This is the core io-throttle kernel infrastructure. It creates the basic
interfaces to cgroups and implements the I/O measurement and throttling
functions.

Signed-off-by: Andrea Righi <righi.andrea@gmail.com>
---
 block/Makefile                  |    2 +
 block/blk-io-throttle.c         |  549 +++++++++++++++++++++++++++++++++++++++
 include/linux/blk-io-throttle.h |   41 +++
 include/linux/cgroup_subsys.h   |    6 +
 init/Kconfig                    |   10 +
 5 files changed, 608 insertions(+), 0 deletions(-)
 create mode 100644 block/blk-io-throttle.c
 create mode 100644 include/linux/blk-io-throttle.h

diff --git a/block/Makefile b/block/Makefile
index 208000b..b3afc86 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -13,6 +13,8 @@ obj-$(CONFIG_IOSCHED_AS)	+= as-iosched.o
 obj-$(CONFIG_IOSCHED_DEADLINE)	+= deadline-iosched.o
 obj-$(CONFIG_IOSCHED_CFQ)	+= cfq-iosched.o
 
+obj-$(CONFIG_CGROUP_IO_THROTTLE)	+= blk-io-throttle.o
+
 obj-$(CONFIG_BLK_DEV_IO_TRACE)	+= blktrace.o
 obj-$(CONFIG_BLOCK_COMPAT)	+= compat_ioctl.o
 obj-$(CONFIG_BLK_DEV_INTEGRITY)	+= blk-integrity.o
diff --git a/block/blk-io-throttle.c b/block/blk-io-throttle.c
new file mode 100644
index 0000000..82100a6
--- /dev/null
+++ b/block/blk-io-throttle.c
@@ -0,0 +1,549 @@
+/*
+ * blk-io-throttle.c
+ *
+ * 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 021110-1307, USA.
+ *
+ * Copyright (C) 2008 Andrea Righi <righi.andrea@gmail.com>
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/cgroup.h>
+#include <linux/slab.h>
+#include <linux/gfp.h>
+#include <linux/err.h>
+#include <linux/sched.h>
+#include <linux/genhd.h>
+#include <linux/fs.h>
+#include <linux/jiffies.h>
+#include <linux/hardirq.h>
+#include <linux/list.h>
+#include <linux/seq_file.h>
+#include <linux/spinlock.h>
+#include <linux/uaccess.h>
+#include <linux/blk-io-throttle.h>
+
+/* The various types of throttling algorithms */
+enum iothrottle_strategy {
+	IOTHROTTLE_LEAKY_BUCKET,
+	IOTHROTTLE_TOKEN_BUCKET,
+};
+
+/**
+ * struct iothrottle_node - throttling rule of a single block device
+ * @node: list of per block device throttling rules
+ * @dev: block device number, used as key in the list
+ * @iorate: max i/o bandwidth (in bytes/s)
+ * @strategy: throttling strategy
+ * @timestamp: timestamp of the last I/O request (in jiffies)
+ * @stat: i/o activity counter (leaky bucket only)
+ * @bucket_size: bucket size in bytes (token bucket only)
+ * @token: token counter (token bucket only)
+ *
+ * Define a i/o throttling rule for a single block device.
+ *
+ * NOTE: limiting rules always refer to dev_t; if a block device is unplugged
+ * the limiting rules defined for that device persist and they are still valid
+ * if a new device is plugged and it uses the same dev_t number.
+ */
+struct iothrottle_node {
+	struct list_head node;
+	dev_t dev;
+	u64 iorate;
+	enum iothrottle_strategy strategy;
+	unsigned long timestamp;
+	atomic_long_t stat;
+	s64 bucket_size;
+	atomic_long_t token;
+};
+
+/**
+ * struct iothrottle - throttling rules for a cgroup
+ * @css: pointer to the cgroup state
+ * @lock: spinlock used to protect write operations in the list
+ * @list: list of iothrottle_node elements
+ *
+ * Define multiple per-block device i/o throttling rules.
+ * Note: the list of the throttling rules is protected by RCU locking.
+ */
+struct iothrottle {
+	struct cgroup_subsys_state css;
+	spinlock_t lock;
+	struct list_head list;
+};
+
+static inline struct iothrottle *cgroup_to_iothrottle(struct cgroup *cgrp)
+{
+	return container_of(cgroup_subsys_state(cgrp, iothrottle_subsys_id),
+			    struct iothrottle, css);
+}
+
+static inline struct iothrottle *task_to_iothrottle(struct task_struct *task)
+{
+	return container_of(task_subsys_state(task, iothrottle_subsys_id),
+			    struct iothrottle, css);
+}
+
+/*
+ * Note: called with rcu_read_lock() or iot->lock held.
+ */
+static struct iothrottle_node *
+iothrottle_search_node(const struct iothrottle *iot, dev_t dev)
+{
+	struct iothrottle_node *n;
+
+	list_for_each_entry_rcu(n, &iot->list, node)
+		if (n->dev == dev)
+			return n;
+	return NULL;
+}
+
+/*
+ * Note: called with iot->lock held.
+ */
+static inline void iothrottle_insert_node(struct iothrottle *iot,
+						struct iothrottle_node *n)
+{
+	list_add_rcu(&n->node, &iot->list);
+}
+
+/*
+ * Note: called with iot->lock held.
+ */
+static inline void
+iothrottle_replace_node(struct iothrottle *iot, struct iothrottle_node *old,
+			struct iothrottle_node *new)
+{
+	list_replace_rcu(&old->node, &new->node);
+}
+
+/*
+ * Note: called with iot->lock held.
+ */
+static struct iothrottle_node *
+iothrottle_delete_node(struct iothrottle *iot, dev_t dev)
+{
+	struct iothrottle_node *n;
+
+	list_for_each_entry(n, &iot->list, node)
+		if (n->dev == dev) {
+			list_del_rcu(&n->node);
+			return n;
+		}
+	return NULL;
+}
+
+/*
+ * Note: called from kernel/cgroup.c with cgroup_lock() held.
+ */
+static struct cgroup_subsys_state *
+iothrottle_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+	struct iothrottle *iot;
+
+	iot = kmalloc(sizeof(*iot), GFP_KERNEL);
+	if (unlikely(!iot))
+		return ERR_PTR(-ENOMEM);
+
+	INIT_LIST_HEAD(&iot->list);
+	spin_lock_init(&iot->lock);
+
+	return &iot->css;
+}
+
+/*
+ * Note: called from kernel/cgroup.c with cgroup_lock() held.
+ */
+static void iothrottle_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+	struct iothrottle_node *n, *p;
+	struct iothrottle *iot = cgroup_to_iothrottle(cgrp);
+
+	/*
+	 * don't worry about locking here, at this point there must be not any
+	 * reference to the list.
+	 */
+	list_for_each_entry_safe(n, p, &iot->list, node)
+		kfree(n);
+	kfree(iot);
+}
+
+static int iothrottle_read(struct cgroup *cgrp, struct cftype *cft,
+				struct seq_file *m)
+{
+	struct iothrottle *iot = cgroup_to_iothrottle(cgrp);
+	struct iothrottle_node *n;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(n, &iot->list, node) {
+		unsigned long delta;
+
+		BUG_ON(!n->dev);
+		delta = jiffies_to_msecs((long)jiffies - (long)n->timestamp);
+		seq_printf(m, "%u %u %llu %u %li %lli %li %lu\n",
+				MAJOR(n->dev), MINOR(n->dev), n->iorate,
+				n->strategy, atomic_long_read(&n->stat),
+				n->bucket_size, atomic_long_read(&n->token),
+				delta);
+	}
+	rcu_read_unlock();
+	return 0;
+}
+
+static dev_t devname2dev_t(const char *buf)
+{
+	struct block_device *bdev;
+	dev_t dev = 0;
+	struct gendisk *disk;
+	int part;
+
+	/* use a lookup to validate the block device */
+	bdev = lookup_bdev(buf);
+	if (IS_ERR(bdev))
+		return 0;
+
+	/* only entire devices are allowed, not single partitions */
+	disk = get_gendisk(bdev->bd_dev, &part);
+	if (disk && !part) {
+		BUG_ON(!bdev->bd_inode);
+		dev = bdev->bd_inode->i_rdev;
+	}
+	bdput(bdev);
+
+	return dev;
+}
+
+/*
+ * The userspace input string must use one of the following syntax:
+ *
+ * dev:0			<- delete a limiting rule
+ * dev:bw-limit:0		<- leaky bucket throttling rule
+ * dev:bw-limit:1:bucket-size	<- token bucket throttling rule
+ */
+static int iothrottle_parse_args(char *buf, size_t nbytes, dev_t *dev,
+					u64 *iorate,
+					enum iothrottle_strategy *strategy,
+					s64 *bucket_size)
+{
+	char *p = buf;
+	int count = 0;
+	char *s[3];
+	unsigned long strategy_val;
+	int ret;
+
+	/* split the colon-delimited input string into its elements */
+	memset(s, 0, sizeof(s));
+	while (count < ARRAY_SIZE(s)) {
+		p = memchr(p, ':', buf + nbytes - p);
+		if (!p)
+			break;
+		*p++ = '\0';
+		if (p >= buf + nbytes)
+			break;
+		s[count++] = p;
+	}
+
+	/* i/o bandwidth limit */
+	if (!s[0])
+		return -EINVAL;
+	ret = strict_strtoull(s[0], 10, iorate);
+	if (ret < 0)
+		return ret;
+	if (!*iorate) {
+		/*
+		 * we're deleting a limiting rule, so just ignore the other
+		 * parameters
+		 */
+		*strategy = 0;
+		*bucket_size = 0;
+		goto out;
+	}
+	*iorate = ALIGN(*iorate, 1024);
+
+	/* throttling strategy */
+	if (!s[1])
+		return -EINVAL;
+	ret = strict_strtoul(s[1], 10, &strategy_val);
+	if (ret < 0)
+		return ret;
+	*strategy = (enum iothrottle_strategy)strategy_val;
+	switch (*strategy) {
+	case IOTHROTTLE_LEAKY_BUCKET:
+		/* leaky bucket ignores bucket size */
+		*bucket_size = 0;
+		goto out;
+	case IOTHROTTLE_TOKEN_BUCKET:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* bucket size */
+	if (!s[2])
+		return -EINVAL;
+	ret = strict_strtoll(s[2], 10, bucket_size);
+	if (ret < 0)
+		return ret;
+	if (*bucket_size < 0)
+		return -EINVAL;
+	*bucket_size = ALIGN(*bucket_size, 1024);
+out:
+
+	/* block device number */
+	*dev = devname2dev_t(buf);
+	return *dev ? 0 : -EINVAL;
+}
+
+static int iothrottle_write(struct cgroup *cgrp, struct cftype *cft,
+				const char *buffer)
+{
+	struct iothrottle *iot;
+	struct iothrottle_node *n, *newn = NULL;
+	dev_t dev;
+	u64 iorate;
+	enum iothrottle_strategy strategy;
+	s64 bucket_size;
+	char *buf;
+	size_t nbytes = strlen(buffer);
+	int ret = 0;
+
+	buf = kmalloc(nbytes + 1, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+	memcpy(buf, buffer, nbytes + 1);
+
+	ret = iothrottle_parse_args(buf