3
J s z*SchedulerUtils.__init__..c s s | ]}t t||fV qd S )N)r( r) )r* r, r r r r- L s )dict_dict_schedcfg2schedconstitems_dict_schedcfg2numvalues_dict_num2schedconst)r
r r r r H s zSchedulerUtils.__init__c C s | j j|S )N)r1 get)r
str_schedulerr r r sched_cfg_to_numN s z SchedulerUtils.sched_cfg_to_numc C s | j j|S )N)r3 r4 )r
r r r r sched_num_to_constR s z!SchedulerUtils.sched_num_to_constc C s
t j|S )N)r) sched_getscheduler)r
pidr r r
get_schedulerU s z SchedulerUtils.get_schedulerc C s t j||t j| d S )N)r) sched_setscheduler sched_param)r
r9 schedprior r r
set_schedulerX s z SchedulerUtils.set_schedulerc C s
t j|S )N)r) sched_getaffinity)r
r9 r r r get_affinity[ s zSchedulerUtils.get_affinityc C s t j|| d S )N)r) sched_setaffinity)r
r9 r r r r set_affinity^ s zSchedulerUtils.set_affinityc C s t j|jS )N)r) sched_getparamsched_priority)r
r9 r r r get_prioritya s zSchedulerUtils.get_priorityc C s
t j|S )N)r) sched_get_priority_min)r
r= r r r get_priority_mind s z SchedulerUtils.get_priority_minc C s
t j|S )N)r) sched_get_priority_max)r
r= r r r get_priority_maxg s z SchedulerUtils.get_priority_maxN)r r r __doc__r/ r r6 r7 r: r? rA rC rF rH rJ r r r r r ; s r c @ sP e Zd ZdZdd Zdd Zdd Zdd Zd
d Zd d
Z dd Z
dd Z dS )SchedulerUtilsSchedutilszE
Class encapsulating scheduler implementation in schedutils module
c C s8 t dd | jj D | _t dd | jj D | _d S )Nc s s | ]\}}|t t|fV qd S )N)r(
schedutils)r* r+ r, r r r r- p s z4SchedulerUtilsSchedutils.__init__..c s s | ]}t t||fV qd S )N)r( rM )r* r, r r r r- r s )r. r/ r0 r1 r2 r3 )r
r r r r n s z!SchedulerUtilsSchedutils.__init__c C s
t j|S )N)rM r: )r
r9 r r r r: t s z&SchedulerUtilsSchedutils.get_schedulerc C s t j||| d S )N)rM r? )r
r9 r= r> r r r r? w s z&SchedulerUtilsSchedutils.set_schedulerc C s
t j|S )N)rM rA )r
r9 r r r rA z s z%SchedulerUtilsSchedutils.get_affinityc C s t j|| d S )N)rM rC )r
r9 r r r r rC } s z%SchedulerUtilsSchedutils.set_affinityc C s
t j|S )N)rM rF )r
r9 r r r rF s z%SchedulerUtilsSchedutils.get_priorityc C s
t j|S )N)rM rH )r
r= r r r rH s z)SchedulerUtilsSchedutils.get_priority_minc C s
t j|S )N)rM rJ )r
r= r r r rJ s z)SchedulerUtilsSchedutils.get_priority_maxN) r r r rK r r: r? rA rC rF rH rJ r r r r rL j s rL c s e Zd ZdZ fddZdd Zdd Zdd Zed
d Z d d
Z
dd Z dd Z dd Z
dd Zdd Zdd Zdd Zd d Zd d Zdd!d"Zd#d$ Zd%d& Zd'd( Zdd)d*Zd+d, Zd-d. Zd/d0 Zd1d2 Z d3d4 Z d5d6 Z d7d8 Z d9d: Z d;d< Z!d=d> Z"dd?d@Z#dAdB Z$dCdD Z% fdEdFZ&dGdH Z'dIdJ Z(dKdL Z)e*j+f fdMdN Z,dOdP Z-dQdR Z. fdSdTZ/dUdV Z0dWdX Z1dYdZ Z2e3d[d d\d]d^ Z4e3d_d d\d`da Z5e3dbd d\dcdd Z6e3ded d\dfdg Z7e3dhd d\didj Z8e3dkd d\dldm Z9dndo Z:dpdq Z;drds Zdxdy Z?dzd{ Z@d|d} ZAd~d ZBdd ZCdd ZDdd ZEdd ZFe3dd dddd ZGdd ZHdd ZIdddZJeKddd ZLeMddd ZNeKddd ZOeMddd ZPeKddd ZQeMddd ZReKddd ZSeMddd ZTeKddd ZUeMddd ZVeKddd ZWeMddd ZXeKddd ZYeMddd ZZeKddd Z[eMddd Z\eKddd Z]eMddd Z^eKddd Z_eMdddĄ Z` ZaS )SchedulerPlugina]-
`scheduler`::
Allows tuning of scheduling priorities, process/thread/IRQ
affinities, and CPU isolation.
+
To prevent processes/threads/IRQs from using certain CPUs, use
the [option]`isolated_cores` option. It changes process/thread
affinities, IRQs affinities and it sets `default_smp_affinity`
for IRQs. The CPU affinity mask is adjusted for all processes and
threads matching [option]`ps_whitelist` option subject to success
of the `sched_setaffinity()` system call. The default setting of
the [option]`ps_whitelist` regular expression is `.*` to match all
processes and thread names. To exclude certain processes and threads
use [option]`ps_blacklist` option. The value of this option is also
interpreted as a regular expression and process/thread names (`ps -eo
cmd`) are matched against that expression. Profile rollback allows
all matching processes and threads to run on all CPUs and restores
the IRQ settings prior to the profile application.
+
Multiple regular expressions for [option]`ps_whitelist`
and [option]`ps_blacklist` options are allowed and separated by
`;`. Quoted semicolon `\;` is taken literally.
+
.Isolate CPUs 2-4
====
----
[scheduler]
isolated_cores=2-4
ps_blacklist=.*pmd.*;.*PMD.*;^DPDK;.*qemu-kvm.*
----
Isolate CPUs 2-4 while ignoring processes and threads matching
`ps_blacklist` regular expressions.
====
The [option]`irq_process` option controls whether the scheduler plugin
applies the `isolated_cores` parameter to IRQ affinities. The default
value is `true`, which means that the scheduler plugin will move all
possible IRQs away from the isolated cores. When `irq_process` is set
to `false`, the plugin will not change any IRQ affinities.
====
The [option]`default_irq_smp_affinity` option controls the values
*TuneD* writes to `/proc/irq/default_smp_affinity`. The file specifies
default affinity mask that applies to all non-active IRQs. Once an
IRQ is allocated/activated its affinity bitmask will be set to the
default mask.
+
The following values are supported:
+
--
`calc`::
Content of `/proc/irq/default_smp_affinity` will be calculated
from the `isolated_cores` parameter. Non-isolated cores
are calculated as an inversion of the `isolated_cores`. Then
the intersection of the non-isolated cores and the previous
content of `/proc/irq/default_smp_affinity` is written to
`/proc/irq/default_smp_affinity`. If the intersection is
an empty set, then just the non-isolated cores are written to
`/proc/irq/default_smp_affinity`. This behavior is the default if
the parameter `default_irq_smp_affinity` is omitted.
`ignore`::
*TuneD* will not touch `/proc/irq/default_smp_affinity`.
explicit cpulist::
The cpulist (such as 1,3-4) is unpacked and written directly to
`/proc/irq/default_smp_affinity`.
--
+
.An explicit CPU list to set the default IRQ smp affinity to CPUs 0 and 2
====
----
[scheduler]
isolated_cores=1,3
default_irq_smp_affinity=0,2
----
====
To adjust scheduling policy, priority and affinity for a group of
processes/threads, use the following syntax.
+
[subs="+quotes,+macros"]
----
group.__groupname__=__rule_prio__:__sched__:__prio__:__affinity__:__regex__
----
+
where `__rule_prio__` defines internal *TuneD* priority of the
rule. Rules are sorted based on priority. This is needed for
inheritence to be able to reorder previously defined rules. Equal
`__rule_prio__` rules should be processed in the order they were
defined. However, this is Python interpreter dependant. To disable
an inherited rule for `__groupname__` use:
+
[subs="+quotes,+macros"]
----
group.__groupname__=
----
+
`__sched__` must be one of:
*`f`* for FIFO,
*`b`* for batch,
*`r`* for round robin,
*`o`* for other,
*`*`* do not change.
+
`__affinity__` is CPU affinity in hexadecimal. Use `*` for no change.
+
`__prio__` scheduling priority (see `chrt -m`).
+
`__regex__` is Python regular expression. It is matched against the output of
+
[subs="+quotes,+macros"]
----
ps -eo cmd
----
+
Any given process name may match more than one group. In such a case,
the priority and scheduling policy are taken from the last matching
`__regex__`.
+
.Setting scheduling policy and priorities to kernel threads and watchdog
====
----
[scheduler]
group.kthreads=0:*:1:*:\[.*\]$
group.watchdog=0:f:99:*:\[watchdog.*\]
----
====
+
The scheduler plug-in uses perf event loop to catch newly created
processes. By default it listens to `perf.RECORD_COMM` and
`perf.RECORD_EXIT` events. By setting [option]`perf_process_fork`
option to `true`, `perf.RECORD_FORK` events will be also listened
to. In other words, child processes created by the `fork()` system
call will be processed. Since child processes inherit CPU affinity
from their parents, the scheduler plug-in usually does not need to
explicitly process these events. As processing perf events can
pose a significant CPU overhead, the [option]`perf_process_fork`
option parameter is set to `false` by default. Due to this, child
processes are not processed by the scheduler plug-in.
+
The CPU overhead of the scheduler plugin can be mitigated by using
the scheduler [option]`runtime` option and setting it to `0`. This
will completely disable the dynamic scheduler functionality and the
perf events will not be monitored and acted upon. The disadvantage
ot this approach is the procees/thread tuning will be done only at
profile application.
+
.Disabling the scheduler dynamic functionality
====
----
[scheduler]
runtime=0
isolated_cores=1,3
----
====
+
NOTE: For perf events, memory mapped buffer is used. Under heavy load
the buffer may overflow. In such cases the `scheduler` plug-in
may start missing events and failing to process some newly created
processes. Increasing the buffer size may help. The buffer size can
be set with the [option]`perf_mmap_pages` option. The value of this
parameter has to expressed in powers of 2. If it is not the power
of 2, the nearest higher power of 2 value is calculated from it
and this calculated value used. If the [option]`perf_mmap_pages`
option is omitted, the default kernel value is used.
+
The scheduler plug-in supports process/thread confinement using
cgroups v1.
+
[option]`cgroup_mount_point` option specifies the path to mount the
cgroup filesystem or where *TuneD* expects it to be mounted. If unset,
`/sys/fs/cgroup/cpuset` is expected.
+
If [option]`cgroup_groups_init` option is set to `1` *TuneD*
will create (and remove) all cgroups defined with the `cgroup*`
options. This is the default behavior. If it is set to `0` the
cgroups need to be preset by other means.
+
If [option]`cgroup_mount_point_init` option is set to `1`,
*TuneD* will create (and remove) the cgroup mountpoint. It implies
`cgroup_groups_init = 1`. If set to `0` the cgroups mount point
needs to be preset by other means. This is the default behavior.
+
The [option]`cgroup_for_isolated_cores` option is the cgroup
name used for the [option]`isolated_cores` option functionality. For
example, if a system has 4 CPUs, `isolated_cores=1` means that all
processes/threads will be moved to CPUs 0,2-3.
The scheduler plug-in will isolate the specified core by writing
the calculated CPU affinity to the `cpuset.cpus` control file of
the specified cgroup and move all the matching processes/threads to
this group. If this option is unset, classic cpuset affinity using
`sched_setaffinity()` will be used.
+
[option]`cgroup.__cgroup_name__` option defines affinities for
arbitrary cgroups. Even hierarchic cgroups can be used, but the
hieararchy needs to be specified in the correct order. Also *TuneD*
does not do any sanity checks here, with the exception that it forces
the cgroup to be under [option]`cgroup_mount_point`.
+
The syntax of the scheduler option starting with `group.` has been
augmented to use `cgroup.__cgroup_name__` instead of the hexadecimal
`__affinity__`. The matching processes will be moved to the cgroup
`__cgroup_name__`. It is also possible to use cgroups which have
not been defined by the [option]`cgroup.` option as described above,
i.e. cgroups not managed by *TuneD*.
+
All cgroup names are sanitized by replacing all all dots (`.`) with
slashes (`/`). This is to prevent the plug-in from writing outside
[option]`cgroup_mount_point`.
+
.Using cgroups v1 with the scheduler plug-in
====
----
[scheduler]
cgroup_mount_point=/sys/fs/cgroup/cpuset
cgroup_mount_point_init=1
cgroup_groups_init=1
cgroup_for_isolated_cores=group
cgroup.group1=2
cgroup.group2=0,2
group.ksoftirqd=0:f:2:cgroup.group1:ksoftirqd.*
ps_blacklist=ksoftirqd.*;rcuc.*;rcub.*;ktimersoftd.*
isolated_cores=1
----
Cgroup `group1` has the affinity set to CPU 2 and the cgroup `group2`
to CPUs 0,2. Given a 4 CPU setup, the [option]`isolated_cores=1`
option causes all processes/threads to be moved to CPU
cores 0,2-3. Processes/threads that are blacklisted by the
[option]`ps_blacklist` regular expression will not be moved.
The scheduler plug-in will isolate the specified core by writing the
CPU affinity 0,2-3 to the `cpuset.cpus` control file of the `group`
and move all the matching processes/threads to this cgroup.
====
Option [option]`cgroup_ps_blacklist` allows excluding processes
which belong to the blacklisted cgroups. The regular expression specified
by this option is matched against cgroup hierarchies from
`/proc/PID/cgroups`. Cgroups v1 hierarchies from `/proc/PID/cgroups`
are separated by commas ',' prior to regular expression matching. The
following is an example of content against which the regular expression
is matched against: `10:hugetlb:/,9:perf_event:/,8:blkio:/`
+
Multiple regular expressions can be separated by semicolon ';'. The
semicolon represents a logical 'or' operator.
+
.Cgroup-based exclusion of processes from the scheduler
====
----
[scheduler]
isolated_cores=1
cgroup_ps_blacklist=:/daemons\b
----
The scheduler plug-in will move all processes away from core 1 except processes which
belong to cgroup '/daemons'. The '\b' is a regular expression
metacharacter that matches a word boundary.
----
[scheduler]
isolated_cores=1
cgroup_ps_blacklist=\b8:blkio:
----
The scheduler plug-in will exclude all processes which belong to a cgroup
with hierarchy-ID 8 and controller-list blkio.
====
Recent kernels moved some `sched_` and `numa_balancing_` kernel run-time
parameters from `/proc/sys/kernel`, managed by the `sysctl` utility, to
`debugfs`, typically mounted under `/sys/kernel/debug`. TuneD provides an
abstraction mechanism for the following parameters via the scheduler plug-in:
[option]`sched_min_granularity_ns`, [option]`sched_latency_ns`,
[option]`sched_wakeup_granularity_ns`, [option]`sched_tunable_scaling`,
[option]`sched_migration_cost_ns`, [option]`sched_nr_migrate`,
[option]`numa_balancing_scan_delay_ms`,
[option]`numa_balancing_scan_period_min_ms`,
[option]`numa_balancing_scan_period_max_ms` and
[option]`numa_balancing_scan_size_mb`.
Based on the kernel used, TuneD will write the specified value to the correct
location.
+
.Set tasks' "cache hot" value for migration decisions.
====
----
[scheduler]
sched_migration_cost_ns=500000
----
On the old kernels, this is equivalent to:
----
[sysctl]
kernel.sched_migration_cost_ns=500000
----
that is, value `500000` will be written to `/proc/sys/kernel/sched_migration_cost_ns`.
However, on more recent kernels, the value `500000` will be written to
`/sys/kernel/debug/sched/migration_cost_ns`.
====
c s t t| j|||||||| d| _tj| _ttj| _ |d k rh|j
tj tj| _t|j tj
tj| _ t | _d | _i | _d| _d| _d| _tj | _| jdd| _d| _| jdd| _d | _ y t | _ W n t k
r t | _ Y nX d S )NTz.* r )Z command_nameirq)!superrN r Z_has_dynamic_optionsconstsZCFG_DEF_DAEMON_daemonintZCFG_DEF_SLEEP_INTERVAL_sleep_intervalget_boolZ
CFG_DAEMONr4 ZCFG_SLEEP_INTERVALr r _secure_boot_hint_sched_knob_paths_cache
_ps_whitelist
_ps_blacklist_cgroup_ps_blacklist_reperfZcpu_map_cpusZ _storage_key_scheduler_storage_key _irq_process_irq_storage_key_evlistr _scheduler_utilsAttributeErrorrL ) r
Zmonitor_repositoryZstorage_factoryZhardware_inventoryZdevice_matcherZdevice_matcher_udevZplugin_instance_factoryZ
global_cfg variables) __class__r r r s0
zSchedulerPlugin.__init__c C sT |d kr d S y t |}W n tk
r, dS X |dkr:dS t dtjtj|d S )Nr )rT
ValueErrormathZceillog)r
Z
mmap_pagesZmpr r r _calc_mmap_pages s z SchedulerPlugin._calc_mmap_pagesc s d |_ d|_d|_d|_ jj ji _t jdkr^t j
d j i _ jj j t
_d _d _d _tj fdd|jj D _|j|_ jj|jd } j|}|dkrt jd| d }|d k rt ||krt j
d ||f x(|jD ] } jj|j| |j|< qW j j |jjd
d d krHd|_t j |_! j"r |jr yt#j$ |_%t#j&t#j't#j(d d ddd d t#j)t#j*B d
}|j+ j,|j%d t#j- j,|j%|_ |j j.| |d kr|j j/ n|j j/|d W n d|_Y nX d S )NFTr z0recovering scheduling settings from previous runc sJ g | ]B\}}|d d dkrt |dkr j|dd jj|fqS )N zcgroup.)len_sanitize_cgroup_path
_variablesexpand)r* optionr )r
r r
s z2SchedulerPlugin._instance_init..perf_mmap_pageszKInvalid 'perf_mmap_pages' value specified: '%s', using default kernel valuezL'perf_mmap_pages' value has to be power of two, specified: '%s', using: '%d'Zruntimer 0) typeconfigZtaskcommmmapZfreqZ
wakeup_eventsZ watermarkZ sample_type)Zcpusthreads)Zpages)0ra Z_has_dynamic_tuningZ_has_static_tuning_runtime_tuning_storager4 r^ _scheduler_originalrl ri info_restore_ps_affinityunsetr. _cgroups_original_affinityr _cgroup_affinity_initialized_cgroup collections OrderedDictoptionsr0 _cgroups
_schedulerrn ro rj errorstrr rV threadingZEvent
_terminaterS r\ Z
thread_mapZ_threadsevselZ
TYPE_SOFTWAREZCOUNT_SW_DUMMYZ
SAMPLE_TIDZ
SAMPLE_CPUopenr] Zevlistaddrw )r
instanceZperf_mmap_pages_rawrr r+ r r )r
r _instance_init s^
z SchedulerPlugin._instance_initc C s* |j r&x |j j D ]}tj|j qW d S )N)ra
get_pollfdr) closer, )r
r fdr r r _instance_cleanup s z!SchedulerPlugin._instance_cleanupc C s4 d t jddd d d d ddd dd d d d d d d d d d dS )NFTcalcZfalse)isolated_corescgroup_mount_pointcgroup_mount_point_initcgroup_groups_initcgroup_for_isolated_corescgroup_ps_blacklist ps_whitelist ps_blacklist irq_processdefault_irq_smp_affinityrr perf_process_forksched_min_granularity_nssched_latency_nssched_wakeup_granularity_nssched_tunable_scalingsched_migration_cost_nssched_nr_migrate numa_balancing_scan_delay_ms!numa_balancing_scan_period_min_ms!numa_balancing_scan_period_max_msnuma_balancing_scan_size_mb)rR ZDEF_CGROUP_MOUNT_POINT)clsr r r _get_config_options s, z#SchedulerPlugin._get_config_optionsc C s |d k rt |jddS d S )N./)r replace)r
r r r r rm 9 s z%SchedulerPlugin._sanitize_cgroup_pathc C s> t |tjs|}tj|}tj|}| j|r:d| d }|S )N[])
isinstanceprocfsprocessZprocess_cmdline _is_kthread)r
r r9 r r r r _get_cmdline= s
z SchedulerPlugin._get_cmdlinec C s t j }|j i }x|j D ]}yN| j|}|d }|||< d|krnx&|d j D ]}| j|}|||< qTW W q ttfk
r } z$|jtj ks|jtj
krw n W Y d d }~X q X q W |S )Nr9 rx ) r pidstatsreload_threadsr2 r keysOSErrorIOErrorerrnoENOENTESRCH)r
ps processesprocr r9 er r r
get_processesG s$
z SchedulerPlugin.get_processesc C s@ | j j|}| j j|}| j j|}tjd|||f ||fS )Nz8Read scheduler policy '%s' and priority '%d' of PID '%d')rb r: r7 rF ri debug)r
r9 r sched_strr
r r r _get_rt` s zSchedulerPlugin._get_rtc C s | j j|}tjd|||f yB| j j|}| j j|}||k sJ||kr`tjd||||f W n4 ttfk
r } ztjd| W Y d d }~X nX y| j j ||| W n` ttfk
r } z>t
|dr|j t j krtjd| ntjd||f W Y d d }~X nX d S )NzBSetting scheduler policy to '%s' and priority to '%d' of PID '%d'.z9Priority for %s must be in range %d - %d. '%d' was given.z(Failed to get allowed priority range: %sr zAFailed to set scheduling parameters of PID %d, the task vanished.z1Failed to set scheduling parameters of PID %d: %s)
rb r7 ri r rH rJ r SystemErrorr r? hasattrr r )r
r9 r= r> r Zprio_minZprio_maxr r r r _set_rth s*
zSchedulerPlugin._set_rtc C s |d d t jj@ dkS )Nstatflagsr )r ZpidstatZ
PF_KTHREAD)r
r r r r r s zSchedulerPlugin._is_kthreadc C s yjt j|}|d j rd|d d dkr8tjd| n(| j|rRtjd| ntjd| dS dS W n ttfk
r } zF|j t j
ks|j t j krtjd | d S tj d
||f d
S W Y d d }~X n8 t
tfk
r } ztj d
||f dS d }~X nX d S )Nr stateZzYAffinity of zombie task with PID %d cannot be changed, the task's affinity mask is fixed.z[Affinity of kernel thread with PID %d cannot be changed, the task's affinity mask is fixed.zRAffinity of task with PID %d cannot be changed, the task's affinity mask is fixed.r r z6Failed to get task info for PID %d, the task vanished.z&Failed to get task info for PID %d: %srf r )r r Zis_bound_to_cpuri r r warnr r r r r r rc KeyError)r
r9 r r r r r _affinity_changeable s2
z$SchedulerPlugin._affinity_changeablec C s\ y| j | }W n( tk
r6 t| j}|| j |< Y nX |jd krX|jd krX||_||_d S )N)r{ r r r r r
)r
r9 r r
paramsr r r _store_orig_process_rt s
z&SchedulerPlugin._store_orig_process_rtc C s d}|d kr|d kr|S y:| j |\}}|d kr4|}| j||| | j||| W nt ttfk
r } zTt|dr|jtjkrtj d| || j
kr| j
|= d}ntj d||f W Y d d }~X nX |S )NTr z=Failed to read scheduler policy of PID %d, the task vanished.FzcRefusing to set scheduler and priority of PID %d, reading original scheduling parameters failed: %s) r r r r r r r r ri r r{ r )r
r9 r= r> contZ
prev_schedZ prev_prior r r r _tune_process_rt s&
z SchedulerPlugin._tune_process_rtc C s t |d d dkS )Nrk zcgroup.)r )r
r r r r _is_cgroup_affinity s z#SchedulerPlugin._is_cgroup_affinityFc C sb y| j | }W n( tk
r6 t| j}|| j |< Y nX |jd kr^|jd kr^|rX||_n||_d S )N)r{ r r r r r )r
r9 r is_cgroupr r r r _store_orig_process_affinity s
z,SchedulerPlugin._store_orig_process_affinityc C sp xj| j jdtjt|df ddjdD ]@}y&|jdd dd }|dkrP|S d S tk
rf Y q(X q(W d S )
Nz%s/%s/%sr T)no_error
z:cpuset:r rO r )r read_filerR ZPROCFS_MOUNT_POINTr split
IndexError)r
r9 lr r r r _get_cgroup_affinity s ,
z$SchedulerPlugin._get_cgroup_affinityc C sB | j |}| j}|dkr$d||f }| jjd| t|dd d S )Nr z%s/%sz%s/tasksT)r )rm _cgroup_mount_pointr
write_to_filer )r
r9 r pathr r r _set_cgroup s
zSchedulerPlugin._set_cgroupc C s, |dd }t |t o"t|dk}||fS )Nrk r )r listrl )r
r r r r r _parse_cgroup_affinity s z&SchedulerPlugin._parse_cgroup_affinityc C s d}|d kr|S yd| j |\}}|r<| j|}| j|| n(| j|}|rX| j|||}| j|| | j||| W nt ttfk
r } zTt |dr|j
t
j krt j
d| || jkr| j|= d}nt jd||f W Y d d }~X nX |S )NTr z5Failed to read affinity of PID %d, the task vanished.FzLRefusing to set CPU affinity of PID %d, reading original affinity failed: %s)r r r
_get_affinity_get_intersect_affinity
_set_affinityr r r r r r ri r r{ r ) r
r9 r intersectr r r
prev_affinityr r r r _tune_process_affinity s4
z&SchedulerPlugin._tune_process_affinityc C sF | j |||}|sd S | j||}| s2|| jkr6d S || j| _d S )N)r r r{ r )r
r9 r r= r> r r r r r
_tune_process s z SchedulerPlugin._tune_processc C sf | j j|}|d kr.|dkr.tjd| dS y t|}W n" tk
r\ tjd| dS X ||fS )Nr z>Invalid scheduler: %s. Scheduler and priority will be ignored.z=Invalid priority: %s. Scheduler and priority will be ignored.)NN)NN)rb r6 ri r rT rg )r
r5 Z str_priorityr r
r r r _convert_sched_params s z%SchedulerPlugin._convert_sched_paramsc C sD |dkrd }n2| j |r |}n"| jj|}|s@tjd| d }|S )Nr z)Invalid affinity: %s. It will be ignored.)r r hex2cpulistri r )r
Z str_affinityr r r r _convert_affinity+ s
z!SchedulerPlugin._convert_affinityc C s6 |\}}}}}| j ||\}}| j|}|||||fS )N)r r )r
vals rule_prior r
r regexr r r _convert_sched_cfg8 s
z"SchedulerPlugin._convert_sched_cfgc C s d| j |f }ytj|tj W n4 tk
rT } ztjd||f W Y d d }~X nX | jj d|df | jj
d| j df ddddstjd| d S )Nz%s/%sz Unable to create cgroup '%s': %sz cpuset.memsT)r z3Unable to initialize 'cpuset.mems ' for cgroup '%s') r r) mkdirrR DEF_CGROUP_MODEr ri r r r r )r
r r r r r r _cgroup_create_group? s $z$SchedulerPlugin._cgroup_create_groupc C s@ | j d k r"| j | jkr"| j| j x| jD ]}| j| q*W d S )N)r r r )r
cgr r r _cgroup_initialize_groupsJ s z)SchedulerPlugin._cgroup_initialize_groupsc C s t jd ytj| jtj W n0 tk
rN } zt jd| W Y d d }~X nX | j j
dddddd| jg\}}|dkrt jd | j d S )
Nz Initializing cgroups settingsz'Unable to create cgroup mount point: %sZmountz-tr z-oZcpusetr zUnable to mount '%s') ri r r) makedirsr rR r r r r execute)r
r retoutr r r _cgroup_initializeP s
z"SchedulerPlugin._cgroup_initializec C sH yt j| W n4 tk
rB } ztjd||f W Y d d }~X nX d S )Nz#Unable to remove directory '%s': %s)r) rmdirr ri r )r
r r r r r _remove_dirZ s zSchedulerPlugin._remove_dirc C sX x&t | jD ]}| jd| j|f q W | jd k rT| j| jkrT| jd| j| jf d S )Nz%s/%s)reversedr r r r )r
r r r r _cgroup_finalize_groups` s z'SchedulerPlugin._cgroup_finalize_groupsc C sl t jd | jjd| jg\}}|dkr.c s6 g | ].\}}t jd |rt|dkr| j|fqS )zgroup\. )rematchrl r )r* rp r )r
r r rq s c S s | d d S )Nr r r )Z option_valsr r r s z8SchedulerPlugin._instance_apply_static..)keyz(error compiling regular expression: '%s'c s( g | ] \}}t j |d k r||fqS )N)r search)r* r9 r )r% r r rq s c s$ g | ] \}}|| ffqS r r )r* r9 r )r rp r
r r r r rq s z(? } ztjd| d S d }~X nX x| jj D ]x\}}||ksL|| |jkrlqL|jd k r|j d k r| j
||j|j |j d k r| j ||j qL|j
d k rL| j||j
qLW i | _| jj| j d S )NzKerror unapplying tuning, cannot get information about running processes: %s)r r r ri r r{ r0 r r r
r r r r r rz r~ r^ )r
r r r9 Z orig_paramsr r r r} s&
z$SchedulerPlugin._restore_ps_affinityc C s t tj}d}xr|dkr|dkr| jjd| j|df ddd}|d
krvx.|jdD ] }| jjd | jdf |dd
qRW |d 8 }qW |dkrtj d | d S )N rO r z%s/%s/%sZtasksT)r r r z%s/%s)r r z(Unable to cleanup tasks from cgroup '%s')rO r# )
rT rR ZCGROUP_CLEANUP_TASKS_RETRYr r r r r ri r )r
r Zcntdatar r r r _cgroup_cleanup_tasks_one s
z)SchedulerPlugin._cgroup_cleanup_tasks_onec C s@ | j d k r"| j | jkr"| j| j x| jD ]}| j| q*W d S )N)r r r% )r
r r r r _cgroup_cleanup_tasks s z%SchedulerPlugin._cgroup_cleanup_tasksc sp t t| j|| | jr2|jr2|jj |jj | j | j
| j | j sV| j
r^| j | j
rl| j d S )N)rQ rN _instance_unapply_staticrS ry r r r! joinr} r r& r r r r )r
r Zrollback)re r r r' s
z(SchedulerPlugin._instance_unapply_staticc C s t jd| d| j|df }| jj|ddd}|dkr|jtjks<|jtjkrLtjd| ntjd||f d S d }~X nX | j j
|j ||}|d k r|| j krtjd||t
|f |\}}} | j||||| | jj| j| j d S )Nz3Failed to get cmdline of PID %d, the task vanished.z#Failed to get cmdline of PID %d: %sz-tuning new process '%s' with PID '%d' by '%s')r r r r r r ri r r r Z re_lookupr r{ r r rz r r^ )
r
r r9 r% r r vr= r> r r r r _add_pid$ s$
zSchedulerPlugin._add_pidc C s6 || j kr2| j |= tjd| | jj| j| j d S )Nz)removed PID %d from the rollback database)r{ ri r rz r r^ )r
r r9 r r r _remove_pid9 s
zSchedulerPlugin._remove_pidc C s | j j|j}tj }|jj }x|D ]}|j| q&W x|jj st
|j| j d dkr:|jj r:d}x|rd}xt| j D ]j}|jj
|}|r~d}|jtjks| jr|jtjkr| j|t|j| q~|jtjkr~| j|t|j q~W qnW q:W d S )Ni r TF)r Zre_lookup_compiler selectpollra r registerr Zis_setrl rU r] Z read_on_cpurt r\ Z RECORD_COMM_perf_process_fork_valueZ RECORD_FORKr2 rT tidZ RECORD_EXITr3 ) r
r r% r5 Zfdsr Z read_eventsZcpuZeventr r r r @ s&
$ z SchedulerPlugin._thread_coder )
per_devicec C s: |rd S |r6|d k r6dj dd tjdt|D | _d S )N|c S s g | ] }d | qS )z(%s)r )r* r1 r r r rq _ s z8SchedulerPlugin._cgroup_ps_blacklist..z(?.z(?.z(? ) _default_irq_smp_affinity_valuer cpulist_unpack)r
r; r r< r0 r r r _default_irq_smp_affinityy s z)SchedulerPlugin._default_irq_smp_affinityr c C s* |rd S |r&|d k r&| j j|dk| _d S )Nr )r rV r7 )r
r; r r< r0 r r r _perf_process_fork s z"SchedulerPlugin._perf_process_forkc C s" | j j|}tjd||f |S )Nz Read affinity '%s' of PID %d)rb rA ri r )r
r9 resr r r r s z SchedulerPlugin._get_affinityc C s t jd||f y| jj|| dS ttfk
r } zXt|dr`|jtjkr`t jd| n.| j |}|dksz|d krt j
d|||f dS d }~X nX d S )
Nz'Setting CPU affinity of PID %d to '%s'.Tr z4Failed to set affinity of PID %d, the task vanished.r rf z,Failed to set affinity of PID %d to '%s': %sFr ) ri r rb rC r r r r r r r )r
r9 r r rC r r r r s
z SchedulerPlugin._set_affinityc C s" t |jt |}|r t|S |S )N)r intersectionr )r
Z affinity1Z affinity2Z affinity3Zaffr r r r s z'SchedulerPlugin._get_intersect_affinityc
s> fdd|D } j dkr. fdd|D } jdkrJ fdd|D }tdd |D }x|D ]}y j|| }W nb ttfk
r } zB|jtjks|jtjkrt j
d| nt j d||f wbW Y d d }~X nX j ||d d
} | sqb| j
kr
| j
| _| rbd || krb j|| d j |d qbW d S ) Nc s( g | ] }t j j j|d k r|qS )N)r r rY _get_stat_comm)r* r1 )r
r r rq s
z9SchedulerPlugin._set_all_obj_affinity..rO c s( g | ] }t j j j|d kr|qS )N)r r rZ rE )r* r1 )r
r r rq s
c s( g | ] }t j j j|d kr|qS )N)r r r[ _get_stat_cgroup)r* r1 )r
r r rq s
c S s g | ]}|j |fqS r )r9 )r* r1 r r r rq s z3Failed to get cmdline of PID %d, the task vanished.zARefusing to set affinity of PID %d, failed to get its cmdline: %sT)r rx )rZ r[ r. r r r r r r ri r r r r{ r _set_all_obj_affinityr2 )
r
Zobjsr rx ZpslZpsdr9 r r r r )r
r rG s6
z%SchedulerPlugin._set_all_obj_affinityc
C s( y|d S t ttfk
r" dS X d S )NZcgroupsrO )r r r )r
r& r r r rF s z SchedulerPlugin._get_stat_cgroupc
C s, y |d d S t ttfk
r& dS X d S )Nr rv rO )r r r )r
r& r r r rE s z SchedulerPlugin._get_stat_commc C s` y&t j }|j | j|j |d W n4 ttfk
rZ } ztjd| W Y d d }~X nX d S )NFzIerror applying tuning, cannot get information about running processes: %s) r r r rG r2 r r ri r )r
r r r r r r _set_ps_affinity s z SchedulerPlugin._set_ps_affinityc C s yJ| j j|}tjd||f d| }t|d}|j| W d Q R X dS ttfk
r } zLt|dr|j t j
kr| rtjd| d
S tj d|||f d S W Y d d }~X nX d S ) Nz&Setting SMP affinity of IRQ %s to '%s'z/proc/irq/%s/smp_affinitywr r z/Setting SMP affinity of IRQ %s is not supportedrf z0Failed to set SMP affinity of IRQ %s to '%s': %sr r r ) r cpulist2hexri r r writer r r r ZEIOr )r
rP r Z restoring affinity_hexfilenamer# r r r r _set_irq_affinity s" z!SchedulerPlugin._set_irq_affinityc C s| y>| j j|}tjd| tdd}|j| W d Q R X W n8 ttfk
rv } ztjd||f W Y d d }~X nX d S )Nz(Setting default SMP IRQ affinity to '%s'z /proc/irq/default_smp_affinityrI z2Failed to set default SMP IRQ affinity to '%s': %s) r rJ ri r r rK r r r )r
r rL r# r r r r _set_default_irq_affinity s z)SchedulerPlugin._set_default_irq_affinityc
C s" t }tj }x|j D ]}y"|| d }tjd||f W n tk
rT wY nX | j|||}t|t|krvq| j ||d}|dkr||j
|< q|d kr|j j | qW | j
jd}| j
j|}| jdkr| j|||}n| jdkr| j}| jdkr| j| ||_| jj| j| d S )
Nr z Read affinity of IRQ '%s': '%s'Fr rf z /proc/irq/default_smp_affinityr r> r )r r
interruptsr ri r r r r rN r r appendr r r r? rO r rz r` ) r
r irq_originalr rP r r rC Zprev_affinity_hexr r r _set_all_irq_affinity
s6
z%SchedulerPlugin._set_all_irq_affinityc C sn | j j| jd }|d kr d S x$|jj D ]\}}| j||d q(W | jdkr\|j}| j| | j j | j d S )NTr> )
rz r4 r` r r0 rN r? r rO r~ )r
rR rP r r r r _restore_all_irq_affinity) s
z)SchedulerPlugin._restore_all_irq_affinityc C sF t |jt |}|r,tjtj||f ntjtj|||f |S )N)r issubsetri r| rR r* r r+ )r
irq_descriptioncorrect_affinityr, rC r r r _verify_irq_affinity4 s
z$SchedulerPlugin._verify_irq_affinityc C s | j j| jd }tj }d}x|j D ]}||jkrR|rRd| }tjt j
| q&y<|| d }tj d||f d| } | j | ||sd}W q& t
k
r w&Y q&X q&W | jjd}
| jj|
}| jdkr| j d || jd
kr|n| j rd}|S ) NTz-IRQ %s does not support changing SMP affinityr z#Read SMP affinity of IRQ '%s': '%s'zSMP affinity of IRQ %sFz /proc/irq/default_smp_affinityr> zdefault IRQ SMP affinityr )rz r4 r` r rP r r ri r| rR Z STR_VERIFY_PROFILE_VALUE_MISSINGr rX r r r r r? ) r
rW r0 rR r rC rP descriptionr, rV Zcurrent_affinity_hexr r r _verify_all_irq_affinity@ s8
z(SchedulerPlugin._verify_all_irq_affinityr
)r9 r
c
C s d }d | _ |d k rrt| jj|}t| j}|j|rRt|| }| jj|| _ n | jj| j}tj d||f |sz|r|d krd S |r| j
r| j ||S dS |r| j r| j
d| j } n|} | j| | j
r| j| n| j
r| j d S )NzJInvalid isolated_cores specified, '%s' does not match available cores '%s'Tz cgroup.%s)r r r r@ r] rU r r) ri r r_ rZ r r
rH rS rT )
r
r; r r< r0 r isolatedZpresentZstr_cpusZ ps_affinityr r r _isolated_cores_ s6
z SchedulerPlugin._isolated_coresc C s d|||f }| j j|}|r"|S d||f }tjj|sv|dkrPd||f }nd|||f }d| }| jd krvd| _|| j |< |S )Nz%s_%s_%sz/proc/sys/kernel/%s_%srO z%s/%sz%s/%s/%sz/sys/kernel/debug/%sT)rX r4 r) r existsrW )r
prefix namespaceknobr r r r r _get_sched_knob_path s
z$SchedulerPlugin._get_sched_knob_pathc C sJ | j j| j|||d d}|d krFtjd| | jrFtjd d| _|S )N)r zError reading '%s'zUThis may not work with Secure Boot or kernel_lockdown (this hint is logged only once)F)r r rb ri r rW )r
r_ r` ra r$ r r r _get_sched_knob s
z SchedulerPlugin._get_sched_knobc C sN |d kr d S |sJ| j j| j|||||r0tjgnddsJtjd||f |S )NF)r z Error writing value '%s' to '%s')r r rb r r ri r )r
r_ r` ra r simremover r r _set_sched_knob s z SchedulerPlugin._set_sched_knobr c C s | j dddS )NrO r= min_granularity_ns)rc )r
r r r _get_sched_min_granularity_ns s z-SchedulerPlugin._get_sched_min_granularity_nsc C s | j ddd|||S )NrO r= rg )rf )r
r rd re r r r _set_sched_min_granularity_ns s z-SchedulerPlugin._set_sched_min_granularity_nsr c C s | j dddS )NrO r=
latency_ns)rc )r
r r r _get_sched_latency_ns s z%SchedulerPlugin._get_sched_latency_nsc C s | j ddd|||S )NrO r= rj )rf )r
r rd re r r r _set_sched_latency_ns s z%SchedulerPlugin._set_sched_latency_nsr c C s | j dddS )NrO r= wakeup_granularity_ns)rc )r
r r r _get_sched_wakeup_granularity_ns s z0SchedulerPlugin._get_sched_wakeup_granularity_nsc C s | j ddd|||S )NrO r= rm )rf )r
r rd re r r r _set_sched_wakeup_granularity_ns s z0SchedulerPlugin._set_sched_wakeup_granularity_nsr c C s | j dddS )NrO r= tunable_scaling)rc )r
r r r _get_sched_tunable_scaling s z*SchedulerPlugin._get_sched_tunable_scalingc C s | j ddd|||S )NrO r= rp )rf )r
r rd re r r r _set_sched_tunable_scaling s z*SchedulerPlugin._set_sched_tunable_scalingr c C s | j dddS )NrO r= migration_cost_ns)rc )r
r r r _get_sched_migration_cost_ns s z,SchedulerPlugin._get_sched_migration_cost_nsc C s | j ddd|||S )NrO r= rs )rf )r
r rd re r r r _set_sched_migration_cost_ns s z,SchedulerPlugin._set_sched_migration_cost_nsr c C s | j dddS )NrO r=
nr_migrate)rc )r
r r r _get_sched_nr_migrate s z%SchedulerPlugin._get_sched_nr_migratec C s | j ddd|||S )NrO r= rv )rf )r
r rd re r r r _set_sched_nr_migrate s z%SchedulerPlugin._set_sched_nr_migrater c C s | j dddS )Nr= numa_balancing
scan_delay_ms)rc )r
r r r !_get_numa_balancing_scan_delay_ms s z1SchedulerPlugin._get_numa_balancing_scan_delay_msc C s | j ddd|||S )Nr= ry rz )rf )r
r rd re r r r !_set_numa_balancing_scan_delay_ms s z1SchedulerPlugin._set_numa_balancing_scan_delay_msr c C s | j dddS )Nr= ry scan_period_min_ms)rc )r
r r r &_get_numa_balancing_scan_period_min_ms s z6SchedulerPlugin._get_numa_balancing_scan_period_min_msc C s | j ddd|||S )Nr= ry r} )rf )r
r rd re r r r &_set_numa_balancing_scan_period_min_ms s z6SchedulerPlugin._set_numa_balancing_scan_period_min_msr c C s | j dddS )Nr= ry scan_period_max_ms)rc )r
r r r &_get_numa_balancing_scan_period_max_ms s z6SchedulerPlugin._get_numa_balancing_scan_period_max_msc C s | j ddd|||S )Nr= ry r )rf )r
r rd re r r r &_set_numa_balancing_scan_period_max_ms s z6SchedulerPlugin._set_numa_balancing_scan_period_max_msr c C s | j dddS )Nr= ry scan_size_mb)rc )r
r r r _get_numa_balancing_scan_size_mb s z0SchedulerPlugin._get_numa_balancing_scan_size_mbc C s | j ddd|||S )Nr= ry r )rf )r
r rd re r r r _set_numa_balancing_scan_size_mb s z0SchedulerPlugin._set_numa_balancing_scan_size_mb)F)F)F)F)F)br r r rK r rj r r classmethodr rm r r r r r r r r r r r r r r r r r r r r r r r r r r
r r r} r% r& rR Z
ROLLBACK_SOFTr' r- r. r/ r2 r3 r Zcommand_customr= rY rZ r_ rA rB r r r rG rF rE rH rN rO rS rT rX rZ r] rb rc rf Z command_getrh Z command_setri rk rl rn ro rq rr rt ru rw rx r{ r| r~ r r r r r
__classcell__r r )re r rN s ( ?
<
" $
rN ) rO r Z
decoratorsZ
tuned.logsZtunedr
subprocessr r\ r4 Z tuned.constsrR r Ztuned.utils.commandsr r r) r rh r rc rM Zlogsr4 ri objectr r r rL ZPluginrN r r r r s0
/