Driver core: convert block from raw kobjects to core devices
[linux-2.6.git] / block / ll_rw_blk.c
index 8e27d0a..3887b2a 100644 (file)
@@ -10,7 +10,6 @@
 /*
  * This handles all read/write requests to block devices
  */
-#include <linux/config.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/backing-dev.h>
 #include <linux/slab.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
+#include <linux/task_io_accounting_ops.h>
 #include <linux/interrupt.h>
 #include <linux/cpu.h>
+#include <linux/blktrace_api.h>
+#include <linux/fault-inject.h>
+#include <linux/scatterlist.h>
 
 /*
  * for max sense size
  */
 #include <scsi/scsi_cmnd.h>
 
-static void blk_unplug_work(void *data);
+static void blk_unplug_work(struct work_struct *work);
 static void blk_unplug_timeout(unsigned long data);
-static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io);
+static void drive_stat_acct(struct request *rq, int new_io);
 static void init_request_from_bio(struct request *req, struct bio *bio);
-static int __make_request(request_queue_t *q, struct bio *bio);
+static int __make_request(struct request_queue *q, struct bio *bio);
+static struct io_context *current_io_context(gfp_t gfp_flags, int node);
+static void blk_recalc_rq_segments(struct request *rq);
+static void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
+                           struct bio *bio);
 
 /*
  * For the allocated request tables
  */
-static kmem_cache_t *request_cachep;
+static struct kmem_cache *request_cachep;
 
 /*
  * For queue allocation
  */
-static kmem_cache_t *requestq_cachep;
+static struct kmem_cache *requestq_cachep;
 
 /*
  * For io context allocations
  */
-static kmem_cache_t *iocontext_cachep;
-
-static wait_queue_head_t congestion_wqh[2] = {
-               __WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[0]),
-               __WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[1])
-       };
+static struct kmem_cache *iocontext_cachep;
 
 /*
  * Controlling structure to kblockd
@@ -111,35 +113,6 @@ static void blk_queue_congestion_threshold(struct request_queue *q)
        q->nr_congestion_off = nr;
 }
 
-/*
- * A queue has just exitted congestion.  Note this in the global counter of
- * congested queues, and wake up anyone who was waiting for requests to be
- * put back.
- */
-static void clear_queue_congested(request_queue_t *q, int rw)
-{
-       enum bdi_state bit;
-       wait_queue_head_t *wqh = &congestion_wqh[rw];
-
-       bit = (rw == WRITE) ? BDI_write_congested : BDI_read_congested;
-       clear_bit(bit, &q->backing_dev_info.state);
-       smp_mb__after_clear_bit();
-       if (waitqueue_active(wqh))
-               wake_up(wqh);
-}
-
-/*
- * A queue has just entered congestion.  Flag that in the queue's VM-visible
- * state flags and increment the global gounter of congested queues.
- */
-static void set_queue_congested(request_queue_t *q, int rw)
-{
-       enum bdi_state bit;
-
-       bit = (rw == WRITE) ? BDI_write_congested : BDI_read_congested;
-       set_bit(bit, &q->backing_dev_info.state);
-}
-
 /**
  * blk_get_backing_dev_info - get the address of a queue's backing_dev_info
  * @bdev:      device
@@ -152,23 +125,14 @@ static void set_queue_congested(request_queue_t *q, int rw)
 struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev)
 {
        struct backing_dev_info *ret = NULL;
-       request_queue_t *q = bdev_get_queue(bdev);
+       struct request_queue *q = bdev_get_queue(bdev);
 
        if (q)
                ret = &q->backing_dev_info;
        return ret;
 }
-
 EXPORT_SYMBOL(blk_get_backing_dev_info);
 
-void blk_queue_activity_fn(request_queue_t *q, activity_fn *fn, void *data)
-{
-       q->activity_fn = fn;
-       q->activity_data = data;
-}
-
-EXPORT_SYMBOL(blk_queue_activity_fn);
-
 /**
  * blk_queue_prep_rq - set a prepare_request function for queue
  * @q:         queue
@@ -180,7 +144,7 @@ EXPORT_SYMBOL(blk_queue_activity_fn);
  * cdb from the request data for instance.
  *
  */
-void blk_queue_prep_rq(request_queue_t *q, prep_rq_fn *pfn)
+void blk_queue_prep_rq(struct request_queue *q, prep_rq_fn *pfn)
 {
        q->prep_rq_fn = pfn;
 }
@@ -203,14 +167,14 @@ EXPORT_SYMBOL(blk_queue_prep_rq);
  * no merge_bvec_fn is defined for a queue, and only the fixed limits are
  * honored.
  */
-void blk_queue_merge_bvec(request_queue_t *q, merge_bvec_fn *mbfn)
+void blk_queue_merge_bvec(struct request_queue *q, merge_bvec_fn *mbfn)
 {
        q->merge_bvec_fn = mbfn;
 }
 
 EXPORT_SYMBOL(blk_queue_merge_bvec);
 
-void blk_queue_softirq_done(request_queue_t *q, softirq_done_fn *fn)
+void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn)
 {
        q->softirq_done_fn = fn;
 }
@@ -239,7 +203,7 @@ EXPORT_SYMBOL(blk_queue_softirq_done);
  *    __bio_kmap_atomic() to get a temporary kernel mapping, or by calling
  *    blk_queue_bounce() to create a buffer in normal memory.
  **/
-void blk_queue_make_request(request_queue_t * q, make_request_fn * mfn)
+void blk_queue_make_request(struct request_queue * q, make_request_fn * mfn)
 {
        /*
         * set defaults
@@ -262,7 +226,7 @@ void blk_queue_make_request(request_queue_t * q, make_request_fn * mfn)
        if (q->unplug_delay == 0)
                q->unplug_delay = 1;
 
-       INIT_WORK(&q->unplug_work, blk_unplug_work, q);
+       INIT_WORK(&q->unplug_work, blk_unplug_work);
 
        q->unplug_timer.function = blk_unplug_timeout;
        q->unplug_timer.data = (unsigned long)q;
@@ -271,25 +235,23 @@ void blk_queue_make_request(request_queue_t * q, make_request_fn * mfn)
         * by default assume old behaviour and bounce for any highmem page
         */
        blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH);
-
-       blk_queue_activity_fn(q, NULL, NULL);
 }
 
 EXPORT_SYMBOL(blk_queue_make_request);
 
-static inline void rq_init(request_queue_t *q, struct request *rq)
+static void rq_init(struct request_queue *q, struct request *rq)
 {
        INIT_LIST_HEAD(&rq->queuelist);
        INIT_LIST_HEAD(&rq->donelist);
 
        rq->errors = 0;
-       rq->rq_status = RQ_ACTIVE;
        rq->bio = rq->biotail = NULL;
+       INIT_HLIST_NODE(&rq->hash);
+       RB_CLEAR_NODE(&rq->rb_node);
        rq->ioprio = 0;
        rq->buffer = NULL;
        rq->ref_count = 1;
        rq->q = q;
-       rq->waiting = NULL;
        rq->special = NULL;
        rq->data_len = 0;
        rq->data = NULL;
@@ -298,12 +260,14 @@ static inline void rq_init(request_queue_t *q, struct request *rq)
        rq->end_io = NULL;
        rq->end_io_data = NULL;
        rq->completion_data = NULL;
+       rq->next_rq = NULL;
 }
 
 /**
  * blk_queue_ordered - does this queue support ordered writes
  * @q:        the request queue
  * @ordered:  one of QUEUE_ORDERED_*
+ * @prepare_flush_fn: rq setup helper for cache flush ordered writes
  *
  * Description:
  *   For journalled file systems, doing ordered writes on a commit
@@ -312,7 +276,7 @@ static inline void rq_init(request_queue_t *q, struct request *rq)
  *   feature should call this function and indicate so.
  *
  **/
-int blk_queue_ordered(request_queue_t *q, unsigned ordered,
+int blk_queue_ordered(struct request_queue *q, unsigned ordered,
                      prepare_flush_fn *prepare_flush_fn)
 {
        if (ordered & (QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH) &&
@@ -332,6 +296,7 @@ int blk_queue_ordered(request_queue_t *q, unsigned ordered,
                return -EINVAL;
        }
 
+       q->ordered = ordered;
        q->next_ordered = ordered;
        q->prepare_flush_fn = prepare_flush_fn;
 
@@ -340,27 +305,10 @@ int blk_queue_ordered(request_queue_t *q, unsigned ordered,
 
 EXPORT_SYMBOL(blk_queue_ordered);
 
-/**
- * blk_queue_issue_flush_fn - set function for issuing a flush
- * @q:     the request queue
- * @iff:   the function to be called issuing the flush
- *
- * Description:
- *   If a driver supports issuing a flush command, the support is notified
- *   to the block layer by defining it through this call.
- *
- **/
-void blk_queue_issue_flush_fn(request_queue_t *q, issue_flush_fn *iff)
-{
-       q->issue_flush_fn = iff;
-}
-
-EXPORT_SYMBOL(blk_queue_issue_flush_fn);
-
 /*
  * Cache flushing for ordered writes handling
  */
-inline unsigned blk_ordered_cur_seq(request_queue_t *q)
+inline unsigned blk_ordered_cur_seq(struct request_queue *q)
 {
        if (!q->ordseq)
                return 0;
@@ -369,7 +317,7 @@ inline unsigned blk_ordered_cur_seq(request_queue_t *q)
 
 unsigned blk_ordered_req_seq(struct request *rq)
 {
-       request_queue_t *q = rq->q;
+       struct request_queue *q = rq->q;
 
        BUG_ON(q->ordseq == 0);
 
@@ -380,14 +328,23 @@ unsigned blk_ordered_req_seq(struct request *rq)
        if (rq == &q->post_flush_rq)
                return QUEUE_ORDSEQ_POSTFLUSH;
 
-       if ((rq->flags & REQ_ORDERED_COLOR) ==
-           (q->orig_bar_rq->flags & REQ_ORDERED_COLOR))
+       /*
+        * !fs requests don't need to follow barrier ordering.  Always
+        * put them at the front.  This fixes the following deadlock.
+        *
+        * http://thread.gmane.org/gmane.linux.kernel/537473
+        */
+       if (!blk_fs_request(rq))
+               return QUEUE_ORDSEQ_DRAIN;
+
+       if ((rq->cmd_flags & REQ_ORDERED_COLOR) ==
+           (q->orig_bar_rq->cmd_flags & REQ_ORDERED_COLOR))
                return QUEUE_ORDSEQ_DRAIN;
        else
                return QUEUE_ORDSEQ_DONE;
 }
 
-void blk_ordered_complete_seq(request_queue_t *q, unsigned seq, int error)
+void blk_ordered_complete_seq(struct request_queue *q, unsigned seq, int error)
 {
        struct request *rq;
        int uptodate;
@@ -404,10 +361,12 @@ void blk_ordered_complete_seq(request_queue_t *q, unsigned seq, int error)
        /*
         * Okay, sequence complete.
         */
-       rq = q->orig_bar_rq;
-       uptodate = q->orderr ? q->orderr : 1;
+       uptodate = 1;
+       if (q->orderr)
+               uptodate = q->orderr;
 
        q->ordseq = 0;
+       rq = q->orig_bar_rq;
 
        end_that_request_first(rq, uptodate, rq->hard_nr_sectors);
        end_that_request_last(rq, uptodate);
@@ -431,7 +390,7 @@ static void post_flush_end_io(struct request *rq, int error)
        blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_POSTFLUSH, error);
 }
 
-static void queue_flush(request_queue_t *q, unsigned which)
+static void queue_flush(struct request_queue *q, unsigned which)
 {
        struct request *rq;
        rq_end_io_fn *end_io;
@@ -444,21 +403,20 @@ static void queue_flush(request_queue_t *q, unsigned which)
                end_io = post_flush_end_io;
        }
 
+       rq->cmd_flags = REQ_HARDBARRIER;
        rq_init(q, rq);
-       rq->flags = REQ_HARDBARRIER;
        rq->elevator_private = NULL;
+       rq->elevator_private2 = NULL;
        rq->rq_disk = q->bar_rq.rq_disk;
-       rq->rl = NULL;
        rq->end_io = end_io;
        q->prepare_flush_fn(q, rq);
 
-       __elv_add_request(q, rq, ELEVATOR_INSERT_FRONT, 0);
+       elv_insert(q, rq, ELEVATOR_INSERT_FRONT);
 }
 
-static inline struct request *start_ordered(request_queue_t *q,
+static inline struct request *start_ordered(struct request_queue *q,
                                            struct request *rq)
 {
-       q->bi_size = 0;
        q->orderr = 0;
        q->ordered = q->next_ordered;
        q->ordseq |= QUEUE_ORDSEQ_STARTED;
@@ -469,11 +427,14 @@ static inline struct request *start_ordered(request_queue_t *q,
        blkdev_dequeue_request(rq);
        q->orig_bar_rq = rq;
        rq = &q->bar_rq;
+       rq->cmd_flags = 0;
        rq_init(q, rq);
-       rq->flags = bio_data_dir(q->orig_bar_rq->bio);
-       rq->flags |= q->ordered & QUEUE_ORDERED_FUA ? REQ_FUA : 0;
+       if (bio_data_dir(q->orig_bar_rq->bio) == WRITE)
+               rq->cmd_flags |= REQ_RW;
+       if (q->ordered & QUEUE_ORDERED_FUA)
+               rq->cmd_flags |= REQ_FUA;
        rq->elevator_private = NULL;
-       rq->rl = NULL;
+       rq->elevator_private2 = NULL;
        init_request_from_bio(rq, q->orig_bar_rq->bio);
        rq->end_io = bar_end_io;
 
@@ -481,14 +442,17 @@ static inline struct request *start_ordered(request_queue_t *q,
         * Queue ordered sequence.  As we stack them at the head, we
         * need to queue in reverse order.  Note that we rely on that
         * no fs request uses ELEVATOR_INSERT_FRONT and thus no fs
-        * request gets inbetween ordered sequence.
+        * request gets inbetween ordered sequence. If this request is
+        * an empty barrier, we don't need to do a postflush ever since
+        * there will be no data written between the pre and post flush.
+        * Hence a single flush will suffice.
         */
-       if (q->ordered & QUEUE_ORDERED_POSTFLUSH)
+       if ((q->ordered & QUEUE_ORDERED_POSTFLUSH) && !blk_empty_barrier(rq))
                queue_flush(q, QUEUE_ORDERED_POSTFLUSH);
        else
                q->ordseq |= QUEUE_ORDSEQ_POSTFLUSH;
 
-       __elv_add_request(q, rq, ELEVATOR_INSERT_FRONT, 0);
+       elv_insert(q, rq, ELEVATOR_INSERT_FRONT);
 
        if (q->ordered & QUEUE_ORDERED_PREFLUSH) {
                queue_flush(q, QUEUE_ORDERED_PREFLUSH);
@@ -504,10 +468,10 @@ static inline struct request *start_ordered(request_queue_t *q,
        return rq;
 }
 
-int blk_do_ordered(request_queue_t *q, struct request **rqp)
+int blk_do_ordered(struct request_queue *q, struct request **rqp)
 {
-       struct request *rq = *rqp, *allowed_rq;
-       int is_barrier = blk_fs_request(rq) && blk_barrier_rq(rq);
+       struct request *rq = *rqp;
+       const int is_barrier = blk_fs_request(rq) && blk_barrier_rq(rq);
 
        if (!q->ordseq) {
                if (!is_barrier)
@@ -530,94 +494,59 @@ int blk_do_ordered(request_queue_t *q, struct request **rqp)
                }
        }
 
-       if (q->ordered & QUEUE_ORDERED_TAG) {
-               if (is_barrier && rq != &q->bar_rq)
-                       *rqp = NULL;
-               return 1;
-       }
-
-       switch (blk_ordered_cur_seq(q)) {
-       case QUEUE_ORDSEQ_PREFLUSH:
-               allowed_rq = &q->pre_flush_rq;
-               break;
-       case QUEUE_ORDSEQ_BAR:
-               allowed_rq = &q->bar_rq;
-               break;
-       case QUEUE_ORDSEQ_POSTFLUSH:
-               allowed_rq = &q->post_flush_rq;
-               break;
-       default:
-               allowed_rq = NULL;
-               break;
-       }
-
-       if (rq != allowed_rq &&
-           (blk_fs_request(rq) || rq == &q->pre_flush_rq ||
-            rq == &q->post_flush_rq))
-               *rqp = NULL;
-
-       return 1;
-}
-
-static int flush_dry_bio_endio(struct bio *bio, unsigned int bytes, int error)
-{
-       request_queue_t *q = bio->bi_private;
-       struct bio_vec *bvec;
-       int i;
-
        /*
-        * This is dry run, restore bio_sector and size.  We'll finish
-        * this request again with the original bi_end_io after an
-        * error occurs or post flush is complete.
+        * Ordered sequence in progress
         */
-       q->bi_size += bytes;
 
-       if (bio->bi_size)
+       /* Special requests are not subject to ordering rules. */
+       if (!blk_fs_request(rq) &&
+           rq != &q->pre_flush_rq && rq != &q->post_flush_rq)
                return 1;
 
-       /* Rewind bvec's */
-       bio->bi_idx = 0;
-       bio_for_each_segment(bvec, bio, i) {
-               bvec->bv_len += bvec->bv_offset;
-               bvec->bv_offset = 0;
+       if (q->ordered & QUEUE_ORDERED_TAG) {
+               /* Ordered by tag.  Blocking the next barrier is enough. */
+               if (is_barrier && rq != &q->bar_rq)
+                       *rqp = NULL;
+       } else {
+               /* Ordered by draining.  Wait for turn. */
+               WARN_ON(blk_ordered_req_seq(rq) < blk_ordered_cur_seq(q));
+               if (blk_ordered_req_seq(rq) > blk_ordered_cur_seq(q))
+                       *rqp = NULL;
        }
 
-       /* Reset bio */
-       set_bit(BIO_UPTODATE, &bio->bi_flags);
-       bio->bi_size = q->bi_size;
-       bio->bi_sector -= (q->bi_size >> 9);
-       q->bi_size = 0;
-
-       return 0;
+       return 1;
 }
 
-static inline int ordered_bio_endio(struct request *rq, struct bio *bio,
-                                   unsigned int nbytes, int error)
+static void req_bio_endio(struct request *rq, struct bio *bio,
+                         unsigned int nbytes, int error)
 {
-       request_queue_t *q = rq->q;
-       bio_end_io_t *endio;
-       void *private;
-
-       if (&q->bar_rq != rq)
-               return 0;
-
-       /*
-        * Okay, this is the barrier request in progress, dry finish it.
-        */
-       if (error && !q->orderr)
-               q->orderr = error;
+       struct request_queue *q = rq->q;
 
-       endio = bio->bi_end_io;
-       private = bio->bi_private;
-       bio->bi_end_io = flush_dry_bio_endio;
-       bio->bi_private = q;
+       if (&q->bar_rq != rq) {
+               if (error)
+                       clear_bit(BIO_UPTODATE, &bio->bi_flags);
+               else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
+                       error = -EIO;
 
-       bio_endio(bio, nbytes, error);
+               if (unlikely(nbytes > bio->bi_size)) {
+                       printk("%s: want %u bytes done, only %u left\n",
+                              __FUNCTION__, nbytes, bio->bi_size);
+                       nbytes = bio->bi_size;
+               }
 
-       bio->bi_end_io = endio;
-       bio->bi_private = private;
+               bio->bi_size -= nbytes;
+               bio->bi_sector += (nbytes >> 9);
+               if (bio->bi_size == 0)
+                       bio_endio(bio, error);
+       } else {
 
-       return 1;
+               /*
+                * Okay, this is the barrier request in progress, just
+                * record the error;
+                */
+               if (error && !q->orderr)
+                       q->orderr = error;
+       }
 }
 
 /**
@@ -629,26 +558,31 @@ static inline int ordered_bio_endio(struct request *rq, struct bio *bio,
  *    Different hardware can have different requirements as to what pages
  *    it can do I/O directly to. A low level driver can call
  *    blk_queue_bounce_limit to have lower memory pages allocated as bounce
- *    buffers for doing I/O to pages residing above @page. By default
- *    the block layer sets this to the highest numbered "low" memory page.
+ *    buffers for doing I/O to pages residing above @page.
  **/
-void blk_queue_bounce_limit(request_queue_t *q, u64 dma_addr)
+void blk_queue_bounce_limit(struct request_queue *q, u64 dma_addr)
 {
        unsigned long bounce_pfn = dma_addr >> PAGE_SHIFT;
-
-       /*
-        * set appropriate bounce gfp mask -- unfortunately we don't have a
-        * full 4GB zone, so we have to resort to low memory for any bounces.
-        * ISA has its own < 16MB zone.
-        */
-       if (bounce_pfn < blk_max_low_pfn) {
-               BUG_ON(dma_addr < BLK_BOUNCE_ISA);
+       int dma = 0;
+
+       q->bounce_gfp = GFP_NOIO;
+#if BITS_PER_LONG == 64
+       /* Assume anything <= 4GB can be handled by IOMMU.
+          Actually some IOMMUs can handle everything, but I don't
+          know of a way to test this here. */
+       if (bounce_pfn < (min_t(u64,0xffffffff,BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
+               dma = 1;
+       q->bounce_pfn = max_low_pfn;
+#else
+       if (bounce_pfn < blk_max_low_pfn)
+               dma = 1;
+       q->bounce_pfn = bounce_pfn;
+#endif
+       if (dma) {
                init_emergency_isa_pool();
                q->bounce_gfp = GFP_NOIO | GFP_DMA;
-       } else
-               q->bounce_gfp = GFP_NOIO;
-
-       q->bounce_pfn = bounce_pfn;
+               q->bounce_pfn = bounce_pfn;
+       }
 }
 
 EXPORT_SYMBOL(blk_queue_bounce_limit);
@@ -662,7 +596,7 @@ EXPORT_SYMBOL(blk_queue_bounce_limit);
  *    Enables a low level driver to set an upper limit on the size of
  *    received requests.
  **/
-void blk_queue_max_sectors(request_queue_t *q, unsigned short max_sectors)
+void blk_queue_max_sectors(struct request_queue *q, unsigned int max_sectors)
 {
        if ((max_sectors << 9) < PAGE_CACHE_SIZE) {
                max_sectors = 1 << (PAGE_CACHE_SHIFT - 9);
@@ -689,7 +623,8 @@ EXPORT_SYMBOL(blk_queue_max_sectors);
  *    physical data segments in a request.  This would be the largest sized
  *    scatter list the driver could handle.
  **/
-void blk_queue_max_phys_segments(request_queue_t *q, unsigned short max_segments)
+void blk_queue_max_phys_segments(struct request_queue *q,
+                                unsigned short max_segments)
 {
        if (!max_segments) {
                max_segments = 1;
@@ -712,7 +647,8 @@ EXPORT_SYMBOL(blk_queue_max_phys_segments);
  *    address/length pairs the host adapter can actually give as once
  *    to the device.
  **/
-void blk_queue_max_hw_segments(request_queue_t *q, unsigned short max_segments)
+void blk_queue_max_hw_segments(struct request_queue *q,
+                              unsigned short max_segments)
 {
        if (!max_segments) {
                max_segments = 1;
@@ -733,7 +669,7 @@ EXPORT_SYMBOL(blk_queue_max_hw_segments);
  *    Enables a low level driver to set an upper limit on the size of a
  *    coalesced segment
  **/
-void blk_queue_max_segment_size(request_queue_t *q, unsigned int max_size)
+void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
 {
        if (max_size < PAGE_CACHE_SIZE) {
                max_size = PAGE_CACHE_SIZE;
@@ -756,7 +692,7 @@ EXPORT_SYMBOL(blk_queue_max_segment_size);
  *   even internal read-modify-write operations). Usually the default
  *   of 512 covers most hardware.
  **/
-void blk_queue_hardsect_size(request_queue_t *q, unsigned short size)
+void blk_queue_hardsect_size(struct request_queue *q, unsigned short size)
 {
        q->hardsect_size = size;
 }
@@ -773,7 +709,7 @@ EXPORT_SYMBOL(blk_queue_hardsect_size);
  * @t: the stacking driver (top)
  * @b:  the underlying device (bottom)
  **/
-void blk_queue_stack_limits(request_queue_t *t, request_queue_t *b)
+void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
 {
        /* zero is "infinity" */
        t->max_sectors = min_not_zero(t->max_sectors,b->max_sectors);
@@ -783,6 +719,8 @@ void blk_queue_stack_limits(request_queue_t *t, request_queue_t *b)
        t->max_hw_segments = min(t->max_hw_segments,b->max_hw_segments);
        t->max_segment_size = min(t->max_segment_size,b->max_segment_size);
        t->hardsect_size = max(t->hardsect_size,b->hardsect_size);
+       if (!test_bit(QUEUE_FLAG_CLUSTER, &b->queue_flags))
+               clear_bit(QUEUE_FLAG_CLUSTER, &t->queue_flags);
 }
 
 EXPORT_SYMBOL(blk_queue_stack_limits);
@@ -792,7 +730,7 @@ EXPORT_SYMBOL(blk_queue_stack_limits);
  * @q:  the request queue for the device
  * @mask:  the memory boundary mask
  **/
-void blk_queue_segment_boundary(request_queue_t *q, unsigned long mask)
+void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
 {
        if (mask < PAGE_CACHE_SIZE - 1) {
                mask = PAGE_CACHE_SIZE - 1;
@@ -814,7 +752,7 @@ EXPORT_SYMBOL(blk_queue_segment_boundary);
  *    this is used when buiding direct io requests for the queue.
  *
  **/
-void blk_queue_dma_alignment(request_queue_t *q, int mask)
+void blk_queue_dma_alignment(struct request_queue *q, int mask)
 {
        q->dma_alignment = mask;
 }
@@ -832,36 +770,27 @@ EXPORT_SYMBOL(blk_queue_dma_alignment);
  *
  *    no locks need be held.
  **/
-struct request *blk_queue_find_tag(request_queue_t *q, int tag)
+struct request *blk_queue_find_tag(struct request_queue *q, int tag)
 {
-       struct blk_queue_tag *bqt = q->queue_tags;
-
-       if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
-               return NULL;
-
-       return bqt->tag_index[tag];
+       return blk_map_queue_find_tag(q->queue_tags, tag);
 }
 
 EXPORT_SYMBOL(blk_queue_find_tag);
 
 /**
- * __blk_queue_free_tags - release tag maintenance info
- * @q:  the request queue for the device
+ * __blk_free_tags - release a given set of tag maintenance info
+ * @bqt:       the tag map to free
  *
- *  Notes:
- *    blk_cleanup_queue() will take care of calling this function, if tagging
- *    has been used. So there's no need to call this directly.
- **/
-static void __blk_queue_free_tags(request_queue_t *q)
+ * Tries to free the specified @bqt@.  Returns true if it was
+ * actually freed and false if there are still references using it
+ */
+static int __blk_free_tags(struct blk_queue_tag *bqt)
 {
-       struct blk_queue_tag *bqt = q->queue_tags;
+       int retval;
 
-       if (!bqt)
-               return;
-
-       if (atomic_dec_and_test(&bqt->refcnt)) {
+       retval = atomic_dec_and_test(&bqt->refcnt);
+       if (retval) {
                BUG_ON(bqt->busy);
-               BUG_ON(!list_empty(&bqt->busy_list));
 
                kfree(bqt->tag_index);
                bqt->tag_index = NULL;
@@ -870,12 +799,49 @@ static void __blk_queue_free_tags(request_queue_t *q)
                bqt->tag_map = NULL;
 
                kfree(bqt);
+
        }
 
+       return retval;
+}
+
+/**
+ * __blk_queue_free_tags - release tag maintenance info
+ * @q:  the request queue for the device
+ *
+ *  Notes:
+ *    blk_cleanup_queue() will take care of calling this function, if tagging
+ *    has been used. So there's no need to call this directly.
+ **/
+static void __blk_queue_free_tags(struct request_queue *q)
+{
+       struct blk_queue_tag *bqt = q->queue_tags;
+
+       if (!bqt)
+               return;
+
+       __blk_free_tags(bqt);
+
        q->queue_tags = NULL;
        q->queue_flags &= ~(1 << QUEUE_FLAG_QUEUED);
 }
 
+
+/**
+ * blk_free_tags - release a given set of tag maintenance info
+ * @bqt:       the tag map to free
+ *
+ * For externally managed @bqt@ frees the map.  Callers of this
+ * function must guarantee to have released all the queues that
+ * might have been using this tag map.
+ */
+void blk_free_tags(struct blk_queue_tag *bqt)
+{
+       if (unlikely(!__blk_free_tags(bqt)))
+               BUG();
+}
+EXPORT_SYMBOL(blk_free_tags);
+
 /**
  * blk_queue_free_tags - release tag maintenance info
  * @q:  the request queue for the device
@@ -884,7 +850,7 @@ static void __blk_queue_free_tags(request_queue_t *q)
  *     This is used to disabled tagged queuing to a device, yet leave
  *     queue in function.
  **/
-void blk_queue_free_tags(request_queue_t *q)
+void blk_queue_free_tags(struct request_queue *q)
 {
        clear_bit(QUEUE_FLAG_QUEUED, &q->queue_flags);
 }
@@ -892,29 +858,27 @@ void blk_queue_free_tags(request_queue_t *q)
 EXPORT_SYMBOL(blk_queue_free_tags);
 
 static int
-init_tag_map(request_queue_t *q, struct blk_queue_tag *tags, int depth)
+init_tag_map(struct request_queue *q, struct blk_queue_tag *tags, int depth)
 {
        struct request **tag_index;
        unsigned long *tag_map;
        int nr_ulongs;
 
-       if (depth > q->nr_requests * 2) {
+       if (q && depth > q->nr_requests * 2) {
                depth = q->nr_requests * 2;
                printk(KERN_ERR "%s: adjusted depth to %d\n",
                                __FUNCTION__, depth);
        }
 
-       tag_index = kmalloc(depth * sizeof(struct request *), GFP_ATOMIC);
+       tag_index = kzalloc(depth * sizeof(struct request *), GFP_ATOMIC);
        if (!tag_index)
                goto fail;
 
        nr_ulongs = ALIGN(depth, BITS_PER_LONG) / BITS_PER_LONG;
-       tag_map = kmalloc(nr_ulongs * sizeof(unsigned long), GFP_ATOMIC);
+       tag_map = kzalloc(nr_ulongs * sizeof(unsigned long), GFP_ATOMIC);
        if (!tag_map)
                goto fail;
 
-       memset(tag_index, 0, depth * sizeof(struct request *));
-       memset(tag_map, 0, nr_ulongs * sizeof(unsigned long));
        tags->real_max_depth = depth;
        tags->max_depth = depth;
        tags->tag_index = tag_index;
@@ -926,13 +890,44 @@ fail:
        return -ENOMEM;
 }
 
+static struct blk_queue_tag *__blk_queue_init_tags(struct request_queue *q,
+                                                  int depth)
+{
+       struct blk_queue_tag *tags;
+
+       tags = kmalloc(sizeof(struct blk_queue_tag), GFP_ATOMIC);
+       if (!tags)
+               goto fail;
+
+       if (init_tag_map(q, tags, depth))
+               goto fail;
+
+       tags->busy = 0;
+       atomic_set(&tags->refcnt, 1);
+       return tags;
+fail:
+       kfree(tags);
+       return NULL;
+}
+
+/**
+ * blk_init_tags - initialize the tag info for an external tag map
+ * @depth:     the maximum queue depth supported
+ * @tags: the tag to use
+ **/
+struct blk_queue_tag *blk_init_tags(int depth)
+{
+       return __blk_queue_init_tags(NULL, depth);
+}
+EXPORT_SYMBOL(blk_init_tags);
+
 /**
  * blk_queue_init_tags - initialize the queue tag info
  * @q:  the request queue for the device
  * @depth:  the maximum queue depth supported
  * @tags: the tag to use
  **/
-int blk_queue_init_tags(request_queue_t *q, int depth,
+int blk_queue_init_tags(struct request_queue *q, int depth,
                        struct blk_queue_tag *tags)
 {
        int rc;
@@ -940,16 +935,10 @@ int blk_queue_init_tags(request_queue_t *q, int depth,
        BUG_ON(tags && q->queue_tags && tags != q->queue_tags);
 
        if (!tags && !q->queue_tags) {
-               tags = kmalloc(sizeof(struct blk_queue_tag), GFP_ATOMIC);
-               if (!tags)
-                       goto fail;
+               tags = __blk_queue_init_tags(q, depth);
 
-               if (init_tag_map(q, tags, depth))
+               if (!tags)
                        goto fail;
-
-               INIT_LIST_HEAD(&tags->busy_list);
-               tags->busy = 0;
-               atomic_set(&tags->refcnt, 1);
        } else if (q->queue_tags) {
                if ((rc = blk_queue_resize_tags(q, depth)))
                        return rc;
@@ -963,6 +952,7 @@ int blk_queue_init_tags(request_queue_t *q, int depth,
         */
        q->queue_tags = tags;
        q->queue_flags |= (1 << QUEUE_FLAG_QUEUED);
+       INIT_LIST_HEAD(&q->tag_busy_list);
        return 0;
 fail:
        kfree(tags);
@@ -979,7 +969,7 @@ EXPORT_SYMBOL(blk_queue_init_tags);
  *  Notes:
  *    Must be called with the queue lock held.
  **/
-int blk_queue_resize_tags(request_queue_t *q, int new_depth)
+int blk_queue_resize_tags(struct request_queue *q, int new_depth)
 {
        struct blk_queue_tag *bqt = q->queue_tags;
        struct request **tag_index;
@@ -1001,6 +991,13 @@ int blk_queue_resize_tags(request_queue_t *q, int new_depth)
        }
 
        /*
+        * Currently cannot replace a shared tag map with a new
+        * one, so error out if this is the case
+        */
+       if (atomic_read(&bqt->refcnt) != 1)
+               return -EBUSY;
+
+       /*
         * save the old state info, so we can copy it back
         */
        tag_index = bqt->tag_index;
@@ -1035,7 +1032,7 @@ EXPORT_SYMBOL(blk_queue_resize_tags);
  *  Notes:
  *   queue lock must be held.
  **/
-void blk_queue_end_tag(request_queue_t *q, struct request *rq)
+void blk_queue_end_tag(struct request_queue *q, struct request *rq)
 {
        struct blk_queue_tag *bqt = q->queue_tags;
        int tag = rq->tag;
@@ -1049,14 +1046,8 @@ void blk_queue_end_tag(request_queue_t *q, struct request *rq)
                 */
                return;
 
-       if (unlikely(!__test_and_clear_bit(tag, bqt->tag_map))) {
-               printk(KERN_ERR "%s: attempt to clear non-busy tag (%d)\n",
-                      __FUNCTION__, tag);
-               return;
-       }
-
        list_del_init(&rq->queuelist);
-       rq->flags &= ~REQ_QUEUED;
+       rq->cmd_flags &= ~REQ_QUEUED;
        rq->tag = -1;
 
        if (unlikely(bqt->tag_index[tag] == NULL))
@@ -1064,6 +1055,17 @@ void blk_queue_end_tag(request_queue_t *q, struct request *rq)
                       __FUNCTION__, tag);
 
        bqt->tag_index[tag] = NULL;
+
+       if (unlikely(!test_bit(tag, bqt->tag_map))) {
+               printk(KERN_ERR "%s: attempt to clear non-busy tag (%d)\n",
+                      __FUNCTION__, tag);
+               return;
+       }
+       /*
+        * The tag_map bit acts as a lock for tag_index[bit], so we need
+        * unlock memory barrier semantics.
+        */
+       clear_bit_unlock(tag, bqt->tag_map);
        bqt->busy--;
 }
 
@@ -1087,12 +1089,12 @@ EXPORT_SYMBOL(blk_queue_end_tag);
  *  Notes:
  *   queue lock must be held.
  **/
-int blk_queue_start_tag(request_queue_t *q, struct request *rq)
+int blk_queue_start_tag(struct request_queue *q, struct request *rq)
 {
        struct blk_queue_tag *bqt = q->queue_tags;
        int tag;
 
-       if (unlikely((rq->flags & REQ_QUEUED))) {
+       if (unlikely((rq->cmd_flags & REQ_QUEUED))) {
                printk(KERN_ERR 
                       "%s: request %p for device [%s] already tagged %d",
                       __FUNCTION__, rq,
@@ -1100,17 +1102,26 @@ int blk_queue_start_tag(request_queue_t *q, struct request *rq)
                BUG();
        }
 
-       tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth);
-       if (tag >= bqt->max_depth)
-               return 1;
+       /*
+        * Protect against shared tag maps, as we may not have exclusive
+        * access to the tag map.
+        */
+       do {
+               tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth);
+               if (tag >= bqt->max_depth)
+                       return 1;
 
-       __set_bit(tag, bqt->tag_map);
+       } while (test_and_set_bit_lock(tag, bqt->tag_map));
+       /*
+        * We need lock ordering semantics given by test_and_set_bit_lock.
+        * See blk_queue_end_tag for details.
+        */
 
-       rq->flags |= REQ_QUEUED;
+       rq->cmd_flags |= REQ_QUEUED;
        rq->tag = tag;
        bqt->tag_index[tag] = rq;
        blkdev_dequeue_request(rq);
-       list_add(&rq->queuelist, &bqt->busy_list);
+       list_add(&rq->queuelist, &q->tag_busy_list);
        bqt->busy++;
        return 0;
 }
@@ -1129,78 +1140,30 @@ EXPORT_SYMBOL(blk_queue_start_tag);
  *  Notes:
  *   queue lock must be held.
  **/
-void blk_queue_invalidate_tags(request_queue_t *q)
+void blk_queue_invalidate_tags(struct request_queue *q)
 {
-       struct blk_queue_tag *bqt = q->queue_tags;
        struct list_head *tmp, *n;
-       struct request *rq;
-
-       list_for_each_safe(tmp, n, &bqt->busy_list) {
-               rq = list_entry_rq(tmp);
-
-               if (rq->tag == -1) {
-                       printk(KERN_ERR
-                              "%s: bad tag found on list\n", __FUNCTION__);
-                       list_del_init(&rq->queuelist);
-                       rq->flags &= ~REQ_QUEUED;
-               } else
-                       blk_queue_end_tag(q, rq);
 
-               rq->flags &= ~REQ_STARTED;
-               __elv_add_request(q, rq, ELEVATOR_INSERT_BACK, 0);
-       }
+       list_for_each_safe(tmp, n, &q->tag_busy_list)
+               blk_requeue_request(q, list_entry_rq(tmp));
 }
 
 EXPORT_SYMBOL(blk_queue_invalidate_tags);
 
-static const char * const rq_flags[] = {
-       "REQ_RW",
-       "REQ_FAILFAST",
-       "REQ_SORTED",
-       "REQ_SOFTBARRIER",
-       "REQ_HARDBARRIER",
-       "REQ_FUA",
-       "REQ_CMD",
-       "REQ_NOMERGE",
-       "REQ_STARTED",
-       "REQ_DONTPREP",
-       "REQ_QUEUED",
-       "REQ_ELVPRIV",
-       "REQ_PC",
-       "REQ_BLOCK_PC",
-       "REQ_SENSE",
-       "REQ_FAILED",
-       "REQ_QUIET",
-       "REQ_SPECIAL",
-       "REQ_DRIVE_CMD",
-       "REQ_DRIVE_TASK",
-       "REQ_DRIVE_TASKFILE",
-       "REQ_PREEMPT",
-       "REQ_PM_SUSPEND",
-       "REQ_PM_RESUME",
-       "REQ_PM_SHUTDOWN",
-       "REQ_ORDERED_COLOR",
-};
-
 void blk_dump_rq_flags(struct request *rq, char *msg)
 {
        int bit;
 
-       printk("%s: dev %s: flags = ", msg,
-               rq->rq_disk ? rq->rq_disk->disk_name : "?");
-       bit = 0;
-       do {
-               if (rq->flags & (1 << bit))
-                       printk("%s ", rq_flags[bit]);
-               bit++;
-       } while (bit < __REQ_NR_BITS);
+       printk("%s: dev %s: type=%x, flags=%x\n", msg,
+               rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->cmd_type,
+               rq->cmd_flags);
 
        printk("\nsector %llu, nr/cnr %lu/%u\n", (unsigned long long)rq->sector,
                                                       rq->nr_sectors,
                                                       rq->current_nr_sectors);
        printk("bio %p, biotail %p, buffer %p, data %p, len %u\n", rq->bio, rq->biotail, rq->buffer, rq->data, rq->data_len);
 
-       if (rq->flags & (REQ_BLOCK_PC | REQ_PC)) {
+       if (blk_pc_request(rq)) {
                printk("cdb: ");
                for (bit = 0; bit < sizeof(rq->cmd); bit++)
                        printk("%02x ", rq->cmd[bit]);
@@ -1210,24 +1173,48 @@ void blk_dump_rq_flags(struct request *rq, char *msg)
 
 EXPORT_SYMBOL(blk_dump_rq_flags);
 
-void blk_recount_segments(request_queue_t *q, struct bio *bio)
+void blk_recount_segments(struct request_queue *q, struct bio *bio)
+{
+       struct request rq;
+       struct bio *nxt = bio->bi_next;
+       rq.q = q;
+       rq.bio = rq.biotail = bio;
+       bio->bi_next = NULL;
+       blk_recalc_rq_segments(&rq);
+       bio->bi_next = nxt;
+       bio->bi_phys_segments = rq.nr_phys_segments;
+       bio->bi_hw_segments = rq.nr_hw_segments;
+       bio->bi_flags |= (1 << BIO_SEG_VALID);
+}
+EXPORT_SYMBOL(blk_recount_segments);
+
+static void blk_recalc_rq_segments(struct request *rq)
 {
+       int nr_phys_segs;
+       int nr_hw_segs;
+       unsigned int phys_size;
+       unsigned int hw_size;
        struct bio_vec *bv, *bvprv = NULL;
-       int i, nr_phys_segs, nr_hw_segs, seg_size, hw_seg_size, cluster;
+       int seg_size;
+       int hw_seg_size;
+       int cluster;
+       struct req_iterator iter;
        int high, highprv = 1;
+       struct request_queue *q = rq->q;
 
-       if (unlikely(!bio->bi_io_vec))
+       if (!rq->bio)
                return;
 
        cluster = q->queue_flags & (1 << QUEUE_FLAG_CLUSTER);
-       hw_seg_size = seg_size = nr_phys_segs = nr_hw_segs = 0;
-       bio_for_each_segment(bv, bio, i) {
+       hw_seg_size = seg_size = 0;
+       phys_size = hw_size = nr_phys_segs = nr_hw_segs = 0;
+       rq_for_each_segment(bv, rq, iter) {
                /*
                 * the trick here is making sure that a high page is never
                 * considered part of another segment, since that might
                 * change with the bounce page.
                 */
-               high = page_to_pfn(bv->bv_page) >= q->bounce_pfn;
+               high = page_to_pfn(bv->bv_page) > q->bounce_pfn;
                if (high || highprv)
                        goto new_hw_segment;
                if (cluster) {
@@ -1247,12 +1234,13 @@ void blk_recount_segments(request_queue_t *q, struct bio *bio)
                }
 new_segment:
                if (BIOVEC_VIRT_MERGEABLE(bvprv, bv) &&
-                   !BIOVEC_VIRT_OVERSIZE(hw_seg_size + bv->bv_len)) {
+                   !BIOVEC_VIRT_OVERSIZE(hw_seg_size + bv->bv_len))
                        hw_seg_size += bv->bv_len;
-               } else {
+               else {
 new_hw_segment:
-                       if (hw_seg_size > bio->bi_hw_front_size)
-                               bio->bi_hw_front_size = hw_seg_size;
+                       if (nr_hw_segs == 1 &&
+                           hw_seg_size > rq->bio->bi_hw_front_size)
+                               rq->bio->bi_hw_front_size = hw_seg_size;
                        hw_seg_size = BIOVEC_VIRT_START_SIZE(bv) + bv->bv_len;
                        nr_hw_segs++;
                }
@@ -1262,17 +1250,17 @@ new_hw_segment:
                seg_size = bv->bv_len;
                highprv = high;
        }
-       if (hw_seg_size > bio->bi_hw_back_size)
-               bio->bi_hw_back_size = hw_seg_size;
-       if (nr_hw_segs == 1 && hw_seg_size > bio->bi_hw_front_size)
-               bio->bi_hw_front_size = hw_seg_size;
-       bio->bi_phys_segments = nr_phys_segs;
-       bio->bi_hw_segments = nr_hw_segs;
-       bio->bi_flags |= (1 << BIO_SEG_VALID);
-}
 
+       if (nr_hw_segs == 1 &&
+           hw_seg_size > rq->bio->bi_hw_front_size)
+               rq->bio->bi_hw_front_size = hw_seg_size;
+       if (hw_seg_size > rq->biotail->bi_hw_back_size)
+               rq->biotail->bi_hw_back_size = hw_seg_size;
+       rq->nr_phys_segments = nr_phys_segs;
+       rq->nr_hw_segments = nr_hw_segs;
+}
 
-static int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
+static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
                                   struct bio *nxt)
 {
        if (!(q->queue_flags & (1 << QUEUE_FLAG_CLUSTER)))
@@ -1293,7 +1281,7 @@ static int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
        return 0;
 }
 
-static int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
+static int blk_hw_contig_segment(struct request_queue *q, struct bio *bio,
                                 struct bio *nxt)
 {
        if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
@@ -1301,9 +1289,9 @@ static int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
        if (unlikely(!bio_flagged(nxt, BIO_SEG_VALID)))
                blk_recount_segments(q, nxt);
        if (!BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt)) ||
-           BIOVEC_VIRT_OVERSIZE(bio->bi_hw_front_size + bio->bi_hw_back_size))
+           BIOVEC_VIRT_OVERSIZE(bio->bi_hw_back_size + nxt->bi_hw_front_size))
                return 0;
-       if (bio->bi_size + nxt->bi_size > q->max_segment_size)
+       if (bio->bi_hw_back_size + nxt->bi_hw_front_size > q->max_segment_size)
                return 0;
 
        return 1;
@@ -1313,11 +1301,13 @@ static int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
  * map a request to scatterlist, return number of sg entries setup. Caller
  * must make sure sg can hold rq->nr_phys_segments entries
  */
-int blk_rq_map_sg(request_queue_t *q, struct request *rq, struct scatterlist *sg)
+int blk_rq_map_sg(struct request_queue *q, struct request *rq,
+                 struct scatterlist *sglist)
 {
        struct bio_vec *bvec, *bvprv;
-       struct bio *bio;
-       int nsegs, i, cluster;
+       struct req_iterator iter;
+       struct scatterlist *sg;
+       int nsegs, cluster;
 
        nsegs = 0;
        cluster = q->queue_flags & (1 << QUEUE_FLAG_CLUSTER);
@@ -1326,35 +1316,47 @@ int blk_rq_map_sg(request_queue_t *q, struct request *rq, struct scatterlist *sg
         * for each bio in rq
         */
        bvprv = NULL;
-       rq_for_each_bio(bio, rq) {
-               /*
-                * for each segment in bio
-                */
-               bio_for_each_segment(bvec, bio, i) {
-                       int nbytes = bvec->bv_len;
+       sg = NULL;
+       rq_for_each_segment(bvec, rq, iter) {
+               int nbytes = bvec->bv_len;
 
-                       if (bvprv && cluster) {
-                               if (sg[nsegs - 1].length + nbytes > q->max_segment_size)
-                                       goto new_segment;
+               if (bvprv && cluster) {
+                       if (sg->length + nbytes > q->max_segment_size)
+                               goto new_segment;
 
-                               if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
-                                       goto new_segment;
-                               if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
-                                       goto new_segment;
+                       if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
+                               goto new_segment;
+                       if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
+                               goto new_segment;
 
-                               sg[nsegs - 1].length += nbytes;
-                       } else {
+                       sg->length += nbytes;
+               } else {
 new_segment:
-                               memset(&sg[nsegs],0,sizeof(struct scatterlist));
-                               sg[nsegs].page = bvec->bv_page;
-                               sg[nsegs].length = nbytes;
-                               sg[nsegs].offset = bvec->bv_offset;
-
-                               nsegs++;
+                       if (!sg)
+                               sg = sglist;
+                       else {
+                               /*
+                                * If the driver previously mapped a shorter
+                                * list, we could see a termination bit
+                                * prematurely unless it fully inits the sg
+                                * table on each mapping. We KNOW that there
+                                * must be more entries here or the driver
+                                * would be buggy, so force clear the
+                                * termination bit to avoid doing a full
+                                * sg_init_table() in drivers for each command.
+                                */
+                               sg->page_link &= ~0x02;
+                               sg = sg_next(sg);
                        }
-                       bvprv = bvec;
-               } /* segments in bio */
-       } /* bios in rq */
+
+                       sg_set_page(sg, bvec->bv_page, nbytes, bvec->bv_offset);
+                       nsegs++;
+               }
+               bvprv = bvec;
+       } /* segments in rq */
+
+       if (sg)
+               sg_mark_end(sg);
 
        return nsegs;
 }
@@ -1366,14 +1368,14 @@ EXPORT_SYMBOL(blk_rq_map_sg);
  * specific ones if so desired
  */
 
-static inline int ll_new_mergeable(request_queue_t *q,
+static inline int ll_new_mergeable(struct request_queue *q,
                                   struct request *req,
                                   struct bio *bio)
 {
        int nr_phys_segs = bio_phys_segments(q, bio);
 
        if (req->nr_phys_segments + nr_phys_segs > q->max_phys_segments) {
-               req->flags |= REQ_NOMERGE;
+               req->cmd_flags |= REQ_NOMERGE;
                if (req == q->last_merge)
                        q->last_merge = NULL;
                return 0;
@@ -1387,7 +1389,7 @@ static inline int ll_new_mergeable(request_queue_t *q,
        return 1;
 }
 
-static inline int ll_new_hw_segment(request_queue_t *q,
+static inline int ll_new_hw_segment(struct request_queue *q,
                                    struct request *req,
                                    struct bio *bio)
 {
@@ -1396,7 +1398,7 @@ static inline int ll_new_hw_segment(request_queue_t *q,
 
        if (req->nr_hw_segments + nr_hw_segs > q->max_hw_segments
            || req->nr_phys_segments + nr_phys_segs > q->max_phys_segments) {
-               req->flags |= REQ_NOMERGE;
+               req->cmd_flags |= REQ_NOMERGE;
                if (req == q->last_merge)
                        q->last_merge = NULL;
                return 0;
@@ -1411,7 +1413,7 @@ static inline int ll_new_hw_segment(request_queue_t *q,
        return 1;
 }
 
-static int ll_back_merge_fn(request_queue_t *q, struct request *req, 
+static int ll_back_merge_fn(struct request_queue *q, struct request *req,
                            struct bio *bio)
 {
        unsigned short max_sectors;
@@ -1423,7 +1425,7 @@ static int ll_back_merge_fn(request_queue_t *q, struct request *req,
                max_sectors = q->max_sectors;
 
        if (req->nr_sectors + bio_sectors(bio) > max_sectors) {
-               req->flags |= REQ_NOMERGE;
+               req->cmd_flags |= REQ_NOMERGE;
                if (req == q->last_merge)
                        q->last_merge = NULL;
                return 0;
@@ -1449,7 +1451,7 @@ static int ll_back_merge_fn(request_queue_t *q, struct request *req,
        return ll_new_hw_segment(q, req, bio);
 }
 
-static int ll_front_merge_fn(request_queue_t *q, struct request *req, 
+static int ll_front_merge_fn(struct request_queue *q, struct request *req, 
                             struct bio *bio)
 {
        unsigned short max_sectors;
@@ -1462,7 +1464,7 @@ static int ll_front_merge_fn(request_queue_t *q, struct request *req,
 
 
        if (req->nr_sectors + bio_sectors(bio) > max_sectors) {
-               req->flags |= REQ_NOMERGE;
+               req->cmd_flags |= REQ_NOMERGE;
                if (req == q->last_merge)
                        q->last_merge = NULL;
                return 0;
@@ -1488,7 +1490,7 @@ static int ll_front_merge_fn(request_queue_t *q, struct request *req,
        return ll_new_hw_segment(q, req, bio);
 }
 
-static int ll_merge_requests_fn(request_queue_t *q, struct request *req,
+static int ll_merge_requests_fn(struct request_queue *q, struct request *req,
                                struct request *next)
 {
        int total_phys_segments;
@@ -1544,7 +1546,7 @@ static int ll_merge_requests_fn(request_queue_t *q, struct request *req,
  * This is called with interrupts off and no requests on the queue and
  * with the queue lock held.
  */
-void blk_plug_device(request_queue_t *q)
+void blk_plug_device(struct request_queue *q)
 {
        WARN_ON(!irqs_disabled());
 
@@ -1552,11 +1554,13 @@ void blk_plug_device(request_queue_t *q)
         * don't plug a stopped queue, it must be paired with blk_start_queue()
         * which will restart the queueing
         */
-       if (test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags))
+       if (blk_queue_stopped(q))
                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);
@@ -1565,7 +1569,7 @@ EXPORT_SYMBOL(blk_plug_device);
  * remove the queue from the plugged list, if present. called with
  * queue lock held and interrupts disabled.
  */
-int blk_remove_plug(request_queue_t *q)
+int blk_remove_plug(struct request_queue *q)
 {
        WARN_ON(!irqs_disabled());
 
@@ -1581,9 +1585,9 @@ EXPORT_SYMBOL(blk_remove_plug);
 /*
  * remove the plug and let it rip..
  */
-void __generic_unplug_device(request_queue_t *q)
+void __generic_unplug_device(struct request_queue *q)
 {
-       if (unlikely(test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags)))
+       if (unlikely(blk_queue_stopped(q)))
                return;
 
        if (!blk_remove_plug(q))
@@ -1595,7 +1599,7 @@ EXPORT_SYMBOL(__generic_unplug_device);
 
 /**
  * generic_unplug_device - fire a request queue
- * @q:    The &request_queue_t in question
+ * @q:    The &struct request_queue in question
  *
  * Description:
  *   Linux uses plugging to build bigger requests queues before letting
@@ -1604,7 +1608,7 @@ EXPORT_SYMBOL(__generic_unplug_device);
  *   gets unplugged, the request_fn defined for the queue is invoked and
  *   transfers started.
  **/
-void generic_unplug_device(request_queue_t *q)
+void generic_unplug_device(struct request_queue *q)
 {
        spin_lock_irq(q->queue_lock);
        __generic_unplug_device(q);
@@ -1615,40 +1619,59 @@ EXPORT_SYMBOL(generic_unplug_device);
 static void blk_backing_dev_unplug(struct backing_dev_info *bdi,
                                   struct page *page)
 {
-       request_queue_t *q = bdi->unplug_io_data;
+       struct request_queue *q = bdi->unplug_io_data;
 
-       /*
-        * devices don't necessarily have an ->unplug_fn defined
-        */
-       if (q->unplug_fn)
-               q->unplug_fn(q);
+       blk_unplug(q);
 }
 
-static void blk_unplug_work(void *data)
+static void blk_unplug_work(struct work_struct *work)
 {
-       request_queue_t *q = data;
+       struct request_queue *q =
+               container_of(work, struct request_queue, unplug_work);
+
+       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_timeout(unsigned long data)
 {
-       request_queue_t *q = (request_queue_t *)data;
+       struct request_queue *q = (struct request_queue *)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);
 }
 
+void blk_unplug(struct request_queue *q)
+{
+       /*
+        * devices don't necessarily have an ->unplug_fn defined
+        */
+       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);
+       }
+}
+EXPORT_SYMBOL(blk_unplug);
+
 /**
  * blk_start_queue - restart a previously stopped queue
- * @q:    The &request_queue_t in question
+ * @q:    The &struct request_queue in question
  *
  * Description:
  *   blk_start_queue() will clear the stop flag on the queue, and call
  *   the request_fn for the queue if it was in a stopped state when
  *   entered. Also see blk_stop_queue(). Queue lock must be held.
  **/
-void blk_start_queue(request_queue_t *q)
+void blk_start_queue(struct request_queue *q)
 {
+       WARN_ON(!irqs_disabled());
+
        clear_bit(QUEUE_FLAG_STOPPED, &q->queue_flags);
 
        /*
@@ -1668,7 +1691,7 @@ EXPORT_SYMBOL(blk_start_queue);
 
 /**
  * blk_stop_queue - stop a queue
- * @q:    The &request_queue_t in question
+ * @q:    The &struct request_queue in question
  *
  * Description:
  *   The Linux block layer assumes that a block driver will consume all
@@ -1680,7 +1703,7 @@ EXPORT_SYMBOL(blk_start_queue);
  *   the driver has signalled it's ready to go again. This happens by calling
  *   blk_start_queue() to restart queue operations. Queue lock must be held.
  **/
-void blk_stop_queue(request_queue_t *q)
+void blk_stop_queue(struct request_queue *q)
 {
        blk_remove_plug(q);
        set_bit(QUEUE_FLAG_STOPPED, &q->queue_flags);
@@ -1696,7 +1719,7 @@ EXPORT_SYMBOL(blk_stop_queue);
  *     on a queue, such as calling the unplug function after a timeout.
  *     A block device may call blk_sync_queue to ensure that any
  *     such activity is cancelled, thus allowing it to release resources
- *     the the callbacks might use. The caller must already have made sure
+ *     that the callbacks might use. The caller must already have made sure
  *     that its ->make_request_fn will not re-add plugging prior to calling
  *     this function.
  *
@@ -1704,7 +1727,7 @@ EXPORT_SYMBOL(blk_stop_queue);
 void blk_sync_queue(struct request_queue *q)
 {
        del_timer_sync(&q->unplug_timer);
-       kblockd_flush();
+       kblockd_flush_work(&q->unplug_work);
 }
 EXPORT_SYMBOL(blk_sync_queue);
 
@@ -1718,15 +1741,28 @@ void blk_run_queue(struct request_queue *q)
 
        spin_lock_irqsave(q->queue_lock, flags);
        blk_remove_plug(q);
-       if (!elv_queue_empty(q))
-               q->request_fn(q);
+
+       /*
+        * Only recurse once to avoid overrunning the stack, let the unplug
+        * handling reinvoke the handler shortly if we already got there.
+        */
+       if (!elv_queue_empty(q)) {
+               if (!test_and_set_bit(QUEUE_FLAG_REENTER, &q->queue_flags)) {
+                       q->request_fn(q);
+                       clear_bit(QUEUE_FLAG_REENTER, &q->queue_flags);
+               } else {
+                       blk_plug_device(q);
+                       kblockd_schedule_work(&q->unplug_work);
+               }
+       }
+
        spin_unlock_irqrestore(q->queue_lock, flags);
 }
 EXPORT_SYMBOL(blk_run_queue);
 
 /**
- * blk_cleanup_queue: - release a &request_queue_t when it is no longer needed
- * @q:    the request queue to be released
+ * blk_cleanup_queue: - release a &struct request_queue when it is no longer needed
+ * @kobj:    the kobj belonging of the request queue to be released
  *
  * Description:
  *     blk_cleanup_queue is the pair to blk_init_queue() or
@@ -1739,16 +1775,12 @@ EXPORT_SYMBOL(blk_run_queue);
  *     Hopefully the low level driver will have finished any
  *     outstanding requests first...
  **/
-void blk_cleanup_queue(request_queue_t * q)
+static void blk_release_queue(struct kobject *kobj)
 {
+       struct request_queue *q =
+               container_of(kobj, struct request_queue, kobj);
        struct request_list *rl = &q->rq;
 
-       if (!atomic_dec_and_test(&q->refcnt))
-               return;
-
-       if (q->elevator)
-               elevator_exit(q->elevator);
-
        blk_sync_queue(q);
 
        if (rl->rq_pool)
@@ -1757,12 +1789,33 @@ void blk_cleanup_queue(request_queue_t * q)
        if (q->queue_tags)
                __blk_queue_free_tags(q);
 
+       blk_trace_shutdown(q);
+
+       bdi_destroy(&q->backing_dev_info);
        kmem_cache_free(requestq_cachep, q);
 }
 
+void blk_put_queue(struct request_queue *q)
+{
+       kobject_put(&q->kobj);
+}
+EXPORT_SYMBOL(blk_put_queue);
+
+void blk_cleanup_queue(struct request_queue * q)
+{
+       mutex_lock(&q->sysfs_lock);
+       set_bit(QUEUE_FLAG_DEAD, &q->queue_flags);
+       mutex_unlock(&q->sysfs_lock);
+
+       if (q->elevator)
+               elevator_exit(q->elevator);
+
+       blk_put_queue(q);
+}
+
 EXPORT_SYMBOL(blk_cleanup_queue);
 
-static int blk_init_free_list(request_queue_t *q)
+static int blk_init_free_list(struct request_queue *q)
 {
        struct request_list *rl = &q->rq;
 
@@ -1781,26 +1834,39 @@ static int blk_init_free_list(request_queue_t *q)
        return 0;
 }
 
-request_queue_t *blk_alloc_queue(gfp_t gfp_mask)
+struct request_queue *blk_alloc_queue(gfp_t gfp_mask)
 {
        return blk_alloc_queue_node(gfp_mask, -1);
 }
 EXPORT_SYMBOL(blk_alloc_queue);
 
-request_queue_t *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
+static struct kobj_type queue_ktype;
+
+struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
 {
-       request_queue_t *q;
+       struct request_queue *q;
+       int err;
 
-       q = kmem_cache_alloc_node(requestq_cachep, gfp_mask, node_id);
+       q = kmem_cache_alloc_node(requestq_cachep,
+                               gfp_mask | __GFP_ZERO, node_id);
        if (!q)
                return NULL;
 
-       memset(q, 0, sizeof(*q));
-       init_timer(&q->unplug_timer);
-       atomic_set(&q->refcnt, 1);
-
        q->backing_dev_info.unplug_io_fn = blk_backing_dev_unplug;
        q->backing_dev_info.unplug_io_data = q;
+       err = bdi_init(&q->backing_dev_info);
+       if (err) {
+               kmem_cache_free(requestq_cachep, q);
+               return NULL;
+       }
+
+       init_timer(&q->unplug_timer);
+
+       kobject_set_name(&q->kobj, "%s", "queue");
+       q->kobj.ktype = &queue_ktype;
+       kobject_init(&q->kobj);
+
+       mutex_init(&q->sysfs_lock);
 
        return q;
 }
@@ -1828,7 +1894,8 @@ EXPORT_SYMBOL(blk_alloc_queue_node);
  *    get dealt with eventually.
  *
  *    The queue spin lock must be held while manipulating the requests on the
- *    request queue.
+ *    request queue; this lock will be taken also from interrupt context, so irq
+ *    disabling is needed for it.
  *
  *    Function returns a pointer to the initialized request queue, or NULL if
  *    it didn't succeed.
@@ -1838,23 +1905,25 @@ EXPORT_SYMBOL(blk_alloc_queue_node);
  *    when the block device is deactivated (such as at module unload).
  **/
 
-request_queue_t *blk_init_queue(request_fn_proc *rfn, spinlock_t *lock)
+struct request_queue *blk_init_queue(request_fn_proc *rfn, spinlock_t *lock)
 {
        return blk_init_queue_node(rfn, lock, -1);
 }
 EXPORT_SYMBOL(blk_init_queue);
 
-request_queue_t *
+struct request_queue *
 blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
 {
-       request_queue_t *q = blk_alloc_queue_node(GFP_KERNEL, node_id);
+       struct request_queue *q = blk_alloc_queue_node(GFP_KERNEL, node_id);
 
        if (!q)
                return NULL;
 
        q->node = node_id;
-       if (blk_init_free_list(q))
-               goto out_init;
+       if (blk_init_free_list(q)) {
+               kmem_cache_free(requestq_cachep, q);
+               return NULL;
+       }
 
        /*
         * if caller didn't supply a lock, they get per-queue locking with
@@ -1866,9 +1935,6 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
        }
 
        q->request_fn           = rfn;
-       q->back_merge_fn        = ll_back_merge_fn;
-       q->front_merge_fn       = ll_front_merge_fn;
-       q->merge_requests_fn    = ll_merge_requests_fn;
        q->prep_rq_fn           = NULL;
        q->unplug_fn            = generic_unplug_device;
        q->queue_flags          = (1 << QUEUE_FLAG_CLUSTER);
@@ -1882,6 +1948,8 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
        blk_queue_max_hw_segments(q, MAX_HW_SEGMENTS);
        blk_queue_max_phys_segments(q, MAX_PHYS_SEGMENTS);
 
+       q->sg_reserved_size = INT_MAX;
+
        /*
         * all done
         */
@@ -1890,17 +1958,15 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
                return q;
        }
 
-       blk_cleanup_queue(q);
-out_init:
-       kmem_cache_free(requestq_cachep, q);
+       blk_put_queue(q);
        return NULL;
 }
 EXPORT_SYMBOL(blk_init_queue_node);
 
-int blk_get_queue(request_queue_t *q)
+int blk_get_queue(struct request_queue *q)
 {
        if (likely(!test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) {
-               atomic_inc(&q->refcnt);
+               kobject_get(&q->kobj);
                return 0;
        }
 
@@ -1909,16 +1975,15 @@ int blk_get_queue(request_queue_t *q)
 
 EXPORT_SYMBOL(blk_get_queue);
 
-static inline void blk_free_request(request_queue_t *q, struct request *rq)
+static inline void blk_free_request(struct request_queue *q, struct request *rq)
 {
-       if (rq->flags & REQ_ELVPRIV)
+       if (rq->cmd_flags & REQ_ELVPRIV)
                elv_put_request(q, rq);
        mempool_free(rq, q->rq.rq_pool);
 }
 
-static inline struct request *
-blk_alloc_request(request_queue_t *q, int rw, struct bio *bio,
-                 int priv, gfp_t gfp_mask)
+static struct request *
+blk_alloc_request(struct request_queue *q, int rw, int priv, gfp_t gfp_mask)
 {
        struct request *rq = mempool_alloc(q->rq.rq_pool, gfp_mask);
 
@@ -1926,17 +1991,17 @@ blk_alloc_request(request_queue_t *q, int rw, struct bio *bio,
                return NULL;
 
        /*
-        * first three bits are identical in rq->flags and bio->bi_rw,
+        * first three bits are identical in rq->cmd_flags and bio->bi_rw,
         * see bio.h and blkdev.h
         */
-       rq->flags = rw;
+       rq->cmd_flags = rw | REQ_ALLOCED;
 
        if (priv) {
-               if (unlikely(elv_set_request(q, rq, bio, gfp_mask))) {
+               if (unlikely(elv_set_request(q, rq, gfp_mask))) {
                        mempool_free(rq, q->rq.rq_pool);
                        return NULL;
                }
-               rq->flags |= REQ_ELVPRIV;
+               rq->cmd_flags |= REQ_ELVPRIV;
        }
 
        return rq;
@@ -1946,7 +2011,7 @@ blk_alloc_request(request_queue_t *q, int rw, struct bio *bio,
  * ioc_batching returns true if the ioc is a valid batching request and
  * should be given priority access to a request.
  */
-static inline int ioc_batching(request_queue_t *q, struct io_context *ioc)
+static inline int ioc_batching(struct request_queue *q, struct io_context *ioc)
 {
        if (!ioc)
                return 0;
@@ -1967,7 +2032,7 @@ static inline int ioc_batching(request_queue_t *q, struct io_context *ioc)
  * is the behaviour we want though - once it gets a wakeup it should be given
  * a nice run.
  */
-static void ioc_set_batching(request_queue_t *q, struct io_context *ioc)
+static void ioc_set_batching(struct request_queue *q, struct io_context *ioc)
 {
        if (!ioc || ioc_batching(q, ioc))
                return;
@@ -1976,12 +2041,12 @@ static void ioc_set_batching(request_queue_t *q, struct io_context *ioc)
        ioc->last_waited = jiffies;
 }
 
-static void __freed_request(request_queue_t *q, int rw)
+static void __freed_request(struct request_queue *q, int rw)
 {
        struct request_list *rl = &q->rq;
 
        if (rl->count[rw] < queue_congestion_off_threshold(q))
-               clear_queue_congested(q, rw);
+               blk_clear_queue_congested(q, rw);
 
        if (rl->count[rw] + 1 <= q->nr_requests) {
                if (waitqueue_active(&rl->wait[rw]))
@@ -1995,7 +2060,7 @@ static void __freed_request(request_queue_t *q, int rw)
  * A request has just been released.  Account for it, update the full and
  * congestion status, wake up any waiters.   Called under q->queue_lock.
  */
-static void freed_request(request_queue_t *q, int rw, int priv)
+static void freed_request(struct request_queue *q, int rw, int priv)
 {
        struct request_list *rl = &q->rq;
 
@@ -2015,21 +2080,22 @@ static void freed_request(request_queue_t *q, int rw, int priv)
  * Returns NULL on failure, with queue_lock held.
  * Returns !NULL on success, with queue_lock *not held*.
  */
-static struct request *get_request(request_queue_t *q, int rw, struct bio *bio,
-                                  gfp_t gfp_mask)
+static struct request *get_request(struct request_queue *q, int rw_flags,
+                                  struct bio *bio, gfp_t gfp_mask)
 {
        struct request *rq = NULL;
        struct request_list *rl = &q->rq;
        struct io_context *ioc = NULL;
+       const int rw = rw_flags & 0x01;
        int may_queue, priv;
 
-       may_queue = elv_may_queue(q, rw, bio);
+       may_queue = elv_may_queue(q, rw_flags);
        if (may_queue == ELV_MQUEUE_NO)
                goto rq_starved;
 
        if (rl->count[rw]+1 >= queue_congestion_on_threshold(q)) {
                if (rl->count[rw]+1 >= q->nr_requests) {
-                       ioc = current_io_context(GFP_ATOMIC);
+                       ioc = current_io_context(GFP_ATOMIC, q->node);
                        /*
                         * The queue will fill after this allocation, so set
                         * it as full, and mark this process as "batching".
@@ -2051,7 +2117,7 @@ static struct request *get_request(request_queue_t *q, int rw, struct bio *bio,
                                }
                        }
                }
-               set_queue_congested(q, rw);
+               blk_set_queue_congested(q, rw);
        }
 
        /*
@@ -2071,7 +2137,7 @@ static struct request *get_request(request_queue_t *q, int rw, struct bio *bio,
 
        spin_unlock_irq(q->queue_lock);
 
-       rq = blk_alloc_request(q, rw, bio, priv, gfp_mask);
+       rq = blk_alloc_request(q, rw_flags, priv, gfp_mask);
        if (unlikely(!rq)) {
                /*
                 * Allocation failed presumably due to memory. Undo anything
@@ -2107,7 +2173,8 @@ rq_starved:
                ioc->nr_batch_requests--;
        
        rq_init(q, rq);
-       rq->rl = rl;
+
+       blk_add_trace_generic(q, bio, rw, BLK_TA_GETRQ);
 out:
        return rq;
 }
@@ -2118,12 +2185,13 @@ out:
  *
  * Called with q->queue_lock held, and returns with it unlocked.
  */
-static struct request *get_request_wait(request_queue_t *q, int rw,
+static struct request *get_request_wait(struct request_queue *q, int rw_flags,
                                        struct bio *bio)
 {
+       const int rw = rw_flags & 0x01;
        struct request *rq;
 
-       rq = get_request(q, rw, bio, GFP_NOIO);
+       rq = get_request(q, rw_flags, bio, GFP_NOIO);
        while (!rq) {
                DEFINE_WAIT(wait);
                struct request_list *rl = &q->rq;
@@ -2131,11 +2199,13 @@ static struct request *get_request_wait(request_queue_t *q, int rw,
                prepare_to_wait_exclusive(&rl->wait[rw], &wait,
                                TASK_UNINTERRUPTIBLE);
 
-               rq = get_request(q, rw, bio, GFP_NOIO);
+               rq = get_request(q, rw_flags, bio, GFP_NOIO);
 
                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();
@@ -2146,7 +2216,7 @@ static struct request *get_request_wait(request_queue_t *q, int rw,
                         * up to a big batch of them for a small period time.
                         * See ioc_batching, ioc_set_batching
                         */
-                       ioc = current_io_context(GFP_NOIO);
+                       ioc = current_io_context(GFP_NOIO, q->node);
                        ioc_set_batching(q, ioc);
 
                        spin_lock_irq(q->queue_lock);
@@ -2157,7 +2227,7 @@ static struct request *get_request_wait(request_queue_t *q, int rw,
        return rq;
 }
 
-struct request *blk_get_request(request_queue_t *q, int rw, gfp_t gfp_mask)
+struct request *blk_get_request(struct request_queue *q, int rw, gfp_t gfp_mask)
 {
        struct request *rq;
 
@@ -2178,6 +2248,25 @@ struct request *blk_get_request(request_queue_t *q, int rw, gfp_t gfp_mask)
 EXPORT_SYMBOL(blk_get_request);
 
 /**
+ * blk_start_queueing - initiate dispatch of requests to device
+ * @q:         request queue to kick into gear
+ *
+ * This is basically a helper to remove the need to know whether a queue
+ * is plugged or not if someone just wants to initiate dispatch of requests
+ * for this queue.
+ *
+ * The queue lock must be held with interrupts disabled.
+ */
+void blk_start_queueing(struct request_queue *q)
+{
+       if (!blk_queue_plugged(q))
+               q->request_fn(q);
+       else
+               __generic_unplug_device(q);
+}
+EXPORT_SYMBOL(blk_start_queueing);
+
+/**
  * blk_requeue_request - put a request back on queue
  * @q:         request queue where request should be inserted
  * @rq:                request to be inserted
@@ -2187,8 +2276,10 @@ EXPORT_SYMBOL(blk_get_request);
  *    more, when that condition happens we need to put the request back
  *    on the queue. Must be called with queue lock held.
  */
-void blk_requeue_request(request_queue_t *q, struct request *rq)
+void blk_requeue_request(struct request_queue *q, struct request *rq)
 {
+       blk_add_trace_rq(q, rq, BLK_TA_REQUEUE);
+
        if (blk_rq_tagged(rq))
                blk_queue_end_tag(q, rq);
 
@@ -2216,7 +2307,7 @@ EXPORT_SYMBOL(blk_requeue_request);
  *    of the queue for things like a QUEUE_FULL message from a device, or a
  *    host that is unable to accept a particular command.
  */
-void blk_insert_request(request_queue_t *q, struct request *rq,
+void blk_insert_request(struct request_queue *q, struct request *rq,
                        int at_head, void *data)
 {
        int where = at_head ? ELEVATOR_INSERT_FRONT : ELEVATOR_INSERT_BACK;
@@ -2227,7 +2318,8 @@ void blk_insert_request(request_queue_t *q, struct request *rq,
         * must not attempt merges on this) and that it acts as a soft
         * barrier
         */
-       rq->flags |= REQ_SPECIAL | REQ_SOFTBARRIER;
+       rq->cmd_type = REQ_TYPE_SPECIAL;
+       rq->cmd_flags |= REQ_SOFTBARRIER;
 
        rq->special = data;
 
@@ -2239,18 +2331,87 @@ void blk_insert_request(request_queue_t *q, struct request *rq,
        if (blk_rq_tagged(rq))
                blk_queue_end_tag(q, rq);
 
-       drive_stat_acct(rq, rq->nr_sectors, 1);
+       drive_stat_acct(rq, 1);
        __elv_add_request(q, rq, where, 0);
-
-       if (blk_queue_plugged(q))
-               __generic_unplug_device(q);
-       else
-               q->request_fn(q);
+       blk_start_queueing(q);
        spin_unlock_irqrestore(q->queue_lock, flags);
 }
 
 EXPORT_SYMBOL(blk_insert_request);
 
+static int __blk_rq_unmap_user(struct bio *bio)
+{
+       int ret = 0;
+
+       if (bio) {
+               if (bio_flagged(bio, BIO_USER_MAPPED))
+                       bio_unmap_user(bio);
+               else
+                       ret = bio_uncopy_user(bio);
+       }
+
+       return ret;
+}
+
+int blk_rq_append_bio(struct request_queue *q, struct request *rq,
+                     struct bio *bio)
+{
+       if (!rq->bio)
+               blk_rq_bio_prep(q, rq, bio);
+       else if (!ll_back_merge_fn(q, rq, bio))
+               return -EINVAL;
+       else {
+               rq->biotail->bi_next = bio;
+               rq->biotail = bio;
+
+               rq->data_len += bio->bi_size;
+       }
+       return 0;
+}
+EXPORT_SYMBOL(blk_rq_append_bio);
+
+static int __blk_rq_map_user(struct request_queue *q, struct request *rq,
+                            void __user *ubuf, unsigned int len)
+{
+       unsigned long uaddr;
+       struct bio *bio, *orig_bio;
+       int reading, ret;
+
+       reading = rq_data_dir(rq) == READ;
+
+       /*
+        * if alignment requirement is satisfied, map in user pages for
+        * direct dma. else, set up kernel bounce buffers
+        */
+       uaddr = (unsigned long) ubuf;
+       if (!(uaddr & queue_dma_alignment(q)) && !(len & queue_dma_alignment(q)))
+               bio = bio_map_user(q, NULL, uaddr, len, reading);
+       else
+               bio = bio_copy_user(q, uaddr, len, reading);
+
+       if (IS_ERR(bio))
+               return PTR_ERR(bio);
+
+       orig_bio = bio;
+       blk_queue_bounce(q, &bio);
+
+       /*
+        * We link the bounce buffer in and could have to traverse it
+        * later so we have to get a ref to prevent it from being freed
+        */
+       bio_get(bio);
+
+       ret = blk_rq_append_bio(q, rq, bio);
+       if (!ret)
+               return bio->bi_size;
+
+       /* if it was boucned we must call the end io function */
+       bio_endio(bio, 0);
+       __blk_rq_unmap_user(orig_bio);
+       bio_put(bio);
+       return ret;
+}
+
 /**
  * blk_rq_map_user - map user data to a request, for REQ_BLOCK_PC usage
  * @q:         request queue where request should be inserted
@@ -2271,43 +2432,48 @@ EXPORT_SYMBOL(blk_insert_request);
  *    original bio must be passed back in to blk_rq_unmap_user() for proper
  *    unmapping.
  */
-int blk_rq_map_user(request_queue_t *q, struct request *rq, void __user *ubuf,
-                   unsigned int len)
+int blk_rq_map_user(struct request_queue *q, struct request *rq,
+                   void __user *ubuf, unsigned long len)
 {
-       unsigned long uaddr;
-       struct bio *bio;
-       int reading;
+       unsigned long bytes_read = 0;
+       struct bio *bio = NULL;
+       int ret;
 
        if (len > (q->max_hw_sectors << 9))
                return -EINVAL;
        if (!len || !ubuf)
                return -EINVAL;
 
-       reading = rq_data_dir(rq) == READ;
-
-       /*
-        * if alignment requirement is satisfied, map in user pages for
-        * direct dma. else, set up kernel bounce buffers
-        */
-       uaddr = (unsigned long) ubuf;
-       if (!(uaddr & queue_dma_alignment(q)) && !(len & queue_dma_alignment(q)))
-               bio = bio_map_user(q, NULL, uaddr, len, reading);
-       else
-               bio = bio_copy_user(q, uaddr, len, reading);
+       while (bytes_read != len) {
+               unsigned long map_len, end, start;
 
-       if (!IS_ERR(bio)) {
-               rq->bio = rq->biotail = bio;
-               blk_rq_bio_prep(q, rq, bio);
+               map_len = min_t(unsigned long, len - bytes_read, BIO_MAX_SIZE);
+               end = ((unsigned long)ubuf + map_len + PAGE_SIZE - 1)
+                                                               >> PAGE_SHIFT;
+               start = (unsigned long)ubuf >> PAGE_SHIFT;
 
-               rq->buffer = rq->data = NULL;
-               rq->data_len = len;
-               return 0;
+               /*
+                * A bad offset could cause us to require BIO_MAX_PAGES + 1
+                * pages. If this happens we just lower the requested
+                * mapping len by a page so that we can fit
+                */
+               if (end - start > BIO_MAX_PAGES)
+                       map_len -= PAGE_SIZE;
+
+               ret = __blk_rq_map_user(q, rq, ubuf, map_len);
+               if (ret < 0)
+                       goto unmap_rq;
+               if (!bio)
+                       bio = rq->bio;
+               bytes_read += ret;
+               ubuf += ret;
        }
 
-       /*
-        * bio is the err-ptr
-        */
-       return PTR_ERR(bio);
+       rq->buffer = rq->data = NULL;
+       return 0;
+unmap_rq:
+       blk_rq_unmap_user(bio);
+       return ret;
 }
 
 EXPORT_SYMBOL(blk_rq_map_user);
@@ -2318,6 +2484,7 @@ EXPORT_SYMBOL(blk_rq_map_user);
  * @rq:                request to map data to
  * @iov:       pointer to the iovec
  * @iov_count: number of elements in the iovec
+ * @len:       I/O byte count
  *
  * Description:
  *    Data will be mapped directly for zero copy io, if possible. Otherwise
@@ -2332,8 +2499,8 @@ EXPORT_SYMBOL(blk_rq_map_user);
  *    original bio must be passed back in to blk_rq_unmap_user() for proper
  *    unmapping.
  */
-int blk_rq_map_user_iov(request_queue_t *q, struct request *rq,
-                       struct sg_iovec *iov, int iov_count)
+int blk_rq_map_user_iov(struct request_queue *q, struct request *rq,
+                       struct sg_iovec *iov, int iov_count, unsigned int len)
 {
        struct bio *bio;
 
@@ -2347,10 +2514,15 @@ int blk_rq_map_user_iov(request_queue_t *q, struct request *rq,
        if (IS_ERR(bio))
                return PTR_ERR(bio);
 
-       rq->bio = rq->biotail = bio;
+       if (bio->bi_size != len) {
+               bio_endio(bio, 0);
+               bio_unmap_user(bio);
+               return -EINVAL;
+       }
+
+       bio_get(bio);
        blk_rq_bio_prep(q, rq, bio);
        rq->buffer = rq->data = NULL;
-       rq->data_len = bio->bi_size;
        return 0;
 }
 
@@ -2358,24 +2530,33 @@ EXPORT_SYMBOL(blk_rq_map_user_iov);
 
 /**
  * blk_rq_unmap_user - unmap a request with user data
- * @bio:       bio to be unmapped
- * @ulen:      length of user buffer
+ * @bio:              start of bio list
  *
  * Description:
- *    Unmap a bio previously mapped by blk_rq_map_user().
+ *    Unmap a rq previously mapped by blk_rq_map_user(). The caller must
+ *    supply the original rq->bio from the blk_rq_map_user() return, since
+ *    the io completion may have changed rq->bio.
  */
-int blk_rq_unmap_user(struct bio *bio, unsigned int ulen)
+int blk_rq_unmap_user(struct bio *bio)
 {
-       int ret = 0;
+       struct bio *mapped_bio;
+       int ret = 0, ret2;
 
-       if (bio) {
-               if (bio_flagged(bio, BIO_USER_MAPPED))
-                       bio_unmap_user(bio);
-               else
-                       ret = bio_uncopy_user(bio);
+       while (bio) {
+               mapped_bio = bio;
+               if (unlikely(bio_flagged(bio, BIO_BOUNCED)))
+                       mapped_bio = bio->bi_private;
+
+               ret2 = __blk_rq_unmap_user(mapped_bio);
+               if (ret2 && !ret)
+                       ret = ret2;
+
+               mapped_bio = bio;
+               bio = bio->bi_next;
+               bio_put(mapped_bio);
        }
 
-       return 0;
+       return ret;
 }
 
 EXPORT_SYMBOL(blk_rq_unmap_user);
@@ -2388,7 +2569,7 @@ EXPORT_SYMBOL(blk_rq_unmap_user);
  * @len:       length of user data
  * @gfp_mask:  memory allocation flags
  */
-int blk_rq_map_kern(request_queue_t *q, struct request *rq, void *kbuf,
+int blk_rq_map_kern(struct request_queue *q, struct request *rq, void *kbuf,
                    unsigned int len, gfp_t gfp_mask)
 {
        struct bio *bio;
@@ -2405,11 +2586,9 @@ int blk_rq_map_kern(request_queue_t *q, struct request *rq, void *kbuf,
        if (rq_data_dir(rq) == WRITE)
                bio->bi_rw |= (1 << BIO_RW);
 
-       rq->bio = rq->biotail = bio;
        blk_rq_bio_prep(q, rq, bio);
-
+       blk_queue_bounce(q, &rq->bio);
        rq->buffer = rq->data = NULL;
-       rq->data_len = len;
        return 0;
 }
 
@@ -2427,19 +2606,21 @@ EXPORT_SYMBOL(blk_rq_map_kern);
  *    Insert a fully prepared request at the back of the io scheduler queue
  *    for execution.  Don't wait for completion.
  */
-void blk_execute_rq_nowait(request_queue_t *q, struct gendisk *bd_disk,
+void blk_execute_rq_nowait(struct request_queue *q, struct gendisk *bd_disk,
                           struct request *rq, int at_head,
                           rq_end_io_fn *done)
 {
        int where = at_head ? ELEVATOR_INSERT_FRONT : ELEVATOR_INSERT_BACK;
 
        rq->rq_disk = bd_disk;
-       rq->flags |= REQ_NOMERGE;
+       rq->cmd_flags |= REQ_NOMERGE;
        rq->end_io = done;
-       elv_add_request(q, rq, where, 1);
-       generic_unplug_device(q);
+       WARN_ON(irqs_disabled());
+       spin_lock_irq(q->queue_lock);
+       __elv_add_request(q, rq, where, 1);
+       __generic_unplug_device(q);
+       spin_unlock_irq(q->queue_lock);
 }
-
 EXPORT_SYMBOL_GPL(blk_execute_rq_nowait);
 
 /**
@@ -2453,10 +2634,10 @@ EXPORT_SYMBOL_GPL(blk_execute_rq_nowait);
  *    Insert a fully prepared request at the back of the io scheduler queue
  *    for execution and wait for completion.
  */
-int blk_execute_rq(request_queue_t *q, struct gendisk *bd_disk,
+int blk_execute_rq(struct request_queue *q, struct gendisk *bd_disk,
                   struct request *rq, int at_head)
 {
-       DECLARE_COMPLETION(wait);
+       DECLARE_COMPLETION_ONSTACK(wait);
        char sense[SCSI_SENSE_BUFFERSIZE];
        int err = 0;
 
@@ -2472,10 +2653,9 @@ int blk_execute_rq(request_queue_t *q, struct gendisk *bd_disk,
                rq->sense_len = 0;
        }
 
-       rq->waiting = &wait;
+       rq->end_io_data = &wait;
        blk_execute_rq_nowait(q, bd_disk, rq, at_head, blk_end_sync_rq);
        wait_for_completion(&wait);
-       rq->waiting = NULL;
 
        if (rq->errors)
                err = -EIO;
@@ -2485,6 +2665,14 @@ int blk_execute_rq(request_queue_t *q, struct gendisk *bd_disk,
 
 EXPORT_SYMBOL(blk_execute_rq);
 
+static void bio_end_empty_barrier(struct bio *bio, int err)
+{
+       if (err)
+               clear_bit(BIO_UPTODATE, &bio->bi_flags);
+
+       complete(bio->bi_private);
+}
+
 /**
  * blkdev_issue_flush - queue a flush
  * @bdev:      blockdev to issue flush for
@@ -2497,7 +2685,10 @@ EXPORT_SYMBOL(blk_execute_rq);
  */
 int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector)
 {
-       request_queue_t *q;
+       DECLARE_COMPLETION_ONSTACK(wait);
+       struct request_queue *q;
+       struct bio *bio;
+       int ret;
 
        if (bdev->bd_disk == NULL)
                return -ENXIO;
@@ -2505,15 +2696,37 @@ int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector)
        q = bdev_get_queue(bdev);
        if (!q)
                return -ENXIO;
-       if (!q->issue_flush_fn)
-               return -EOPNOTSUPP;
 
-       return q->issue_flush_fn(q, bdev->bd_disk, error_sector);
+       bio = bio_alloc(GFP_KERNEL, 0);
+       if (!bio)
+               return -ENOMEM;
+
+       bio->bi_end_io = bio_end_empty_barrier;
+       bio->bi_private = &wait;
+       bio->bi_bdev = bdev;
+       submit_bio(1 << BIO_RW_BARRIER, bio);
+
+       wait_for_completion(&wait);
+
+       /*
+        * The driver must store the error location in ->bi_sector, if
+        * it supports it. For non-stacked drivers, this should be copied
+        * from rq->sector.
+        */
+       if (error_sector)
+               *error_sector = bio->bi_sector;
+
+       ret = 0;
+       if (!bio_flagged(bio, BIO_UPTODATE))
+               ret = -EIO;
+
+       bio_put(bio);
+       return ret;
 }
 
 EXPORT_SYMBOL(blkdev_issue_flush);
 
-static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io)
+static void drive_stat_acct(struct request *rq, int new_io)
 {
        int rw = rq_data_dir(rq);
 
@@ -2533,12 +2746,9 @@ static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io)
  * queue lock is held and interrupts disabled, as we muck with the
  * request queue list.
  */
-static inline void add_request(request_queue_t * q, struct request * req)
+static inline void add_request(struct request_queue * q, struct request * req)
 {
-       drive_stat_acct(req, req->nr_sectors, 1);
-
-       if (q->activity_fn)
-               q->activity_fn(q->activity_data, rq_data_dir(req));
+       drive_stat_acct(req, 1);
 
        /*
         * elevator indicated where it wants this request to be
@@ -2577,13 +2787,13 @@ void disk_round_stats(struct gendisk *disk)
        disk->stamp = now;
 }
 
+EXPORT_SYMBOL_GPL(disk_round_stats);
+
 /*
  * queue lock must be held
  */
-void __blk_put_request(request_queue_t *q, struct request *req)
+void __blk_put_request(struct request_queue *q, struct request *req)
 {
-       struct request_list *rl = req->rl;
-
        if (unlikely(!q))
                return;
        if (unlikely(--req->ref_count))
@@ -2591,18 +2801,16 @@ void __blk_put_request(request_queue_t *q, struct request *req)
 
        elv_completed_request(q, req);
 
-       req->rq_status = RQ_INACTIVE;
-       req->rl = NULL;
-
        /*
         * Request may not have originated from ll_rw_blk. if not,
         * it didn't come out of our reserved rq pools
         */
-       if (rl) {
+       if (req->cmd_flags & REQ_ALLOCED) {
                int rw = rq_data_dir(req);
-               int priv = req->flags & REQ_ELVPRIV;
+               int priv = req->cmd_flags & REQ_ELVPRIV;
 
                BUG_ON(!list_empty(&req->queuelist));
+               BUG_ON(!hlist_unhashed(&req->hash));
 
                blk_free_request(q, req);
                freed_request(q, rw, priv);
@@ -2614,7 +2822,7 @@ EXPORT_SYMBOL_GPL(__blk_put_request);
 void blk_put_request(struct request *req)
 {
        unsigned long flags;
-       request_queue_t *q = req->q;
+       struct request_queue *q = req->q;
 
        /*
         * Gee, IDE calls in w/ NULL q.  Fix IDE and remove the
@@ -2632,12 +2840,13 @@ EXPORT_SYMBOL(blk_put_request);
 /**
  * blk_end_sync_rq - executes a completion event on a request
  * @rq: request to complete
+ * @error: end io status of the request
  */
 void blk_end_sync_rq(struct request *rq, int error)
 {
-       struct completion *waiting = rq->waiting;
+       struct completion *waiting = rq->end_io_data;
 
-       rq->waiting = NULL;
+       rq->end_io_data = NULL;
        __blk_put_request(rq->q, rq);
 
        /*
@@ -2648,47 +2857,24 @@ void blk_end_sync_rq(struct request *rq, int error)
 }
 EXPORT_SYMBOL(blk_end_sync_rq);
 
-/**
- * blk_congestion_wait - wait for a queue to become uncongested
- * @rw: READ or WRITE
- * @timeout: timeout in jiffies
- *
- * Waits for up to @timeout jiffies for a queue (any queue) to exit congestion.
- * If no queues are congested then just wait for the next request to be
- * returned.
- */
-long blk_congestion_wait(int rw, long timeout)
-{
-       long ret;
-       DEFINE_WAIT(wait);
-       wait_queue_head_t *wqh = &congestion_wqh[rw];
-
-       prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
-       ret = io_schedule_timeout(timeout);
-       finish_wait(wqh, &wait);
-       return ret;
-}
-
-EXPORT_SYMBOL(blk_congestion_wait);
-
 /*
  * Has to be called with the request spinlock acquired
  */
-static int attempt_merge(request_queue_t *q, struct request *req,
+static int attempt_merge(struct request_queue *q, struct request *req,
                          struct request *next)
 {
        if (!rq_mergeable(req) || !rq_mergeable(next))
                return 0;
 
        /*
-        * not contigious
+        * not contiguous
         */
        if (req->sector + req->nr_sectors != next->sector)
                return 0;
 
        if (rq_data_dir(req) != rq_data_dir(next)
            || req->rq_disk != next->rq_disk
-           || next->waiting || next->special)
+           || next->special)
                return 0;
 
        /*
@@ -2697,7 +2883,7 @@ static int attempt_merge(request_queue_t *q, struct request *req,
         * will have updated segment counts, update sector
         * counts here.
         */
-       if (!q->merge_requests_fn(q, req, next))
+       if (!ll_merge_requests_fn(q, req, next))
                return 0;
 
        /*
@@ -2727,7 +2913,8 @@ static int attempt_merge(request_queue_t *q, struct request *req,
        return 1;
 }
 
-static inline int attempt_back_merge(request_queue_t *q, struct request *rq)
+static inline int attempt_back_merge(struct request_queue *q,
+                                    struct request *rq)
 {
        struct request *next = elv_latter_request(q, rq);
 
@@ -2737,7 +2924,8 @@ static inline int attempt_back_merge(request_queue_t *q, struct request *rq)
        return 0;
 }
 
-static inline int attempt_front_merge(request_queue_t *q, struct request *rq)
+static inline int attempt_front_merge(struct request_queue *q,
+                                     struct request *rq)
 {
        struct request *prev = elv_former_request(q, rq);
 
@@ -2749,48 +2937,41 @@ static inline int attempt_front_merge(request_queue_t *q, struct request *rq)
 
 static void init_request_from_bio(struct request *req, struct bio *bio)
 {
-       req->flags |= REQ_CMD;
+       req->cmd_type = REQ_TYPE_FS;
 
        /*
         * inherit FAILFAST from bio (for read-ahead, and explicit FAILFAST)
         */
        if (bio_rw_ahead(bio) || bio_failfast(bio))
-               req->flags |= REQ_FAILFAST;
+               req->cmd_flags |= REQ_FAILFAST;
 
        /*
         * REQ_BARRIER implies no merging, but lets make it explicit
         */
        if (unlikely(bio_barrier(bio)))
-               req->flags |= (REQ_HARDBARRIER | REQ_NOMERGE);
+               req->cmd_flags |= (REQ_HARDBARRIER | REQ_NOMERGE);
+
+       if (bio_sync(bio))
+               req->cmd_flags |= REQ_RW_SYNC;
+       if (bio_rw_meta(bio))
+               req->cmd_flags |= REQ_RW_META;
 
        req->errors = 0;
        req->hard_sector = req->sector = bio->bi_sector;
-       req->hard_nr_sectors = req->nr_sectors = bio_sectors(bio);
-       req->current_nr_sectors = req->hard_cur_sectors = bio_cur_sectors(bio);
-       req->nr_phys_segments = bio_phys_segments(req->q, bio);
-       req->nr_hw_segments = bio_hw_segments(req->q, bio);
-       req->buffer = bio_data(bio);    /* see ->buffer comment above */
-       req->waiting = NULL;
-       req->bio = req->biotail = bio;
        req->ioprio = bio_prio(bio);
-       req->rq_disk = bio->bi_bdev->bd_disk;
        req->start_time = jiffies;
+       blk_rq_bio_prep(req->q, req, bio);
 }
 
-static int __make_request(request_queue_t *q, struct bio *bio)
+static int __make_request(struct request_queue *q, struct bio *bio)
 {
        struct request *req;
-       int el_ret, rw, nr_sectors, cur_nr_sectors, barrier, err, sync;
-       unsigned short prio;
-       sector_t sector;
+       int el_ret, nr_sectors, barrier, err;
+       const unsigned short prio = bio_prio(bio);
+       const int sync = bio_sync(bio);
+       int rw_flags;
 
-       sector = bio->bi_sector;
        nr_sectors = bio_sectors(bio);
-       cur_nr_sectors = bio_cur_sectors(bio);
-       prio = bio_prio(bio);
-
-       rw = bio_data_dir(bio);
-       sync = bio_sync(bio);
 
        /*
         * low level driver can indicate that it wants pages above a
@@ -2799,8 +2980,6 @@ static int __make_request(request_queue_t *q, struct bio *bio)
         */
        blk_queue_bounce(q, &bio);
 
-       spin_lock_prefetch(q->queue_lock);
-
        barrier = bio_barrier(bio);
        if (unlikely(barrier) && (q->next_ordered == QUEUE_ORDERED_NONE)) {
                err = -EOPNOTSUPP;
@@ -2817,24 +2996,28 @@ static int __make_request(request_queue_t *q, struct bio *bio)
                case ELEVATOR_BACK_MERGE:
                        BUG_ON(!rq_mergeable(req));
 
-                       if (!q->back_merge_fn(q, req, bio))
+                       if (!ll_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;
                        req->ioprio = ioprio_best(req->ioprio, prio);
-                       drive_stat_acct(req, nr_sectors, 0);
+                       drive_stat_acct(req, 0);
                        if (!attempt_back_merge(q, req))
-                               elv_merged_request(q, req);
+                               elv_merged_request(q, req, el_ret);
                        goto out;
 
                case ELEVATOR_FRONT_MERGE:
                        BUG_ON(!rq_mergeable(req));
 
-                       if (!q->front_merge_fn(q, req, bio))
+                       if (!ll_front_merge_fn(q, req, bio))
                                break;
 
+                       blk_add_trace_bio(q, bio, BLK_TA_FRONTMERGE);
+
                        bio->bi_next = req->bio;
                        req->bio = bio;
 
@@ -2844,14 +3027,14 @@ static int __make_request(request_queue_t *q, struct bio *bio)
                         * not touch req->buffer either...
                         */
                        req->buffer = bio_data(bio);
-                       req->current_nr_sectors = cur_nr_sectors;
-                       req->hard_cur_sectors = cur_nr_sectors;
-                       req->sector = req->hard_sector = sector;
+                       req->current_nr_sectors = bio_cur_sectors(bio);
+                       req->hard_cur_sectors = req->current_nr_sectors;
+                       req->sector = req->hard_sector = bio->bi_sector;
                        req->nr_sectors = req->hard_nr_sectors += nr_sectors;
                        req->ioprio = ioprio_best(req->ioprio, prio);
-                       drive_stat_acct(req, nr_sectors, 0);
+                       drive_stat_acct(req, 0);
                        if (!attempt_front_merge(q, req))
-                               elv_merged_request(q, req);
+                               elv_merged_request(q, req, el_ret);
                        goto out;
 
                /* ELV_NO_MERGE: elevator says don't/can't merge. */
@@ -2861,10 +3044,19 @@ static int __make_request(request_queue_t *q, struct bio *bio)
 
 get_rq:
        /*
+        * This sync check and mask will be re-done in init_request_from_bio(),
+        * but we need to set it earlier to expose the sync flag to the
+        * rq allocator and io schedulers.
+        */
+       rw_flags = bio_data_dir(bio);
+       if (sync)
+               rw_flags |= REQ_RW_SYNC;
+
+       /*
         * Grab a free request. This is might sleep but can not fail.
         * Returns with the queue unlocked.
         */
-       req = get_request_wait(q, rw, bio);
+       req = get_request_wait(q, rw_flags, bio);
 
        /*
         * After dropping the lock and possibly sleeping here, our request
@@ -2886,7 +3078,7 @@ out:
        return 0;
 
 end_io:
-       bio_endio(bio, nr_sectors << 9, err);
+       bio_endio(bio, err);
        return 0;
 }
 
@@ -2897,7 +3089,7 @@ static inline void blk_partition_remap(struct bio *bio)
 {
        struct block_device *bdev = bio->bi_bdev;
 
-       if (bdev != bdev->bd_contains) {
+       if (bio_sectors(bio) && bdev != bdev->bd_contains) {
                struct hd_struct *p = bdev->bd_part;
                const int rw = bio_data_dir(bio);
 
@@ -2906,6 +3098,10 @@ static inline void blk_partition_remap(struct bio *bio)
 
                bio->bi_sector += p->start_sect;
                bio->bi_bdev = bdev->bd_contains;
+
+               blk_add_trace_remap(bdev_get_queue(bio->bi_bdev), bio,
+                                   bdev->bd_dev, bio->bi_sector,
+                                   bio->bi_sector - p->start_sect);
        }
 }
 
@@ -2923,6 +3119,71 @@ static void handle_bad_sector(struct bio *bio)
        set_bit(BIO_EOF, &bio->bi_flags);
 }
 
+#ifdef CONFIG_FAIL_MAKE_REQUEST
+
+static DECLARE_FAULT_ATTR(fail_make_request);
+
+static int __init setup_fail_make_request(char *str)
+{
+       return setup_fault_attr(&fail_make_request, str);
+}
+__setup("fail_make_request=", setup_fail_make_request);
+
+static int should_fail_request(struct bio *bio)
+{
+       if ((bio->bi_bdev->bd_disk->flags & GENHD_FL_FAIL) ||
+           (bio->bi_bdev->bd_part && bio->bi_bdev->bd_part->make_it_fail))
+               return should_fail(&fail_make_request, bio->bi_size);
+
+       return 0;
+}
+
+static int __init fail_make_request_debugfs(void)
+{
+       return init_fault_attr_dentries(&fail_make_request,
+                                       "fail_make_request");
+}
+
+late_initcall(fail_make_request_debugfs);
+
+#else /* CONFIG_FAIL_MAKE_REQUEST */
+
+static inline int should_fail_request(struct bio *bio)
+{
+       return 0;
+}
+
+#endif /* CONFIG_FAIL_MAKE_REQUEST */
+
+/*
+ * Check whether this bio extends beyond the end of the device.
+ */
+static inline int bio_check_eod(struct bio *bio, unsigned int nr_sectors)
+{
+       sector_t maxsector;
+
+       if (!nr_sectors)
+               return 0;
+
+       /* Test device or partition size, when known. */
+       maxsector = bio->bi_bdev->bd_inode->i_size >> 9;
+       if (maxsector) {
+               sector_t sector = bio->bi_sector;
+
+               if (maxsector < nr_sectors || maxsector - nr_sectors < sector) {
+                       /*
+                        * This may well happen - the kernel calls bread()
+                        * without checking the size of the device, e.g., when
+                        * mounting a device.
+                        */
+                       handle_bad_sector(bio);
+                       return 1;
+               }
+       }
+
+       return 0;
+}
+
 /**
  * generic_make_request: hand a buffer to its device driver for I/O
  * @bio:  The bio describing the location in memory and on the device.
@@ -2947,28 +3208,18 @@ static void handle_bad_sector(struct bio *bio)
  * bi_sector for remaps as it sees fit.  So the values of these fields
  * should NOT be depended on after the call to generic_make_request.
  */
-void generic_make_request(struct bio *bio)
+static inline void __generic_make_request(struct bio *bio)
 {
-       request_queue_t *q;
-       sector_t maxsector;
+       struct request_queue *q;
+       sector_t old_sector;
        int ret, nr_sectors = bio_sectors(bio);
+       dev_t old_dev;
+       int err = -EIO;
 
        might_sleep();
-       /* Test device or partition size, when known. */
-       maxsector = bio->bi_bdev->bd_inode->i_size >> 9;
-       if (maxsector) {
-               sector_t sector = bio->bi_sector;
 
-               if (maxsector < nr_sectors || maxsector - nr_sectors < sector) {
-                       /*
-                        * This may well happen - the kernel calls bread()
-                        * without checking the size of the device, e.g., when
-                        * mounting a device.
-                        */
-                       handle_bad_sector(bio);
-                       goto end_io;
-               }
-       }
+       if (bio_check_eod(bio, nr_sectors))
+               goto end_io;
 
        /*
         * Resolve the mapping until finished. (drivers are
@@ -2978,6 +3229,8 @@ 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.
         */
+       old_sector = -1;
+       old_dev = 0;
        do {
                char b[BDEVNAME_SIZE];
 
@@ -2989,11 +3242,11 @@ void generic_make_request(struct bio *bio)
                                bdevname(bio->bi_bdev, b),
                                (long long) bio->bi_sector);
 end_io:
-                       bio_endio(bio, bio->bi_size, -EIO);
+                       bio_endio(bio, err);
                        break;
                }
 
-               if (unlikely(bio_sectors(bio) > q->max_hw_sectors)) {
+               if (unlikely(nr_sectors > q->max_hw_sectors)) {
                        printk("bio too big device %s (%u > %u)\n", 
                                bdevname(bio->bi_bdev, b),
                                bio_sectors(bio),
@@ -3004,16 +3257,86 @@ end_io:
                if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)))
                        goto end_io;
 
+               if (should_fail_request(bio))
+                       goto end_io;
+
                /*
                 * If this device has partitions, remap block n
                 * of partition p to block n+start(p) of the disk.
                 */
                blk_partition_remap(bio);
 
+               if (old_sector != -1)
+                       blk_add_trace_remap(q, bio, old_dev, bio->bi_sector,
+                                           old_sector);
+
+               blk_add_trace_bio(q, bio, BLK_TA_QUEUE);
+
+               old_sector = bio->bi_sector;
+               old_dev = bio->bi_bdev->bd_dev;
+
+               if (bio_check_eod(bio, nr_sectors))
+                       goto end_io;
+               if (bio_empty_barrier(bio) && !q->prepare_flush_fn) {
+                       err = -EOPNOTSUPP;
+                       goto end_io;
+               }
+
                ret = q->make_request_fn(q, bio);
        } while (ret);
 }
 
+/*
+ * We only want one ->make_request_fn to be active at a time,
+ * else stack usage with stacked devices could be a problem.
+ * So use current->bio_{list,tail} to keep a list of requests
+ * submited by a make_request_fn function.
+ * current->bio_tail is also used as a flag to say if
+ * generic_make_request is currently active in this task or not.
+ * If it is NULL, then no make_request is active.  If it is non-NULL,
+ * then a make_request is active, and new requests should be added
+ * at the tail
+ */
+void generic_make_request(struct bio *bio)
+{
+       if (current->bio_tail) {
+               /* make_request is active */
+               *(current->bio_tail) = bio;
+               bio->bi_next = NULL;
+               current->bio_tail = &bio->bi_next;
+               return;
+       }
+       /* following loop may be a bit non-obvious, and so deserves some
+        * explanation.
+        * Before entering the loop, bio->bi_next is NULL (as all callers
+        * ensure that) so we have a list with a single bio.
+        * We pretend that we have just taken it off a longer list, so
+        * we assign bio_list to the next (which is NULL) and bio_tail
+        * to &bio_list, thus initialising the bio_list of new bios to be
+        * added.  __generic_make_request may indeed add some more bios
+        * through a recursive call to generic_make_request.  If it
+        * did, we find a non-NULL value in bio_list and re-enter the loop
+        * from the top.  In this case we really did just take the bio
+        * of the top of the list (no pretending) and so fixup bio_list and
+        * bio_tail or bi_next, and call into __generic_make_request again.
+        *
+        * The loop was structured like this to make only one call to
+        * __generic_make_request (which is important as it is large and
+        * inlined) and to keep the structure simple.
+        */
+       BUG_ON(bio->bi_next);
+       do {
+               current->bio_list = bio->bi_next;
+               if (bio->bi_next == NULL)
+                       current->bio_tail = &current->bio_list;
+               else
+                       bio->bi_next = NULL;
+               __generic_make_request(bio);
+               bio = current->bio_list;
+       } while (bio);
+       current->bio_tail = NULL; /* deactivate */
+}
+
 EXPORT_SYMBOL(generic_make_request);
 
 /**
@@ -3030,70 +3353,39 @@ void submit_bio(int rw, struct bio *bio)
 {
        int count = bio_sectors(bio);
 
-       BIO_BUG_ON(!bio->bi_size);
-       BIO_BUG_ON(!bio->bi_io_vec);
        bio->bi_rw |= rw;
-       if (rw & WRITE)
-               mod_page_state(pgpgout, count);
-       else
-               mod_page_state(pgpgin, count);
-
-       if (unlikely(block_dump)) {
-               char b[BDEVNAME_SIZE];
-               printk(KERN_DEBUG "%s(%d): %s block %Lu on %s\n",
-                       current->comm, current->pid,
-                       (rw & WRITE) ? "WRITE" : "READ",
-                       (unsigned long long)bio->bi_sector,
-                       bdevname(bio->bi_bdev,b));
-       }
 
-       generic_make_request(bio);
-}
-
-EXPORT_SYMBOL(submit_bio);
+       /*
+        * If it's a regular read/write or a barrier with data attached,
+        * go through the normal accounting stuff before submission.
+        */
+       if (!bio_empty_barrier(bio)) {
 
-static void blk_recalc_rq_segments(struct request *rq)
-{
-       struct bio *bio, *prevbio = NULL;
-       int nr_phys_segs, nr_hw_segs;
-       unsigned int phys_size, hw_size;
-       request_queue_t *q = rq->q;
+               BIO_BUG_ON(!bio->bi_size);
+               BIO_BUG_ON(!bio->bi_io_vec);
 
-       if (!rq->bio)
-               return;
+               if (rw & WRITE) {
+                       count_vm_events(PGPGOUT, count);
+               } else {
+                       task_io_account_read(bio->bi_size);
+                       count_vm_events(PGPGIN, count);
+               }
 
-       phys_size = hw_size = nr_phys_segs = nr_hw_segs = 0;
-       rq_for_each_bio(bio, rq) {
-               /* Force bio hw/phys segs to be recalculated. */
-               bio->bi_flags &= ~(1 << BIO_SEG_VALID);
-
-               nr_phys_segs += bio_phys_segments(q, bio);
-               nr_hw_segs += bio_hw_segments(q, bio);
-               if (prevbio) {
-                       int pseg = phys_size + prevbio->bi_size + bio->bi_size;
-                       int hseg = hw_size + prevbio->bi_size + bio->bi_size;
-
-                       if (blk_phys_contig_segment(q, prevbio, bio) &&
-                           pseg <= q->max_segment_size) {
-                               nr_phys_segs--;
-                               phys_size += prevbio->bi_size + bio->bi_size;
-                       } else
-                               phys_size = 0;
-
-                       if (blk_hw_contig_segment(q, prevbio, bio) &&
-                           hseg <= q->max_segment_size) {
-                               nr_hw_segs--;
-                               hw_size += prevbio->bi_size + bio->bi_size;
-                       } else
-                               hw_size = 0;
+               if (unlikely(block_dump)) {
+                       char b[BDEVNAME_SIZE];
+                       printk(KERN_DEBUG "%s(%d): %s block %Lu on %s\n",
+                       current->comm, task_pid_nr(current),
+                               (rw & WRITE) ? "WRITE" : "READ",
+                               (unsigned long long)bio->bi_sector,
+                               bdevname(bio->bi_bdev,b));
                }
-               prevbio = bio;
        }
 
-       rq->nr_phys_segments = nr_phys_segs;
-       rq->nr_hw_segments = nr_hw_segs;
+       generic_make_request(bio);
 }
 
+EXPORT_SYMBOL(submit_bio);
+
 static void blk_recalc_rq_sectors(struct request *rq, int nsect)
 {
        if (blk_fs_request(rq)) {
@@ -3129,6 +3421,8 @@ 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
         */
@@ -3144,7 +3438,7 @@ static int __end_that_request_first(struct request *req, int uptodate,
                req->errors = 0;
 
        if (!uptodate) {
-               if (blk_fs_request(req) && !(req->flags & REQ_QUIET))
+               if (blk_fs_request(req) && !(req->cmd_flags & REQ_QUIET))
                        printk("end_request: I/O error, dev %s, sector %llu\n",
                                req->rq_disk ? req->rq_disk->disk_name : "?",
                                (unsigned long long)req->sector);
@@ -3153,18 +3447,25 @@ static int __end_that_request_first(struct request *req, int uptodate,
        if (blk_fs_request(req) && req->rq_disk) {
                const int rw = rq_data_dir(req);
 
-               __disk_stat_add(req->rq_disk, sectors[rw], nr_bytes >> 9);
+               disk_stat_add(req->rq_disk, sectors[rw], nr_bytes >> 9);
        }
 
        total_bytes = bio_nbytes = 0;
        while ((bio = req->bio) != NULL) {
                int nbytes;
 
+               /*
+                * For an empty barrier request, the low level driver must
+                * store a potential error location in ->sector. We pass
+                * that back up in ->bi_sector.
+                */
+               if (blk_empty_barrier(req))
+                       bio->bi_sector = req->sector;
+
                if (nr_bytes >= bio->bi_size) {
                        req->bio = bio->bi_next;
                        nbytes = bio->bi_size;
-                       if (!ordered_bio_endio(req, bio, nbytes, error))
-                               bio_endio(bio, nbytes, error);
+                       req_bio_endio(req, bio, nbytes, error);
                        next_idx = 0;
                        bio_nbytes = 0;
                } else {
@@ -3219,8 +3520,7 @@ static int __end_that_request_first(struct request *req, int uptodate,
         * if the request wasn't completed, update state
         */
        if (bio_nbytes) {
-               if (!ordered_bio_endio(req, bio, bio_nbytes, error))
-                       bio_endio(bio, bio_nbytes, error);
+               req_bio_endio(req, bio, bio_nbytes, error);
                bio->bi_idx += next_idx;
                bio_iovec(bio)->bv_offset += nr_bytes;
                bio_iovec(bio)->bv_len -= nr_bytes;
@@ -3280,12 +3580,11 @@ EXPORT_SYMBOL(end_that_request_chunk);
  */
 static void blk_done_softirq(struct softirq_action *h)
 {
-       struct list_head *cpu_list;
-       LIST_HEAD(local_list);
+       struct list_head *cpu_list, local_list;
 
        local_irq_disable();
        cpu_list = &__get_cpu_var(blk_cpu_done);
-       list_splice_init(cpu_list, &local_list);
+       list_replace_init(cpu_list, &local_list);
        local_irq_enable();
 
        while (!list_empty(&local_list)) {
@@ -3296,16 +3595,14 @@ static void blk_done_softirq(struct softirq_action *h)
        }
 }
 
-#ifdef CONFIG_HOTPLUG_CPU
-
-static int blk_cpu_notify(struct notifier_block *self, unsigned long action,
+static int __cpuinit blk_cpu_notify(struct notifier_block *self, unsigned long action,
                          void *hcpu)
 {
        /*
         * If a CPU goes away, splice its entries to the current CPU
         * and trigger a run of the softirq
         */
-       if (action == CPU_DEAD) {
+       if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
                int cpu = (unsigned long) hcpu;
 
                local_irq_disable();
@@ -3319,20 +3616,18 @@ static int blk_cpu_notify(struct notifier_block *self, unsigned long action,
 }
 
 
-static struct notifier_block __devinitdata blk_cpu_notifier = {
+static struct notifier_block blk_cpu_notifier __cpuinitdata = {
        .notifier_call  = blk_cpu_notify,
 };
 
-#endif /* CONFIG_HOTPLUG_CPU */
-
 /**
  * blk_complete_request - end I/O on a request
  * @req:      the request being processed
  *
  * Description:
  *     Ends all I/O on a request. It does not handle partial completions,
- *     unless the driver actually implements this in its completionc callback
- *     through requeueing. Theh actual completion happens out-of-order,
+ *     unless the driver actually implements this in its completion callback
+ *     through requeueing. The actual completion happens out-of-order,
  *     through a softirq handler. The user must have registered a completion
  *     callback through blk_queue_softirq_done().
  **/
@@ -3373,7 +3668,12 @@ void end_that_request_last(struct request *req, int uptodate)
        if (unlikely(laptop_mode) && blk_fs_request(req))
                laptop_io_completion();
 
-       if (disk && blk_fs_request(req)) {
+       /*
+        * Account IO completion.  bar_rq isn't accounted as a normal
+        * IO on queueing nor completion.  Accounting the containing
+        * request is enough.
+        */
+       if (disk && blk_fs_request(req) && req != &req->q->bar_rq) {
                unsigned long duration = jiffies - req->start_time;
                const int rw = rq_data_dir(req);
 
@@ -3390,21 +3690,90 @@ void end_that_request_last(struct request *req, int uptodate)
 
 EXPORT_SYMBOL(end_that_request_last);
 
-void end_request(struct request *req, int uptodate)
+static inline void __end_request(struct request *rq, int uptodate,
+                                unsigned int nr_bytes, int dequeue)
 {
-       if (!end_that_request_first(req, uptodate, req->hard_cur_sectors)) {
-               add_disk_randomness(req->rq_disk);
-               blkdev_dequeue_request(req);
-               end_that_request_last(req, uptodate);
+       if (!end_that_request_chunk(rq, uptodate, nr_bytes)) {
+               if (dequeue)
+                       blkdev_dequeue_request(rq);
+               add_disk_randomness(rq->rq_disk);
+               end_that_request_last(rq, uptodate);
        }
 }
 
+static unsigned int rq_byte_size(struct request *rq)
+{
+       if (blk_fs_request(rq))
+               return rq->hard_nr_sectors << 9;
+
+       return rq->data_len;
+}
+
+/**
+ * end_queued_request - end all I/O on a queued request
+ * @rq:                the request being processed
+ * @uptodate:  error value or 0/1 uptodate flag
+ *
+ * Description:
+ *     Ends all I/O on a request, and removes it from the block layer queues.
+ *     Not suitable for normal IO completion, unless the driver still has
+ *     the request attached to the block layer.
+ *
+ **/
+void end_queued_request(struct request *rq, int uptodate)
+{
+       __end_request(rq, uptodate, rq_byte_size(rq), 1);
+}
+EXPORT_SYMBOL(end_queued_request);
+
+/**
+ * end_dequeued_request - end all I/O on a dequeued request
+ * @rq:                the request being processed
+ * @uptodate:  error value or 0/1 uptodate flag
+ *
+ * Description:
+ *     Ends all I/O on a request. The request must already have been
+ *     dequeued using blkdev_dequeue_request(), as is normally the case
+ *     for most drivers.
+ *
+ **/
+void end_dequeued_request(struct request *rq, int uptodate)
+{
+       __end_request(rq, uptodate, rq_byte_size(rq), 0);
+}
+EXPORT_SYMBOL(end_dequeued_request);
+
+
+/**
+ * end_request - end I/O on the current segment of the request
+ * @req:       the request being processed
+ * @uptodate:  error value or 0/1 uptodate flag
+ *
+ * Description:
+ *     Ends I/O on the current segment of a request. If that is the only
+ *     remaining segment, the request is also completed and freed.
+ *
+ *     This is a remnant of how older block drivers handled IO completions.
+ *     Modern drivers typically end IO on the full request in one go, unless
+ *     they have a residual value to account for. For that case this function
+ *     isn't really useful, unless the residual just happens to be the
+ *     full current segment. In other words, don't use this function in new
+ *     code. Either use end_request_completely(), or the
+ *     end_that_request_chunk() (along with end_that_request_last()) for
+ *     partial completions.
+ *
+ **/
+void end_request(struct request *req, int uptodate)
+{
+       __end_request(req, uptodate, req->hard_cur_sectors << 9, 1);
+}
 EXPORT_SYMBOL(end_request);
 
-void blk_rq_bio_prep(request_queue_t *q, struct request *rq, struct bio *bio)
+static void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
+                           struct bio *bio)
 {
-       /* first three bits are identical in rq->flags and bio->bi_rw */
-       rq->flags |= (bio->bi_rw & 7);
+       /* first two bits are identical in rq->cmd_flags and bio->bi_rw */
+       rq->cmd_flags |= (bio->bi_rw & 3);
 
        rq->nr_phys_segments = bio_phys_segments(q, bio);
        rq->nr_hw_segments = bio_hw_segments(q, bio);
@@ -3412,11 +3781,13 @@ void blk_rq_bio_prep(request_queue_t *q, struct request *rq, struct bio *bio)
        rq->hard_cur_sectors = rq->current_nr_sectors;
        rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio);
        rq->buffer = bio_data(bio);
+       rq->data_len = bio->bi_size;
 
        rq->bio = rq->biotail = bio;
-}
 
-EXPORT_SYMBOL(blk_rq_bio_prep);
+       if (bio->bi_bdev)
+               rq->rq_disk = bio->bi_bdev->bd_disk;
+}
 
 int kblockd_schedule_work(struct work_struct *work)
 {
@@ -3425,11 +3796,11 @@ int kblockd_schedule_work(struct work_struct *work)
 
 EXPORT_SYMBOL(kblockd_schedule_work);
 
-void kblockd_flush(void)
+void kblockd_flush_work(struct work_struct *work)
 {
-       flush_workqueue(kblockd_workqueue);
+       cancel_work_sync(work);
 }
-EXPORT_SYMBOL(kblockd_flush);
+EXPORT_SYMBOL(kblockd_flush_work);
 
 int __init blk_dev_init(void)
 {
@@ -3440,24 +3811,22 @@ int __init blk_dev_init(void)
                panic("Failed to create kblockd\n");
 
        request_cachep = kmem_cache_create("blkdev_requests",
-                       sizeof(struct request), 0, SLAB_PANIC, NULL, NULL);
+                       sizeof(struct request), 0, SLAB_PANIC, NULL);
 
        requestq_cachep = kmem_cache_create("blkdev_queue",
-                       sizeof(request_queue_t), 0, SLAB_PANIC, NULL, NULL);
+                       sizeof(struct request_queue), 0, SLAB_PANIC, NULL);
 
        iocontext_cachep = kmem_cache_create("blkdev_ioc",
-                       sizeof(struct io_context), 0, SLAB_PANIC, NULL, NULL);
+                       sizeof(struct io_context), 0, SLAB_PANIC, NULL);
 
-       for (i = 0; i < NR_CPUS; i++)
+       for_each_possible_cpu(i)
                INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i));
 
        open_softirq(BLOCK_SOFTIRQ, blk_done_softirq, NULL);
-#ifdef CONFIG_HOTPLUG_CPU
-       register_cpu_notifier(&blk_cpu_notifier);
-#endif
+       register_hotcpu_notifier(&blk_cpu_notifier);
 
-       blk_max_low_pfn = max_low_pfn;
-       blk_max_pfn = max_pfn;
+       blk_max_low_pfn = max_low_pfn - 1;
+       blk_max_pfn = max_pfn - 1;
 
        return 0;
 }
@@ -3473,10 +3842,18 @@ void put_io_context(struct io_context *ioc)
        BUG_ON(atomic_read(&ioc->refcount) == 0);
 
        if (atomic_dec_and_test(&ioc->refcount)) {
+               struct cfq_io_context *cic;
+
+               rcu_read_lock();
                if (ioc->aic && ioc->aic->dtor)
                        ioc->aic->dtor(ioc->aic);
-               if (ioc->cic && ioc->cic->dtor)
-                       ioc->cic->dtor(ioc->cic);
+               if (ioc->cic_root.rb_node != NULL) {
+                       struct rb_node *n = rb_first(&ioc->cic_root);
+
+                       cic = rb_entry(n, struct cfq_io_context, rb_node);
+                       cic->dtor(ioc);
+               }
+               rcu_read_unlock();
 
                kmem_cache_free(iocontext_cachep, ioc);
        }
@@ -3486,21 +3863,21 @@ EXPORT_SYMBOL(put_io_context);
 /* Called by the exitting task */
 void exit_io_context(void)
 {
-       unsigned long flags;
        struct io_context *ioc;
+       struct cfq_io_context *cic;
 
-       local_irq_save(flags);
        task_lock(current);
        ioc = current->io_context;
        current->io_context = NULL;
-       ioc->task = NULL;
        task_unlock(current);
-       local_irq_restore(flags);
 
+       ioc->task = NULL;
        if (ioc->aic && ioc->aic->exit)
                ioc->aic->exit(ioc->aic);
-       if (ioc->cic && ioc->cic->exit)
-               ioc->cic->exit(ioc->cic);
+       if (ioc->cic_root.rb_node != NULL) {
+               cic = rb_entry(rb_first(&ioc->cic_root), struct cfq_io_context, rb_node);
+               cic->exit(ioc);
+       }
 
        put_io_context(ioc);
 }
@@ -3513,7 +3890,7 @@ void exit_io_context(void)
  * but since the current task itself holds a reference, the context can be
  * used in general code, so long as it stays within `current` context.
  */
-struct io_context *current_io_context(gfp_t gfp_flags)
+static struct io_context *current_io_context(gfp_t gfp_flags, int node)
 {
        struct task_struct *tsk = current;
        struct io_context *ret;
@@ -3522,21 +3899,23 @@ struct io_context *current_io_context(gfp_t gfp_flags)
        if (likely(ret))
                return ret;
 
-       ret = kmem_cache_alloc(iocontext_cachep, gfp_flags);
+       ret = kmem_cache_alloc_node(iocontext_cachep, gfp_flags, node);
        if (ret) {
                atomic_set(&ret->refcount, 1);
                ret->task = current;
-               ret->set_ioprio = NULL;
+               ret->ioprio_changed = 0;
                ret->last_waited = jiffies; /* doesn't matter... */
                ret->nr_batch_requests = 0; /* because this is 0 */
                ret->aic = NULL;
-               ret->cic = NULL;
+               ret->cic_root.rb_node = NULL;
+               ret->ioc_data = NULL;
+               /* make sure set_task_ioprio() sees the settings above */
+               smp_wmb();
                tsk->io_context = ret;
        }
 
        return ret;
 }
-EXPORT_SYMBOL(current_io_context);
 
 /*
  * If the current task has no IO context then create one and initialise it.
@@ -3544,10 +3923,10 @@ EXPORT_SYMBOL(current_io_context);
  *
  * This is always called in the context of the task which submitted the I/O.
  */
-struct io_context *get_io_context(gfp_t gfp_flags)
+struct io_context *get_io_context(gfp_t gfp_flags, int node)
 {
        struct io_context *ret;
-       ret = current_io_context(gfp_flags);
+       ret = current_io_context(gfp_flags, node);
        if (likely(ret))
                atomic_inc(&ret->refcount);
        return ret;
@@ -3610,21 +3989,24 @@ static ssize_t
 queue_requests_store(struct request_queue *q, const char *page, size_t count)
 {
        struct request_list *rl = &q->rq;
+       unsigned long nr;
+       int ret = queue_var_store(&nr, page, count);
+       if (nr < BLKDEV_MIN_RQ)
+               nr = BLKDEV_MIN_RQ;
 
-       int ret = queue_var_store(&q->nr_requests, page, count);
-       if (q->nr_requests < BLKDEV_MIN_RQ)
-               q->nr_requests = BLKDEV_MIN_RQ;
+       spin_lock_irq(q->queue_lock);
+       q->nr_requests = nr;
        blk_queue_congestion_threshold(q);
 
        if (rl->count[READ] >= queue_congestion_on_threshold(q))
-               set_queue_congested(q, READ);
+               blk_set_queue_congested(q, READ);
        else if (rl->count[READ] < queue_congestion_off_threshold(q))
-               clear_queue_congested(q, READ);
+               blk_clear_queue_congested(q, READ);
 
        if (rl->count[WRITE] >= queue_congestion_on_threshold(q))
-               set_queue_congested(q, WRITE);
+               blk_set_queue_congested(q, WRITE);
        else if (rl->count[WRITE] < queue_congestion_off_threshold(q))
-               clear_queue_congested(q, WRITE);
+               blk_clear_queue_congested(q, WRITE);
 
        if (rl->count[READ] >= q->nr_requests) {
                blk_set_queue_full(q, READ);
@@ -3639,6 +4021,7 @@ queue_requests_store(struct request_queue *q, const char *page, size_t count)
                blk_clear_queue_full(q, WRITE);
                wake_up(&rl->wait[WRITE]);
        }
+       spin_unlock_irq(q->queue_lock);
        return ret;
 }
 
@@ -3656,9 +4039,6 @@ queue_ra_store(struct request_queue *q, const char *page, size_t count)
        ssize_t ret = queue_var_store(&ra_kb, page, count);
 
        spin_lock_irq(q->queue_lock);
-       if (ra_kb > (q->max_sectors >> 1))
-               ra_kb = (q->max_sectors >> 1);
-
        q->backing_dev_info.ra_pages = ra_kb >> (PAGE_CACHE_SHIFT - 10);
        spin_unlock_irq(q->queue_lock);
 
@@ -3679,7 +4059,6 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count)
                        max_hw_sectors_kb = q->max_hw_sectors >> 1,
                        page_kb = 1 << (PAGE_CACHE_SHIFT - 10);
        ssize_t ret = queue_var_store(&max_sectors_kb, page, count);
-       int ra_kb;
 
        if (max_sectors_kb > max_hw_sectors_kb || max_sectors_kb < page_kb)
                return -EINVAL;
@@ -3688,14 +4067,6 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count)
         * values synchronously:
         */
        spin_lock_irq(q->queue_lock);
-       /*
-        * Trim readahead window as well, if necessary:
-        */
-       ra_kb = q->backing_dev_info.ra_pages << (PAGE_CACHE_SHIFT - 10);
-       if (ra_kb > max_sectors_kb)
-               q->backing_dev_info.ra_pages =
-                               max_sectors_kb >> (PAGE_CACHE_SHIFT - 10);
-
        q->max_sectors = max_sectors_kb << 1;
        spin_unlock_irq(q->queue_lock);
 
@@ -3754,13 +4125,20 @@ static ssize_t
 queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 {
        struct queue_sysfs_entry *entry = to_queue(attr);
-       struct request_queue *q;
+       struct request_queue *q =
+               container_of(kobj, struct request_queue, kobj);
+       ssize_t res;
 
-       q = container_of(kobj, struct request_queue, kobj);
        if (!entry->show)
                return -EIO;
-
-       return entry->show(q, page);
+       mutex_lock(&q->sysfs_lock);
+       if (test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)) {
+               mutex_unlock(&q->sysfs_lock);
+               return -ENOENT;
+       }
+       res = entry->show(q, page);
+       mutex_unlock(&q->sysfs_lock);
+       return res;
 }
 
 static ssize_t
@@ -3768,13 +4146,20 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr,
                    const char *page, size_t length)
 {
        struct queue_sysfs_entry *entry = to_queue(attr);
-       struct request_queue *q;
+       struct request_queue *q = container_of(kobj, struct request_queue, kobj);
+
+       ssize_t res;
 
-       q = container_of(kobj, struct request_queue, kobj);
        if (!entry->store)
                return -EIO;
-
-       return entry->store(q, page, length);
+       mutex_lock(&q->sysfs_lock);
+       if (test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)) {
+               mutex_unlock(&q->sysfs_lock);
+               return -ENOENT;
+       }
+       res = entry->store(q, page, length);
+       mutex_unlock(&q->sysfs_lock);
+       return res;
 }
 
 static struct sysfs_ops queue_sysfs_ops = {
@@ -3785,31 +4170,30 @@ static struct sysfs_ops queue_sysfs_ops = {
 static struct kobj_type queue_ktype = {
        .sysfs_ops      = &queue_sysfs_ops,
        .default_attrs  = default_attrs,
+       .release        = blk_release_queue,
 };
 
 int blk_register_queue(struct gendisk *disk)
 {
        int ret;
 
-       request_queue_t *q = disk->queue;
+       struct request_queue *q = disk->queue;
 
        if (!q || !q->request_fn)
                return -ENXIO;
 
-       q->kobj.parent = kobject_get(&disk->kobj);
-       if (!q->kobj.parent)
-               return -EBUSY;
-
-       snprintf(q->kobj.name, KOBJ_NAME_LEN, "%s", "queue");
-       q->kobj.ktype = &queue_ktype;
+       q->kobj.parent = kobject_get(&disk->dev.kobj);
 
-       ret = kobject_register(&q->kobj);
+       ret = kobject_add(&q->kobj);
        if (ret < 0)
                return ret;
 
+       kobject_uevent(&q->kobj, KOBJ_ADD);
+
        ret = elv_register_queue(q);
        if (ret) {
-               kobject_unregister(&q->kobj);
+               kobject_uevent(&q->kobj, KOBJ_REMOVE);
+               kobject_del(&q->kobj);
                return ret;
        }
 
@@ -3818,12 +4202,13 @@ int blk_register_queue(struct gendisk *disk)
 
 void blk_unregister_queue(struct gendisk *disk)
 {
-       request_queue_t *q = disk->queue;
+       struct request_queue *q = disk->queue;
 
        if (q && q->request_fn) {
                elv_unregister_queue(q);
 
-               kobject_unregister(&q->kobj);
-               kobject_put(&disk->kobj);
+               kobject_uevent(&q->kobj, KOBJ_REMOVE);
+               kobject_del(&q->kobj);
+               kobject_put(&disk->dev.kobj);
        }
 }