palloc-3.15.patch

diff --git a/include/linux/cgroup_subsys.h b/include/linux/cgroup_subsys.h
index 768fe44..8238c7c 100644
--- a/include/linux/cgroup_subsys.h
+++ b/include/linux/cgroup_subsys.h
@@ -50,6 +50,9 @@ SUBSYS(net_prio)
 #if IS_ENABLED(CONFIG_CGROUP_HUGETLB)
 SUBSYS(hugetlb)
 #endif
-/*
- * DO NOT ADD ANY SUBSYSTEM WITHOUT EXPLICIT ACKS FROM CGROUP MAINTAINERS.
- */
+
+/* */
+
+#if IS_ENABLED(CONFIG_CGROUP_PALLOC)
+SUBSYS(palloc)
+#endif
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index fac5509..3656155 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -69,6 +69,14 @@ enum {
 #  define is_migrate_cma(migratetype) false
 #endif
 
+#ifdef CONFIG_CGROUP_PALLOC
+/* Determine the number of bins according to the bits required for
+   each component of the address*/
+#  define MAX_PALLOC_BITS 8
+#  define MAX_PALLOC_BINS (1 << MAX_PALLOC_BITS)
+#  define COLOR_BITMAP(name) DECLARE_BITMAP(name, MAX_PALLOC_BINS)
+#endif
+
 #define for_each_migratetype_order(order, type) \
 	for (order = 0; order < MAX_ORDER; order++) \
 		for (type = 0; type < MIGRATE_TYPES; type++)
@@ -370,6 +378,14 @@ struct zone {
 #endif
 	struct free_area	free_area[MAX_ORDER];
 
+#ifdef CONFIG_CGROUP_PALLOC
+	/*
+	 * Color page cache. for movable type free pages of order-0
+	 */
+	struct list_head        color_list[MAX_PALLOC_BINS];
+	COLOR_BITMAP(color_bitmap);
+#endif
+
 #ifndef CONFIG_SPARSEMEM
 	/*
 	 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
diff --git a/include/linux/palloc.h b/include/linux/palloc.h
new file mode 100644
index 0000000..ec4c092
--- /dev/null
+++ b/include/linux/palloc.h
@@ -0,0 +1,33 @@
+#ifndef _LINUX_PALLOC_H
+#define _LINUX_PALLOC_H
+
+/*
+ * kernel/palloc.h
+ *
+ * PHysical memory aware allocator
+ */
+
+#include <linux/types.h>
+#include <linux/cgroup.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+
+#ifdef CONFIG_CGROUP_PALLOC
+
+struct palloc {
+	struct cgroup_subsys_state css;
+	COLOR_BITMAP(cmap);
+};
+
+/* Retrieve the palloc group corresponding to this cgroup container */
+struct palloc *cgroup_ph(struct cgroup *cgrp);
+
+/* Retrieve the palloc group corresponding to this subsys */
+struct palloc * ph_from_subsys(struct cgroup_subsys_state * subsys);
+
+/* return #of palloc bins */
+int palloc_bins(void);
+
+#endif /* CONFIG_CGROUP_PALLOC */
+
+#endif /* _LINUX_PALLOC_H */
diff --git a/init/Kconfig b/init/Kconfig
index 765018c..84776c1 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1089,6 +1089,12 @@ config DEBUG_BLK_CGROUP
 	Enable some debugging help. Currently it exports additional stat
 	files in a cgroup which can be useful for debugging.
 
+config CGROUP_PALLOC
+	bool "Enable PALLOC"
+	help
+	  Enables PALLOC: physical address based page allocator that 
+	  replaces the buddy allocator.
+
 endif # CGROUPS
 
 config CHECKPOINT_RESTORE
diff --git a/mm/Makefile b/mm/Makefile
index b484452..cc1e594 100644
--- a/mm/Makefile
+++ b/mm/Makefile
@@ -63,3 +63,5 @@ obj-$(CONFIG_MEMORY_ISOLATION) += page_isolation.o
 obj-$(CONFIG_ZBUD)	+= zbud.o
 obj-$(CONFIG_ZSMALLOC)	+= zsmalloc.o
 obj-$(CONFIG_GENERIC_EARLY_IOREMAP) += early_ioremap.o
+obj-$(CONFIG_CGROUP_PALLOC) += palloc.o
+
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 5dba293..b5b0f09 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1,3 +1,4 @@
+
 /*
  *  linux/mm/page_alloc.c
  *
@@ -63,12 +64,194 @@
 #include <linux/sched/rt.h>
 
 #include <asm/sections.h>
+#include <linux/debugfs.h>
 #include <asm/tlbflush.h>
 #include <asm/div64.h>
 #include "internal.h"
 
+
 /* prevent >1 _updater_ of zone percpu pageset ->high and ->batch fields */
 static DEFINE_MUTEX(pcp_batch_high_lock);
+#ifdef CONFIG_CGROUP_PALLOC
+#include <linux/palloc.h>
+
+int memdbg_enable = 0;
+EXPORT_SYMBOL(memdbg_enable);
+
+static int sysctl_alloc_balance = 0;
+/* palloc address bitmask */
+static unsigned long sysctl_palloc_mask = 0x0;
+
+static int mc_xor_bits[64];
+static int use_mc_xor = 0;
+static int use_palloc = 0;
+
+DEFINE_PER_CPU(unsigned long, palloc_rand_seed);
+
+#define memdbg(lvl, fmt, ...)					\
+        do {                                                    \
+		if(memdbg_enable >= lvl)			\
+			trace_printk(fmt, ##__VA_ARGS__);       \
+        } while(0)
+
+struct palloc_stat {
+	s64 max_ns;
+	s64 min_ns;
+	s64 tot_ns;
+
+	s64 tot_cnt;
+	s64 iter_cnt;      /* avg_iter = iter_cnt/tot_cnt */
+
+	s64 cache_hit_cnt; /* hit rate = cache_hit_cnt / cache_acc_cnt */
+	s64 cache_acc_cnt;
+
+	s64 flush_cnt;
+
+	s64 alloc_balance;
+	s64 alloc_balance_timeout;
+	ktime_t start;     /* start time of the current iteration */
+};
+
+static struct {
+        u32 enabled;
+	int colors;
+	struct palloc_stat stat[3]; /* 0 - color, 1 - normal, 2 - fail */
+} palloc;
+
+static void palloc_flush(struct zone *zone);
+
+static ssize_t palloc_write(struct file *filp, const char __user *ubuf,
+				      size_t cnt, loff_t *ppos)
+{
+        char buf[64];
+	int i;
+        if (cnt > 63) cnt = 63;
+        if (copy_from_user(&buf, ubuf, cnt))
+                return -EFAULT;
+
+	if (!strncmp(buf, "reset", 5)) {
+		printk(KERN_INFO "reset statistics...\n");
+		for (i = 0; i < ARRAY_SIZE(palloc.stat); i++) {
+			memset(&palloc.stat[i], 0, sizeof(struct palloc_stat));
+			palloc.stat[i].min_ns = 0x7fffffff;
+		}
+	} else if (!strncmp(buf, "flush", 5)) {
+		struct zone *zone;
+		printk(KERN_INFO "flush color cache...\n");
+		for_each_populated_zone(zone) {
+			unsigned long flags;
+			if (!zone)
+				continue;
+			spin_lock_irqsave(&zone->lock, flags);
+			palloc_flush(zone);
+			spin_unlock_irqrestore(&zone->lock, flags);
+		}
+	} else if (!strncmp(buf, "xor", 3)) {
+		int bit, xor_bit;
+		sscanf(buf + 4, "%d %d", &bit, &xor_bit);
+		if ((bit > 0 && bit < 64) &&
+		    (xor_bit > 0 && xor_bit < 64) &&
+		    bit != xor_bit)
+		{
+			mc_xor_bits[bit] = xor_bit;
+		}
+	}
+
+        *ppos += cnt;
+        return cnt;
+}
+
+static int palloc_show(struct seq_file *m, void *v)
+{
+	int i, tmp;
+	char *desc[] = { "Color", "Normal", "Fail" };
+	char buf[256];
+	for (i = 0; i < 3; i++) {
+		struct palloc_stat *stat = &palloc.stat[i];
+		seq_printf(m, "statistics %s:\n", desc[i]);
+		seq_printf(m, "  min(ns)/max(ns)/avg(ns)/tot_cnt: %lld %lld %lld %lld\n",
+			   stat->min_ns,
+			   stat->max_ns,
+			   (stat->tot_cnt) ? div64_u64(stat->tot_ns, stat->tot_cnt) : 0,
+			   stat->tot_cnt);
+		seq_printf(m, "  hit rate: %lld/%lld (%lld %%)\n",
+			   stat->cache_hit_cnt, stat->cache_acc_cnt,
+			   (stat->cache_acc_cnt) ?
+			   div64_u64(stat->cache_hit_cnt*100, stat->cache_acc_cnt) : 0);
+		seq_printf(m, "  avg iter: %lld (%lld/%lld)\n",
+			   (stat->tot_cnt) ?
+			   div64_u64(stat->iter_cnt, stat->tot_cnt) : 0,
+			   stat->iter_cnt, stat->tot_cnt);
+		seq_printf(m, "  flush cnt: %lld\n", stat->flush_cnt);
+
+		seq_printf(m, "  balance: %lld | fail: %lld\n",
+			   stat->alloc_balance, stat->alloc_balance_timeout);
+	}
+	seq_printf(m, "mask: 0x%lx\n", sysctl_palloc_mask);
+	tmp = bitmap_weight(&sysctl_palloc_mask, sizeof(unsigned long)*8);
+	seq_printf(m, "weight: %d  (bins: %d)\n", tmp, 1<<tmp);
+	bitmap_scnlistprintf(buf, 256, &sysctl_palloc_mask, sizeof(unsigned long)*8);
+	seq_printf(m, "bits: %s\n", buf);
+
+	seq_printf(m, "XOR bits: %s\n", (use_mc_xor) ? "enabled" : "disabled");
+	for (i = 0; i < 64; i++) {
+		if (mc_xor_bits[i] > 0)
+			seq_printf(m, "   %3d <-> %3d\n", i, mc_xor_bits[i]);
+	}
+
+	seq_printf(m, "Use PALLOC: %s\n", (use_palloc) ? "enabled" : "disabled");
+        return 0;
+}
+static int palloc_open(struct inode *inode, struct file *filp)
+{
+        return single_open(filp, palloc_show, NULL);
+}
+
+static const struct file_operations palloc_fops = {
+        .open           = palloc_open,
+        .write          = palloc_write,
+        .read           = seq_read,
+        .llseek         = seq_lseek,
+        .release        = single_release,
+};
+
+static int __init palloc_debugfs(void)
+{
+        umode_t mode = S_IFREG | S_IRUSR | S_IWUSR;
+        struct dentry *dir;
+	int i;
+
+        dir = debugfs_create_dir("palloc", NULL);
+
+	/* statistics initialization */
+	for (i = 0; i < ARRAY_SIZE(palloc.stat); i++) {
+		memset(&palloc.stat[i], 0, sizeof(struct palloc_stat));
+		palloc.stat[i].min_ns = 0x7fffffff;
+	}
+
+        if (!dir)
+                return PTR_ERR(dir);
+        if (!debugfs_create_file("control", mode, dir, NULL, &palloc_fops))
+                goto fail;
+	if (!debugfs_create_u64("palloc_mask", mode, dir, (u64 *)&sysctl_palloc_mask))
+		goto fail;
+        if (!debugfs_create_u32("use_mc_xor", mode, dir, &use_mc_xor))
+                goto fail;
+        if (!debugfs_create_u32("use_palloc", mode, dir, &use_palloc))
+                goto fail;
+        if (!debugfs_create_u32("debug_level", mode, dir, &memdbg_enable))
+                goto fail;
+	if (!debugfs_create_u32("alloc_balance", mode, dir, &sysctl_alloc_balance))
+		goto fail;
+        return 0;
+fail:
+        debugfs_remove_recursive(dir);
+        return -ENOMEM;
+}
+
+late_initcall(palloc_debugfs);
+
+#endif /* CONFIG_CGROUP_PALLOC */
 
 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
 DEFINE_PER_CPU(int, numa_node);
@@ -907,12 +1090,314 @@ static int prep_new_page(struct page *page, int order, gfp_t gfp_flags)
 	return 0;
 }
 
+#ifdef CONFIG_CGROUP_PALLOC
+
+int palloc_bins(void)
+{
+	return min((1 << bitmap_weight(&sysctl_palloc_mask, 8*sizeof (unsigned long))),
+		   MAX_PALLOC_BINS);
+}
+
+static inline int page_to_color(struct page *page)
+{
+	int color = 0;
+	int idx = 0;
+	int c;
+	unsigned long paddr = page_to_phys(page);
+	for_each_set_bit(c, &sysctl_palloc_mask, sizeof(unsigned long) * 8) {
+		if (use_mc_xor) {
+			if (((paddr >> c) & 0x1) ^ ((paddr >> mc_xor_bits[c]) & 0x1))
+				color |= (1<<idx);
+		} else {
+			if ((paddr >> c) & 0x1)
+				color |= (1<<idx);
+		}
+		idx++;
+	}
+	return color;
+}
+
+/* debug */
+static inline unsigned long list_count(struct list_head *head)
+{
+	unsigned long n = 0;
+	struct list_head *curr;
+	list_for_each(curr, head)
+		n++;
+	return n;
+}
+
+/* move all color_list pages into free_area[0].freelist[2]
+ * zone->lock must be hold before calling this function
+ */
+static void palloc_flush(struct zone *zone)
+{
+	int c;
+	struct page *page;
+	memdbg(2, "flush the ccache for zone %s\n", zone->name);
+
+	while (1) {
+		for (c = 0; c < MAX_PALLOC_BINS; c++) {
+			if (!list_empty(&zone->color_list[c])) {
+				page = list_entry(zone->color_list[c].next,
+						  struct page, lru);
+				list_del_init(&page->lru);
+				__free_one_page(page, zone, 0, get_pageblock_migratetype(page));
+				zone->free_area[0].nr_free--;
+			}
+
+			if (list_empty(&zone->color_list[c])) {
+				bitmap_clear(zone->color_bitmap, c, 1);
+				INIT_LIST_HEAD(&zone->color_list[c]);
+			}
+		}
+
+		if (bitmap_weight(zone->color_bitmap, MAX_PALLOC_BINS) == 0)
+			break;
+	}
+}
+
+/* move a page (size=1<<order) into a order-0 colored cache */
+static void palloc_insert(struct zone *zone, struct page *page, int order)
+{
+	int i, color;
+	/* 1 page (2^order) -> 2^order x pages of colored cache. */
+
+	/* remove from zone->free_area[order].free_list[mt] */
+	list_del(&page->lru);
+	zone->free_area[order].nr_free--;
+
+	/* insert pages to zone->color_list[] (all order-0) */
+	for (i = 0; i < (1<<order); i++) {
+		color = page_to_color(&page[i]);
+		/* add to zone->color_list[color] */
+		memdbg(5, "- add pfn %ld (0x%08llx) to color_list[%d]\n",
+		       page_to_pfn(&page[i]), (u64)page_to_phys(&page[i]), color);
+		INIT_LIST_HEAD(&page[i].lru);
+		list_add_tail(&page[i].lru, &zone->color_list[color]);
+		bitmap_set(zone->color_bitmap, color, 1);
+		zone->free_area[0].nr_free++;
+		rmv_page_order(&page[i]);
+	}
+	memdbg(4, "add order=%d zone=%s\n", order, zone->name);
+}
+
+/* return a colored page (order-0) and remove it from the colored cache */
+static inline struct page *palloc_find_cmap(struct zone *zone, COLOR_BITMAP(cmap),
+				     int order,
+				     struct palloc_stat *stat)
+{
+	struct page *page;
+	COLOR_BITMAP(tmpmask);
+	int c;
+	unsigned int tmp_idx;
+	int found_w, want_w;
+	unsigned long rand_seed;
+	/* cache statistics */
+	if (stat) stat->cache_acc_cnt++;
+
+	/* find color cache entry */
+	if (!bitmap_intersects(zone->color_bitmap, cmap, MAX_PALLOC_BINS))
+		return NULL;
+	
+	bitmap_and(tmpmask, zone->color_bitmap, cmap, MAX_PALLOC_BINS);
+
+	/* must have a balance. */
+	found_w = bitmap_weight(tmpmask, MAX_PALLOC_BINS);
+	want_w  = bitmap_weight(cmap, MAX_PALLOC_BINS);
+	if (sysctl_alloc_balance &&
+	    found_w < want_w &&
+	    found_w < min(sysctl_alloc_balance, want_w) &&
+	    memdbg_enable)
+	{
+		ktime_t dur = ktime_sub(ktime_get(), stat->start);
+		if (dur.tv64 < 1000000) {
+			/* try to balance unless order=MAX-2 or 1ms has passed */
+			memdbg(4, "found_w=%d want_w=%d order=%d elapsed=%lld ns\n",
+			       found_w, want_w, order, dur.tv64);
+			stat->alloc_balance++;
+
+			return NULL;
+		}
+		stat->alloc_balance_timeout++;
+	}
+
+	/* choose a bit among the candidates */
+	if (sysctl_alloc_balance && memdbg_enable) {
+		rand_seed = (unsigned long)stat->start.tv64;
+	} else {
+		rand_seed = per_cpu(palloc_rand_seed, smp_processor_id())++;
+		if (rand_seed > MAX_PALLOC_BINS)
+			per_cpu(palloc_rand_seed, smp_processor_id()) = 0;
+	}
+
+	tmp_idx = rand_seed % found_w;
+	for_each_set_bit(c, tmpmask, MAX_PALLOC_BINS) {
+		if (tmp_idx-- <= 0)
+			break;
+	}
+
+
+	BUG_ON(c >= MAX_PALLOC_BINS);
+	BUG_ON(list_empty(&zone->color_list[c]));
+
+	page = list_entry(zone->color_list[c].next, struct page, lru);
+
+	memdbg(1, "Found colored page pfn %ld color %d seed %ld found/want %d/%d\n",
+	       page_to_pfn(page), c, rand_seed, found_w, want_w);
+
+	/* remove from the zone->color_list[color] */
+	list_del(&page->lru);
+	if (list_empty(&zone->color_list[c]))
+		bitmap_clear(zone->color_bitmap, c, 1);
+	zone->free_area[0].nr_free--;
+
+	memdbg(5, "- del pfn %ld from color_list[%d]\n",
+	       page_to_pfn(page), c);
+
+	if (stat) stat->cache_hit_cnt++;
+	return page;
+}
+
+static inline void
+update_stat(struct palloc_stat *stat, struct page *page, int iters)
+{
+	ktime_t dur;
+
+	if (memdbg_enable == 0)
+		return;
+
+	dur = ktime_sub(ktime_get(), stat->start);
+
+	if(dur.tv64 > 0) {
+		stat->min_ns = min(dur.tv64, stat->min_ns);
+		stat->max_ns = max(dur.tv64, stat->max_ns);
+
+		stat->tot_ns += dur.tv64;
+		stat->iter_cnt += iters;
+
+		stat->tot_cnt++;
+
+		memdbg(2, "order %ld pfn %ld(0x%08llx) color %d iters %d in %lld ns\n",
+		       page_order(page), page_to_pfn(page), (u64)page_to_phys(page),
+		       (int)page_to_color(page),
+		       iters, dur.tv64);
+	} else {
+		memdbg(5, "dur %lld is < 0\n", dur.tv64);
+	}
+}
+
 /*
  * Go through the free lists for the given migratetype and remove
  * the smallest available page from the freelists
  */
 static inline
 struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
+				int migratetype)
+{
+	unsigned int current_order;
+	struct free_area *area;
+	struct list_head *curr, *tmp;
+	struct page *page;
+
+	struct palloc *ph;
+	struct palloc_stat *c_stat = &palloc.stat[0];
+	struct palloc_stat *n_stat = &palloc.stat[1];
+	struct palloc_stat *f_stat = &palloc.stat[2];
+	int iters = 0;
+	COLOR_BITMAP(tmpcmap);
+	unsigned long *cmap;
+
+	if (memdbg_enable)
+		c_stat->start = n_stat->start = f_stat->start = ktime_get();
+
+	if (!use_palloc)
+		goto normal_buddy_alloc;
+
+	/* cgroup information */
+	ph = ph_from_subsys(current->cgroups->subsys[palloc_cgrp_id]);
+	if (ph && bitmap_weight(ph->cmap, MAX_PALLOC_BINS) > 0)
+		cmap = ph->cmap;
+	else {
+		bitmap_fill(tmpcmap, MAX_PALLOC_BINS);
+		cmap = tmpcmap;
+	}
+
+	page = NULL;
+	if (order == 0) {
+		/* find in the cache */
+		memdbg(5, "check color cache (mt=%d)\n", migratetype);
+		page = palloc_find_cmap(zone, cmap, 0, c_stat);
+
+		if (page) {
+			update_stat(c_stat, page, iters);
+			return page;
+		}
+	}
+
+	if (order == 0) {
+		/* build color cache */
+		iters++;
+		/* search the entire list. make color cache in the process  */
+		for (current_order = 0;
+		     current_order < MAX_ORDER; ++current_order)
+		{
+			area = &(zone->free_area[current_order]);
+			if (list_empty(&area->free_list[migratetype]))
+				continue;
+			memdbg(3, " order=%d (nr_free=%ld)\n",
+			       current_order, area->nr_free);
+			list_for_each_safe(curr, tmp,
+					   &area->free_list[migratetype])
+			{
+				iters++;
+				page = list_entry(curr, struct page, lru);
+				palloc_insert(zone, page, current_order);
+				page = palloc_find_cmap(zone, cmap, current_order, c_stat);
+				if (page) {
+					update_stat(c_stat, page, iters);
+					memdbg(1, "Found at Zone %s pfn 0x%lx\n",
+					       zone->name,
+					       page_to_pfn(page));
+					return page;
+				}
+			}
+		}
+		memdbg(1, "Failed to find a matching color\n");
+	} else {
+	normal_buddy_alloc:
+		/* normal buddy */
+		/* Find a page of the appropriate size in the preferred list */
+		for (current_order = order;
+		     current_order < MAX_ORDER; ++current_order)
+		{
+			area = &(zone->free_area[current_order]);
+			iters++;
+			if (list_empty(&area->free_list[migratetype]))
+				continue;
+			page = list_entry(area->free_list[migratetype].next,
+					  struct page, lru);
+
+			list_del(&page->lru);
+			rmv_page_order(page);
+			area->nr_free--;
+			expand(zone, page, order,
+			       current_order, area, migratetype);
+
+			update_stat(n_stat, page, iters);
+			return page;
+		}
+	}
+	/* no memory (color or normal) found in this zone */
+	memdbg(1, "No memory in Zone %s: order %d mt %d\n",
+	       zone->name, order, migratetype);
+
+	return NULL;
+}
+#else /* !CONFIG_CGROUP_PALLOC */
+
+static inline
+struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
 						int migratetype)
 {
 	unsigned int current_order;
@@ -936,7 +1421,7 @@ struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
 
 	return NULL;
 }
-
+#endif /* CONFIG_CGROUP_PALLOC */
 
 /*
  * This array describes the order lists are fallen back to when
@@ -1528,9 +2013,13 @@ struct page *buffered_rmqueue(struct zone *preferred_zone,
 	unsigned long flags;
 	struct page *page;
 	int cold = !!(gfp_flags & __GFP_COLD);
+	struct palloc * ph;
+
+	ph = ph_from_subsys(current->cgroups->subsys[palloc_cgrp_id]);
 
 again:
-	if (likely(order == 0)) {
+	/* Skip PCP when physically-aware allocation is requested */
+	if (likely(order == 0) && !ph) {
 		struct per_cpu_pages *pcp;
 		struct list_head *list;
 
@@ -4096,6 +4585,15 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
 static void __meminit zone_init_free_lists(struct zone *zone)
 {
 	int order, t;
+
+#ifdef CONFIG_CGROUP_PALLOC
+	int c;
+	for (c = 0; c < MAX_PALLOC_BINS; c++) {
+		INIT_LIST_HEAD(&zone->color_list[c]);
+	}
+	bitmap_zero(zone->color_bitmap, MAX_PALLOC_BINS);
+#endif /* CONFIG_CGROUP_PALLOC */
+
 	for_each_migratetype_order(order, t) {
 		INIT_LIST_HEAD(&zone->free_area[order].free_list[t]);
 		zone->free_area[order].nr_free = 0;
@@ -6420,6 +6918,9 @@ __offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
 		return;
 	zone = page_zone(pfn_to_page(pfn));
 	spin_lock_irqsave(&zone->lock, flags);
+#ifdef CONFIG_CGROUP_PALLOC
+	palloc_flush(zone);
+#endif
 	pfn = start_pfn;
 	while (pfn < end_pfn) {
 		if (!pfn_valid(pfn)) {
diff --git a/mm/palloc.c b/mm/palloc.c
new file mode 100644
index 0000000..b9acae1
--- /dev/null
+++ b/mm/palloc.c
@@ -0,0 +1,166 @@
+/*
+ * kernel/palloc.c
+ *
+ * Physical driven User Space Allocator info for a set of tasks.
+ */
+
+#include <linux/types.h>
+#include <linux/cgroup.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/palloc.h>
+#include <linux/mm.h>
+#include <linux/err.h>
+#include <linux/fs.h>
+#include <linux/bitmap.h>
+#include <linux/module.h>
+
+/*
+ * Check if a page is compliant to the policy defined for the given vma
+ */
+#ifdef CONFIG_CGROUP_PALLOC
+
+#define MAX_LINE_LEN (6*128)
+/*
+ * Types of files in a palloc group
+ * FILE_PALLOC - contain list of palloc bins allowed
+*/
+typedef enum {
+	FILE_PALLOC,
+} palloc_filetype_t;
+
+/*
+ * Top level palloc - mask initialized to zero implying no restriction on
+ * physical pages
+*/
+
+static struct palloc top_palloc;
+
+/* Retrieve the palloc group corresponding to this cgroup container */
+struct palloc *cgroup_ph(struct cgroup *cgrp)
+{
+	return container_of(cgrp->subsys[palloc_cgrp_id],
+			    struct palloc, css);
+}
+
+struct palloc * ph_from_subsys(struct cgroup_subsys_state * subsys)
+{
+	return container_of(subsys, struct palloc, css);
+}
+
+/*
+ * Common write function for files in palloc cgroup
+ */
+static int update_bitmask(unsigned long *bitmap, const char *buf, int maxbits)
+{
+	int retval = 0;
+
+	if (!*buf)
+		bitmap_clear(bitmap, 0, maxbits);
+	else
+		retval = bitmap_parselist(buf, bitmap, maxbits);
+
+	return retval;
+}
+
+
+static int palloc_file_write(struct cgroup_subsys_state *css, struct cftype *cft,
+                char *buffer)
+{
+	int retval = 0;
+	struct palloc *ph = container_of(css, struct palloc, css);
+
+	switch (cft->private) {
+	case FILE_PALLOC:
+		retval = update_bitmask(ph->cmap, buffer, palloc_bins());
+		printk(KERN_INFO "Bins : %s\n", buffer);
+		break;
+	default:
+		retval = -EINVAL;
+		break;
+	}
+
+	return retval;
+}
+
+static int palloc_file_read(struct seq_file *sf, void *v)
+{
+	struct cgroup_subsys_state *css = seq_css(sf);
+	struct cftype *cft = seq_cft(sf);
+	struct palloc *ph = container_of(css, struct palloc, css);
+	char *page;
+	ssize_t retval = 0;
+	char *s;
+
+	if (!(page = (char *)__get_free_page(GFP_TEMPORARY|__GFP_ZERO)))
+		return -ENOMEM;
+
+	s = page;
+
+	switch (cft->private) {
+	case FILE_PALLOC:
+		s += bitmap_scnlistprintf(s, PAGE_SIZE, ph->cmap, palloc_bins());
+		*s++ = '\n';
+		printk(KERN_INFO "Bins : %s", page);
+		break;
+	default:
+		retval = -EINVAL;
+		goto out;
+	}
+
+	retval = seq_printf(sf, "%s", page);
+out:
+	free_page((unsigned long)page);
+	return retval;
+}
+
+
+/*
+ * struct cftype: handler definitions for cgroup control files
+ *
+ * for the common functions, 'private' gives the type of the file
+ */
+static struct cftype files[] = {
+	{
+		.name = "bins",
+		.seq_show = palloc_file_read,
+		.write_string = palloc_file_write,
+		.max_write_len = MAX_LINE_LEN,
+		.private = FILE_PALLOC,
+	},
+	{ }	/* terminate */
+};
+
+/*
+ * palloc_create - create a palloc group
+ */
+static struct cgroup_subsys_state *palloc_create(struct cgroup_subsys_state *css)
+{
+	struct palloc * ph_child;
+	ph_child = kmalloc(sizeof(struct palloc), GFP_KERNEL);
+	if(!ph_child)
+		return ERR_PTR(-ENOMEM);
+
+	bitmap_clear(ph_child->cmap, 0, MAX_PALLOC_BINS);
+	return &ph_child->css;
+}
+
+
+/*
+ * Destroy an existing palloc group
+ */
+static void palloc_destroy(struct cgroup_subsys_state *css)
+{
+	struct palloc *ph = container_of(css, struct palloc, css);
+	kfree(ph);
+}
+
+struct cgroup_subsys palloc_cgrp_subsys = {
+	.name = "palloc",
+	.css_alloc = palloc_create,
+	.css_free = palloc_destroy,
+	.id = palloc_cgrp_id,
+	.base_cftypes = files,
+};
+
+#endif /* CONFIG_CGROUP_PALLOC */
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 302dd07..b5753cd 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -23,6 +23,8 @@
 
 #include "internal.h"
 
+#include <linux/palloc.h>
+
 #ifdef CONFIG_VM_EVENT_COUNTERS
 DEFINE_PER_CPU(struct vm_event_state, vm_event_states) = {{0}};
 EXPORT_PER_CPU_SYMBOL(vm_event_states);
@@ -876,6 +878,38 @@ static void frag_show_print(struct seq_file *m, pg_data_t *pgdat,
 						struct zone *zone)
 {
 	int order;
+#ifdef CONFIG_CGROUP_PALLOC
+#include <linux/palloc.h>
+	int color, mt;
+	int cnt, bins;
+	struct free_area *area;
+	struct list_head *curr;
+
+	seq_printf(m, "-------\n");
+	/* order by memory type */
+	for (mt = 0; mt < MIGRATE_ISOLATE; mt++) {
+		seq_printf(m, "- %17s[%d]", "mt", mt);
+		for (order = 0; order < MAX_ORDER; order++) {
+			area = &(zone->free_area[order]);
+			cnt = 0;
+			list_for_each(curr, &area->free_list[mt])
+				cnt++;
+			seq_printf(m, "%6d ", cnt);
+		}
+		seq_printf(m, "\n");
+	}
+	/* order by color */
+	seq_printf(m, "-------\n");
+	bins = palloc_bins();
+
+	for (color = 0; color < bins; color++) {
+		seq_printf(m, "- color [%d:%0x]", color, color);
+		cnt = 0;
+		list_for_each(curr, &zone->color_list[color])
+			cnt++;
+		seq_printf(m, "%6d\n", cnt);
+	}
+#endif /* !CONFIG_CGROUP_PALLOC */
 
 	seq_printf(m, "Node %d, zone %8s ", pgdat->node_id, zone->name);
 	for (order = 0; order < MAX_ORDER; ++order)
@@ -883,6 +917,7 @@ static void frag_show_print(struct seq_file *m, pg_data_t *pgdat,
 	seq_putc(m, '\n');
 }
 
+
 /*
  * This walks the free areas for each zone.
  */