Systemd/src/core/cgroup.h
Lennart Poettering 5af8805872 cgroup: drastically simplify caching of cgroups members mask
Previously we tried to be smart: when a new unit appeared and it only
added controllers to the cgroup mask we'd update the cached members mask
in all parents by ORing in the controller flags in their cached values.
Unfortunately this was quite broken, as we missed some conditions when
this cache had to be reset (for example, when a unit got unloaded),
moreover the optimization doesn't work when a controller is removed
anyway (as in that case there's no other way for the parent to iterate
though all children if any other, remaining child unit still needs it).
Hence, let's simplify the logic substantially: instead of updating the
cache on the right events (which we didn't get right), let's simply
invalidate the cache, and generate it lazily when we encounter it later.
This should actually result in better behaviour as we don't have to
calculate the new members mask for a whole subtree whever we have the
suspicion something changed, but can delay it to the point where we
actually need the members mask.

This allows us to simplify things quite a bit, which is good, since
validating this cache for correctness is hard enough.

Fixes: #9512
2018-11-23 13:41:37 +01:00

217 lines
6.8 KiB
C

/* SPDX-License-Identifier: LGPL-2.1+ */
#pragma once
#include <stdbool.h>
#include "cgroup-util.h"
#include "ip-address-access.h"
#include "list.h"
#include "time-util.h"
typedef struct CGroupContext CGroupContext;
typedef struct CGroupDeviceAllow CGroupDeviceAllow;
typedef struct CGroupIODeviceWeight CGroupIODeviceWeight;
typedef struct CGroupIODeviceLimit CGroupIODeviceLimit;
typedef struct CGroupIODeviceLatency CGroupIODeviceLatency;
typedef struct CGroupBlockIODeviceWeight CGroupBlockIODeviceWeight;
typedef struct CGroupBlockIODeviceBandwidth CGroupBlockIODeviceBandwidth;
typedef enum CGroupDevicePolicy {
/* When devices listed, will allow those, plus built-in ones,
if none are listed will allow everything. */
CGROUP_AUTO,
/* Everything forbidden, except built-in ones and listed ones. */
CGROUP_CLOSED,
/* Everythings forbidden, except for the listed devices */
CGROUP_STRICT,
_CGROUP_DEVICE_POLICY_MAX,
_CGROUP_DEVICE_POLICY_INVALID = -1
} CGroupDevicePolicy;
struct CGroupDeviceAllow {
LIST_FIELDS(CGroupDeviceAllow, device_allow);
char *path;
bool r:1;
bool w:1;
bool m:1;
};
struct CGroupIODeviceWeight {
LIST_FIELDS(CGroupIODeviceWeight, device_weights);
char *path;
uint64_t weight;
};
struct CGroupIODeviceLimit {
LIST_FIELDS(CGroupIODeviceLimit, device_limits);
char *path;
uint64_t limits[_CGROUP_IO_LIMIT_TYPE_MAX];
};
struct CGroupIODeviceLatency {
LIST_FIELDS(CGroupIODeviceLatency, device_latencies);
char *path;
usec_t target_usec;
};
struct CGroupBlockIODeviceWeight {
LIST_FIELDS(CGroupBlockIODeviceWeight, device_weights);
char *path;
uint64_t weight;
};
struct CGroupBlockIODeviceBandwidth {
LIST_FIELDS(CGroupBlockIODeviceBandwidth, device_bandwidths);
char *path;
uint64_t rbps;
uint64_t wbps;
};
struct CGroupContext {
bool cpu_accounting;
bool io_accounting;
bool blockio_accounting;
bool memory_accounting;
bool tasks_accounting;
bool ip_accounting;
/* For unified hierarchy */
uint64_t cpu_weight;
uint64_t startup_cpu_weight;
usec_t cpu_quota_per_sec_usec;
uint64_t io_weight;
uint64_t startup_io_weight;
LIST_HEAD(CGroupIODeviceWeight, io_device_weights);
LIST_HEAD(CGroupIODeviceLimit, io_device_limits);
LIST_HEAD(CGroupIODeviceLatency, io_device_latencies);
uint64_t memory_min;
uint64_t memory_low;
uint64_t memory_high;
uint64_t memory_max;
uint64_t memory_swap_max;
LIST_HEAD(IPAddressAccessItem, ip_address_allow);
LIST_HEAD(IPAddressAccessItem, ip_address_deny);
/* For legacy hierarchies */
uint64_t cpu_shares;
uint64_t startup_cpu_shares;
uint64_t blockio_weight;
uint64_t startup_blockio_weight;
LIST_HEAD(CGroupBlockIODeviceWeight, blockio_device_weights);
LIST_HEAD(CGroupBlockIODeviceBandwidth, blockio_device_bandwidths);
uint64_t memory_limit;
CGroupDevicePolicy device_policy;
LIST_HEAD(CGroupDeviceAllow, device_allow);
/* Common */
uint64_t tasks_max;
bool delegate;
CGroupMask delegate_controllers;
};
/* Used when querying IP accounting data */
typedef enum CGroupIPAccountingMetric {
CGROUP_IP_INGRESS_BYTES,
CGROUP_IP_INGRESS_PACKETS,
CGROUP_IP_EGRESS_BYTES,
CGROUP_IP_EGRESS_PACKETS,
_CGROUP_IP_ACCOUNTING_METRIC_MAX,
_CGROUP_IP_ACCOUNTING_METRIC_INVALID = -1,
} CGroupIPAccountingMetric;
typedef struct Unit Unit;
typedef struct Manager Manager;
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_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a);
void cgroup_context_free_io_device_weight(CGroupContext *c, CGroupIODeviceWeight *w);
void cgroup_context_free_io_device_limit(CGroupContext *c, CGroupIODeviceLimit *l);
void cgroup_context_free_io_device_latency(CGroupContext *c, CGroupIODeviceLatency *l);
void cgroup_context_free_blockio_device_weight(CGroupContext *c, CGroupBlockIODeviceWeight *w);
void cgroup_context_free_blockio_device_bandwidth(CGroupContext *c, CGroupBlockIODeviceBandwidth *b);
int cgroup_add_device_allow(CGroupContext *c, const char *dev, const char *mode);
CGroupMask unit_get_own_mask(Unit *u);
CGroupMask unit_get_delegate_mask(Unit *u);
CGroupMask unit_get_members_mask(Unit *u);
CGroupMask unit_get_siblings_mask(Unit *u);
CGroupMask unit_get_subtree_mask(Unit *u);
CGroupMask unit_get_target_mask(Unit *u);
CGroupMask unit_get_enable_mask(Unit *u);
void unit_invalidate_cgroup_members_masks(Unit *u);
void unit_add_to_cgroup_realize_queue(Unit *u);
const char *unit_get_realized_cgroup_path(Unit *u, CGroupMask mask);
char *unit_default_cgroup_path(Unit *u);
int unit_set_cgroup_path(Unit *u, const char *path);
int unit_pick_cgroup_path(Unit *u);
int unit_realize_cgroup(Unit *u);
void unit_release_cgroup(Unit *u);
void unit_prune_cgroup(Unit *u);
int unit_watch_cgroup(Unit *u);
void unit_add_to_cgroup_empty_queue(Unit *u);
int unit_attach_pids_to_cgroup(Unit *u, Set *pids, const char *suffix_path);
int manager_setup_cgroup(Manager *m);
void manager_shutdown_cgroup(Manager *m, bool delete);
unsigned manager_dispatch_cgroup_realize_queue(Manager *m);
Unit *manager_get_unit_by_cgroup(Manager *m, const char *cgroup);
Unit *manager_get_unit_by_pid_cgroup(Manager *m, pid_t pid);
Unit* manager_get_unit_by_pid(Manager *m, pid_t pid);
int unit_search_main_pid(Unit *u, pid_t *ret);
int unit_watch_all_pids(Unit *u);
int unit_synthesize_cgroup_empty_event(Unit *u);
int unit_get_memory_current(Unit *u, uint64_t *ret);
int unit_get_tasks_current(Unit *u, uint64_t *ret);
int unit_get_cpu_usage(Unit *u, nsec_t *ret);
int unit_get_ip_accounting(Unit *u, CGroupIPAccountingMetric metric, uint64_t *ret);
int unit_reset_cpu_accounting(Unit *u);
int unit_reset_ip_accounting(Unit *u);
#define UNIT_CGROUP_BOOL(u, name) \
({ \
CGroupContext *cc = unit_get_cgroup_context(u); \
cc ? cc->name : false; \
})
bool manager_owns_host_root_cgroup(Manager *m);
bool unit_has_host_root_cgroup(Unit *u);
int manager_notify_cgroup_empty(Manager *m, const char *group);
void unit_invalidate_cgroup(Unit *u, CGroupMask m);
void unit_invalidate_cgroup_bpf(Unit *u);
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_;
bool unit_cgroup_delegate(Unit *u);