Hand merged.

arch/i386/kernel/cpu/mtrr/generic.c

@@ -69,7 +69,9 @@ void generic_set_mtrr(unsigned int reg, unsigned long base,
 {
 	u32 cr0, cr4 = 0;
 	u32 deftype_lo, deftype_hi;
+	static spinlock_t set_atomicity_lock = SPIN_LOCK_UNLOCKED;
 
+	spin_lock(&set_atomicity_lock);
 	/*  Save value of CR4 and clear Page Global Enable (bit 7)  */
 	if ( cpu_has_pge ) {
 		cr4 = read_cr4();
@@ -112,6 +114,7 @@ void generic_set_mtrr(unsigned int reg, unsigned long base,
 	/*  Restore value of CR4  */
 	if ( cpu_has_pge )
 		write_cr4(cr4);
+	spin_unlock(&set_atomicity_lock);
 }
 
 int generic_validate_add_page(unsigned long base, unsigned long size, unsigned int type)

arch/i386/mm/fault.c

@@ -184,7 +184,7 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long error_code)
 	 * If we're in an interrupt, have no user context or are running in an
 	 * atomic region then we must not take the fault..
 	 */
-	if (preempt_count() || !mm)
+	if (in_atomic() || !mm)
 		goto no_context;
 
 	down_read(&mm->mmap_sem);

drivers/md/linear.c

@@ -55,12 +55,12 @@ static int linear_run (mddev_t *mddev)
 		int j = rdev->raid_disk;
 		dev_info_t *disk = conf->disks + j;
 
-		if (j < 0 || j > mddev->raid_disks || disk->bdev) {
+		if (j < 0 || j > mddev->raid_disks || disk->rdev) {
 			printk("linear: disk numbering problem. Aborting!\n");
 			goto out;
 		}
 
-		disk->bdev = rdev->bdev;
+		disk->rdev = rdev;
 		disk->size = rdev->size;
 
 		if (!conf->smallest || (disk->size < conf->smallest->size))
@@ -153,11 +153,11 @@ static int linear_make_request (request_queue_t *q, struct bio *bio)
     
 	if (block >= (tmp_dev->size + tmp_dev->offset)
 				|| block < tmp_dev->offset) {
-		printk ("linear_make_request: Block %ld out of bounds on dev %s size %ld offset %ld\n", block, bdevname(tmp_dev->bdev), tmp_dev->size, tmp_dev->offset);
+		printk ("linear_make_request: Block %ld out of bounds on dev %s size %ld offset %ld\n", block, bdevname(tmp_dev->rdev->bdev), tmp_dev->size, tmp_dev->offset);
 		bio_io_error(bio);
 		return 0;
 	}
-	bio->bi_bdev = tmp_dev->bdev;
+	bio->bi_bdev = tmp_dev->rdev->bdev;
 	bio->bi_sector = bio->bi_sector - (tmp_dev->offset << 1);
 
 	return 1;
@@ -176,11 +176,11 @@ static int linear_status (char *page, mddev_t *mddev)
 	for (j = 0; j < conf->nr_zones; j++)
 	{
 		sz += sprintf(page+sz, "[%s",
-			bdev_partition_name(conf->hash_table[j].dev0->bdev));
+			bdev_partition_name(conf->hash_table[j].dev0->rdev->bdev));
 
 		if (conf->hash_table[j].dev1)
 			sz += sprintf(page+sz, "/%s] ",
-			  bdev_partition_name(conf->hash_table[j].dev1->bdev));
+			  bdev_partition_name(conf->hash_table[j].dev1->rdev->bdev));
 		else
 			sz += sprintf(page+sz, "] ");
 	}

drivers/md/md.c

@@ -233,7 +233,7 @@ mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr)
 	struct list_head *tmp;
 
 	ITERATE_RDEV(mddev,rdev,tmp) {
-		if (rdev->raid_disk == nr)
+		if (rdev->desc_nr == nr)
 			return rdev;
 	}
 	return NULL;
@@ -251,18 +251,6 @@ static mdk_rdev_t * find_rdev(mddev_t * mddev, dev_t dev)
 	return NULL;
 }
 
-static mdk_rdev_t * find_rdev_bdev(mddev_t * mddev, struct block_device *bdev)
-{
-	struct list_head *tmp;
-	mdk_rdev_t *rdev;
-
-	ITERATE_RDEV(mddev,rdev,tmp) {
-		if (rdev->bdev == bdev)
-			return rdev;
-	}
-	return NULL;
-}
-
 static LIST_HEAD(device_names);
 
 char * partition_name(kdev_t dev)
@@ -377,9 +365,6 @@ static void free_disk_sb(mdk_rdev_t * rdev)
 		rdev->sb_page = NULL;
 		rdev->sb_offset = 0;
 		rdev->size = 0;
-	} else {
-		if (!rdev->faulty)
-			MD_BUG();
 	}
 }
 
@@ -594,11 +579,10 @@ static void export_rdev(mdk_rdev_t * rdev)
 		MD_BUG();
 	free_disk_sb(rdev);
 	list_del_init(&rdev->same_set);
-	unlock_rdev(rdev);
 #ifndef MODULE
 	md_autodetect_dev(rdev->bdev->bd_dev);
 #endif
-	rdev->faulty = 0;
+	unlock_rdev(rdev);
 	kfree(rdev);
 }
 
@@ -683,9 +667,9 @@ static void print_sb(mdp_super_t *sb)
 
 static void print_rdev(mdk_rdev_t *rdev)
 {
-	printk(KERN_INFO "md: rdev %s, SZ:%08ld F:%d DN:%d ",
+	printk(KERN_INFO "md: rdev %s, SZ:%08ld F:%d S:%d DN:%d ",
 		bdev_partition_name(rdev->bdev),
-		rdev->size, rdev->faulty, rdev->desc_nr);
+		rdev->size, rdev->faulty, rdev->in_sync, rdev->desc_nr);
 	if (rdev->sb) {
 		printk(KERN_INFO "md: rdev superblock:\n");
 		print_sb(rdev->sb);
@@ -816,6 +800,7 @@ static void sync_sbs(mddev_t * mddev)
 	mdk_rdev_t *rdev;
 	mdp_super_t *sb;
 	struct list_head *tmp;
+	int next_spare = mddev->raid_disks;
 
 	/* make all rdev->sb match mddev data..
 	 * we setup the data in the first rdev and copy it
@@ -868,12 +853,20 @@ static void sync_sbs(mddev_t * mddev)
 
 	sb->disks[0].state = (1<<MD_DISK_REMOVED);
 	ITERATE_RDEV(mddev,rdev,tmp) {
-		mdp_disk_t *d = &sb->disks[rdev->desc_nr];
+		mdp_disk_t *d;
+		if (rdev->raid_disk >= 0)
+			rdev->desc_nr = rdev->raid_disk;
+		else
+			rdev->desc_nr = next_spare++;
+		d = &sb->disks[rdev->desc_nr];
 		nr_disks++;
 		d->number = rdev->desc_nr;
 		d->major = MAJOR(rdev->bdev->bd_dev);
 		d->minor = MINOR(rdev->bdev->bd_dev);
+		if (rdev->raid_disk >= 0)
 			d->raid_disk = rdev->raid_disk;
+		else
+			d->raid_disk = rdev->desc_nr; /* compatability */
 		if (rdev->faulty) {
 			d->state = (1<<MD_DISK_FAULTY);
 			failed++;
@@ -909,8 +902,6 @@ static void sync_sbs(mddev_t * mddev)
 	ITERATE_RDEV(mddev,rdev,tmp) {
 		mdp_super_t *this_sb;
 		
-		if (rdev->faulty || rdev->alias_device)
-			continue;
 		this_sb = rdev->sb;
 		if (this_sb != sb)
 			*this_sb = *sb;
@@ -956,16 +947,17 @@ static void md_update_sb(mddev_t * mddev)
 		printk(KERN_INFO "md: ");
 		if (rdev->faulty)
 			printk("(skipping faulty ");
-		if (rdev->alias_device)
-			printk("(skipping alias ");
 
 		printk("%s ", bdev_partition_name(rdev->bdev));
-		if (!rdev->faulty && !rdev->alias_device) {
+		if (!rdev->faulty) {
 			printk("[events: %08lx]",
 				(unsigned long)rdev->sb->events_lo);
 			err += write_disk_sb(rdev);
 		} else
 			printk(")\n");
+		if (!err && mddev->level == LEVEL_MULTIPATH)
+			/* only need to write one superblock... */
+			break;
 	}
 	if (err) {
 		if (--count) {
@@ -1010,6 +1002,8 @@ static mdk_rdev_t *md_import_device(dev_t newdev, int on_disk)
 	}
 	rdev->desc_nr = -1;
 	rdev->faulty = 0;
+	rdev->in_sync = 0;
+	atomic_set(&rdev->nr_pending, 0);
 
 	size = rdev->bdev->bd_inode->i_size >> BLOCK_SIZE_BITS;
 	if (!size) {
@@ -1198,7 +1192,6 @@ static int analyze_sbs(mddev_t * mddev)
 	i = 0;
 	ITERATE_RDEV(mddev,rdev,tmp) {
 		if (mddev->level == LEVEL_MULTIPATH) {
-			rdev->alias_device = !!i;
 			rdev->desc_nr = i++;
 			rdev->raid_disk = rdev->desc_nr;
 			rdev->in_sync = 1;
@@ -1206,15 +1199,17 @@ static int analyze_sbs(mddev_t * mddev)
 			mdp_disk_t *desc;
 			rdev->desc_nr = rdev->sb->this_disk.number;
 			desc = sb->disks + rdev->desc_nr;
-			rdev->raid_disk = desc->raid_disk;
+			rdev->raid_disk = -1;
 			rdev->in_sync = rdev->faulty = 0;
 
 			if (desc->state & (1<<MD_DISK_FAULTY)) {
 				rdev->faulty = 1;
 				kick_rdev_from_array(rdev);
 			} else if (desc->state & (1<<MD_DISK_SYNC) &&
-				 rdev->raid_disk < mddev->raid_disks)
+				   desc->raid_disk < mddev->raid_disks) {
 				rdev->in_sync = 1;
+				rdev->raid_disk = desc->raid_disk;
+			}
 		}
 	}
 
@@ -1345,6 +1340,8 @@ static int do_md_run(mddev_t * mddev)
 	struct list_head *tmp;
 	mdk_rdev_t *rdev;
 	struct gendisk *disk;
+	char *major_name;
+
 
 	if (list_empty(&mddev->disks)) {
 		MD_BUG();
@@ -1397,10 +1394,7 @@ static int do_md_run(mddev_t * mddev)
 			printk(TOO_SMALL_CHUNKSIZE, chunk_size, PAGE_SIZE);
 			return -EINVAL;
 		}
-	} else
-		if (chunk_size)
-			printk(KERN_INFO "md: RAID level %d does not need chunksize! Continuing anyway.\n",
-			       mddev->level);
+	}
 
 	if (pnum >= MAX_PERSONALITY) {
 		MD_BUG();
@@ -1454,15 +1448,16 @@ static int do_md_run(mddev_t * mddev)
 	if (!disk)
 		return -ENOMEM;
 	memset(disk, 0, sizeof(struct gendisk));
-	disk->major_name = kmalloc(6, GFP_KERNEL);
-	if (!disk->major_name) {
+	major_name = kmalloc(6, GFP_KERNEL);
+	if (!major_name) {
 		kfree(disk);
 		return -ENOMEM;
 	}
 	disk->major = MD_MAJOR;
 	disk->first_minor = mdidx(mddev);
 	disk->minor_shift = 0;
-	sprintf(disk->major_name, "md%d", mdidx(mddev));
+	sprintf(major_name, "md%d", mdidx(mddev));
+	disk->major_name = major_name;
 	disk->part = md_hd_struct + mdidx(mddev);
 	disk->nr_real = 1;
 	disk->fops = &md_fops;
@@ -1559,10 +1554,6 @@ static int do_md_stop(mddev_t * mddev, int ro)
 				mddev->recovery_running = -EINTR;
 			md_unregister_thread(mddev->sync_thread);
 			mddev->sync_thread = NULL;
-			if (mddev->spare) {
-				mddev->pers->spare_inactive(mddev);
-				mddev->spare = NULL;
-			}
 		}
 
 		invalidate_device(dev, 1);
@@ -1933,7 +1924,7 @@ static int get_disk_info(mddev_t * mddev, void * arg)
 		}
 	} else {
 		info.major = info.minor = 0;
-		info.raid_disk = 0;
+		info.raid_disk = -1;
 		info.state = (1<<MD_DISK_REMOVED);
 	}
 
@@ -1983,7 +1974,11 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 			return PTR_ERR(rdev);
 		}
 		rdev->desc_nr = info->number;
+		if (info->raid_disk < mddev->raid_disks)
 			rdev->raid_disk = info->raid_disk;
+		else
+			rdev->raid_disk = -1;
+
 		rdev->faulty = 0;
 		if (rdev->raid_disk < mddev->raid_disks)
 			rdev->in_sync = (info->state & (1<<MD_DISK_SYNC));
@@ -2042,7 +2037,6 @@ static int hot_generate_error(mddev_t * mddev, dev_t dev)
 
 static int hot_remove_disk(mddev_t * mddev, dev_t dev)
 {
-	int err;
 	mdk_rdev_t *rdev;
 
 	if (!mddev->pers)
@@ -2051,28 +2045,12 @@ static int hot_remove_disk(mddev_t * mddev, dev_t dev)
 	printk(KERN_INFO "md: trying to remove %s from md%d ... \n",
 		partition_name(to_kdev_t(dev)), mdidx(mddev));
 
-	if (!mddev->pers->hot_remove_disk) {
-		printk(KERN_WARNING "md%d: personality does not support diskops!\n",
-		       mdidx(mddev));
-		return -EINVAL;
-	}
-
 	rdev = find_rdev(mddev, dev);
 	if (!rdev)
 		return -ENXIO;
 
-	if (rdev->in_sync && ! rdev->faulty)
-		goto busy;
-
-	err = mddev->pers->hot_remove_disk(mddev, rdev->raid_disk);
-	if (err == -EBUSY) {
-		MD_BUG();
+	if (rdev->raid_disk >= 0)
 		goto busy;
-	}
-	if (err) {
-		MD_BUG();
-		return -EINVAL;
-	}
 
 	kick_rdev_from_array(rdev);
 	md_update_sb(mddev);
@@ -2145,13 +2123,7 @@ static int hot_add_disk(mddev_t * mddev, dev_t dev)
 	}
 
 	rdev->desc_nr = i;
-	rdev->raid_disk = i;
-
-	if (mddev->pers->hot_add_disk(mddev, rdev)) {
-		MD_BUG();
-		err = -EINVAL;
-		goto abort_unbind_export;
-	}
+	rdev->raid_disk = -1;
 
 	md_update_sb(mddev);
 
@@ -2204,14 +2176,13 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
 static int set_disk_faulty(mddev_t *mddev, dev_t dev)
 {
 	mdk_rdev_t *rdev;
-	int ret;
 
 	rdev = find_rdev(mddev, dev);
 	if (!rdev)
 		return 0;
 
-	ret = md_error(mddev, rdev->bdev);
-	return ret;
+	md_error(mddev, rdev);
+	return 1;
 }
 
 static int md_ioctl(struct inode *inode, struct file *file,
@@ -2434,6 +2405,7 @@ static int md_ioctl(struct inode *inode, struct file *file,
 		}
 
 		default:
+			if (_IOC_TYPE(cmd) == MD_MAJOR)
 				printk(KERN_WARNING "md: %s(pid %d) used obsolete MD ioctl, "
 				       "upgrade your software to use new ictls.\n",
 				       current->comm, current->pid);
@@ -2626,10 +2598,8 @@ static void md_recover_arrays(void)
 }
 
 
-int md_error(mddev_t *mddev, struct block_device *bdev)
+void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
 {
-	mdk_rdev_t * rrdev;
-
 	dprintk("md_error dev:(%d:%d), rdev:(%d:%d), (caller: %p,%p,%p,%p).\n",
 		MD_MAJOR,mdidx(mddev),MAJOR(bdev->bd_dev),MINOR(bdev->bd_dev),
 		__builtin_return_address(0),__builtin_return_address(1),
@@ -2637,25 +2607,15 @@ int md_error(mddev_t *mddev, struct block_device *bdev)
 
 	if (!mddev) {
 		MD_BUG();
-		return 0;
+		return;
 	}
-	rrdev = find_rdev_bdev(mddev, bdev);
-	if (!rrdev || rrdev->faulty)
-		return 0;
-	if (!mddev->pers->error_handler
-			|| mddev->pers->error_handler(mddev,bdev) <= 0) {
-		rrdev->faulty = 1;
-		rrdev->in_sync = 0;
-	} else
-		return 1;
-	/*
-	 * if recovery was running, stop it now.
-	 */
-	if (mddev->recovery_running) 
-		mddev->recovery_running = -EIO;
-	md_recover_arrays();
 
-	return 0;
+	if (!rdev || rdev->faulty)
+		return;
+	if (!mddev->pers->error_handler)
+		return;
+	mddev->pers->error_handler(mddev,rdev);
+	md_recover_arrays();
 }
 
 static int status_unused(char * page)
@@ -2706,7 +2666,7 @@ static int status_resync(char * page, mddev_t * mddev)
 		sz += sprintf(page + sz, "] ");
 	}
 	sz += sprintf(page + sz, " %s =%3lu.%lu%% (%lu/%lu)",
-		      (mddev->spare ? "recovery" : "resync"),
+		      (mddev->spares ? "recovery" : "resync"),
 		      res/10, res % 10, resync, max_blocks);
 
 	/*
@@ -2824,26 +2784,10 @@ int unregister_md_personality(int pnum)
 	return 0;
 }
 
-static mdk_rdev_t *get_spare(mddev_t *mddev)
-{
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
-
-	ITERATE_RDEV(mddev,rdev,tmp) {
-		if (rdev->faulty)
-			continue;
-		if (rdev->in_sync)
-			continue;
-
-		return rdev;
-	}
-	return NULL;
-}
-
 static unsigned int sync_io[DK_MAX_MAJOR][DK_MAX_DISK];
-void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
+void md_sync_acct(mdk_rdev_t *rdev, unsigned long nr_sectors)
 {
-	kdev_t dev = to_kdev_t(bdev->bd_dev);
+	kdev_t dev = to_kdev_t(rdev->bdev->bd_dev);
 	unsigned int major = major(dev);
 	unsigned int index;
 
@@ -3057,19 +3001,30 @@ static void md_do_sync(void *data)
 
 
 /*
- * This is the kernel thread that watches all md arrays for re-sync action
- * that might be needed.
+ * This is the kernel thread that watches all md arrays for re-sync and other
+ * action that might be needed.
  * It does not do any resync itself, but rather "forks" off other threads
  * to do that as needed.
  * When it is determined that resync is needed, we set "->recovery_running" and
  * create a thread at ->sync_thread.
- * When the thread finishes is clears recovery_running (or set and error)
+ * When the thread finishes it clears recovery_running (or sets an error)
  * and wakeup up this thread which will reap the thread and finish up.
+ * This thread also removes any faulty devices (with nr_pending == 0).
+ *
+ * The overall approach is:
+ *  1/ if the superblock needs updating, update it.
+ *  2/ If a recovery thread is running, don't do anything else.
+ *  3/ If recovery has finished, clean up, possibly marking spares active.
+ *  4/ If there are any faulty devices, remove them.
+ *  5/ If array is degraded, try to add spares devices
+ *  6/ If array has spares or is not in-sync, start a resync thread.
  */
 void md_do_recovery(void *data)
 {
 	mddev_t *mddev;
-	struct list_head *tmp;
+	mdk_rdev_t *rdev;
+	struct list_head *tmp, *rtmp;
+
 
 	dprintk(KERN_INFO "md: recovery thread got woken up ...\n");
 
@@ -3085,26 +3040,11 @@ void md_do_recovery(void *data)
 			/* resync has finished, collect result */
 			md_unregister_thread(mddev->sync_thread);
 			mddev->sync_thread = NULL;
-			if (mddev->recovery_running < 0) {
-				/* some sort of failure.
-				 * If we were doing a reconstruction,
-				 * we need to retrieve the spare
-				 */
-				if (!mddev->pers->spare_inactive)
-					goto unlock;
-				if (mddev->spare) {
-					mddev->pers->spare_inactive(mddev);
-					mddev->spare = NULL;
-				}
-			} else {
-				if (!mddev->pers->spare_active)
-					goto unlock;
+			if (mddev->recovery_running == 0) {
 				/* success...*/
-				if (mddev->spare) {
+				/* activate any spares */
 				mddev->pers->spare_active(mddev);
-					mddev->spare->in_sync = 1;
-					mddev->spare = NULL;
-				}
+				mddev->spares = 0;
 			}
 			md_update_sb(mddev);
 			mddev->recovery_running = 0;
@@ -3117,16 +3057,33 @@ void md_do_recovery(void *data)
 			wake_up(&resync_wait);
 		}
 
+		/* no recovery is running.
+		 * remove any failed drives, then
+		 * add spares if possible
+		 */
+		mddev->spares = 0;
+		ITERATE_RDEV(mddev,rdev,rtmp) {
+			if (rdev->raid_disk >= 0 &&
+			    rdev->faulty &&
+			    atomic_read(&rdev->nr_pending)==0) {
+				mddev->pers->hot_remove_disk(mddev, rdev->raid_disk);
+				rdev->raid_disk = -1;
+			}
+			if (!rdev->faulty && rdev->raid_disk >= 0 && !rdev->in_sync)
+				mddev->spares++;
+		}
 		if (mddev->degraded) {
-			mddev->spare = get_spare(mddev);
-			if (!mddev->spare)
-				printk(KERN_ERR "md%d: no spare disk to reconstruct array! "
-				       "-- continuing in degraded mode\n", mdidx(mddev));
+			ITERATE_RDEV(mddev,rdev,rtmp)
+				if (rdev->raid_disk < 0
+				    && !rdev->faulty) {
+					if (mddev->pers->hot_add_disk(mddev,rdev))
+						mddev->spares++;
 					else
-				printk(KERN_INFO "md%d: resyncing spare disk %s to replace failed disk\n",
-				       mdidx(mddev), bdev_partition_name(mddev->spare->bdev));
+						break;
+				}
 		}
-		if (!mddev->spare && mddev->in_sync) {
+
+		if (!mddev->spares && mddev->in_sync) {
 			/* nothing we can do ... */
 			goto unlock;
 		}
@@ -3136,13 +3093,9 @@ void md_do_recovery(void *data)
 								"md_resync");
 			if (!mddev->sync_thread) {
 				printk(KERN_ERR "md%d: could not start resync thread...\n", mdidx(mddev));
-				if (mddev->spare)
-					mddev->pers->spare_inactive(mddev);
-				mddev->spare = NULL;
+				/* leave the spares where they are, it shouldn't hurt */
 				mddev->recovery_running = 0;
 			} else {
-				if (mddev->spare)
-					mddev->pers->spare_write(mddev);
 				mddev->recovery_running = 1;
 				md_wakeup_thread(mddev->sync_thread);
 			}
@@ -3540,7 +3493,7 @@ static int __init raid_setup(char *str)
 	return 1;
 }
 
-int __init md_run_setup(void)
+static int __init md_run_setup(void)
 {
 	if (raid_setup_args.noautodetect)
 		printk(KERN_INFO "md: Skipping autodetection of RAID arrays. (raid=noautodetect)\n");
@@ -3604,6 +3557,5 @@ EXPORT_SYMBOL(md_register_thread);
 EXPORT_SYMBOL(md_unregister_thread);
 EXPORT_SYMBOL(md_wakeup_thread);
 EXPORT_SYMBOL(md_print_devices);
-EXPORT_SYMBOL(find_rdev_nr);
 EXPORT_SYMBOL(md_interrupt_thread);
 MODULE_LICENSE("GPL");

drivers/md/multipath.c

@@ -70,7 +70,7 @@ static void mp_pool_free(void *mpb, void *data)
 	kfree(mpb);
 }
 
-static int multipath_map (mddev_t *mddev, struct block_device **bdev)
+static int multipath_map (mddev_t *mddev, mdk_rdev_t **rdevp)
 {
 	multipath_conf_t *conf = mddev_to_conf(mddev);
 	int i, disks = MD_SB_DISKS;
@@ -80,12 +80,17 @@ static int multipath_map (mddev_t *mddev, struct block_device **bdev)
 	 * now we use the first available disk.
 	 */
 
+	spin_lock_irq(&conf->device_lock);
 	for (i = 0; i < disks; i++) {
-		if (conf->multipaths[i].operational) {
-			*bdev = conf->multipaths[i].bdev;
-			return (0);
+		mdk_rdev_t *rdev = conf->multipaths[i].rdev;
+		if (rdev && rdev->in_sync) {
+			*rdevp = rdev;
+			atomic_inc(&rdev->nr_pending);
+			spin_unlock_irq(&conf->device_lock);
+			return 0;
 		}
 	}
+	spin_unlock_irq(&conf->device_lock);
 
 	printk (KERN_ERR "multipath_map(): no more operational IO paths?\n");
 	return (-1);
@@ -126,21 +131,21 @@ void multipath_end_request(struct bio *bio)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
 	struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private);
-	multipath_conf_t *conf;
-	struct block_device *bdev;
-	if (uptodate) {
+	multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev);
+	mdk_rdev_t *rdev = conf->multipaths[mp_bh->path].rdev;
+
+	if (uptodate)
 		multipath_end_bh_io(mp_bh, uptodate);
-		return;
-	}
+	else {
 		/*
 		 * oops, IO error:
 		 */
-	conf = mddev_to_conf(mp_bh->mddev);
-	bdev = conf->multipaths[mp_bh->path].bdev;
-	md_error (mp_bh->mddev, bdev);
+		md_error (mp_bh->mddev, rdev);
 		printk(KERN_ERR "multipath: %s: rescheduling sector %lu\n", 
-		 bdev_partition_name(bdev), bio->bi_sector);
+		       bdev_partition_name(rdev->bdev), bio->bi_sector);
 		multipath_reschedule_retry(mp_bh);
+	}
+	atomic_dec(&rdev->nr_pending);
 	return;
 }
 
@@ -153,9 +158,11 @@ static int multipath_read_balance (multipath_conf_t *conf)
 {
 	int disk;
 
-	for (disk = 0; disk < MD_SB_DISKS; disk++)	
-		if (conf->multipaths[disk].operational)
+	for (disk = 0; disk < MD_SB_DISKS; disk++) {
+		mdk_rdev_t *rdev = conf->multipaths[disk].rdev;
+		if (rdev && rdev->in_sync)
 			return disk;
+	}
 	BUG();
 	return 0;
 }
@@ -175,11 +182,14 @@ static int multipath_make_request (request_queue_t *q, struct bio * bio)
 	/*
 	 * read balancing logic:
 	 */
+	spin_lock_irq(&conf->device_lock);
 	mp_bh->path = multipath_read_balance(conf);
 	multipath = conf->multipaths + mp_bh->path;
+	atomic_inc(&multipath->rdev->nr_pending);
+	spin_unlock_irq(&conf->device_lock);
 
 	mp_bh->bio = *bio;
-	mp_bh->bio.bi_bdev = multipath->bdev;
+	mp_bh->bio.bi_bdev = multipath->rdev->bdev;
 	mp_bh->bio.bi_end_io = multipath_end_request;
 	mp_bh->bio.bi_private = mp_bh;
 	generic_make_request(&mp_bh->bio);
@@ -195,7 +205,8 @@ static int multipath_status (char *page, mddev_t *mddev)
 						 conf->working_disks);
 	for (i = 0; i < conf->raid_disks; i++)
 		sz += sprintf (page+sz, "%s",
-			conf->multipaths[i].operational ? "U" : "_");
+			       conf->multipaths[i].rdev && 
+			       conf->multipaths[i].rdev->in_sync ? "U" : "_");
 	sz += sprintf (page+sz, "]");
 	return sz;
 }
@@ -210,28 +221,13 @@ static int multipath_status (char *page, mddev_t *mddev)
 "multipath: IO failure on %s, disabling IO path. \n" \
 "	Operation continuing on %d IO paths.\n"
 
-static void mark_disk_bad (mddev_t *mddev, int failed)
-{
-	multipath_conf_t *conf = mddev_to_conf(mddev);
-	struct multipath_info *multipath = conf->multipaths+failed;
-
-	multipath->operational = 0;
-	mddev->sb_dirty = 1;
-	conf->working_disks--;
-	printk (DISK_FAILED, bdev_partition_name (multipath->bdev),
-				 conf->working_disks);
-}
 
 /*
  * Careful, this can execute in IRQ contexts as well!
  */
-static int multipath_error (mddev_t *mddev, struct block_device *bdev)
+static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	multipath_conf_t *conf = mddev_to_conf(mddev);
-	struct multipath_info * multipaths = conf->multipaths;
-	int disks = MD_SB_DISKS;
-	int i;
-
 
 	if (conf->working_disks <= 1) {
 		/*
@@ -239,24 +235,21 @@ static int multipath_error (mddev_t *mddev, struct block_device *bdev)
 		 * first check if this is a queued request for a device
 		 * which has just failed.
 		 */
-		for (i = 0; i < disks; i++) {
-			if (multipaths[i].bdev == bdev && !multipaths[i].operational)
-				return 0;
-		}
 		printk (LAST_DISK);
-		return 1; /* leave it active... it's all we have */
+		/* leave it active... it's all we have */
 	} else {
 		/*
 		 * Mark disk as unusable
 		 */
-		for (i = 0; i < disks; i++) {
-			if (multipaths[i].bdev == bdev && multipaths[i].operational) {
-				mark_disk_bad(mddev, i);
-				break;
-			}
+		if (!rdev->faulty) {
+			rdev->in_sync = 0;
+			rdev->faulty = 1;
+			mddev->sb_dirty = 1;
+			conf->working_disks--;
+			printk (DISK_FAILED, bdev_partition_name (rdev->bdev),
+				conf->working_disks);
 		}
 	}
-	return 0;
 }
 
 #undef LAST_DISK
@@ -279,11 +272,10 @@ static void print_multipath_conf (multipath_conf_t *conf)
 
 	for (i = 0; i < MD_SB_DISKS; i++) {
 		tmp = conf->multipaths + i;
-		if (tmp->operational || tmp->used_slot)
-			printk(" disk%d, o:%d, us:%d dev:%s\n",
-				i,tmp->operational,
-				tmp->used_slot,
-				bdev_partition_name(tmp->bdev));
+		if (tmp->rdev)
+			printk(" disk%d, o:%d, dev:%s\n",
+				i,!tmp->rdev->faulty,
+			       bdev_partition_name(tmp->rdev->bdev));
 	}
 }
 
@@ -291,24 +283,23 @@ static void print_multipath_conf (multipath_conf_t *conf)
 static int multipath_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	multipath_conf_t *conf = mddev->private;
-	int err = 1;
-	struct multipath_info *p = conf->multipaths + rdev->raid_disk;
+	int found = 0;
+	int path;
+	struct multipath_info *p;
 
 	print_multipath_conf(conf);
 	spin_lock_irq(&conf->device_lock);
-	if (!p->used_slot) {
-		p->bdev = rdev->bdev;
-		p->operational = 1;
-		p->used_slot = 1;
+	for (path=0; path<mddev->raid_disks; path++) 
+		if ((p=conf->multipaths+path)->rdev == NULL) {
+			p->rdev = rdev;
 			conf->working_disks++;
-		err = 0;
+			rdev->raid_disk = path;
+			found = 1;
 		}
-	if (err)
-		MD_BUG();
 	spin_unlock_irq(&conf->device_lock);
 
 	print_multipath_conf(conf);
-	return err;
+	return found;
 }
 
 static int multipath_remove_disk(mddev_t *mddev, int number)
@@ -320,14 +311,14 @@ static int multipath_remove_disk(mddev_t *mddev, int number)
 	print_multipath_conf(conf);
 	spin_lock_irq(&conf->device_lock);
 
-	if (p->used_slot) {
-		if (p->operational) {
+	if (p->rdev) {
+		if (p->rdev->in_sync ||
+		    atomic_read(&p->rdev->nr_pending)) {
 			printk(KERN_ERR "hot-remove-disk, slot %d is identified but is still operational!\n", number);
 			err = -EBUSY;
 			goto abort;
 		}
-		p->bdev = NULL;
-		p->used_slot = 0;
+		p->rdev = NULL;
 		err = 0;
 	}
 	if (err)
@@ -359,7 +350,7 @@ static void multipathd (void *data)
 	struct bio *bio;
 	unsigned long flags;
 	mddev_t *mddev;
-	struct block_device *bdev;
+	mdk_rdev_t *rdev;
 
 	for (;;) {
 		spin_lock_irqsave(&retry_list_lock, flags);
@@ -372,16 +363,16 @@ static void multipathd (void *data)
 		mddev = mp_bh->mddev;
 		bio = &mp_bh->bio;
 		bio->bi_sector = mp_bh->master_bio->bi_sector;
-		bdev = bio->bi_bdev;
 		
-		multipath_map (mddev, &bio->bi_bdev);
-		if (bio->bi_bdev == bdev) {
+		rdev = NULL;
+		if (multipath_map (mddev, &rdev)<0) {
 			printk(IO_ERROR,
 				bdev_partition_name(bio->bi_bdev), bio->bi_sector);
 			multipath_end_bh_io(mp_bh, 0);
 		} else {
 			printk(REDIRECT_SECTOR,
 				bdev_partition_name(bio->bi_bdev), bio->bi_sector);
+			bio->bi_bdev = rdev->bdev;
 			generic_make_request(bio);
 		}
 	}
@@ -436,7 +427,6 @@ static int multipath_run (mddev_t *mddev)
 	struct multipath_info *disk;
 	mdk_rdev_t *rdev;
 	struct list_head *tmp;
-	int num_rdevs = 0;
 
 	MOD_INC_USE_COUNT;
 
@@ -458,40 +448,20 @@ static int multipath_run (mddev_t *mddev)
 	}
 	memset(conf, 0, sizeof(*conf));
 
+	conf->working_disks = 0;
 	ITERATE_RDEV(mddev,rdev,tmp) {
-		if (rdev->faulty) {
-			/* this is a "should never happen" case and if it */
-			/* ever does happen, a continue; won't help */
-			printk(ERRORS, bdev_partition_name(rdev->bdev));
-			continue;
-		} else {
-			/* this is a "should never happen" case and if it */
-			/* ever does happen, a continue; won't help */
-			if (!rdev->sb) {
-				MD_BUG();
-				continue;
-			}
-		}
-		if (rdev->desc_nr == -1) {
-			MD_BUG();
+		disk_idx = rdev->raid_disk;
+		if (disk_idx < 0 ||
+		    disk_idx >= mddev->raid_disks)
 			continue;
-		}
 
-		disk_idx = rdev->raid_disk;
 		disk = conf->multipaths + disk_idx;
-
-		/*
-		 * Mark all disks as active to start with, there are no
-		 * spares.  multipath_read_balance deals with choose
-		 * the "best" operational device.
-		 */
-		disk->bdev = rdev->bdev;
-		disk->operational = 1;
-		disk->used_slot = 1;
-		num_rdevs++;
+		disk->rdev = rdev;
+		if (!rdev->faulty) 
+			conf->working_disks++;
 	}
 
-	conf->raid_disks = mddev->raid_disks = num_rdevs;
+	conf->raid_disks = mddev->raid_disks;
 	mddev->sb_dirty = 1;
 	conf->mddev = mddev;
 	conf->device_lock = SPIN_LOCK_UNLOCKED;
@@ -500,6 +470,7 @@ static int multipath_run (mddev_t *mddev)
 		printk(NONE_OPERATIONAL, mdidx(mddev));
 		goto out_free_conf;
 	}
+	mddev->degraded = conf->raid_disks = conf->working_disks;
 
 	conf->pool = mempool_create(NR_RESERVED_BUFS,
 				    mp_pool_alloc, mp_pool_free,

drivers/md/raid0.c

@@ -87,7 +87,7 @@ static int create_strip_zones (mddev_t *mddev)
 	cnt = 0;
 	smallest = NULL;
 	ITERATE_RDEV(mddev, rdev1, tmp1) {
-		int j = rdev1->sb->this_disk.raid_disk;
+		int j = rdev1->raid_disk;
 
 		if (j < 0 || j >= mddev->raid_disks) {
 			printk("raid0: bad disk number %d - aborting!\n", j);

drivers/md/raid1.c

@@ -135,7 +135,7 @@ static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
 		bio_put(r1_bio->read_bio);
 		r1_bio->read_bio = NULL;
 	}
-	for (i = 0; i < MD_SB_DISKS; i++) {
+	for (i = 0; i < conf->raid_disks; i++) {
 		struct bio **bio = r1_bio->write_bios + i;
 		if (*bio) {
 			if (atomic_read(&(*bio)->bi_cnt) != 1)
@@ -188,22 +188,27 @@ static inline void put_buf(r1bio_t *r1_bio)
 	mempool_free(r1_bio, conf->r1buf_pool);
 }
 
-static int map(mddev_t *mddev, struct block_device **bdev)
+static int map(mddev_t *mddev, mdk_rdev_t **rdevp)
 {
 	conf_t *conf = mddev_to_conf(mddev);
-	int i, disks = MD_SB_DISKS;
+	int i, disks = conf->raid_disks;
 
 	/*
 	 * Later we do read balancing on the read side
 	 * now we use the first available disk.
 	 */
 
+	spin_lock_irq(&conf->device_lock);
 	for (i = 0; i < disks; i++) {
-		if (conf->mirrors[i].operational) {
-			*bdev = conf->mirrors[i].bdev;
+		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
+		if (rdev && rdev->in_sync) {
+			*rdevp = rdev;
+			atomic_inc(&rdev->nr_pending);
+			spin_unlock_irq(&conf->device_lock);
 			return 0;
 		}
 	}
+	spin_unlock_irq(&conf->device_lock);
 
 	printk (KERN_ERR "raid1_map(): huh, no more operational devices?\n");
 	return -1;
@@ -244,7 +249,6 @@ static void inline update_head_pos(int disk, r1bio_t *r1_bio)
 
 	conf->mirrors[disk].head_position =
 		r1_bio->sector + (r1_bio->master_bio->bi_size >> 9);
-	atomic_dec(&conf->mirrors[disk].nr_pending);
 }
 
 static void end_request(struct bio *bio)
@@ -257,7 +261,7 @@ static void end_request(struct bio *bio)
 	if (r1_bio->cmd == READ || r1_bio->cmd == READA)
 		mirror = r1_bio->read_disk;
 	else {
-		for (mirror = 0; mirror < MD_SB_DISKS; mirror++)
+		for (mirror = 0; mirror < conf->raid_disks; mirror++)
 			if (r1_bio->write_bios[mirror] == bio)
 				break;
 	}
@@ -265,7 +269,7 @@ static void end_request(struct bio *bio)
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
 	if (!uptodate)
-		md_error(r1_bio->mddev, conf->mirrors[mirror].bdev);
+		md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
 	else
 		/*
 		 * Set R1BIO_Uptodate in our master bio, so that
@@ -285,18 +289,17 @@ static void end_request(struct bio *bio)
 		/*
 		 * we have only one bio on the read side
 		 */
-		if (uptodate) {
+		if (uptodate)
 			raid_end_bio_io(r1_bio, uptodate);
-			return;
-		}
+		else {
 			/*
 			 * oops, read error:
 			 */
 			printk(KERN_ERR "raid1: %s: rescheduling sector %lu\n",
-			bdev_partition_name(conf->mirrors[mirror].bdev), r1_bio->sector);
+			       bdev_partition_name(conf->mirrors[mirror].rdev->bdev), r1_bio->sector);
 			reschedule_retry(r1_bio);
-		return;
 		}
+	} else {
 
 		if (r1_bio->read_bio)
 			BUG();
@@ -308,6 +311,8 @@ static void end_request(struct bio *bio)
 		 */
 		if (atomic_dec_and_test(&r1_bio->remaining))
 			raid_end_bio_io(r1_bio, uptodate);
+	}
+	atomic_dec(&conf->mirrors[mirror].rdev->nr_pending);
 }
 
 /*
@@ -321,6 +326,8 @@ static void end_request(struct bio *bio)
  *
  * If there are 2 mirrors in the same 2 devices, performance degrades
  * because position is mirror, not device based.
+ *
+ * The rdev for the device selected will have nr_pending incremented.
  */
 static int read_balance(conf_t *conf, struct bio *bio, r1bio_t *r1_bio)
 {
@@ -329,6 +336,7 @@ static int read_balance(conf_t *conf, struct bio *bio, r1bio_t *r1_bio)
 	const int sectors = bio->bi_size >> 9;
 	sector_t new_distance, current_distance;
 
+	spin_lock_irq(&conf->device_lock);
 	/*
 	 * Check if it if we can balance. We can balance on the whole
 	 * device if no resync is going on, or below the resync window.
@@ -337,7 +345,9 @@ static int read_balance(conf_t *conf, struct bio *bio, r1bio_t *r1_bio)
 	if (!conf->mddev->in_sync && (this_sector + sectors >= conf->next_resync)) {
 		/* make sure that disk is operational */
 		new_disk = 0;
-		while (!conf->mirrors[new_disk].operational || conf->mirrors[new_disk].write_only) {
+
+		while (!conf->mirrors[new_disk].rdev ||
+		       !conf->mirrors[new_disk].rdev->in_sync) {
 			new_disk++;
 			if (new_disk == conf->raid_disks) {
 				new_disk = 0;
@@ -349,7 +359,8 @@ static int read_balance(conf_t *conf, struct bio *bio, r1bio_t *r1_bio)
 
 
 	/* make sure the disk is operational */
-	while (!conf->mirrors[new_disk].operational) {
+	while (!conf->mirrors[new_disk].rdev ||
+	       !conf->mirrors[new_disk].rdev->in_sync) {
 		if (new_disk <= 0)
 			new_disk = conf->raid_disks;
 		new_disk--;
@@ -378,11 +389,11 @@ static int read_balance(conf_t *conf, struct bio *bio, r1bio_t *r1_bio)
 			disk = conf->raid_disks;
 		disk--;
 
-		if ((conf->mirrors[disk].write_only) ||
-				(!conf->mirrors[disk].operational))
+		if (!conf->mirrors[disk].rdev ||
+		    !conf->mirrors[disk].rdev->in_sync)
 			continue;
 
-		if (!atomic_read(&conf->mirrors[disk].nr_pending)) {
+		if (!atomic_read(&conf->mirrors[disk].rdev->nr_pending)) {
 			new_disk = disk;
 			break;
 		}
@@ -399,6 +410,10 @@ static int read_balance(conf_t *conf, struct bio *bio, r1bio_t *r1_bio)
 
 	conf->last_used = new_disk;
 
+	if (conf->mirrors[new_disk].rdev)
+		atomic_inc(&conf->mirrors[new_disk].rdev->nr_pending);
+	spin_unlock_irq(&conf->device_lock);
+
 	return new_disk;
 }
 
@@ -441,7 +456,7 @@ static int make_request(request_queue_t *q, struct bio * bio)
 	mirror_info_t *mirror;
 	r1bio_t *r1_bio;
 	struct bio *read_bio;
-	int i, sum_bios = 0, disks = MD_SB_DISKS;
+	int i, sum_bios = 0, disks = conf->raid_disks;
 
 	/*
 	 * Register the new request and wait if the reconstruction
@@ -478,31 +493,42 @@ static int make_request(request_queue_t *q, struct bio * bio)
 		r1_bio->read_bio = read_bio;
 
 		read_bio->bi_sector = r1_bio->sector;
-		read_bio->bi_bdev = mirror->bdev;
+		read_bio->bi_bdev = mirror->rdev->bdev;
 		read_bio->bi_end_io = end_request;
 		read_bio->bi_rw = r1_bio->cmd;
 		read_bio->bi_private = r1_bio;
 
 		generic_make_request(read_bio);
-		atomic_inc(&conf->mirrors[r1_bio->read_disk].nr_pending);
 		return 0;
 	}
 
 	/*
 	 * WRITE:
 	 */
+	/* first select target devices under spinlock and
+	 * inc refcount on their rdev.  Record them by setting
+	 * write_bios[x] to bio
+	 */
+	spin_lock_irq(&conf->device_lock);
+	for (i = 0;  i < disks; i++) {
+		if (conf->mirrors[i].rdev &&
+		    !conf->mirrors[i].rdev->faulty) {
+			atomic_inc(&conf->mirrors[i].rdev->nr_pending);
+			r1_bio->write_bios[i] = bio;
+		} else
+			r1_bio->write_bios[i] = NULL;
+	}
+	spin_unlock_irq(&conf->device_lock);
 	for (i = 0; i < disks; i++) {
 		struct bio *mbio;
-		if (!conf->mirrors[i].operational)
+		if (!r1_bio->write_bios[i])
 			continue;
 
 		mbio = bio_clone(bio, GFP_NOIO);
-		if (r1_bio->write_bios[i])
-			BUG();
 		r1_bio->write_bios[i] = mbio;
 
 		mbio->bi_sector	= r1_bio->sector;
-		mbio->bi_bdev = conf->mirrors[i].bdev;
+		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
 		mbio->bi_end_io	= end_request;
 		mbio->bi_rw = r1_bio->cmd;
 		mbio->bi_private = r1_bio;
@@ -529,14 +555,13 @@ static int make_request(request_queue_t *q, struct bio * bio)
 	 * do end_request by hand if all requests finish until we had a
 	 * chance to set up the semaphore correctly ... lots of races).
 	 */
-	for (i = 0; i < disks; i++) {
+	for (i=disks; i--; ) {
 		struct bio *mbio;
 		mbio = r1_bio->write_bios[i];
 		if (!mbio)
 			continue;
 
 		generic_make_request(mbio);
-		atomic_inc(&conf->mirrors[i].nr_pending);
 	}
 	return 0;
 }
@@ -550,7 +575,8 @@ static int status(char *page, mddev_t *mddev)
 						conf->working_disks);
 	for (i = 0; i < conf->raid_disks; i++)
 		sz += sprintf(page+sz, "%s",
-			conf->mirrors[i].operational ? "U" : "_");
+			      conf->mirrors[i].rdev &&
+			      conf->mirrors[i].rdev->in_sync ? "U" : "_");
 	sz += sprintf (page+sz, "]");
 	return sz;
 }
@@ -571,48 +597,37 @@ static int status(char *page, mddev_t *mddev)
 #define ALREADY_SYNCING KERN_INFO \
 "raid1: syncing already in progress.\n"
 
-static void mark_disk_bad(mddev_t *mddev, int failed)
-{
-	conf_t *conf = mddev_to_conf(mddev);
-	mirror_info_t *mirror = conf->mirrors+failed;
 
-	mirror->operational = 0;
-	if (!mirror->write_only) {
-		mddev->degraded++;
-		conf->working_disks--;
-	}
-	mddev->sb_dirty = 1;
-	printk(DISK_FAILED, bdev_partition_name(mirror->bdev), conf->working_disks);
-}
-
-static int error(mddev_t *mddev, struct block_device *bdev)
+static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	conf_t *conf = mddev_to_conf(mddev);
-	mirror_info_t * mirrors = conf->mirrors;
-	int disks = MD_SB_DISKS;
-	int i;
 
 	/*
-	 * Find the drive.
 	 * If it is not operational, then we have already marked it as dead
 	 * else if it is the last working disks, ignore the error, let the
 	 * next level up know.
 	 * else mark the drive as failed
 	 */
-	for (i = 0; i < disks; i++)
-		if (mirrors[i].bdev == bdev && mirrors[i].operational)
-			break;
-	if (i == disks)
-		return 0;
-
-	if (i < conf->raid_disks && conf->working_disks == 1)
+	if (rdev->in_sync
+	    && conf->working_disks == 1)
 		/*
 		 * Don't fail the drive, act as though we were just a
 		 * normal single drive
 		 */
-		return 1;
-	mark_disk_bad(mddev, i);
-	return 0;
+		return;
+	if (rdev->in_sync) {
+		mddev->degraded++;
+		conf->working_disks--;
+		/*
+		 * if recovery was running, stop it now.
+		 */
+		if (mddev->recovery_running) 
+			mddev->recovery_running = -EIO;
+	}
+	rdev->in_sync = 0;
+	rdev->faulty = 1;
+	mddev->sb_dirty = 1;
+	printk(DISK_FAILED, bdev_partition_name(rdev->bdev), conf->working_disks);
 }
 
 static void print_conf(conf_t *conf)
@@ -628,12 +643,12 @@ static void print_conf(conf_t *conf)
 	printk(" --- wd:%d rd:%d\n", conf->working_disks,
 			conf->raid_disks);
 
-	for (i = 0; i < MD_SB_DISKS; i++) {
+	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->mirrors + i;
-		printk(" disk %d, s:%d, o:%d, us:%d dev:%s\n",
-			i, tmp->spare, tmp->operational,
-			tmp->used_slot,
-			bdev_partition_name(tmp->bdev));
+		if (tmp->rdev)
+			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
+			       i, !tmp->rdev->in_sync, !tmp->rdev->faulty,
+			       bdev_partition_name(tmp->rdev->bdev));
 	}
 }
 
@@ -653,160 +668,52 @@ static void close_sync(conf_t *conf)
 
 static int raid1_spare_active(mddev_t *mddev)
 {
-	int err = 0;
-	int i, failed_disk = -1, spare_disk = -1;
+	int i;
 	conf_t *conf = mddev->private;
-	mirror_info_t *tmp, *sdisk, *fdisk;
-	mdk_rdev_t *spare_rdev, *failed_rdev;
+	mirror_info_t *tmp;
 
-	print_conf(conf);
 	spin_lock_irq(&conf->device_lock);
 	/*
-	 * Find the failed disk within the RAID1 configuration ...
-	 * (this can only be in the first conf->working_disks part)
+	 * Find all failed disks within the RAID1 configuration 
+	 * and mark them readable
 	 */
 	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->mirrors + i;
-		if ((!tmp->operational && !tmp->spare) ||
-				!tmp->used_slot) {
-			failed_disk = i;
-			break;
-		}
-	}
-	/*
-	 * When we activate a spare disk we _must_ have a disk in
-	 * the lower (active) part of the array to replace.
-	 */
-	if (failed_disk == -1) {
-		MD_BUG();
-		err = 1;
-		goto abort;
-	}
-	/*
-	 * Find the spare disk ... (can only be in the 'high'
-	 * area of the array)
-	 */
-	spare_disk = mddev->spare->raid_disk;
-
-	sdisk = conf->mirrors + spare_disk;
-	fdisk = conf->mirrors + failed_disk;
-
-	/*
-	 * do the switch finally
-	 */
-	spare_rdev = find_rdev_nr(mddev, spare_disk);
-	failed_rdev = find_rdev_nr(mddev, failed_disk);
-
-	/*
-	 * There must be a spare_rdev, but there may not be a
-	 * failed_rdev. That slot might be empty...
-	 */
-	spare_rdev->desc_nr = failed_disk;
-	spare_rdev->raid_disk = failed_disk;
-	if (failed_rdev) {
-		failed_rdev->desc_nr = spare_disk;
-		failed_rdev->raid_disk = spare_disk;
-	}
-
-	xchg_values(*fdisk, *sdisk);
-
-	/*
-	 * (careful, 'failed' and 'spare' are switched from now on)
-	 *
-	 * we want to preserve linear numbering and we want to
-	 * give the proper raid_disk number to the now activated
-	 * disk. (this means we switch back these values)
-	 */
-
-	if (!sdisk->bdev)
-		sdisk->used_slot = 0;
-	/*
-	 * this really activates the spare.
-	 */
-	fdisk->spare = 0;
-	fdisk->write_only = 0;
-
-	/*
-	 * if we activate a spare, we definitely replace a
-	 * non-operational disk slot in the 'low' area of
-	 * the disk array.
-	 */
-
+		if (tmp->rdev 
+		    && !tmp->rdev->faulty
+		    && !tmp->rdev->in_sync) {
 			conf->working_disks++;
 			mddev->degraded--;
-abort:
-	spin_unlock_irq(&conf->device_lock);
-
-	print_conf(conf);
-	return err;
-}
-
-static int raid1_spare_inactive(mddev_t *mddev)
-{
-	conf_t *conf = mddev->private;
-	mirror_info_t *p;
-	int err = 0;
-
-	print_conf(conf);
-	spin_lock_irq(&conf->device_lock);
-	p = conf->mirrors + mddev->spare->raid_disk;
-	if (p) {
-		p->operational = 0;
-		p->write_only = 0;
-	} else {
-		MD_BUG();
-		err = 1;
+			tmp->rdev->in_sync = 1;
 		}
-	spin_unlock_irq(&conf->device_lock);
-	print_conf(conf);
-	return err;
-}
-
-static int raid1_spare_write(mddev_t *mddev)
-{
-	conf_t *conf = mddev->private;
-	mirror_info_t *p;
-	int err = 0;
-
-	print_conf(conf);
-	spin_lock_irq(&conf->device_lock);
-	p = conf->mirrors + mddev->spare->raid_disk;
-	if (p) {
-		p->operational = 1;
-		p->write_only = 1;
-	} else {
-		MD_BUG();
-		err = 1;
 	}
 	spin_unlock_irq(&conf->device_lock);
+
 	print_conf(conf);
-	return err;
+	return 0;
 }
 
+
 static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	conf_t *conf = mddev->private;
-	int err = 1;
-	mirror_info_t *p = conf->mirrors + rdev->raid_disk;
+	int found = 0;
+	int mirror;
+	mirror_info_t *p;
 
-	print_conf(conf);
 	spin_lock_irq(&conf->device_lock);
-	if (!p->used_slot) {
-		/* it will be held open by rdev */
-		p->bdev = rdev->bdev;
-		p->operational = 0;
-		p->write_only = 0;
-		p->spare = 1;
-		p->used_slot = 1;
+	for (mirror=0; mirror < mddev->raid_disks; mirror++)
+		if ( !(p=conf->mirrors+mirror)->rdev) {
+			p->rdev = rdev;
 			p->head_position = 0;
-		err = 0;
+			rdev->raid_disk = mirror;
+			found = 1;
+			break;
 		}
-	if (err)
-		MD_BUG();
 	spin_unlock_irq(&conf->device_lock);
 
 	print_conf(conf);
-	return err;
+	return found;
 }
 
 static int raid1_remove_disk(mddev_t *mddev, int number)
@@ -817,13 +724,13 @@ static int raid1_remove_disk(mddev_t *mddev, int number)
 
 	print_conf(conf);
 	spin_lock_irq(&conf->device_lock);
-	if (p->used_slot) {
-		if (p->operational) {
+	if (p->rdev) {
+		if (p->rdev->in_sync ||
+		    atomic_read(&p->rdev->nr_pending)) {
 			err = -EBUSY;
 			goto abort;
 		}
-		p->bdev = NULL;
-		p->used_slot = 0;
+		p->rdev = NULL;
 		err = 0;
 	}
 	if (err)
@@ -857,9 +764,10 @@ static void end_sync_read(struct bio *bio)
 	 */
 	if (!uptodate)
 		md_error(r1_bio->mddev,
-			 conf->mirrors[r1_bio->read_disk].bdev);
+			 conf->mirrors[r1_bio->read_disk].rdev);
 	else
 		set_bit(R1BIO_Uptodate, &r1_bio->state);
+	atomic_dec(&conf->mirrors[r1_bio->read_disk].rdev->nr_pending);
 	reschedule_retry(r1_bio);
 }
 
@@ -872,13 +780,13 @@ static void end_sync_write(struct bio *bio)
 	int i;
 	int mirror=0;
 
-	for (i = 0; i < MD_SB_DISKS; i++)
+	for (i = 0; i < conf->raid_disks; i++)
 		if (r1_bio->write_bios[i] == bio) {
 			mirror = i;
 			break;
 		}
 	if (!uptodate)
-		md_error(mddev, conf->mirrors[mirror].bdev);
+		md_error(mddev, conf->mirrors[mirror].rdev);
 	update_head_pos(mirror, r1_bio);
 
 	if (atomic_dec_and_test(&r1_bio->remaining)) {
@@ -886,13 +794,14 @@ static void end_sync_write(struct bio *bio)
 		resume_device(conf);
 		put_buf(r1_bio);
 	}
+	atomic_dec(&conf->mirrors[mirror].rdev->nr_pending);
 }
 
 static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 {
 	conf_t *conf = mddev_to_conf(mddev);
 	int i, sum_bios = 0;
-	int disks = MD_SB_DISKS;
+	int disks = conf->raid_disks;
 	struct bio *bio, *mbio;
 
 	bio = r1_bio->master_bio;
@@ -913,25 +822,33 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 		return;
 	}
 
+	spin_lock_irq(&conf->device_lock);
 	for (i = 0; i < disks ; i++) {
-		if (!conf->mirrors[i].operational)
+		r1_bio->write_bios[i] = NULL;
+		if (!conf->mirrors[i].rdev || 
+		    conf->mirrors[i].rdev->faulty)
 			continue;
 		if (i == conf->last_used)
 			/*
 			 * we read from here, no need to write
 			 */
 			continue;
-		if (i < conf->raid_disks && mddev->in_sync)
+		if (conf->mirrors[i].rdev->in_sync && mddev->in_sync)
 			/*
 			 * don't need to write this we are just rebuilding
 			 */
 			continue;
+		atomic_inc(&conf->mirrors[i].rdev->nr_pending);
+		r1_bio->write_bios[i] = bio;
+	}
+	spin_unlock_irq(&conf->device_lock);
 
+	for (i = 0; i < disks ; i++) {
+		if (!r1_bio->write_bios[i])
+			continue;
 		mbio = bio_clone(bio, GFP_NOIO);
-		if (r1_bio->write_bios[i])
-			BUG();
 		r1_bio->write_bios[i] = mbio;
-		mbio->bi_bdev = conf->mirrors[i].bdev;
+		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
 		mbio->bi_sector = r1_bio->sector;
 		mbio->bi_end_io	= end_sync_write;
 		mbio->bi_rw = WRITE;
@@ -949,7 +866,7 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 		 * Nowhere to write this to... I guess we
 		 * must be done
 		 */
-		printk(IO_ERROR, bdev_partition_name(bio->bi_bdev), r1_bio->sector);
+		printk(KERN_ALERT "raid1: sync aborting as there is nowhere to write sector %lu\n", r1_bio->sector);
 		md_done_sync(mddev, r1_bio->master_bio->bi_size >> 9, 0);
 		resume_device(conf);
 		put_buf(r1_bio);
@@ -960,9 +877,8 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 		if (!mbio)
 			continue;
 
-		md_sync_acct(mbio->bi_bdev, mbio->bi_size >> 9);
+		md_sync_acct(conf->mirrors[i].rdev, mbio->bi_size >> 9);
 		generic_make_request(mbio);
-		atomic_inc(&conf->mirrors[i].nr_pending);
 	}
 }
 
@@ -982,7 +898,7 @@ static void raid1d(void *data)
 	unsigned long flags;
 	mddev_t *mddev;
 	conf_t *conf;
-	struct block_device *bdev;
+	mdk_rdev_t *rdev;
 
 
 	for (;;) {
@@ -1002,20 +918,18 @@ static void raid1d(void *data)
 			break;
 		case READ:
 		case READA:
-			bdev = bio->bi_bdev;
-			map(mddev, &bio->bi_bdev);
-			if (bio->bi_bdev == bdev) {
+			if (map(mddev, &rdev) == -1) {
 				printk(IO_ERROR, bdev_partition_name(bio->bi_bdev), r1_bio->sector);
 				raid_end_bio_io(r1_bio, 0);
 				break;
 			}
 			printk(REDIRECT_SECTOR,
-				bdev_partition_name(bio->bi_bdev), r1_bio->sector);
+				bdev_partition_name(rdev->bdev), r1_bio->sector);
+			bio->bi_bdev = rdev->bdev;
 			bio->bi_sector = r1_bio->sector;
 			bio->bi_rw = r1_bio->cmd;
 
 			generic_make_request(bio);
-			atomic_inc(&conf->mirrors[r1_bio->read_disk].nr_pending);
 			break;
 		}
 	}
@@ -1081,7 +995,9 @@ static int sync_request(mddev_t *mddev, sector_t sector_nr, int go_faster)
 	 */
 	disk = conf->last_used;
 	/* make sure disk is operational */
-	while (!conf->mirrors[disk].operational) {
+	spin_lock_irq(&conf->device_lock);
+	while (conf->mirrors[disk].rdev == NULL ||
+	       !conf->mirrors[disk].rdev->in_sync) {
 		if (disk <= 0)
 			disk = conf->raid_disks;
 		disk--;
@@ -1089,6 +1005,8 @@ static int sync_request(mddev_t *mddev, sector_t sector_nr, int go_faster)
 			break;
 	}
 	conf->last_used = disk;
+	atomic_inc(&conf->mirrors[disk].rdev->nr_pending);
+	spin_unlock_irq(&conf->device_lock);
 
 	mirror = conf->mirrors + conf->last_used;
 
@@ -1119,7 +1037,7 @@ static int sync_request(mddev_t *mddev, sector_t sector_nr, int go_faster)
 	read_bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
 
 	read_bio->bi_sector = sector_nr;
-	read_bio->bi_bdev = mirror->bdev;
+	read_bio->bi_bdev = mirror->rdev->bdev;
 	read_bio->bi_end_io = end_sync_read;
 	read_bio->bi_rw = READ;
 	read_bio->bi_private = r1_bio;
@@ -1128,10 +1046,9 @@ static int sync_request(mddev_t *mddev, sector_t sector_nr, int go_faster)
 		BUG();
 	r1_bio->read_bio = read_bio;
 
-	md_sync_acct(read_bio->bi_bdev, nr_sectors);
+	md_sync_acct(mirror->rdev, nr_sectors);
 
 	generic_make_request(read_bio);
-	atomic_inc(&conf->mirrors[conf->last_used].nr_pending);
 
 	return nr_sectors;
 }
@@ -1209,61 +1126,18 @@ static int run(mddev_t *mddev)
 		goto out;
 	}
 
-//	for (tmp = (mddev)->disks.next; rdev = ((mdk_rdev_t *)((char *)(tmp)-(unsigned long)(&((mdk_rdev_t *)0)->same_set))), tmp = tmp->next, tmp->prev != &(mddev)->disks ; ) {
 
 	ITERATE_RDEV(mddev, rdev, tmp) {
-		if (rdev->faulty) {
-			printk(ERRORS, bdev_partition_name(rdev->bdev));
-		} else {
-			if (!rdev->sb) {
-				MD_BUG();
-				continue;
-			}
-		}
-		if (rdev->desc_nr == -1) {
-			MD_BUG();
-			continue;
-		}
 		disk_idx = rdev->raid_disk;
+		if (disk_idx >= mddev->raid_disks
+		    || disk_idx < 0)
+			continue;
 		disk = conf->mirrors + disk_idx;
 
-		if (rdev->faulty) {
-			disk->bdev = rdev->bdev;
-			disk->operational = 0;
-			disk->write_only = 0;
-			disk->spare = 0;
-			disk->used_slot = 1;
-			disk->head_position = 0;
-			continue;
-		}
-		if (rdev->in_sync) {
-			if (disk->operational) {
-				printk(ALREADY_RUNNING,
-					bdev_partition_name(rdev->bdev),
-					disk_idx);
-				continue;
-			}
-			printk(OPERATIONAL, bdev_partition_name(rdev->bdev),
-					disk_idx);
-			disk->bdev = rdev->bdev;
-			disk->operational = 1;
-			disk->write_only = 0;
-			disk->spare = 0;
-			disk->used_slot = 1;
+		disk->rdev = rdev;
 		disk->head_position = 0;
+		if (!rdev->faulty && rdev->in_sync)
 			conf->working_disks++;
-		} else {
-		/*
-		 * Must be a spare disk ..
-		 */
-			printk(SPARE, bdev_partition_name(rdev->bdev));
-			disk->bdev = rdev->bdev;
-			disk->operational = 0;
-			disk->write_only = 0;
-			disk->spare = 1;
-			disk->used_slot = 1;
-			disk->head_position = 0;
-		}
 	}
 	conf->raid_disks = mddev->raid_disks;
 	conf->mddev = mddev;
@@ -1283,23 +1157,19 @@ static int run(mddev_t *mddev)
 
 		disk = conf->mirrors + i;
 
-		if (!disk->used_slot) {
-			disk->bdev = NULL;
-			disk->operational = 0;
-			disk->write_only = 0;
-			disk->spare = 0;
-			disk->used_slot = 1;
+		if (!disk->rdev) {
 			disk->head_position = 0;
-		}
-		if (!disk->used_slot)
 			mddev->degraded++;
 		}
+	}
 
 	/*
 	 * find the first working one and use it as a starting point
 	 * to read balancing.
 	 */
-	for (j = 0; !conf->mirrors[j].operational && j < MD_SB_DISKS; j++)
+	for (j = 0; j < conf->raid_disks &&
+		     (!conf->mirrors[j].rdev ||
+		      !conf->mirrors[j].rdev->in_sync) ; j++)
 		/* nothing */;
 	conf->last_used = j;
 
@@ -1354,8 +1224,6 @@ static mdk_personality_t raid1_personality =
 	.error_handler	= error,
 	.hot_add_disk	= raid1_add_disk,
 	.hot_remove_disk= raid1_remove_disk,
-	.spare_write	= raid1_spare_write,
-	.spare_inactive	= raid1_spare_inactive,
 	.spare_active	= raid1_spare_active,
 	.sync_request	= sync_request,
 };

drivers/md/raid5.c

@@ -371,9 +371,10 @@ static void raid5_end_read_request (struct bio * bi)
 		set_bit(R5_UPTODATE, &sh->dev[i].flags);
 #endif		
 	} else {
-		md_error(conf->mddev, conf->disks[i].bdev);
+		md_error(conf->mddev, conf->disks[i].rdev);
 		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
 	}
+	atomic_dec(&conf->disks[i].rdev->nr_pending);
 #if 0
 	/* must restore b_page before unlocking buffer... */
 	if (sh->bh_page[i] != bh->b_page) {
@@ -407,7 +408,9 @@ static void raid5_end_write_request (struct bio *bi)
 
 	spin_lock_irqsave(&conf->device_lock, flags);
 	if (!uptodate)
-		md_error(conf->mddev, conf->disks[i].bdev);
+		md_error(conf->mddev, conf->disks[i].rdev);
+
+	atomic_dec(&conf->disks[i].rdev->nr_pending);
 	
 	clear_bit(R5_LOCKED, &sh->dev[i].flags);
 	set_bit(STRIPE_HANDLE, &sh->state);
@@ -420,7 +423,6 @@ static unsigned long compute_blocknr(struct stripe_head *sh, int i);
 	
 static void raid5_build_block (struct stripe_head *sh, int i)
 {
-	raid5_conf_t *conf = sh->raid_conf;
 	struct r5dev *dev = &sh->dev[i];
 
 	bio_init(&dev->req);
@@ -430,7 +432,6 @@ static void raid5_build_block (struct stripe_head *sh, int i)
 	dev->vec.bv_len = STRIPE_SIZE;
 	dev->vec.bv_offset = 0;
 
-	dev->req.bi_bdev = conf->disks[i].bdev;
 	dev->req.bi_sector = sh->sector;
 	dev->req.bi_private = sh;
 
@@ -439,54 +440,30 @@ static void raid5_build_block (struct stripe_head *sh, int i)
 		dev->sector = compute_blocknr(sh, i);
 }
 
-static int error(mddev_t *mddev, struct block_device *bdev)
+static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
-	struct disk_info *disk;
-	int i;
-
 	PRINTK("raid5: error called\n");
 
-	for (i = 0, disk = conf->disks; i < conf->raid_disks; i++, disk++) {
-		if (disk->bdev != bdev)
-			continue;
-		if (disk->operational) {
-			disk->operational = 0;
+	if (!rdev->faulty) {
 		mddev->sb_dirty = 1;
-			mddev->degraded++;
 		conf->working_disks--;
+		if (rdev->in_sync) {
+			mddev->degraded++;
 			conf->failed_disks++;
-			printk (KERN_ALERT
-				"raid5: Disk failure on %s, disabling device."
-				" Operation continuing on %d devices\n",
-				bdev_partition_name(bdev), conf->working_disks);
-		}
-		return 0;
-	}
+			rdev->in_sync = 0;
 			/*
-	 * handle errors in spares (during reconstruction)
+			 * if recovery was running, stop it now.
 			 */
-	if (conf->spare) {
-		disk = conf->spare;
-		if (disk->bdev == bdev) {
-			printk (KERN_ALERT
-				"raid5: Disk failure on spare %s\n",
-				bdev_partition_name (bdev));
-			if (!conf->spare->operational) {
-				/* probably a SET_DISK_FAULTY ioctl */
-				return -EIO;
-			}
-			disk->operational = 0;
-			disk->write_only = 0;
-			conf->spare = NULL;
-
-			mddev->sb_dirty = 1;
-
-			return 0;
+			if (mddev->recovery_running) 
+				mddev->recovery_running = -EIO;
 		}
+		rdev->faulty = 1;
+		printk (KERN_ALERT
+			"raid5: Disk failure on %s, disabling device."
+			" Operation continuing on %d devices\n",
+			bdev_partition_name(rdev->bdev), conf->working_disks);
 	}
-	MD_BUG();
-	return -EIO;
 }	
 
 /*
@@ -681,7 +658,7 @@ static void compute_block(struct stripe_head *sh, int dd_idx)
 	}
 	if (count != 1)
 		xor_block(count, STRIPE_SIZE, ptr);
-	set_bit(R5_UPTODATE, &sh->dev[i].flags);
+	set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags);
 }
 
 static void compute_parity(struct stripe_head *sh, int method)
@@ -840,7 +817,6 @@ static void handle_stripe(struct stripe_head *sh)
 	int disks = conf->raid_disks;
 	struct bio *return_bi= NULL;
 	struct bio *bi;
-	int action[MD_SB_DISKS];
 	int i;
 	int syncing;
 	int locked=0, uptodate=0, to_read=0, to_write=0, failed=0, written=0;
@@ -848,7 +824,6 @@ static void handle_stripe(struct stripe_head *sh)
 	struct r5dev *dev;
 
 	PRINTK("handling stripe %ld, cnt=%d, pd_idx=%d\n", sh->sector, atomic_read(&sh->count), sh->pd_idx);
-	memset(action, 0, sizeof(action));
 
 	spin_lock(&sh->lock);
 	clear_bit(STRIPE_HANDLE, &sh->state);
@@ -858,7 +833,13 @@ static void handle_stripe(struct stripe_head *sh)
 	/* Now to look around and see what can be done */
 
 	for (i=disks; i--; ) {
+		mdk_rdev_t *rdev;
 		dev = &sh->dev[i];
+		clear_bit(R5_Wantread, &dev->flags);
+		clear_bit(R5_Wantwrite, &dev->flags);
+		clear_bit(R5_Insync, &dev->flags);
+		clear_bit(R5_Syncio, &dev->flags);
+
 		PRINTK("check %d: state 0x%lx read %p write %p written %p\n", i, 
 		       dev->flags, dev->toread, dev->towrite, dev->written);
 		/* maybe we can reply to a read */
@@ -890,10 +871,12 @@ static void handle_stripe(struct stripe_head *sh)
 		if (dev->toread) to_read++;
 		if (dev->towrite) to_write++;
 		if (dev->written) written++;
-		if (!conf->disks[i].operational) {
+		rdev = conf->disks[i].rdev; /* FIXME, should I be looking rdev */
+		if (!rdev || !rdev->in_sync) {
 			failed++;
 			failed_num = i;
-		}
+		} else
+			set_bit(R5_Insync, &dev->flags);
 	}
 	PRINTK("locked=%d uptodate=%d to_read=%d to_write=%d failed=%d failed_num=%d\n",
 	       locked, uptodate, to_read, to_write, failed, failed_num);
@@ -918,7 +901,7 @@ static void handle_stripe(struct stripe_head *sh)
 				bi = nextbi;
 			}
 			/* fail any reads if this device is non-operational */
-			if (!conf->disks[i].operational) {
+			if (!test_bit(R5_Insync, &sh->dev[i].flags)) {
 				bi = sh->dev[i].toread;
 				sh->dev[i].toread = NULL;
 				if (bi) to_read--;
@@ -946,7 +929,7 @@ static void handle_stripe(struct stripe_head *sh)
 	 */
 	dev = &sh->dev[sh->pd_idx];
 	if ( written &&
-	     ( (conf->disks[sh->pd_idx].operational && !test_bit(R5_LOCKED, &dev->flags) &&
+	     ( (test_bit(R5_Insync, &dev->flags) && !test_bit(R5_LOCKED, &dev->flags) &&
 		test_bit(R5_UPTODATE, &dev->flags))
 	       || (failed == 1 && failed_num == sh->pd_idx))
 	    ) {
@@ -954,7 +937,7 @@ static void handle_stripe(struct stripe_head *sh)
 	    for (i=disks; i--; )
 		if (sh->dev[i].written) {
 		    dev = &sh->dev[i];
-		    if (!conf->disks[sh->pd_idx].operational ||
+		    if (!test_bit(R5_Insync, &dev->flags) &&
 			(!test_bit(R5_LOCKED, &dev->flags) && test_bit(R5_UPTODATE, &dev->flags)) ) {
 			/* maybe we can return some write requests */
 			    struct bio *wbi, *wbi2;
@@ -988,9 +971,9 @@ static void handle_stripe(struct stripe_head *sh)
 					PRINTK("Computing block %d\n", i);
 					compute_block(sh, i);
 					uptodate++;
-				} else if (conf->disks[i].operational) {
+				} else if (test_bit(R5_Insync, &dev->flags)) {
 					set_bit(R5_LOCKED, &dev->flags);
-					action[i] = READ+1;
+					set_bit(R5_Wantread, &dev->flags);
 #if 0
 					/* if I am just reading this block and we don't have
 					   a failed drive, or any pending writes then sidestep the cache */
@@ -1003,7 +986,7 @@ static void handle_stripe(struct stripe_head *sh)
 					locked++;
 					PRINTK("Reading block %d (sync=%d)\n", i, syncing);
 					if (syncing)
-						md_sync_acct(conf->disks[i].bdev, STRIPE_SECTORS);
+						md_sync_acct(conf->disks[i].rdev, STRIPE_SECTORS);
 				}
 			}
 		}
@@ -1023,7 +1006,7 @@ static void handle_stripe(struct stripe_head *sh)
 #endif
 				    ) &&
 			    !test_bit(R5_UPTODATE, &dev->flags)) {
-				if (conf->disks[i].operational 
+				if (test_bit(R5_Insync, &dev->flags)
 /*				    && !(!mddev->insync && i == sh->pd_idx) */
 					)
 					rmw++;
@@ -1037,7 +1020,7 @@ static void handle_stripe(struct stripe_head *sh)
 #endif
 				    ) &&
 			    !test_bit(R5_UPTODATE, &dev->flags)) {
-				if (conf->disks[i].operational) rcw++;
+				if (test_bit(R5_Insync, &dev->flags)) rcw++;
 				else rcw += 2*disks;
 			}
 		}
@@ -1049,12 +1032,12 @@ static void handle_stripe(struct stripe_head *sh)
 				dev = &sh->dev[i];
 				if ((dev->towrite || i == sh->pd_idx) &&
 				    !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) &&
-				    conf->disks[i].operational) {
+				    test_bit(R5_Insync, &dev->flags)) {
 					if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
 					{
 						PRINTK("Read_old block %d for r-m-w\n", i);
 						set_bit(R5_LOCKED, &dev->flags);
-						action[i] = READ+1;
+						set_bit(R5_Wantread, &dev->flags);
 						locked++;
 					} else {
 						set_bit(STRIPE_DELAYED, &sh->state);
@@ -1068,12 +1051,12 @@ static void handle_stripe(struct stripe_head *sh)
 				dev = &sh->dev[i];
 				if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx &&
 				    !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) &&
-				    conf->disks[i].operational) {
+				    test_bit(R5_Insync, &dev->flags)) {
 					if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
 					{
 						PRINTK("Read_old block %d for Reconstruct\n", i);
 						set_bit(R5_LOCKED, &dev->flags);
-						action[i] = READ+1;
+						set_bit(R5_Wantread, &dev->flags);
 						locked++;
 					} else {
 						set_bit(STRIPE_DELAYED, &sh->state);
@@ -1090,8 +1073,8 @@ static void handle_stripe(struct stripe_head *sh)
 				if (test_bit(R5_LOCKED, &sh->dev[i].flags)) {
 					PRINTK("Writing block %d\n", i);
 					locked++;
-					action[i] = WRITE+1;
-					if (!conf->disks[i].operational
+					set_bit(R5_Wantwrite, &sh->dev[i].flags);
+					if (!test_bit(R5_Insync, &sh->dev[i].flags)
 					    || (i==sh->pd_idx && failed == 0))
 						set_bit(STRIPE_INSYNC, &sh->state);
 				}
@@ -1124,7 +1107,6 @@ static void handle_stripe(struct stripe_head *sh)
 			}
 		}
 		if (!test_bit(STRIPE_INSYNC, &sh->state)) {
-			struct disk_info *spare;
 			if (failed==0)
 				failed_num = sh->pd_idx;
 			/* should be able to compute the missing block and write it to spare */
@@ -1138,14 +1120,10 @@ static void handle_stripe(struct stripe_head *sh)
 				BUG();
 			dev = &sh->dev[failed_num];
 			set_bit(R5_LOCKED, &dev->flags);
-			action[failed_num] = WRITE+1;
+			set_bit(R5_Wantwrite, &dev->flags);
 			locked++;
 			set_bit(STRIPE_INSYNC, &sh->state);
-			if (conf->disks[failed_num].operational)
-				md_sync_acct(conf->disks[failed_num].bdev, STRIPE_SECTORS);
-			else if ((spare=conf->spare))
-				md_sync_acct(spare->bdev, STRIPE_SECTORS);
-
+			set_bit(R5_Syncio, &dev->flags);
 		}
 	}
 	if (syncing && locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) {
@@ -1161,27 +1139,34 @@ static void handle_stripe(struct stripe_head *sh)
 		bi->bi_end_io(bi);
 	}
 	for (i=disks; i-- ;) 
-		if (action[i]) {
+		if (sh->dev[i].flags & ((1<<R5_Wantwrite)|(1<<R5_Wantread))) {
 			struct bio *bi = &sh->dev[i].req;
-			struct disk_info *spare = conf->spare;
-			int skip = 0;
-			if (action[i] == READ+1)
+			mdk_rdev_t *rdev ;
+
+			bi->bi_rw = 0;
+			if (test_bit(R5_Wantread, &sh->dev[i].flags))
 				bi->bi_end_io = raid5_end_read_request;
-			else
+			else {
 				bi->bi_end_io = raid5_end_write_request;
-			if (conf->disks[i].operational)
-				bi->bi_bdev = conf->disks[i].bdev;
-			else if (spare && action[i] == WRITE+1)
-				bi->bi_bdev = spare->bdev;
-			else skip=1;
-			if (!skip) {
-				PRINTK("for %ld schedule op %d on disc %d\n", sh->sector, action[i]-1, i);
+				bi->bi_rw = 1;
+			}
+
+			spin_lock_irq(&conf->device_lock);
+			rdev = conf->disks[i].rdev;
+			if (rdev && rdev->faulty)
+				rdev = NULL;
+			if (rdev)
+				atomic_inc(&rdev->nr_pending);
+			spin_unlock_irq(&conf->device_lock);
+
+			if (rdev) {
+				if (test_bit(R5_Syncio, &sh->dev[i].flags))
+					md_sync_acct(rdev, STRIPE_SECTORS);
+
+				bi->bi_bdev = rdev->bdev;
+				PRINTK("for %ld schedule op %d on disc %d\n", sh->sector, bi->bi_rw, i);
 				atomic_inc(&sh->count);
 				bi->bi_sector = sh->sector;
-				if (action[i] == READ+1) 
-					bi->bi_rw = 0;
-				else
-					bi->bi_rw = 1;
 				bi->bi_flags = 0;
 				bi->bi_vcnt = 1;	
 				bi->bi_idx = 0;
@@ -1190,7 +1175,7 @@ static void handle_stripe(struct stripe_head *sh)
 				bi->bi_next = NULL;
 				generic_make_request(bi);
 			} else {
-				PRINTK("skip op %d on disc %d for sector %ld\n", action[i]-1, i, sh->sector);
+				PRINTK("skip op %d on disc %d for sector %ld\n", bi->bi_rw, i, sh->sector);
 				clear_bit(R5_LOCKED, &dev->flags);
 				set_bit(STRIPE_HANDLE, &sh->state);
 			}
@@ -1363,7 +1348,7 @@ static void raid5d (void *data)
 static int run (mddev_t *mddev)
 {
 	raid5_conf_t *conf;
-	int i, raid_disk, memory;
+	int raid_disk, memory;
 	mdk_rdev_t *rdev;
 	struct disk_info *disk;
 	struct list_head *tmp;
@@ -1399,60 +1384,17 @@ static int run (mddev_t *mddev)
 	PRINTK("raid5: run(md%d) called.\n", mdidx(mddev));
 
 	ITERATE_RDEV(mddev,rdev,tmp) {
-		/*
-		 * This is important -- we are using the descriptor on
-		 * the disk only to get a pointer to the descriptor on
-		 * the main superblock, which might be more recent.
-		 */
 		raid_disk = rdev->raid_disk;
+		if (raid_disk > mddev->raid_disks
+		    || raid_disk < 0)
+			continue;
 		disk = conf->disks + raid_disk;
 
-		if (rdev->faulty) {
-			printk(KERN_ERR "raid5: disabled device %s (errors detected)\n", bdev_partition_name(rdev->bdev));
-			disk->bdev = rdev->bdev;
+		disk->rdev = rdev;
 
-			disk->operational = 0;
-			disk->write_only = 0;
-			disk->spare = 0;
-			disk->used_slot = 1;
-			continue;
-		}
 		if (rdev->in_sync) {
-			if (disk->operational) {
-				printk(KERN_ERR "raid5: disabled device %s (device %d already operational)\n", bdev_partition_name(rdev->bdev), raid_disk);
-				continue;
-			}
 			printk(KERN_INFO "raid5: device %s operational as raid disk %d\n", bdev_partition_name(rdev->bdev), raid_disk);
-	
-			disk->bdev = rdev->bdev;
-			disk->operational = 1;
-			disk->used_slot = 1;
-
 			conf->working_disks++;
-		} else {
-			/*
-			 * Must be a spare disk ..
-			 */
-			printk(KERN_INFO "raid5: spare disk %s\n", bdev_partition_name(rdev->bdev));
-			disk->bdev = rdev->bdev;
-
-			disk->operational = 0;
-			disk->write_only = 0;
-			disk->spare = 1;
-			disk->used_slot = 1;
-		}
-	}
-
-	for (i = 0; i < conf->raid_disks; i++) {
-		disk = conf->disks + i;
-
-		if (!disk->used_slot) {
-			disk->bdev = NULL;
-
-			disk->operational = 0;
-			disk->write_only = 0;
-			disk->spare = 0;
-			disk->used_slot = 1;
 		}
 	}
 
@@ -1467,14 +1409,6 @@ static int run (mddev_t *mddev)
 	conf->algorithm = mddev->layout;
 	conf->max_nr_stripes = NR_STRIPES;
 
-#if 0
-	for (i = 0; i < conf->raid_disks; i++) {
-		if (!conf->disks[i].used_slot) {
-			MD_BUG();
-			goto abort;
-		}
-	}
-#endif
 	if (!conf->chunk_size || conf->chunk_size % 4) {
 		printk(KERN_ERR "raid5: invalid chunk size %d for md%d\n", conf->chunk_size, mdidx(mddev));
 		goto abort;
@@ -1519,7 +1453,7 @@ static int run (mddev_t *mddev)
 		       mddev->raid_disks-mddev->degraded, mddev->raid_disks, conf->algorithm);
 	else
 		printk(KERN_ALERT "raid5: raid level %d set md%d active with %d out of %d devices, algorithm %d\n", conf->level, mdidx(mddev),
-		       mddev->raid_disks = mddev->degraded, mddev->raid_disks, conf->algorithm);
+		       mddev->raid_disks - mddev->degraded, mddev->raid_disks, conf->algorithm);
 
 	print_raid5_conf(conf);
 
@@ -1596,7 +1530,9 @@ static int status (char *page, mddev_t *mddev)
 	sz += sprintf (page+sz, " level %d, %dk chunk, algorithm %d", mddev->level, mddev->chunk_size >> 10, mddev->layout);
 	sz += sprintf (page+sz, " [%d/%d] [", conf->raid_disks, conf->working_disks);
 	for (i = 0; i < conf->raid_disks; i++)
-		sz += sprintf (page+sz, "%s", conf->disks[i].operational ? "U" : "_");
+		sz += sprintf (page+sz, "%s",
+			       conf->disks[i].rdev &&
+			       conf->disks[i].rdev->in_sync ? "U" : "_");
 	sz += sprintf (page+sz, "]");
 #if RAID5_DEBUG
 #define D(x) \
@@ -1619,149 +1555,36 @@ static void print_raid5_conf (raid5_conf_t *conf)
 	printk(" --- rd:%d wd:%d fd:%d\n", conf->raid_disks,
 		 conf->working_disks, conf->failed_disks);
 
-#if RAID5_DEBUG
-	for (i = 0; i < MD_SB_DISKS; i++) {
-#else
-	for (i = 0; i < conf->working_disks+conf->failed_disks; i++) {
-#endif
+	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->disks + i;
-		printk(" disk %d, s:%d, o:%d, us:%d dev:%s\n",
-			i, tmp->spare,tmp->operational,
-			tmp->used_slot,
-			bdev_partition_name(tmp->bdev));
+		if (tmp->rdev)
+		printk(" disk %d, o:%d, dev:%s\n",
+			i, !tmp->rdev->faulty,
+			bdev_partition_name(tmp->rdev->bdev));
 	}
 }
 
 static int raid5_spare_active(mddev_t *mddev)
 {
-	int err = 0;
-	int i, failed_disk=-1, spare_disk=-1;
+	int i;
 	raid5_conf_t *conf = mddev->private;
-	struct disk_info *tmp, *sdisk, *fdisk;
-	mdk_rdev_t *spare_rdev, *failed_rdev;
+	struct disk_info *tmp;
 
-	print_raid5_conf(conf);
 	spin_lock_irq(&conf->device_lock);
 	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->disks + i;
-		if ((!tmp->operational && !tmp->spare) ||
-				!tmp->used_slot) {
-			failed_disk = i;
-			break;
-		}
-	}
-	if (failed_disk == -1) {
-		MD_BUG();
-		err = 1;
-		goto abort;
-	}
-	/*
-	 * Find the spare disk ... (can only be in the 'high'
-	 * area of the array)
-	 */
-	spare_disk = mddev->spare->raid_disk;
-
-	if (!conf->spare) {
-		MD_BUG();
-		err = 1;
-		goto abort;
-	}
-	sdisk = conf->disks + spare_disk;
-	fdisk = conf->disks + failed_disk;
-
-	/*
-	 * do the switch finally
-	 */
-	spare_rdev = find_rdev_nr(mddev, spare_disk);
-	failed_rdev = find_rdev_nr(mddev, failed_disk);
-
-	/* There must be a spare_rdev, but there may not be a
-	 * failed_rdev.  That slot might be empty...
-	 */
-	spare_rdev->desc_nr = failed_disk;
-	spare_rdev->raid_disk = failed_disk;
-	if (failed_rdev) {
-		failed_rdev->desc_nr = spare_disk;
-		failed_rdev->raid_disk = spare_disk;
-	}
-	
-	xchg_values(*fdisk, *sdisk);
-
-	/*
-	 * (careful, 'failed' and 'spare' are switched from now on)
-	 *
-	 * we want to preserve linear numbering and we want to
-	 * give the proper raid_disk number to the now activated
-	 * disk. (this means we switch back these values)
-	 */
-
-	if (!sdisk->bdev)
-		sdisk->used_slot = 0;
-
-	/*
-	 * this really activates the spare.
-	 */
-	fdisk->spare = 0;
-	fdisk->write_only = 0;
-
-	/*
-	 * if we activate a spare, we definitely replace a
-	 * non-operational disk slot in the 'low' area of
-	 * the disk array.
-	 */
+		if (tmp->rdev
+		    && !tmp->rdev->faulty
+		    && !tmp->rdev->in_sync) {
 			mddev->degraded--;
 			conf->failed_disks--;
 			conf->working_disks++;
-	conf->spare = NULL;
-abort:
-	spin_unlock_irq(&conf->device_lock);
-	print_raid5_conf(conf);
-	return err;
-}
-
-static int raid5_spare_inactive(mddev_t *mddev)
-{
-	raid5_conf_t *conf = mddev->private;
-	struct disk_info *p;
-	int err = 0;
-
-	print_raid5_conf(conf);
-	spin_lock_irq(&conf->device_lock);
-	p = conf->disks + mddev->spare->raid_disk;
-	if (p) {
-		p->operational = 0;
-		p->write_only = 0;
-		if (conf->spare == p)
-			conf->spare = NULL;
-	} else {
-		MD_BUG();
-		err = 1;
+			tmp->rdev->in_sync = 1;
 		}
-	spin_unlock_irq(&conf->device_lock);
-	print_raid5_conf(conf);
-	return err;
-}
-
-static int raid5_spare_write(mddev_t *mddev)
-{
-	raid5_conf_t *conf = mddev->private;
-	struct disk_info *p;
-	int err = 0;
-
-	print_raid5_conf(conf);
-	spin_lock_irq(&conf->device_lock);
-	p = conf->disks + mddev->spare->raid_disk;
-	if (p && !conf->spare) {
-		p->operational = 1;
-		p->write_only = 1;
-		conf->spare = p;
-	} else {
-		MD_BUG();
-		err = 1;
 	}
 	spin_unlock_irq(&conf->device_lock);
 	print_raid5_conf(conf);
-	return err;
+	return 0;
 }
 
 static int raid5_remove_disk(mddev_t *mddev, int number)
@@ -1773,13 +1596,13 @@ static int raid5_remove_disk(mddev_t *mddev, int number)
 	print_raid5_conf(conf);
 	spin_lock_irq(&conf->device_lock);
 
-	if (p->used_slot) {
-		if (p->operational) {
+	if (p->rdev) {
+		if (p->rdev->in_sync || 
+		    atomic_read(&p->rdev->nr_pending)) {
 			err = -EBUSY;
 			goto abort;
 		}
-		p->bdev = NULL;
-		p->used_slot = 0;
+		p->rdev = NULL;
 		err = 0;
 	}
 	if (err)
@@ -1793,29 +1616,25 @@ static int raid5_remove_disk(mddev_t *mddev, int number)
 static int raid5_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	raid5_conf_t *conf = mddev->private;
-	int err = 1;
-	struct disk_info *p = conf->disks + rdev->raid_disk;
+	int found = 0;
+	int disk;
+	struct disk_info *p;
 
-	print_raid5_conf(conf);
 	spin_lock_irq(&conf->device_lock);
 	/*
 	 * find the disk ...
 	 */
-
-	if (!p->used_slot) {
-		/* it will be held open by rdev */
-		p->bdev = rdev->bdev;
-		p->operational = 0;
-		p->write_only = 0;
-		p->spare = 1;
-		p->used_slot = 1;
-		err = 0;
+	for (disk=0; disk < mddev->raid_disks; disk++)
+		if ((p=conf->disks + disk)->rdev == NULL) {
+			p->rdev = rdev;
+			rdev->in_sync = 0;
+			rdev->raid_disk = disk;
+			found = 1;
+			break;
 		}
-	if (err)
-		MD_BUG();
 	spin_unlock_irq(&conf->device_lock);
 	print_raid5_conf(conf);
-	return err;
+	return found;
 }
 
 static mdk_personality_t raid5_personality=
@@ -1828,8 +1647,6 @@ static mdk_personality_t raid5_personality=
 	.error_handler	= error,
 	.hot_add_disk	= raid5_add_disk,
 	.hot_remove_disk= raid5_remove_disk,
-	.spare_write	= raid5_spare_write,
-	.spare_inactive	= raid5_spare_inactive,
 	.spare_active	= raid5_spare_active,
 	.sync_request	= sync_request,
 };

drivers/net/appletalk/Config.in

@@ -16,3 +16,4 @@ if [ "$CONFIG_DEV_APPLETALK" = "y" ]; then
       bool '    Appletalk-IP to IP Decapsulation support' CONFIG_IPDDP_DECAP
    fi
 fi
+

fs/nfsd/nfs3proc.c

@@ -134,7 +134,7 @@ nfsd3_proc_access(struct svc_rqst *rqstp, struct nfsd3_accessargs *argp,
 
 	fh_copy(&resp->fh, &argp->fh);
 	resp->access = argp->access;
-	nfserr = nfsd_access(rqstp, &resp->fh, &resp->access);
+	nfserr = nfsd_access(rqstp, &resp->fh, &resp->access, NULL);
 	RETURN_STATUS(nfserr);
 }
 
@@ -267,7 +267,7 @@ nfsd3_proc_create(struct svc_rqst *rqstp, struct nfsd3_createargs *argp,
 	/* Now create the file and set attributes */
 	nfserr = nfsd_create_v3(rqstp, dirfhp, argp->name, argp->len,
 				attr, newfhp,
-				argp->createmode, argp->verf);
+				argp->createmode, argp->verf, NULL);
 
 	RETURN_STATUS(nfserr);
 }
@@ -460,7 +460,7 @@ nfsd3_proc_readdir(struct svc_rqst *rqstp, struct nfsd3_readdirargs *argp,
 	fh_copy(&resp->fh, &argp->fh);
 	nfserr = nfsd_readdir(rqstp, &resp->fh, (loff_t) argp->cookie, 
 					nfs3svc_encode_entry,
-					buffer, &count, argp->verf);
+					buffer, &count, argp->verf, NULL);
 	memcpy(resp->verf, argp->verf, 8);
 	resp->count = count;
 
@@ -495,7 +495,7 @@ nfsd3_proc_readdirplus(struct svc_rqst *rqstp, struct nfsd3_readdirargs *argp,
 	fh_copy(&resp->fh, &argp->fh);
 	nfserr = nfsd_readdir(rqstp, &resp->fh, (loff_t) argp->cookie, 
 					nfs3svc_encode_entry_plus,
-					buffer, &count, argp->verf);
+					buffer, &count, argp->verf, NULL);
 	memcpy(resp->verf, argp->verf, 8);
 	resp->count = count;
 

fs/nfsd/nfsfh.c

@@ -107,8 +107,10 @@ fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, int access)
 		int fsid = 0;
 
 		error = nfserr_stale;
-		if (rqstp->rq_vers == 3)
+		if (rqstp->rq_vers > 2)
 			error = nfserr_badhandle;
+		if (rqstp->rq_vers == 4 && fh->fh_size == 0)
+			return nfserr_nofilehandle;
 
 		if (fh->fh_version == 1) {
 			datap = fh->fh_auth;
@@ -171,7 +173,7 @@ fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, int access)
 		 * Look up the dentry using the NFS file handle.
 		 */
 		error = nfserr_stale;
-		if (rqstp->rq_vers == 3)
+		if (rqstp->rq_vers > 2)
 			error = nfserr_badhandle;
 
 		if (fh->fh_version != 1) {
@@ -234,11 +236,23 @@ fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, int access)
 
 	/* Type can be negative when creating hardlinks - not to a dir */
 	if (type > 0 && (inode->i_mode & S_IFMT) != type) {
-		error = (type == S_IFDIR)? nfserr_notdir : nfserr_isdir;
+		if (rqstp->rq_vers == 4 && (inode->i_mode & S_IFMT) == S_IFLNK)
+			error = nfserr_symlink;
+		else if (type == S_IFDIR)
+			error = nfserr_notdir;
+		else if ((inode->i_mode & S_IFMT) == S_IFDIR)
+			error = nfserr_isdir;
+		else
+			error = nfserr_inval;
 		goto out;
 	}
 	if (type < 0 && (inode->i_mode & S_IFMT) == -type) {
-		error = (type == -S_IFDIR)? nfserr_notdir : nfserr_isdir;
+		if (rqstp->rq_vers == 4 && (inode->i_mode & S_IFMT) == S_IFLNK)
+			error = nfserr_symlink;
+		else if (type == -S_IFDIR)
+			error = nfserr_isdir;
+		else
+			error = nfserr_notdir;
 		goto out;
 	}
 
@@ -303,6 +317,8 @@ fh_compose(struct svc_fh *fhp, struct svc_export *exp, struct dentry *dentry, st
 	 *
 	 */
 
+	u8 ref_fh_version = 0;
+	u8 ref_fh_fsid_type = 0;
 	struct inode * inode = dentry->d_inode;
 	struct dentry *parent = dentry->d_parent;
 	__u32 *datap;
@@ -312,6 +328,13 @@ fh_compose(struct svc_fh *fhp, struct svc_export *exp, struct dentry *dentry, st
 		parent->d_name.name, dentry->d_name.name,
 		(inode ? inode->i_ino : 0));
 
+	if (ref_fh) {
+		ref_fh_version = ref_fh->fh_handle.fh_version;
+		ref_fh_fsid_type = ref_fh->fh_handle.fh_fsid_type;
+		if (ref_fh == fhp)
+			fh_put(ref_fh);
+	}
+
 	if (fhp->fh_locked || fhp->fh_dentry) {
 		printk(KERN_ERR "fh_compose: fh %s/%s not initialized!\n",
 			parent->d_name.name, dentry->d_name.name);
@@ -323,8 +346,7 @@ fh_compose(struct svc_fh *fhp, struct svc_export *exp, struct dentry *dentry, st
 	fhp->fh_dentry = dentry; /* our internal copy */
 	fhp->fh_export = exp;
 
-	if (ref_fh &&
-	    ref_fh->fh_handle.fh_version == 0xca) {
+	if (ref_fh_version == 0xca) {
 		/* old style filehandle please */
 		memset(&fhp->fh_handle.fh_base, 0, NFS_FHSIZE);
 		fhp->fh_handle.fh_size = NFS_FHSIZE;
@@ -340,7 +362,7 @@ fh_compose(struct svc_fh *fhp, struct svc_export *exp, struct dentry *dentry, st
 		fhp->fh_handle.fh_auth_type = 0;
 		datap = fhp->fh_handle.fh_auth+0;
 		if ((exp->ex_flags & NFSEXP_FSID) &&
-		    (!ref_fh || ref_fh->fh_handle.fh_fsid_type == 1)) {
+		    (ref_fh_fsid_type == 1)) {
 			fhp->fh_handle.fh_fsid_type = 1;
 			/* fsid_type 1 == 4 bytes filesystem id */
 			*datap++ = exp->ex_fsid;
@@ -424,6 +446,10 @@ fh_put(struct svc_fh *fhp)
 		fh_unlock(fhp);
 		fhp->fh_dentry = NULL;
 		dput(dentry);
+#ifdef CONFIG_NFSD_V3
+		fhp->fh_pre_saved = 0;
+		fhp->fh_post_saved = 0;
+#endif
 		nfsd_nr_put++;
 	}
 	return;

fs/nfsd/nfsproc.c

@@ -492,7 +492,7 @@ nfsd_proc_readdir(struct svc_rqst *rqstp, struct nfsd_readdirargs *argp,
 	/* Read directory and encode entries on the fly */
 	nfserr = nfsd_readdir(rqstp, &argp->fh, (loff_t) argp->cookie, 
 			      nfssvc_encode_entry,
-			      buffer, &count, NULL);
+			      buffer, &count, NULL, NULL);
 	resp->count = count;
 
 	fh_put(&argp->fh);

fs/nfsd/vfs.c

@@ -348,12 +348,12 @@ static struct accessmap	nfs3_anyaccess[] = {
 };
 
 int
-nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access)
+nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
 {
 	struct accessmap	*map;
 	struct svc_export	*export;
 	struct dentry		*dentry;
-	u32			query, result = 0;
+	u32			query, result = 0, sresult = 0;
 	unsigned int		error;
 
 	error = fh_verify(rqstp, fhp, 0, MAY_NOP);
@@ -375,6 +375,9 @@ nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access)
 	for  (; map->access; map++) {
 		if (map->access & query) {
 			unsigned int err2;
+
+			sresult |= map->access;
+
 			err2 = nfsd_permission(export, dentry, map->how);
 			switch (err2) {
 			case nfs_ok:
@@ -395,6 +398,8 @@ nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access)
 		}
 	}
 	*access = result;
+	if (supported)
+		*supported = sresult;
 
  out:
 	return error;
@@ -756,6 +761,9 @@ nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	struct file	file;
 	int		err;
 
+	if ((u64)count > ~(u64)offset)
+		return nfserr_inval;
+
 	if ((err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file)) != 0)
 		return err;
 	if (EX_ISSYNC(fhp->fh_export)) {
@@ -904,7 +912,8 @@ nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
 int
 nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		char *fname, int flen, struct iattr *iap,
-		struct svc_fh *resfhp, int createmode, u32 *verifier)
+		struct svc_fh *resfhp, int createmode, u32 *verifier,
+	        int *truncp)
 {
 	struct dentry	*dentry, *dchild;
 	struct inode	*dirp;
@@ -966,6 +975,16 @@ nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
 		case NFS3_CREATE_UNCHECKED:
 			if (! S_ISREG(dchild->d_inode->i_mode))
 				err = nfserr_exist;
+			else if (truncp) {
+				/* in nfsv4, we need to treat this case a little
+				 * differently.  we don't want to truncate the
+				 * file now; this would be wrong if the OPEN
+				 * fails for some other reason.  furthermore,
+				 * if the size is nonzero, we should ignore it
+				 * according to spec!
+				 */
+				*truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
+			}
 			else {
 				iap->ia_valid &= ATTR_SIZE;
 				goto set_attr;
@@ -1326,6 +1345,9 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 		goto out;
 	}
 
+	if (!type)
+		type = rdentry->d_inode->i_mode & S_IFMT;
+
 	if (type != S_IFDIR) { /* It's UNLINK */
 #ifdef MSNFS
 		if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
@@ -1359,7 +1381,7 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
  */
 int
 nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t offset, 
-             encode_dent_fn func, u32 *buffer, int *countp, u32 *verf)
+             encode_dent_fn func, u32 *buffer, int *countp, u32 *verf, u32 *bmval)
 {
 	u32		*p;
 	int		oldlen, eof, err;
@@ -1380,6 +1402,10 @@ nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t offset,
 	cd.buffer = buffer;
 	cd.buflen = *countp; /* count of words */
 	cd.dirfh  = fhp;
+	if (bmval) {
+		cd.bmval[0] = bmval[0];
+		cd.bmval[1] = bmval[1];
+	}
 
 	/*
 	 * Read the directory entries. This silly loop is necessary because
@@ -1395,13 +1421,20 @@ nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t offset,
 		if (err < 0)
 			goto out_nfserr;
 
+		err = cd.nfserr;
+		if (err)
+			goto out_close;
 	} while (oldlen != cd.buflen && !cd.eob);
 
+	err = nfserr_readdir_nospc;
+	if (rqstp->rq_vers == 4 && cd.eob && cd.buffer == buffer)
+		goto out_close;
+
 	/* If we didn't fill the buffer completely, we're at EOF */
 	eof = !cd.eob;
 
 	if (cd.offset) {
-		if (rqstp->rq_vers == 3)
+		if (rqstp->rq_vers > 2)
 			(void)xdr_encode_hyper(cd.offset, file.f_pos);
 		else
 			*cd.offset = htonl(file.f_pos);

include/asm-i386/hardirq.h

@@ -77,8 +77,10 @@ typedef struct {
 #define irq_enter()		(preempt_count() += HARDIRQ_OFFSET)
 
 #if CONFIG_PREEMPT
+# define in_atomic()	(preempt_count() != kernel_locked())
 # define IRQ_EXIT_OFFSET (HARDIRQ_OFFSET-1)
 #else
+# define in_atomic()	(preempt_count() != 0)
 # define IRQ_EXIT_OFFSET HARDIRQ_OFFSET
 #endif
 #define irq_exit()							\

include/linux/nfs.h

@@ -39,39 +39,68 @@
  * standard, but seem to be widely used nevertheless.
  */
  enum nfs_stat {
-	NFS_OK = 0,			/* v2 v3 */
-	NFSERR_PERM = 1,		/* v2 v3 */
-	NFSERR_NOENT = 2,		/* v2 v3 */
-	NFSERR_IO = 5,			/* v2 v3 */
-	NFSERR_NXIO = 6,		/* v2 v3 */
+	NFS_OK = 0,			/* v2 v3 v4 */
+	NFSERR_PERM = 1,		/* v2 v3 v4 */
+	NFSERR_NOENT = 2,		/* v2 v3 v4 */
+	NFSERR_IO = 5,			/* v2 v3 v4 */
+	NFSERR_NXIO = 6,		/* v2 v3 v4 */
 	NFSERR_EAGAIN = 11,		/* v2 v3 */
-	NFSERR_ACCES = 13,		/* v2 v3 */
-	NFSERR_EXIST = 17,		/* v2 v3 */
-	NFSERR_XDEV = 18,		/*    v3 */
-	NFSERR_NODEV = 19,		/* v2 v3 */
-	NFSERR_NOTDIR = 20,		/* v2 v3 */
-	NFSERR_ISDIR = 21,		/* v2 v3 */
-	NFSERR_INVAL = 22,		/* v2 v3 that Sun forgot */
-	NFSERR_FBIG = 27,		/* v2 v3 */
-	NFSERR_NOSPC = 28,		/* v2 v3 */
-	NFSERR_ROFS = 30,		/* v2 v3 */
-	NFSERR_MLINK = 31,		/*    v3 */
+	NFSERR_ACCES = 13,		/* v2 v3 v4 */
+	NFSERR_EXIST = 17,		/* v2 v3 v4 */
+	NFSERR_XDEV = 18,		/*    v3 v4 */
+	NFSERR_NODEV = 19,		/* v2 v3 v4 */
+	NFSERR_NOTDIR = 20,		/* v2 v3 v4 */
+	NFSERR_ISDIR = 21,		/* v2 v3 v4 */
+	NFSERR_INVAL = 22,		/* v2 v3 v4 */
+	NFSERR_FBIG = 27,		/* v2 v3 v4 */
+	NFSERR_NOSPC = 28,		/* v2 v3 v4 */
+	NFSERR_ROFS = 30,		/* v2 v3 v4 */
+	NFSERR_MLINK = 31,		/*    v3 v4 */
 	NFSERR_OPNOTSUPP = 45,		/* v2 v3 */
-	NFSERR_NAMETOOLONG = 63,	/* v2 v3 */
-	NFSERR_NOTEMPTY = 66,		/* v2 v3 */
-	NFSERR_DQUOT = 69,		/* v2 v3 */
-	NFSERR_STALE = 70,		/* v2 v3 */
+	NFSERR_NAMETOOLONG = 63,	/* v2 v3 v4 */
+	NFSERR_NOTEMPTY = 66,		/* v2 v3 v4 */
+	NFSERR_DQUOT = 69,		/* v2 v3 v4 */
+	NFSERR_STALE = 70,		/* v2 v3 v4 */
 	NFSERR_REMOTE = 71,		/* v2 v3 */
 	NFSERR_WFLUSH = 99,		/* v2    */
-	NFSERR_BADHANDLE = 10001,	/*    v3 */
+	NFSERR_BADHANDLE = 10001,	/*    v3 v4 */
 	NFSERR_NOT_SYNC = 10002,	/*    v3 */
-	NFSERR_BAD_COOKIE = 10003,	/*    v3 */
-	NFSERR_NOTSUPP = 10004,		/*    v3 */
-	NFSERR_TOOSMALL = 10005,	/*    v3 */
-	NFSERR_SERVERFAULT = 10006,	/*    v3 */
-	NFSERR_BADTYPE = 10007,		/*    v3 */
-	NFSERR_JUKEBOX = 10008		/*    v3 */
- };
+	NFSERR_BAD_COOKIE = 10003,	/*    v3 v4 */
+	NFSERR_NOTSUPP = 10004,		/*    v3 v4 */
+	NFSERR_TOOSMALL = 10005,	/*    v3 v4 */
+	NFSERR_SERVERFAULT = 10006,	/*    v3 v4 */
+	NFSERR_BADTYPE = 10007,		/*    v3 v4 */
+	NFSERR_JUKEBOX = 10008,		/*    v3 v4 */
+	NFSERR_SAME = 10009,		/*       v4 */
+	NFSERR_DENIED = 10010,		/*       v4 */
+	NFSERR_EXPIRED = 10011,		/*       v4 */
+	NFSERR_LOCKED = 10012,		/*       v4 */
+	NFSERR_GRACE = 10013,		/*       v4 */
+	NFSERR_FHEXPIRED = 10014,	/*       v4 */
+	NFSERR_SHARE_DENIED = 10015,	/*       v4 */
+	NFSERR_WRONGSEC = 10016,	/*       v4 */
+	NFSERR_CLID_INUSE = 10017,	/*       v4 */
+	NFSERR_RESOURCE = 10018,	/*       v4 */
+	NFSERR_MOVED = 10019,		/*       v4 */
+	NFSERR_NOFILEHANDLE = 10020,	/*       v4 */
+	NFSERR_MINOR_VERS_MISMATCH = 10021,   /* v4 */
+	NFSERR_STALE_CLIENTID = 10022,	/*       v4 */
+	NFSERR_STALE_STATEID = 10023,	/*       v4 */
+	NFSERR_OLD_STATEID = 10024,	/*       v4 */
+	NFSERR_BAD_STATEID = 10025,	/*       v4 */
+	NFSERR_BAD_SEQID = 10026,	/*       v4 */
+	NFSERR_NOT_SAME = 10027,	/*       v4 */
+	NFSERR_LOCK_RANGE = 10028,	/*       v4 */
+	NFSERR_SYMLINK = 10029,		/*       v4 */
+	NFSERR_READDIR_NOSPC = 10030,	/*       v4 */
+	NFSERR_LEASE_MOVED = 10031,	/*       v4 */
+	NFSERR_ATTRNOTSUPP = 10032,	/*       v4 */
+	NFSERR_NO_GRACE = 10033,	/*       v4 */
+	NFSERR_RECLAIM_BAD = 10034,	/*       v4 */
+	NFSERR_RECLAIM_CONFLICT = 10035,/*       v4 */
+	NFSERR_BAD_XDR = 10036,		/*       v4 */
+	NFSERR_LOCKS_HELD = 10037	/*       v4 */
+};
 
 /* NFSv2 file types - beware, these are not the same in NFSv3 */
 

include/linux/nfsd/nfsd.h

@@ -55,6 +55,8 @@ struct readdir_cd {
 	char			plus;		/* readdirplus */
 	char			eob;		/* end of buffer */
 	char			dotonly;
+	int			nfserr;		/* v4 only */
+	u32			bmval[2];	/* v4 only */
 };
 typedef int		(*encode_dent_fn)(struct readdir_cd *, const char *,
 						int, loff_t, ino_t, unsigned int);
@@ -86,11 +88,11 @@ int		nfsd_create(struct svc_rqst *, struct svc_fh *,
 				char *name, int len, struct iattr *attrs,
 				int type, dev_t rdev, struct svc_fh *res);
 #ifdef CONFIG_NFSD_V3
-int		nfsd_access(struct svc_rqst *, struct svc_fh *, u32 *);
+int		nfsd_access(struct svc_rqst *, struct svc_fh *, u32 *, u32 *);
 int		nfsd_create_v3(struct svc_rqst *, struct svc_fh *,
 				char *name, int len, struct iattr *attrs,
 				struct svc_fh *res, int createmode,
-				u32 *verifier);
+				u32 *verifier, int *truncp);
 int		nfsd_commit(struct svc_rqst *, struct svc_fh *,
 				off_t, unsigned long);
 #endif /* CONFIG_NFSD_V3 */
@@ -119,7 +121,8 @@ int		nfsd_truncate(struct svc_rqst *, struct svc_fh *,
 				unsigned long size);
 int		nfsd_readdir(struct svc_rqst *, struct svc_fh *,
 				loff_t, encode_dent_fn,
-				u32 *buffer, int *countp, u32 *verf);
+				u32 *buffer, int *countp, u32 *verf,
+				u32 *bmval);
 int		nfsd_statfs(struct svc_rqst *, struct svc_fh *,
 				struct statfs *);
 
@@ -170,6 +173,16 @@ void		nfsd_lockd_unexport(struct svc_client *);
 #define	nfserr_serverfault	__constant_htonl(NFSERR_SERVERFAULT)
 #define	nfserr_badtype		__constant_htonl(NFSERR_BADTYPE)
 #define	nfserr_jukebox		__constant_htonl(NFSERR_JUKEBOX)
+#define	nfserr_bad_cookie	__constant_htonl(NFSERR_BAD_COOKIE)
+#define	nfserr_same		__constant_htonl(NFSERR_SAME)
+#define	nfserr_clid_inuse	__constant_htonl(NFSERR_CLID_INUSE)
+#define	nfserr_resource		__constant_htonl(NFSERR_RESOURCE)
+#define	nfserr_nofilehandle	__constant_htonl(NFSERR_NOFILEHANDLE)
+#define	nfserr_minor_vers_mismatch	__constant_htonl(NFSERR_MINOR_VERS_MISMATCH)
+#define	nfserr_symlink		__constant_htonl(NFSERR_SYMLINK)
+#define	nfserr_not_same		__constant_htonl(NFSERR_NOT_SAME)
+#define	nfserr_readdir_nospc	__constant_htonl(NFSERR_READDIR_NOSPC)
+#define	nfserr_bad_xdr		__constant_htonl(NFSERR_BAD_XDR)
 
 /* error code for internal use - if a request fails due to
  * kmalloc failure, it gets dropped.  Client should resend eventually

include/linux/raid/linear.h

@@ -4,7 +4,7 @@
 #include <linux/raid/md.h>
 
 struct dev_info {
-	struct block_device *bdev;
+	mdk_rdev_t	*rdev;
 	unsigned long	size;
 	unsigned long	offset;
 };

include/linux/raid/md.h

@@ -76,9 +76,8 @@ extern void md_unregister_thread (mdk_thread_t *thread);
 extern void md_wakeup_thread(mdk_thread_t *thread);
 extern void md_interrupt_thread (mdk_thread_t *thread);
 extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
-extern void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors);
-extern int md_error (mddev_t *mddev, struct block_device *bdev);
-extern int md_run_setup(void);
+extern void md_sync_acct(mdk_rdev_t *rdev, unsigned long nr_sectors);
+extern void md_error (mddev_t *mddev, mdk_rdev_t *rdev);
 
 extern void md_print_devices (void);
 

include/linux/raid/md_k.h

@@ -154,12 +154,26 @@ struct mdk_rdev_s
 	mdp_super_t	*sb;
 	unsigned long	sb_offset;
 
-	int alias_device;		/* device alias to the same disk */
+	/* A device can be in one of three states based on two flags:
+	 * Not working:   faulty==1 in_sync==0
+	 * Fully working: faulty==0 in_sync==1
+	 * Working, but not
+	 * in sync with array
+	 *                faulty==0 in_sync==0
+	 *
+	 * It can never have faulty==1, in_sync==1
+	 * This reduces the burden of testing multiple flags in many cases
+	 */
 	int faulty;			/* if faulty do not issue IO requests */
 	int in_sync;			/* device is a full member of the array */
 
 	int desc_nr;			/* descriptor index in the superblock */
 	int raid_disk;			/* role of device in array */
+
+	atomic_t	nr_pending;	/* number of pending requests.
+					 * only maintained for arrays that
+					 * support hot removal
+					 */
 };
 
 typedef struct mdk_personality_s mdk_personality_t;
@@ -202,7 +216,7 @@ struct mddev_s
 	int				in_sync;	/* know to not need resync */
 	struct semaphore		reconfig_sem;
 	atomic_t			active;
-	mdk_rdev_t			*spare;
+	int				spares;
 
 	int				degraded;	/* whether md should consider
 							 * adding a spare
@@ -223,11 +237,12 @@ struct mdk_personality_s
 	int (*run)(mddev_t *mddev);
 	int (*stop)(mddev_t *mddev);
 	int (*status)(char *page, mddev_t *mddev);
-	int (*error_handler)(mddev_t *mddev, struct block_device *bdev);
+	/* error_handler must set ->faulty and clear ->in_sync
+	 * if appropriate, and should abort recovery if needed 
+	 */
+	void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
 	int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
 	int (*hot_remove_disk) (mddev_t *mddev, int number);
-	int (*spare_write) (mddev_t *mddev);
-	int (*spare_inactive) (mddev_t *mddev);
 	int (*spare_active) (mddev_t *mddev);
 	int (*sync_request)(mddev_t *mddev, sector_t sector_nr, int go_faster);
 };
@@ -272,9 +287,6 @@ extern mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr);
 #define ITERATE_RDEV_PENDING(rdev,tmp)					\
 	ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
 
-#define xchg_values(x,y) do { __typeof__(x) __tmp = x; \
-				x = y; y = __tmp; } while (0)
-
 typedef struct mdk_thread_s {
 	void			(*run) (void *data);
 	void			*data;

include/linux/raid/multipath.h

@@ -5,14 +5,7 @@
 #include <linux/bio.h>
 
 struct multipath_info {
-	struct block_device *bdev;
-
-	/*
-	 * State bits:
-	 */
-	int		operational;
-
-	int		used_slot;
+	mdk_rdev_t	*rdev;
 };
 
 struct multipath_private_data {

include/linux/raid/raid1.h

@@ -6,18 +6,8 @@
 typedef struct mirror_info mirror_info_t;
 
 struct mirror_info {
-	struct block_device *bdev;
+	mdk_rdev_t	*rdev;
 	sector_t	head_position;
-	atomic_t	nr_pending;
-
-	/*
-	 * State bits:
-	 */
-	int		operational;
-	int		write_only;
-	int		spare;
-
-	int		used_slot;
 };
 
 typedef struct r1bio_s r1bio_t;
@@ -30,7 +20,6 @@ struct r1_private_data_s {
 	int			last_used;
 	sector_t		next_seq_sect;
 	mdk_thread_t		*thread;
-	mirror_info_t		*spare;
 	spinlock_t		device_lock;
 
 	/* for use when syncing mirrors: */

include/linux/raid/raid5.h

@@ -148,6 +148,11 @@ struct stripe_head {
 #define	R5_UPTODATE	0	/* page contains current data */
 #define	R5_LOCKED	1	/* IO has been submitted on "req" */
 #define	R5_OVERWRITE	2	/* towrite covers whole page */
+/* and some that are internal to handle_stripe */
+#define	R5_Insync	3	/* rdev && rdev->in_sync at start */
+#define	R5_Wantread	4	/* want to schedule a read */
+#define	R5_Wantwrite	5
+#define	R5_Syncio	6	/* this io need to be accounted as resync io */
 
 /*
  * Write method
@@ -192,11 +197,7 @@ struct stripe_head {
  
 
 struct disk_info {
-	struct block_device *bdev;
-	int	operational;
-	int	write_only;
-	int	spare;
-	int	used_slot;
+	mdk_rdev_t	*rdev;
 };
 
 struct raid5_private_data {

mm/swap.c

@@ -76,9 +76,8 @@ void lru_add_drain(void)
  */
 void __page_cache_release(struct page *page)
 {
-	BUG_ON(page_count(page) != 0);
-	if (PageLRU(page)) {
 	unsigned long flags;
+	BUG_ON(page_count(page) != 0);
 
 	spin_lock_irqsave(&_pagemap_lru_lock, flags);
 	if (TestClearPageLRU(page)) {
@@ -90,7 +89,6 @@ void __page_cache_release(struct page *page)
 	if (page_count(page) != 0)
 		page = NULL;
 	spin_unlock_irqrestore(&_pagemap_lru_lock, flags);
-	}
 	if (page)
 		__free_pages_ok(page, 0);
 }

mm/vmalloc.c

@@ -153,15 +153,20 @@ int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page ***pages)
 	unsigned long address = VMALLOC_VMADDR(area->addr);
 	unsigned long end = address + (area->size-PAGE_SIZE);
 	pgd_t *dir;
+	int err = 0;
 
 	dir = pgd_offset_k(address);
 	spin_lock(&init_mm.page_table_lock);
 	do {
 		pmd_t *pmd = pmd_alloc(&init_mm, dir, address);
-		if (!pmd)
-			return -ENOMEM;
-		if (map_area_pmd(pmd, address, end - address, prot, pages))
-			return -ENOMEM;
+		if (!pmd) {
+			err = -ENOMEM;
+			break;
+		}
+		if (map_area_pmd(pmd, address, end - address, prot, pages)) {
+			err = -ENOMEM;
+			break;
+		}
 
 		address = (address + PGDIR_SIZE) & PGDIR_MASK;
 		dir++;
@@ -169,7 +174,7 @@ int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page ***pages)
 
 	spin_unlock(&init_mm.page_table_lock);
 	flush_cache_all();
-	return 0;
+	return err;
 }
 
 
@@ -379,15 +384,21 @@ void *__vmalloc(unsigned long size, int gfp_mask, pgprot_t prot)
 
 	area->nr_pages = nr_pages;
 	area->pages = pages = kmalloc(array_size, (gfp_mask & ~__GFP_HIGHMEM));
-	if (!area->pages)
+	if (!area->pages) {
+		remove_vm_area(area->addr);
+		kfree(area);
 		return NULL;
+	}
 	memset(area->pages, 0, array_size);
 
 	for (i = 0; i < area->nr_pages; i++) {
 		area->pages[i] = alloc_page(gfp_mask);
-		if (unlikely(!area->pages[i]))
+		if (unlikely(!area->pages[i])) {
+			/* Successfully allocated i pages, free them in __vunmap() */
+			area->nr_pages = i;
 			goto fail;
 		}
+	}
 	
 	if (map_vm_area(area, prot, &pages))
 		goto fail;

net/sunrpc/svcsock.c

@@ -140,7 +140,7 @@ svc_sock_enqueue(struct svc_sock *svsk)
 	    && !test_bit(SK_CLOSE, &svsk->sk_flags)
 	    && !test_bit(SK_CONN, &svsk->sk_flags)) {
 		/* Don't enqueue while not enough space for reply */
-		dprintk("svc: socket %p  no space, %d > %ld, not enqueued\n",
+		dprintk("svc: socket %p  no space, %d*2 > %ld, not enqueued\n",
 			svsk->sk_sk, svsk->sk_reserved+serv->sv_bufsz,
 			sock_wspace(svsk->sk_sk));
 		goto out_unlock;
@@ -574,6 +574,14 @@ svc_udp_init(struct svc_sock *svsk)
 	svsk->sk_recvfrom = svc_udp_recvfrom;
 	svsk->sk_sendto = svc_udp_sendto;
 
+	/* initialise setting must have enough space to
+	 * receive and respond to one request.  
+	 * svc_udp_recvfrom will re-adjust if necessary
+	 */
+	svc_sock_setbufsize(svsk->sk_sock,
+			    3 * svsk->sk_server->sv_bufsz,
+			    3 * svsk->sk_server->sv_bufsz);
+
 	set_bit(SK_CHNGBUF, &svsk->sk_flags);
 
 	return 0;
@@ -679,6 +687,8 @@ svc_tcp_accept(struct svc_sock *svsk)
 		goto failed;		/* aborted connection or whatever */
 	}
 	set_bit(SK_CONN, &svsk->sk_flags);
+	svc_sock_enqueue(svsk);
+
 	slen = sizeof(sin);
 	err = ops->getname(newsock, (struct sockaddr *) &sin, &slen, 1);
 	if (err < 0) {
@@ -944,6 +954,14 @@ svc_tcp_init(struct svc_sock *svsk)
 		svsk->sk_reclen = 0;
 		svsk->sk_tcplen = 0;
 
+		/* initialise setting must have enough space to
+		 * receive and respond to one request.  
+		 * svc_tcp_recvfrom will re-adjust if necessary
+		 */
+		svc_sock_setbufsize(svsk->sk_sock,
+				    3 * svsk->sk_server->sv_bufsz,
+				    3 * svsk->sk_server->sv_bufsz);
+
 		set_bit(SK_CHNGBUF, &svsk->sk_flags);
 	}
 
@@ -1220,7 +1238,7 @@ svc_create_socket(struct svc_serv *serv, int protocol, struct sockaddr_in *sin)
 	}
 
 	if (protocol == IPPROTO_TCP) {
-		if ((error = sock->ops->listen(sock, 5)) < 0)
+		if ((error = sock->ops->listen(sock, 64)) < 0)
 			goto bummer;
 	}