[PATCH 0/3][RFC] Containers: Pagecache accounting and control subsystem (v1)

[PATCH 0/3][RFC] Containers: Pagecache accounting and control subsystem (v1)

[Vaidyanathan Srinivasan]

Containers: Pagecache accounting and control subsystem (v1)
-----------------------------------------------------------

This patch adds pagecache accounting and control on top of
Paul's container subsystem v7 posted at
http://lkml.org/lkml/2007/2/12/88

and Balbir's RSS controller posted at
http://lkml.org/lkml/2007/2/26/8

This patchset depends on Balbir's RSS controller and cannot
work independent of it. The page reclaim code has been merged
with container RSS controller.  However compile time options
can individually enable/disable memory controller and/or
pagecache controller.

Comments, suggestions and criticisms are welcome.

Features:
--------
* New subsystem called 'pagecache_acct' is registered with containers
* Container pointer is added to struct address_space to keep track of
  associated container
* In filemap.c and swap_state.c, the corresponding container's
  pagecache_acct subsystem is charged and uncharged whenever a new
  page is added or removed from pagecache
* The accounting number include pages in swap cache and filesystem
  buffer pages apart from pagecache, basically everything under
  NR_FILE_PAGES is counted as pagecache.  However this excluded
  mapped and anonymous pages
* Limits on pagecache can be set by echo 100000 > pagecache_limit on
  the /container file system.  The unit is in kilobytes
* If the pagecache utilisation limit is exceeded, pagecache reclaim
  code is invoked to recover dirty and clean pagecache pages only.

Advantages:
-----------
* Does not add container pointers in struct page

Limitations:
-----------
* Code is not safe for container deletion/task migration
* Pagecache page reclaim needs performance improvements
* Global LRU is churned in search of pagecache pages

Usage:
-----

* Add patch on top of Paul container (v7) at kernel version 2.6.20
* Enable CONFIG_CONTAINER_PAGECACHE_ACCT in 'General Setup'
* Boot new kernel
* Mount container filesystem
	mount -t container /container
	cd /container
* Create new container
	mkdir mybox
	cd /container/mybox
* Add current shell to container
	echo $$ > tasks
* There are two files pagecache_usage and pagecache_limit
* In order to set limit, echo value in kilobytes to pagecache_limit
	echo 100000 > pagecache_limit
	#This would set 100MB limit on pagecache usage
* Trash the system from current shell using scp/cp/dd/tar etc
* Watch pagecache_usage and /proc/meminfo to verify behavior

* Only unmapped pagecache data will be accounted and controlled.
  These are memory used by cp, scp, tar etc.  While file mmap will
  be controlled by Balbir's RSS controller.

Tests:
------

* Ran kernbench within container with pagecache_limits set

ToDo:
----

* Merge with container RSS controller and eliminate redundant code
* Test and support task migration and container deletion
* Review reclaim performance
* Optimise page reclaim

Patch Series:
-------------
pagecache-controller-setup.patch
pagecache-controller-acct.patch
pagecache-controller-reclaim.patch

---	

[PATCH 0/3][RFC] Containers: Pagecache accounting and control subsystem (v1)

[Vaidyanathan Srinivasan]

Containers: Pagecache accounting and control subsystem (v1)
-----------------------------------------------------------

This patch adds pagecache accounting and control on top of
Paul's container subsystem v7 posted at
http://lkml.org/lkml/2007/2/12/88

and Balbir's RSS controller posted at
http://lkml.org/lkml/2007/2/26/8

This patchset depends on Balbir's RSS controller and cannot
work independent of it. The page reclaim code has been merged
with container RSS controller.  However compile time options
can individually enable/disable memory controller and/or
pagecache controller.

Comments, suggestions and criticisms are welcome.

Features:
--------
* New subsystem called 'pagecache_acct' is registered with containers
* Container pointer is added to struct address_space to keep track of
  associated container
* In filemap.c and swap_state.c, the corresponding container's
  pagecache_acct subsystem is charged and uncharged whenever a new
  page is added or removed from pagecache
* The accounting number include pages in swap cache and filesystem
  buffer pages apart from pagecache, basically everything under
  NR_FILE_PAGES is counted as pagecache.  However this excluded
  mapped and anonymous pages
* Limits on pagecache can be set by echo 100000 > pagecache_limit on
  the /container file system.  The unit is in kilobytes
* If the pagecache utilisation limit is exceeded, pagecache reclaim
  code is invoked to recover dirty and clean pagecache pages only.

Advantages:
-----------
* Does not add container pointers in struct page

Limitations:
-----------
* Code is not safe for container deletion/task migration
* Pagecache page reclaim needs performance improvements
* Global LRU is churned in search of pagecache pages

Usage:
-----

* Add patch on top of Paul container (v7) at kernel version 2.6.20
* Enable CONFIG_CONTAINER_PAGECACHE_ACCT in 'General Setup'
* Boot new kernel
* Mount container filesystem
	mount -t container /container
	cd /container
* Create new container
	mkdir mybox
	cd /container/mybox
* Add current shell to container
	echo $$ > tasks
* There are two files pagecache_usage and pagecache_limit
* In order to set limit, echo value in kilobytes to pagecache_limit
	echo 100000 > pagecache_limit
	#This would set 100MB limit on pagecache usage
* Trash the system from current shell using scp/cp/dd/tar etc
* Watch pagecache_usage and /proc/meminfo to verify behavior

* Only unmapped pagecache data will be accounted and controlled.
  These are memory used by cp, scp, tar etc.  While file mmap will
  be controlled by Balbir's RSS controller.

Tests:
------

* Ran kernbench within container with pagecache_limits set

ToDo:
----

* Merge with container RSS controller and eliminate redundant code
* Test and support task migration and container deletion
* Review reclaim performance
* Optimise page reclaim

Patch Series:
-------------
pagecache-controller-setup.patch
pagecache-controller-acct.patch
pagecache-controller-reclaim.patch

---	

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 1/3][RFC] Containers: Pagecache controller setup

[Vaidyanathan Srinivasan]

This patch adds basic infrastructure for pagecache accounting and
control subsystem within the container framework.

pagecache_usage and pagecache_limit files are created along with
every new container.  Routines to extract subsystem pointers and
container pointers have been added.

Compile time kernel options and have been added along with Makefile
changes.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
---
 include/linux/pagecache_acct.h |   53 +++++
 init/Kconfig                   |    7
 mm/Makefile                    |    1
 mm/pagecache_acct.c            |  368 +++++++++++++++++++++++++++++++++++++++++
 4 files changed, 429 insertions(+)

--- /dev/null
+++ linux-2.6.20/include/linux/pagecache_acct.h
@@ -0,0 +1,53 @@
+/*
+ * Pagecache controller - "Account and control pagecache usage"
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright IBM Corporation, 2007
+ *
+ * Author: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
+ *
+ */
+
+#ifndef _LINUX_PAGECACHE_ACCT_H
+#define _LINUX_PAGECACHE_ACCT_H
+
+#include <linux/container.h>
+
+#ifdef CONFIG_CONTAINER_PAGECACHE_ACCT
+extern void pagecache_acct_init_page_ptr(struct page *page);
+extern void pagecache_acct_charge(struct page *page);
+extern void pagecache_acct_uncharge(struct page *page);
+extern int pagecache_acct_page_overlimit(struct page *page);
+extern int pagecache_acct_mapping_overlimit(struct address_space *mapping);
+extern int pagecache_acct_cont_overlimit(struct container *cont);
+extern int pagecache_acct_shrink_used(unsigned long nr_pages);
+#else
+static inline void pagecache_acct_init_page_ptr(struct page *page) {}
+static inline void pagecache_acct_charge(struct page *page) {}
+static inline void pagecache_acct_uncharge(struct page *page) {}
+static inline int pagecache_acct_page_overlimit(
+		struct page *page) { return 0; }
+static inline int pagecache_acct_mapping_overlimit(
+		struct address_space *mapping) { return 0; }
+static inline int pagecache_acct_cont_overlimit(
+		struct container *cont) { return 0; }
+static inline int pagecache_acct_shrink_used(
+		unsigned long nr_pages) { return 0; }
+#endif /* CONFIG_CONTAINER_PAGECACHE_ACCT */
+
+#endif /* _LINUX_PAGECACHE_ACCT_H */
+
+
--- linux-2.6.20.orig/init/Kconfig
+++ linux-2.6.20/init/Kconfig
@@ -313,6 +313,13 @@ config CONTAINER_MEMCONTROL
 	  Provides a simple Resource Controller for monitoring and
 	  controlling the total Resident Set Size of the tasks in a container

+config CONTAINER_PAGECACHE_ACCT
+	bool "Simple PageCache accounting & control container subsystem"
+	select CONTAINERS
+	help
+	  Provides a simple Resource Controller for monitoring the
+	  total pagecache memory consumed by the tasks in a container
+
 config RELAY
 	bool "Kernel->user space relay support (formerly relayfs)"
 	help
--- linux-2.6.20.orig/mm/Makefile
+++ linux-2.6.20/mm/Makefile
@@ -30,3 +30,4 @@ obj-$(CONFIG_FS_XIP) += filemap_xip.o
 obj-$(CONFIG_MIGRATION) += migrate.o
 obj-$(CONFIG_SMP) += allocpercpu.o
 obj-$(CONFIG_CONTAINER_MEMCONTROL) += memcontrol.o
+obj-$(CONFIG_CONTAINER_PAGECACHE_ACCT) += pagecache_acct.o
--- /dev/null
+++ linux-2.6.20/mm/pagecache_acct.c
@@ -0,0 +1,368 @@
+/*
+ * Pagecache controller - "Account and control pagecache usage"
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright IBM Corporation, 2007
+ *
+ * Author: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/container.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/mm_types.h>
+#include <linux/uaccess.h>
+#include <asm/div64.h>
+#include <linux/pagecache_acct.h>
+
+/*
+ * Convert unit from pages to kilobytes
+ */
+#define K(x) ((x) << (PAGE_SHIFT - 10))
+/*
+ * Convert unit from kilobytes to pages
+ */
+#define K_to_pages(x) ((x) >> (PAGE_SHIFT -10))
+
+/* Limits for user string */
+
+#define MAX_LIMIT_STRING 	25
+
+/* nr_pages above limit to start reclaim */
+
+#define NR_PAGES_RECLAIM_THRESHOLD	64
+
+struct pagecache_acct {
+	struct container_subsys_state css;
+	spinlock_t lock;
+	atomic_t count; /*Pagecache pages added*/
+	atomic_t removed_count; /*Pagecache pages removed*/
+	unsigned int limit; /* Pagecache usage limit in kiB */
+};
+
+/*Failure counters for debugging*/
+static atomic_t failed_count; /*Page charge failures?*/
+static atomic_t failed_removed_count; /*Page uncharge failure?*/
+static atomic_t reclaim_count; /*Overlimit direct page reclaim run count */
+
+static struct container_subsys pagecache_acct_subsys;
+
+static inline struct pagecache_acct *container_pca(struct container *cont)
+{
+	return container_of(
+			container_subsys_state(cont, &pagecache_acct_subsys),
+			    struct pagecache_acct, css);
+}
+
+static int pagecache_acct_create(struct container_subsys *ss,
+				struct container *cont)
+{
+	struct pagecache_acct *pca = kzalloc(sizeof(*pca), GFP_KERNEL);
+	if (!pca)
+		return -ENOMEM;
+	spin_lock_init(&pca->lock);
+	cont->subsys[pagecache_acct_subsys.subsys_id] = &pca->css;
+	return 0;
+}
+
+static void pagecache_acct_destroy(struct container_subsys *ss,
+					struct container *cont)
+{
+	kfree(container_pca(cont));
+}
+
+static unsigned int pagecache_get_usage(struct pagecache_acct *pca)
+{
+	unsigned int count, removed_count, pagecache_used_kB;
+
+	count = (unsigned int) atomic_read(&pca->count);
+	removed_count = (unsigned int) atomic_read(&pca->removed_count);
+	/* Take care of roll over in the counters */
+	if (count >= removed_count)
+		pagecache_used_kB = count - removed_count;
+	else
+		pagecache_used_kB = ~0UL - (removed_count - count) + 1;
+
+	/* Convert unit from pages into kB */
+	pagecache_used_kB = K(pagecache_used_kB);
+
+	return pagecache_used_kB;
+}
+
+static ssize_t pagecache_usage_read(struct container *cont,
+
+			     struct cftype *cft,
+			     struct file *file,
+			     char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char usagebuf[64];
+	char *s = usagebuf;
+
+	s += sprintf(s, "%u kB \n", pagecache_get_usage(pca));
+	return simple_read_from_buffer(buf, nbytes, ppos, usagebuf,
+						s - usagebuf);
+}
+
+static ssize_t pagecache_debug_read(struct container *cont,
+			     struct cftype *cft,
+			     struct file *file,
+			     char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char usagebuf[64];
+	char *s = usagebuf;
+
+	s += sprintf(s, "%u kB Cnt: %u, %u Failed: %u, %u Reclaim: %u\n",
+				pagecache_get_usage(pca),
+				(unsigned int) atomic_read(&pca->count),
+				(unsigned int) atomic_read(&pca->removed_count),
+				(unsigned int) atomic_read(&failed_count),
+				(unsigned int) atomic_read(&failed_removed_count),
+				(unsigned int) atomic_read(&reclaim_count));
+
+	return simple_read_from_buffer(buf, nbytes, ppos, usagebuf,
+						s - usagebuf);
+}
+
+static ssize_t pagecache_limit_read(struct container *cont,
+			     struct cftype *cft,
+			     struct file *file,
+			     char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char usagebuf[64];
+	char *s = usagebuf;
+
+	s += sprintf(s, "%u kB\n", pca->limit);
+
+	return simple_read_from_buffer(buf, nbytes, ppos, usagebuf,
+						s - usagebuf);
+}
+
+static ssize_t pagecache_limit_write(struct container *cont,
+			     struct cftype *cft,
+			     struct file *file,
+			     const char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char buffer[MAX_LIMIT_STRING];
+	unsigned int limit;
+	unsigned int nr_pages;
+
+	if( nbytes > (MAX_LIMIT_STRING-1) || nbytes < 1 )
+		return -EINVAL;
+
+	if (copy_from_user(buffer, buf, nbytes))
+		return -EFAULT;
+
+	buffer[nbytes] = '\0'; /* Null termination */
+	limit = simple_strtoul(buffer, NULL, 10);
+	/* Round it off to lower 4K page boundary */
+	limit &= ~0x3;
+
+	/* Set the value in pca struct (atomic store). No read-modify-update */
+	pca->limit = limit;
+
+	/* Check for overlimit and initiate reclaim if needed */
+	/* The limits have changed now */
+	if ((nr_pages = pagecache_acct_cont_overlimit(cont))) {
+		pagecache_acct_shrink_used(nr_pages);
+	}
+	return nbytes;
+}
+
+static struct cftype cft_usage = {
+	.name = "pagecache_usage",
+	.read = pagecache_usage_read,
+};
+
+static struct cftype cft_debug = {
+	.name = "pagecache_debug",
+	.read = pagecache_debug_read,
+};
+
+static struct cftype cft_limit = {
+	.name = "pagecache_limit",
+	.read = pagecache_limit_read,
+	.write = pagecache_limit_write,
+};
+
+static int pagecache_acct_populate(struct container_subsys *ss,
+					struct container *cont)
+{
+	int rc;
+	rc = container_add_file(cont, &cft_usage);
+	if (rc)
+		return rc;
+	rc = container_add_file(cont, &cft_debug);
+	if (rc)
+		/* Cleanup with container_remove_file()? */
+		return rc;
+	rc = container_add_file(cont, &cft_limit);
+	return rc;
+}
+
+static struct container *mapping_container(struct address_space *mapping)
+{
+	if ((unsigned long) mapping & PAGE_MAPPING_ANON)
+		mapping = NULL;
+	if(!mapping)
+		return NULL;
+	if (!mapping->container) {
+		printk( KERN_DEBUG "Null Container in mapping: %p\n", mapping);
+	}
+	return mapping->container;
+}
+
+static struct container *page_container(struct page *page)
+{
+	return mapping_container(page->mapping);
+}
+
+void pagecache_acct_init_page_ptr(struct page *page)
+{
+	struct address_space *mapping;
+	mapping = page->mapping;
+	if ((unsigned long) mapping & PAGE_MAPPING_ANON)
+		mapping = NULL;
+	BUG_ON(!mapping);
+	if (current) {
+		if(!mapping->container)
+			mapping->container = task_container(current, &pagecache_acct_subsys);
+	} else
+		mapping->container = NULL;
+}
+
+void pagecache_acct_charge(struct page *page)
+{
+	struct container *cont;
+	struct pagecache_acct *pca;
+	unsigned int nr_pages;
+
+	if (pagecache_acct_subsys.subsys_id < 0) return;
+	cont = page_container(page);
+	if (cont) {
+		pca = container_pca(cont);
+		BUG_ON(!pca);
+		atomic_inc(&pca->count);
+	} else {
+		/* page->container is null??? */
+		printk(KERN_WARNING "pca_charge:page_container null\n");
+		atomic_inc(&failed_count);
+	}
+	/* Check for overlimit and initiate reclaim if needed */
+	if ((nr_pages = pagecache_acct_page_overlimit(page))) {
+		pagecache_acct_shrink_used(nr_pages);
+	}
+}
+
+void pagecache_acct_uncharge(struct page * page)
+{
+	struct container *cont;
+	struct pagecache_acct *pca;
+
+	if (pagecache_acct_subsys.subsys_id < 0) return;
+	cont = page_container(page);
+	if (cont) {
+		pca = container_pca(cont);
+		BUG_ON(!pca);
+		atomic_inc(&pca->removed_count);
+	} else {
+		/* page->container is null??? */
+		printk(KERN_WARNING "pca_uncharge:page_container null\n");
+		atomic_inc(&failed_removed_count);
+	}
+}
+
+int pagecache_acct_page_overlimit(struct page *page)
+{
+	struct container *cont;
+
+	if (pagecache_acct_subsys.subsys_id < 0)
+		return 0;
+	cont = page_container(page);
+	if (!cont)
+		return 0;
+	return pagecache_acct_cont_overlimit(cont);
+}
+
+int pagecache_acct_mapping_overlimit(struct address_space *mapping)
+{
+	struct container *cont;
+
+	if (pagecache_acct_subsys.subsys_id < 0)
+		return 0;
+	cont = mapping_container(mapping);
+	if (!cont)
+		return 0;
+	return pagecache_acct_cont_overlimit(cont);
+}
+
+int pagecache_acct_cont_overlimit(struct container *cont)
+{
+	struct pagecache_acct *pca;
+	unsigned int used, limit;
+
+	if (pagecache_acct_subsys.subsys_id < 0)
+		return 0;
+
+	if (!cont)
+		return 0;
+	pca = container_pca(cont);
+	used = pagecache_get_usage(pca);
+	limit = pca->limit;
+	if( limit && (used > limit) )
+		return K_to_pages(used - limit);
+	else
+		return 0;
+}
+
+extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
+
+int pagecache_acct_shrink_used(unsigned long nr_pages)
+{
+	unsigned long ret = 0;
+	atomic_inc(&reclaim_count);
+	return 0;
+}
+
+static struct container_subsys pagecache_acct_subsys = {
+	.name = "pagecache_acct",
+	.create = pagecache_acct_create,
+	.destroy = pagecache_acct_destroy,
+	.populate = pagecache_acct_populate,
+	.subsys_id = -1,
+};
+
+int __init init_pagecache_acct(void)
+{
+	int id = container_register_subsys(&pagecache_acct_subsys);
+	int rc = 0; /* Default to success */
+	/* Flag failure */
+	if (id < 0)
+		rc = id;
+	return rc;
+}
+
+__initcall(init_pagecache_acct);
+

[PATCH 1/3][RFC] Containers: Pagecache controller setup

[Vaidyanathan Srinivasan]

This patch adds basic infrastructure for pagecache accounting and
control subsystem within the container framework.

pagecache_usage and pagecache_limit files are created along with
every new container.  Routines to extract subsystem pointers and
container pointers have been added.

Compile time kernel options and have been added along with Makefile
changes.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
---
 include/linux/pagecache_acct.h |   53 +++++
 init/Kconfig                   |    7
 mm/Makefile                    |    1
 mm/pagecache_acct.c            |  368 +++++++++++++++++++++++++++++++++++++++++
 4 files changed, 429 insertions(+)

--- /dev/null
+++ linux-2.6.20/include/linux/pagecache_acct.h
@@ -0,0 +1,53 @@
+/*
+ * Pagecache controller - "Account and control pagecache usage"
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright IBM Corporation, 2007
+ *
+ * Author: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
+ *
+ */
+
+#ifndef _LINUX_PAGECACHE_ACCT_H
+#define _LINUX_PAGECACHE_ACCT_H
+
+#include <linux/container.h>
+
+#ifdef CONFIG_CONTAINER_PAGECACHE_ACCT
+extern void pagecache_acct_init_page_ptr(struct page *page);
+extern void pagecache_acct_charge(struct page *page);
+extern void pagecache_acct_uncharge(struct page *page);
+extern int pagecache_acct_page_overlimit(struct page *page);
+extern int pagecache_acct_mapping_overlimit(struct address_space *mapping);
+extern int pagecache_acct_cont_overlimit(struct container *cont);
+extern int pagecache_acct_shrink_used(unsigned long nr_pages);
+#else
+static inline void pagecache_acct_init_page_ptr(struct page *page) {}
+static inline void pagecache_acct_charge(struct page *page) {}
+static inline void pagecache_acct_uncharge(struct page *page) {}
+static inline int pagecache_acct_page_overlimit(
+		struct page *page) { return 0; }
+static inline int pagecache_acct_mapping_overlimit(
+		struct address_space *mapping) { return 0; }
+static inline int pagecache_acct_cont_overlimit(
+		struct container *cont) { return 0; }
+static inline int pagecache_acct_shrink_used(
+		unsigned long nr_pages) { return 0; }
+#endif /* CONFIG_CONTAINER_PAGECACHE_ACCT */
+
+#endif /* _LINUX_PAGECACHE_ACCT_H */
+
+
--- linux-2.6.20.orig/init/Kconfig
+++ linux-2.6.20/init/Kconfig
@@ -313,6 +313,13 @@ config CONTAINER_MEMCONTROL
 	  Provides a simple Resource Controller for monitoring and
 	  controlling the total Resident Set Size of the tasks in a container

+config CONTAINER_PAGECACHE_ACCT
+	bool "Simple PageCache accounting & control container subsystem"
+	select CONTAINERS
+	help
+	  Provides a simple Resource Controller for monitoring the
+	  total pagecache memory consumed by the tasks in a container
+
 config RELAY
 	bool "Kernel->user space relay support (formerly relayfs)"
 	help
--- linux-2.6.20.orig/mm/Makefile
+++ linux-2.6.20/mm/Makefile
@@ -30,3 +30,4 @@ obj-$(CONFIG_FS_XIP) += filemap_xip.o
 obj-$(CONFIG_MIGRATION) += migrate.o
 obj-$(CONFIG_SMP) += allocpercpu.o
 obj-$(CONFIG_CONTAINER_MEMCONTROL) += memcontrol.o
+obj-$(CONFIG_CONTAINER_PAGECACHE_ACCT) += pagecache_acct.o
--- /dev/null
+++ linux-2.6.20/mm/pagecache_acct.c
@@ -0,0 +1,368 @@
+/*
+ * Pagecache controller - "Account and control pagecache usage"
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright IBM Corporation, 2007
+ *
+ * Author: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/container.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/mm_types.h>
+#include <linux/uaccess.h>
+#include <asm/div64.h>
+#include <linux/pagecache_acct.h>
+
+/*
+ * Convert unit from pages to kilobytes
+ */
+#define K(x) ((x) << (PAGE_SHIFT - 10))
+/*
+ * Convert unit from kilobytes to pages
+ */
+#define K_to_pages(x) ((x) >> (PAGE_SHIFT -10))
+
+/* Limits for user string */
+
+#define MAX_LIMIT_STRING 	25
+
+/* nr_pages above limit to start reclaim */
+
+#define NR_PAGES_RECLAIM_THRESHOLD	64
+
+struct pagecache_acct {
+	struct container_subsys_state css;
+	spinlock_t lock;
+	atomic_t count; /*Pagecache pages added*/
+	atomic_t removed_count; /*Pagecache pages removed*/
+	unsigned int limit; /* Pagecache usage limit in kiB */
+};
+
+/*Failure counters for debugging*/
+static atomic_t failed_count; /*Page charge failures?*/
+static atomic_t failed_removed_count; /*Page uncharge failure?*/
+static atomic_t reclaim_count; /*Overlimit direct page reclaim run count */
+
+static struct container_subsys pagecache_acct_subsys;
+
+static inline struct pagecache_acct *container_pca(struct container *cont)
+{
+	return container_of(
+			container_subsys_state(cont, &pagecache_acct_subsys),
+			    struct pagecache_acct, css);
+}
+
+static int pagecache_acct_create(struct container_subsys *ss,
+				struct container *cont)
+{
+	struct pagecache_acct *pca = kzalloc(sizeof(*pca), GFP_KERNEL);
+	if (!pca)
+		return -ENOMEM;
+	spin_lock_init(&pca->lock);
+	cont->subsys[pagecache_acct_subsys.subsys_id] = &pca->css;
+	return 0;
+}
+
+static void pagecache_acct_destroy(struct container_subsys *ss,
+					struct container *cont)
+{
+	kfree(container_pca(cont));
+}
+
+static unsigned int pagecache_get_usage(struct pagecache_acct *pca)
+{
+	unsigned int count, removed_count, pagecache_used_kB;
+
+	count = (unsigned int) atomic_read(&pca->count);
+	removed_count = (unsigned int) atomic_read(&pca->removed_count);
+	/* Take care of roll over in the counters */
+	if (count >= removed_count)
+		pagecache_used_kB = count - removed_count;
+	else
+		pagecache_used_kB = ~0UL - (removed_count - count) + 1;
+
+	/* Convert unit from pages into kB */
+	pagecache_used_kB = K(pagecache_used_kB);
+
+	return pagecache_used_kB;
+}
+
+static ssize_t pagecache_usage_read(struct container *cont,
+
+			     struct cftype *cft,
+			     struct file *file,
+			     char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char usagebuf[64];
+	char *s = usagebuf;
+
+	s += sprintf(s, "%u kB \n", pagecache_get_usage(pca));
+	return simple_read_from_buffer(buf, nbytes, ppos, usagebuf,
+						s - usagebuf);
+}
+
+static ssize_t pagecache_debug_read(struct container *cont,
+			     struct cftype *cft,
+			     struct file *file,
+			     char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char usagebuf[64];
+	char *s = usagebuf;
+
+	s += sprintf(s, "%u kB Cnt: %u, %u Failed: %u, %u Reclaim: %u\n",
+				pagecache_get_usage(pca),
+				(unsigned int) atomic_read(&pca->count),
+				(unsigned int) atomic_read(&pca->removed_count),
+				(unsigned int) atomic_read(&failed_count),
+				(unsigned int) atomic_read(&failed_removed_count),
+				(unsigned int) atomic_read(&reclaim_count));
+
+	return simple_read_from_buffer(buf, nbytes, ppos, usagebuf,
+						s - usagebuf);
+}
+
+static ssize_t pagecache_limit_read(struct container *cont,
+			     struct cftype *cft,
+			     struct file *file,
+			     char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char usagebuf[64];
+	char *s = usagebuf;
+
+	s += sprintf(s, "%u kB\n", pca->limit);
+
+	return simple_read_from_buffer(buf, nbytes, ppos, usagebuf,
+						s - usagebuf);
+}
+
+static ssize_t pagecache_limit_write(struct container *cont,
+			     struct cftype *cft,
+			     struct file *file,
+			     const char __user *buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	struct pagecache_acct *pca = container_pca(cont);
+	char buffer[MAX_LIMIT_STRING];
+	unsigned int limit;
+	unsigned int nr_pages;
+
+	if( nbytes > (MAX_LIMIT_STRING-1) || nbytes < 1 )
+		return -EINVAL;
+
+	if (copy_from_user(buffer, buf, nbytes))
+		return -EFAULT;
+
+	buffer[nbytes] = '\0'; /* Null termination */
+	limit = simple_strtoul(buffer, NULL, 10);
+	/* Round it off to lower 4K page boundary */
+	limit &= ~0x3;
+
+	/* Set the value in pca struct (atomic store). No read-modify-update */
+	pca->limit = limit;
+
+	/* Check for overlimit and initiate reclaim if needed */
+	/* The limits have changed now */
+	if ((nr_pages = pagecache_acct_cont_overlimit(cont))) {
+		pagecache_acct_shrink_used(nr_pages);
+	}
+	return nbytes;
+}
+
+static struct cftype cft_usage = {
+	.name = "pagecache_usage",
+	.read = pagecache_usage_read,
+};
+
+static struct cftype cft_debug = {
+	.name = "pagecache_debug",
+	.read = pagecache_debug_read,
+};
+
+static struct cftype cft_limit = {
+	.name = "pagecache_limit",
+	.read = pagecache_limit_read,
+	.write = pagecache_limit_write,
+};
+
+static int pagecache_acct_populate(struct container_subsys *ss,
+					struct container *cont)
+{
+	int rc;
+	rc = container_add_file(cont, &cft_usage);
+	if (rc)
+		return rc;
+	rc = container_add_file(cont, &cft_debug);
+	if (rc)
+		/* Cleanup with container_remove_file()? */
+		return rc;
+	rc = container_add_file(cont, &cft_limit);
+	return rc;
+}
+
+static struct container *mapping_container(struct address_space *mapping)
+{
+	if ((unsigned long) mapping & PAGE_MAPPING_ANON)
+		mapping = NULL;
+	if(!mapping)
+		return NULL;
+	if (!mapping->container) {
+		printk( KERN_DEBUG "Null Container in mapping: %p\n", mapping);
+	}
+	return mapping->container;
+}
+
+static struct container *page_container(struct page *page)
+{
+	return mapping_container(page->mapping);
+}
+
+void pagecache_acct_init_page_ptr(struct page *page)
+{
+	struct address_space *mapping;
+	mapping = page->mapping;
+	if ((unsigned long) mapping & PAGE_MAPPING_ANON)
+		mapping = NULL;
+	BUG_ON(!mapping);
+	if (current) {
+		if(!mapping->container)
+			mapping->container = task_container(current, &pagecache_acct_subsys);
+	} else
+		mapping->container = NULL;
+}
+
+void pagecache_acct_charge(struct page *page)
+{
+	struct container *cont;
+	struct pagecache_acct *pca;
+	unsigned int nr_pages;
+
+	if (pagecache_acct_subsys.subsys_id < 0) return;
+	cont = page_container(page);
+	if (cont) {
+		pca = container_pca(cont);
+		BUG_ON(!pca);
+		atomic_inc(&pca->count);
+	} else {
+		/* page->container is null??? */
+		printk(KERN_WARNING "pca_charge:page_container null\n");
+		atomic_inc(&failed_count);
+	}
+	/* Check for overlimit and initiate reclaim if needed */
+	if ((nr_pages = pagecache_acct_page_overlimit(page))) {
+		pagecache_acct_shrink_used(nr_pages);
+	}
+}
+
+void pagecache_acct_uncharge(struct page * page)
+{
+	struct container *cont;
+	struct pagecache_acct *pca;
+
+	if (pagecache_acct_subsys.subsys_id < 0) return;
+	cont = page_container(page);
+	if (cont) {
+		pca = container_pca(cont);
+		BUG_ON(!pca);
+		atomic_inc(&pca->removed_count);
+	} else {
+		/* page->container is null??? */
+		printk(KERN_WARNING "pca_uncharge:page_container null\n");
+		atomic_inc(&failed_removed_count);
+	}
+}
+
+int pagecache_acct_page_overlimit(struct page *page)
+{
+	struct container *cont;
+
+	if (pagecache_acct_subsys.subsys_id < 0)
+		return 0;
+	cont = page_container(page);
+	if (!cont)
+		return 0;
+	return pagecache_acct_cont_overlimit(cont);
+}
+
+int pagecache_acct_mapping_overlimit(struct address_space *mapping)
+{
+	struct container *cont;
+
+	if (pagecache_acct_subsys.subsys_id < 0)
+		return 0;
+	cont = mapping_container(mapping);
+	if (!cont)
+		return 0;
+	return pagecache_acct_cont_overlimit(cont);
+}
+
+int pagecache_acct_cont_overlimit(struct container *cont)
+{
+	struct pagecache_acct *pca;
+	unsigned int used, limit;
+
+	if (pagecache_acct_subsys.subsys_id < 0)
+		return 0;
+
+	if (!cont)
+		return 0;
+	pca = container_pca(cont);
+	used = pagecache_get_usage(pca);
+	limit = pca->limit;
+	if( limit && (used > limit) )
+		return K_to_pages(used - limit);
+	else
+		return 0;
+}
+
+extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
+
+int pagecache_acct_shrink_used(unsigned long nr_pages)
+{
+	unsigned long ret = 0;
+	atomic_inc(&reclaim_count);
+	return 0;
+}
+
+static struct container_subsys pagecache_acct_subsys = {
+	.name = "pagecache_acct",
+	.create = pagecache_acct_create,
+	.destroy = pagecache_acct_destroy,
+	.populate = pagecache_acct_populate,
+	.subsys_id = -1,
+};
+
+int __init init_pagecache_acct(void)
+{
+	int id = container_register_subsys(&pagecache_acct_subsys);
+	int rc = 0; /* Default to success */
+	/* Flag failure */
+	if (id < 0)
+		rc = id;
+	return rc;
+}
+
+__initcall(init_pagecache_acct);
+

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/3][RFC] Containers: Pagecache controller accounting

[Vaidyanathan Srinivasan]

The accounting framework works by adding a container pointer in
address_space structure.  Each page in pagecache belongs to a
radix tree within the address_space structure corresponding to the inode.

In order to charge the container for pagecache usage, the corresponding
address_space is obtained from struct page which holds the container pointer.
This framework avoids any additional pointers in struct page.

additions and deletions from pagecache are hooked to charge and uncharge
the corresponding container.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
---
 include/linux/fs.h |    4 ++++
 mm/filemap.c       |    8 ++++++++
 2 files changed, 12 insertions(+)

--- linux-2.6.20.orig/include/linux/fs.h
+++ linux-2.6.20/include/linux/fs.h
@@ -447,6 +447,10 @@ struct address_space {
 	spinlock_t		private_lock;	/* for use by the address_space */
 	struct list_head	private_list;	/* ditto */
 	struct address_space	*assoc_mapping;	/* ditto */
+#ifdef CONFIG_CONTAINER_PAGECACHE_ACCT
+	struct container *container; 	/* Charge page to the right container
+					   using page->mapping */
+#endif
 } __attribute__((aligned(sizeof(long))));
 	/*
 	 * On most architectures that alignment is already the case; but
--- linux-2.6.20.orig/mm/filemap.c
+++ linux-2.6.20/mm/filemap.c
@@ -30,6 +30,7 @@
 #include <linux/security.h>
 #include <linux/syscalls.h>
 #include <linux/cpuset.h>
+#include <linux/pagecache_acct.h>
 #include "filemap.h"
 #include "internal.h"

@@ -117,6 +118,8 @@ void __remove_from_page_cache(struct pag
 	struct address_space *mapping = page->mapping;

 	radix_tree_delete(&mapping->page_tree, page->index);
+	/* Uncharge before the mapping is gone */
+	pagecache_acct_uncharge(page);
 	page->mapping = NULL;
 	mapping->nrpages--;
 	__dec_zone_page_state(page, NR_FILE_PAGES);
@@ -451,6 +454,11 @@ int add_to_page_cache(struct page *page,
 			__inc_zone_page_state(page, NR_FILE_PAGES);
 		}
 		write_unlock_irq(&mapping->tree_lock);
+		/* Unlock before charge, because we may reclaim this inline */
+		if (!error) {
+			pagecache_acct_init_page_ptr(page);
+			pagecache_acct_charge(page);
+		}
 		radix_tree_preload_end();
 	}
 	return error;

Re: [PATCH 2/3][RFC] Containers: Pagecache controller accounting

[Vaidyanathan Srinivasan]

 
References: <45ED251C.2010400@linux.vnet.ibm.com>
In-Reply-To: <45ED251C.2010400@linux.vnet.ibm.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit


The accounting framework works by adding a container pointer in
address_space structure.  Each page in pagecache belongs to a
radix tree within the address_space structure corresponding to the inode.

In order to charge the container for pagecache usage, the corresponding
address_space is obtained from struct page which holds the container pointer.
This framework avoids any additional pointers in struct page.

additions and deletions from pagecache are hooked to charge and uncharge
the corresponding container.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
---
 include/linux/fs.h |    4 ++++
 mm/filemap.c       |    8 ++++++++
 2 files changed, 12 insertions(+)

--- linux-2.6.20.orig/include/linux/fs.h
+++ linux-2.6.20/include/linux/fs.h
@@ -447,6 +447,10 @@ struct address_space {
 	spinlock_t		private_lock;	/* for use by the address_space */
 	struct list_head	private_list;	/* ditto */
 	struct address_space	*assoc_mapping;	/* ditto */
+#ifdef CONFIG_CONTAINER_PAGECACHE_ACCT
+	struct container *container; 	/* Charge page to the right container
+					   using page->mapping */
+#endif
 } __attribute__((aligned(sizeof(long))));
 	/*
 	 * On most architectures that alignment is already the case; but
--- linux-2.6.20.orig/mm/filemap.c
+++ linux-2.6.20/mm/filemap.c
@@ -30,6 +30,7 @@
 #include <linux/security.h>
 #include <linux/syscalls.h>
 #include <linux/cpuset.h>
+#include <linux/pagecache_acct.h>
 #include "filemap.h"
 #include "internal.h"

@@ -117,6 +118,8 @@ void __remove_from_page_cache(struct pag
 	struct address_space *mapping = page->mapping;

 	radix_tree_delete(&mapping->page_tree, page->index);
+	/* Uncharge before the mapping is gone */
+	pagecache_acct_uncharge(page);
 	page->mapping = NULL;
 	mapping->nrpages--;
 	__dec_zone_page_state(page, NR_FILE_PAGES);
@@ -451,6 +454,11 @@ int add_to_page_cache(struct page *page,
 			__inc_zone_page_state(page, NR_FILE_PAGES);
 		}
 		write_unlock_irq(&mapping->tree_lock);
+		/* Unlock before charge, because we may reclaim this inline */
+		if (!error) {
+			pagecache_acct_init_page_ptr(page);
+			pagecache_acct_charge(page);
+		}
 		radix_tree_preload_end();
 	}
 	return error;

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

The reclaim code is similar to RSS memory controller.  Scan control is
slightly different since we are targeting different type of pages.

Additionally no mapped pages are touched when scanning for pagecache pages.

RSS memory controller and pagecache controller share common code in reclaim
and hence pagecache controller patches are dependent on RSS memory controller
patch even though the features are independently configurable at compile time.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
---
 include/linux/memcontrol.h |    6 +++
 mm/memcontrol.c            |    4 +-
 mm/pagecache_acct.c        |   11 ++++++
 mm/vmscan.c                |   71 +++++++++++++++++++++++++++++++++++++--------
 4 files changed, 77 insertions(+), 15 deletions(-)

--- linux-2.6.20.orig/include/linux/memcontrol.h
+++ linux-2.6.20/include/linux/memcontrol.h
@@ -36,6 +36,12 @@ enum {
 	MEMCONTROL_DONT_CHECK_LIMIT = false,
 };

+/* Type of memory to reclaim in shrink_container_memory() */
+enum {
+	RECLAIM_MAPPED_MEMORY = 1,
+	RECLAIM_PAGECACHE_MEMORY,
+};
+
 #ifdef CONFIG_CONTAINER_MEMCONTROL
 #include <linux/wait.h>

--- linux-2.6.20.orig/mm/memcontrol.c
+++ linux-2.6.20/mm/memcontrol.c
@@ -165,8 +165,8 @@ static int memcontrol_check_and_reclaim(
 				nr_pages = (pushback * limit) / 100;
 			mem->reclaim_in_progress = true;
 			spin_unlock(&mem->lock);
-			nr_reclaimed +=
-				memcontrol_shrink_mapped_memory(nr_pages, cont);
+			nr_reclaimed += shrink_container_memory(
+					RECLAIM_MAPPED_MEMORY, nr_pages, cont);
 			spin_lock(&mem->lock);
 			mem->reclaim_in_progress = false;
 			wake_up_all(&mem->wq);
--- linux-2.6.20.orig/mm/pagecache_acct.c
+++ linux-2.6.20/mm/pagecache_acct.c
@@ -29,6 +29,7 @@
 #include <linux/uaccess.h>
 #include <asm/div64.h>
 #include <linux/pagecache_acct.h>
+#include <linux/memcontrol.h>

 /*
  * Convert unit from pages to kilobytes
@@ -337,12 +338,20 @@ int pagecache_acct_cont_overlimit(struct
 		return 0;
 }

-extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
+extern unsigned long shrink_container_memory(unsigned int memory_type,
+				unsigned long nr_pages, void *container);

 int pagecache_acct_shrink_used(unsigned long nr_pages)
 {
 	unsigned long ret = 0;
 	atomic_inc(&reclaim_count);
+
+	/* Don't call reclaim for each page above limit */
+	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
+		ret += shrink_container_memory(
+				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
+	}
+
 	return 0;
 }

--- linux-2.6.20.orig/mm/vmscan.c
+++ linux-2.6.20/mm/vmscan.c
@@ -43,6 +43,7 @@

 #include <linux/swapops.h>
 #include <linux/memcontrol.h>
+#include <linux/pagecache_acct.h>

 #include "internal.h"

@@ -70,6 +71,8 @@ struct scan_control {

 	struct container *container;	/* Used by containers for reclaiming */
 					/* pages when the limit is exceeded  */
+	int reclaim_pagecache_only;     /* Set when called from
+					   pagecache controller */
 };

 /*
@@ -474,6 +477,15 @@ static unsigned long shrink_page_list(st
 			goto keep;

 		VM_BUG_ON(PageActive(page));
+		/* Take it easy if we are doing only pagecache pages */
+		if (sc->reclaim_pagecache_only) {
+			/* Check if this is a pagecache page they are not mapped */
+			if (page_mapped(page))
+				goto keep_locked;
+			/* Check if this container has exceeded pagecache limit */
+			if (!pagecache_acct_page_overlimit(page))
+				goto keep_locked;
+		}

 		sc->nr_scanned++;

@@ -522,7 +534,8 @@ static unsigned long shrink_page_list(st
 		}

 		if (PageDirty(page)) {
-			if (referenced)
+			/* Reclaim even referenced pagecache pages if over limit */
+			if (!pagecache_acct_page_overlimit(page) && referenced)
 				goto keep_locked;
 			if (!may_enter_fs)
 				goto keep_locked;
@@ -869,6 +882,13 @@ force_reclaim_mapped:
 		cond_resched();
 		page = lru_to_page(&l_hold);
 		list_del(&page->lru);
+		/* While reclaiming pagecache make it easy */
+		if (sc->reclaim_pagecache_only) {
+			if (page_mapped(page) || !pagecache_acct_page_overlimit(page)) {
+				list_add(&page->lru, &l_active);
+				continue;
+			}
+		}
 		if (page_mapped(page)) {
 			if (!reclaim_mapped ||
 			    (total_swap_pages == 0 && PageAnon(page)) ||
@@ -1064,6 +1084,7 @@ unsigned long try_to_free_pages(struct z
 		.swap_cluster_max = SWAP_CLUSTER_MAX,
 		.may_swap = 1,
 		.swappiness = vm_swappiness,
+		.reclaim_pagecache_only = 0,
 	};

 	count_vm_event(ALLOCSTALL);
@@ -1168,6 +1189,7 @@ static unsigned long balance_pgdat(pg_da
 		.may_swap = 1,
 		.swap_cluster_max = SWAP_CLUSTER_MAX,
 		.swappiness = vm_swappiness,
+		.reclaim_pagecache_only = 0,
 	};
 	/*
 	 * temp_priority is used to remember the scanning priority at which
@@ -1398,7 +1420,9 @@ void wakeup_kswapd(struct zone *zone, in
 	wake_up_interruptible(&pgdat->kswapd_wait);
 }

-#if defined(CONFIG_PM) || defined(CONFIG_CONTAINER_MEMCONTROL)
+#if defined(CONFIG_PM) || defined(CONFIG_CONTAINER_MEMCONTROL) \
+	|| defined(CONFIG_CONTAINER_PAGECACHE_ACCT)
+
 /*
  * Helper function for shrink_all_memory().  Tries to reclaim 'nr_pages' pages
  * from LRU lists system-wide, for given pass and priority, and returns the
@@ -1470,11 +1494,13 @@ unsigned long shrink_all_memory(unsigned
 	int pass;
 	struct reclaim_state reclaim_state;
 	struct scan_control sc = {
-		.gfp_mask = GFP_KERNEL,
+		.gfp_mask = GFdefined(CONFIG_CONTAINER_PAGECACHE_ACCT)
+P_KERNEL,
 		.may_swap = 0,
 		.swap_cluster_max = nr_pages,
 		.may_writepage = 1,
 		.swappiness = vm_swappiness,
+		.reclaim_pagecache_only = 0,
 	};

 	current->reclaim_state = &reclaim_state;
@@ -1551,33 +1577,53 @@ out:
 }
 #endif

-#ifdef CONFIG_CONTAINER_MEMCONTROL
+#if defined(CONFIG_CONTAINER_MEMCTLR) || \
+	defined(CONFIG_CONTAINER_PAGECACHE_ACCT)
+
 /*
  * Try to free `nr_pages' of memory, system-wide, and return the number of
  * freed pages.
  * Modelled after shrink_all_memory()
  */
-unsigned long memcontrol_shrink_mapped_memory(unsigned long nr_pages,
-						struct container *container)
+unsigned long shrink_container_memory(unsigned int memory_type,
+			unsigned long nr_pages, struct container *container)
 {
 	unsigned long ret = 0;
 	int pass;
 	unsigned long nr_total_scanned = 0;
+	struct reclaim_state reclaim_state;

 	struct scan_control sc = {
 		.gfp_mask = GFP_KERNEL,
-		.may_swap = 0,
 		.swap_cluster_max = nr_pages,
 		.may_writepage = 1,
-		.container = container,
-		.may_swap = 1,
-		.swappiness = 100,
 	};

+ 	switch (memory_type) {
+ 		case RECLAIM_PAGECACHE_MEMORY:
+ 			sc.may_swap = 0;
+ 			sc.swappiness = 0; /* Do not swap, only pagecache reclaim */
+ 			sc.reclaim_pagecache_only = 1; /* Flag it */
+ 			break;
+
+ 		case RECLAIM_MAPPED_MEMORY:
+ 			sc.container = container;
+ 			sc.may_swap = 1;
+ 			sc.swappiness = 100; /* Do swap and free memory */
+ 			sc.reclaim_pagecache_only = 0; /* Flag it */
+ 			break;
+
+ 		default:
+ 			BUG();
+ 	}
+ 	current->reclaim_state = &reclaim_state;
+
+
 	/*
 	 * We try to shrink LRUs in 3 passes:
 	 * 0 = Reclaim from inactive_list only
-	 * 1 = Reclaim mapped (normal reclaim)
+ 	 * 1 = Reclaim from active list
+ 	 * 	(Mapped or pagecache pages depending on memory type)
 	 * 2 = 2nd pass of type 1
 	 */
 	for (pass = 0; pass < 3; pass++) {
@@ -1585,7 +1631,6 @@ unsigned long memcontrol_shrink_mapped_m

 		for (prio = DEF_PRIORITY; prio >= 0; prio--) {
 			unsigned long nr_to_scan = nr_pages - ret;
-
 			sc.nr_scanned = 0;
 			ret += shrink_all_zones(nr_to_scan, prio,
 						pass, 1, &sc);
@@ -1598,8 +1643,10 @@ unsigned long memcontrol_shrink_mapped_m
 		}
 	}
 out:
+	current->reclaim_state = NULL;
 	return ret;
 }
+
 #endif

 /* It's optimal to keep kswapds on the same CPUs as their memory, but

[PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

The reclaim code is similar to RSS memory controller.  Scan control is
slightly different since we are targeting different type of pages.

Additionally no mapped pages are touched when scanning for pagecache pages.

RSS memory controller and pagecache controller share common code in reclaim
and hence pagecache controller patches are dependent on RSS memory controller
patch even though the features are independently configurable at compile time.

Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
---
 include/linux/memcontrol.h |    6 +++
 mm/memcontrol.c            |    4 +-
 mm/pagecache_acct.c        |   11 ++++++
 mm/vmscan.c                |   71 +++++++++++++++++++++++++++++++++++++--------
 4 files changed, 77 insertions(+), 15 deletions(-)

--- linux-2.6.20.orig/include/linux/memcontrol.h
+++ linux-2.6.20/include/linux/memcontrol.h
@@ -36,6 +36,12 @@ enum {
 	MEMCONTROL_DONT_CHECK_LIMIT = false,
 };

+/* Type of memory to reclaim in shrink_container_memory() */
+enum {
+	RECLAIM_MAPPED_MEMORY = 1,
+	RECLAIM_PAGECACHE_MEMORY,
+};
+
 #ifdef CONFIG_CONTAINER_MEMCONTROL
 #include <linux/wait.h>

--- linux-2.6.20.orig/mm/memcontrol.c
+++ linux-2.6.20/mm/memcontrol.c
@@ -165,8 +165,8 @@ static int memcontrol_check_and_reclaim(
 				nr_pages = (pushback * limit) / 100;
 			mem->reclaim_in_progress = true;
 			spin_unlock(&mem->lock);
-			nr_reclaimed +=
-				memcontrol_shrink_mapped_memory(nr_pages, cont);
+			nr_reclaimed += shrink_container_memory(
+					RECLAIM_MAPPED_MEMORY, nr_pages, cont);
 			spin_lock(&mem->lock);
 			mem->reclaim_in_progress = false;
 			wake_up_all(&mem->wq);
--- linux-2.6.20.orig/mm/pagecache_acct.c
+++ linux-2.6.20/mm/pagecache_acct.c
@@ -29,6 +29,7 @@
 #include <linux/uaccess.h>
 #include <asm/div64.h>
 #include <linux/pagecache_acct.h>
+#include <linux/memcontrol.h>

 /*
  * Convert unit from pages to kilobytes
@@ -337,12 +338,20 @@ int pagecache_acct_cont_overlimit(struct
 		return 0;
 }

-extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
+extern unsigned long shrink_container_memory(unsigned int memory_type,
+				unsigned long nr_pages, void *container);

 int pagecache_acct_shrink_used(unsigned long nr_pages)
 {
 	unsigned long ret = 0;
 	atomic_inc(&reclaim_count);
+
+	/* Don't call reclaim for each page above limit */
+	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
+		ret += shrink_container_memory(
+				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
+	}
+
 	return 0;
 }

--- linux-2.6.20.orig/mm/vmscan.c
+++ linux-2.6.20/mm/vmscan.c
@@ -43,6 +43,7 @@

 #include <linux/swapops.h>
 #include <linux/memcontrol.h>
+#include <linux/pagecache_acct.h>

 #include "internal.h"

@@ -70,6 +71,8 @@ struct scan_control {

 	struct container *container;	/* Used by containers for reclaiming */
 					/* pages when the limit is exceeded  */
+	int reclaim_pagecache_only;     /* Set when called from
+					   pagecache controller */
 };

 /*
@@ -474,6 +477,15 @@ static unsigned long shrink_page_list(st
 			goto keep;

 		VM_BUG_ON(PageActive(page));
+		/* Take it easy if we are doing only pagecache pages */
+		if (sc->reclaim_pagecache_only) {
+			/* Check if this is a pagecache page they are not mapped */
+			if (page_mapped(page))
+				goto keep_locked;
+			/* Check if this container has exceeded pagecache limit */
+			if (!pagecache_acct_page_overlimit(page))
+				goto keep_locked;
+		}

 		sc->nr_scanned++;

@@ -522,7 +534,8 @@ static unsigned long shrink_page_list(st
 		}

 		if (PageDirty(page)) {
-			if (referenced)
+			/* Reclaim even referenced pagecache pages if over limit */
+			if (!pagecache_acct_page_overlimit(page) && referenced)
 				goto keep_locked;
 			if (!may_enter_fs)
 				goto keep_locked;
@@ -869,6 +882,13 @@ force_reclaim_mapped:
 		cond_resched();
 		page = lru_to_page(&l_hold);
 		list_del(&page->lru);
+		/* While reclaiming pagecache make it easy */
+		if (sc->reclaim_pagecache_only) {
+			if (page_mapped(page) || !pagecache_acct_page_overlimit(page)) {
+				list_add(&page->lru, &l_active);
+				continue;
+			}
+		}
 		if (page_mapped(page)) {
 			if (!reclaim_mapped ||
 			    (total_swap_pages == 0 && PageAnon(page)) ||
@@ -1064,6 +1084,7 @@ unsigned long try_to_free_pages(struct z
 		.swap_cluster_max = SWAP_CLUSTER_MAX,
 		.may_swap = 1,
 		.swappiness = vm_swappiness,
+		.reclaim_pagecache_only = 0,
 	};

 	count_vm_event(ALLOCSTALL);
@@ -1168,6 +1189,7 @@ static unsigned long balance_pgdat(pg_da
 		.may_swap = 1,
 		.swap_cluster_max = SWAP_CLUSTER_MAX,
 		.swappiness = vm_swappiness,
+		.reclaim_pagecache_only = 0,
 	};
 	/*
 	 * temp_priority is used to remember the scanning priority at which
@@ -1398,7 +1420,9 @@ void wakeup_kswapd(struct zone *zone, in
 	wake_up_interruptible(&pgdat->kswapd_wait);
 }

-#if defined(CONFIG_PM) || defined(CONFIG_CONTAINER_MEMCONTROL)
+#if defined(CONFIG_PM) || defined(CONFIG_CONTAINER_MEMCONTROL) \
+	|| defined(CONFIG_CONTAINER_PAGECACHE_ACCT)
+
 /*
  * Helper function for shrink_all_memory().  Tries to reclaim 'nr_pages' pages
  * from LRU lists system-wide, for given pass and priority, and returns the
@@ -1470,11 +1494,13 @@ unsigned long shrink_all_memory(unsigned
 	int pass;
 	struct reclaim_state reclaim_state;
 	struct scan_control sc = {
-		.gfp_mask = GFP_KERNEL,
+		.gfp_mask = GFdefined(CONFIG_CONTAINER_PAGECACHE_ACCT)
+P_KERNEL,
 		.may_swap = 0,
 		.swap_cluster_max = nr_pages,
 		.may_writepage = 1,
 		.swappiness = vm_swappiness,
+		.reclaim_pagecache_only = 0,
 	};

 	current->reclaim_state = &reclaim_state;
@@ -1551,33 +1577,53 @@ out:
 }
 #endif

-#ifdef CONFIG_CONTAINER_MEMCONTROL
+#if defined(CONFIG_CONTAINER_MEMCTLR) || \
+	defined(CONFIG_CONTAINER_PAGECACHE_ACCT)
+
 /*
  * Try to free `nr_pages' of memory, system-wide, and return the number of
  * freed pages.
  * Modelled after shrink_all_memory()
  */
-unsigned long memcontrol_shrink_mapped_memory(unsigned long nr_pages,
-						struct container *container)
+unsigned long shrink_container_memory(unsigned int memory_type,
+			unsigned long nr_pages, struct container *container)
 {
 	unsigned long ret = 0;
 	int pass;
 	unsigned long nr_total_scanned = 0;
+	struct reclaim_state reclaim_state;

 	struct scan_control sc = {
 		.gfp_mask = GFP_KERNEL,
-		.may_swap = 0,
 		.swap_cluster_max = nr_pages,
 		.may_writepage = 1,
-		.container = container,
-		.may_swap = 1,
-		.swappiness = 100,
 	};

+ 	switch (memory_type) {
+ 		case RECLAIM_PAGECACHE_MEMORY:
+ 			sc.may_swap = 0;
+ 			sc.swappiness = 0; /* Do not swap, only pagecache reclaim */
+ 			sc.reclaim_pagecache_only = 1; /* Flag it */
+ 			break;
+
+ 		case RECLAIM_MAPPED_MEMORY:
+ 			sc.container = container;
+ 			sc.may_swap = 1;
+ 			sc.swappiness = 100; /* Do swap and free memory */
+ 			sc.reclaim_pagecache_only = 0; /* Flag it */
+ 			break;
+
+ 		default:
+ 			BUG();
+ 	}
+ 	current->reclaim_state = &reclaim_state;
+
+
 	/*
 	 * We try to shrink LRUs in 3 passes:
 	 * 0 = Reclaim from inactive_list only
-	 * 1 = Reclaim mapped (normal reclaim)
+ 	 * 1 = Reclaim from active list
+ 	 * 	(Mapped or pagecache pages depending on memory type)
 	 * 2 = 2nd pass of type 1
 	 */
 	for (pass = 0; pass < 3; pass++) {
@@ -1585,7 +1631,6 @@ unsigned long memcontrol_shrink_mapped_m

 		for (prio = DEF_PRIORITY; prio >= 0; prio--) {
 			unsigned long nr_to_scan = nr_pages - ret;
-
 			sc.nr_scanned = 0;
 			ret += shrink_all_zones(nr_to_scan, prio,
 						pass, 1, &sc);
@@ -1598,8 +1643,10 @@ unsigned long memcontrol_shrink_mapped_m
 		}
 	}
 out:
+	current->reclaim_state = NULL;
 	return ret;
 }
+
 #endif

 /* It's optimal to keep kswapds on the same CPUs as their memory, but

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Kari Hurtta]

Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> writes
in gmane.linux.kernel,gmane.linux.kernel.mm:

> --- linux-2.6.20.orig/mm/pagecache_acct.c
> +++ linux-2.6.20/mm/pagecache_acct.c
> @@ -29,6 +29,7 @@
>  #include <linux/uaccess.h>
>  #include <asm/div64.h>
>  #include <linux/pagecache_acct.h>
> +#include <linux/memcontrol.h>
> 
>  /*
>   * Convert unit from pages to kilobytes
> @@ -337,12 +338,20 @@ int pagecache_acct_cont_overlimit(struct
>  		return 0;
>  }
> 
> -extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
> +extern unsigned long shrink_container_memory(unsigned int memory_type,
> +				unsigned long nr_pages, void *container);
> 
>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>  {
>  	unsigned long ret = 0;
>  	atomic_inc(&reclaim_count);
> +
> +	/* Don't call reclaim for each page above limit */
> +	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
> +		ret += shrink_container_memory(
> +				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
> +	}
> +
>  	return 0;
>  }
> 

'ret' is not used ?

/ Kari Hurtta

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Kari Hurtta]

Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> writes
in gmane.linux.kernel,gmane.linux.kernel.mm:

> --- linux-2.6.20.orig/mm/pagecache_acct.c
> +++ linux-2.6.20/mm/pagecache_acct.c
> @@ -29,6 +29,7 @@
>  #include <linux/uaccess.h>
>  #include <asm/div64.h>
>  #include <linux/pagecache_acct.h>
> +#include <linux/memcontrol.h>
> 
>  /*
>   * Convert unit from pages to kilobytes
> @@ -337,12 +338,20 @@ int pagecache_acct_cont_overlimit(struct
>  		return 0;
>  }
> 
> -extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
> +extern unsigned long shrink_container_memory(unsigned int memory_type,
> +				unsigned long nr_pages, void *container);
> 
>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>  {
>  	unsigned long ret = 0;
>  	atomic_inc(&reclaim_count);
> +
> +	/* Don't call reclaim for each page above limit */
> +	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
> +		ret += shrink_container_memory(
> +				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
> +	}
> +
>  	return 0;
>  }
> 

'ret' is not used ?

/ Kari Hurtta

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Kari Hurtta wrote:
> Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> writes
> in gmane.linux.kernel,gmane.linux.kernel.mm:
> 
>> --- linux-2.6.20.orig/mm/pagecache_acct.c
>> +++ linux-2.6.20/mm/pagecache_acct.c
>> @@ -29,6 +29,7 @@
>>  #include <linux/uaccess.h>
>>  #include <asm/div64.h>
>>  #include <linux/pagecache_acct.h>
>> +#include <linux/memcontrol.h>
>>
>>  /*
>>   * Convert unit from pages to kilobytes
>> @@ -337,12 +338,20 @@ int pagecache_acct_cont_overlimit(struct
>>  		return 0;
>>  }
>>
>> -extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
>> +extern unsigned long shrink_container_memory(unsigned int memory_type,
>> +				unsigned long nr_pages, void *container);
>>
>>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>>  {
>>  	unsigned long ret = 0;
>>  	atomic_inc(&reclaim_count);
>> +
>> +	/* Don't call reclaim for each page above limit */
>> +	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
>> +		ret += shrink_container_memory(
>> +				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
>> +	}
>> +
>>  	return 0;
>>  }
>>
> 
> 'ret' is not used ?

I have been setting watch points and tracing the value of ret.
Basically that is used while debugging.  I have not removed it since
this is an RFC post and we would go through many cleanups cycles.

I will remember to remove it in the next version or verify that the
compiler does remove 'ret' :)

--Vaidy

> / Kari Hurtta
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Kari Hurtta wrote:
> Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> writes
> in gmane.linux.kernel,gmane.linux.kernel.mm:
> 
>> --- linux-2.6.20.orig/mm/pagecache_acct.c
>> +++ linux-2.6.20/mm/pagecache_acct.c
>> @@ -29,6 +29,7 @@
>>  #include <linux/uaccess.h>
>>  #include <asm/div64.h>
>>  #include <linux/pagecache_acct.h>
>> +#include <linux/memcontrol.h>
>>
>>  /*
>>   * Convert unit from pages to kilobytes
>> @@ -337,12 +338,20 @@ int pagecache_acct_cont_overlimit(struct
>>  		return 0;
>>  }
>>
>> -extern unsigned long shrink_all_pagecache_memory(unsigned long nr_pages);
>> +extern unsigned long shrink_container_memory(unsigned int memory_type,
>> +				unsigned long nr_pages, void *container);
>>
>>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>>  {
>>  	unsigned long ret = 0;
>>  	atomic_inc(&reclaim_count);
>> +
>> +	/* Don't call reclaim for each page above limit */
>> +	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
>> +		ret += shrink_container_memory(
>> +				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
>> +	}
>> +
>>  	return 0;
>>  }
>>
> 
> 'ret' is not used ?

I have been setting watch points and tracing the value of ret.
Basically that is used while debugging.  I have not removed it since
this is an RFC post and we would go through many cleanups cycles.

I will remember to remove it in the next version or verify that the
compiler does remove 'ret' :)

--Vaidy

> / Kari Hurtta
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Aubrey Li]

On 3/6/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>
> The reclaim code is similar to RSS memory controller.  Scan control is
> slightly different since we are targeting different type of pages.
>
> Additionally no mapped pages are touched when scanning for pagecache pages.
>
> RSS memory controller and pagecache controller share common code in reclaim
> and hence pagecache controller patches are dependent on RSS memory controller
> patch even though the features are independently configurable at compile time.
>
> --- linux-2.6.20.orig/mm/vmscan.c
> +++ linux-2.6.20/mm/vmscan.c
> @@ -43,6 +43,7 @@
>
>  #include <linux/swapops.h>
>  #include <linux/memcontrol.h>
> +#include <linux/pagecache_acct.h>
>
>  #include "internal.h"
>
> @@ -70,6 +71,8 @@ struct scan_control {
>
>         struct container *container;    /* Used by containers for reclaiming */
>                                         /* pages when the limit is exceeded  */
> +       int reclaim_pagecache_only;     /* Set when called from
> +                                          pagecache controller */
>  };
>
>  /*
> @@ -474,6 +477,15 @@ static unsigned long shrink_page_list(st
>                         goto keep;
>
>                 VM_BUG_ON(PageActive(page));
> +               /* Take it easy if we are doing only pagecache pages */
> +               if (sc->reclaim_pagecache_only) {
> +                       /* Check if this is a pagecache page they are not mapped */
> +                       if (page_mapped(page))
> +                               goto keep_locked;
> +                       /* Check if this container has exceeded pagecache limit */
> +                       if (!pagecache_acct_page_overlimit(page))
> +                               goto keep_locked;
> +               }
>
>                 sc->nr_scanned++;
>
> @@ -522,7 +534,8 @@ static unsigned long shrink_page_list(st
>                 }
>
>                 if (PageDirty(page)) {
> -                       if (referenced)
> +                       /* Reclaim even referenced pagecache pages if over limit */
> +                       if (!pagecache_acct_page_overlimit(page) && referenced)
>                                 goto keep_locked;
>                         if (!may_enter_fs)
>                                 goto keep_locked;
> @@ -869,6 +882,13 @@ force_reclaim_mapped:
>                 cond_resched();
>                 page = lru_to_page(&l_hold);
>                 list_del(&page->lru);
> +               /* While reclaiming pagecache make it easy */
> +               if (sc->reclaim_pagecache_only) {
> +                       if (page_mapped(page) || !pagecache_acct_page_overlimit(page)) {
> +                               list_add(&page->lru, &l_active);
> +                               continue;
> +                       }
> +               }

Please correct me if I'm wrong.
Here, if page type is mapped or not overlimit, why add it back to active list?
Did  shrink_page_list() is called by shrink_inactive_list()?

-Aubrey

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Aubrey Li]

On 3/6/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>
> The reclaim code is similar to RSS memory controller.  Scan control is
> slightly different since we are targeting different type of pages.
>
> Additionally no mapped pages are touched when scanning for pagecache pages.
>
> RSS memory controller and pagecache controller share common code in reclaim
> and hence pagecache controller patches are dependent on RSS memory controller
> patch even though the features are independently configurable at compile time.
>
> --- linux-2.6.20.orig/mm/vmscan.c
> +++ linux-2.6.20/mm/vmscan.c
> @@ -43,6 +43,7 @@
>
>  #include <linux/swapops.h>
>  #include <linux/memcontrol.h>
> +#include <linux/pagecache_acct.h>
>
>  #include "internal.h"
>
> @@ -70,6 +71,8 @@ struct scan_control {
>
>         struct container *container;    /* Used by containers for reclaiming */
>                                         /* pages when the limit is exceeded  */
> +       int reclaim_pagecache_only;     /* Set when called from
> +                                          pagecache controller */
>  };
>
>  /*
> @@ -474,6 +477,15 @@ static unsigned long shrink_page_list(st
>                         goto keep;
>
>                 VM_BUG_ON(PageActive(page));
> +               /* Take it easy if we are doing only pagecache pages */
> +               if (sc->reclaim_pagecache_only) {
> +                       /* Check if this is a pagecache page they are not mapped */
> +                       if (page_mapped(page))
> +                               goto keep_locked;
> +                       /* Check if this container has exceeded pagecache limit */
> +                       if (!pagecache_acct_page_overlimit(page))
> +                               goto keep_locked;
> +               }
>
>                 sc->nr_scanned++;
>
> @@ -522,7 +534,8 @@ static unsigned long shrink_page_list(st
>                 }
>
>                 if (PageDirty(page)) {
> -                       if (referenced)
> +                       /* Reclaim even referenced pagecache pages if over limit */
> +                       if (!pagecache_acct_page_overlimit(page) && referenced)
>                                 goto keep_locked;
>                         if (!may_enter_fs)
>                                 goto keep_locked;
> @@ -869,6 +882,13 @@ force_reclaim_mapped:
>                 cond_resched();
>                 page = lru_to_page(&l_hold);
>                 list_del(&page->lru);
> +               /* While reclaiming pagecache make it easy */
> +               if (sc->reclaim_pagecache_only) {
> +                       if (page_mapped(page) || !pagecache_acct_page_overlimit(page)) {
> +                               list_add(&page->lru, &l_active);
> +                               continue;
> +                       }
> +               }

Please correct me if I'm wrong.
Here, if page type is mapped or not overlimit, why add it back to active list?
Did  shrink_page_list() is called by shrink_inactive_list()?

-Aubrey

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Aubrey Li wrote:
> On 3/6/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>> The reclaim code is similar to RSS memory controller.  Scan control is
>> slightly different since we are targeting different type of pages.
>>
>> Additionally no mapped pages are touched when scanning for pagecache pages.
>>
>> RSS memory controller and pagecache controller share common code in reclaim
>> and hence pagecache controller patches are dependent on RSS memory controller
>> patch even though the features are independently configurable at compile time.
>>
>> --- linux-2.6.20.orig/mm/vmscan.c
>> +++ linux-2.6.20/mm/vmscan.c
>> @@ -43,6 +43,7 @@
>>
>>  #include <linux/swapops.h>
>>  #include <linux/memcontrol.h>
>> +#include <linux/pagecache_acct.h>
>>
>>  #include "internal.h"
>>
>> @@ -70,6 +71,8 @@ struct scan_control {
>>
>>         struct container *container;    /* Used by containers for reclaiming */
>>                                         /* pages when the limit is exceeded  */
>> +       int reclaim_pagecache_only;     /* Set when called from
>> +                                          pagecache controller */
>>  };
>>
>>  /*
>> @@ -474,6 +477,15 @@ static unsigned long shrink_page_list(st
>>                         goto keep;
>>
>>                 VM_BUG_ON(PageActive(page));
>> +               /* Take it easy if we are doing only pagecache pages */
>> +               if (sc->reclaim_pagecache_only) {
>> +                       /* Check if this is a pagecache page they are not mapped */
>> +                       if (page_mapped(page))
>> +                               goto keep_locked;
>> +                       /* Check if this container has exceeded pagecache limit */
>> +                       if (!pagecache_acct_page_overlimit(page))
>> +                               goto keep_locked;
>> +               }
>>
>>                 sc->nr_scanned++;
>>
>> @@ -522,7 +534,8 @@ static unsigned long shrink_page_list(st
>>                 }
>>
>>                 if (PageDirty(page)) {
>> -                       if (referenced)
>> +                       /* Reclaim even referenced pagecache pages if over limit */
>> +                       if (!pagecache_acct_page_overlimit(page) && referenced)
>>                                 goto keep_locked;
>>                         if (!may_enter_fs)
>>                                 goto keep_locked;
>> @@ -869,6 +882,13 @@ force_reclaim_mapped:
>>                 cond_resched();
>>                 page = lru_to_page(&l_hold);
>>                 list_del(&page->lru);
>> +               /* While reclaiming pagecache make it easy */
>> +               if (sc->reclaim_pagecache_only) {
>> +                       if (page_mapped(page) || !pagecache_acct_page_overlimit(page)) {
>> +                               list_add(&page->lru, &l_active);
>> +                               continue;
>> +                       }
>> +               }
> 
> Please correct me if I'm wrong.
> Here, if page type is mapped or not overlimit, why add it back to active list?
> Did  shrink_page_list() is called by shrink_inactive_list()?

Correct, shrink_page_list() is called from shrink_inactive_list() but
the above code is patched in shrink_active_list().  The
'force_reclaim_mapped' label is from function shrink_active_list() and
not in shrink_page_list() as it may seem in the patch file.

While removing pages from active_list, we want to select only
pagecache pages and leave the remaining in the active_list.
page_mapped() pages are _not_ of interest to pagecache controller
(they will be taken care by rss controller) and hence we put it back.
 Also if the pagecache controller is below limit, no need to reclaim
so we put back all pages and come out.

--Vaidy

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Aubrey Li wrote:
> On 3/6/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>> The reclaim code is similar to RSS memory controller.  Scan control is
>> slightly different since we are targeting different type of pages.
>>
>> Additionally no mapped pages are touched when scanning for pagecache pages.
>>
>> RSS memory controller and pagecache controller share common code in reclaim
>> and hence pagecache controller patches are dependent on RSS memory controller
>> patch even though the features are independently configurable at compile time.
>>
>> --- linux-2.6.20.orig/mm/vmscan.c
>> +++ linux-2.6.20/mm/vmscan.c
>> @@ -43,6 +43,7 @@
>>
>>  #include <linux/swapops.h>
>>  #include <linux/memcontrol.h>
>> +#include <linux/pagecache_acct.h>
>>
>>  #include "internal.h"
>>
>> @@ -70,6 +71,8 @@ struct scan_control {
>>
>>         struct container *container;    /* Used by containers for reclaiming */
>>                                         /* pages when the limit is exceeded  */
>> +       int reclaim_pagecache_only;     /* Set when called from
>> +                                          pagecache controller */
>>  };
>>
>>  /*
>> @@ -474,6 +477,15 @@ static unsigned long shrink_page_list(st
>>                         goto keep;
>>
>>                 VM_BUG_ON(PageActive(page));
>> +               /* Take it easy if we are doing only pagecache pages */
>> +               if (sc->reclaim_pagecache_only) {
>> +                       /* Check if this is a pagecache page they are not mapped */
>> +                       if (page_mapped(page))
>> +                               goto keep_locked;
>> +                       /* Check if this container has exceeded pagecache limit */
>> +                       if (!pagecache_acct_page_overlimit(page))
>> +                               goto keep_locked;
>> +               }
>>
>>                 sc->nr_scanned++;
>>
>> @@ -522,7 +534,8 @@ static unsigned long shrink_page_list(st
>>                 }
>>
>>                 if (PageDirty(page)) {
>> -                       if (referenced)
>> +                       /* Reclaim even referenced pagecache pages if over limit */
>> +                       if (!pagecache_acct_page_overlimit(page) && referenced)
>>                                 goto keep_locked;
>>                         if (!may_enter_fs)
>>                                 goto keep_locked;
>> @@ -869,6 +882,13 @@ force_reclaim_mapped:
>>                 cond_resched();
>>                 page = lru_to_page(&l_hold);
>>                 list_del(&page->lru);
>> +               /* While reclaiming pagecache make it easy */
>> +               if (sc->reclaim_pagecache_only) {
>> +                       if (page_mapped(page) || !pagecache_acct_page_overlimit(page)) {
>> +                               list_add(&page->lru, &l_active);
>> +                               continue;
>> +                       }
>> +               }
> 
> Please correct me if I'm wrong.
> Here, if page type is mapped or not overlimit, why add it back to active list?
> Did  shrink_page_list() is called by shrink_inactive_list()?

Correct, shrink_page_list() is called from shrink_inactive_list() but
the above code is patched in shrink_active_list().  The
'force_reclaim_mapped' label is from function shrink_active_list() and
not in shrink_page_list() as it may seem in the patch file.

While removing pages from active_list, we want to select only
pagecache pages and leave the remaining in the active_list.
page_mapped() pages are _not_ of interest to pagecache controller
(they will be taken care by rss controller) and hence we put it back.
 Also if the pagecache controller is below limit, no need to reclaim
so we put back all pages and come out.

--Vaidy

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Aubrey Li]

On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
> Correct, shrink_page_list() is called from shrink_inactive_list() but
> the above code is patched in shrink_active_list().  The
> 'force_reclaim_mapped' label is from function shrink_active_list() and
> not in shrink_page_list() as it may seem in the patch file.
>
> While removing pages from active_list, we want to select only
> pagecache pages and leave the remaining in the active_list.
> page_mapped() pages are _not_ of interest to pagecache controller
> (they will be taken care by rss controller) and hence we put it back.
>  Also if the pagecache controller is below limit, no need to reclaim
> so we put back all pages and come out.

Oh, I just read the patch, not apply it to my local tree, I'm working
on 2.6.19 now.
So the question is, when vfs pagecache limit is hit, the current
implementation just reclaim few pages, so it's quite possible the
limit is hit again, and hence the reclaim code will be called again
and again, that will impact application performance.

-Aubrey

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Aubrey Li]

On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
> Correct, shrink_page_list() is called from shrink_inactive_list() but
> the above code is patched in shrink_active_list().  The
> 'force_reclaim_mapped' label is from function shrink_active_list() and
> not in shrink_page_list() as it may seem in the patch file.
>
> While removing pages from active_list, we want to select only
> pagecache pages and leave the remaining in the active_list.
> page_mapped() pages are _not_ of interest to pagecache controller
> (they will be taken care by rss controller) and hence we put it back.
>  Also if the pagecache controller is below limit, no need to reclaim
> so we put back all pages and come out.

Oh, I just read the patch, not apply it to my local tree, I'm working
on 2.6.19 now.
So the question is, when vfs pagecache limit is hit, the current
implementation just reclaim few pages, so it's quite possible the
limit is hit again, and hence the reclaim code will be called again
and again, that will impact application performance.

-Aubrey

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Aubrey Li wrote:
> On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>> Correct, shrink_page_list() is called from shrink_inactive_list() but
>> the above code is patched in shrink_active_list().  The
>> 'force_reclaim_mapped' label is from function shrink_active_list() and
>> not in shrink_page_list() as it may seem in the patch file.
>>
>> While removing pages from active_list, we want to select only
>> pagecache pages and leave the remaining in the active_list.
>> page_mapped() pages are _not_ of interest to pagecache controller
>> (they will be taken care by rss controller) and hence we put it back.
>>  Also if the pagecache controller is below limit, no need to reclaim
>> so we put back all pages and come out.
> 
> Oh, I just read the patch, not apply it to my local tree, I'm working
> on 2.6.19 now.
> So the question is, when vfs pagecache limit is hit, the current
> implementation just reclaim few pages, so it's quite possible the
> limit is hit again, and hence the reclaim code will be called again
> and again, that will impact application performance.

Yes, you are correct.  So if we start reclaiming one page at a time,
then the cost of reclaim is very high and we would be calling the
reclaim code too often.  Hence we have a 'buffer zone' or 'reclaim
threshold' or 'push back' around the limit.  In the patch we have a 64
page (256KB) NR_PAGES_RECLAIM_THRESHOLD:

 int pagecache_acct_shrink_used(unsigned long nr_pages)
 {
 	unsigned long ret = 0;
 	atomic_inc(&reclaim_count);
+
+	/* Don't call reclaim for each page above limit */
+	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
+		ret += shrink_container_memory(
+				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
+	}
+
 	return 0;
 }

Hence we do not call the reclaimer if the threshold is exceeded by
just 1 page... we wait for 64 pages or 256KB of pagecache memory to go
 overlimit and then call the reclaimer which will reclaim all 64 pages
in one shot.

This prevents the reclaim code from being called too often and it also
keeps the cost of reclaim low.

In future patches we are planing to have a percentage based reclaim
threshold so that it would scale well with the container size.

--Vaidy

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Aubrey Li wrote:
> On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>> Correct, shrink_page_list() is called from shrink_inactive_list() but
>> the above code is patched in shrink_active_list().  The
>> 'force_reclaim_mapped' label is from function shrink_active_list() and
>> not in shrink_page_list() as it may seem in the patch file.
>>
>> While removing pages from active_list, we want to select only
>> pagecache pages and leave the remaining in the active_list.
>> page_mapped() pages are _not_ of interest to pagecache controller
>> (they will be taken care by rss controller) and hence we put it back.
>>  Also if the pagecache controller is below limit, no need to reclaim
>> so we put back all pages and come out.
> 
> Oh, I just read the patch, not apply it to my local tree, I'm working
> on 2.6.19 now.
> So the question is, when vfs pagecache limit is hit, the current
> implementation just reclaim few pages, so it's quite possible the
> limit is hit again, and hence the reclaim code will be called again
> and again, that will impact application performance.

Yes, you are correct.  So if we start reclaiming one page at a time,
then the cost of reclaim is very high and we would be calling the
reclaim code too often.  Hence we have a 'buffer zone' or 'reclaim
threshold' or 'push back' around the limit.  In the patch we have a 64
page (256KB) NR_PAGES_RECLAIM_THRESHOLD:

 int pagecache_acct_shrink_used(unsigned long nr_pages)
 {
 	unsigned long ret = 0;
 	atomic_inc(&reclaim_count);
+
+	/* Don't call reclaim for each page above limit */
+	if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
+		ret += shrink_container_memory(
+				RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
+	}
+
 	return 0;
 }

Hence we do not call the reclaimer if the threshold is exceeded by
just 1 page... we wait for 64 pages or 256KB of pagecache memory to go
 overlimit and then call the reclaimer which will reclaim all 64 pages
in one shot.

This prevents the reclaim code from being called too often and it also
keeps the cost of reclaim low.

In future patches we are planing to have a percentage based reclaim
threshold so that it would scale well with the container size.

--Vaidy

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Aubrey Li]

On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>
>
> Aubrey Li wrote:
> > On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
> >> Correct, shrink_page_list() is called from shrink_inactive_list() but
> >> the above code is patched in shrink_active_list().  The
> >> 'force_reclaim_mapped' label is from function shrink_active_list() and
> >> not in shrink_page_list() as it may seem in the patch file.
> >>
> >> While removing pages from active_list, we want to select only
> >> pagecache pages and leave the remaining in the active_list.
> >> page_mapped() pages are _not_ of interest to pagecache controller
> >> (they will be taken care by rss controller) and hence we put it back.
> >>  Also if the pagecache controller is below limit, no need to reclaim
> >> so we put back all pages and come out.
> >
> > Oh, I just read the patch, not apply it to my local tree, I'm working
> > on 2.6.19 now.
> > So the question is, when vfs pagecache limit is hit, the current
> > implementation just reclaim few pages, so it's quite possible the
> > limit is hit again, and hence the reclaim code will be called again
> > and again, that will impact application performance.
>
> Yes, you are correct.  So if we start reclaiming one page at a time,
> then the cost of reclaim is very high and we would be calling the
> reclaim code too often.  Hence we have a 'buffer zone' or 'reclaim
> threshold' or 'push back' around the limit.  In the patch we have a 64
> page (256KB) NR_PAGES_RECLAIM_THRESHOLD:
>
>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>  {
>         unsigned long ret = 0;
>         atomic_inc(&reclaim_count);
> +
> +       /* Don't call reclaim for each page above limit */
> +       if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
> +               ret += shrink_container_memory(
> +                               RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
> +       }
> +
>         return 0;
>  }
>
> Hence we do not call the reclaimer if the threshold is exceeded by
> just 1 page... we wait for 64 pages or 256KB of pagecache memory to go
>  overlimit and then call the reclaimer which will reclaim all 64 pages
> in one shot.
>
> This prevents the reclaim code from being called too often and it also
> keeps the cost of reclaim low.
>
> In future patches we are planing to have a percentage based reclaim
> threshold so that it would scale well with the container size.
>
Actually it's not a good idea IMHO. No matter how big the threshold
is, it's not suitable. If it's too small, application performance will
be impacted seriously after pagecache limit is hit. If it's too large,
Limiting pagecache is useless.

Why not reclaim pages as much as possible when the pagecache limit is hit?

-Aubrey

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Aubrey Li]

On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>
>
> Aubrey Li wrote:
> > On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
> >> Correct, shrink_page_list() is called from shrink_inactive_list() but
> >> the above code is patched in shrink_active_list().  The
> >> 'force_reclaim_mapped' label is from function shrink_active_list() and
> >> not in shrink_page_list() as it may seem in the patch file.
> >>
> >> While removing pages from active_list, we want to select only
> >> pagecache pages and leave the remaining in the active_list.
> >> page_mapped() pages are _not_ of interest to pagecache controller
> >> (they will be taken care by rss controller) and hence we put it back.
> >>  Also if the pagecache controller is below limit, no need to reclaim
> >> so we put back all pages and come out.
> >
> > Oh, I just read the patch, not apply it to my local tree, I'm working
> > on 2.6.19 now.
> > So the question is, when vfs pagecache limit is hit, the current
> > implementation just reclaim few pages, so it's quite possible the
> > limit is hit again, and hence the reclaim code will be called again
> > and again, that will impact application performance.
>
> Yes, you are correct.  So if we start reclaiming one page at a time,
> then the cost of reclaim is very high and we would be calling the
> reclaim code too often.  Hence we have a 'buffer zone' or 'reclaim
> threshold' or 'push back' around the limit.  In the patch we have a 64
> page (256KB) NR_PAGES_RECLAIM_THRESHOLD:
>
>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>  {
>         unsigned long ret = 0;
>         atomic_inc(&reclaim_count);
> +
> +       /* Don't call reclaim for each page above limit */
> +       if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
> +               ret += shrink_container_memory(
> +                               RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
> +       }
> +
>         return 0;
>  }
>
> Hence we do not call the reclaimer if the threshold is exceeded by
> just 1 page... we wait for 64 pages or 256KB of pagecache memory to go
>  overlimit and then call the reclaimer which will reclaim all 64 pages
> in one shot.
>
> This prevents the reclaim code from being called too often and it also
> keeps the cost of reclaim low.
>
> In future patches we are planing to have a percentage based reclaim
> threshold so that it would scale well with the container size.
>
Actually it's not a good idea IMHO. No matter how big the threshold
is, it's not suitable. If it's too small, application performance will
be impacted seriously after pagecache limit is hit. If it's too large,
Limiting pagecache is useless.

Why not reclaim pages as much as possible when the pagecache limit is hit?

-Aubrey

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Aubrey Li wrote:
> On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>>
>> Aubrey Li wrote:
>>> On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>>>> Correct, shrink_page_list() is called from shrink_inactive_list() but
>>>> the above code is patched in shrink_active_list().  The
>>>> 'force_reclaim_mapped' label is from function shrink_active_list() and
>>>> not in shrink_page_list() as it may seem in the patch file.
>>>>
>>>> While removing pages from active_list, we want to select only
>>>> pagecache pages and leave the remaining in the active_list.
>>>> page_mapped() pages are _not_ of interest to pagecache controller
>>>> (they will be taken care by rss controller) and hence we put it back.
>>>>  Also if the pagecache controller is below limit, no need to reclaim
>>>> so we put back all pages and come out.
>>> Oh, I just read the patch, not apply it to my local tree, I'm working
>>> on 2.6.19 now.
>>> So the question is, when vfs pagecache limit is hit, the current
>>> implementation just reclaim few pages, so it's quite possible the
>>> limit is hit again, and hence the reclaim code will be called again
>>> and again, that will impact application performance.
>> Yes, you are correct.  So if we start reclaiming one page at a time,
>> then the cost of reclaim is very high and we would be calling the
>> reclaim code too often.  Hence we have a 'buffer zone' or 'reclaim
>> threshold' or 'push back' around the limit.  In the patch we have a 64
>> page (256KB) NR_PAGES_RECLAIM_THRESHOLD:
>>
>>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>>  {
>>         unsigned long ret = 0;
>>         atomic_inc(&reclaim_count);
>> +
>> +       /* Don't call reclaim for each page above limit */
>> +       if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
>> +               ret += shrink_container_memory(
>> +                               RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
>> +       }
>> +
>>         return 0;
>>  }
>>
>> Hence we do not call the reclaimer if the threshold is exceeded by
>> just 1 page... we wait for 64 pages or 256KB of pagecache memory to go
>>  overlimit and then call the reclaimer which will reclaim all 64 pages
>> in one shot.
>>
>> This prevents the reclaim code from being called too often and it also
>> keeps the cost of reclaim low.
>>
>> In future patches we are planing to have a percentage based reclaim
>> threshold so that it would scale well with the container size.
>>
> Actually it's not a good idea IMHO. No matter how big the threshold
> is, it's not suitable. If it's too small, application performance will
> be impacted seriously after pagecache limit is hit. If it's too large,
> Limiting pagecache is useless.
> 
> Why not reclaim pages as much as possible when the pagecache limit is hit?
> 

Well, that seems to be a good suggestion.  We will try it out by
asking the reclaimer to do as much as possible in minimum time/effort.
 However we have to figure out how hard we want to push the reclaimer.
 In fact we can push the shrink_active_list() and
shrink_inactive_list() routines to reclaim the _all_ container pages.
 We do have reclaim priority to play with.  Let see if we can comeup
with some automatic method to reclaim 'good' number of pages each time.

--Vaidy

Re: [PATCH 3/3][RFC] Containers: Pagecache controller reclaim

[Vaidyanathan Srinivasan]

Aubrey Li wrote:
> On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>>
>> Aubrey Li wrote:
>>> On 3/27/07, Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>>>> Correct, shrink_page_list() is called from shrink_inactive_list() but
>>>> the above code is patched in shrink_active_list().  The
>>>> 'force_reclaim_mapped' label is from function shrink_active_list() and
>>>> not in shrink_page_list() as it may seem in the patch file.
>>>>
>>>> While removing pages from active_list, we want to select only
>>>> pagecache pages and leave the remaining in the active_list.
>>>> page_mapped() pages are _not_ of interest to pagecache controller
>>>> (they will be taken care by rss controller) and hence we put it back.
>>>>  Also if the pagecache controller is below limit, no need to reclaim
>>>> so we put back all pages and come out.
>>> Oh, I just read the patch, not apply it to my local tree, I'm working
>>> on 2.6.19 now.
>>> So the question is, when vfs pagecache limit is hit, the current
>>> implementation just reclaim few pages, so it's quite possible the
>>> limit is hit again, and hence the reclaim code will be called again
>>> and again, that will impact application performance.
>> Yes, you are correct.  So if we start reclaiming one page at a time,
>> then the cost of reclaim is very high and we would be calling the
>> reclaim code too often.  Hence we have a 'buffer zone' or 'reclaim
>> threshold' or 'push back' around the limit.  In the patch we have a 64
>> page (256KB) NR_PAGES_RECLAIM_THRESHOLD:
>>
>>  int pagecache_acct_shrink_used(unsigned long nr_pages)
>>  {
>>         unsigned long ret = 0;
>>         atomic_inc(&reclaim_count);
>> +
>> +       /* Don't call reclaim for each page above limit */
>> +       if (nr_pages > NR_PAGES_RECLAIM_THRESHOLD) {
>> +               ret += shrink_container_memory(
>> +                               RECLAIM_PAGECACHE_MEMORY, nr_pages, NULL);
>> +       }
>> +
>>         return 0;
>>  }
>>
>> Hence we do not call the reclaimer if the threshold is exceeded by
>> just 1 page... we wait for 64 pages or 256KB of pagecache memory to go
>>  overlimit and then call the reclaimer which will reclaim all 64 pages
>> in one shot.
>>
>> This prevents the reclaim code from being called too often and it also
>> keeps the cost of reclaim low.
>>
>> In future patches we are planing to have a percentage based reclaim
>> threshold so that it would scale well with the container size.
>>
> Actually it's not a good idea IMHO. No matter how big the threshold
> is, it's not suitable. If it's too small, application performance will
> be impacted seriously after pagecache limit is hit. If it's too large,
> Limiting pagecache is useless.
> 
> Why not reclaim pages as much as possible when the pagecache limit is hit?
> 

Well, that seems to be a good suggestion.  We will try it out by
asking the reclaimer to do as much as possible in minimum time/effort.
 However we have to figure out how hard we want to push the reclaimer.
 In fact we can push the shrink_active_list() and
shrink_inactive_list() routines to reclaim the _all_ container pages.
 We do have reclaim priority to play with.  Let see if we can comeup
with some automatic method to reclaim 'good' number of pages each time.

--Vaidy

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 0/3][RFC] Containers: Pagecache accounting and control subsystem (v1)

[Vaidyanathan Srinivasan]

-----------------------------------------------------------

This patch adds pagecache accounting and control on top of 
Paul's container subsystem v7 posted at 
http://lkml.org/lkml/2007/2/12/88

and Balbir's RSS controller posted at
http://lkml.org/lkml/2007/2/26/8

This patchset depends on Balbir's RSS controller and cannot 
work independent of it. The page reclaim code has been merged 
with container RSS controller.  However compile time options 
can individually enable/disable memory controller and/or 
pagecache controller.

Comments, suggestions and criticisms are welcome.

Features:
--------
* New subsystem called 'pagecache_acct' is registered with containers 
* Container pointer is added to struct address_space to keep track of  
  associated container
* In filemap.c and swap_state.c, the corresponding container's 
  pagecache_acct subsystem is charged and uncharged whenever a new 
  page is added or removed from pagecache
* The accounting number include pages in swap cache and filesystem 
  buffer pages apart from pagecache, basically everything under 
  NR_FILE_PAGES is counted as pagecache.  However this excluded 
  mapped and anonymous pages
* Limits on pagecache can be set by echo 100000 > pagecache_limit on 
  the /container file system.  The unit is in kilobytes  
* If the pagecache utilisation limit is exceeded, pagecache reclaim
  code is invoked to recover dirty and clean pagecache pages only.

Advantages:
-----------
* Does not add container pointers in struct page

Limitations:
-----------
* Code is not safe for container deletion/task migration
* Pagecache page reclaim needs performance improvements
* Global LRU is churned in search of pagecache pages

Usage:
-----

* Add patch on top of Paul container (v7) at kernel version 2.6.20
* Enable CONFIG_CONTAINER_PAGECACHE_ACCT in 'General Setup'
* Boot new kernel
* Mount container filesystem 
	mount -t container /container
	cd /container
* Create new container
	mkdir mybox 
	cd /container/mybox
* Add current shell to container
	echo $$ > tasks
* There are two files pagecache_usage and pagecache_limit
* In order to set limit, echo value in kilobytes to pagecache_limit
	echo 100000 > pagecache_limit 
	#This would set 100MB limit on pagecache usage
* Trash the system from current shell using scp/cp/dd/tar etc
* Watch pagecache_usage and /proc/meminfo to verify behavior

* Only unmapped pagecache data will be accounted and controlled.  
  These are memory used by cp, scp, tar etc.  While file mmap will 
  be controlled by Balbir's RSS controller.

Tests:
------

* Ran kernbench within container with pagecache_limits set

ToDo:
----

* Merge with container RSS controller and eliminate redundant code
* Test and support task migration and container deletion
* Review reclaim performance
* Optimise page reclaim

Patch Series:
-------------
pagecache-controller-setup.patch
pagecache-controller-acct.patch
pagecache-controller-reclaim.patch

-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 0/3][RFC] Containers: Pagecache accounting and control subsystem (v1)

[Vaidyanathan Srinivasan]

-----------------------------------------------------------

This patch adds pagecache accounting and control on top of 
Paul's container subsystem v7 posted at 
http://lkml.org/lkml/2007/2/12/88

and Balbir's RSS controller posted at
http://lkml.org/lkml/2007/2/19/10

This patchset depends on Balbir's RSS controller and cannot 
work independent of it. The page reclaim code has been merged 
with container RSS controller.  However compile time options 
can individually enable/disable memory controller and/or 
pagecache controller.

Comments, suggestions and criticisms are welcome.

Features:
--------
* New subsystem called 'pagecache_acct' is registered with containers 
* Container pointer is added to struct address_space to keep track of  
  associated container
* In filemap.c and swap_state.c, the corresponding container's 
  pagecache_acct subsystem is charged and uncharged whenever a new 
  page is added or removed from pagecache
* The accounting number include pages in swap cache and filesystem 
  buffer pages apart from pagecache, basically everything under 
  NR_FILE_PAGES is counted as pagecache.  However this excluded 
  mapped and anonymous pages
* Limits on pagecache can be set by echo 100000 > pagecache_limit on 
  the /container file system.  The unit is in kilobytes  
* If the pagecache utilisation limit is exceeded, pagecache reclaim
  code is invoked to recover dirty and clean pagecache pages only.

Advantages:
-----------
* Does not add container pointers in struct page

Limitations:
-----------
* Code is not safe for container deletion/task migration
* Pagecache page reclaim needs performance improvements
* Global LRU is churned in search of pagecache pages

Usage:
-----

* Add patch on top of Paul container (v7) at kernel version 2.6.20
* Enable CONFIG_CONTAINER_PAGECACHE_ACCT in 'General Setup'
* Boot new kernel
* Mount container filesystem 
	mount -t container /container
	cd /container
* Create new container
	mkdir mybox 
	cd /container/mybox
* Add current shell to container
	echo $$ > tasks
* There are two files pagecache_usage and pagecache_limit
* In order to set limit, echo value in kilobytes to pagecache_limit
	echo 100000 > pagecache_limit 
	#This would set 100MB limit on pagecache usage
* Trash the system from current shell using scp/cp/dd/tar etc
* Watch pagecache_usage and /proc/meminfo to verify behavior

* Only unmapped pagecache data will be accounted and controlled.  
  These are memory used by cp, scp, tar etc.  While file mmap will 
  be controlled by Balbir's RSS controller.

ToDo:
----

* Merge with container RSS controller and eliminate redundant code
* Support and test task migration and container deletion
* Review reclaim performance
* Optimise page reclaim
	
-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>