kernel_optimize_test/kernel/cgroup_freezer.c
Tejun Heo 1f7dd3e5a6 cgroup: fix handling of multi-destination migration from subtree_control enabling
Consider the following v2 hierarchy.

  P0 (+memory) --- P1 (-memory) --- A
                                 \- B
       
P0 has memory enabled in its subtree_control while P1 doesn't.  If
both A and B contain processes, they would belong to the memory css of
P1.  Now if memory is enabled on P1's subtree_control, memory csses
should be created on both A and B and A's processes should be moved to
the former and B's processes the latter.  IOW, enabling controllers
can cause atomic migrations into different csses.

The core cgroup migration logic has been updated accordingly but the
controller migration methods haven't and still assume that all tasks
migrate to a single target css; furthermore, the methods were fed the
css in which subtree_control was updated which is the parent of the
target csses.  pids controller depends on the migration methods to
move charges and this made the controller attribute charges to the
wrong csses often triggering the following warning by driving a
counter negative.

 WARNING: CPU: 1 PID: 1 at kernel/cgroup_pids.c:97 pids_cancel.constprop.6+0x31/0x40()
 Modules linked in:
 CPU: 1 PID: 1 Comm: systemd Not tainted 4.4.0-rc1+ #29
 ...
  ffffffff81f65382 ffff88007c043b90 ffffffff81551ffc 0000000000000000
  ffff88007c043bc8 ffffffff810de202 ffff88007a752000 ffff88007a29ab00
  ffff88007c043c80 ffff88007a1d8400 0000000000000001 ffff88007c043bd8
 Call Trace:
  [<ffffffff81551ffc>] dump_stack+0x4e/0x82
  [<ffffffff810de202>] warn_slowpath_common+0x82/0xc0
  [<ffffffff810de2fa>] warn_slowpath_null+0x1a/0x20
  [<ffffffff8118e031>] pids_cancel.constprop.6+0x31/0x40
  [<ffffffff8118e0fd>] pids_can_attach+0x6d/0xf0
  [<ffffffff81188a4c>] cgroup_taskset_migrate+0x6c/0x330
  [<ffffffff81188e05>] cgroup_migrate+0xf5/0x190
  [<ffffffff81189016>] cgroup_attach_task+0x176/0x200
  [<ffffffff8118949d>] __cgroup_procs_write+0x2ad/0x460
  [<ffffffff81189684>] cgroup_procs_write+0x14/0x20
  [<ffffffff811854e5>] cgroup_file_write+0x35/0x1c0
  [<ffffffff812e26f1>] kernfs_fop_write+0x141/0x190
  [<ffffffff81265f88>] __vfs_write+0x28/0xe0
  [<ffffffff812666fc>] vfs_write+0xac/0x1a0
  [<ffffffff81267019>] SyS_write+0x49/0xb0
  [<ffffffff81bcef32>] entry_SYSCALL_64_fastpath+0x12/0x76

This patch fixes the bug by removing @css parameter from the three
migration methods, ->can_attach, ->cancel_attach() and ->attach() and
updating cgroup_taskset iteration helpers also return the destination
css in addition to the task being migrated.  All controllers are
updated accordingly.

* Controllers which don't care whether there are one or multiple
  target csses can be converted trivially.  cpu, io, freezer, perf,
  netclassid and netprio fall in this category.

* cpuset's current implementation assumes that there's single source
  and destination and thus doesn't support v2 hierarchy already.  The
  only change made by this patchset is how that single destination css
  is obtained.

* memory migration path already doesn't do anything on v2.  How the
  single destination css is obtained is updated and the prep stage of
  mem_cgroup_can_attach() is reordered to accomodate the change.

* pids is the only controller which was affected by this bug.  It now
  correctly handles multi-destination migrations and no longer causes
  counter underflow from incorrect accounting.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Cc: Aleksa Sarai <cyphar@cyphar.com>
2015-12-03 10:18:21 -05:00

482 lines
12 KiB
C

/*
* cgroup_freezer.c - control group freezer subsystem
*
* Copyright IBM Corporation, 2007
*
* Author : Cedric Le Goater <clg@fr.ibm.com>
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of version 2.1 of the GNU Lesser General Public License
* as published by the Free Software Foundation.
*
* This program is distributed in the hope that it would be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
*/
#include <linux/export.h>
#include <linux/slab.h>
#include <linux/cgroup.h>
#include <linux/fs.h>
#include <linux/uaccess.h>
#include <linux/freezer.h>
#include <linux/seq_file.h>
#include <linux/mutex.h>
/*
* A cgroup is freezing if any FREEZING flags are set. FREEZING_SELF is
* set if "FROZEN" is written to freezer.state cgroupfs file, and cleared
* for "THAWED". FREEZING_PARENT is set if the parent freezer is FREEZING
* for whatever reason. IOW, a cgroup has FREEZING_PARENT set if one of
* its ancestors has FREEZING_SELF set.
*/
enum freezer_state_flags {
CGROUP_FREEZER_ONLINE = (1 << 0), /* freezer is fully online */
CGROUP_FREEZING_SELF = (1 << 1), /* this freezer is freezing */
CGROUP_FREEZING_PARENT = (1 << 2), /* the parent freezer is freezing */
CGROUP_FROZEN = (1 << 3), /* this and its descendants frozen */
/* mask for all FREEZING flags */
CGROUP_FREEZING = CGROUP_FREEZING_SELF | CGROUP_FREEZING_PARENT,
};
struct freezer {
struct cgroup_subsys_state css;
unsigned int state;
};
static DEFINE_MUTEX(freezer_mutex);
static inline struct freezer *css_freezer(struct cgroup_subsys_state *css)
{
return css ? container_of(css, struct freezer, css) : NULL;
}
static inline struct freezer *task_freezer(struct task_struct *task)
{
return css_freezer(task_css(task, freezer_cgrp_id));
}
static struct freezer *parent_freezer(struct freezer *freezer)
{
return css_freezer(freezer->css.parent);
}
bool cgroup_freezing(struct task_struct *task)
{
bool ret;
rcu_read_lock();
ret = task_freezer(task)->state & CGROUP_FREEZING;
rcu_read_unlock();
return ret;
}
static const char *freezer_state_strs(unsigned int state)
{
if (state & CGROUP_FROZEN)
return "FROZEN";
if (state & CGROUP_FREEZING)
return "FREEZING";
return "THAWED";
};
static struct cgroup_subsys_state *
freezer_css_alloc(struct cgroup_subsys_state *parent_css)
{
struct freezer *freezer;
freezer = kzalloc(sizeof(struct freezer), GFP_KERNEL);
if (!freezer)
return ERR_PTR(-ENOMEM);
return &freezer->css;
}
/**
* freezer_css_online - commit creation of a freezer css
* @css: css being created
*
* We're committing to creation of @css. Mark it online and inherit
* parent's freezing state while holding both parent's and our
* freezer->lock.
*/
static int freezer_css_online(struct cgroup_subsys_state *css)
{
struct freezer *freezer = css_freezer(css);
struct freezer *parent = parent_freezer(freezer);
mutex_lock(&freezer_mutex);
freezer->state |= CGROUP_FREEZER_ONLINE;
if (parent && (parent->state & CGROUP_FREEZING)) {
freezer->state |= CGROUP_FREEZING_PARENT | CGROUP_FROZEN;
atomic_inc(&system_freezing_cnt);
}
mutex_unlock(&freezer_mutex);
return 0;
}
/**
* freezer_css_offline - initiate destruction of a freezer css
* @css: css being destroyed
*
* @css is going away. Mark it dead and decrement system_freezing_count if
* it was holding one.
*/
static void freezer_css_offline(struct cgroup_subsys_state *css)
{
struct freezer *freezer = css_freezer(css);
mutex_lock(&freezer_mutex);
if (freezer->state & CGROUP_FREEZING)
atomic_dec(&system_freezing_cnt);
freezer->state = 0;
mutex_unlock(&freezer_mutex);
}
static void freezer_css_free(struct cgroup_subsys_state *css)
{
kfree(css_freezer(css));
}
/*
* Tasks can be migrated into a different freezer anytime regardless of its
* current state. freezer_attach() is responsible for making new tasks
* conform to the current state.
*
* Freezer state changes and task migration are synchronized via
* @freezer->lock. freezer_attach() makes the new tasks conform to the
* current state and all following state changes can see the new tasks.
*/
static void freezer_attach(struct cgroup_taskset *tset)
{
struct task_struct *task;
struct cgroup_subsys_state *new_css;
mutex_lock(&freezer_mutex);
/*
* Make the new tasks conform to the current state of @new_css.
* For simplicity, when migrating any task to a FROZEN cgroup, we
* revert it to FREEZING and let update_if_frozen() determine the
* correct state later.
*
* Tasks in @tset are on @new_css but may not conform to its
* current state before executing the following - !frozen tasks may
* be visible in a FROZEN cgroup and frozen tasks in a THAWED one.
*/
cgroup_taskset_for_each(task, new_css, tset) {
struct freezer *freezer = css_freezer(new_css);
if (!(freezer->state & CGROUP_FREEZING)) {
__thaw_task(task);
} else {
freeze_task(task);
/* clear FROZEN and propagate upwards */
while (freezer && (freezer->state & CGROUP_FROZEN)) {
freezer->state &= ~CGROUP_FROZEN;
freezer = parent_freezer(freezer);
}
}
}
mutex_unlock(&freezer_mutex);
}
/**
* freezer_fork - cgroup post fork callback
* @task: a task which has just been forked
*
* @task has just been created and should conform to the current state of
* the cgroup_freezer it belongs to. This function may race against
* freezer_attach(). Losing to freezer_attach() means that we don't have
* to do anything as freezer_attach() will put @task into the appropriate
* state.
*/
static void freezer_fork(struct task_struct *task, void *private)
{
struct freezer *freezer;
/*
* The root cgroup is non-freezable, so we can skip locking the
* freezer. This is safe regardless of race with task migration.
* If we didn't race or won, skipping is obviously the right thing
* to do. If we lost and root is the new cgroup, noop is still the
* right thing to do.
*/
if (task_css_is_root(task, freezer_cgrp_id))
return;
mutex_lock(&freezer_mutex);
rcu_read_lock();
freezer = task_freezer(task);
if (freezer->state & CGROUP_FREEZING)
freeze_task(task);
rcu_read_unlock();
mutex_unlock(&freezer_mutex);
}
/**
* update_if_frozen - update whether a cgroup finished freezing
* @css: css of interest
*
* Once FREEZING is initiated, transition to FROZEN is lazily updated by
* calling this function. If the current state is FREEZING but not FROZEN,
* this function checks whether all tasks of this cgroup and the descendant
* cgroups finished freezing and, if so, sets FROZEN.
*
* The caller is responsible for grabbing RCU read lock and calling
* update_if_frozen() on all descendants prior to invoking this function.
*
* Task states and freezer state might disagree while tasks are being
* migrated into or out of @css, so we can't verify task states against
* @freezer state here. See freezer_attach() for details.
*/
static void update_if_frozen(struct cgroup_subsys_state *css)
{
struct freezer *freezer = css_freezer(css);
struct cgroup_subsys_state *pos;
struct css_task_iter it;
struct task_struct *task;
lockdep_assert_held(&freezer_mutex);
if (!(freezer->state & CGROUP_FREEZING) ||
(freezer->state & CGROUP_FROZEN))
return;
/* are all (live) children frozen? */
rcu_read_lock();
css_for_each_child(pos, css) {
struct freezer *child = css_freezer(pos);
if ((child->state & CGROUP_FREEZER_ONLINE) &&
!(child->state & CGROUP_FROZEN)) {
rcu_read_unlock();
return;
}
}
rcu_read_unlock();
/* are all tasks frozen? */
css_task_iter_start(css, &it);
while ((task = css_task_iter_next(&it))) {
if (freezing(task)) {
/*
* freezer_should_skip() indicates that the task
* should be skipped when determining freezing
* completion. Consider it frozen in addition to
* the usual frozen condition.
*/
if (!frozen(task) && !freezer_should_skip(task))
goto out_iter_end;
}
}
freezer->state |= CGROUP_FROZEN;
out_iter_end:
css_task_iter_end(&it);
}
static int freezer_read(struct seq_file *m, void *v)
{
struct cgroup_subsys_state *css = seq_css(m), *pos;
mutex_lock(&freezer_mutex);
rcu_read_lock();
/* update states bottom-up */
css_for_each_descendant_post(pos, css) {
if (!css_tryget_online(pos))
continue;
rcu_read_unlock();
update_if_frozen(pos);
rcu_read_lock();
css_put(pos);
}
rcu_read_unlock();
mutex_unlock(&freezer_mutex);
seq_puts(m, freezer_state_strs(css_freezer(css)->state));
seq_putc(m, '\n');
return 0;
}
static void freeze_cgroup(struct freezer *freezer)
{
struct css_task_iter it;
struct task_struct *task;
css_task_iter_start(&freezer->css, &it);
while ((task = css_task_iter_next(&it)))
freeze_task(task);
css_task_iter_end(&it);
}
static void unfreeze_cgroup(struct freezer *freezer)
{
struct css_task_iter it;
struct task_struct *task;
css_task_iter_start(&freezer->css, &it);
while ((task = css_task_iter_next(&it)))
__thaw_task(task);
css_task_iter_end(&it);
}
/**
* freezer_apply_state - apply state change to a single cgroup_freezer
* @freezer: freezer to apply state change to
* @freeze: whether to freeze or unfreeze
* @state: CGROUP_FREEZING_* flag to set or clear
*
* Set or clear @state on @cgroup according to @freeze, and perform
* freezing or thawing as necessary.
*/
static void freezer_apply_state(struct freezer *freezer, bool freeze,
unsigned int state)
{
/* also synchronizes against task migration, see freezer_attach() */
lockdep_assert_held(&freezer_mutex);
if (!(freezer->state & CGROUP_FREEZER_ONLINE))
return;
if (freeze) {
if (!(freezer->state & CGROUP_FREEZING))
atomic_inc(&system_freezing_cnt);
freezer->state |= state;
freeze_cgroup(freezer);
} else {
bool was_freezing = freezer->state & CGROUP_FREEZING;
freezer->state &= ~state;
if (!(freezer->state & CGROUP_FREEZING)) {
if (was_freezing)
atomic_dec(&system_freezing_cnt);
freezer->state &= ~CGROUP_FROZEN;
unfreeze_cgroup(freezer);
}
}
}
/**
* freezer_change_state - change the freezing state of a cgroup_freezer
* @freezer: freezer of interest
* @freeze: whether to freeze or thaw
*
* Freeze or thaw @freezer according to @freeze. The operations are
* recursive - all descendants of @freezer will be affected.
*/
static void freezer_change_state(struct freezer *freezer, bool freeze)
{
struct cgroup_subsys_state *pos;
/*
* Update all its descendants in pre-order traversal. Each
* descendant will try to inherit its parent's FREEZING state as
* CGROUP_FREEZING_PARENT.
*/
mutex_lock(&freezer_mutex);
rcu_read_lock();
css_for_each_descendant_pre(pos, &freezer->css) {
struct freezer *pos_f = css_freezer(pos);
struct freezer *parent = parent_freezer(pos_f);
if (!css_tryget_online(pos))
continue;
rcu_read_unlock();
if (pos_f == freezer)
freezer_apply_state(pos_f, freeze,
CGROUP_FREEZING_SELF);
else
freezer_apply_state(pos_f,
parent->state & CGROUP_FREEZING,
CGROUP_FREEZING_PARENT);
rcu_read_lock();
css_put(pos);
}
rcu_read_unlock();
mutex_unlock(&freezer_mutex);
}
static ssize_t freezer_write(struct kernfs_open_file *of,
char *buf, size_t nbytes, loff_t off)
{
bool freeze;
buf = strstrip(buf);
if (strcmp(buf, freezer_state_strs(0)) == 0)
freeze = false;
else if (strcmp(buf, freezer_state_strs(CGROUP_FROZEN)) == 0)
freeze = true;
else
return -EINVAL;
freezer_change_state(css_freezer(of_css(of)), freeze);
return nbytes;
}
static u64 freezer_self_freezing_read(struct cgroup_subsys_state *css,
struct cftype *cft)
{
struct freezer *freezer = css_freezer(css);
return (bool)(freezer->state & CGROUP_FREEZING_SELF);
}
static u64 freezer_parent_freezing_read(struct cgroup_subsys_state *css,
struct cftype *cft)
{
struct freezer *freezer = css_freezer(css);
return (bool)(freezer->state & CGROUP_FREEZING_PARENT);
}
static struct cftype files[] = {
{
.name = "state",
.flags = CFTYPE_NOT_ON_ROOT,
.seq_show = freezer_read,
.write = freezer_write,
},
{
.name = "self_freezing",
.flags = CFTYPE_NOT_ON_ROOT,
.read_u64 = freezer_self_freezing_read,
},
{
.name = "parent_freezing",
.flags = CFTYPE_NOT_ON_ROOT,
.read_u64 = freezer_parent_freezing_read,
},
{ } /* terminate */
};
struct cgroup_subsys freezer_cgrp_subsys = {
.css_alloc = freezer_css_alloc,
.css_online = freezer_css_online,
.css_offline = freezer_css_offline,
.css_free = freezer_css_free,
.attach = freezer_attach,
.fork = freezer_fork,
.legacy_cftypes = files,
};