src/core/cgroup.c

	cgroup.c revision e41969e3d1a8bda8b0b2f7eb012cef1c8835509a
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek/***
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  This file is part of systemd.
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  Copyright 2013 Lennart Poettering
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  systemd is free software; you can redistribute it and/or modify it
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  under the terms of the GNU Lesser General Public License as published by
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  the Free Software Foundation; either version 2.1 of the License, or
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  (at your option) any later version.
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  systemd is distributed in the hope that it will be useful, but
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  WITHOUT ANY WARRANTY; without even the implied warranty of
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  Lesser General Public License for more details.
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  You should have received a copy of the GNU Lesser General Public License
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek  along with systemd; If not, see <http://www.gnu.org/licenses/>.
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek***/
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek#include <fcntl.h>
613b411c947635136637f8cdd66b94512f761eabLennart Poettering#include <fnmatch.h>
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek#include "path-util.h"
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek#include "special.h"
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek#include "cgroup-util.h"
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek#include "cgroup.h"
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
613b411c947635136637f8cdd66b94512f761eabLennart Poetteringvoid cgroup_context_init(CGroupContext *c) {
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(c);
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        /* Initialize everything to the kernel defaults, assuming the
613b411c947635136637f8cdd66b94512f761eabLennart Poettering         * structure is preinitialized to 0 */
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek        c->cpu_shares = 1024;
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        c->memory_limit = (uint64_t) -1;
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        c->blockio_weight = 1000;
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek}
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poetteringvoid cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a) {
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek        assert(c);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(a);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        LIST_REMOVE(device_allow, c->device_allow, a);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        free(a->path);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        free(a);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering}
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poetteringvoid cgroup_context_free_blockio_device_weight(CGroupContext *c, CGroupBlockIODeviceWeight *w) {
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(c);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(w);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        LIST_REMOVE(device_weights, c->blockio_device_weights, w);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        free(w->path);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        free(w);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering}
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poetteringvoid cgroup_context_free_blockio_device_bandwidth(CGroupContext *c, CGroupBlockIODeviceBandwidth *b) {
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(c);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(b);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        LIST_REMOVE(device_bandwidths, c->blockio_device_bandwidths, b);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        free(b->path);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        free(b);
3d94f76c99da13e5603831d0b278f8c8c21bcb02Lennart Poettering}
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poetteringvoid cgroup_context_done(CGroupContext *c) {
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(c);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        while (c->blockio_device_weights)
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        while (c->blockio_device_bandwidths)
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                cgroup_context_free_blockio_device_bandwidth(c, c->blockio_device_bandwidths);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        while (c->device_allow)
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                cgroup_context_free_device_allow(c, c->device_allow);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering}
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poetteringvoid cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) {
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        CGroupBlockIODeviceBandwidth *b;
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        CGroupBlockIODeviceWeight *w;
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        CGroupDeviceAllow *a;
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(c);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        assert(f);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        prefix = strempty(prefix);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        fprintf(f,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sCPUAccounting=%s\n"
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sBlockIOAccounting=%s\n"
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sMemoryAccounting=%s\n"
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sCPUShares=%lu\n"
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sBlockIOWeight=%lu\n"
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sMemoryLimit=%" PRIu64 "\n"
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                "%sDevicePolicy=%s\n",
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, yes_no(c->cpu_accounting),
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, yes_no(c->blockio_accounting),
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, yes_no(c->memory_accounting),
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, c->cpu_shares,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, c->blockio_weight,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, c->memory_limit,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                prefix, cgroup_device_policy_to_string(c->device_policy));
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        LIST_FOREACH(device_allow, a, c->device_allow)
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                fprintf(f,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        "%sDeviceAllow=%s %s%s%s\n",
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        prefix,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        a->path,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        a->r ? "r" : "", a->w ? "w" : "", a->m ? "m" : "");
613b411c947635136637f8cdd66b94512f761eabLennart Poettering
613b411c947635136637f8cdd66b94512f761eabLennart Poettering        LIST_FOREACH(device_weights, w, c->blockio_device_weights)
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                fprintf(f,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        "%sBlockIODeviceWeight=%s %lu",
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        prefix,
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                        w->path,
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek                        w->weight);
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek        LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) {
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering                char buf[FORMAT_BYTES_MAX];
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering                fprintf(f,
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek                        "%s%s=%s %s\n",
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering                        prefix,
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering                        b->read ? "BlockIOReadBandwidth" : "BlockIOWriteBandwidth",
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering                        b->path,
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering                        format_bytes(buf, sizeof(buf), b->bandwidth));
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering        }
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering}
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poetteringstatic int lookup_blkio_device(const char *p, dev_t *dev) {
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering        struct stat st;
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering        int r;
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering        assert(p);
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering        assert(dev);
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering
6b46ea73e3b1d8a1e65f58ac04772821bd4a72fbLennart Poettering        r = stat(p, &st);
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek        if (r < 0) {
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                log_warning("Couldn't stat device %s: %m", p);
613b411c947635136637f8cdd66b94512f761eabLennart Poettering                return -errno;
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek        }
d8c9d3a468e61ee2a2b2c3454e662398b0885411Zbigniew Jędrzejewski-Szmek
        if (S_ISBLK(st.st_mode))
                *dev = st.st_rdev;
        else if (major(st.st_dev) != 0) {
                /* If this is not a device node then find the block
                 * device this file is stored on */
                *dev = st.st_dev;

                /* If this is a partition, try to get the originating
                 * block device */
                block_get_whole_disk(*dev, dev);
        } else {
                log_warning("%s is not a block device and file system block device cannot be determined or is not local.", p);
                return -ENODEV;
        }

        return 0;
}

static int whitelist_device(const char *path, const char *node, const char *acc) {
        char buf[2+DECIMAL_STR_MAX(dev_t)*2+2+4];
        struct stat st;
        int r;

        assert(path);
        assert(acc);

        if (stat(node, &st) < 0) {
                log_warning("Couldn't stat device %s", node);
                return -errno;
        }

        if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
                log_warning("%s is not a device.", node);
                return -ENODEV;
        }

        sprintf(buf,
                "%c %u:%u %s",
                S_ISCHR(st.st_mode) ? 'c' : 'b',
                major(st.st_rdev), minor(st.st_rdev),
                acc);

        r = cg_set_attribute("devices", path, "devices.allow", buf);
        if (r < 0)
                log_warning("Failed to set devices.allow on %s: %s", path, strerror(-r));

        return r;
}

static int whitelist_major(const char *path, const char *name, char type, const char *acc) {
        _cleanup_fclose_ FILE *f = NULL;
        char line[LINE_MAX];
        bool good = false;
        int r;

        assert(path);
        assert(acc);
        assert(type == 'b' || type == 'c');

        f = fopen("/proc/devices", "re");
        if (!f) {
                log_warning("Cannot open /proc/devices to resolve %s (%c): %m", name, type);
                return -errno;
        }

        FOREACH_LINE(line, f, goto fail) {
                char buf[2+DECIMAL_STR_MAX(unsigned)+3+4], *p, *w;
                unsigned maj;

                truncate_nl(line);

                if (type == 'c' && streq(line, "Character devices:")) {
                        good = true;
                        continue;
                }

                if (type == 'b' && streq(line, "Block devices:")) {
                        good = true;
                        continue;
                }

                if (isempty(line)) {
                        good = false;
                        continue;
                }

                if (!good)
                        continue;

                p = strstrip(line);

                w = strpbrk(p, WHITESPACE);
                if (!w)
                        continue;
                *w = 0;

                r = safe_atou(p, &maj);
                if (r < 0)
                        continue;
                if (maj <= 0)
                        continue;

                w++;
                w += strspn(w, WHITESPACE);

                if (fnmatch(name, w, 0) != 0)
                        continue;

                sprintf(buf,
                        "%c %u:* %s",
                        type,
                        maj,
                        acc);

                r = cg_set_attribute("devices", path, "devices.allow", buf);
                if (r < 0)
                        log_warning("Failed to set devices.allow on %s: %s", path, strerror(-r));
        }

        return 0;

fail:
        log_warning("Failed to read /proc/devices: %m");
        return -errno;
}

void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const char *path) {
        bool is_root;
        int r;

        assert(c);
        assert(path);

        if (mask == 0)
                return;

        /* Some cgroup attributes are not support on the root cgroup,
         * hence silently ignore */
        is_root = isempty(path) || path_equal(path, "/");

        if ((mask & CGROUP_CPU) && !is_root) {
                char buf[DECIMAL_STR_MAX(unsigned long) + 1];

                sprintf(buf, "%lu\n", c->cpu_shares);
                r = cg_set_attribute("cpu", path, "cpu.shares", buf);
                if (r < 0)
                        log_warning("Failed to set cpu.shares on %s: %s", path, strerror(-r));
        }

        if (mask & CGROUP_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)];
                CGroupBlockIODeviceWeight *w;
                CGroupBlockIODeviceBandwidth *b;

                if (!is_root) {
                        sprintf(buf, "%lu\n", c->blockio_weight);
                        r = cg_set_attribute("blkio", path, "blkio.weight", buf);
                        if (r < 0)
                                log_warning("Failed to set blkio.weight on %s: %s", path, strerror(-r));

                        /* FIXME: no way to reset this list */
                        LIST_FOREACH(device_weights, w, c->blockio_device_weights) {
                                dev_t dev;

                                r = lookup_blkio_device(w->path, &dev);
                                if (r < 0)
                                        continue;

                                sprintf(buf, "%u:%u %lu", major(dev), minor(dev), w->weight);
                                r = cg_set_attribute("blkio", path, "blkio.weight_device", buf);
                                if (r < 0)
                                        log_error("Failed to set blkio.weight_device on %s: %s", path, strerror(-r));
                        }
                }

                /* FIXME: no way to reset this list */
                LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) {
                        const char *a;
                        dev_t dev;

                        r = lookup_blkio_device(b->path, &dev);
                        if (r < 0)
                                continue;

                        a = b->read ? "blkio.throttle.read_bps_device" : "blkio.throttle.write_bps_device";

                        sprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), b->bandwidth);
                        r = cg_set_attribute("blkio", path, a, buf);
                        if (r < 0)
                                log_error("Failed to set %s on %s: %s", a, path, strerror(-r));
                }
        }

        if (mask & CGROUP_MEMORY) {
                if (c->memory_limit != (uint64_t) -1) {
                        char buf[DECIMAL_STR_MAX(uint64_t) + 1];

                        sprintf(buf, "%" PRIu64 "\n", c->memory_limit);
                        r = cg_set_attribute("memory", path, "memory.limit_in_bytes", buf);
                } else
                        r = cg_set_attribute("memory", path, "memory.limit_in_bytes", "-1");

                if (r < 0)
                        log_error("Failed to set memory.limit_in_bytes on %s: %s", path, strerror(-r));
        }

        if ((mask & CGROUP_DEVICE) && !is_root) {
                CGroupDeviceAllow *a;

                if (c->device_allow || c->device_policy != CGROUP_AUTO)
                        r = cg_set_attribute("devices", path, "devices.deny", "a");
                else
                        r = cg_set_attribute("devices", path, "devices.allow", "a");
                if (r < 0)
                        log_warning("Failed to reset devices.list on %s: %s", path, strerror(-r));

                if (c->device_policy == CGROUP_CLOSED ||
                    (c->device_policy == CGROUP_AUTO && c->device_allow)) {
                        static const char auto_devices[] =
                                "/dev/null\0" "rw\0"
                                "/dev/zero\0" "rw\0"
                                "/dev/full\0" "rw\0"
                                "/dev/random\0" "rw\0"
                                "/dev/urandom\0" "rw\0";

                        const char *x, *y;

                        NULSTR_FOREACH_PAIR(x, y, auto_devices)
                                whitelist_device(path, x, y);
                }

                LIST_FOREACH(device_allow, a, c->device_allow) {
                        char acc[4];
                        unsigned k = 0;

                        if (a->r)
                                acc[k++] = 'r';
                        if (a->w)
                                acc[k++] = 'w';
                        if (a->m)
                                acc[k++] = 'm';

                        if (k == 0)
                                continue;

                        acc[k++] = 0;

                        if (startswith(a->path, "/dev/"))
                                whitelist_device(path, a->path, acc);
                        else if (startswith(a->path, "block-"))
                                whitelist_major(path, a->path + 6, 'b', acc);
                        else if (startswith(a->path, "char-"))
                                whitelist_major(path, a->path + 5, 'c', acc);
                        else
                                log_debug("Ignoring device %s while writing cgroup attribute.", a->path);
                }
        }
}

CGroupControllerMask cgroup_context_get_mask(CGroupContext *c) {
        CGroupControllerMask mask = 0;

        /* Figure out which controllers we need */

        if (c->cpu_accounting || c->cpu_shares != 1024)
                mask |= CGROUP_CPUACCT | CGROUP_CPU;

        if (c->blockio_accounting ||
            c->blockio_weight != 1000 ||
            c->blockio_device_weights ||
            c->blockio_device_bandwidths)
                mask |= CGROUP_BLKIO;

        if (c->memory_accounting ||
            c->memory_limit != (uint64_t) -1)
                mask |= CGROUP_MEMORY;

        if (c->device_allow || c->device_policy != CGROUP_AUTO)
                mask |= CGROUP_DEVICE;

        return mask;
}

CGroupControllerMask unit_get_cgroup_mask(Unit *u) {
        CGroupContext *c;

        c = unit_get_cgroup_context(u);
        if (!c)
                return 0;

        return cgroup_context_get_mask(c);
}

CGroupControllerMask unit_get_members_mask(Unit *u) {
        assert(u);

        if (u->cgroup_members_mask_valid)
                return u->cgroup_members_mask;

        u->cgroup_members_mask = 0;

        if (u->type == UNIT_SLICE) {
                Unit *member;
                Iterator i;

                SET_FOREACH(member, u->dependencies[UNIT_BEFORE], i) {

                        if (member == u)
                                continue;

                        if (UNIT_DEREF(member->slice) != u)
                                continue;

                        u->cgroup_members_mask |=
                                unit_get_cgroup_mask(member) |
                                unit_get_members_mask(member);
                }
        }

        u->cgroup_members_mask_valid = true;
        return u->cgroup_members_mask;
}

CGroupControllerMask unit_get_siblings_mask(Unit *u) {
        CGroupControllerMask m;

        assert(u);

        if (UNIT_ISSET(u->slice))
                m = unit_get_members_mask(UNIT_DEREF(u->slice));
        else
                m = unit_get_cgroup_mask(u) | unit_get_members_mask(u);

        /* Sibling propagation is only relevant for weight-based
         * controllers, so let's mask out everything else */
        return m & (CGROUP_CPU|CGROUP_BLKIO|CGROUP_CPUACCT);
}

CGroupControllerMask unit_get_target_mask(Unit *u) {
        CGroupControllerMask mask;

        mask = unit_get_cgroup_mask(u) | unit_get_members_mask(u) | unit_get_siblings_mask(u);
        mask &= u->manager->cgroup_supported;

        return mask;
}

/* Recurse from a unit up through its containing slices, propagating
 * mask bits upward. A unit is also member of itself. */
void unit_update_cgroup_members_masks(Unit *u) {
        CGroupControllerMask m;
        bool more;

        assert(u);

        /* Calculate subtree mask */
        m = unit_get_cgroup_mask(u) | unit_get_members_mask(u);

        /* See if anything changed from the previous invocation. If
         * not, we're done. */
        if (u->cgroup_subtree_mask_valid && m == u->cgroup_subtree_mask)
                return;

        more =
                u->cgroup_subtree_mask_valid &&
                ((m & ~u->cgroup_subtree_mask) != 0) &&
                ((~m & u->cgroup_subtree_mask) == 0);

        u->cgroup_subtree_mask = m;
        u->cgroup_subtree_mask_valid = true;

        if (UNIT_ISSET(u->slice)) {
                Unit *s = UNIT_DEREF(u->slice);

                if (more)
                        /* There's more set now than before. We
                         * propagate the new mask to the parent's mask
                         * (not caring if it actually was valid or
                         * not). */

                        s->cgroup_members_mask |= m;

                else
                        /* There's less set now than before (or we
                         * don't know), we need to recalculate
                         * everything, so let's invalidate the
                         * parent's members mask */

                        s->cgroup_members_mask_valid = false;

                /* And now make sure that this change also hits our
                 * grandparents */
                unit_update_cgroup_members_masks(s);
        }
}

static const char *migrate_callback(CGroupControllerMask mask, void *userdata) {
        Unit *u = userdata;

        assert(mask != 0);
        assert(u);

        while (u) {
                if (u->cgroup_path &&
                    u->cgroup_realized &&
                    (u->cgroup_realized_mask & mask) == mask)
                        return u->cgroup_path;

                u = UNIT_DEREF(u->slice);
        }

        return NULL;
}

static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) {
        _cleanup_free_ char *path = NULL;
        int r;

        assert(u);

        path = unit_default_cgroup_path(u);
        if (!path)
                return log_oom();

        r = hashmap_put(u->manager->cgroup_unit, path, u);
        if (r < 0) {
                log_error(r == -EEXIST ? "cgroup %s exists already: %s" : "hashmap_put failed for %s: %s", path, strerror(-r));
                return r;
        }
        if (r > 0) {
                u->cgroup_path = path;
                path = NULL;
        }

        /* First, create our own group */
        r = cg_create_everywhere(u->manager->cgroup_supported, mask, u->cgroup_path);
        if (r < 0) {
                log_error("Failed to create cgroup %s: %s", u->cgroup_path, strerror(-r));
                return r;
        }

        /* Keep track that this is now realized */
        u->cgroup_realized = true;
        u->cgroup_realized_mask = mask;

        /* Then, possibly move things over */
        r = cg_migrate_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->cgroup_path, migrate_callback, u);
        if (r < 0)
                log_warning("Failed to migrate cgroup from to %s: %s", u->cgroup_path, strerror(-r));

        return 0;
}

static bool unit_has_mask_realized(Unit *u, CGroupControllerMask mask) {
        assert(u);

        return u->cgroup_realized && u->cgroup_realized_mask == mask;
}

/* Check if necessary controllers and attributes for a unit are in place.
 *
 * If so, do nothing.
 * If not, create paths, move processes over, and set attributes.
 *
 * Returns 0 on success and < 0 on failure. */
static int unit_realize_cgroup_now(Unit *u) {
        CGroupControllerMask mask;
        int r;

        assert(u);

        if (u->in_cgroup_queue) {
                LIST_REMOVE(cgroup_queue, u->manager->cgroup_queue, u);
                u->in_cgroup_queue = false;
        }

        mask = unit_get_target_mask(u);

        if (unit_has_mask_realized(u, mask))
                return 0;

        /* First, realize parents */
        if (UNIT_ISSET(u->slice)) {
                r = unit_realize_cgroup_now(UNIT_DEREF(u->slice));
                if (r < 0)
                        return r;
        }

        /* And then do the real work */
        r = unit_create_cgroups(u, mask);
        if (r < 0)
                return r;

        /* Finally, apply the necessary attributes. */
        cgroup_context_apply(unit_get_cgroup_context(u), mask, u->cgroup_path);

        return 0;
}

static void unit_add_to_cgroup_queue(Unit *u) {

        if (u->in_cgroup_queue)
                return;

        LIST_PREPEND(cgroup_queue, u->manager->cgroup_queue, u);
        u->in_cgroup_queue = true;
}

unsigned manager_dispatch_cgroup_queue(Manager *m) {
        Unit *i;
        unsigned n = 0;
        int r;

        while ((i = m->cgroup_queue)) {
                assert(i->in_cgroup_queue);

                r = unit_realize_cgroup_now(i);
                if (r < 0)
                        log_warning("Failed to realize cgroups for queued unit %s: %s", i->id, strerror(-r));

                n++;
        }

        return n;
}

static void unit_queue_siblings(Unit *u) {
        Unit *slice;

        /* This adds the siblings of the specified unit and the
         * siblings of all parent units to the cgroup queue. (But
         * neither the specified unit itself nor the parents.) */

        while ((slice = UNIT_DEREF(u->slice))) {
                Iterator i;
                Unit *m;

                SET_FOREACH(m, slice->dependencies[UNIT_BEFORE], i) {
                        if (m == u)
                                continue;

                        /* Skip units that have a dependency on the slice
                         * but aren't actually in it. */
                        if (UNIT_DEREF(m->slice) != slice)
                                continue;

                        /* No point in doing cgroup application for units
                         * without active processes. */
                        if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(m)))
                                continue;

                        /* If the unit doesn't need any new controllers
                         * and has current ones realized, it doesn't need
                         * any changes. */
                        if (unit_has_mask_realized(m, unit_get_target_mask(m)))
                                continue;

                        unit_add_to_cgroup_queue(m);
                }

                u = slice;
        }
}

int unit_realize_cgroup(Unit *u) {
        CGroupContext *c;

        assert(u);

        c = unit_get_cgroup_context(u);
        if (!c)
                return 0;

        /* So, here's the deal: when realizing the cgroups for this
         * unit, we need to first create all parents, but there's more
         * actually: for the weight-based controllers we also need to
         * make sure that all our siblings (i.e. units that are in the
         * same slice as we are) have cgroups, too. Otherwise, things
         * would become very uneven as each of their processes would
         * get as much resources as all our group together. This call
         * will synchronously create the parent cgroups, but will
         * defer work on the siblings to the next event loop
         * iteration. */

        /* Add all sibling slices to the cgroup queue. */
        unit_queue_siblings(u);

        /* And realize this one now (and apply the values) */
        return unit_realize_cgroup_now(u);
}

void unit_destroy_cgroup(Unit *u) {
        int r;

        assert(u);

        if (!u->cgroup_path)
                return;

        r = cg_trim_everywhere(u->manager->cgroup_supported, u->cgroup_path, !unit_has_name(u, SPECIAL_ROOT_SLICE));
        if (r < 0)
                log_debug("Failed to destroy cgroup %s: %s", u->cgroup_path, strerror(-r));

        hashmap_remove(u->manager->cgroup_unit, u->cgroup_path);

        free(u->cgroup_path);
        u->cgroup_path = NULL;
        u->cgroup_realized = false;
        u->cgroup_realized_mask = 0;

}

pid_t unit_search_main_pid(Unit *u) {
        _cleanup_fclose_ FILE *f = NULL;
        pid_t pid = 0, npid, mypid;

        assert(u);

        if (!u->cgroup_path)
                return 0;

        if (cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, &f) < 0)
                return 0;

        mypid = getpid();
        while (cg_read_pid(f, &npid) > 0)  {
                pid_t ppid;

                if (npid == pid)
                        continue;

                /* Ignore processes that aren't our kids */
                if (get_parent_of_pid(npid, &ppid) >= 0 && ppid != mypid)
                        continue;

                if (pid != 0) {
                        /* Dang, there's more than one daemonized PID
                        in this group, so we don't know what process
                        is the main process. */
                        pid = 0;
                        break;
                }

                pid = npid;
        }

        return pid;
}

int manager_setup_cgroup(Manager *m) {
        _cleanup_free_ char *path = NULL;
        char *e;
        int r;

        assert(m);

        /* 0. Be nice to Ingo Molnar #628004 */
        if (path_is_mount_point("/sys/fs/cgroup/systemd", false) <= 0) {
                log_warning("No control group support available, not creating root group.");
                return 0;
        }

        /* 1. Determine hierarchy */
        free(m->cgroup_root);
        m->cgroup_root = NULL;

        r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, &m->cgroup_root);
        if (r < 0) {
                log_error("Cannot determine cgroup we are running in: %s", strerror(-r));
                return r;
        }

        /* LEGACY: Already in /system.slice? If so, let's cut this
         * off. This is to support live upgrades from older systemd
         * versions where PID 1 was moved there. */
        if (m->running_as == SYSTEMD_SYSTEM) {
                e = endswith(m->cgroup_root, "/" SPECIAL_SYSTEM_SLICE);
                if (!e)
                        e = endswith(m->cgroup_root, "/system");
                if (e)
                        *e = 0;
        }

        /* And make sure to store away the root value without trailing
         * slash, even for the root dir, so that we can easily prepend
         * it everywhere. */
        if (streq(m->cgroup_root, "/"))
                m->cgroup_root[0] = 0;

        /* 2. Show data */
        r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, NULL, &path);
        if (r < 0) {
                log_error("Cannot find cgroup mount point: %s", strerror(-r));
                return r;
        }

        log_debug("Using cgroup controller " SYSTEMD_CGROUP_CONTROLLER ". File system hierarchy is at %s.", path);

        /* 3. Install agent */
        if (m->running_as == SYSTEMD_SYSTEM) {
                r = cg_install_release_agent(SYSTEMD_CGROUP_CONTROLLER, SYSTEMD_CGROUP_AGENT_PATH);
                if (r < 0)
                        log_warning("Failed to install release agent, ignoring: %s", strerror(-r));
                else if (r > 0)
                        log_debug("Installed release agent.");
                else
                        log_debug("Release agent already installed.");
        }

        /* 4. Make sure we are in the root cgroup */
        r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, 0);
        if (r < 0) {
                log_error("Failed to create root cgroup hierarchy: %s", strerror(-r));
                return r;
        }

        /* 5. And pin it, so that it cannot be unmounted */
        if (m->pin_cgroupfs_fd >= 0)
                close_nointr_nofail(m->pin_cgroupfs_fd);

        m->pin_cgroupfs_fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY|O_NOCTTY|O_NONBLOCK);
        if (r < 0) {
                log_error("Failed to open pin file: %m");
                return -errno;
        }

        /* 6. Figure out which controllers are supported */
        m->cgroup_supported = cg_mask_supported();

        /* 7.  Always enable hierarchial support if it exists... */
        cg_set_attribute("memory", "/", "memory.use_hierarchy", "1");

        return 0;
}

void manager_shutdown_cgroup(Manager *m, bool delete) {
        assert(m);

        /* We can't really delete the group, since we are in it. But
         * let's trim it. */
        if (delete && m->cgroup_root)
                cg_trim(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, false);

        if (m->pin_cgroupfs_fd >= 0) {
                close_nointr_nofail(m->pin_cgroupfs_fd);
                m->pin_cgroupfs_fd = -1;
        }

        free(m->cgroup_root);
        m->cgroup_root = NULL;
}

Unit* manager_get_unit_by_cgroup(Manager *m, const char *cgroup) {
        char *p;
        Unit *u;

        assert(m);
        assert(cgroup);

        u = hashmap_get(m->cgroup_unit, cgroup);
        if (u)
                return u;

        p = strdupa(cgroup);
        for (;;) {
                char *e;

                e = strrchr(p, '/');
                if (e == p || !e)
                        return NULL;

                *e = 0;

                u = hashmap_get(m->cgroup_unit, p);
                if (u)
                        return u;
        }
}

Unit *manager_get_unit_by_pid(Manager *m, pid_t pid) {
        _cleanup_free_ char *cgroup = NULL;
        int r;

        assert(m);

        if (pid <= 1)
                return NULL;

        r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &cgroup);
        if (r < 0)
                return NULL;

        return manager_get_unit_by_cgroup(m, cgroup);
}

int manager_notify_cgroup_empty(Manager *m, const char *cgroup) {
        Unit *u;
        int r;

        assert(m);
        assert(cgroup);

        u = manager_get_unit_by_cgroup(m, cgroup);
        if (u) {
                r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, true);
                if (r > 0) {
                        if (UNIT_VTABLE(u)->notify_cgroup_empty)
                                UNIT_VTABLE(u)->notify_cgroup_empty(u);

                        unit_add_to_gc_queue(u);
                }
        }

        return 0;
}

static const char* const cgroup_device_policy_table[_CGROUP_DEVICE_POLICY_MAX] = {
        [CGROUP_AUTO] = "auto",
        [CGROUP_CLOSED] = "closed",
        [CGROUP_STRICT] = "strict",
};

DEFINE_STRING_TABLE_LOOKUP(cgroup_device_policy, CGroupDevicePolicy);