults.pressure[PRESSURE_MEMORY].watch)), + JSON_BUILD_PAIR_FINITE_USEC("DefaultCPUPressureThresholdUSec", m->defaults.pressure[PRESSURE_CPU].threshold_usec), + SD_JSON_BUILD_PAIR_STRING("DefaultCPUPressureWatch", cgroup_pressure_watch_to_string(m->defaults.pressure[PRESSURE_CPU].watch)), JSON_BUILD_PAIR_FINITE_USEC("RuntimeWatchdogUSec", manager_get_watchdog(m, WATCHDOG_RUNTIME)), JSON_BUILD_PAIR_FINITE_USEC("RebootWatchdogUSec", manager_get_watchdog(m, WATCHDOG_REBOOT)), JSON_BUILD_PAIR_FINITE_USEC("KExecWatchdogUSec", manager_get_watchdog(m, WATCHDOG_KEXEC)), diff --git a/src/libsystemd/sd-event/sd-event.c b/src/libsystemd/sd-event/sd-event.c index 4b539a35cf..aba6bf9b47 100644 --- a/src/libsystemd/sd-event/sd-event.c +++ b/src/libsystemd/sd-event/sd-event.c @@ -1949,8 +1949,7 @@ static int event_add_pressure( assert_return(e->state != SD_EVENT_FINISHED, -ESTALE); assert_return(!event_origin_changed(e), -ECHILD); - assert(resource >= 0 && resource < _PRESSURE_RESOURCE_MAX); - const PressureResourceInfo *info = &pressure_resource_info[resource]; + const PressureResourceInfo *info = pressure_resource_get_info(resource); if (!callback) callback = default_callback; diff --git a/src/shared/bus-unit-util.c b/src/shared/bus-unit-util.c index 84de3478a7..9c732543fa 100644 --- a/src/shared/bus-unit-util.c +++ b/src/shared/bus-unit-util.c @@ -2383,6 +2383,7 @@ static const BusProperty cgroup_properties[] = { { "ManagedOOMMemoryPressure", bus_append_string }, { "ManagedOOMPreference", bus_append_string }, { "MemoryPressureWatch", bus_append_string }, + { "CPUPressureWatch", bus_append_string }, { "DelegateSubgroup", bus_append_string }, { "ManagedOOMMemoryPressureLimit", bus_append_parse_permyriad }, { "MemoryAccounting", bus_append_parse_boolean }, @@ -2421,6 +2422,7 @@ static const BusProperty cgroup_properties[] = { { "SocketBindAllow", bus_append_socket_filter }, { "SocketBindDeny", bus_append_socket_filter }, { "MemoryPressureThresholdSec", bus_append_parse_sec_rename }, + { "CPUPressureThresholdSec", bus_append_parse_sec_rename }, { "NFTSet", bus_append_nft_set }, { "BindNetworkInterface", bus_append_string }, diff --git a/src/shared/varlink-io.systemd.Manager.c b/src/shared/varlink-io.systemd.Manager.c index ddf15b173e..f947c0a056 100644 --- a/src/shared/varlink-io.systemd.Manager.c +++ b/src/shared/varlink-io.systemd.Manager.c @@ -64,6 +64,10 @@ static SD_VARLINK_DEFINE_STRUCT_TYPE( SD_VARLINK_DEFINE_FIELD(DefaultMemoryPressureThresholdUSec, SD_VARLINK_INT, 0), SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd-system.conf.html#DefaultMemoryPressureWatch="), SD_VARLINK_DEFINE_FIELD(DefaultMemoryPressureWatch, SD_VARLINK_STRING, 0), + SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd-system.conf.html#DefaultCPUPressureThresholdUSec="), + SD_VARLINK_DEFINE_FIELD(DefaultCPUPressureThresholdUSec, SD_VARLINK_INT, 0), + SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd-system.conf.html#DefaultCPUPressureWatch="), + SD_VARLINK_DEFINE_FIELD(DefaultCPUPressureWatch, SD_VARLINK_STRING, 0), SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd-system.conf.html#RuntimeWatchdogSec="), SD_VARLINK_DEFINE_FIELD(RuntimeWatchdogUSec, SD_VARLINK_INT, SD_VARLINK_NULLABLE), SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd-system.conf.html#RebootWatchdogSec="), diff --git a/src/shared/varlink-io.systemd.Unit.c b/src/shared/varlink-io.systemd.Unit.c index 05676210ef..a230f29dab 100644 --- a/src/shared/varlink-io.systemd.Unit.c +++ b/src/shared/varlink-io.systemd.Unit.c @@ -228,6 +228,10 @@ static SD_VARLINK_DEFINE_STRUCT_TYPE( SD_VARLINK_DEFINE_FIELD(MemoryPressureWatch, SD_VARLINK_STRING, 0), SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd.resource-control.html#MemoryPressureThresholdSec="), SD_VARLINK_DEFINE_FIELD(MemoryPressureThresholdUSec, SD_VARLINK_INT, SD_VARLINK_NULLABLE), + SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd.resource-control.html#CPUPressureWatch="), + SD_VARLINK_DEFINE_FIELD(CPUPressureWatch, SD_VARLINK_STRING, 0), + SD_VARLINK_FIELD_COMMENT("https://www.freedesktop.org/software/systemd/man/"PROJECT_VERSION_STR"/systemd.resource-control.html#CPUPressureThresholdSec="), + SD_VARLINK_DEFINE_FIELD(CPUPressureThresholdUSec, SD_VARLINK_INT, SD_VARLINK_NULLABLE), /* Others */ SD_VARLINK_FIELD_COMMENT("Reflects whether to forward coredumps for processes that crash within this cgroup"), diff --git a/test/integration-tests/TEST-79-MEMPRESS/meson.build b/test/integration-tests/TEST-79-PRESSURE/meson.build similarity index 100% rename from test/integration-tests/TEST-79-MEMPRESS/meson.build rename to test/integration-tests/TEST-79-PRESSURE/meson.build diff --git a/test/integration-tests/meson.build b/test/integration-tests/meson.build index 5d71e87c79..198371ccae 100644 --- a/test/integration-tests/meson.build +++ b/test/integration-tests/meson.build @@ -91,7 +91,7 @@ foreach dirname : [ 'TEST-74-AUX-UTILS', 'TEST-75-RESOLVED', 'TEST-78-SIGQUEUE', - 'TEST-79-MEMPRESS', + 'TEST-79-PRESSURE', 'TEST-80-NOTIFYACCESS', 'TEST-81-GENERATORS', 'TEST-82-SOFTREBOOT', diff --git a/test/units/TEST-79-MEMPRESS.sh b/test/units/TEST-79-PRESSURE.sh similarity index 56% rename from test/units/TEST-79-MEMPRESS.sh rename to test/units/TEST-79-PRESSURE.sh index 0659160966..d4e4a9e06b 100755 --- a/test/units/TEST-79-MEMPRESS.sh +++ b/test/units/TEST-79-PRESSURE.sh @@ -13,7 +13,7 @@ if ! cat /proc/pressure/memory >/dev/null ; then exit 0 fi -CGROUP=/sys/fs/cgroup/"$(systemctl show TEST-79-MEMPRESS.service -P ControlGroup)" +CGROUP=/sys/fs/cgroup/"$(systemctl show TEST-79-PRESSURE.service -P ControlGroup)" test -d "$CGROUP" if ! test -f "$CGROUP"/memory.pressure ; then @@ -61,4 +61,57 @@ systemd-run \ rm "$SCRIPT" +# Now test CPU pressure + +if ! cat /proc/pressure/cpu >/dev/null ; then + echo "kernel has no CPU PSI support." >&2 + echo OK >/testok + exit 0 +fi + +if ! test -f "$CGROUP"/cpu.pressure ; then + echo "No CPU accounting/PSI delegated via cgroup, can't test." >&2 + echo OK >/testok + exit 0 +fi + +UNIT="test-cpupress-$RANDOM.service" +SCRIPT="/tmp/cpupress-$RANDOM.sh" + +cat >"$SCRIPT" <<'EOF' +#!/usr/bin/env bash + +set -ex + +export +id + +test -n "$CPU_PRESSURE_WATCH" +test "$CPU_PRESSURE_WATCH" != /dev/null +test -w "$CPU_PRESSURE_WATCH" + +ls -al "$CPU_PRESSURE_WATCH" + +EXPECTED="$(echo -n -e "some 123000 2000000\x00" | base64)" + +test "$EXPECTED" = "$CPU_PRESSURE_WRITE" + +EOF + +chmod +x "$SCRIPT" + +systemd-run \ + -u "$UNIT" \ + -p Type=exec \ + -p ProtectControlGroups=1 \ + -p DynamicUser=1 \ + -p CPUPressureWatch=on \ + -p CPUPressureThresholdSec=123ms \ + -p BindPaths=$SCRIPT \ + `# Make sanitizers happy when DynamicUser=1 pulls in instrumented systemd NSS modules` \ + -p EnvironmentFile=-/usr/lib/systemd/systemd-asan-env \ + --wait "$SCRIPT" + +rm "$SCRIPT" + touch /testok