[PATCH] Block queue IO tracing support (blktrace) as of 2006-03-23

Signed-off-by: Jens Axboe <axboe@suse.de>

17 files changed, 916 insertions, 3 deletions

diff --git a/block/Kconfig b/block/Kconfig
index 377f6dd20e17..96783645092d 100644
--- a/block/Kconfig
+++ b/block/Kconfig
@@ -11,4 +11,16 @@ 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 RELAY
+        select DEBUG_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 7e4f93e2b44e..c05de0e0037f 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -8,3 +8,5 @@ obj-$(CONFIG_IOSCHED_NOOP)	+= noop-iosched.o
 obj-$(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 000000000000..36f3a172275f
--- /dev/null
+++ b/block/blktrace.c
@@ -0,0 +1,538 @@
+/*
+ * Copyright (C) 2006 Jens Axboe <axboe@suse.de>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * 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., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ */
+#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 <linux/debugfs.h>
+#include <asm/uaccess.h>
+
+static DEFINE_PER_CPU(unsigned long long, blk_trace_cpu_offset) = { 0, };
+static unsigned int blktrace_seq __read_mostly = 1;
+
+/*
+ * Send out a notify for this process, if we haven't done so since a trace
+ * started
+ */
+static void trace_note_tsk(struct blk_trace *bt, struct task_struct *tsk)
+{
+        struct blk_io_trace *t;
+
+        t = relay_reserve(bt->rchan, sizeof(*t) + sizeof(tsk->comm));
+        if (t) {
+                t->magic = BLK_IO_TRACE_MAGIC | BLK_IO_TRACE_VERSION;
+                t->device = bt->dev;
+                t->action = BLK_TC_ACT(BLK_TC_NOTIFY);
+                t->pid = tsk->pid;
+                t->cpu = smp_processor_id();
+                t->pdu_len = sizeof(tsk->comm);
+                memcpy((void *) t + sizeof(*t), tsk->comm, t->pdu_len);
+                tsk->btrace_seq = blktrace_seq;
+        }
+}
+
+static int act_log_check(struct blk_trace *bt, u32 what, sector_t sector,
+                         pid_t pid)
+{
+        if (((bt->act_mask << BLK_TC_SHIFT) & what) == 0)
+                return 1;
+        if (sector < bt->start_lba || sector > bt->end_lba)
+                return 1;
+        if (bt->pid && pid != bt->pid)
+                return 1;
+
+        return 0;
+}
+
+/*
+ * Data direction bit lookup
+ */
+static u32 ddir_act[2] __read_mostly = { BLK_TC_ACT(BLK_TC_READ), BLK_TC_ACT(BLK_TC_WRITE) };
+
+/*
+ * Bio action bits of interest
+ */
+static u32 bio_act[3] __read_mostly = { 0, BLK_TC_ACT(BLK_TC_BARRIER), BLK_TC_ACT(BLK_TC_SYNC) };
+
+/*
+ * More could be added as needed, taking care to increment the decrementer
+ * to get correct indexing
+ */
+#define trace_barrier_bit(rw)   \
+        (((rw) & (1 << BIO_RW_BARRIER)) >> (BIO_RW_BARRIER - 0))
+#define trace_sync_bit(rw)      \
+        (((rw) & (1 << BIO_RW_SYNC)) >> (BIO_RW_SYNC - 1))
+
+/*
+ * The worker for the various blk_add_trace*() types. Fills out a
+ * blk_io_trace structure and places it in a per-cpu subbuffer.
+ */
+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 task_struct *tsk = current;
+        struct blk_io_trace *t;
+        unsigned long flags;
+        unsigned long *sequence;
+        pid_t pid;
+        int cpu;
+
+        if (unlikely(bt->trace_state != Blktrace_running))
+                return;
+
+        what |= ddir_act[rw & WRITE];
+        what |= bio_act[trace_barrier_bit(rw)];
+        what |= bio_act[trace_sync_bit(rw)];
+
+        pid = tsk->pid;
+        if (unlikely(act_log_check(bt, what, sector, pid)))
+                return;
+
+        /*
+         * A word about the locking here - we disable interrupts to reserve
+         * some space in the relay per-cpu buffer, to prevent an irq
+         * from coming in and stepping on our toes. Once reserved, it's
+         * enough to get preemption disabled to prevent read of this data
+         * before we are through filling it. get_cpu()/put_cpu() does this
+         * for us
+         */
+        local_irq_save(flags);
+
+        if (unlikely(tsk->btrace_seq != blktrace_seq))
+                trace_note_tsk(bt, tsk);
+
+        t = relay_reserve(bt->rchan, sizeof(*t) + pdu_len);
+        if (t) {
+                cpu = smp_processor_id();
+                sequence = per_cpu_ptr(bt->sequence, cpu);
+
+                t->magic = BLK_IO_TRACE_MAGIC | BLK_IO_TRACE_VERSION;
+                t->sequence = ++(*sequence);
+                t->time = sched_clock() - per_cpu(blk_trace_cpu_offset, cpu);
+                t->sector = sector;
+                t->bytes = bytes;
+                t->action = what;
+                t->pid = pid;
+                t->device = bt->dev;
+                t->cpu = cpu;
+                t->error = error;
+                t->pdu_len = pdu_len;
+
+                if (pdu_len)
+                        memcpy((void *) t + sizeof(*t), pdu_data, pdu_len);
+        }
+
+        local_irq_restore(flags);
+}
+
+EXPORT_SYMBOL_GPL(__blk_add_trace);
+
+static struct dentry *blk_tree_root;
+static struct mutex blk_tree_mutex;
+static unsigned int root_users;
+
+static inline void blk_remove_root(void)
+{
+        if (blk_tree_root) {
+                debugfs_remove(blk_tree_root);
+                blk_tree_root = NULL;
+        }
+}
+
+static void blk_remove_tree(struct dentry *dir)
+{
+        mutex_lock(&blk_tree_mutex);
+        debugfs_remove(dir);
+        if (--root_users == 0)
+                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 = debugfs_create_dir("block", NULL);
+                if (!blk_tree_root)
+                        goto err;
+        }
+
+        dir = debugfs_create_dir(blk_name, blk_tree_root);
+        if (dir)
+                root_users++;
+        else
+                blk_remove_root();
+
+err:
+        mutex_unlock(&blk_tree_mutex);
+        return dir;
+}
+
+static void blk_trace_cleanup(struct blk_trace *bt)
+{
+        relay_close(bt->rchan);
+        debugfs_remove(bt->dropped_file);
+        blk_remove_tree(bt->dir);
+        free_percpu(bt->sequence);
+        kfree(bt);
+}
+
+static int blk_trace_remove(request_queue_t *q)
+{
+        struct blk_trace *bt;
+
+        bt = xchg(&q->blk_trace, NULL);
+        if (!bt)
+                return -EINVAL;
+
+        if (bt->trace_state == Blktrace_setup ||
+            bt->trace_state == Blktrace_stopped)
+                blk_trace_cleanup(bt);
+
+        return 0;
+}
+
+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,
+};
+
+/*
+ * Keep track of how many times we encountered a full subbuffer, to aid
+ * the user space app in telling how many lost events there were.
+ */
+static int blk_subbuf_start_callback(struct rchan_buf *buf, void *subbuf,
+                                     void *prev_subbuf, size_t prev_padding)
+{
+        struct blk_trace *bt;
+
+        if (!relay_buf_full(buf))
+                return 1;
+
+        bt = buf->chan->private_data;
+        atomic_inc(&bt->dropped);
+        return 0;
+}
+
+static int blk_remove_buf_file_callback(struct dentry *dentry)
+{
+        debugfs_remove(dentry);
+        return 0;
+}
+
+static struct dentry *blk_create_buf_file_callback(const char *filename,
+                                                   struct dentry *parent,
+                                                   int mode,
+                                                   struct rchan_buf *buf,
+                                                   int *is_global)
+{
+        return debugfs_create_file(filename, mode, parent, buf,
+                                        &relay_file_operations);
+}
+
+static struct rchan_callbacks blk_relay_callbacks = {
+        .subbuf_start           = blk_subbuf_start_callback,
+        .create_buf_file        = blk_create_buf_file_callback,
+        .remove_buf_file        = blk_remove_buf_file_callback,
+};
+
+/*
+ * Setup everything required to start tracing
+ */
+static int blk_trace_setup(request_queue_t *q, struct block_device *bdev,
+                           char __user *arg)
+{
+        struct blk_user_trace_setup buts;
+        struct blk_trace *old_bt, *bt = NULL;
+        struct dentry *dir = NULL;
+        char b[BDEVNAME_SIZE];
+        int ret, i;
+
+        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;
+
+        ret = -ENOMEM;
+        bt = kzalloc(sizeof(*bt), GFP_KERNEL);
+        if (!bt)
+                goto err;
+
+        bt->sequence = alloc_percpu(unsigned long);
+        if (!bt->sequence)
+                goto err;
+
+        ret = -ENOENT;
+        dir = blk_create_tree(buts.name);
+        if (!dir)
+                goto err;
+
+        bt->dir = dir;
+        bt->dev = bdev->bd_dev;
+        atomic_set(&bt->dropped, 0);
+
+        ret = -EIO;
+        bt->dropped_file = debugfs_create_file("dropped", 0444, dir, bt, &blk_dropped_fops);
+        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;
+        bt->trace_state = Blktrace_setup;
+
+        ret = -EBUSY;
+        old_bt = xchg(&q->blk_trace, bt);
+        if (old_bt) {
+                (void) xchg(&q->blk_trace, old_bt);
+                goto err;
+        }
+
+        return 0;
+err:
+        if (dir)
+                blk_remove_tree(dir);
+        if (bt) {
+                if (bt->dropped_file)
+                        debugfs_remove(bt->dropped_file);
+                if (bt->sequence)
+                        free_percpu(bt->sequence);
+                if (bt->rchan)
+                        relay_close(bt->rchan);
+                kfree(bt);
+        }
+        return ret;
+}
+
+static int blk_trace_startstop(request_queue_t *q, int start)
+{
+        struct blk_trace *bt;
+        int ret;
+
+        if ((bt = q->blk_trace) == NULL)
+                return -EINVAL;
+
+        /*
+         * For starting a trace, we can transition from a setup or stopped
+         * trace. For stopping a trace, the state must be running
+         */
+        ret = -EINVAL;
+        if (start) {
+                if (bt->trace_state == Blktrace_setup ||
+                    bt->trace_state == Blktrace_stopped) {
+                        blktrace_seq++;
+                        smp_mb();
+                        bt->trace_state = Blktrace_running;
+                        ret = 0;
+                }
+        } else {
+                if (bt->trace_state == Blktrace_running) {
+                        bt->trace_state = Blktrace_stopped;
+                        relay_flush(bt->rchan);
+                        ret = 0;
+                }
+        }
+
+        return ret;
+}
+
+/**
+ * blk_trace_ioctl: - handle the ioctls associated with tracing
+ * @bdev:       the block device
+ * @cmd:        the ioctl cmd
+ * @arg:        the argument data, if any
+ *
+ **/
+int blk_trace_ioctl(struct block_device *bdev, unsigned cmd, char __user *arg)
+{
+        request_queue_t *q;
+        int ret, start = 0;
+
+        q = bdev_get_queue(bdev);
+        if (!q)
+                return -ENXIO;
+
+        mutex_lock(&bdev->bd_mutex);
+
+        switch (cmd) {
+        case BLKTRACESETUP:
+                ret = blk_trace_setup(q, bdev, arg);
+                break;
+        case BLKTRACESTART:
+                start = 1;
+        case BLKTRACESTOP:
+                ret = blk_trace_startstop(q, start);
+                break;
+        case BLKTRACETEARDOWN:
+                ret = blk_trace_remove(q);
+                break;
+        default:
+                ret = -ENOTTY;
+                break;
+        }
+
+        mutex_unlock(&bdev->bd_mutex);
+        return ret;
+}
+
+/**
+ * blk_trace_shutdown: - stop and cleanup trace structures
+ * @q:    the request queue associated with the device
+ *
+ **/
+void blk_trace_shutdown(request_queue_t *q)
+{
+        blk_trace_startstop(q, 0);
+        blk_trace_remove(q);
+}
+
+/*
+ * 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 db3d0d8296a0..5e558c4689a4 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_insert(request_queue_t *q, struct request *rq, int where)
         struct list_head *pos;
         unsigned ordseq;
 
+        blk_add_trace_rq(q, rq, BLK_TA_INSERT);
+
         rq->q = q;
 
         switch (where) {
@@ -499,6 +502,7 @@ struct request *elv_next_request(request_queue_t *q)
                          * 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 35fdb7dc6512..9cfa2e1ecb24 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 +190,11 @@ static int blkdev_locked_ioctl(struct file *file, struct block_device *bdev,
                 return put_ulong(arg, bdev->bd_inode->i_size >> 9);
         case BLKGETSIZE64:
                 return put_u64(arg, bdev->bd_inode->i_size);
+        case BLKTRACESTART:
+        case BLKTRACESTOP:
+        case BLKTRACESETUP:
+        case BLKTRACETEARDOWN:
+                return blk_trace_ioctl(bdev, cmd, (char __user *) arg);
         }
         return -ENOIOCTLCMD;
 }
diff --git a/block/ll_rw_blk.c b/block/ll_rw_blk.c
index 6c793b196aa9..062067fa7ead 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
@@ -1556,8 +1557,10 @@ 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);
@@ -1621,14 +1624,21 @@ static void blk_backing_dev_unplug(struct backing_dev_info *bdi,
         /*
          * 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);
 }
 
@@ -1636,6 +1646,9 @@ static void blk_unplug_timeout(unsigned long data)
 {
         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);
 }
 
@@ -1753,6 +1766,9 @@ static void blk_release_queue(struct kobject *kobj)
         if (q->queue_tags)
                 __blk_queue_free_tags(q);
 
+        if (q->blk_trace)
+                blk_trace_shutdown(q);
+
         kmem_cache_free(requestq_cachep, q);
 }
 
@@ -2129,6 +2145,8 @@ rq_starved:
         
         rq_init(q, rq);
         rq->rl = rl;
+
+        blk_add_trace_generic(q, bio, rw, BLK_TA_GETRQ);
 out:
         return rq;
 }
@@ -2157,6 +2175,8 @@ static struct request *get_request_wait(request_queue_t *q, int rw,
                 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();
@@ -2210,6 +2230,8 @@ 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);
 
@@ -2844,6 +2866,8 @@ static int __make_request(request_queue_t *q, struct bio *bio)
                         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;
@@ -2859,6 +2883,8 @@ static int __make_request(request_queue_t *q, struct bio *bio)
                         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;
 
@@ -2976,6 +3002,7 @@ void generic_make_request(struct bio *bio)
         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. */
@@ -3002,6 +3029,8 @@ void generic_make_request(struct bio *bio)
          * 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];
 
@@ -3034,6 +3063,15 @@ 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);
 }
@@ -3153,6 +3191,8 @@ static int __end_that_request_first(struct request *req, int uptodate,
         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 e29b8926f80e..1f2890989b56 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>
 
@@ -2331,6 +2332,7 @@ static inline void complete_command( ctlr_info_t *h, CommandList_struct *cmd,
 
         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 26b08ee425c7..8c82373f7ff3 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;
 
@@ -334,6 +335,8 @@ static void dec_pending(struct dm_io *io, int error)
                         /* 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);
         }
@@ -392,6 +395,7 @@ static void __map_bio(struct dm_target *ti, struct bio *clone,
                       struct target_io *tio)
 {
         int r;
+        sector_t sector;
 
         /*
          * Sanity checks.
@@ -407,10 +411,17 @@ static void __map_bio(struct dm_target *ti, struct bio *clone,
          * 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 8f1d2e815c96..0a8c59cb68f5 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
@@ -1095,6 +1096,9 @@ struct bio_pair *bio_split(struct bio *bi, mempool_t *pool, int first_sectors)
         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 c666769a875d..7c031f00fd79 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 56bb6a4e15f3..c179966f1a2f 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -22,6 +22,7 @@ typedef struct request_queue request_queue_t;
 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 000000000000..b34d3e73d5ea
--- /dev/null
+++ b/include/linux/blktrace_api.h
@@ -0,0 +1,277 @@
+#ifndef BLKTRACE_H
+#define BLKTRACE_H
+
+#include <linux/config.h>
+#include <linux/blkdev.h>
+#include <linux/relay.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_NOTIFY   = 1 << 10,      /* special message */
+
+        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    0x07
+
+/*
+ * 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 device;             /* device number */
+        u32 cpu;                /* on what cpu did it happen */
+        u16 error;              /* completion error */
+        u16 pdu_len;            /* length of data after this trace */
+};
+
+/*
+ * The remap event
+ */
+struct blk_io_trace_remap {
+        u32 device;
+        u32 __pad;
+        u64 sector;
+};
+
+enum {
+        Blktrace_setup = 1,
+        Blktrace_running,
+        Blktrace_stopped,
+};
+
+struct blk_trace {
+        int trace_state;
+        struct rchan *rchan;
+        unsigned long *sequence;
+        u16 act_mask;
+        u64 start_lba;
+        u64 end_lba;
+        u32 pid;
+        u32 dev;
+        struct dentry *dir;
+        struct dentry *dropped_file;
+        atomic_t dropped;
+};
+
+/*
+ * User setup structure passed with BLKTRACESTART
+ */
+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_trace_ioctl(struct block_device *, unsigned, char __user *);
+extern void blk_trace_shutdown(request_queue_t *);
+extern void __blk_add_trace(struct blk_trace *, sector_t, int, int, u32, int, int, void *);
+
+/**
+ * blk_add_trace_rq - Add a trace for a request oriented action
+ * @q:          queue the io is for
+ * @rq:         the source request
+ * @what:       the action
+ *
+ * Description:
+ *     Records an action against a request. Will log the bio offset + size.
+ *
+ **/
+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);
+        }
+}
+
+/**
+ * blk_add_trace_bio - Add a trace for a bio oriented action
+ * @q:          queue the io is for
+ * @bio:        the source bio
+ * @what:       the action
+ *
+ * Description:
+ *     Records an action against a bio. Will log the bio offset + size.
+ *
+ **/
+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);
+}
+
+/**
+ * blk_add_trace_generic - Add a trace for a generic action
+ * @q:          queue the io is for
+ * @bio:        the source bio
+ * @rw:         the data direction
+ * @what:       the action
+ *
+ * Description:
+ *     Records a simple trace
+ *
+ **/
+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);
+}
+
+/**
+ * blk_add_trace_pdu_int - Add a trace for a bio with an integer payload
+ * @q:          queue the io is for
+ * @what:       the action
+ * @bio:        the source bio
+ * @pdu:        the integer payload
+ *
+ * Description:
+ *     Adds a trace with some integer payload. This might be an unplug
+ *     option given as the action, with the depth at unplug time given
+ *     as the payload
+ *
+ **/
+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);
+}
+
+/**
+ * blk_add_trace_remap - Add a trace for a remap operation
+ * @q:          queue the io is for
+ * @bio:        the source bio
+ * @dev:        target device
+ * @from:       source sector
+ * @to:         target sector
+ *
+ * Description:
+ *     Device mapper or raid target sometimes need to split a bio because
+ *     it spans a stripe (or similar). Add a trace for that action.
+ *
+ **/
+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_trace_ioctl(bdev, cmd, arg)         (-ENOTTY)
+#define blk_trace_shutdown(q)                   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 ae7dfb790df3..efb518f16bb3 100644
--- a/include/linux/compat_ioctl.h
+++ b/include/linux/compat_ioctl.h
@@ -97,6 +97,10 @@ COMPATIBLE_IOCTL(BLKRRPART)
 COMPATIBLE_IOCTL(BLKFLSBUF)
 COMPATIBLE_IOCTL(BLKSECTSET)
 COMPATIBLE_IOCTL(BLKSSZGET)
+COMPATIBLE_IOCTL(BLKTRACESTART)
+COMPATIBLE_IOCTL(BLKTRACESTOP)
+COMPATIBLE_IOCTL(BLKTRACESETUP)
+COMPATIBLE_IOCTL(BLKTRACETEARDOWN)
 ULONG_IOCTL(BLKRASET)
 ULONG_IOCTL(BLKFRASET)
 /* RAID */
diff --git a/include/linux/fs.h b/include/linux/fs.h
index f9c9dea636d0..9b34a1b03455 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -197,6 +197,10 @@ 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 BLKTRACESETUP _IOWR(0x12,115,struct blk_user_trace_setup)
+#define BLKTRACESTART _IO(0x12,116)
+#define BLKTRACESTOP _IO(0x12,117)
+#define BLKTRACETEARDOWN _IO(0x12,118)
 
 #define BMAP_IOCTL 1            /* obsolete - kept for compatibility */
 #define FIBMAP     _IO(0x00,1)  /* bmap access */
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 62e6314382f0..e60a91d5b369 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -706,6 +706,7 @@ struct task_struct {
         prio_array_t *array;
 
         unsigned short ioprio;
+        unsigned int btrace_seq;
 
         unsigned long sleep_avg;
         unsigned long long timestamp, last_ran;
diff --git a/kernel/fork.c b/kernel/fork.c
index c79ae0b19a49..c21bae8c93b9 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -181,6 +181,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig)
         /* One for us, one for whoever does the "release_task()" (usually parent) */
         atomic_set(&tsk->usage,2);
         atomic_set(&tsk->fs_excl, 0);
+        tsk->btrace_seq = 0;
         return tsk;
 }
 
diff --git a/mm/highmem.c b/mm/highmem.c
index ce2e7e8bbfa7..d0ea1eec6a9a 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, struct bio **bio_orig)
                 pool = isa_page_pool;
         }
 
+        blk_add_trace_bio(q, *bio_orig, BLK_TA_BOUNCE);
+
         /*
          * slow path
          */