[PATCH] blktrace: remove check_time, we always need to check it

[blktrace.git] / kernel / blk-trace-2.6.16-rc1-git-T0

diff --git a/block/Kconfig b/block/Kconfig
index 377f6dd..27eaed9 100644
--- a/block/Kconfig
+++ b/block/Kconfig
@@ -11,4 +11,15 @@ config LBD
 	  your machine, or if you want to have a raid or loopback device
 	  bigger than 2TB.  Otherwise say N.
 
+config BLK_DEV_IO_TRACE
+	bool "Support for tracing block io actions"
+	select RELAYFS_FS
+	help
+	  Say Y here, if you want to be able to trace the block layer actions
+	  on a given queue. Tracing allows you to see any traffic happening
+	  on a block device queue. For more information (and the user space
+	  support tools needed), fetch the blktrace app from:
+
+	  git://brick.kernel.dk/data/git/blktrace.git
+
 source block/Kconfig.iosched
diff --git a/block/Makefile b/block/Makefile
index 7e4f93e..c05de0e 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -8,3 +8,5 @@ obj-$(CONFIG_IOSCHED_NOOP)	+= noop-iosch
 obj-$(CONFIG_IOSCHED_AS)	+= as-iosched.o
 obj-$(CONFIG_IOSCHED_DEADLINE)	+= deadline-iosched.o
 obj-$(CONFIG_IOSCHED_CFQ)	+= cfq-iosched.o
+
+obj-$(CONFIG_BLK_DEV_IO_TRACE)	+= blktrace.o
diff --git a/block/blktrace.c b/block/blktrace.c
new file mode 100644
index 0000000..d12d166
--- /dev/null
+++ b/block/blktrace.c
@@ -0,0 +1,360 @@
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/blkdev.h>
+#include <linux/blktrace_api.h>
+#include <linux/percpu.h>
+#include <linux/init.h>
+#include <linux/mutex.h>
+#include <asm/uaccess.h>
+
+static DEFINE_PER_CPU(unsigned long long, blk_trace_cpu_offset) = { 0, };
+
+void __blk_add_trace(struct blk_trace *bt, sector_t sector, int bytes,
+		     int rw, u32 what, int error, int pdu_len, void *pdu_data)
+{
+	struct blk_io_trace t;
+	unsigned long flags;
+	void *rbuf;
+	pid_t pid;
+	int cpu;
+
+	if (rw & (1 << BIO_RW_BARRIER))
+		what |= BLK_TC_ACT(BLK_TC_BARRIER);
+	if (rw & (1 << BIO_RW_SYNC))
+		what |= BLK_TC_ACT(BLK_TC_SYNC);
+
+	if (rw & WRITE)
+		what |= BLK_TC_ACT(BLK_TC_WRITE);
+	else
+		what |= BLK_TC_ACT(BLK_TC_READ);
+
+	if (((bt->act_mask << BLK_TC_SHIFT) & what) == 0)
+		return;
+	if (sector < bt->start_lba || sector > bt->end_lba)
+		return;
+
+	pid = current->pid;
+	if (bt->pid && pid != bt->pid)
+		return;
+
+	t.magic		= BLK_IO_TRACE_MAGIC | BLK_IO_TRACE_VERSION;
+
+	t.device	= bt->dev;
+	t.sector	= sector;
+	t.bytes		= bytes;
+	t.action	= what;
+	t.error		= error;
+	t.pdu_len	= pdu_len;
+
+	t.pid		= pid;
+	memcpy(t.comm, current->comm, sizeof(t.comm));
+
+	/*
+	 * need to serialize this part completely to prevent multiple CPUs
+	 * from misordering events
+	 */
+	spin_lock_irqsave(&bt->lock, flags);
+
+	t.sequence	= ++bt->sequence;
+
+	cpu = smp_processor_id();
+	t.cpu		= cpu;
+	t.time		= sched_clock() - per_cpu(blk_trace_cpu_offset, cpu);
+
+	rbuf = relay_reserve(bt->rchan, sizeof(t) + pdu_len);
+	if (rbuf) {
+		memcpy(rbuf, &t, sizeof(t));
+		if (pdu_len)
+			memcpy(rbuf + sizeof(t), pdu_data, pdu_len);
+	}
+
+	spin_unlock_irqrestore(&bt->lock, flags);
+}
+
+EXPORT_SYMBOL_GPL(__blk_add_trace);
+
+static struct dentry *blk_tree_root;
+static struct mutex blk_tree_mutex;
+
+static inline void blk_remove_root(void)
+{
+	if (relayfs_remove_dir(blk_tree_root) != -ENOTEMPTY)
+		blk_tree_root = NULL;
+}
+
+static void blk_remove_tree(struct dentry *dir)
+{
+	mutex_lock(&blk_tree_mutex);
+	relayfs_remove_dir(dir);
+	blk_remove_root();
+	mutex_unlock(&blk_tree_mutex);
+}
+
+static struct dentry *blk_create_tree(const char *blk_name)
+{
+	struct dentry *dir = NULL;
+
+	mutex_lock(&blk_tree_mutex);
+
+	if (!blk_tree_root) {
+		blk_tree_root = relayfs_create_dir("block", NULL);
+		if (!blk_tree_root)
+			goto err;
+	}
+
+	dir = relayfs_create_dir(blk_name, blk_tree_root);
+	if (!dir)
+		blk_remove_root();
+
+err:
+	mutex_unlock(&blk_tree_mutex);
+	return dir;
+}
+
+void blk_cleanup_trace(struct blk_trace *bt)
+{
+	relay_close(bt->rchan);
+	relayfs_remove_file(bt->dropped_file);
+	blk_remove_tree(bt->dir);
+	kfree(bt);
+}
+
+int blk_stop_trace(struct block_device *bdev)
+{
+	request_queue_t *q = bdev_get_queue(bdev);
+	struct blk_trace *bt = NULL;
+	int ret = -EINVAL;
+
+	if (!q)
+		return -ENXIO;
+
+	down(&bdev->bd_sem);
+
+	if (q->blk_trace) {
+		bt = q->blk_trace;
+		q->blk_trace = NULL;
+		ret = 0;
+	}
+
+	up(&bdev->bd_sem);
+
+	if (bt)
+		blk_cleanup_trace(bt);
+
+	return ret;
+}
+
+static int blk_dropped_open(struct inode *inode, struct file *filp)
+{
+	filp->private_data = inode->u.generic_ip;
+	
+	return 0;
+}
+
+static ssize_t blk_dropped_read(struct file *filp, char __user *buffer,
+				size_t count, loff_t *ppos)
+{
+	struct blk_trace *bt = filp->private_data;
+	char buf[16];
+
+	snprintf(buf, sizeof(buf), "%u\n", atomic_read(&bt->dropped));
+	
+	return simple_read_from_buffer(buffer, count, ppos, buf, strlen(buf));
+}
+
+static struct file_operations blk_dropped_fops = {
+	.owner =	THIS_MODULE,
+	.open =		blk_dropped_open,
+	.read =		blk_dropped_read,
+};
+
+static int blk_subbuf_start_callback (struct rchan_buf *buf,
+				      void *subbuf,
+				      void *prev_subbuf,
+				      size_t prev_padding)
+{
+	if (relay_buf_full(buf)) {
+		struct blk_trace *bt = buf->chan->private_data;
+
+		atomic_inc(&bt->dropped);
+		return 0;
+	}
+
+	return 1;
+}
+
+static struct rchan_callbacks blk_relay_callbacks = {
+	.subbuf_start = blk_subbuf_start_callback,
+};
+
+int blk_start_trace(struct block_device *bdev, char __user *arg)
+{
+	request_queue_t *q = bdev_get_queue(bdev);
+	struct blk_user_trace_setup buts;
+	struct blk_trace *bt = NULL;
+	struct dentry *dir = NULL;
+	char b[BDEVNAME_SIZE];
+	int ret, i;
+
+	if (!q)
+		return -ENXIO;
+
+	if (copy_from_user(&buts, arg, sizeof(buts)))
+		return -EFAULT;
+
+	if (!buts.buf_size || !buts.buf_nr)
+		return -EINVAL;
+
+	strcpy(buts.name, bdevname(bdev, b));
+
+	/*
+	 * some device names have larger paths - convert the slashes
+	 * to underscores for this to work as expected
+	 */
+	for (i = 0; i < strlen(buts.name); i++)
+		if (buts.name[i] == '/')
+			buts.name[i] = '_';
+
+	if (copy_to_user(arg, &buts, sizeof(buts)))
+		return -EFAULT;
+
+	down(&bdev->bd_sem);
+	ret = -EBUSY;
+	if (q->blk_trace)
+		goto err;
+
+	ret = -ENOMEM;
+	bt = kmalloc(sizeof(*bt), GFP_KERNEL);
+	if (!bt)
+		goto err;
+
+	ret = -ENOENT;
+	dir = blk_create_tree(buts.name);
+	if (!dir)
+		goto err;
+
+	bt->dir = dir;
+	bt->dev = bdev->bd_dev;
+	bt->sequence = 0;
+	spin_lock_init(&bt->lock);
+	atomic_set(&bt->dropped, 0);
+
+	ret = -EIO;
+	bt->dropped_file = relayfs_create_file("dropped", dir, 0, &blk_dropped_fops, bt);
+	if (!bt->dropped_file)
+		goto err;
+
+	bt->rchan = relay_open("trace", dir, buts.buf_size, buts.buf_nr, &blk_relay_callbacks);
+	if (!bt->rchan)
+		goto err;
+	bt->rchan->private_data = bt;
+
+	bt->act_mask = buts.act_mask;
+	if (!bt->act_mask)
+		bt->act_mask = (u16) -1;
+
+	bt->start_lba = buts.start_lba;
+	bt->end_lba = buts.end_lba;
+	if (!bt->end_lba)
+		bt->end_lba = -1ULL;
+
+	bt->pid = buts.pid;
+
+	q->blk_trace = bt;
+	up(&bdev->bd_sem);
+	return 0;
+err:
+	up(&bdev->bd_sem);
+	if (bt && bt->dropped_file)
+		relayfs_remove_file(bt->dropped_file);
+	if (dir)
+		blk_remove_tree(dir);
+	if (bt)
+		kfree(bt);
+	return ret;
+}
+
+/*
+ * Average offset over two calls to sched_clock() with a gettimeofday()
+ * in the middle
+ */
+static void blk_check_time(unsigned long long *t)
+{
+	unsigned long long a, b;
+	struct timeval tv;
+
+	a = sched_clock();
+	do_gettimeofday(&tv);
+	b = sched_clock();
+
+	*t = tv.tv_sec * 1000000000 + tv.tv_usec * 1000;
+	*t -= (a + b) / 2;
+}
+
+static void blk_trace_check_cpu_time(void *data)
+{
+	unsigned long long *t;
+	int cpu = get_cpu();
+
+	t = &per_cpu(blk_trace_cpu_offset, cpu);
+
+	/*
+	 * Just call it twice, hopefully the second call will be cache hot
+	 * and a little more precise
+	 */
+	blk_check_time(t);
+	blk_check_time(t);
+
+	put_cpu();
+}
+
+/*
+ * Call blk_trace_check_cpu_time() on each CPU to calibrate our inter-CPU
+ * timings
+ */
+static void blk_trace_calibrate_offsets(void)
+{
+	unsigned long flags;
+
+	smp_call_function(blk_trace_check_cpu_time, NULL, 1, 1);
+	local_irq_save(flags);
+	blk_trace_check_cpu_time(NULL);
+	local_irq_restore(flags);
+}
+
+static void blk_trace_set_ht_offsets(void)
+{
+#if defined(CONFIG_SCHED_SMT)
+	int cpu, i;
+
+	/*
+	 * now make sure HT siblings have the same time offset
+	 */
+	preempt_disable();
+	for_each_online_cpu(cpu) {
+		unsigned long long *cpu_off, *sibling_off;
+
+		for_each_cpu_mask(i, cpu_sibling_map[cpu]) {
+			if (i == cpu)
+				continue;
+
+			cpu_off = &per_cpu(blk_trace_cpu_offset, cpu);
+			sibling_off = &per_cpu(blk_trace_cpu_offset, i);
+			*sibling_off = *cpu_off;
+		}
+	}
+	preempt_enable();
+#endif
+}
+
+static __init int blk_trace_init(void)
+{
+	mutex_init(&blk_tree_mutex);
+	blk_trace_calibrate_offsets();
+	blk_trace_set_ht_offsets();
+
+	return 0;
+}
+
+module_init(blk_trace_init);
+
diff --git a/block/elevator.c b/block/elevator.c
index c9f424d..793c686 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -33,6 +33,7 @@
 #include <linux/init.h>
 #include <linux/compiler.h>
 #include <linux/delay.h>
+#include <linux/blktrace_api.h>
 
 #include <asm/uaccess.h>
 
@@ -333,6 +334,8 @@ void __elv_add_request(request_queue_t *
 	struct list_head *pos;
 	unsigned ordseq;
 
+	blk_add_trace_rq(q, rq, BLK_TA_INSERT);
+
 	if (q->ordcolor)
 		rq->flags |= REQ_ORDERED_COLOR;
 
@@ -491,6 +494,7 @@ struct request *elv_next_request(request
 			 * not be passed by new incoming requests
 			 */
 			rq->flags |= REQ_STARTED;
+			blk_add_trace_rq(q, rq, BLK_TA_ISSUE);
 		}
 
 		if (!q->boundary_rq || q->boundary_rq == rq) {
diff --git a/block/ioctl.c b/block/ioctl.c
index e110949..63e67a2 100644
--- a/block/ioctl.c
+++ b/block/ioctl.c
@@ -5,6 +5,7 @@
 #include <linux/backing-dev.h>
 #include <linux/buffer_head.h>
 #include <linux/smp_lock.h>
+#include <linux/blktrace_api.h>
 #include <asm/uaccess.h>
 
 static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user *arg)
@@ -189,6 +190,10 @@ static int blkdev_locked_ioctl(struct fi
 		return put_ulong(arg, bdev->bd_inode->i_size >> 9);
 	case BLKGETSIZE64:
 		return put_u64(arg, bdev->bd_inode->i_size);
+	case BLKSTARTTRACE:
+		return blk_start_trace(bdev, (char __user *) arg);
+	case BLKSTOPTRACE:
+		return blk_stop_trace(bdev);
 	}
 	return -ENOIOCTLCMD;
 }
diff --git a/block/ll_rw_blk.c b/block/ll_rw_blk.c
index 8e27d0a..bfcde0f 100644
--- a/block/ll_rw_blk.c
+++ b/block/ll_rw_blk.c
@@ -28,6 +28,7 @@
 #include <linux/writeback.h>
 #include <linux/interrupt.h>
 #include <linux/cpu.h>
+#include <linux/blktrace_api.h>
 
 /*
  * for max sense size
@@ -1555,8 +1556,10 @@ void blk_plug_device(request_queue_t *q)
 	if (test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags))
 		return;
 
-	if (!test_and_set_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags))
+	if (!test_and_set_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags)) {
 		mod_timer(&q->unplug_timer, jiffies + q->unplug_delay);
+		blk_add_trace_generic(q, NULL, 0, BLK_TA_PLUG);
+	}
 }
 
 EXPORT_SYMBOL(blk_plug_device);
@@ -1620,14 +1623,21 @@ static void blk_backing_dev_unplug(struc
 	/*
 	 * devices don't necessarily have an ->unplug_fn defined
 	 */
-	if (q->unplug_fn)
+	if (q->unplug_fn) {
+		blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_IO, NULL,
+					q->rq.count[READ] + q->rq.count[WRITE]);
+
 		q->unplug_fn(q);
+	}
 }
 
 static void blk_unplug_work(void *data)
 {
 	request_queue_t *q = data;
 
+	blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_IO, NULL,
+				q->rq.count[READ] + q->rq.count[WRITE]);
+
 	q->unplug_fn(q);
 }
 
@@ -1635,6 +1645,9 @@ static void blk_unplug_timeout(unsigned 
 {
 	request_queue_t *q = (request_queue_t *)data;
 
+	blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_TIMER, NULL,
+				q->rq.count[READ] + q->rq.count[WRITE]);
+
 	kblockd_schedule_work(&q->unplug_work);
 }
 
@@ -1757,6 +1770,11 @@ void blk_cleanup_queue(request_queue_t *
 	if (q->queue_tags)
 		__blk_queue_free_tags(q);
 
+	if (q->blk_trace) {
+		blk_cleanup_trace(q->blk_trace);
+		q->blk_trace = NULL;
+	}
+
 	kmem_cache_free(requestq_cachep, q);
 }
 
@@ -2108,6 +2126,8 @@ rq_starved:
 	
 	rq_init(q, rq);
 	rq->rl = rl;
+
+	blk_add_trace_generic(q, bio, rw, BLK_TA_GETRQ);
 out:
 	return rq;
 }
@@ -2136,6 +2156,8 @@ static struct request *get_request_wait(
 		if (!rq) {
 			struct io_context *ioc;
 
+			blk_add_trace_generic(q, bio, rw, BLK_TA_SLEEPRQ);
+
 			__generic_unplug_device(q);
 			spin_unlock_irq(q->queue_lock);
 			io_schedule();
@@ -2189,6 +2211,8 @@ EXPORT_SYMBOL(blk_get_request);
  */
 void blk_requeue_request(request_queue_t *q, struct request *rq)
 {
+	blk_add_trace_rq(q, rq, BLK_TA_REQUEUE);
+
 	if (blk_rq_tagged(rq))
 		blk_queue_end_tag(q, rq);
 
@@ -2820,6 +2844,8 @@ static int __make_request(request_queue_
 			if (!q->back_merge_fn(q, req, bio))
 				break;
 
+			blk_add_trace_bio(q, bio, BLK_TA_BACKMERGE);
+
 			req->biotail->bi_next = bio;
 			req->biotail = bio;
 			req->nr_sectors = req->hard_nr_sectors += nr_sectors;
@@ -2835,6 +2861,8 @@ static int __make_request(request_queue_
 			if (!q->front_merge_fn(q, req, bio))
 				break;
 
+			blk_add_trace_bio(q, bio, BLK_TA_FRONTMERGE);
+
 			bio->bi_next = req->bio;
 			req->bio = bio;
 
@@ -2952,6 +2980,7 @@ void generic_make_request(struct bio *bi
 	request_queue_t *q;
 	sector_t maxsector;
 	int ret, nr_sectors = bio_sectors(bio);
+	dev_t old_dev;
 
 	might_sleep();
 	/* Test device or partition size, when known. */
@@ -2978,6 +3007,8 @@ void generic_make_request(struct bio *bi
 	 * NOTE: we don't repeat the blk_size check for each new device.
 	 * Stacking drivers are expected to know what they are doing.
 	 */
+	maxsector = -1;
+	old_dev = 0;
 	do {
 		char b[BDEVNAME_SIZE];
 
@@ -3010,6 +3041,15 @@ end_io:
 		 */
 		blk_partition_remap(bio);
 
+		if (maxsector != -1)
+			blk_add_trace_remap(q, bio, old_dev, bio->bi_sector, 
+					    maxsector);
+
+		blk_add_trace_bio(q, bio, BLK_TA_QUEUE);
+
+		maxsector = bio->bi_sector;
+		old_dev = bio->bi_bdev->bd_dev;
+
 		ret = q->make_request_fn(q, bio);
 	} while (ret);
 }
@@ -3129,6 +3169,8 @@ static int __end_that_request_first(stru
 	int total_bytes, bio_nbytes, error, next_idx = 0;
 	struct bio *bio;
 
+	blk_add_trace_rq(req->q, req, BLK_TA_COMPLETE);
+
 	/*
 	 * extend uptodate bool to allow < 0 value to be direct io error
 	 */
diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
index 12d7b9b..880892a 100644
--- a/drivers/block/cciss.c
+++ b/drivers/block/cciss.c
@@ -38,6 +38,7 @@
 #include <linux/hdreg.h>
 #include <linux/spinlock.h>
 #include <linux/compat.h>
+#include <linux/blktrace_api.h>
 #include <asm/uaccess.h>
 #include <asm/io.h>
 
@@ -2330,6 +2331,7 @@ static inline void complete_command( ctl
 
 	cmd->rq->completion_data = cmd;
 	cmd->rq->errors = status;
+	blk_add_trace_rq(cmd->rq->q, cmd->rq, BLK_TA_COMPLETE);
 	blk_complete_request(cmd->rq);
 }
 
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 8c16359..8c979c2 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -17,6 +17,7 @@
 #include <linux/mempool.h>
 #include <linux/slab.h>
 #include <linux/idr.h>
+#include <linux/blktrace_api.h>
 
 static const char *_name = DM_NAME;
 
@@ -303,6 +304,8 @@ static void dec_pending(struct dm_io *io
 			/* nudge anyone waiting on suspend queue */
 			wake_up(&io->md->wait);
 
+		blk_add_trace_bio(io->md->queue, io->bio, BLK_TA_COMPLETE);
+
 		bio_endio(io->bio, io->bio->bi_size, io->error);
 		free_io(io->md, io);
 	}
@@ -361,6 +364,7 @@ static void __map_bio(struct dm_target *
 		      struct target_io *tio)
 {
 	int r;
+	sector_t sector;
 
 	/*
 	 * Sanity checks.
@@ -376,10 +380,17 @@ static void __map_bio(struct dm_target *
 	 * this io.
 	 */
 	atomic_inc(&tio->io->io_count);
+	sector = clone->bi_sector;
 	r = ti->type->map(ti, clone, &tio->info);
-	if (r > 0)
+	if (r > 0) {
 		/* the bio has been remapped so dispatch it */
+
+		blk_add_trace_remap(bdev_get_queue(clone->bi_bdev), clone, 
+				    tio->io->bio->bi_bdev->bd_dev, sector, 
+				    clone->bi_sector);
+
 		generic_make_request(clone);
+	}
 
 	else if (r < 0) {
 		/* error the io and bail out */
diff --git a/fs/bio.c b/fs/bio.c
index bbc442b..8a1b0b6 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -25,6 +25,7 @@
 #include <linux/module.h>
 #include <linux/mempool.h>
 #include <linux/workqueue.h>
+#include <linux/blktrace_api.h>
 #include <scsi/sg.h>		/* for struct sg_iovec */
 
 #define BIO_POOL_SIZE 256
@@ -1094,6 +1095,9 @@ struct bio_pair *bio_split(struct bio *b
 	if (!bp)
 		return bp;
 
+	blk_add_trace_pdu_int(bdev_get_queue(bi->bi_bdev), BLK_TA_SPLIT, bi,
+				bi->bi_sector + first_sectors);
+
 	BUG_ON(bi->bi_vcnt != 1);
 	BUG_ON(bi->bi_idx != 0);
 	atomic_set(&bp->cnt, 3);
diff --git a/fs/compat_ioctl.c b/fs/compat_ioctl.c
index 5dd0207..010e02b 100644
--- a/fs/compat_ioctl.c
+++ b/fs/compat_ioctl.c
@@ -72,6 +72,7 @@
 #include <linux/i2c-dev.h>
 #include <linux/wireless.h>
 #include <linux/atalk.h>
+#include <linux/blktrace_api.h>
 
 #include <net/sock.h>          /* siocdevprivate_ioctl */
 #include <net/bluetooth/bluetooth.h>
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 02a585f..195c3b9 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -22,6 +22,7 @@ typedef struct request_queue request_que
 struct elevator_queue;
 typedef struct elevator_queue elevator_t;
 struct request_pm_state;
+struct blk_trace;
 
 #define BLKDEV_MIN_RQ	4
 #define BLKDEV_MAX_RQ	128	/* Default maximum */
@@ -416,6 +417,8 @@ struct request_queue
 	unsigned int		sg_reserved_size;
 	int			node;
 
+	struct blk_trace	*blk_trace;
+
 	/*
 	 * reserved for flush operations
 	 */
diff --git a/include/linux/blktrace_api.h b/include/linux/blktrace_api.h
new file mode 100644
index 0000000..d6b4317
--- /dev/null
+++ b/include/linux/blktrace_api.h
@@ -0,0 +1,216 @@
+#ifndef BLKTRACE_H
+#define BLKTRACE_H
+
+#include <linux/config.h>
+#include <linux/blkdev.h>
+#include <linux/relayfs_fs.h>
+
+/*
+ * Trace categories
+ */
+enum blktrace_cat {
+	BLK_TC_READ	= 1 << 0,	/* reads */
+	BLK_TC_WRITE	= 1 << 1,	/* writes */
+	BLK_TC_BARRIER	= 1 << 2,	/* barrier */
+	BLK_TC_SYNC	= 1 << 3,	/* barrier */
+	BLK_TC_QUEUE	= 1 << 4,	/* queueing/merging */
+	BLK_TC_REQUEUE	= 1 << 5,	/* requeueing */
+	BLK_TC_ISSUE	= 1 << 6,	/* issue */
+	BLK_TC_COMPLETE	= 1 << 7,	/* completions */
+	BLK_TC_FS	= 1 << 8,	/* fs requests */
+	BLK_TC_PC	= 1 << 9,	/* pc requests */
+
+	BLK_TC_END	= 1 << 15,	/* only 16-bits, reminder */
+};
+
+#define BLK_TC_SHIFT		(16)
+#define BLK_TC_ACT(act)		((act) << BLK_TC_SHIFT)
+
+/*
+ * Basic trace actions
+ */
+enum blktrace_act {
+	__BLK_TA_QUEUE = 1,		/* queued */
+	__BLK_TA_BACKMERGE,		/* back merged to existing rq */
+	__BLK_TA_FRONTMERGE,		/* front merge to existing rq */
+	__BLK_TA_GETRQ,			/* allocated new request */
+	__BLK_TA_SLEEPRQ,		/* sleeping on rq allocation */
+	__BLK_TA_REQUEUE,		/* request requeued */
+	__BLK_TA_ISSUE,			/* sent to driver */
+	__BLK_TA_COMPLETE,		/* completed by driver */
+	__BLK_TA_PLUG,			/* queue was plugged */
+	__BLK_TA_UNPLUG_IO,		/* queue was unplugged by io */
+	__BLK_TA_UNPLUG_TIMER,		/* queue was unplugged by timer */
+	__BLK_TA_INSERT,		/* insert request */
+	__BLK_TA_SPLIT,			/* bio was split */
+	__BLK_TA_BOUNCE,		/* bio was bounced */
+	__BLK_TA_REMAP,			/* bio was remapped */
+};
+
+/*
+ * Trace actions in full. Additionally, read or write is masked
+ */
+#define BLK_TA_QUEUE		(__BLK_TA_QUEUE | BLK_TC_ACT(BLK_TC_QUEUE))
+#define BLK_TA_BACKMERGE	(__BLK_TA_BACKMERGE | BLK_TC_ACT(BLK_TC_QUEUE))
+#define BLK_TA_FRONTMERGE	(__BLK_TA_FRONTMERGE | BLK_TC_ACT(BLK_TC_QUEUE))
+#define	BLK_TA_GETRQ		(__BLK_TA_GETRQ | BLK_TC_ACT(BLK_TC_QUEUE))
+#define	BLK_TA_SLEEPRQ		(__BLK_TA_SLEEPRQ | BLK_TC_ACT(BLK_TC_QUEUE))
+#define	BLK_TA_REQUEUE		(__BLK_TA_REQUEUE | BLK_TC_ACT(BLK_TC_REQUEUE))
+#define BLK_TA_ISSUE		(__BLK_TA_ISSUE | BLK_TC_ACT(BLK_TC_ISSUE))
+#define BLK_TA_COMPLETE		(__BLK_TA_COMPLETE| BLK_TC_ACT(BLK_TC_COMPLETE))
+#define BLK_TA_PLUG		(__BLK_TA_PLUG | BLK_TC_ACT(BLK_TC_QUEUE))
+#define BLK_TA_UNPLUG_IO	(__BLK_TA_UNPLUG_IO | BLK_TC_ACT(BLK_TC_QUEUE))
+#define BLK_TA_UNPLUG_TIMER	(__BLK_TA_UNPLUG_TIMER | BLK_TC_ACT(BLK_TC_QUEUE))
+#define BLK_TA_INSERT		(__BLK_TA_INSERT | BLK_TC_ACT(BLK_TC_QUEUE))
+#define BLK_TA_SPLIT		(__BLK_TA_SPLIT)
+#define BLK_TA_BOUNCE		(__BLK_TA_BOUNCE)
+#define BLK_TA_REMAP		(__BLK_TA_REMAP | BLK_TC_ACT(BLK_TC_QUEUE))
+
+#define BLK_IO_TRACE_MAGIC	0x65617400
+#define BLK_IO_TRACE_VERSION	0x05
+
+/*
+ * The trace itself
+ */
+struct blk_io_trace {
+	u32 magic;		/* MAGIC << 8 | version */
+	u32 sequence;		/* event number */
+	u64 time;		/* in microseconds */
+	u64 sector;		/* disk offset */
+	u32 bytes;		/* transfer length */
+	u32 action;		/* what happened */
+	u32 pid;		/* who did it */
+	u32 cpu;		/* on what cpu did it happen */
+	u16 error;		/* completion error */
+	u16 pdu_len;		/* length of data after this trace */
+	u32 device;		/* device number */
+	char comm[16];		/* task command name (TASK_COMM_LEN) */
+};
+
+/*
+ * The remap event
+ */
+struct blk_io_trace_remap {
+	u32 device;
+	u32 __pad;
+	u64 sector;
+};
+
+struct blk_trace {
+	struct dentry *dir;
+	struct rchan *rchan;
+	struct dentry *dropped_file;
+	atomic_t dropped;
+	spinlock_t lock;
+	unsigned long sequence;
+	u32 dev;
+	u16 act_mask;
+	u64 start_lba;
+	u64 end_lba;
+	u32 pid;
+};
+
+/*
+ * User setup structure passed with BLKSTARTTRACE
+ */
+struct blk_user_trace_setup {
+	char name[BDEVNAME_SIZE];	/* output */
+	u16 act_mask;			/* input */
+	u32 buf_size;			/* input */
+	u32 buf_nr;			/* input */
+	u64 start_lba;
+	u64 end_lba;
+	u32 pid;
+};
+
+#if defined(CONFIG_BLK_DEV_IO_TRACE)
+extern int blk_start_trace(struct block_device *, char __user *);
+extern int blk_stop_trace(struct block_device *);
+extern void blk_cleanup_trace(struct blk_trace *);
+extern void __blk_add_trace(struct blk_trace *, sector_t, int, int, u32, int, int, void *);
+
+static inline void blk_add_trace_rq(struct request_queue *q, struct request *rq,
+				    u32 what)
+{
+	struct blk_trace *bt = q->blk_trace;
+	int rw = rq->flags & 0x07;
+
+	if (likely(!bt))
+		return;
+
+	if (blk_pc_request(rq)) {
+		what |= BLK_TC_ACT(BLK_TC_PC);
+		__blk_add_trace(bt, 0, rq->data_len, rw, what, rq->errors, sizeof(rq->cmd), rq->cmd);
+	} else  {
+		what |= BLK_TC_ACT(BLK_TC_FS);
+		__blk_add_trace(bt, rq->hard_sector, rq->hard_nr_sectors << 9, rw, what, rq->errors, 0, NULL);
+	}
+}
+
+static inline void blk_add_trace_bio(struct request_queue *q, struct bio *bio,
+				     u32 what)
+{
+	struct blk_trace *bt = q->blk_trace;
+
+	if (likely(!bt))
+		return;
+
+	__blk_add_trace(bt, bio->bi_sector, bio->bi_size, bio->bi_rw, what, !bio_flagged(bio, BIO_UPTODATE), 0, NULL);
+}
+
+static inline void blk_add_trace_generic(struct request_queue *q,
+					 struct bio *bio, int rw, u32 what)
+{
+	struct blk_trace *bt = q->blk_trace;
+
+	if (likely(!bt))
+		return;
+
+	if (bio)
+		blk_add_trace_bio(q, bio, what);
+	else
+		__blk_add_trace(bt, 0, 0, rw, what, 0, 0, NULL);
+}
+
+static inline void blk_add_trace_pdu_int(struct request_queue *q, u32 what,
+					 struct bio *bio, unsigned int pdu)
+{
+	struct blk_trace *bt = q->blk_trace;
+	u64 rpdu = cpu_to_be64(pdu);
+
+	if (likely(!bt))
+		return;
+
+	if (bio)
+		__blk_add_trace(bt, bio->bi_sector, bio->bi_size, bio->bi_rw, what, !bio_flagged(bio, BIO_UPTODATE), sizeof(rpdu), &rpdu);
+	else
+		__blk_add_trace(bt, 0, 0, 0, what, 0, sizeof(rpdu), &rpdu);
+}
+
+static inline void blk_add_trace_remap(struct request_queue *q, struct bio *bio,
+				       dev_t dev, sector_t from, sector_t to)
+{
+	struct blk_trace *bt = q->blk_trace;
+	struct blk_io_trace_remap r;
+
+	if (likely(!bt))
+		return;
+
+	r.device = cpu_to_be32(dev);
+	r.sector = cpu_to_be64(to);
+
+	__blk_add_trace(bt, from, bio->bi_size, bio->bi_rw, BLK_TA_REMAP, !bio_flagged(bio, BIO_UPTODATE), sizeof(r), &r);
+}
+
+#else /* !CONFIG_BLK_DEV_IO_TRACE */
+#define blk_start_trace(bdev, arg)		(-EINVAL)
+#define blk_stop_trace(bdev)			(-EINVAL)
+#define blk_cleanup_trace(bt)			do { } while (0)
+#define blk_add_trace_rq(q, rq, what)		do { } while (0)
+#define blk_add_trace_bio(q, rq, what)		do { } while (0)
+#define blk_add_trace_generic(q, rq, rw, what)	do { } while (0)
+#define blk_add_trace_pdu_int(q, what, bio, pdu)	do { } while (0)
+#define blk_add_trace_remap(q, bio, dev, f, t)	do {} while (0)
+#endif /* CONFIG_BLK_DEV_IO_TRACE */
+
+#endif
diff --git a/include/linux/compat_ioctl.h b/include/linux/compat_ioctl.h
index 8fad50f..5bed09a 100644
--- a/include/linux/compat_ioctl.h
+++ b/include/linux/compat_ioctl.h
@@ -97,6 +97,8 @@ COMPATIBLE_IOCTL(BLKRRPART)
 COMPATIBLE_IOCTL(BLKFLSBUF)
 COMPATIBLE_IOCTL(BLKSECTSET)
 COMPATIBLE_IOCTL(BLKSSZGET)
+COMPATIBLE_IOCTL(BLKSTARTTRACE)
+COMPATIBLE_IOCTL(BLKSTOPTRACE)
 ULONG_IOCTL(BLKRASET)
 ULONG_IOCTL(BLKFRASET)
 /* RAID */
diff --git a/include/linux/fs.h b/include/linux/fs.h
index b77f260..5575284 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -196,6 +196,8 @@ extern int dir_notify_enable;
 #define BLKBSZGET  _IOR(0x12,112,size_t)
 #define BLKBSZSET  _IOW(0x12,113,size_t)
 #define BLKGETSIZE64 _IOR(0x12,114,size_t)	/* return device size in bytes (u64 *arg) */
+#define BLKSTARTTRACE _IOWR(0x12,115,struct blk_user_trace_setup)
+#define BLKSTOPTRACE _IO(0x12,116)
 
 #define BMAP_IOCTL 1		/* obsolete - kept for compatibility */
 #define FIBMAP	   _IO(0x00,1)	/* bmap access */
diff --git a/mm/highmem.c b/mm/highmem.c
index ce2e7e8..d0ea1ee 100644
--- a/mm/highmem.c
+++ b/mm/highmem.c
@@ -26,6 +26,7 @@
 #include <linux/init.h>
 #include <linux/hash.h>
 #include <linux/highmem.h>
+#include <linux/blktrace_api.h>
 #include <asm/tlbflush.h>
 
 static mempool_t *page_pool, *isa_page_pool;
@@ -483,6 +484,8 @@ void blk_queue_bounce(request_queue_t *q
 		pool = isa_page_pool;
 	}
 
+	blk_add_trace_bio(q, *bio_orig, BLK_TA_BOUNCE);
+
 	/*
 	 * slow path
 	 */