2013-06-17 23:33:26 +04:00
/***
This file is part of systemd .
Copyright 2013 Lennart Poettering
systemd 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 .
systemd 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 systemd ; If not , see < http : //www.gnu.org/licenses/>.
* * */
# include <errno.h>
2015-10-27 05:01:06 +03:00
# include "alloc-util.h"
2015-10-24 23:58:24 +03:00
# include "dbus-slice.h"
2013-06-17 23:33:26 +04:00
# include "log.h"
2015-11-17 00:09:36 +03:00
# include "slice.h"
2013-06-17 23:33:26 +04:00
# include "special.h"
2015-10-24 23:58:24 +03:00
# include "string-util.h"
# include "strv.h"
2013-06-17 23:33:26 +04:00
# include "unit-name.h"
core: unified cgroup hierarchy support
This patch set adds full support the new unified cgroup hierarchy logic
of modern kernels.
A new kernel command line option "systemd.unified_cgroup_hierarchy=1" is
added. If specified the unified hierarchy is mounted to /sys/fs/cgroup
instead of a tmpfs. No further hierarchies are mounted. The kernel
command line option defaults to off. We can turn it on by default as
soon as the kernel's APIs regarding this are stabilized (but even then
downstream distros might want to turn this off, as this will break any
tools that access cgroupfs directly).
It is possibly to choose for each boot individually whether the unified
or the legacy hierarchy is used. nspawn will by default provide the
legacy hierarchy to containers if the host is using it, and the unified
otherwise. However it is possible to run containers with the unified
hierarchy on a legacy host and vice versa, by setting the
$UNIFIED_CGROUP_HIERARCHY environment variable for nspawn to 1 or 0,
respectively.
The unified hierarchy provides reliable cgroup empty notifications for
the first time, via inotify. To make use of this we maintain one
manager-wide inotify fd, and each cgroup to it.
This patch also removes cg_delete() which is unused now.
On kernel 4.2 only the "memory" controller is compatible with the
unified hierarchy, hence that's the only controller systemd exposes when
booted in unified heirarchy mode.
This introduces a new enum for enumerating supported controllers, plus a
related enum for the mask bits mapping to it. The core is changed to
make use of this everywhere.
This moves PID 1 into a new "init.scope" implicit scope unit in the root
slice. This is necessary since on the unified hierarchy cgroups may
either contain subgroups or processes but not both. PID 1 hence has to
move out of the root cgroup (strictly speaking the root cgroup is the
only one where processes and subgroups are still allowed, but in order
to support containers nicey, we move PID 1 into the new scope in all
cases.) This new unit is also used on legacy hierarchy setups. It's
actually pretty useful on all systems, as it can then be used to filter
journal messages coming from PID 1, and so on.
The root slice ("-.slice") is now implicitly created and started (and
does not require a unit file on disk anymore), since
that's where "init.scope" is located and the slice needs to be started
before the scope can.
To check whether we are in unified or legacy hierarchy mode we use
statfs() on /sys/fs/cgroup. If the .f_type field reports tmpfs we are in
legacy mode, if it reports cgroupfs we are in unified mode.
This patch set carefuly makes sure that cgls and cgtop continue to work
as desired.
When invoking nspawn as a service it will implicitly create two
subcgroups in the cgroup it is using, one to move the nspawn process
into, the other to move the actual container processes into. This is
done because of the requirement that cgroups may either contain
processes or other subgroups.
2015-09-01 20:22:36 +03:00
# include "unit.h"
2013-06-17 23:33:26 +04:00
static const UnitActiveState state_translation_table [ _SLICE_STATE_MAX ] = {
[ SLICE_DEAD ] = UNIT_INACTIVE ,
[ SLICE_ACTIVE ] = UNIT_ACTIVE
} ;
2016-02-19 00:51:23 +03:00
static void slice_init ( Unit * u ) {
assert ( u ) ;
assert ( u - > load_state = = UNIT_STUB ) ;
u - > ignore_on_isolate = true ;
}
2013-06-17 23:33:26 +04:00
static void slice_set_state ( Slice * t , SliceState state ) {
SliceState old_state ;
assert ( t ) ;
old_state = t - > state ;
t - > state = state ;
if ( state ! = old_state )
log_debug ( " %s changed %s -> %s " ,
UNIT ( t ) - > id ,
slice_state_to_string ( old_state ) ,
slice_state_to_string ( state ) ) ;
unit_notify ( UNIT ( t ) , state_translation_table [ old_state ] , state_translation_table [ state ] , true ) ;
}
2013-06-27 06:14:27 +04:00
static int slice_add_parent_slice ( Slice * s ) {
2013-06-17 23:33:26 +04:00
char * a , * dash ;
Unit * parent ;
2013-06-27 06:14:27 +04:00
int r ;
2013-06-17 23:33:26 +04:00
assert ( s ) ;
2013-06-27 06:14:27 +04:00
if ( UNIT_ISSET ( UNIT ( s ) - > slice ) )
2013-06-17 23:33:26 +04:00
return 0 ;
2013-06-27 06:14:27 +04:00
if ( unit_has_name ( UNIT ( s ) , SPECIAL_ROOT_SLICE ) )
2013-06-17 23:33:26 +04:00
return 0 ;
2013-06-27 06:14:27 +04:00
a = strdupa ( UNIT ( s ) - > id ) ;
dash = strrchr ( a , ' - ' ) ;
if ( dash )
strcpy ( dash , " .slice " ) ;
else
a = ( char * ) SPECIAL_ROOT_SLICE ;
2013-06-17 23:33:26 +04:00
r = manager_load_unit ( UNIT ( s ) - > manager , a , NULL , NULL , & parent ) ;
if ( r < 0 )
return r ;
unit_ref_set ( & UNIT ( s ) - > slice , parent ) ;
return 0 ;
}
static int slice_add_default_dependencies ( Slice * s ) {
int r ;
assert ( s ) ;
2015-11-11 22:42:39 +03:00
if ( ! UNIT ( s ) - > default_dependencies )
return 0 ;
2013-06-17 23:33:26 +04:00
/* Make sure slices are unloaded on shutdown */
2013-07-01 02:03:57 +04:00
r = unit_add_two_dependencies_by_name (
UNIT ( s ) ,
UNIT_BEFORE , UNIT_CONFLICTS ,
SPECIAL_SHUTDOWN_TARGET , NULL , true ) ;
2013-06-17 23:33:26 +04:00
if ( r < 0 )
return r ;
return 0 ;
}
static int slice_verify ( Slice * s ) {
2015-05-05 23:39:14 +03:00
_cleanup_free_ char * parent = NULL ;
int r ;
2013-06-17 23:33:26 +04:00
assert ( s ) ;
if ( UNIT ( s ) - > load_state ! = UNIT_LOADED )
return 0 ;
2015-05-05 23:39:14 +03:00
if ( ! slice_name_is_valid ( UNIT ( s ) - > id ) ) {
core,network: major per-object logging rework
This changes log_unit_info() (and friends) to take a real Unit* object
insted of just a unit name as parameter. The call will now prefix all
logged messages with the unit name, thus allowing the unit name to be
dropped from the various passed romat strings, simplifying invocations
drastically, and unifying log output across messages. Also, UNIT= vs.
USER_UNIT= is now derived from the Manager object attached to the Unit
object, instead of getpid(). This has the benefit of correcting the
field for --test runs.
Also contains a couple of other logging improvements:
- Drops a couple of strerror() invocations in favour of using %m.
- Not only .mount units now warn if a symlinks exist for the mount
point already, .automount units do that too, now.
- A few invocations of log_struct() that didn't actually pass any
additional structured data have been replaced by simpler invocations
of log_unit_info() and friends.
- For structured data a new LOG_UNIT_MESSAGE() macro has been added,
that works like LOG_MESSAGE() but prefixes the message with the unit
name. Similar, there's now LOG_LINK_MESSAGE() and
LOG_NETDEV_MESSAGE().
- For structured data new LOG_UNIT_ID(), LOG_LINK_INTERFACE(),
LOG_NETDEV_INTERFACE() macros have been added that generate the
necessary per object fields. The old log_unit_struct() call has been
removed in favour of these new macros used in raw log_struct()
invocations. In addition to removing one more function call this
allows generated structured log messages that contain two object
fields, as necessary for example for network interfaces that are
joined into another network interface, and whose messages shall be
indexed by both.
- The LOG_ERRNO() macro has been removed, in favour of
log_struct_errno(). The latter has the benefit of ensuring that %m in
format strings is properly resolved to the specified error number.
- A number of logging messages have been converted to use
log_unit_info() instead of log_info()
- The client code in sysv-generator no longer #includes core code from
src/core/.
- log_unit_full_errno() has been removed, log_unit_full() instead takes
an errno now, too.
- log_unit_info(), log_link_info(), log_netdev_info() and friends, now
avoid double evaluation of their parameters
2015-05-11 21:38:21 +03:00
log_unit_error ( UNIT ( s ) , " Slice name %s is not valid. Refusing. " , UNIT ( s ) - > id ) ;
2015-05-05 23:39:14 +03:00
return - EINVAL ;
}
r = slice_build_parent_slice ( UNIT ( s ) - > id , & parent ) ;
if ( r < 0 )
core,network: major per-object logging rework
This changes log_unit_info() (and friends) to take a real Unit* object
insted of just a unit name as parameter. The call will now prefix all
logged messages with the unit name, thus allowing the unit name to be
dropped from the various passed romat strings, simplifying invocations
drastically, and unifying log output across messages. Also, UNIT= vs.
USER_UNIT= is now derived from the Manager object attached to the Unit
object, instead of getpid(). This has the benefit of correcting the
field for --test runs.
Also contains a couple of other logging improvements:
- Drops a couple of strerror() invocations in favour of using %m.
- Not only .mount units now warn if a symlinks exist for the mount
point already, .automount units do that too, now.
- A few invocations of log_struct() that didn't actually pass any
additional structured data have been replaced by simpler invocations
of log_unit_info() and friends.
- For structured data a new LOG_UNIT_MESSAGE() macro has been added,
that works like LOG_MESSAGE() but prefixes the message with the unit
name. Similar, there's now LOG_LINK_MESSAGE() and
LOG_NETDEV_MESSAGE().
- For structured data new LOG_UNIT_ID(), LOG_LINK_INTERFACE(),
LOG_NETDEV_INTERFACE() macros have been added that generate the
necessary per object fields. The old log_unit_struct() call has been
removed in favour of these new macros used in raw log_struct()
invocations. In addition to removing one more function call this
allows generated structured log messages that contain two object
fields, as necessary for example for network interfaces that are
joined into another network interface, and whose messages shall be
indexed by both.
- The LOG_ERRNO() macro has been removed, in favour of
log_struct_errno(). The latter has the benefit of ensuring that %m in
format strings is properly resolved to the specified error number.
- A number of logging messages have been converted to use
log_unit_info() instead of log_info()
- The client code in sysv-generator no longer #includes core code from
src/core/.
- log_unit_full_errno() has been removed, log_unit_full() instead takes
an errno now, too.
- log_unit_info(), log_link_info(), log_netdev_info() and friends, now
avoid double evaluation of their parameters
2015-05-11 21:38:21 +03:00
return log_unit_error_errno ( UNIT ( s ) , r , " Failed to determine parent slice: %m " ) ;
2015-05-05 23:39:14 +03:00
if ( parent ? ! unit_has_name ( UNIT_DEREF ( UNIT ( s ) - > slice ) , parent ) : UNIT_ISSET ( UNIT ( s ) - > slice ) ) {
core,network: major per-object logging rework
This changes log_unit_info() (and friends) to take a real Unit* object
insted of just a unit name as parameter. The call will now prefix all
logged messages with the unit name, thus allowing the unit name to be
dropped from the various passed romat strings, simplifying invocations
drastically, and unifying log output across messages. Also, UNIT= vs.
USER_UNIT= is now derived from the Manager object attached to the Unit
object, instead of getpid(). This has the benefit of correcting the
field for --test runs.
Also contains a couple of other logging improvements:
- Drops a couple of strerror() invocations in favour of using %m.
- Not only .mount units now warn if a symlinks exist for the mount
point already, .automount units do that too, now.
- A few invocations of log_struct() that didn't actually pass any
additional structured data have been replaced by simpler invocations
of log_unit_info() and friends.
- For structured data a new LOG_UNIT_MESSAGE() macro has been added,
that works like LOG_MESSAGE() but prefixes the message with the unit
name. Similar, there's now LOG_LINK_MESSAGE() and
LOG_NETDEV_MESSAGE().
- For structured data new LOG_UNIT_ID(), LOG_LINK_INTERFACE(),
LOG_NETDEV_INTERFACE() macros have been added that generate the
necessary per object fields. The old log_unit_struct() call has been
removed in favour of these new macros used in raw log_struct()
invocations. In addition to removing one more function call this
allows generated structured log messages that contain two object
fields, as necessary for example for network interfaces that are
joined into another network interface, and whose messages shall be
indexed by both.
- The LOG_ERRNO() macro has been removed, in favour of
log_struct_errno(). The latter has the benefit of ensuring that %m in
format strings is properly resolved to the specified error number.
- A number of logging messages have been converted to use
log_unit_info() instead of log_info()
- The client code in sysv-generator no longer #includes core code from
src/core/.
- log_unit_full_errno() has been removed, log_unit_full() instead takes
an errno now, too.
- log_unit_info(), log_link_info(), log_netdev_info() and friends, now
avoid double evaluation of their parameters
2015-05-11 21:38:21 +03:00
log_unit_error ( UNIT ( s ) , " Located outside of parent slice. Refusing. " ) ;
2015-05-05 23:39:14 +03:00
return - EINVAL ;
2013-06-17 23:33:26 +04:00
}
return 0 ;
}
static int slice_load ( Unit * u ) {
Slice * s = SLICE ( u ) ;
int r ;
assert ( s ) ;
2016-04-07 16:43:59 +03:00
assert ( u - > load_state = = UNIT_STUB ) ;
2013-06-17 23:33:26 +04:00
2013-06-27 06:14:27 +04:00
r = unit_load_fragment_and_dropin_optional ( u ) ;
2013-06-17 23:33:26 +04:00
if ( r < 0 )
return r ;
/* This is a new unit? Then let's add in some extras */
if ( u - > load_state = = UNIT_LOADED ) {
2014-03-19 23:40:05 +04:00
r = unit_patch_contexts ( u ) ;
if ( r < 0 )
return r ;
2013-06-27 06:14:27 +04:00
r = slice_add_parent_slice ( s ) ;
2013-06-17 23:33:26 +04:00
if ( r < 0 )
return r ;
2015-11-11 22:42:39 +03:00
r = slice_add_default_dependencies ( s ) ;
if ( r < 0 )
return r ;
2013-06-17 23:33:26 +04:00
}
return slice_verify ( s ) ;
}
2015-04-24 16:27:19 +03:00
static int slice_coldplug ( Unit * u ) {
2013-06-17 23:33:26 +04:00
Slice * t = SLICE ( u ) ;
assert ( t ) ;
assert ( t - > state = = SLICE_DEAD ) ;
if ( t - > deserialized_state ! = t - > state )
slice_set_state ( t , t - > deserialized_state ) ;
return 0 ;
}
static void slice_dump ( Unit * u , FILE * f , const char * prefix ) {
Slice * t = SLICE ( u ) ;
assert ( t ) ;
assert ( f ) ;
fprintf ( f ,
" %sSlice State: %s \n " ,
prefix , slice_state_to_string ( t - > state ) ) ;
2013-06-27 06:14:27 +04:00
cgroup_context_dump ( & t - > cgroup_context , f , prefix ) ;
2013-06-17 23:33:26 +04:00
}
static int slice_start ( Unit * u ) {
Slice * t = SLICE ( u ) ;
assert ( t ) ;
assert ( t - > state = = SLICE_DEAD ) ;
2015-03-01 18:24:19 +03:00
( void ) unit_realize_cgroup ( u ) ;
( void ) unit_reset_cpu_usage ( u ) ;
2013-06-17 23:33:26 +04:00
slice_set_state ( t , SLICE_ACTIVE ) ;
2015-01-28 17:07:13 +03:00
return 1 ;
2013-06-17 23:33:26 +04:00
}
static int slice_stop ( Unit * u ) {
Slice * t = SLICE ( u ) ;
assert ( t ) ;
assert ( t - > state = = SLICE_ACTIVE ) ;
2013-06-27 06:14:27 +04:00
/* We do not need to destroy the cgroup explicitly,
* unit_notify ( ) will do that for us anyway . */
2013-06-17 23:33:26 +04:00
slice_set_state ( t , SLICE_DEAD ) ;
2015-01-28 17:07:13 +03:00
return 1 ;
2013-06-17 23:33:26 +04:00
}
2013-11-20 00:12:59 +04:00
static int slice_kill ( Unit * u , KillWho who , int signo , sd_bus_error * error ) {
2013-06-17 23:33:26 +04:00
return unit_kill_common ( u , who , signo , - 1 , - 1 , error ) ;
}
static int slice_serialize ( Unit * u , FILE * f , FDSet * fds ) {
Slice * s = SLICE ( u ) ;
assert ( s ) ;
assert ( f ) ;
assert ( fds ) ;
unit_serialize_item ( u , f , " state " , slice_state_to_string ( s - > state ) ) ;
return 0 ;
}
static int slice_deserialize_item ( Unit * u , const char * key , const char * value , FDSet * fds ) {
Slice * s = SLICE ( u ) ;
assert ( u ) ;
assert ( key ) ;
assert ( value ) ;
assert ( fds ) ;
if ( streq ( key , " state " ) ) {
SliceState state ;
state = slice_state_from_string ( value ) ;
if ( state < 0 )
log_debug ( " Failed to parse state value %s " , value ) ;
else
s - > deserialized_state = state ;
} else
log_debug ( " Unknown serialization key '%s' " , key ) ;
return 0 ;
}
_pure_ static UnitActiveState slice_active_state ( Unit * u ) {
assert ( u ) ;
return state_translation_table [ SLICE ( u ) - > state ] ;
}
_pure_ static const char * slice_sub_state_to_string ( Unit * u ) {
assert ( u ) ;
return slice_state_to_string ( SLICE ( u ) - > state ) ;
}
2015-11-10 22:36:37 +03:00
static void slice_enumerate ( Manager * m ) {
core: unified cgroup hierarchy support
This patch set adds full support the new unified cgroup hierarchy logic
of modern kernels.
A new kernel command line option "systemd.unified_cgroup_hierarchy=1" is
added. If specified the unified hierarchy is mounted to /sys/fs/cgroup
instead of a tmpfs. No further hierarchies are mounted. The kernel
command line option defaults to off. We can turn it on by default as
soon as the kernel's APIs regarding this are stabilized (but even then
downstream distros might want to turn this off, as this will break any
tools that access cgroupfs directly).
It is possibly to choose for each boot individually whether the unified
or the legacy hierarchy is used. nspawn will by default provide the
legacy hierarchy to containers if the host is using it, and the unified
otherwise. However it is possible to run containers with the unified
hierarchy on a legacy host and vice versa, by setting the
$UNIFIED_CGROUP_HIERARCHY environment variable for nspawn to 1 or 0,
respectively.
The unified hierarchy provides reliable cgroup empty notifications for
the first time, via inotify. To make use of this we maintain one
manager-wide inotify fd, and each cgroup to it.
This patch also removes cg_delete() which is unused now.
On kernel 4.2 only the "memory" controller is compatible with the
unified hierarchy, hence that's the only controller systemd exposes when
booted in unified heirarchy mode.
This introduces a new enum for enumerating supported controllers, plus a
related enum for the mask bits mapping to it. The core is changed to
make use of this everywhere.
This moves PID 1 into a new "init.scope" implicit scope unit in the root
slice. This is necessary since on the unified hierarchy cgroups may
either contain subgroups or processes but not both. PID 1 hence has to
move out of the root cgroup (strictly speaking the root cgroup is the
only one where processes and subgroups are still allowed, but in order
to support containers nicey, we move PID 1 into the new scope in all
cases.) This new unit is also used on legacy hierarchy setups. It's
actually pretty useful on all systems, as it can then be used to filter
journal messages coming from PID 1, and so on.
The root slice ("-.slice") is now implicitly created and started (and
does not require a unit file on disk anymore), since
that's where "init.scope" is located and the slice needs to be started
before the scope can.
To check whether we are in unified or legacy hierarchy mode we use
statfs() on /sys/fs/cgroup. If the .f_type field reports tmpfs we are in
legacy mode, if it reports cgroupfs we are in unified mode.
This patch set carefuly makes sure that cgls and cgtop continue to work
as desired.
When invoking nspawn as a service it will implicitly create two
subcgroups in the cgroup it is using, one to move the nspawn process
into, the other to move the actual container processes into. This is
done because of the requirement that cgroups may either contain
processes or other subgroups.
2015-09-01 20:22:36 +03:00
Unit * u ;
int r ;
assert ( m ) ;
u = manager_get_unit ( m , SPECIAL_ROOT_SLICE ) ;
if ( ! u ) {
u = unit_new ( m , sizeof ( Slice ) ) ;
2015-11-10 22:36:37 +03:00
if ( ! u ) {
log_oom ( ) ;
return ;
}
core: unified cgroup hierarchy support
This patch set adds full support the new unified cgroup hierarchy logic
of modern kernels.
A new kernel command line option "systemd.unified_cgroup_hierarchy=1" is
added. If specified the unified hierarchy is mounted to /sys/fs/cgroup
instead of a tmpfs. No further hierarchies are mounted. The kernel
command line option defaults to off. We can turn it on by default as
soon as the kernel's APIs regarding this are stabilized (but even then
downstream distros might want to turn this off, as this will break any
tools that access cgroupfs directly).
It is possibly to choose for each boot individually whether the unified
or the legacy hierarchy is used. nspawn will by default provide the
legacy hierarchy to containers if the host is using it, and the unified
otherwise. However it is possible to run containers with the unified
hierarchy on a legacy host and vice versa, by setting the
$UNIFIED_CGROUP_HIERARCHY environment variable for nspawn to 1 or 0,
respectively.
The unified hierarchy provides reliable cgroup empty notifications for
the first time, via inotify. To make use of this we maintain one
manager-wide inotify fd, and each cgroup to it.
This patch also removes cg_delete() which is unused now.
On kernel 4.2 only the "memory" controller is compatible with the
unified hierarchy, hence that's the only controller systemd exposes when
booted in unified heirarchy mode.
This introduces a new enum for enumerating supported controllers, plus a
related enum for the mask bits mapping to it. The core is changed to
make use of this everywhere.
This moves PID 1 into a new "init.scope" implicit scope unit in the root
slice. This is necessary since on the unified hierarchy cgroups may
either contain subgroups or processes but not both. PID 1 hence has to
move out of the root cgroup (strictly speaking the root cgroup is the
only one where processes and subgroups are still allowed, but in order
to support containers nicey, we move PID 1 into the new scope in all
cases.) This new unit is also used on legacy hierarchy setups. It's
actually pretty useful on all systems, as it can then be used to filter
journal messages coming from PID 1, and so on.
The root slice ("-.slice") is now implicitly created and started (and
does not require a unit file on disk anymore), since
that's where "init.scope" is located and the slice needs to be started
before the scope can.
To check whether we are in unified or legacy hierarchy mode we use
statfs() on /sys/fs/cgroup. If the .f_type field reports tmpfs we are in
legacy mode, if it reports cgroupfs we are in unified mode.
This patch set carefuly makes sure that cgls and cgtop continue to work
as desired.
When invoking nspawn as a service it will implicitly create two
subcgroups in the cgroup it is using, one to move the nspawn process
into, the other to move the actual container processes into. This is
done because of the requirement that cgroups may either contain
processes or other subgroups.
2015-09-01 20:22:36 +03:00
r = unit_add_name ( u , SPECIAL_ROOT_SLICE ) ;
if ( r < 0 ) {
unit_free ( u ) ;
2015-11-10 22:36:37 +03:00
log_error_errno ( r , " Failed to add -.slice name " ) ;
return ;
core: unified cgroup hierarchy support
This patch set adds full support the new unified cgroup hierarchy logic
of modern kernels.
A new kernel command line option "systemd.unified_cgroup_hierarchy=1" is
added. If specified the unified hierarchy is mounted to /sys/fs/cgroup
instead of a tmpfs. No further hierarchies are mounted. The kernel
command line option defaults to off. We can turn it on by default as
soon as the kernel's APIs regarding this are stabilized (but even then
downstream distros might want to turn this off, as this will break any
tools that access cgroupfs directly).
It is possibly to choose for each boot individually whether the unified
or the legacy hierarchy is used. nspawn will by default provide the
legacy hierarchy to containers if the host is using it, and the unified
otherwise. However it is possible to run containers with the unified
hierarchy on a legacy host and vice versa, by setting the
$UNIFIED_CGROUP_HIERARCHY environment variable for nspawn to 1 or 0,
respectively.
The unified hierarchy provides reliable cgroup empty notifications for
the first time, via inotify. To make use of this we maintain one
manager-wide inotify fd, and each cgroup to it.
This patch also removes cg_delete() which is unused now.
On kernel 4.2 only the "memory" controller is compatible with the
unified hierarchy, hence that's the only controller systemd exposes when
booted in unified heirarchy mode.
This introduces a new enum for enumerating supported controllers, plus a
related enum for the mask bits mapping to it. The core is changed to
make use of this everywhere.
This moves PID 1 into a new "init.scope" implicit scope unit in the root
slice. This is necessary since on the unified hierarchy cgroups may
either contain subgroups or processes but not both. PID 1 hence has to
move out of the root cgroup (strictly speaking the root cgroup is the
only one where processes and subgroups are still allowed, but in order
to support containers nicey, we move PID 1 into the new scope in all
cases.) This new unit is also used on legacy hierarchy setups. It's
actually pretty useful on all systems, as it can then be used to filter
journal messages coming from PID 1, and so on.
The root slice ("-.slice") is now implicitly created and started (and
does not require a unit file on disk anymore), since
that's where "init.scope" is located and the slice needs to be started
before the scope can.
To check whether we are in unified or legacy hierarchy mode we use
statfs() on /sys/fs/cgroup. If the .f_type field reports tmpfs we are in
legacy mode, if it reports cgroupfs we are in unified mode.
This patch set carefuly makes sure that cgls and cgtop continue to work
as desired.
When invoking nspawn as a service it will implicitly create two
subcgroups in the cgroup it is using, one to move the nspawn process
into, the other to move the actual container processes into. This is
done because of the requirement that cgroups may either contain
processes or other subgroups.
2015-09-01 20:22:36 +03:00
}
}
u - > default_dependencies = false ;
u - > no_gc = true ;
2015-10-09 18:18:18 +03:00
u - > ignore_on_isolate = true ;
u - > refuse_manual_start = true ;
u - > refuse_manual_stop = true ;
core: unified cgroup hierarchy support
This patch set adds full support the new unified cgroup hierarchy logic
of modern kernels.
A new kernel command line option "systemd.unified_cgroup_hierarchy=1" is
added. If specified the unified hierarchy is mounted to /sys/fs/cgroup
instead of a tmpfs. No further hierarchies are mounted. The kernel
command line option defaults to off. We can turn it on by default as
soon as the kernel's APIs regarding this are stabilized (but even then
downstream distros might want to turn this off, as this will break any
tools that access cgroupfs directly).
It is possibly to choose for each boot individually whether the unified
or the legacy hierarchy is used. nspawn will by default provide the
legacy hierarchy to containers if the host is using it, and the unified
otherwise. However it is possible to run containers with the unified
hierarchy on a legacy host and vice versa, by setting the
$UNIFIED_CGROUP_HIERARCHY environment variable for nspawn to 1 or 0,
respectively.
The unified hierarchy provides reliable cgroup empty notifications for
the first time, via inotify. To make use of this we maintain one
manager-wide inotify fd, and each cgroup to it.
This patch also removes cg_delete() which is unused now.
On kernel 4.2 only the "memory" controller is compatible with the
unified hierarchy, hence that's the only controller systemd exposes when
booted in unified heirarchy mode.
This introduces a new enum for enumerating supported controllers, plus a
related enum for the mask bits mapping to it. The core is changed to
make use of this everywhere.
This moves PID 1 into a new "init.scope" implicit scope unit in the root
slice. This is necessary since on the unified hierarchy cgroups may
either contain subgroups or processes but not both. PID 1 hence has to
move out of the root cgroup (strictly speaking the root cgroup is the
only one where processes and subgroups are still allowed, but in order
to support containers nicey, we move PID 1 into the new scope in all
cases.) This new unit is also used on legacy hierarchy setups. It's
actually pretty useful on all systems, as it can then be used to filter
journal messages coming from PID 1, and so on.
The root slice ("-.slice") is now implicitly created and started (and
does not require a unit file on disk anymore), since
that's where "init.scope" is located and the slice needs to be started
before the scope can.
To check whether we are in unified or legacy hierarchy mode we use
statfs() on /sys/fs/cgroup. If the .f_type field reports tmpfs we are in
legacy mode, if it reports cgroupfs we are in unified mode.
This patch set carefuly makes sure that cgls and cgtop continue to work
as desired.
When invoking nspawn as a service it will implicitly create two
subcgroups in the cgroup it is using, one to move the nspawn process
into, the other to move the actual container processes into. This is
done because of the requirement that cgroups may either contain
processes or other subgroups.
2015-09-01 20:22:36 +03:00
SLICE ( u ) - > deserialized_state = SLICE_ACTIVE ;
if ( ! u - > description )
u - > description = strdup ( " Root Slice " ) ;
if ( ! u - > documentation )
( void ) strv_extend ( & u - > documentation , " man:systemd.special(7) " ) ;
unit_add_to_load_queue ( u ) ;
unit_add_to_dbus_queue ( u ) ;
}
2013-06-17 23:33:26 +04:00
const UnitVTable slice_vtable = {
. object_size = sizeof ( Slice ) ,
2013-11-20 00:12:59 +04:00
. cgroup_context_offset = offsetof ( Slice , cgroup_context ) ,
2013-06-17 23:33:26 +04:00
. sections =
" Unit \0 "
" Slice \0 "
" Install \0 " ,
2013-06-27 06:14:27 +04:00
. private_section = " Slice " ,
2015-11-13 20:46:50 +03:00
. can_transient = true ,
2013-06-17 23:33:26 +04:00
2016-02-19 00:51:23 +03:00
. init = slice_init ,
2013-06-17 23:33:26 +04:00
. load = slice_load ,
2013-06-27 06:14:27 +04:00
2013-06-17 23:33:26 +04:00
. coldplug = slice_coldplug ,
. dump = slice_dump ,
. start = slice_start ,
. stop = slice_stop ,
. kill = slice_kill ,
. serialize = slice_serialize ,
. deserialize_item = slice_deserialize_item ,
. active_state = slice_active_state ,
. sub_state_to_string = slice_sub_state_to_string ,
2013-11-20 00:12:59 +04:00
. bus_vtable = bus_slice_vtable ,
2013-06-27 23:14:56 +04:00
. bus_set_property = bus_slice_set_property ,
. bus_commit_properties = bus_slice_commit_properties ,
2013-06-17 23:33:26 +04:00
core: unified cgroup hierarchy support
This patch set adds full support the new unified cgroup hierarchy logic
of modern kernels.
A new kernel command line option "systemd.unified_cgroup_hierarchy=1" is
added. If specified the unified hierarchy is mounted to /sys/fs/cgroup
instead of a tmpfs. No further hierarchies are mounted. The kernel
command line option defaults to off. We can turn it on by default as
soon as the kernel's APIs regarding this are stabilized (but even then
downstream distros might want to turn this off, as this will break any
tools that access cgroupfs directly).
It is possibly to choose for each boot individually whether the unified
or the legacy hierarchy is used. nspawn will by default provide the
legacy hierarchy to containers if the host is using it, and the unified
otherwise. However it is possible to run containers with the unified
hierarchy on a legacy host and vice versa, by setting the
$UNIFIED_CGROUP_HIERARCHY environment variable for nspawn to 1 or 0,
respectively.
The unified hierarchy provides reliable cgroup empty notifications for
the first time, via inotify. To make use of this we maintain one
manager-wide inotify fd, and each cgroup to it.
This patch also removes cg_delete() which is unused now.
On kernel 4.2 only the "memory" controller is compatible with the
unified hierarchy, hence that's the only controller systemd exposes when
booted in unified heirarchy mode.
This introduces a new enum for enumerating supported controllers, plus a
related enum for the mask bits mapping to it. The core is changed to
make use of this everywhere.
This moves PID 1 into a new "init.scope" implicit scope unit in the root
slice. This is necessary since on the unified hierarchy cgroups may
either contain subgroups or processes but not both. PID 1 hence has to
move out of the root cgroup (strictly speaking the root cgroup is the
only one where processes and subgroups are still allowed, but in order
to support containers nicey, we move PID 1 into the new scope in all
cases.) This new unit is also used on legacy hierarchy setups. It's
actually pretty useful on all systems, as it can then be used to filter
journal messages coming from PID 1, and so on.
The root slice ("-.slice") is now implicitly created and started (and
does not require a unit file on disk anymore), since
that's where "init.scope" is located and the slice needs to be started
before the scope can.
To check whether we are in unified or legacy hierarchy mode we use
statfs() on /sys/fs/cgroup. If the .f_type field reports tmpfs we are in
legacy mode, if it reports cgroupfs we are in unified mode.
This patch set carefuly makes sure that cgls and cgtop continue to work
as desired.
When invoking nspawn as a service it will implicitly create two
subcgroups in the cgroup it is using, one to move the nspawn process
into, the other to move the actual container processes into. This is
done because of the requirement that cgroups may either contain
processes or other subgroups.
2015-09-01 20:22:36 +03:00
. enumerate = slice_enumerate ,
2013-06-17 23:33:26 +04:00
. status_message_formats = {
. finished_start_job = {
2013-06-27 06:14:27 +04:00
[ JOB_DONE ] = " Created slice %s. " ,
2013-06-17 23:33:26 +04:00
} ,
. finished_stop_job = {
2013-06-27 06:14:27 +04:00
[ JOB_DONE ] = " Removed slice %s. " ,
2013-06-17 23:33:26 +04:00
} ,
} ,
} ;