Merge branch 'blk-softirq' of git://brick.kernel.dk/data/git/linux-2.6-block

Manual merge for trivial #include changes

block/ll_rw_blk.c

@@ -26,6 +26,8 @@
 #include <linux/slab.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
+#include <linux/interrupt.h>
+#include <linux/cpu.h>
 
 /*
  * for max sense size
@@ -68,6 +70,8 @@ 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)
 
@@ -206,6 +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
@@ -269,6 +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;
@@ -285,6 +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;
 }
 
 /**
@@ -3261,6 +3274,87 @@ 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
  */
@@ -3339,6 +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");
@@ -3352,6 +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;
 

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);
 }
 
 /* 
@@ -3209,6 +3227,8 @@ static int __devinit cciss_init_one(struct pci_dev *pdev,
 
 		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);
 		sprintf(disk->devfs_name, "cciss/host%d/target%d", i, j);

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);
 	}
 
+	/*
+	 * 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;
+		blk_complete_request(rq);
+		ret = 0;
+	} else {
 		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);
-
 			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;
 

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;
 

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);
 
-	/*
-	 * Next, enqueue the command into the done queue.
-	 * It is a per-CPU queue, so we just disable local interrupts
-	 * and need no spinlock.
-	 */
-	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);
-}
+	BUG_ON(!rq);
 
-/**
- * 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.
+	/*
+	 * The uptodate/nbytes values don't matter, as we allow partial
+	 * completes and thus will check this in the softirq callback
 	 */
-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);

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);

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);

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
@@ -645,6 +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
@@ -693,6 +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 *);

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

include/linux/interrupt.h

@@ -112,7 +112,7 @@ enum
 	TIMER_SOFTIRQ,
 	NET_TX_SOFTIRQ,
 	NET_RX_SOFTIRQ,
-	SCSI_SOFTIRQ,
+	BLOCK_SOFTIRQ,
 	TASKLET_SOFTIRQ
 };