OpenVZ Forum


Home » Mailing lists » Devel » [RFC][PATCH 0/7] Resource controllers based on process containers
Re: [RFC][PATCH 2/7] RSS controller core [message #10914 is a reply to message #10909] Wed, 07 March 2007 07:25 Go to previous messageGo to previous message
xemul is currently offline  xemul
Messages: 248
Registered: November 2005
Senior Member
Balbir Singh wrote:
> Pavel Emelianov wrote:
>> This includes setup of RSS container within generic
>> process containers, all the declarations used in RSS
>> accounting, and core code responsible for accounting.
>>
>>
>> ------------------------------------------------------------ ------------
>>
>> diff -upr linux-2.6.20.orig/include/linux/rss_container.h
>> linux-2.6.20-0/include/linux/rss_container.h
>> --- linux-2.6.20.orig/include/linux/rss_container.h 2007-03-06
>> 13:39:17.000000000 +0300
>> +++ linux-2.6.20-0/include/linux/rss_container.h 2007-03-06
>> 13:33:28.000000000 +0300
>> @@ -0,0 +1,68 @@
>> +#ifndef __RSS_CONTAINER_H__
>> +#define __RSS_CONTAINER_H__
>> +/*
>> + * RSS container
>> + *
>> + * Copyright 2007 OpenVZ SWsoft Inc
>> + *
>> + * Author: Pavel Emelianov <xemul@openvz.org>
>> + *
>> + */
>> +
>> +struct page_container;
>> +struct rss_container;
>> +
>> +#ifdef CONFIG_RSS_CONTAINER
>> +int container_rss_prepare(struct page *, struct vm_area_struct *vma,
>> + struct page_container **);
>> +
>> +void container_rss_add(struct page_container *);
>> +void container_rss_del(struct page_container *);
>> +void container_rss_release(struct page_container *);
>> +
>> +int mm_init_container(struct mm_struct *mm, struct task_struct *tsk);
>> +void mm_free_container(struct mm_struct *mm);
>> +
>> +unsigned long container_isolate_pages(unsigned long nr_to_scan,
>> + struct rss_container *rss, struct list_head *dst,
>> + int active, unsigned long *scanned);
>> +unsigned long container_nr_physpages(struct rss_container *rss);
>> +
>> +unsigned long container_try_to_free_pages(struct rss_container *);
>> +void container_out_of_memory(struct rss_container *);
>> +
>> +void container_rss_init_early(void);
>> +#else
>> +static inline int container_rss_prepare(struct page *pg,
>> + struct vm_area_struct *vma, struct page_container **pc)
>> +{
>> + *pc = NULL; /* to make gcc happy */
>> + return 0;
>> +}
>> +
>> +static inline void container_rss_add(struct page_container *pc)
>> +{
>> +}
>> +
>> +static inline void container_rss_del(struct page_container *pc)
>> +{
>> +}
>> +
>> +static inline void container_rss_release(struct page_container *pc)
>> +{
>> +}
>> +
>> +static inline int mm_init_container(struct mm_struct *mm, struct
>> task_struct *t)
>> +{
>> + return 0;
>> +}
>> +
>> +static inline void mm_free_container(struct mm_struct *mm)
>> +{
>> +}
>> +
>> +static inline void container_rss_init_early(void)
>> +{
>> +}
>> +#endif
>> +#endif
>> diff -upr linux-2.6.20.orig/init/Kconfig linux-2.6.20-0/init/Kconfig
>> --- linux-2.6.20.orig/init/Kconfig 2007-03-06 13:33:28.000000000 +0300
>> +++ linux-2.6.20-0/init/Kconfig 2007-03-06 13:33:28.000000000 +0300
>> @@ -265,6 +265,13 @@ config CPUSETS
>> bool
>> select CONTAINERS
>>
>> +config RSS_CONTAINER
>> + bool "RSS accounting container"
>> + select RESOURCE_COUNTERS
>> + help
>> + Provides a simple Resource Controller for monitoring and
>> + controlling the total Resident Set Size of the tasks in a
>> container
>> +
>
> The wording looks very familiar :-). It would be useful to add
> "The reclaim logic is now container aware, when the container goes
> overlimit
> the page reclaimer reclaims pages belonging to this container. If we are
> unable to reclaim enough pages to satisfy the request, the process is
> killed with an out of memory warning"

OK. Thanks.

>
>> config SYSFS_DEPRECATED
>> bool "Create deprecated sysfs files"
>> default y
>> diff -upr linux-2.6.20.orig/mm/Makefile linux-2.6.20-0/mm/Makefile
>> --- linux-2.6.20.orig/mm/Makefile 2007-02-04 21:44:54.000000000 +0300
>> +++ linux-2.6.20-0/mm/Makefile 2007-03-06 13:33:28.000000000 +0300
>> @@ -29,3 +29,5 @@ obj-$(CONFIG_MEMORY_HOTPLUG) += memory_h
>> obj-$(CONFIG_FS_XIP) += filemap_xip.o
>> obj-$(CONFIG_MIGRATION) += migrate.o
>> obj-$(CONFIG_SMP) += allocpercpu.o
>> +
>> +obj-$(CONFIG_RSS_CONTAINER) += rss_container.o
>> diff -upr linux-2.6.20.orig/mm/rss_container.c
>> linux-2.6.20-0/mm/rss_container.c
>> --- linux-2.6.20.orig/mm/rss_container.c 2007-03-06
>> 13:39:17.000000000 +0300
>> +++ linux-2.6.20-0/mm/rss_container.c 2007-03-06 13:33:28.000000000
>> +0300
>> @@ -0,0 +1,307 @@
>> +/*
>> + * RSS accounting container
>> + *
>> + * Copyright 2007 OpenVZ SWsoft Inc
>> + *
>> + * Author: Pavel Emelianov <xemul@openvz.org>
>> + *
>> + */
>> +
>> +#include <linux/list.h>
>> +#include <linux/sched.h>
>> +#include <linux/mm.h>
>> +#include <linux/res_counter.h>
>> +#include <linux/rss_container.h>
>> +
>> +static struct container_subsys rss_subsys;
>> +
>> +struct rss_container {
>> + struct res_counter res;
>> + struct list_head page_list;
>> + struct container_subsys_state css;
>> +};
>> +
>> +struct page_container {
>> + struct page *page;
>> + struct rss_container *cnt;
>> + struct list_head list;
>> +};
>> +
>
> Yes, this is what I was planning to get to -- a per container LRU list.
> But you have just one list, don't you need active and inactive lists?
> When the global LRU is manipulated, shouldn't this list be updated as
> well, so that reclaim will pick the best pages.
>
>> +static inline struct rss_container *rss_from_cont(struct container *cnt)
>> +{
>> + return container_of(container_subsys_state(cnt, &rss_subsys),
>> + struct rss_container, css);
>> +}
>> +
>> +int mm_init_container(struct mm_struct *mm, struct task_struct *tsk)
>> +{
>> + struct rss_container *cnt;
>> +
>> + cnt = rss_from_cont(task_container(tsk, &rss_subsys));
>> + if (css_get(&cnt->css))
>> + return -EBUSY;
>> +
>> + mm->rss_container = cnt;
>> + return 0;
>> +}
>> +
>> +void mm_free_container(struct mm_struct *mm)
>> +{
>> + css_put(&mm->rss_container->css);
>> +}
>> +
>> +int container_rss_prepare(struct page *page, struct vm_area_struct *vma,
>> + struct page_container **ppc)
>> +{
>> + struct rss_container *rss;
>> + struct page_container *pc;
>> +
>> + rcu_read_lock();
>> + rss = rcu_dereference(vma->vm_mm->rss_container);
>> + css_get_current(&rss->css);
>> + rcu_read_unlock();
>> +
>> + pc = kmalloc(sizeof(struct page_container), GFP_KERNEL);
>> + if (pc == NULL)
>> + goto out_nomem;
>> +
>> + while (res_counter_charge(&rss->res, 1)) {
>> + if (container_try_to_free_pages(rss))
>> + continue;
>> +
>
> The return codes of the functions is a bit confusing, ideally
> container_try_to_free_pages() should return 0 on success. Also

This returns exactly what try_to_free_pages() does.

> res_counter_charge() has a WARN_ON(1) if the limit is exceeded.

Nope - res_counter_uncharge() has - this is an absolutely
sane check that we haven't over-uncharged resources.

> The system administrator can figure out the details from failcnt,
> I suspect when the container is running close to it's limit,
> dmesg will have too many WARNING messages.
>
> How much memory do you try to reclaim in container_try_to_free_pages()?

At least one page. This is enough to make one page charge.
That's the difference from general try_to_free_pages() that
returns success if it freed swap_cluster_max pages at least.

> With my patches, I was planning to export this knob to userspace with
> a default value. This will help the administrator decide how much
> of the working set/container LRU should be freed on reaching the limit.
> I cannot find the definition of container_try_to_free_pages() in
> this patch.

This is in patch #5.
Sorry for such a bad split - I'll make it cleaner next time :)

>
>
>> + container_out_of_memory(rss);
>> + if (test_thread_flag(TIF_MEMDIE))
>> + goto out_charge;
>> + }
>> +
>> + pc->page = page;
>> + pc->cnt = rss;
>> + *ppc = pc;
>> + return 0;
>> +
>> +out_charge:
>> + kfree(pc);
>> +out_nomem:
>> + css_put(&rss->css);
>> + return -ENOMEM;
>> +}
>> +
>> +void container_rss_release(struct page_container *pc)
>> +{
>> + struct rss_container *rss;
>> +
>> + rss = pc->cnt;
>> + res_counter_uncharge(&rss->res, 1);
>> + css_put(&rss->css);
>> + kfree(pc);
>> +}
>> +
>> +void container_rss_add(struct page_container *pc)
>> +{
>> + struct page *pg;
>> + struct rss_container *rss;
>> +
>> + pg = pc->page;
>> + rss = pc->cnt;
>> +
>> + spin_lock(&r ...

[ Show the rest of the message ]

Report message to a moderator

[Message index]
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: Re: [ckrm-tech] [PATCH 7/7] containers (V7): Container interface to nsproxy subsystem
Next Topic: Linux-VServer example results for sharing vs. separate mappings ...
Goto Forum:
  

[ Syndicate this forum (XML) ] [ RSS ] [ PDF ]

Current Time: Tue Nov 19 01:23:53 GMT 2024

Total time taken to generate the page: 0.03040 seconds
Powered by FUDforum Powered by Virtuozzo Containers