External/libvirt
1
0
Fork 0
mirror of https://gitlab.com/libvirt/libvirt.git synced 2024-07-03 16:42:37 +00:00
Code Issues Packages Projects Releases Wiki Activity
9eb68903f5
libvirt/tests/vircgroupmock.c

608 lines
21 KiB
C
Raw Normal View History

Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
/*
* Copyright (C) 2013 Red Hat, Inc.
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library. If not, see
* <http://www.gnu.org/licenses/>.
*
* Author: Daniel P. Berrange <berrange@redhat.com>
*/
#include <config.h>
#ifdef __linux__
# include "internal.h"
# include <stdio.h>
# include <dlfcn.h>
# include <stdlib.h>
# include <unistd.h>
# include <fcntl.h>
# include <sys/stat.h>
# include <stdarg.h>
static int (*realopen)(const char *path, int flags, ...);
static FILE *(*realfopen)(const char *path, const char *mode);
static int (*realaccess)(const char *path, int mode);
Track symlinks for co-mounted cgroup controllers If a cgroup controller is co-mounted with another, eg /sys/fs/cgroup/cpu,cpuacct Then it is a requirement that there exist symlinks at /sys/fs/cgroup/cpu /sys/fs/cgroup/cpuacct pointing to the real mount point. Add support to virCgroupPtr to detect and track these symlinks Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-05 10:28:04 +00:00
static int (*reallstat)(const char *path, struct stat *sb);
static int (*real__lxstat)(int ver, const char *path, struct stat *sb);
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
static int (*realmkdir)(const char *path, mode_t mode);
Workaround issue with clang and inline functions with static vars Clang does not like it when you pass a static variable to an inline function vircgroupmock.c:462:22: error: static variable 'fakesysfsdir' is used in an inline function with external linkage [-Werror,-Wstatic-in-inline] Just make the var non-static to avoid this Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-05-13 11:43:08 +00:00
/* Don't make static, since it causes problems with clang
* when passed as an arg to asprintf()
* vircgroupmock.c:462:22: error: static variable 'fakesysfsdir' is used in an inline function with external linkage [-Werror,-Wstatic-in-inline]
*/
char *fakesysfsdir;
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
# define SYSFS_PREFIX "/not/really/sys/fs/cgroup/"
/*
* The plan:
*
* We fake out /proc/mounts, so make it look as is cgroups
* are mounted on /not/really/sys/fs/cgroup. We don't
* use /sys/fs/cgroup, because we want to make it easy to
* detect places where we've not mocked enough syscalls.
*
* In any open/acces/mkdir calls we look at path and if
* it starts with /not/really/sys/fs/cgroup, we rewrite
* the path to point at a temporary directory referred
* to by LIBVIRT_FAKE_SYSFS_DIR env variable that is
* set by the main test suite
*
* In mkdir() calls, we simulate the cgroups behaviour
* whereby creating the directory auto-creates a bunch
* of files beneath it
*/
/*
* Intentionally missing the 'devices' mount.
* Co-mounting cpu & cpuacct controllers
* An anonymous controller for systemd
*/
Do proper escaping of cgroup resource partitions If a user cgroup name begins with "cgroup.", "_" or with any of the controllers from /proc/cgroups followed by a dot, then they need to be prefixed with a single underscore. eg if there is an object "cpu.service", then this would end up as "_cpu.service" in the cgroup filesystem tree, however, "waldo.service" would stay "waldo.service", at least as long as nobody comes up with a cgroup controller called "waldo". Since we require a '.XXXX' suffix on all partitions, there is no scope for clashing with the kernel 'tasks' and 'release_agent' files. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-26 09:50:24 +00:00
const char *procmounts =
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
"rootfs / rootfs rw 0 0\n"
"tmpfs /run tmpfs rw,seclabel,nosuid,nodev,mode=755 0 0\n"
"tmpfs /not/really/sys/fs/cgroup tmpfs rw,seclabel,nosuid,nodev,noexec,mode=755 0 0\n"
"cgroup /not/really/sys/fs/cgroup/systemd cgroup rw,nosuid,nodev,noexec,relatime,release_agent=/usr/lib/systemd/systemd-cgroups-agent,name=systemd 0 0\n"
"cgroup /not/really/sys/fs/cgroup/cpuset cgroup rw,nosuid,nodev,noexec,relatime,cpuset 0 0\n"
"cgroup /not/really/sys/fs/cgroup/cpu,cpuacct cgroup rw,nosuid,nodev,noexec,relatime,cpuacct,cpu 0 0\n"
"cgroup /not/really/sys/fs/cgroup/freezer cgroup rw,nosuid,nodev,noexec,relatime,freezer 0 0\n"
"cgroup /not/really/sys/fs/cgroup/blkio cgroup rw,nosuid,nodev,noexec,relatime,blkio 0 0\n"
"cgroup /not/really/sys/fs/cgroup/memory cgroup rw,nosuid,nodev,noexec,relatime,memory 0 0\n"
"/dev/sda1 /boot ext4 rw,seclabel,relatime,data=ordered 0 0\n"
"tmpfs /tmp tmpfs rw,seclabel,relatime,size=1024000k 0 0\n";
Do proper escaping of cgroup resource partitions If a user cgroup name begins with "cgroup.", "_" or with any of the controllers from /proc/cgroups followed by a dot, then they need to be prefixed with a single underscore. eg if there is an object "cpu.service", then this would end up as "_cpu.service" in the cgroup filesystem tree, however, "waldo.service" would stay "waldo.service", at least as long as nobody comes up with a cgroup controller called "waldo". Since we require a '.XXXX' suffix on all partitions, there is no scope for clashing with the kernel 'tasks' and 'release_agent' files. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-26 09:50:24 +00:00
const char *procselfcgroups =
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
"115:memory:/\n"
"8:blkio:/\n"
"6:freezer:/\n"
"3:cpuacct,cpu:/system\n"
"2:cpuset:/\n"
"1:name=systemd:/user/berrange/123\n";
Do proper escaping of cgroup resource partitions If a user cgroup name begins with "cgroup.", "_" or with any of the controllers from /proc/cgroups followed by a dot, then they need to be prefixed with a single underscore. eg if there is an object "cpu.service", then this would end up as "_cpu.service" in the cgroup filesystem tree, however, "waldo.service" would stay "waldo.service", at least as long as nobody comes up with a cgroup controller called "waldo". Since we require a '.XXXX' suffix on all partitions, there is no scope for clashing with the kernel 'tasks' and 'release_agent' files. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-26 09:50:24 +00:00
const char *proccgroups =
"#subsys_name hierarchy num_cgroups enabled\n"
"cpuset 2 4 1\n"
"cpu 3 48 1\n"
"cpuacct 3 48 1\n"
"memory 4 4 1\n"
"devices 5 4 1\n"
"freezer 6 4 1\n"
"blkio 8 4 1\n";
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
const char *procmountsallinone =
"rootfs / rootfs rw 0 0\n"
"sysfs /sys sysfs rw,nosuid,nodev,noexec,relatime 0 0\n"
"proc /proc proc rw,nosuid,nodev,noexec,relatime 0 0\n"
"udev /dev devtmpfs rw,relatime,size=16458560k,nr_inodes=4114640,mode=755 0 0\n"
"devpts /dev/pts devpts rw,nosuid,noexec,relatime,gid=5,mode=620,ptmxmode=000 0 0\n"
"nfsd /proc/fs/nfsd nfsd rw,relatime 0 0\n"
"cgroup /not/really/sys/fs/cgroup cgroup rw,relatime,blkio,devices,memory,cpuacct,cpu,cpuset 0 0\n";
const char *procselfcgroupsallinone =
"6:blkio,devices,memory,cpuacct,cpu,cpuset:/";
const char *proccgroupsallinone =
"#subsys_name hierarchy num_cgroups enabled\n"
"cpuset 6 1 1\n"
"cpu 6 1 1\n"
"cpuacct 6 1 1\n"
"memory 6 1 1\n"
"devices 6 1 1\n"
"blkio 6 1 1\n";
Fix launching of VMs on when only logind part of systemd is present Debian systems may run the 'systemd-logind' daemon, which causes the /sys/fs/cgroup/systemd mount to be setup, but no other cgroup controllers are created. While the LXC driver considers cgroups to be mandatory, the QEMU driver is supposed to accept them as optional. We detect whether they are present by looking in /proc/mounts for any mounts of type 'cgroups', but this is not sufficient. We need to skip any named mounts (as seen by a name=XXX string in the mount options), so that we only detect actual resource controllers. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721979 Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-11 18:15:52 +00:00
const char *procmountslogind =
"none /not/really/sys/fs/cgroup tmpfs rw,rootcontext=system_u:object_r:sysfs_t:s0,seclabel,relatime,size=4k,mode=755 0 0\n"
"systemd /not/really/sys/fs/cgroup/systemd cgroup rw,nosuid,nodev,noexec,relatime,name=systemd 0 0\n";
const char *procselfcgroupslogind =
"1:name=systemd:/\n";
const char *proccgroupslogind =
"#subsys_name hierarchy num_cgroups enabled\n"
"cpuset 0 1 1\n"
"cpu 0 1 1\n"
"cpuacct 0 1 1\n"
"memory 0 1 0\n"
"devices 0 1 1\n"
"freezer 0 1 1\n"
"net_cls 0 1 1\n"
"blkio 0 1 1\n"
"perf_event 0 1 1\n";
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
static int make_file(const char *path,
const char *name,
const char *value)
{
int fd = -1;
int ret = -1;
char *filepath = NULL;
if (asprintf(&filepath, "%s/%s", path, name) < 0)
return -1;
if ((fd = realopen(filepath, O_CREAT|O_WRONLY, 0600)) < 0)
goto cleanup;
if (write(fd, value, strlen(value)) != strlen(value))
goto cleanup;
ret = 0;
cleanup:
if (fd != -1 && close(fd) < 0)
ret = -1;
free(filepath);
return ret;
}
static int make_controller(const char *path, mode_t mode)
{
int ret = -1;
const char *controller;
if (!STRPREFIX(path, fakesysfsdir)) {
errno = EINVAL;
return -1;
}
controller = path + strlen(fakesysfsdir) + 1;
if (STREQ(controller, "cpu"))
return symlink("cpu,cpuacct", path);
if (STREQ(controller, "cpuacct"))
return symlink("cpu,cpuacct", path);
if (realmkdir(path, mode) < 0)
goto cleanup;
# define MAKE_FILE(name, value) \
do { \
if (make_file(path, name, value) < 0) \
goto cleanup; \
} while (0)
if (STRPREFIX(controller, "cpu,cpuacct")) {
MAKE_FILE("cpu.cfs_period_us", "100000\n");
MAKE_FILE("cpu.cfs_quota_us", "-1\n");
MAKE_FILE("cpu.rt_period_us", "1000000\n");
MAKE_FILE("cpu.rt_runtime_us", "950000\n");
MAKE_FILE("cpu.shares", "1024\n");
MAKE_FILE("cpu.stat",
"nr_periods 0\n"
"nr_throttled 0\n"
"throttled_time 0\n");
MAKE_FILE("cpuacct.stat",
"user 216687025\n"
"system 43421396\n");
MAKE_FILE("cpuacct.usage", "2787788855799582\n");
MAKE_FILE("cpuacct.usage_per_cpu", "1413142688153030 1374646168910542\n");
} else if (STRPREFIX(controller, "cpuset")) {
MAKE_FILE("cpuset.cpu_exclusive", "1\n");
if (STREQ(controller, "cpuset"))
MAKE_FILE("cpuset.cpus", "0-1");
else
MAKE_FILE("cpuset.cpus", ""); /* Values don't inherit */
MAKE_FILE("cpuset.mem_exclusive", "1\n");
MAKE_FILE("cpuset.mem_hardwall", "0\n");
MAKE_FILE("cpuset.memory_migrate", "0\n");
MAKE_FILE("cpuset.memory_pressure", "0\n");
MAKE_FILE("cpuset.memory_pressure_enabled", "0\n");
MAKE_FILE("cpuset.memory_spread_page", "0\n");
MAKE_FILE("cpuset.memory_spread_slab", "0\n");
if (STREQ(controller, "cpuset"))
MAKE_FILE("cpuset.mems", "0");
else
MAKE_FILE("cpuset.mems", ""); /* Values don't inherit */
MAKE_FILE("cpuset.sched_load_balance", "1\n");
MAKE_FILE("cpuset.sched_relax_domain_level", "-1\n");
} else if (STRPREFIX(controller, "memory")) {
MAKE_FILE("memory.failcnt", "0\n");
MAKE_FILE("memory.force_empty", ""); /* Write only */
MAKE_FILE("memory.kmem.tcp.failcnt", "0\n");
MAKE_FILE("memory.kmem.tcp.limit_in_bytes", "9223372036854775807\n");
MAKE_FILE("memory.kmem.tcp.max_usage_in_bytes", "0\n");
MAKE_FILE("memory.kmem.tcp.usage_in_bytes", "16384\n");
MAKE_FILE("memory.limit_in_bytes", "9223372036854775807\n");
MAKE_FILE("memory.max_usage_in_bytes", "0\n");
MAKE_FILE("memory.memsw.failcnt", ""); /* Not supported */
MAKE_FILE("memory.memsw.limit_in_bytes", ""); /* Not supported */
MAKE_FILE("memory.memsw.max_usage_in_bytes", ""); /* Not supported */
MAKE_FILE("memory.memsw.usage_in_bytes", ""); /* Not supported */
MAKE_FILE("memory.move_charge_at_immigrate", "0\n");
MAKE_FILE("memory.numa_stat",
"total=367664 N0=367664\n"
"file=314764 N0=314764\n"
"anon=51999 N0=51999\n"
"unevictable=901 N0=901\n");
MAKE_FILE("memory.oom_control",
"oom_kill_disable 0\n"
"under_oom 0\n");
MAKE_FILE("memory.soft_limit_in_bytes", "9223372036854775807\n");
MAKE_FILE("memory.stat",
"cache 1336619008\n"
"rss 97792000\n"
"mapped_file 42090496\n"
"pgpgin 13022605027\n"
"pgpgout 13023820533\n"
"pgfault 54429417056\n"
"pgmajfault 315715\n"
"inactive_anon 145887232\n"
"active_anon 67100672\n"
"inactive_file 627400704\n"
"active_file 661872640\n"
"unevictable 3690496\n"
"hierarchical_memory_limit 9223372036854775807\n"
"total_cache 1336635392\n"
"total_rss 118689792\n"
"total_mapped_file 42106880\n"
"total_pgpgin 13022606816\n"
"total_pgpgout 13023820793\n"
"total_pgfault 54429422313\n"
"total_pgmajfault 315715\n"
"total_inactive_anon 145891328\n"
"total_active_anon 88010752\n"
"total_inactive_file 627400704\n"
"total_active_file 661872640\n"
"total_unevictable 3690496\n"
"recent_rotated_anon 112807028\n"
"recent_rotated_file 2547948\n"
"recent_scanned_anon 113796164\n"
"recent_scanned_file 8199863\n");
MAKE_FILE("memory.swappiness", "60\n");
MAKE_FILE("memory.usage_in_bytes", "1455321088\n");
MAKE_FILE("memory.use_hierarchy", "0\n");
} else if (STRPREFIX(controller, "freezer")) {
MAKE_FILE("freezer.state", "THAWED");
} else if (STRPREFIX(controller, "blkio")) {
MAKE_FILE("blkio.io_merged",
"8:0 Read 1100949\n"
"8:0 Write 2248076\n"
"8:0 Sync 63063\n"
"8:0 Async 3285962\n"
"8:0 Total 3349025\n");
MAKE_FILE("blkio.io_queued",
"8:0 Read 0\n"
"8:0 Write 0\n"
"8:0 Sync 0\n"
"8:0 Async 0\n"
"8:0 Total 0\n");
MAKE_FILE("blkio.io_service_bytes",
"8:0 Read 59542078464\n"
"8:0 Write 397369182208\n"
"8:0 Sync 234080922624\n"
"8:0 Async 222830338048\n"
"8:0 Total 456911260672\n");
MAKE_FILE("blkio.io_serviced",
"8:0 Read 3402504\n"
"8:0 Write 14966516\n"
"8:0 Sync 12064031\n"
"8:0 Async 6304989\n"
"8:0 Total 18369020\n");
MAKE_FILE("blkio.io_service_time",
"8:0 Read 10747537542349\n"
"8:0 Write 9200028590575\n"
"8:0 Sync 6449319855381\n"
"8:0 Async 13498246277543\n"
"8:0 Total 19947566132924\n");
MAKE_FILE("blkio.io_wait_time",
"8:0 Read 14687514824889\n"
"8:0 Write 357748452187691\n"
"8:0 Sync 55296974349413\n"
"8:0 Async 317138992663167\n"
"8:0 Total 372435967012580\n");
MAKE_FILE("blkio.reset_stats", ""); /* Write only */
MAKE_FILE("blkio.sectors", "8:0 892404806\n");
MAKE_FILE("blkio.throttle.io_service_bytes",
"8:0 Read 59542107136\n"
"8:0 Write 411440480256\n"
"8:0 Sync 248486822912\n"
"8:0 Async 222495764480\n"
"8:0 Total 470982587392\n");
MAKE_FILE("blkio.throttle.io_serviced",
"8:0 Read 4832583\n"
"8:0 Write 36641903\n"
"8:0 Sync 30723171\n"
"8:0 Async 10751315\n"
"8:0 Total 41474486\n");
MAKE_FILE("blkio.throttle.read_bps_device", "");
MAKE_FILE("blkio.throttle.read_iops_device", "");
MAKE_FILE("blkio.throttle.write_bps_device", "");
MAKE_FILE("blkio.throttle.write_iops_device", "");
MAKE_FILE("blkio.time", "8:0 61019089\n");
MAKE_FILE("blkio.weight", "1000\n");
MAKE_FILE("blkio.weight_device", "");
} else {
errno = EINVAL;
goto cleanup;
}
ret = 0;
cleanup:
return ret;
}
static void init_syms(void)
{
if (realfopen)
return;
# define LOAD_SYM(name) \
do { \
if (!(real ## name = dlsym(RTLD_NEXT, #name))) { \
fprintf(stderr, "Cannot find real '%s' symbol\n", #name); \
abort(); \
} \
} while (0)
Track symlinks for co-mounted cgroup controllers If a cgroup controller is co-mounted with another, eg /sys/fs/cgroup/cpu,cpuacct Then it is a requirement that there exist symlinks at /sys/fs/cgroup/cpu /sys/fs/cgroup/cpuacct pointing to the real mount point. Add support to virCgroupPtr to detect and track these symlinks Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-05 10:28:04 +00:00
# define LOAD_SYM_ALT(name1, name2) \
do { \
if (!(real ## name1 = dlsym(RTLD_NEXT, #name1)) && \
!(real ## name2 = dlsym(RTLD_NEXT, #name2))) { \
fprintf(stderr, "Cannot find real '%s' or '%s' symbol\n", #name1, #name2); \
abort(); \
} \
} while (0)
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
LOAD_SYM(fopen);
LOAD_SYM(access);
Track symlinks for co-mounted cgroup controllers If a cgroup controller is co-mounted with another, eg /sys/fs/cgroup/cpu,cpuacct Then it is a requirement that there exist symlinks at /sys/fs/cgroup/cpu /sys/fs/cgroup/cpuacct pointing to the real mount point. Add support to virCgroupPtr to detect and track these symlinks Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-05 10:28:04 +00:00
LOAD_SYM_ALT(lstat, __lxstat);
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
LOAD_SYM(mkdir);
LOAD_SYM(open);
}
static void init_sysfs(void)
{
if (fakesysfsdir)
return;
if (!(fakesysfsdir = getenv("LIBVIRT_FAKE_SYSFS_DIR"))) {
fprintf(stderr, "Missing LIBVIRT_FAKE_SYSFS_DIR env variable\n");
abort();
}
# define MAKE_CONTROLLER(subpath) \
do { \
char *path; \
if (asprintf(&path,"%s/%s", fakesysfsdir, subpath) < 0) \
abort(); \
if (make_controller(path, 0755) < 0) { \
fprintf(stderr, "Cannot initialize %s\n", path); \
abort(); \
} \
} while (0)
MAKE_CONTROLLER("cpu");
MAKE_CONTROLLER("cpuacct");
MAKE_CONTROLLER("cpu,cpuacct");
MAKE_CONTROLLER("cpu,cpuacct/system");
MAKE_CONTROLLER("cpuset");
MAKE_CONTROLLER("blkio");
MAKE_CONTROLLER("memory");
MAKE_CONTROLLER("freezer");
}
FILE *fopen(const char *path, const char *mode)
{
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
const char *mock;
Fix launching of VMs on when only logind part of systemd is present Debian systems may run the 'systemd-logind' daemon, which causes the /sys/fs/cgroup/systemd mount to be setup, but no other cgroup controllers are created. While the LXC driver considers cgroups to be mandatory, the QEMU driver is supposed to accept them as optional. We detect whether they are present by looking in /proc/mounts for any mounts of type 'cgroups', but this is not sufficient. We need to skip any named mounts (as seen by a name=XXX string in the mount options), so that we only detect actual resource controllers. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721979 Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-11 18:15:52 +00:00
bool allinone = false, logind = false;
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
init_syms();
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
mock = getenv("VIR_CGROUP_MOCK_MODE");
Fix launching of VMs on when only logind part of systemd is present Debian systems may run the 'systemd-logind' daemon, which causes the /sys/fs/cgroup/systemd mount to be setup, but no other cgroup controllers are created. While the LXC driver considers cgroups to be mandatory, the QEMU driver is supposed to accept them as optional. We detect whether they are present by looking in /proc/mounts for any mounts of type 'cgroups', but this is not sufficient. We need to skip any named mounts (as seen by a name=XXX string in the mount options), so that we only detect actual resource controllers. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721979 Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-11 18:15:52 +00:00
if (mock) {
if (STREQ(mock, "allinone"))
allinone = true;
else if (STREQ(mock, "logind"))
logind = true;
}
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
if (STREQ(path, "/proc/mounts")) {
if (STREQ(mode, "r")) {
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
if (allinone)
return fmemopen((void *)procmountsallinone,
strlen(procmountsallinone), mode);
Fix launching of VMs on when only logind part of systemd is present Debian systems may run the 'systemd-logind' daemon, which causes the /sys/fs/cgroup/systemd mount to be setup, but no other cgroup controllers are created. While the LXC driver considers cgroups to be mandatory, the QEMU driver is supposed to accept them as optional. We detect whether they are present by looking in /proc/mounts for any mounts of type 'cgroups', but this is not sufficient. We need to skip any named mounts (as seen by a name=XXX string in the mount options), so that we only detect actual resource controllers. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721979 Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-11 18:15:52 +00:00
else if (logind)
return fmemopen((void *)procmountslogind,
strlen(procmountslogind), mode);
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
else
return fmemopen((void *)procmounts, strlen(procmounts), mode);
Do proper escaping of cgroup resource partitions If a user cgroup name begins with "cgroup.", "_" or with any of the controllers from /proc/cgroups followed by a dot, then they need to be prefixed with a single underscore. eg if there is an object "cpu.service", then this would end up as "_cpu.service" in the cgroup filesystem tree, however, "waldo.service" would stay "waldo.service", at least as long as nobody comes up with a cgroup controller called "waldo". Since we require a '.XXXX' suffix on all partitions, there is no scope for clashing with the kernel 'tasks' and 'release_agent' files. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-26 09:50:24 +00:00
} else {
errno = EACCES;
return NULL;
}
}
if (STREQ(path, "/proc/cgroups")) {
if (STREQ(mode, "r")) {
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
if (allinone)
return fmemopen((void *)proccgroupsallinone,
strlen(proccgroupsallinone), mode);
Fix launching of VMs on when only logind part of systemd is present Debian systems may run the 'systemd-logind' daemon, which causes the /sys/fs/cgroup/systemd mount to be setup, but no other cgroup controllers are created. While the LXC driver considers cgroups to be mandatory, the QEMU driver is supposed to accept them as optional. We detect whether they are present by looking in /proc/mounts for any mounts of type 'cgroups', but this is not sufficient. We need to skip any named mounts (as seen by a name=XXX string in the mount options), so that we only detect actual resource controllers. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721979 Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-11 18:15:52 +00:00
else if (logind)
return fmemopen((void *)proccgroupslogind,
strlen(proccgroupslogind), mode);
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
else
return fmemopen((void *)proccgroups, strlen(proccgroups), mode);
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
} else {
errno = EACCES;
return NULL;
}
}
if (STREQ(path, "/proc/self/cgroup")) {
if (STREQ(mode, "r")) {
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
if (allinone)
return fmemopen((void *)procselfcgroupsallinone,
strlen(procselfcgroupsallinone), mode);
Fix launching of VMs on when only logind part of systemd is present Debian systems may run the 'systemd-logind' daemon, which causes the /sys/fs/cgroup/systemd mount to be setup, but no other cgroup controllers are created. While the LXC driver considers cgroups to be mandatory, the QEMU driver is supposed to accept them as optional. We detect whether they are present by looking in /proc/mounts for any mounts of type 'cgroups', but this is not sufficient. We need to skip any named mounts (as seen by a name=XXX string in the mount options), so that we only detect actual resource controllers. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721979 Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-11 18:15:52 +00:00
else if (logind)
return fmemopen((void *)procselfcgroupslogind,
strlen(procselfcgroupslogind), mode);
Fix cgroups when all are mounted on /sys/fs/cgroup Some users in Ubuntu/Debian seem to have a setup where all the cgroup controllers are mounted on /sys/fs/cgroup rather than any /sys/fs/cgroup/<controller> name. In the loop which detects which controllers are present for a mount point we were modifying 'mnt_dir' field in the 'struct mntent' var, but not always restoring the original value. This caused detection to break in the all-in-one mount setup. Fix that logic bug and add test case coverage for this mount setup. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-09-10 13:31:53 +00:00
else
return fmemopen((void *)procselfcgroups, strlen(procselfcgroups), mode);
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
} else {
errno = EACCES;
return NULL;
}
}
return realfopen(path, mode);
}
int access(const char *path, int mode)
{
int ret;
init_syms();
if (STRPREFIX(path, SYSFS_PREFIX)) {
init_sysfs();
char *newpath;
if (asprintf(&newpath, "%s/%s",
fakesysfsdir,
path + strlen(SYSFS_PREFIX)) < 0) {
errno = ENOMEM;
return -1;
}
ret = realaccess(newpath, mode);
free(newpath);
} else {
ret = realaccess(path, mode);
}
return ret;
}
Track symlinks for co-mounted cgroup controllers If a cgroup controller is co-mounted with another, eg /sys/fs/cgroup/cpu,cpuacct Then it is a requirement that there exist symlinks at /sys/fs/cgroup/cpu /sys/fs/cgroup/cpuacct pointing to the real mount point. Add support to virCgroupPtr to detect and track these symlinks Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-04-05 10:28:04 +00:00
int __lxstat(int ver, const char *path, struct stat *sb)
{
int ret;
init_syms();
if (STRPREFIX(path, SYSFS_PREFIX)) {
init_sysfs();
char *newpath;
if (asprintf(&newpath, "%s/%s",
fakesysfsdir,
path + strlen(SYSFS_PREFIX)) < 0) {
errno = ENOMEM;
return -1;
}
ret = real__lxstat(ver, newpath, sb);
free(newpath);
} else {
ret = real__lxstat(ver, path, sb);
}
return ret;
}
int lstat(const char *path, struct stat *sb)
{
int ret;
init_syms();
if (STRPREFIX(path, SYSFS_PREFIX)) {
init_sysfs();
char *newpath;
if (asprintf(&newpath, "%s/%s",
fakesysfsdir,
path + strlen(SYSFS_PREFIX)) < 0) {
errno = ENOMEM;
return -1;
}
ret = reallstat(newpath, sb);
free(newpath);
} else {
ret = reallstat(path, sb);
}
return ret;
}
Add a test suite for cgroups functionality Some aspects of the cgroups setup / detection code are quite subtle and easy to break. It would greatly benefit from unit testing, but this is difficult because the test suite won't have privileges to play around with cgroups. The solution is to use monkey patching via LD_PRELOAD to override the fopen, open, mkdir, access functions to redirect access of cgroups files to some magic stubs in the test suite. Using this we provide custom content for the /proc/cgroup and /proc/self/mounts files which report a fixed cgroup setup. We then override open/mkdir/access so that access to the cgroups filesystem gets redirected into files in a temporary directory tree in the test suite build dir. Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2013-03-28 14:36:52 +00:00
int mkdir(const char *path, mode_t mode)
{
int ret;
init_syms();
if (STRPREFIX(path, SYSFS_PREFIX)) {
init_sysfs();
char *newpath;
if (asprintf(&newpath, "%s/%s",
fakesysfsdir,
path + strlen(SYSFS_PREFIX)) < 0) {
errno = ENOMEM;
return -1;
}
ret = make_controller(newpath, mode);
free(newpath);
} else {
ret = realmkdir(path, mode);
}
return ret;
}
int open(const char *path, int flags, ...)
{
int ret;
char *newpath = NULL;
init_syms();
if (STRPREFIX(path, SYSFS_PREFIX)) {
init_sysfs();
if (asprintf(&newpath, "%s/%s",
fakesysfsdir,
path + strlen(SYSFS_PREFIX)) < 0) {
errno = ENOMEM;
return -1;
}
}
if (flags & O_CREAT) {
va_list ap;
mode_t mode;
va_start(ap, flags);
mode = va_arg(ap, mode_t);
va_end(ap);
ret = realopen(newpath ? newpath : path, flags, mode);
} else {
ret = realopen(newpath ? newpath : path, flags);
}
free(newpath);
return ret;
}
#else
/* Nothing to override on non-__linux__ platforms */
#endif
Reference in New Issue Copy Permalink