ChangeLog
========================================
v1 -> v2
  o implement can_attach() and attach() method.
  o remvoe can_fork member from cgroup_subsys.
    instead, copy_process() call task_cgroup_can_fork() directly.
  o fixed dead lock bug.
  o add Documentation/contoroller/task.txt


benefit
========================================
1. prevent fork bomb.

We already have "/prox/sys/kernel/threads-max".
but it isn't perfect solution.
because threads-max prevent any process creation.
then, System-administrator can't login and restore trouble.

task limit cgroup is better solution.
it can prevent fork by network service daemon, but allow fork by interactive process.


2. help implement batch processing

in general, batch processing environment need support #task limit.
this patch help implement it.


usage
========================================

# mount -t cgroup -o task none /dev/cgroup
# mkdir /dev/cgroup/foo
# cd /dev/cgroup/foo
# ls 
notify_on_release  task.max_tasks  task.nr_tasks  tasks
# echo 100 > task.max_tasks
# echo $$ > tasks
# fork_bomb 1000 &  <- try create 1000 process
# pgrep fork_bomb | wc -l
98


Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
CC: Li Zefan <lizf@cn.fujitsu.com>
CC: Paul Menage <menage@google.com>

---
 Documentation/controllers/task.txt |   18 +++
 include/linux/cgroup_subsys.h      |    4 
 include/linux/cgroup_task.h        |   33 +++++
 init/Kconfig                       |   10 +
 kernel/Makefile                    |    1 
 kernel/cgroup_task.c               |  213 +++++++++++++++++++++++++++++++++++++
 kernel/fork.c                      |    4 
 7 files changed, 283 insertions(+)

Index: b/init/Kconfig
===================================================================
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -289,6 +289,16 @@ config CGROUP_DEBUG
 
 	  Say N if unsure
 
+config CGROUP_TASK
+	bool "Simple number of task accounting cgroup subsystem"
+	depends on CGROUPS && EXPERIMENTAL
+	default n
+	help
+          Provides a simple number of task accounting cgroup subsystem for
+	  prevent fork bomb.
+
+	  Say N if unsure
+
 config CGROUP_NS
         bool "Namespace cgroup subsystem"
         depends on CGROUPS
Index: b/kernel/Makefile
===================================================================
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -46,6 +46,7 @@ obj-$(CONFIG_KEXEC) += kexec.o
 obj-$(CONFIG_COMPAT) += compat.o
 obj-$(CONFIG_CGROUPS) += cgroup.o
 obj-$(CONFIG_CGROUP_DEBUG) += cgroup_debug.o
+obj-$(CONFIG_CGROUP_TASK) += cgroup_task.o
 obj-$(CONFIG_CPUSETS) += cpuset.o
 obj-$(CONFIG_CGROUP_NS) += ns_cgroup.o
 obj-$(CONFIG_UTS_NS) += utsname.o
Index: b/kernel/cgroup_task.c
===================================================================
--- /dev/null
+++ b/kernel/cgroup_task.c
@@ -0,0 +1,213 @@
+/* cgroup_task.c - #task control group
+ *
+ * Copyright (C) 2008 KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ */
+
+#include <linux/cgroup.h>
+#include <linux/err.h>
+#include <linux/cgroup_task.h>
+
+struct task_cgroup {
+	struct cgroup_subsys_state css;
+	/*
+	 * the counter to account for number of thread.
+	 */
+	int max_tasks;
+	int nr_tasks;
+
+	spinlock_t lock;
+};
+
+struct cgroup_subsys task_cgroup_subsys __read_mostly;
+
+static inline struct task_cgroup *task_cgroup_from_task(struct task_struct *p)
+{
+	return container_of(task_subsys_state(p, task_cgroup_subsys_id),
+				struct task_cgroup, css);
+}
+
+static struct task_cgroup *task_cgroup_from_cgrp(struct cgroup *cgrp)
+{
+	return container_of(cgroup_subsys_state(cgrp,
+				task_cgroup_subsys_id), struct task_cgroup,
+				css);
+}
+
+static int task_cgroup_max_tasks_write(struct cgroup *cgrp,
+				       struct cftype *cftype,
+				       s64 new_max_tasks)
+{
+	struct task_cgroup *taskcg;
+	int ret = -EBUSY;
+
+	if ((new_max_tasks > INT_MAX) ||
+	    (new_max_tasks < -1))
+		return -EINVAL;
+
+	taskcg = task_cgroup_from_cgrp(cgrp);
+
+	spin_lock(&taskcg->lock);
+	if (taskcg->nr_tasks <= new_max_tasks) {
+		taskcg->max_tasks = new_max_tasks;
+		ret = 0;
+	}
+	spin_unlock(&taskcg->lock);
+
+	return ret;
+}
+
+static s64 task_cgroup_max_tasks_read(struct cgroup *cgrp, struct cftype *cft)
+{
+	s64 max_tasks;
+	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
+
+	spin_lock(&taskcg->lock);
+	max_tasks = taskcg->max_tasks;
+	spin_unlock(&taskcg->lock);
+
+	return max_tasks;
+}
+
+static s64 task_cgroup_nr_tasks_read(struct cgroup *cgrp, struct cftype *cft)
+{
+	s64 nr_tasks;
+	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
+
+	spin_lock(&taskcg->lock);
+	nr_tasks = taskcg->nr_tasks;
+	spin_unlock(&taskcg->lock);
+
+	return nr_tasks;
+}
+
+static struct cftype task_cgroup_files[] =  {
+	{
+		.name = "max_tasks",
+		.write_s64 = task_cgroup_max_tasks_write,
+		.read_s64  = task_cgroup_max_tasks_read,
+	},
+	{
+		.name = "nr_tasks",
+		.read_s64  = task_cgroup_nr_tasks_read,
+	},
+};
+
+static struct cgroup_subsys_state *task_cgroup_create(struct cgroup_subsys *ss,
+						   struct cgroup *cgrp)
+{
+	struct task_cgroup *taskcg;
+
+	taskcg = kzalloc(sizeof(struct task_cgroup), GFP_KERNEL);
+	if (!taskcg)
+		return ERR_PTR(-ENOMEM);
+
+	taskcg->max_tasks = -1;   /* infinite */
+	spin_lock_init(&taskcg->lock);
+
+	return &taskcg->css;
+}
+
+static void task_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+	kfree(cgrp->subsys[task_cgroup_subsys_id]);
+}
+
+
+static int task_cgroup_can_attach(struct cgroup_subsys *ss,
+				  struct cgroup *cgrp, struct task_struct *tsk)
+{
+	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
+	int ret = 0;
+
+	spin_lock(&taskcg->lock);
+	if (taskcg->nr_tasks == taskcg->max_tasks)
+		ret = -EBUSY;
+	spin_unlock(&taskcg->lock);
+
+	return ret;
+}
+
+static void task_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
+			       struct cgroup *old_cgrp, struct task_struct *tsk)
+{
+	struct task_cgroup *old_taskcg = task_cgroup_from_cgrp(old_cgrp);
+	struct task_cgroup *new_taskcg = task_cgroup_from_cgrp(cgrp);
+
+	spin_lock(&old_taskcg->lock);
+	old_taskcg->nr_tasks--;
+	spin_unlock(&old_taskcg->lock);
+
+	spin_lock(&new_taskcg->lock);
+	new_taskcg->nr_tasks++;
+	spin_unlock(&new_taskcg->lock);
+}
+
+
+static int task_cgroup_populate(struct cgroup_subsys *ss,
+				struct cgroup *cgrp)
+{
+	return cgroup_add_files(cgrp, ss, task_cgroup_files,
+				ARRAY_SIZE(task_cgroup_files));
+}
+
+int task_cgroup_can_fork(struct task_struct *task)
+{
+	struct task_cgroup *taskcg;
+	int max_tasks;
+	int ret = 0;
+
+	if (task_cgroup_subsys.disabled)
+		return 0;
+
+	taskcg = task_cgroup_from_task(task);
+
+	spin_lock(&taskcg->lock);
+	max_tasks = taskcg->max_tasks;
+
+	if ((max_tasks >= 0) &&
+	    (taskcg->nr_tasks >= max_tasks))
+		ret = -EAGAIN;
+	else
+		taskcg->nr_tasks++;
+	spin_unlock(&taskcg->lock);
+
+	return ret;
+}
+
+static void task_cgroup_exit(struct cgroup_subsys *ss, struct task_struct *task)
+{
+	struct task_cgroup *taskcg;
+
+	if (task_cgroup_subsys.disabled)
+		return;
+
+	taskcg = task_cgroup_from_task(task);
+
+	spin_lock(&taskcg->lock);
+	taskcg->nr_tasks--;
+	spin_unlock(&taskcg->lock);
+}
+
+
+struct cgroup_subsys task_cgroup_subsys = {
+	.name = "task",
+	.subsys_id = task_cgroup_subsys_id,
+	.create = task_cgroup_create,
+	.destroy = task_cgroup_destroy,
+	.can_attach = task_cgroup_can_attach,
+	.attach = task_cgroup_attach,
+	.populate = task_cgroup_populate,
+	.exit = task_cgroup_exit,
+	.early_init = 0,
+};
+
Index: b/kernel/fork.c
===================================================================
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -54,6 +54,7 @@
 #include <linux/tty.h>
 #include <linux/proc_fs.h>
 #include <linux/blkdev.h>
+#include <linux/cgroup_task.h>
 
 #include <asm/pgtable.h>
 #include <asm/pgalloc.h>
@@ -920,6 +921,8 @@ static struct task_struct *copy_process(
 		    p->user != current->nsproxy->user_ns->root_user)
 			goto bad_fork_free;
 	}
+	if (task_cgroup_can_fork(p))
+		goto bad_fork_free;
 
 	atomic_inc(&p->user->__count);
 	atomic_inc(&p->user->processes);
@@ -994,6 +997,7 @@ static struct task_struct *copy_process(
 	p->io_context = NULL;
 	p->audit_context = NULL;
 	cgroup_fork(p);
+
 #ifdef CONFIG_NUMA
 	p->mempolicy = mpol_dup(p->mempolicy);
  	if (IS_ERR(p->mempolicy)) {
Index: b/include/linux/cgroup_subsys.h
===================================================================
--- a/include/linux/cgroup_subsys.h
+++ b/include/linux/cgroup_subsys.h
@@ -48,3 +48,7 @@ SUBSYS(devices)
 #endif
 
 /* */
+
+#ifdef CONFIG_CGROUP_TASK
+SUBSYS(task_cgroup)
+#endif
Index: b/include/linux/cgroup_task.h
===================================================================
--- /dev/null
+++ b/include/linux/cgroup_task.h
@@ -0,0 +1,33 @@
+/* cgroup_task.h - #task control group
+ *
+ * Copyright (C) 2008 KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at you

KOSAKI Motohiro wrote:
> ChangeLog
> ========================================
> v1 -> v2
>   o implement can_attach() and attach() method.
>   o remvoe can_fork member from cgroup_subsys.
>     instead, copy_process() call task_cgroup_can_fork() directly.
>   o fixed dead lock bug.
>   o add Documentation/contoroller/task.txt
> 

I think Documentation/kernel-parameters.txt should also be updated.

	cgroup_disable= [KNL] Disable a particular controller
			Format: {name of the controller(s) to disable}
				{Currently supported controllers - "memory"}

> 
> benefit
> ========================================
> 1. prevent fork bomb.
> 
> We already have "/prox/sys/kernel/threads-max".
> but it isn't perfect solution.
> because threads-max prevent any process creation.
> then, System-administrator can't login and restore trouble.
> 
> task limit cgroup is better solution.
> it can prevent fork by network service daemon, but allow fork by interactive process.
> 
> 
> 2. help implement batch processing
> 
> in general, batch processing environment need support #task limit.
> this patch help implement it.
> 
> 
> usage
> ========================================
> 
> # mount -t cgroup -o task none /dev/cgroup
> # mkdir /dev/cgroup/foo
> # cd /dev/cgroup/foo
> # ls 
> notify_on_release  task.max_tasks  task.nr_tasks  tasks
> # echo 100 > task.max_tasks
> # echo $$ > tasks
> # fork_bomb 1000 &  <- try create 1000 process
> # pgrep fork_bomb | wc -l
> 98
> 
> 
> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
> CC: Li Zefan <lizf@cn.fujitsu.com>
> CC: Paul Menage <menage@google.com>
> 
> ---
>  Documentation/controllers/task.txt |   18 +++
>  include/linux/cgroup_subsys.h      |    4 
>  include/linux/cgroup_task.h        |   33 +++++
>  init/Kconfig                       |   10 +
>  kernel/Makefile                    |    1 
>  kernel/cgroup_task.c               |  213 +++++++++++++++++++++++++++++++++++++
>  kernel/fork.c                      |    4 
>  7 files changed, 283 insertions(+)
> 
> Index: b/init/Kconfig
> ===================================================================
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -289,6 +289,16 @@ config CGROUP_DEBUG
>  
>  	  Say N if unsure
>  
> +config CGROUP_TASK
> +	bool "Simple number of task accounting cgroup subsystem"
> +	depends on CGROUPS && EXPERIMENTAL
> +	default n
> +	help
> +          Provides a simple number of task accounting cgroup subsystem for

should use TAB

> +	  prevent fork bomb.
> +
> +	  Say N if unsure
> +
>  config CGROUP_NS
>          bool "Namespace cgroup subsystem"
>          depends on CGROUPS
> Index: b/kernel/Makefile
> ===================================================================
> --- a/kernel/Makefile
> +++ b/kernel/Makefile
> @@ -46,6 +46,7 @@ obj-$(CONFIG_KEXEC) += kexec.o
>  obj-$(CONFIG_COMPAT) += compat.o
>  obj-$(CONFIG_CGROUPS) += cgroup.o
>  obj-$(CONFIG_CGROUP_DEBUG) += cgroup_debug.o
> +obj-$(CONFIG_CGROUP_TASK) += cgroup_task.o
>  obj-$(CONFIG_CPUSETS) += cpuset.o
>  obj-$(CONFIG_CGROUP_NS) += ns_cgroup.o
>  obj-$(CONFIG_UTS_NS) += utsname.o
> Index: b/kernel/cgroup_task.c
> ===================================================================
> --- /dev/null
> +++ b/kernel/cgroup_task.c
> @@ -0,0 +1,213 @@
> +/* cgroup_task.c - #task control group
> + *
> + * Copyright (C) 2008 KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program 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 General Public License for more details.
> + */
> +
> +#include <linux/cgroup.h>
> +#include <linux/err.h>
> +#include <linux/cgroup_task.h>
> +
> +struct task_cgroup {
> +	struct cgroup_subsys_state css;
> +	/*
> +	 * the counter to account for number of thread.
> +	 */
> +	int max_tasks;
> +	int nr_tasks;
> +
> +	spinlock_t lock;
> +};
> +
> +struct cgroup_subsys task_cgroup_subsys __read_mostly;
> +
> +static inline struct task_cgroup *task_cgroup_from_task(struct task_struct *p)
> +{
> +	return container_of(task_subsys_state(p, task_cgroup_subsys_id),
> +				struct task_cgroup, css);
> +}
> +
> +static struct task_cgroup *task_cgroup_from_cgrp(struct cgroup *cgrp)
> +{
> +	return container_of(cgroup_subsys_state(cgrp,
> +				task_cgroup_subsys_id), struct task_cgroup,
> +				css);
> +}
> +
> +static int task_cgroup_max_tasks_write(struct cgroup *cgrp,
> +				       struct cftype *cftype,
> +				       s64 new_max_tasks)
> +{
> +	struct task_cgroup *taskcg;
> +	int ret = -EBUSY;
> +
> +	if ((new_max_tasks > INT_MAX) ||
> +	    (new_max_tasks < -1))
> +		return -EINVAL;
> +
> +	taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&taskcg->lock);
> +	if (taskcg->nr_tasks <= new_max_tasks) {
> +		taskcg->max_tasks = new_max_tasks;
> +		ret = 0;
> +	}
> +	spin_unlock(&taskcg->lock);
> +
> +	return ret;
> +}
> +
> +static s64 task_cgroup_max_tasks_read(struct cgroup *cgrp, struct cftype *cft)
> +{
> +	s64 max_tasks;
> +	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&taskcg->lock);
> +	max_tasks = taskcg->max_tasks;
> +	spin_unlock(&taskcg->lock);
> +
> +	return max_tasks;
> +}
> +
> +static s64 task_cgroup_nr_tasks_read(struct cgroup *cgrp, struct cftype *cft)
> +{
> +	s64 nr_tasks;
> +	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&taskcg->lock);
> +	nr_tasks = taskcg->nr_tasks;
> +	spin_unlock(&taskcg->lock);
> +
> +	return nr_tasks;
> +}
> +
> +static struct cftype task_cgroup_files[] =  {
> +	{
> +		.name = "max_tasks",
> +		.write_s64 = task_cgroup_max_tasks_write,
> +		.read_s64  = task_cgroup_max_tasks_read,
> +	},
> +	{
> +		.name = "nr_tasks",
> +		.read_s64  = task_cgroup_nr_tasks_read,
> +	},
> +};
> +
> +static struct cgroup_subsys_state *task_cgroup_create(struct cgroup_subsys *ss,
> +						   struct cgroup *cgrp)
> +{
> +	struct task_cgroup *taskcg;
> +
> +	taskcg = kzalloc(sizeof(struct task_cgroup), GFP_KERNEL);
> +	if (!taskcg)
> +		return ERR_PTR(-ENOMEM);
> +
> +	taskcg->max_tasks = -1;   /* infinite */
> +	spin_lock_init(&taskcg->lock);
> +
> +	return &taskcg->css;
> +}
> +
> +static void task_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
> +{
> +	kfree(cgrp->subsys[task_cgroup_subsys_id]);

You should kfree task_cgroup* instead of cgroup_subsys_state*.

> +}
> +
> +
> +static int task_cgroup_can_attach(struct cgroup_subsys *ss,
> +				  struct cgroup *cgrp, struct task_struct *tsk)
> +{
> +	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
> +	int ret = 0;
> +
> +	spin_lock(&taskcg->lock);
> +	if (taskcg->nr_tasks == taskcg->max_tasks)
> +		ret = -EBUSY;
> +	spin_unlock(&taskcg->lock);
> +
> +	return ret;
> +}
> +
> +static void task_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
> +			       struct cgroup *old_cgrp, struct task_struct *tsk)
> +{
> +	struct task_cgroup *old_taskcg = task_cgroup_from_cgrp(old_cgrp);
> +	struct task_cgroup *new_taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&old_taskcg->lock);
> +	old_taskcg->nr_tasks--;
> +	spin_unlock(&old_taskcg->lock);
> +
> +	spin_lock(&new_taskcg->lock);
> +	new_taskcg->nr_tasks++;
> +	spin_unlock(&new_taskcg->lock);
> +}
> +
> +
> +static int task_cgroup_populate(struct cgroup_subsys *ss,
> +				struct cgroup *cgrp)
> +{
> +	return cgroup_add_files(cgrp, ss, task_cgroup_files,
> +				ARRAY_SIZE(task_cgroup_files));
> +}
> +
> +int task_cgroup_can_fork(struct task_struct *task)
> +{
> +	struct task_cgroup *taskcg;
> +	int max_tasks;
> +	int ret = 0;
> +
> +	if (task_cgroup_subsys.disabled)
> +		return 0;
> +
> +	taskcg = task_cgroup_from_task(task);
> +
> +	spin_lock(&taskcg->lock);
> +	max_tasks = taskcg->max_tasks;
> +
> +	if ((max_tasks >= 0) &&
> +	    (taskcg->nr_tasks >= max_tasks))
> +		ret = -EAGAIN;
> +	else
> +		taskcg->nr_tasks++;
> +	spin_unlock(&taskcg->lock);
> +
> +	return ret;
> +}
> +
> +static void task_cgroup_exit(struct cgroup_subsys *ss, struct task_struct *task)
> +{
> +	struct task_cgroup *taskcg;
> +
> +	if (task_cgroup_subsys.disabled)
> +		return;
> +
> +	taskcg = task_cgroup_from_task(task);
> +
> +	spin_lock(&taskcg->lock);
> +	taskcg->nr_tasks--;
> +	spin_unlock(&taskcg->lock);
> +}
> +
> +
> +struct cgroup_subsys task_cgroup_subsys = {
> +	.name = "task",
> +	.subsys_id = task_cgroup_subsys_id,
> +	.create = task_cgroup_create,
> +	.destroy = task_cgroup_destroy,
> +	.can_

On Sat, Jun 7, 2008 at 4:00 AM, KOSAKI Motohiro
<kosaki.motohiro@jp.fujitsu.com> wrote:
> +
> +#include <linux/cgroup.h>
> +#include <linux/err.h>
> +#include <linux/cgroup_task.h>
> +
> +struct task_cgroup {
> +       struct cgroup_subsys_state css;
> +       /*
> +        * the counter to account for number of thread.
> +        */
> +       int max_tasks;
> +       int nr_tasks;
> +
> +       spinlock_t lock;
> +};

This looks rather like a res_counter. Can you resuse that rather than
implementing your own read/write/charge/uncharge routines?

Paul
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

On Sat, 07 Jun 2008 20:00:00 +0900 KOSAKI Motohiro wrote:

> ---
>  Documentation/controllers/task.txt |   18 +++
>  include/linux/cgroup_subsys.h      |    4 
>  include/linux/cgroup_task.h        |   33 +++++
>  init/Kconfig                       |   10 +
>  kernel/Makefile                    |    1 
>  kernel/cgroup_task.c               |  213 +++++++++++++++++++++++++++++++++++++
>  kernel/fork.c                      |    4 
>  7 files changed, 283 insertions(+)
> 
> Index: b/init/Kconfig
> ===================================================================
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -289,6 +289,16 @@ config CGROUP_DEBUG
>  
>  	  Say N if unsure
>  
> +config CGROUP_TASK
> +	bool "Simple number of task accounting cgroup subsystem"
> +	depends on CGROUPS && EXPERIMENTAL
> +	default n
> +	help
> +          Provides a simple number of task accounting cgroup subsystem for

Indent help test one tab + 2 spaces, please.
s/for/to/

> +	  prevent fork bomb.

s/bomb/bombs/

> +
> +	  Say N if unsure

end with period (".").

> +
>  config CGROUP_NS
>          bool "Namespace cgroup subsystem"
>          depends on CGROUPS

> Index: b/Documentation/controllers/task.txt
> ===================================================================
> --- /dev/null
> +++ b/Documentation/controllers/task.txt
> @@ -0,0 +1,18 @@
> +Task limit Controller
> +
> +1. Description
> +
> +Implement a cgroup to track number of tasks (process and thread).
> +this feature is useful for prevent fork bomb attack.

s/this/This/
s/for/to/
s/attack/attacks/ or s/prevent fork/prevent a fork/

> +
> +
> +2. User Interface
> +
> +ristrict number of tasks to 100.

   Restrict

> +
> +	# echo 100 > /cgroups/foo/task.max_tasks
> +
> +
> +display current number of tasks.

   Display

> +
> +        # cat > /cgroups/foo/task.nr_tasks
> 
> 
> --


---
~Randy
'"Daemon' is an old piece of jargon from the UNIX operating system,
where it referred to a piece of low-level utility software, a
fundamental part of the operating system."
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

Hi Paul,

very sorry, late responce.

> > +struct task_cgroup {
> > +       struct cgroup_subsys_state css;
> > +       /*
> > +        * the counter to account for number of thread.
> > +        */
> > +       int max_tasks;
> > +       int nr_tasks;
> > +
> > +       spinlock_t lock;
> > +};
> 
> This looks rather like a res_counter. Can you resuse that rather than
> implementing your own read/write/charge/uncharge routines?

honestly, I used res_counter on early version.
but I got bad performance.
tus, I changed to current implementation.

Of cource, if res_counter become faster, I'll use it.



_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

Hi Randy,

sorry for so late responce.
REALLY REALLY Thank you for your professional review!

I'll merge it.

> > Index: b/init/Kconfig
> > ===================================================================
> > --- a/init/Kconfig
> > +++ b/init/Kconfig
> > @@ -289,6 +289,16 @@ config CGROUP_DEBUG
> >  
> >  	  Say N if unsure
> >  
> > +config CGROUP_TASK
> > +	bool "Simple number of task accounting cgroup subsystem"
> > +	depends on CGROUPS && EXPERIMENTAL
> > +	default n
> > +	help
> > +          Provides a simple number of task accounting cgroup subsystem for
> 
> Indent help test one tab + 2 spaces, please.
> s/for/to/

(snip)



_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

Hi

> > v1 -> v2
> >   o implement can_attach() and attach() method.
> >   o remvoe can_fork member from cgroup_subsys.
> >     instead, copy_process() call task_cgroup_can_fork() directly.
> >   o fixed dead lock bug.
> >   o add Documentation/contoroller/task.txt
> > 
> 
> I think Documentation/kernel-parameters.txt should also be updated.
> 
> 	cgroup_disable= [KNL] Disable a particular controller
> 			Format: {name of the controller(s) to disable}
> 				{Currently supported controllers - "memory"}

Yup, You are right.
I'll fix it.

> > +config CGROUP_TASK
> > +	bool "Simple number of task accounting cgroup subsystem"
> > +	depends on CGROUPS && EXPERIMENTAL
> > +	default n
> > +	help
> > +          Provides a simple number of task accounting cgroup subsystem for
> 
> should use TAB

Ah, Thanks.


> > +static void task_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
> > +{
> > +	kfree(cgrp->subsys[task_cgroup_subsys_id]);
> 
> You should kfree task_cgroup* instead of cgroup_subsys_state*.

There point to the same address.
but, your opinion has better code clearness, of cource.

Thanks.



_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

On Sat, 2008-06-07 at 20:00 +0900, KOSAKI Motohiro wrote:
> ChangeLog
> ========================================
> v1 -> v2
>   o implement can_attach() and attach() method.
>   o remvoe can_fork member from cgroup_subsys.
>     instead, copy_process() call task_cgroup_can_fork() directly.
>   o fixed dead lock bug.
>   o add Documentation/contoroller/task.txt
> 
> 
> benefit
> ========================================
> 1. prevent fork bomb.
> 
> We already have "/prox/sys/kernel/threads-max".
> but it isn't perfect solution.
> because threads-max prevent any process creation.
> then, System-administrator can't login and restore trouble.
> 
> task limit cgroup is better solution.
> it can prevent fork by network service daemon, but allow fork by interactive process.
> 
> 
> 2. help implement batch processing
> 
> in general, batch processing environment need support #task limit.
> this patch help implement it.
> 
> 
> usage
> ========================================
> 
> # mount -t cgroup -o task none /dev/cgroup
> # mkdir /dev/cgroup/foo
> # cd /dev/cgroup/foo
> # ls 
> notify_on_release  task.max_tasks  task.nr_tasks  tasks
> # echo 100 > task.max_tasks
> # echo $$ > tasks
> # fork_bomb 1000 &  <- try create 1000 process
> # pgrep fork_bomb | wc -l
> 98

fwiw I think the name sucks!

cgroups are about grouping tasks, so what's a task cgroup?


> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
> CC: Li Zefan <lizf@cn.fujitsu.com>
> CC: Paul Menage <menage@google.com>
> 
> ---
>  Documentation/controllers/task.txt |   18 +++
>  include/linux/cgroup_subsys.h      |    4 
>  include/linux/cgroup_task.h        |   33 +++++
>  init/Kconfig                       |   10 +
>  kernel/Makefile                    |    1 
>  kernel/cgroup_task.c               |  213 +++++++++++++++++++++++++++++++++++++
>  kernel/fork.c                      |    4 
>  7 files changed, 283 insertions(+)
> 
> Index: b/init/Kconfig
> ===================================================================
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -289,6 +289,16 @@ config CGROUP_DEBUG
>  
>  	  Say N if unsure
>  
> +config CGROUP_TASK
> +	bool "Simple number of task accounting cgroup subsystem"
> +	depends on CGROUPS && EXPERIMENTAL
> +	default n
> +	help
> +          Provides a simple number of task accounting cgroup subsystem for
> +	  prevent fork bomb.
> +
> +	  Say N if unsure
> +
>  config CGROUP_NS
>          bool "Namespace cgroup subsystem"
>          depends on CGROUPS
> Index: b/kernel/Makefile
> ===================================================================
> --- a/kernel/Makefile
> +++ b/kernel/Makefile
> @@ -46,6 +46,7 @@ obj-$(CONFIG_KEXEC) += kexec.o
>  obj-$(CONFIG_COMPAT) += compat.o
>  obj-$(CONFIG_CGROUPS) += cgroup.o
>  obj-$(CONFIG_CGROUP_DEBUG) += cgroup_debug.o
> +obj-$(CONFIG_CGROUP_TASK) += cgroup_task.o
>  obj-$(CONFIG_CPUSETS) += cpuset.o
>  obj-$(CONFIG_CGROUP_NS) += ns_cgroup.o
>  obj-$(CONFIG_UTS_NS) += utsname.o
> Index: b/kernel/cgroup_task.c
> ===================================================================
> --- /dev/null
> +++ b/kernel/cgroup_task.c
> @@ -0,0 +1,213 @@
> +/* cgroup_task.c - #task control group
> + *
> + * Copyright (C) 2008 KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program 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 General Public License for more details.
> + */
> +
> +#include <linux/cgroup.h>
> +#include <linux/err.h>
> +#include <linux/cgroup_task.h>
> +
> +struct task_cgroup {
> +	struct cgroup_subsys_state css;
> +	/*
> +	 * the counter to account for number of thread.
> +	 */
> +	int max_tasks;
> +	int nr_tasks;
> +
> +	spinlock_t lock;
> +};
> +
> +struct cgroup_subsys task_cgroup_subsys __read_mostly;
> +
> +static inline struct task_cgroup *task_cgroup_from_task(struct task_struct *p)
> +{
> +	return container_of(task_subsys_state(p, task_cgroup_subsys_id),
> +				struct task_cgroup, css);
> +}
> +
> +static struct task_cgroup *task_cgroup_from_cgrp(struct cgroup *cgrp)
> +{
> +	return container_of(cgroup_subsys_state(cgrp,
> +				task_cgroup_subsys_id), struct task_cgroup,
> +				css);
> +}
> +
> +static int task_cgroup_max_tasks_write(struct cgroup *cgrp,
> +				       struct cftype *cftype,
> +				       s64 new_max_tasks)
> +{
> +	struct task_cgroup *taskcg;
> +	int ret = -EBUSY;
> +
> +	if ((new_max_tasks > INT_MAX) ||
> +	    (new_max_tasks < -1))
> +		return -EINVAL;
> +
> +	taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&taskcg->lock);
> +	if (taskcg->nr_tasks <= new_max_tasks) {
> +		taskcg->max_tasks = new_max_tasks;
> +		ret = 0;
> +	}
> +	spin_unlock(&taskcg->lock);
> +
> +	return ret;
> +}
> +
> +static s64 task_cgroup_max_tasks_read(struct cgroup *cgrp, struct cftype *cft)
> +{
> +	s64 max_tasks;
> +	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&taskcg->lock);
> +	max_tasks = taskcg->max_tasks;
> +	spin_unlock(&taskcg->lock);
> +
> +	return max_tasks;
> +}
> +
> +static s64 task_cgroup_nr_tasks_read(struct cgroup *cgrp, struct cftype *cft)
> +{
> +	s64 nr_tasks;
> +	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&taskcg->lock);
> +	nr_tasks = taskcg->nr_tasks;
> +	spin_unlock(&taskcg->lock);
> +
> +	return nr_tasks;
> +}
> +
> +static struct cftype task_cgroup_files[] =  {
> +	{
> +		.name = "max_tasks",
> +		.write_s64 = task_cgroup_max_tasks_write,
> +		.read_s64  = task_cgroup_max_tasks_read,
> +	},
> +	{
> +		.name = "nr_tasks",
> +		.read_s64  = task_cgroup_nr_tasks_read,
> +	},
> +};
> +
> +static struct cgroup_subsys_state *task_cgroup_create(struct cgroup_subsys *ss,
> +						   struct cgroup *cgrp)
> +{
> +	struct task_cgroup *taskcg;
> +
> +	taskcg = kzalloc(sizeof(struct task_cgroup), GFP_KERNEL);
> +	if (!taskcg)
> +		return ERR_PTR(-ENOMEM);
> +
> +	taskcg->max_tasks = -1;   /* infinite */
> +	spin_lock_init(&taskcg->lock);
> +
> +	return &taskcg->css;
> +}
> +
> +static void task_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
> +{
> +	kfree(cgrp->subsys[task_cgroup_subsys_id]);
> +}
> +
> +
> +static int task_cgroup_can_attach(struct cgroup_subsys *ss,
> +				  struct cgroup *cgrp, struct task_struct *tsk)
> +{
> +	struct task_cgroup *taskcg = task_cgroup_from_cgrp(cgrp);
> +	int ret = 0;
> +
> +	spin_lock(&taskcg->lock);
> +	if (taskcg->nr_tasks == taskcg->max_tasks)
> +		ret = -EBUSY;
> +	spin_unlock(&taskcg->lock);
> +
> +	return ret;
> +}
> +
> +static void task_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
> +			       struct cgroup *old_cgrp, struct task_struct *tsk)
> +{
> +	struct task_cgroup *old_taskcg = task_cgroup_from_cgrp(old_cgrp);
> +	struct task_cgroup *new_taskcg = task_cgroup_from_cgrp(cgrp);
> +
> +	spin_lock(&old_taskcg->lock);
> +	old_taskcg->nr_tasks--;
> +	spin_unlock(&old_taskcg->lock);
> +
> +	spin_lock(&new_taskcg->lock);
> +	new_taskcg->nr_tasks++;
> +	spin_unlock(&new_taskcg->lock);
> +}
> +
> +
> +static int task_cgroup_populate(struct cgroup_subsys *ss,
> +				struct cgroup *cgrp)
> +{
> +	return cgroup_add_files(cgrp, ss, task_cgroup_files,
> +				ARRAY_SIZE(task_cgroup_files));
> +}
> +
> +int task_cgroup_can_fork(struct task_struct *task)
> +{
> +	struct task_cgroup *taskcg;
> +	int max_tasks;
> +	int ret = 0;
> +
> +	if (task_cgroup_subsys.disabled)
> +		return 0;
> +
> +	taskcg = task_cgroup_from_task(task);
> +
> +	spin_lock(&taskcg->lock);
> +	max_tasks = taskcg->max_tasks;
> +
> +	if ((max_tasks >= 0) &&
> +	    (taskcg->nr_tasks >= max_tasks))
> +		ret = -EAGAIN;
> +	else
> +		taskcg->nr_tasks++;
> +	spin_unlock(&taskcg->lock);
> +
> +	return ret;
> +}
> +
> +static void task_cgroup_exit(struct cgroup_subsys *ss, struct task_struct *task)
> +{
> +	struct task_cgroup *taskcg;
> +
> +	if (task_cgroup_subsys.disabled)
> +		return;
> +
> +	taskcg = task_cgroup_from_task(task);
> +
> +	spin_lock(&taskcg->lock);
> +	taskcg->nr_tasks--;
> +	spin_unlock(&taskcg->lock);
> +}
> +
> +
> +struct cgroup_subsys task_cgroup_subsys = {
> +	.name = "task",
> +	.subsys_id = task_cgroup_subsys_id,
> +	.create = task_cgroup_create,
> +	.destroy = task_cgroup_destroy,
> +	.can_attach = task_cgroup_can_attach,
> +	.attach = task_cgroup_attach,
> +	.populate = task_cgroup_populate,
> +	.exit = task_cgroup_exit,
> +	.early_init = 0,
> +};
> +
> Index: b/kern

Hi Peter,

> > # mount -t cgroup -o task none /dev/cgroup
> > # mkdir /dev/cgroup/foo
> > # cd /dev/cgroup/foo
> > # ls 
> > notify_on_release  task.max_tasks  task.nr_tasks  tasks
> > # echo 100 > task.max_tasks
> > # echo $$ > tasks
> > # fork_bomb 1000 &  <- try create 1000 process
> > # pgrep fork_bomb | wc -l
> > 98
> 
> fwiw I think the name sucks!
> cgroups are about grouping tasks, so what's a task cgroup?

Sorry, my english skill is very low.
I hope restrict number of tasks.

Do you have any name idea?



_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

I guess you didn't notice this comment ? :)

>> --- a/kernel/fork.c
>> > +++ b/kernel/fork.c
>> > @@ -54,6 +54,7 @@
>> >  #include <linux/tty.h>
>> >  #include <linux/proc_fs.h>
>> >  #include <linux/blkdev.h>
>> > +#include <linux/cgroup_task.h>
>> >  
>> >  #include <asm/pgtable.h>
>> >  #include <asm/pgalloc.h>
>> > @@ -920,6 +921,8 @@ static struct task_struct *copy_process(
>> >  		    p->user != current->nsproxy->user_ns->root_user)
>> >  			goto bad_fork_free;
>> >  	}
>> > +	if (task_cgroup_can_fork(p))
>> > +		goto bad_fork_free;
> 
> If task_cgroup_can_fork() returns 0, but copy_process() fails afterwords,
> taskcg->nr_tasks will be in a wrong state.
> 

_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

On Mon, 2008-06-16 at 15:59 +0900, KOSAKI Motohiro wrote:
> Hi Peter,
> 
> > > # mount -t cgroup -o task none /dev/cgroup
> > > # mkdir /dev/cgroup/foo
> > > # cd /dev/cgroup/foo
> > > # ls 
> > > notify_on_release  task.max_tasks  task.nr_tasks  tasks
> > > # echo 100 > task.max_tasks
> > > # echo $$ > tasks
> > > # fork_bomb 1000 &  <- try create 1000 process
> > > # pgrep fork_bomb | wc -l
> > > 98
> > 
> > fwiw I think the name sucks!
> > cgroups are about grouping tasks, so what's a task cgroup?
> 
> Sorry, my english skill is very low.
> I hope restrict number of tasks.
> 
> Do you have any name idea?

Sadly I suck at naming too ;-(

perhaps task_limit controller, or nr_tasks controller.

We should preferably pick one that transfers to most other rlimit like
controls.

_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

> I guess you didn't notice this comment ? :)

Agghhh, you are right.
sorry.

> >> --- a/kernel/fork.c
> >> > +++ b/kernel/fork.c
> >> > @@ -54,6 +54,7 @@
> >> >  #include <linux/tty.h>
> >> >  #include <linux/proc_fs.h>
> >> >  #include <linux/blkdev.h>
> >> > +#include <linux/cgroup_task.h>
> >> >  
> >> >  #include <asm/pgtable.h>
> >> >  #include <asm/pgalloc.h>
> >> > @@ -920,6 +921,8 @@ static struct task_struct *copy_process(
> >> >  		    p->user != current->nsproxy->user_ns->root_user)
> >> >  			goto bad_fork_free;
> >> >  	}
> >> > +	if (task_cgroup_can_fork(p))
> >> > +		goto bad_fork_free;
> > 
> > If task_cgroup_can_fork() returns 0, but copy_process() fails afterwords,
> > taskcg->nr_tasks will be in a wrong state.

Sure.
I'll fix it, of cource.

Thanks!



_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

Peter Zijlstra wrote:
> On Mon, 2008-06-16 at 15:59 +0900, KOSAKI Motohiro wrote:
>> Hi Peter,
>>
>>>> # mount -t cgroup -o task none /dev/cgroup
>>>> # mkdir /dev/cgroup/foo
>>>> # cd /dev/cgroup/foo
>>>> # ls 
>>>> notify_on_release  task.max_tasks  task.nr_tasks  tasks
>>>> # echo 100 > task.max_tasks
>>>> # echo $$ > tasks
>>>> # fork_bomb 1000 &  <- try create 1000 process
>>>> # pgrep fork_bomb | wc -l
>>>> 98
>>> fwiw I think the name sucks!
>>> cgroups are about grouping tasks, so what's a task cgroup?
>> Sorry, my english skill is very low.
>> I hope restrict number of tasks.
>>
>> Do you have any name idea?
> 
> Sadly I suck at naming too ;-(
> 
> perhaps task_limit controller, or nr_tasks controller.
> 
> We should preferably pick one that transfers to most other rlimit like
> controls.
> 

We have memrlimit in -mm currently, which controls address space.
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

> > > fwiw I think the name sucks!
> > > cgroups are about grouping tasks, so what's a task cgroup?
> > 
> > Sorry, my english skill is very low.
> > I hope restrict number of tasks.
> > 
> > Do you have any name idea?
> 
> Sadly I suck at naming too ;-(
> 
> perhaps task_limit controller, or nr_tasks controller.
> 
> We should preferably pick one that transfers to most other rlimit like
> controls.

Yeah, thanks.
I like "task_limit" controller.

I'll change to it.

very thanks.


_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

On Sun, Jun 15, 2008 at 6:32 PM, KOSAKI Motohiro
<kosaki.motohiro@jp.fujitsu.com> wrote:
>
> honestly, I used res_counter on early version.
> but I got bad performance.

Bad performance on the charge/uncharge?

The only difference I can see is that res_counter uses
spin_lock_irqsave()/spin_unlock_irqrestore(), and you're using plain
spin_lock()/spin_unlock().

Is the overhead of a pushf/cli/popf really going to matter compared
with the overhead of forking/exiting a task?

Or approaching this from the other side, does res_counter really need
irq-safe locking, or is it just being cautious?

Paul
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

Paul Menage wrote:
> On Sun, Jun 15, 2008 at 6:32 PM, KOSAKI Motohiro
> <kosaki.motohiro@jp.fujitsu.com> wrote:
>> honestly, I used res_counter on early version.
>> but I got bad performance.
> 
> Bad performance on the charge/uncharge?
> 
> The only difference I can see is that res_counter uses
> spin_lock_irqsave()/spin_unlock_irqrestore(), and you're using plain
> spin_lock()/spin_unlock().
> 
> Is the overhead of a pushf/cli/popf really going to matter compared
> with the overhead of forking/exiting a task?
> 
> Or approaching this from the other side, does res_counter really need
> irq-safe locking, or is it just being cautious?

We really need irq-safe locking. We can end up uncharging from reclaim context
(called under zone->lru_lock and mem->zone->lru_lock - held with interrupts
disabled)

I am going to convert the spin lock to a reader writers lock, so that reads from
user space do not cause contention. I'll experiment and look at the overhead.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

> > Bad performance on the charge/uncharge?
> > 
> > The only difference I can see is that res_counter uses
> > spin_lock_irqsave()/spin_unlock_irqrestore(), and you're using plain
> > spin_lock()/spin_unlock().
> > 
> > Is the overhead of a pushf/cli/popf really going to matter compared
> > with the overhead of forking/exiting a task?
> > 
> > Or approaching this from the other side, does res_counter really need
> > irq-safe locking, or is it just being cautious?
> 
> We really need irq-safe locking. We can end up uncharging from reclaim context
> (called under zone->lru_lock and mem->zone->lru_lock - held with interrupts
> disabled)
> 
> I am going to convert the spin lock to a reader writers lock, so that reads from
> user space do not cause contention. I'll experiment and look at the overhead.

Sorry, late responce.
I'm working on fix current -mm tree regression recently ;)

Note:

I am going to convert spinlock in task limit cgroup to atomic_t.
task limit cgroup has following caractatics.
	- many write (fork, exit)
	- few read 
	- fork() is performance sensitive systemcall.
	  if increase fork overhead, system total performance cause degression.



_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

On Sat, Jun 21, 2008 at 2:10 AM, KOSAKI Motohiro
<kosaki.motohiro@jp.fujitsu.com> wrote:
>
> I am going to convert spinlock in task limit cgroup to atomic_t.
> task limit cgroup has following caractatics.
>        - many write (fork, exit)
>        - few read
>        - fork() is performance sensitive systemcall.

This is true, but I don't see how it can be more performance-sensitive
than the overhead of allocating/freeing a page.

What kinds of performance regressions did you see?

>          if increase fork overhead, system total performance cause degression.

What kind of overhead were you seeing? How about if you delay doing
any task accounting until the task_limit subsystem is bound to a
hierarchy? That way there's no noticeable overhead for people who
aren't using your subsystem.

Paul
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

>> I am going to convert spinlock in task limit cgroup to atomic_t.
>> task limit cgroup has following caractatics.
>>        - many write (fork, exit)
>>        - few read
>>        - fork() is performance sensitive systemcall.
>
> This is true, but I don't see how it can be more performance-sensitive
> than the overhead of allocating/freeing a page.
>
> What kinds of performance regressions did you see?

I ran spawn test of unix bench, thus

implement way                   performance degression
-------------------------------------------------------------------------
use res_counter                  15-20%
use spin_lock()                    nealy 10%
use atomic_t                       nealy 5%

Yes, this is really roughly number.
Of cource, I'll post more detail result at next week.

>>          if increase fork overhead, system total performance cause degression.
>
> What kind of overhead were you seeing? How about if you delay doing
> any task accounting until the task_limit subsystem is bound to a
> hierarchy? That way there's no noticeable overhead for people who
> aren't using your subsystem.

honestly, I am seeing it on micro-benchmark only.
but, I'm afraid to performance degression because many people check
performance regression periodically.
So if my implement cause performance regression, they never used mine.

Or, if you strongly want to task_limit subsystem use res_counter,
I can be working on improve to res_counter performance instead.
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

KOSAKI Motohiro wrote:
> Or, if you strongly want to task_limit subsystem use res_counter,
> I can be working on improve to res_counter performance instead.

I would prefer that as it would help a common cause and avoid
duplication/maintenance overhead.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

>> Or, if you strongly want to task_limit subsystem use res_counter,
>> I can be working on improve to res_counter performance instead.
>
> I would prefer that as it would help a common cause and avoid
> duplication/maintenance overhead.

okey :)
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers