18 files changed, 296 insertions, 214 deletions

diff --git a/block/elevator.c b/block/elevator.c
index 39dcccc82a..99a4d7b2f8 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -64,7 +64,7 @@ inline int elv_rq_merge_ok(struct request *rq, struct bio *bio)
 }
 EXPORT_SYMBOL(elv_rq_merge_ok);
 
-inline int elv_try_merge(struct request *__rq, struct bio *bio)
+static inline int elv_try_merge(struct request *__rq, struct bio *bio)
 {
         int ret = ELEVATOR_NO_MERGE;
 
@@ -80,7 +80,6 @@ inline int elv_try_merge(struct request *__rq, struct bio *bio)
 
         return ret;
 }
-EXPORT_SYMBOL(elv_try_merge);
 
 static struct elevator_type *elevator_find(const char *name)
 {
diff --git a/block/ll_rw_blk.c b/block/ll_rw_blk.c
index 91d3b4828c..8e27d0ab0d 100644
--- a/block/ll_rw_blk.c
+++ b/block/ll_rw_blk.c
@@ -26,7 +26,8 @@
 #include <linux/slab.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
-#include <linux/blkdev.h>
+#include <linux/interrupt.h>
+#include <linux/cpu.h>
 
 /*
  * for max sense size
@@ -62,13 +63,15 @@ static wait_queue_head_t congestion_wqh[2] = {
 /*
  * Controlling structure to kblockd
  */
-static struct workqueue_struct *kblockd_workqueue; 
+static struct workqueue_struct *kblockd_workqueue;
 
 unsigned long blk_max_low_pfn, blk_max_pfn;
 
 EXPORT_SYMBOL(blk_max_low_pfn);
 EXPORT_SYMBOL(blk_max_pfn);
 
+static DEFINE_PER_CPU(struct list_head, blk_cpu_done);
+
 /* Amount of time in which a process may batch requests */
 #define BLK_BATCH_TIME  (HZ/50UL)
 
@@ -207,6 +210,13 @@ void blk_queue_merge_bvec(request_queue_t *q, merge_bvec_fn *mbfn)
 
 EXPORT_SYMBOL(blk_queue_merge_bvec);
 
+void blk_queue_softirq_done(request_queue_t *q, softirq_done_fn *fn)
+{
+        q->softirq_done_fn = fn;
+}
+
+EXPORT_SYMBOL(blk_queue_softirq_done);
+
 /**
  * blk_queue_make_request - define an alternate make_request function for a device
  * @q:  the request queue for the device to be affected
@@ -270,6 +280,7 @@ EXPORT_SYMBOL(blk_queue_make_request);
 static inline void rq_init(request_queue_t *q, struct request *rq)
 {
         INIT_LIST_HEAD(&rq->queuelist);
+        INIT_LIST_HEAD(&rq->donelist);
 
         rq->errors = 0;
         rq->rq_status = RQ_ACTIVE;
@@ -286,6 +297,7 @@ static inline void rq_init(request_queue_t *q, struct request *rq)
         rq->sense = NULL;
         rq->end_io = NULL;
         rq->end_io_data = NULL;
+        rq->completion_data = NULL;
 }
 
 /**
@@ -2735,30 +2747,6 @@ static inline int attempt_front_merge(request_queue_t *q, struct request *rq)
         return 0;
 }
 
-/**
- * blk_attempt_remerge  - attempt to remerge active head with next request
- * @q:    The &request_queue_t belonging to the device
- * @rq:   The head request (usually)
- *
- * Description:
- *    For head-active devices, the queue can easily be unplugged so quickly
- *    that proper merging is not done on the front request. This may hurt
- *    performance greatly for some devices. The block layer cannot safely
- *    do merging on that first request for these queues, but the driver can
- *    call this function and make it happen any way. Only the driver knows
- *    when it is safe to do so.
- **/
-void blk_attempt_remerge(request_queue_t *q, struct request *rq)
-{
-        unsigned long flags;
-
-        spin_lock_irqsave(q->queue_lock, flags);
-        attempt_back_merge(q, rq);
-        spin_unlock_irqrestore(q->queue_lock, flags);
-}
-
-EXPORT_SYMBOL(blk_attempt_remerge);
-
 static void init_request_from_bio(struct request *req, struct bio *bio)
 {
         req->flags |= REQ_CMD;
@@ -3287,6 +3275,87 @@ int end_that_request_chunk(struct request *req, int uptodate, int nr_bytes)
 EXPORT_SYMBOL(end_that_request_chunk);
 
 /*
+ * splice the completion data to a local structure and hand off to
+ * process_completion_queue() to complete the requests
+ */
+static void blk_done_softirq(struct softirq_action *h)
+{
+        struct list_head *cpu_list;
+        LIST_HEAD(local_list);
+
+        local_irq_disable();
+        cpu_list = &__get_cpu_var(blk_cpu_done);
+        list_splice_init(cpu_list, &local_list);
+        local_irq_enable();
+
+        while (!list_empty(&local_list)) {
+                struct request *rq = list_entry(local_list.next, struct request, donelist);
+
+                list_del_init(&rq->donelist);
+                rq->q->softirq_done_fn(rq);
+        }
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+
+static int 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) {
+                int cpu = (unsigned long) hcpu;
+
+                local_irq_disable();
+                list_splice_init(&per_cpu(blk_cpu_done, cpu),
+                                 &__get_cpu_var(blk_cpu_done));
+                raise_softirq_irqoff(BLOCK_SOFTIRQ);
+                local_irq_enable();
+        }
+
+        return NOTIFY_OK;
+}
+
+
+static struct notifier_block __devinitdata blk_cpu_notifier = {
+        .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,
+ *     through a softirq handler. The user must have registered a completion
+ *     callback through blk_queue_softirq_done().
+ **/
+
+void blk_complete_request(struct request *req)
+{
+        struct list_head *cpu_list;
+        unsigned long flags;
+
+        BUG_ON(!req->q->softirq_done_fn);
+                
+        local_irq_save(flags);
+
+        cpu_list = &__get_cpu_var(blk_cpu_done);
+        list_add_tail(&req->donelist, cpu_list);
+        raise_softirq_irqoff(BLOCK_SOFTIRQ);
+
+        local_irq_restore(flags);
+}
+
+EXPORT_SYMBOL(blk_complete_request);
+        
+/*
  * queue lock must be held
  */
 void end_that_request_last(struct request *req, int uptodate)
@@ -3364,6 +3433,8 @@ EXPORT_SYMBOL(kblockd_flush);
 
 int __init blk_dev_init(void)
 {
+        int i;
+
         kblockd_workqueue = create_workqueue("kblockd");
         if (!kblockd_workqueue)
                 panic("Failed to create kblockd\n");
@@ -3377,6 +3448,14 @@ int __init blk_dev_init(void)
         iocontext_cachep = kmem_cache_create("blkdev_ioc",
                         sizeof(struct io_context), 0, SLAB_PANIC, NULL, NULL);
 
+        for (i = 0; i < NR_CPUS; 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
+
         blk_max_low_pfn = max_low_pfn;
         blk_max_pfn = max_pfn;
 
diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index c2ac36dfe4..18de84c8cc 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -190,16 +190,21 @@ static int verify_command(struct file *file, unsigned char *cmd)
                 safe_for_write(GPCMD_SET_STREAMING),
         };
         unsigned char type = cmd_type[cmd[0]];
+        int has_write_perm = 0;
 
         /* Anybody who can open the device can do a read-safe command */
         if (type & CMD_READ_SAFE)
                 return 0;
 
+        /*
+         * file can be NULL from ioctl_by_bdev()...
+         */
+        if (file)
+                has_write_perm = file->f_mode & FMODE_WRITE;
+
         /* Write-safe commands just require a writable open.. */
-        if (type & CMD_WRITE_SAFE) {
-                if (file->f_mode & FMODE_WRITE)
-                        return 0;
-        }
+        if ((type & CMD_WRITE_SAFE) && has_write_perm)
+                return 0;
 
         /* And root can do any command.. */
         if (capable(CAP_SYS_RAWIO))
diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
index 88452c79fb..e4e9f255bd 100644
--- a/drivers/block/cciss.c
+++ b/drivers/block/cciss.c
@@ -2178,16 +2178,48 @@ static inline void resend_cciss_cmd( ctlr_info_t *h, CommandList_struct *c)
 
         start_io(h);
 }
+
+static void cciss_softirq_done(struct request *rq)
+{
+        CommandList_struct *cmd = rq->completion_data;
+        ctlr_info_t *h = hba[cmd->ctlr];
+        u64bit temp64;
+        int i, ddir;
+
+        if (cmd->Request.Type.Direction == XFER_READ)
+                ddir = PCI_DMA_FROMDEVICE;
+        else
+                ddir = PCI_DMA_TODEVICE;
+
+        /* command did not need to be retried */
+        /* unmap the DMA mapping for all the scatter gather elements */
+        for(i=0; i<cmd->Header.SGList; i++) {
+                temp64.val32.lower = cmd->SG[i].Addr.lower;
+                temp64.val32.upper = cmd->SG[i].Addr.upper;
+                pci_unmap_page(h->pdev, temp64.val, cmd->SG[i].Len, ddir);
+        }
+
+        complete_buffers(rq->bio, rq->errors);
+
+#ifdef CCISS_DEBUG
+        printk("Done with %p\n", rq);
+#endif /* CCISS_DEBUG */ 
+
+        spin_lock_irq(&h->lock);
+        end_that_request_last(rq, rq->errors);
+        cmd_free(h, cmd,1);
+        spin_unlock_irq(&h->lock);
+}
+
 /* checks the status of the job and calls complete buffers to mark all 
- * buffers for the completed job. 
+ * buffers for the completed job. Note that this function does not need
+ * to hold the hba/queue lock.
  */ 
 static inline void complete_command( ctlr_info_t *h, CommandList_struct *cmd,
                 int timeout)
 {
         int status = 1;
-        int i;
         int retry_cmd = 0;
-        u64bit temp64;
                 
         if (timeout)
                 status = 0; 
@@ -2295,24 +2327,10 @@ static inline void complete_command( ctlr_info_t *h, CommandList_struct *cmd,
                 resend_cciss_cmd(h,cmd);
                 return;
         }       
-        /* command did not need to be retried */
-        /* unmap the DMA mapping for all the scatter gather elements */
-        for(i=0; i<cmd->Header.SGList; i++) {
-                temp64.val32.lower = cmd->SG[i].Addr.lower;
-                temp64.val32.upper = cmd->SG[i].Addr.upper;
-                pci_unmap_page(hba[cmd->ctlr]->pdev,
-                        temp64.val, cmd->SG[i].Len,
-                        (cmd->Request.Type.Direction == XFER_READ) ?
-                                PCI_DMA_FROMDEVICE : PCI_DMA_TODEVICE);
-        }
-        complete_buffers(cmd->rq->bio, status);
-
-#ifdef CCISS_DEBUG
-        printk("Done with %p\n", cmd->rq);
-#endif /* CCISS_DEBUG */ 
 
-        end_that_request_last(cmd->rq, status ? 1 : -EIO);
-        cmd_free(h,cmd,1);
+        cmd->rq->completion_data = cmd;
+        cmd->rq->errors = status;
+        blk_complete_request(cmd->rq);
 }
 
 /* 
@@ -3199,15 +3217,17 @@ static int __devinit cciss_init_one(struct pci_dev *pdev,
                 drv->queue = q;
 
                 q->backing_dev_info.ra_pages = READ_AHEAD;
-        blk_queue_bounce_limit(q, hba[i]->pdev->dma_mask);
+                blk_queue_bounce_limit(q, hba[i]->pdev->dma_mask);
+
+                /* This is a hardware imposed limit. */
+                blk_queue_max_hw_segments(q, MAXSGENTRIES);
 
-        /* This is a hardware imposed limit. */
-        blk_queue_max_hw_segments(q, MAXSGENTRIES);
+                /* This is a limit in the driver and could be eliminated. */
+                blk_queue_max_phys_segments(q, MAXSGENTRIES);
 
-        /* This is a limit in the driver and could be eliminated. */
-        blk_queue_max_phys_segments(q, MAXSGENTRIES);
+                blk_queue_max_sectors(q, 512);
 
-        blk_queue_max_sectors(q, 512);
+                blk_queue_softirq_done(q, cciss_softirq_done);
 
                 q->queuedata = hba[i];
                 sprintf(disk->disk_name, "cciss/c%dd%d", i, j);
diff --git a/drivers/ide/ide-cd.c b/drivers/ide/ide-cd.c
index d31117eb95..e4d55ad32d 100644
--- a/drivers/ide/ide-cd.c
+++ b/drivers/ide/ide-cd.c
@@ -1332,8 +1332,6 @@ static ide_startstop_t cdrom_start_read (ide_drive_t *drive, unsigned int block)
         if (cdrom_read_from_buffer(drive))
                 return ide_stopped;
 
-        blk_attempt_remerge(drive->queue, rq);
-
         /* Clear the local sector buffer. */
         info->nsectors_buffered = 0;
 
@@ -1874,14 +1872,6 @@ static ide_startstop_t cdrom_start_write(ide_drive_t *drive, struct request *rq)
                 return ide_stopped;
         }
 
-        /*
-         * for dvd-ram and such media, it's a really big deal to get
-         * big writes all the time. so scour the queue and attempt to
-         * remerge requests, often the plugging will not have had time
-         * to do this properly
-         */
-        blk_attempt_remerge(drive->queue, rq);
-
         info->nsectors_buffered = 0;
 
         /* use dma, if possible. we don't need to check more, since we
diff --git a/drivers/ide/ide-io.c b/drivers/ide/ide-io.c
index b5dc6df8e6..dea2d4dcc6 100644
--- a/drivers/ide/ide-io.c
+++ b/drivers/ide/ide-io.c
@@ -55,9 +55,22 @@
 #include <asm/io.h>
 #include <asm/bitops.h>
 
+void ide_softirq_done(struct request *rq)
+{
+        request_queue_t *q = rq->q;
+
+        add_disk_randomness(rq->rq_disk);
+        end_that_request_chunk(rq, rq->errors, rq->data_len);
+
+        spin_lock_irq(q->queue_lock);
+        end_that_request_last(rq, rq->errors);
+        spin_unlock_irq(q->queue_lock);
+}
+
 int __ide_end_request(ide_drive_t *drive, struct request *rq, int uptodate,
                       int nr_sectors)
 {
+        unsigned int nbytes;
         int ret = 1;
 
         BUG_ON(!(rq->flags & REQ_STARTED));
@@ -81,17 +94,28 @@ int __ide_end_request(ide_drive_t *drive, struct request *rq, int uptodate,
                 HWGROUP(drive)->hwif->ide_dma_on(drive);
         }
 
-        if (!end_that_request_first(rq, uptodate, nr_sectors)) {
-                add_disk_randomness(rq->rq_disk);
-
-                if (blk_rq_tagged(rq))
-                        blk_queue_end_tag(drive->queue, rq);
-
+        /*
+         * For partial completions (or non fs/pc requests), use the regular
+         * direct completion path.
+         */
+        nbytes = nr_sectors << 9;
+        if (rq_all_done(rq, nbytes)) {
+                rq->errors = uptodate;
+                rq->data_len = nbytes;
                 blkdev_dequeue_request(rq);
                 HWGROUP(drive)->rq = NULL;
-                end_that_request_last(rq, uptodate);
+                blk_complete_request(rq);
                 ret = 0;
+        } else {
+                if (!end_that_request_first(rq, uptodate, nr_sectors)) {
+                        add_disk_randomness(rq->rq_disk);
+                        blkdev_dequeue_request(rq);
+                        HWGROUP(drive)->rq = NULL;
+                        end_that_request_last(rq, uptodate);
+                        ret = 0;
+                }
         }
+
         return ret;
 }
 EXPORT_SYMBOL(__ide_end_request);
@@ -113,6 +137,10 @@ int ide_end_request (ide_drive_t *drive, int uptodate, int nr_sectors)
         unsigned long flags;
         int ret = 1;
 
+        /*
+         * room for locking improvements here, the calls below don't
+         * need the queue lock held at all
+         */
         spin_lock_irqsave(&ide_lock, flags);
         rq = HWGROUP(drive)->rq;
 
diff --git a/drivers/ide/ide-probe.c b/drivers/ide/ide-probe.c
index 02167a5b75..1ddaa71a8f 100644
--- a/drivers/ide/ide-probe.c
+++ b/drivers/ide/ide-probe.c
@@ -1011,6 +1011,8 @@ static int ide_init_queue(ide_drive_t *drive)
         blk_queue_max_hw_segments(q, max_sg_entries);
         blk_queue_max_phys_segments(q, max_sg_entries);
 
+        blk_queue_softirq_done(q, ide_softirq_done);
+
         /* assign drive queue */
         drive->queue = q;
 
diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
index 180676d711..ee5f4dfdab 100644
--- a/drivers/scsi/scsi.c
+++ b/drivers/scsi/scsi.c
@@ -69,7 +69,6 @@
 #include "scsi_logging.h"
 
 static void scsi_done(struct scsi_cmnd *cmd);
-static int scsi_retry_command(struct scsi_cmnd *cmd);
 
 /*
  * Definitions and constants.
@@ -752,7 +751,7 @@ static void scsi_done(struct scsi_cmnd *cmd)
  * isn't running --- used by scsi_times_out */
 void __scsi_done(struct scsi_cmnd *cmd)
 {
-        unsigned long flags;
+        struct request *rq = cmd->request;
 
         /*
          * Set the serial numbers back to zero
@@ -763,71 +762,14 @@ void __scsi_done(struct scsi_cmnd *cmd)
         if (cmd->result)
                 atomic_inc(&cmd->device->ioerr_cnt);
 
+        BUG_ON(!rq);
+
         /*
-         * Next, enqueue the command into the done queue.
-         * It is a per-CPU queue, so we just disable local interrupts
-         * and need no spinlock.
+         * The uptodate/nbytes values don't matter, as we allow partial
+         * completes and thus will check this in the softirq callback
          */
-        local_irq_save(flags);
-        list_add_tail(&cmd->eh_entry, &__get_cpu_var(scsi_done_q));
-        raise_softirq_irqoff(SCSI_SOFTIRQ);
-        local_irq_restore(flags);
-}
-
-/**
- * scsi_softirq - Perform post-interrupt processing of finished SCSI commands.
- *
- * This is the consumer of the done queue.
- *
- * This is called with all interrupts enabled.  This should reduce
- * interrupt latency, stack depth, and reentrancy of the low-level
- * drivers.
- */
-static void scsi_softirq(struct softirq_action *h)
-{
-        int disposition;
-        LIST_HEAD(local_q);
-
-        local_irq_disable();
-        list_splice_init(&__get_cpu_var(scsi_done_q), &local_q);
-        local_irq_enable();
-
-        while (!list_empty(&local_q)) {
-                struct scsi_cmnd *cmd = list_entry(local_q.next,
-                                                   struct scsi_cmnd, eh_entry);
-                /* The longest time any command should be outstanding is the
-                 * per command timeout multiplied by the number of retries.
-                 *
-                 * For a typical command, this is 2.5 minutes */
-                unsigned long wait_for 
-                        = cmd->allowed * cmd->timeout_per_command;
-                list_del_init(&cmd->eh_entry);
-
-                disposition = scsi_decide_disposition(cmd);
-                if (disposition != SUCCESS &&
-                    time_before(cmd->jiffies_at_alloc + wait_for, jiffies)) {
-                        sdev_printk(KERN_ERR, cmd->device,
-                                    "timing out command, waited %lus\n",
-                                    wait_for/HZ);
-                        disposition = SUCCESS;
-                }
-                        
-                scsi_log_completion(cmd, disposition);
-                switch (disposition) {
-                case SUCCESS:
-                        scsi_finish_command(cmd);
-                        break;
-                case NEEDS_RETRY:
-                        scsi_retry_command(cmd);
-                        break;
-                case ADD_TO_MLQUEUE:
-                        scsi_queue_insert(cmd, SCSI_MLQUEUE_DEVICE_BUSY);
-                        break;
-                default:
-                        if (!scsi_eh_scmd_add(cmd, 0))
-                                scsi_finish_command(cmd);
-                }
-        }
+        rq->completion_data = cmd;
+        blk_complete_request(rq);
 }
 
 /*
@@ -840,7 +782,7 @@ static void scsi_softirq(struct softirq_action *h)
  *              level drivers should not become re-entrant as a result of
  *              this.
  */
-static int scsi_retry_command(struct scsi_cmnd *cmd)
+int scsi_retry_command(struct scsi_cmnd *cmd)
 {
         /*
          * Restore the SCSI command state.
@@ -1273,38 +1215,6 @@ int scsi_device_cancel(struct scsi_device *sdev, int recovery)
 }
 EXPORT_SYMBOL(scsi_device_cancel);
 
-#ifdef CONFIG_HOTPLUG_CPU
-static int scsi_cpu_notify(struct notifier_block *self,
-                           unsigned long action, void *hcpu)
-{
-        int cpu = (unsigned long)hcpu;
-
-        switch(action) {
-        case CPU_DEAD:
-                /* Drain scsi_done_q. */
-                local_irq_disable();
-                list_splice_init(&per_cpu(scsi_done_q, cpu),
-                                 &__get_cpu_var(scsi_done_q));
-                raise_softirq_irqoff(SCSI_SOFTIRQ);
-                local_irq_enable();
-                break;
-        default:
-                break;
-        }
-        return NOTIFY_OK;
-}
-
-static struct notifier_block __devinitdata scsi_cpu_nb = {
-        .notifier_call  = scsi_cpu_notify,
-};
-
-#define register_scsi_cpu() register_cpu_notifier(&scsi_cpu_nb)
-#define unregister_scsi_cpu() unregister_cpu_notifier(&scsi_cpu_nb)
-#else
-#define register_scsi_cpu()
-#define unregister_scsi_cpu()
-#endif /* CONFIG_HOTPLUG_CPU */
-
 MODULE_DESCRIPTION("SCSI core");
 MODULE_LICENSE("GPL");
 
@@ -1338,8 +1248,6 @@ static int __init init_scsi(void)
                 INIT_LIST_HEAD(&per_cpu(scsi_done_q, i));
 
         devfs_mk_dir("scsi");
-        open_softirq(SCSI_SOFTIRQ, scsi_softirq, NULL);
-        register_scsi_cpu();
         printk(KERN_NOTICE "SCSI subsystem initialized\n");
         return 0;
 
@@ -1367,7 +1275,6 @@ static void __exit exit_scsi(void)
         devfs_remove("scsi");
         scsi_exit_procfs();
         scsi_exit_queue();
-        unregister_scsi_cpu();
 }
 
 subsys_initcall(init_scsi);
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index ba93d6e66d..00c9bf383e 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1493,6 +1493,41 @@ static void scsi_kill_request(struct request *req, request_queue_t *q)
         __scsi_done(cmd);
 }
 
+static void scsi_softirq_done(struct request *rq)
+{
+        struct scsi_cmnd *cmd = rq->completion_data;
+        unsigned long wait_for = cmd->allowed * cmd->timeout_per_command;
+        int disposition;
+
+        INIT_LIST_HEAD(&cmd->eh_entry);
+
+        disposition = scsi_decide_disposition(cmd);
+        if (disposition != SUCCESS &&
+            time_before(cmd->jiffies_at_alloc + wait_for, jiffies)) {
+                sdev_printk(KERN_ERR, cmd->device,
+                            "timing out command, waited %lus\n",
+                            wait_for/HZ);
+                disposition = SUCCESS;
+        }
+                        
+        scsi_log_completion(cmd, disposition);
+
+        switch (disposition) {
+                case SUCCESS:
+                        scsi_finish_command(cmd);
+                        break;
+                case NEEDS_RETRY:
+                        scsi_retry_command(cmd);
+                        break;
+                case ADD_TO_MLQUEUE:
+                        scsi_queue_insert(cmd, SCSI_MLQUEUE_DEVICE_BUSY);
+                        break;
+                default:
+                        if (!scsi_eh_scmd_add(cmd, 0))
+                                scsi_finish_command(cmd);
+        }
+}
+
 /*
  * Function:    scsi_request_fn()
  *
@@ -1667,6 +1702,7 @@ struct request_queue *scsi_alloc_queue(struct scsi_device *sdev)
         blk_queue_bounce_limit(q, scsi_calculate_bounce_limit(shost));
         blk_queue_segment_boundary(q, shost->dma_boundary);
         blk_queue_issue_flush_fn(q, scsi_issue_flush_fn);
+        blk_queue_softirq_done(q, scsi_softirq_done);
 
         if (!shost->use_clustering)
                 clear_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
diff --git a/drivers/scsi/scsi_priv.h b/drivers/scsi/scsi_priv.h
index f04e7e11f5..14a6198cb8 100644
--- a/drivers/scsi/scsi_priv.h
+++ b/drivers/scsi/scsi_priv.h
@@ -44,6 +44,7 @@ extern void scsi_init_cmd_from_req(struct scsi_cmnd *cmd,
                 struct scsi_request *sreq);
 extern void __scsi_release_request(struct scsi_request *sreq);
 extern void __scsi_done(struct scsi_cmnd *cmd);
+extern int scsi_retry_command(struct scsi_cmnd *cmd);
 #ifdef CONFIG_SCSI_LOGGING
 void scsi_log_send(struct scsi_cmnd *cmd);
 void scsi_log_completion(struct scsi_cmnd *cmd, int disposition);
diff --git a/fs/bio.c b/fs/bio.c
index dfe242a21e..7b30695899 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -126,6 +126,7 @@ static void bio_fs_destructor(struct bio *bio)
 inline void bio_init(struct bio *bio)
 {
         bio->bi_next = NULL;
+        bio->bi_bdev = NULL;
         bio->bi_flags = 1 << BIO_UPTODATE;
         bio->bi_rw = 0;
         bio->bi_vcnt = 0;
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index fb09853774..02a585faa6 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -118,9 +118,9 @@ struct request_list {
  * try to put the fields that are referenced together in the same cacheline
  */
 struct request {
-        struct list_head queuelist; /* looking for ->queue? you must _not_
-                                     * access it directly, use
-                                     * blkdev_dequeue_request! */
+        struct list_head queuelist;
+        struct list_head donelist;
+
         unsigned long flags;            /* see REQ_ bits below */
 
         /* Maintain bio traversal state for part by part I/O submission.
@@ -141,6 +141,7 @@ struct request {
         struct bio *biotail;
 
         void *elevator_private;
+        void *completion_data;
 
         unsigned short ioprio;
 
@@ -291,6 +292,7 @@ typedef int (merge_bvec_fn) (request_queue_t *, struct bio *, struct bio_vec *);
 typedef void (activity_fn) (void *data, int rw);
 typedef int (issue_flush_fn) (request_queue_t *, struct gendisk *, sector_t *);
 typedef void (prepare_flush_fn) (request_queue_t *, struct request *);
+typedef void (softirq_done_fn)(struct request *);
 
 enum blk_queue_state {
         Queue_down,
@@ -332,6 +334,7 @@ struct request_queue
         activity_fn             *activity_fn;
         issue_flush_fn          *issue_flush_fn;
         prepare_flush_fn        *prepare_flush_fn;
+        softirq_done_fn         *softirq_done_fn;
 
         /*
          * Dispatch queue sorting
@@ -592,7 +595,6 @@ extern void generic_make_request(struct bio *bio);
 extern void blk_put_request(struct request *);
 extern void __blk_put_request(request_queue_t *, struct request *);
 extern void blk_end_sync_rq(struct request *rq, int error);
-extern void blk_attempt_remerge(request_queue_t *, struct request *);
 extern struct request *blk_get_request(request_queue_t *, int, gfp_t);
 extern void blk_insert_request(request_queue_t *, struct request *, int, void *);
 extern void blk_requeue_request(request_queue_t *, struct request *);
@@ -646,6 +648,17 @@ extern int end_that_request_first(struct request *, int, int);
 extern int end_that_request_chunk(struct request *, int, int);
 extern void end_that_request_last(struct request *, int);
 extern void end_request(struct request *req, int uptodate);
+extern void blk_complete_request(struct request *);
+
+static inline int rq_all_done(struct request *rq, unsigned int nr_bytes)
+{
+        if (blk_fs_request(rq))
+                return (nr_bytes >= (rq->hard_nr_sectors << 9));
+        else if (blk_pc_request(rq))
+                return nr_bytes >= rq->data_len;
+
+        return 0;
+}
 
 /*
  * end_that_request_first/chunk() takes an uptodate argument. we account
@@ -694,6 +707,7 @@ extern void blk_queue_segment_boundary(request_queue_t *, unsigned long);
 extern void blk_queue_prep_rq(request_queue_t *, prep_rq_fn *pfn);
 extern void blk_queue_merge_bvec(request_queue_t *, merge_bvec_fn *);
 extern void blk_queue_dma_alignment(request_queue_t *, int);
+extern void blk_queue_softirq_done(request_queue_t *, softirq_done_fn *);
 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
 extern int blk_queue_ordered(request_queue_t *, unsigned, prepare_flush_fn *);
 extern void blk_queue_issue_flush_fn(request_queue_t *, issue_flush_fn *);
diff --git a/include/linux/elevator.h b/include/linux/elevator.h
index fb80fa44c4..4a6f50e31c 100644
--- a/include/linux/elevator.h
+++ b/include/linux/elevator.h
@@ -114,8 +114,6 @@ extern ssize_t elv_iosched_store(request_queue_t *, const char *, size_t);
 extern int elevator_init(request_queue_t *, char *);
 extern void elevator_exit(elevator_t *);
 extern int elv_rq_merge_ok(struct request *, struct bio *);
-extern int elv_try_merge(struct request *, struct bio *);
-extern int elv_try_last_merge(request_queue_t *, struct bio *);
 
 /*
  * Return values from elevator merger
diff --git a/include/linux/ide.h b/include/linux/ide.h
index 4dd6694963..ef8d0cbb83 100644
--- a/include/linux/ide.h
+++ b/include/linux/ide.h
@@ -1001,6 +1001,7 @@ extern int noautodma;
 
 extern int ide_end_request (ide_drive_t *drive, int uptodate, int nrsecs);
 extern int __ide_end_request (ide_drive_t *drive, struct request *rq, int uptodate, int nrsecs);
+extern void ide_softirq_done(struct request *rq);
 
 /*
  * This is used on exit from the driver to designate the next irq handler
diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
index e50a95fbeb..2c08fdc2bd 100644
--- a/include/linux/interrupt.h
+++ b/include/linux/interrupt.h
@@ -112,7 +112,7 @@ enum
         TIMER_SOFTIRQ,
         NET_TX_SOFTIRQ,
         NET_RX_SOFTIRQ,
-        SCSI_SOFTIRQ,
+        BLOCK_SOFTIRQ,
         TASKLET_SOFTIRQ
 };
 
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 51747cd88d..a1d26cb289 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -125,36 +125,7 @@ static inline void rcu_bh_qsctr_inc(int cpu)
         rdp->passed_quiesc = 1;
 }
 
-static inline int __rcu_pending(struct rcu_ctrlblk *rcp,
-                                                struct rcu_data *rdp)
-{
-        /* This cpu has pending rcu entries and the grace period
-         * for them has completed.
-         */
-        if (rdp->curlist && !rcu_batch_before(rcp->completed, rdp->batch))
-                return 1;
-
-        /* This cpu has no pending entries, but there are new entries */
-        if (!rdp->curlist && rdp->nxtlist)
-                return 1;
-
-        /* This cpu has finished callbacks to invoke */
-        if (rdp->donelist)
-                return 1;
-
-        /* The rcu core waits for a quiescent state from the cpu */
-        if (rdp->quiescbatch != rcp->cur || rdp->qs_pending)
-                return 1;
-
-        /* nothing to do */
-        return 0;
-}
-
-static inline int rcu_pending(int cpu)
-{
-        return __rcu_pending(&rcu_ctrlblk, &per_cpu(rcu_data, cpu)) ||
-                __rcu_pending(&rcu_bh_ctrlblk, &per_cpu(rcu_bh_data, cpu));
-}
+extern int rcu_pending(int cpu);
 
 /**
  * rcu_read_lock - mark the beginning of an RCU read-side critical section.
diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c
index 30b0bba038..ccc45d49ce 100644
--- a/kernel/rcupdate.c
+++ b/kernel/rcupdate.c
@@ -429,6 +429,36 @@ static void rcu_process_callbacks(unsigned long unused)
                                 &__get_cpu_var(rcu_bh_data));
 }
 
+static int __rcu_pending(struct rcu_ctrlblk *rcp, struct rcu_data *rdp)
+{
+        /* This cpu has pending rcu entries and the grace period
+         * for them has completed.
+         */
+        if (rdp->curlist && !rcu_batch_before(rcp->completed, rdp->batch))
+                return 1;
+
+        /* This cpu has no pending entries, but there are new entries */
+        if (!rdp->curlist && rdp->nxtlist)
+                return 1;
+
+        /* This cpu has finished callbacks to invoke */
+        if (rdp->donelist)
+                return 1;
+
+        /* The rcu core waits for a quiescent state from the cpu */
+        if (rdp->quiescbatch != rcp->cur || rdp->qs_pending)
+                return 1;
+
+        /* nothing to do */
+        return 0;
+}
+
+int rcu_pending(int cpu)
+{
+        return __rcu_pending(&rcu_ctrlblk, &per_cpu(rcu_data, cpu)) ||
+                __rcu_pending(&rcu_bh_ctrlblk, &per_cpu(rcu_bh_data, cpu));
+}
+
 void rcu_check_callbacks(int cpu, int user)
 {
         if (user || 
diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 7849cac14d..a67f1b44c9 100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -402,7 +402,7 @@ static int netlink_create(struct socket *sock, int protocol)
         groups = nl_table[protocol].groups;
         netlink_unlock_table();
 
-        if ((err = __netlink_create(sock, protocol) < 0))
+        if ((err = __netlink_create(sock, protocol)) < 0)
                 goto out_module;
 
         nlk = nlk_sk(sock->sk);