Skip to contentSkip to conversation

md: Introduction

PATCH New NeilBrown 2 participants 24 messages Mar 17, 2006

Subsystems (2)
lore
Share & tools
lore

NeilBrown

author
#0 lore raw

PATCH 1/1 md: Introduction

Following are 13 patches for md in 2.6.last (created against 2.6.16-rc6-mm1). They are NOT appropriate for 2.6.16 but should be OK for .17. Please include them in -mm.

The first three are assorted bug fixes, none really serious

The remainder implement raid5 reshaping. Currently the only shape change that is supported is added a device, but it is envisioned that changing the chunksize and layout will also be supported, as well as changing the level (e.g. 1->5, 5->6).

The reshape process naturally has to move all of the data in the array, and so should be used with caution. It is believed to work, and some testing does support this, but wider testing would be great for increasing my confidence.

You will need a version of mdadm newer than 2.3.1 to make use of raid5 growth. This is because mdadm need to take a copy of a 'critical section' at the start of the array incase there is a crash at an awkward moment. On restart, mdadm will restore the critical section and allow reshape to continue.

I hope to release a 2.4-pre by early next week - it still needs a little more polishing.

NeilBrown

[PATCH 001 of 13] md: Add '4' to the list of levels for which bitmaps are supported. [PATCH 002 of 13] md: Fix the 'failed' count for version-0 superblocks. [PATCH 003 of 13] md: Update status_resync to handle LARGE devices.

[PATCH 004 of 13] md: Split disks array out of raid5 conf structure so it is easier to grow. [PATCH 005 of 13] md: Allow stripes to be expanded in preparation for expanding an array. [PATCH 006 of 13] md: Infrastructure to allow normal IO to continue while array is expanding. [PATCH 007 of 13] md: Core of raid5 resize process
[PATCH 008 of 13] md: Final stages of raid5 expand code.
[PATCH 009 of 13] md: Checkpoint and allow restart of raid5 reshape
[PATCH 010 of 13] md: Only checkpoint expansion progress occasionally.
[PATCH 011 of 13] md: Split reshape handler in check_reshape and start_reshape. [PATCH 012 of 13] md: Make 'reshape' a possible sync_action action.
[PATCH 013 of 13] md: Support suspending of IO to regions of an md array.

author
#1 lore raw

PATCH 1/1 md: Add '4' to the list of levels for which bitmaps are supported.

I really should make this a function of the personality....

### Diffstat output

 ./drivers/md/md.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +2 -1
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -763,7 +763,8 @@ static int super_90_validate(mddev_t *md
 
 		if (sb->state & (1<<MD_SB_BITMAP_PRESENT) &&
 		    mddev->bitmap_file == NULL) {
-			if (mddev->level != 1 && mddev->level != 5 && mddev->level != 6
+			if (mddev->level != 1 && mddev->level != 4
+			    && mddev->level != 5 && mddev->level != 6
 			    && mddev->level != 10) {
 				/* FIXME use a better test */
 				printk(KERN_WARNING "md: bitmaps not supported for this level.\n");
author
#2 lore raw

PATCH 1/1 md: Fix the 'failed' count for version-0 superblocks.

In reply to NeilBrown

We are counting failed devices twice, once of the device that is failed, and once for the hole that has been left in the array. Remove the former so 'failed' matches 'missing'. Storing these counts in the superblock is a bit silly anyway....

### Diffstat output

 ./drivers/md/md.c |    5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +2 -3
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -893,10 +893,9 @@ static void super_90_sync(mddev_t *mddev
 			d->raid_disk = rdev2->raid_disk;
 		else
 			d->raid_disk = rdev2->desc_nr; /* compatibility */
-		if (test_bit(Faulty, &rdev2->flags)) {
+		if (test_bit(Faulty, &rdev2->flags))
 			d->state = (1<<MD_DISK_FAULTY);
-			failed++;
-		} else if (test_bit(In_sync, &rdev2->flags)) {
+		else if (test_bit(In_sync, &rdev2->flags)) {
 			d->state = (1<<MD_DISK_ACTIVE);
 			d->state |= (1<<MD_DISK_SYNC);
 			active++;
author
#3 lore raw

PATCH 1/1 md: Update status_resync to handle LARGE devices.

In reply to NeilBrown

status_resync - used by /proc/mdstat to report the status of a resync, assumes that device sizes will always fit into an 'unsigned long' This is no longer the case...

### Diffstat output

 ./drivers/md/md.c |   30 ++++++++++++++++++++++++------
 1 file changed, 24 insertions(+), 6 deletions(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +24 -6
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -4040,7 +4040,10 @@ static void status_unused(struct seq_fil
 
 static void status_resync(struct seq_file *seq, mddev_t * mddev)
 {
-	unsigned long max_blocks, resync, res, dt, db, rt;
+	sector_t max_blocks, resync, res;
+	unsigned long dt, db, rt;
+	int scale;
+	unsigned int per_milli;
 
 	resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active))/2;
 
@@ -4056,9 +4059,22 @@ static void status_resync(struct seq_fil
 		MD_BUG();
 		return;
 	}
-	res = (resync/1024)*1000/(max_blocks/1024 + 1);
+	/* Pick 'scale' such that (resync>>scale)*1000 will fit
+	 * in a sector_t, and (max_blocks>>scale) will fit in a
+	 * u32, as those are the requirements for sector_div.
+	 * Thus 'scale' must be at least 10
+	 */
+	scale = 10;
+	if (sizeof(sector_t) > sizeof(unsigned long)) {
+		while ( max_blocks/2 > (1ULL<<(scale+32)))
+			scale++;
+	}
+	res = (resync>>scale)*1000;
+	sector_div(res, (u32)((max_blocks>>scale)+1));
+
+	per_milli = res;
 	{
-		int i, x = res/50, y = 20-x;
+		int i, x = per_milli/50, y = 20-x;
 		seq_printf(seq, "[");
 		for (i = 0; i < x; i++)
 			seq_printf(seq, "=");
@@ -4067,10 +4083,12 @@ static void status_resync(struct seq_fil
 			seq_printf(seq, ".");
 		seq_printf(seq, "] ");
 	}
-	seq_printf(seq, " %s =%3lu.%lu%% (%lu/%lu)",
+	seq_printf(seq, " %s =%3u.%u%% (%llu/%llu)",
 		      (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ?
 		       "resync" : "recovery"),
-		      res/10, res % 10, resync, max_blocks);
+		      per_milli/10, per_milli % 10,
+		   (unsigned long long) resync,
+		   (unsigned long long) max_blocks);
 
 	/*
 	 * We do not want to overflow, so the order of operands and
@@ -4084,7 +4102,7 @@ static void status_resync(struct seq_fil
 	dt = ((jiffies - mddev->resync_mark) / HZ);
 	if (!dt) dt++;
 	db = resync - (mddev->resync_mark_cnt/2);
-	rt = (dt * ((max_blocks-resync) / (db/100+1)))/100;
+	rt = (dt * ((unsigned long)(max_blocks-resync) / (db/100+1)))/100;
 
 	seq_printf(seq, " finish=%lu.%lumin", rt / 60, (rt % 60)/6);
 
author
#4 lore raw

PATCH 1/1 md: Split disks array out of raid5 conf structure so it is easier to grow.

In reply to NeilBrown

Previously the array of disk information was included in the raid5 'conf' structure which was allocated to an appropriate size. This makes it awkward to change the size of that array. So we split it off into a separate kmalloced array which will require a little extra indexing, but is much easier to grow.

### Diffstat output

 ./drivers/md/raid5.c         |   10 +++++++---
 ./drivers/md/raid6main.c     |   10 +++++++---
 ./include/linux/raid/raid5.h |    2 +-
 3 files changed, 15 insertions(+), 7 deletions(-)
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +7 -3
diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
@@ -1822,11 +1822,13 @@ static int run(mddev_t *mddev)
 		return -EIO;
 	}
 
-	mddev->private = kzalloc(sizeof (raid5_conf_t)
-				 + mddev->raid_disks * sizeof(struct disk_info),
-				 GFP_KERNEL);
+	mddev->private = kzalloc(sizeof (raid5_conf_t), GFP_KERNEL);
 	if ((conf = mddev->private) == NULL)
 		goto abort;
+	conf->disks = kzalloc(mddev->raid_disks * sizeof(struct disk_info),
+			      GFP_KERNEL);
+	if (!conf->disks)
+		goto abort;
 
 	conf->mddev = mddev;
 
@@ -1966,6 +1968,7 @@ static int run(mddev_t *mddev)
 abort:
 	if (conf) {
 		print_raid5_conf(conf);
+		kfree(conf->disks);
 		kfree(conf->stripe_hashtbl);
 		kfree(conf);
 	}
@@ -1986,6 +1989,7 @@ static int stop(mddev_t *mddev)
 	kfree(conf->stripe_hashtbl);
 	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
 	sysfs_remove_group(&mddev->kobj, &raid5_attrs_group);
+	kfree(conf->disks);
 	kfree(conf);
 	mddev->private = NULL;
 	return 0;
drivers/md/raid6main.c renamed from drivers/md/raid6main.c~current~ +7 -3
diff ./drivers/md/raid6main.c~current~ ./drivers/md/raid6main.c
@@ -2006,11 +2006,14 @@ static int run(mddev_t *mddev)
 		return -EIO;
 	}
 
-	mddev->private = kzalloc(sizeof (raid6_conf_t)
-				 + mddev->raid_disks * sizeof(struct disk_info),
-				 GFP_KERNEL);
+	mddev->private = kzalloc(sizeof (raid6_conf_t), GFP_KERNEL);
 	if ((conf = mddev->private) == NULL)
 		goto abort;
+	conf->disks = kzalloc(mddev->raid_disks * sizeof(struct disk_info),
+				 GFP_KERNEL);
+	if (!conf->disks)
+		goto abort;
+
 	conf->mddev = mddev;
 
 	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
@@ -2158,6 +2161,7 @@ abort:
 		print_raid6_conf(conf);
 		safe_put_page(conf->spare_page);
 		kfree(conf->stripe_hashtbl);
+		kfree(conf->disks);
 		kfree(conf);
 	}
 	mddev->private = NULL;
include/linux/raid/raid5.h renamed from include/linux/raid/raid5.h~current~ +1 -1
diff ./include/linux/raid/raid5.h~current~ ./include/linux/raid/raid5.h
@@ -240,7 +240,7 @@ struct raid5_private_data {
 							 * waiting for 25% to be free
 							 */        
 	spinlock_t		device_lock;
-	struct disk_info	disks[0];
+	struct disk_info	*disks;
 };
 
 typedef struct raid5_private_data raid5_conf_t;
author
#5 lore raw

PATCH 1/1 md: Allow stripes to be expanded in preparation for expanding an array.

In reply to NeilBrown

Before a RAID-5 can be expanded, we need to be able to expand the stripe-cache data structure.
This requires allocating new stripes in a new kmem_cache. If this succeeds, we copy cache pages over and release the old stripes and kmem_cache.
We then allocate new pages. If that fails, we leave the stripe cache at it's new size. It isn't worth the effort to shrink it back again.

Unfortuanately this means we need two kmem_cache names as we, for a short period of time, we have two kmem_caches. So they are raid5/%s and raid5/%s-alt

### Diffstat output

 ./drivers/md/raid5.c         |  118 +++++++++++++++++++++++++++++++++++++++++--
 ./drivers/md/raid6main.c     |    4 -
 ./include/linux/raid/raid5.h |    9 ++-
 3 files changed, 123 insertions(+), 8 deletions(-)
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +114 -4
diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
@@ -313,20 +313,130 @@ static int grow_stripes(raid5_conf_t *co
 	kmem_cache_t *sc;
 	int devs = conf->raid_disks;
 
-	sprintf(conf->cache_name, "raid5/%s", mdname(conf->mddev));
-
-	sc = kmem_cache_create(conf->cache_name, 
+	sprintf(conf->cache_name[0], "raid5/%s", mdname(conf->mddev));
+	sprintf(conf->cache_name[1], "raid5/%s-alt", mdname(conf->mddev));
+	conf->active_name = 0;
+	sc = kmem_cache_create(conf->cache_name[conf->active_name],
 			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
 			       0, 0, NULL, NULL);
 	if (!sc)
 		return 1;
 	conf->slab_cache = sc;
+	conf->pool_size = devs;
 	while (num--) {
 		if (!grow_one_stripe(conf))
 			return 1;
 	}
 	return 0;
 }
+static int resize_stripes(raid5_conf_t *conf, int newsize)
+{
+	/* make all the stripes able to hold 'newsize' devices.
+	 * New slots in each stripe get 'page' set to a new page.
+	 * We allocate all the new stripes first, then if that succeeds,
+	 * copy everything across.
+	 * Finally we add new pages.  This could fail, but we leave
+	 * the stripe cache at it's new size, just with some pages empty.
+	 *
+	 * We use GFP_NOIO allocations as IO to the raid5 is blocked
+	 * at some points in this operation.
+	 */
+	struct stripe_head *osh, *nsh;
+	struct list_head newstripes, oldstripes;
+	struct disk_info *ndisks;
+	int err = 0;
+	kmem_cache_t *sc;
+	int i;
+
+	if (newsize <= conf->pool_size)
+		return 0; /* never bother to shrink */
+
+	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
+			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
+			       0, 0, NULL, NULL);
+	if (!sc)
+		return -ENOMEM;
+	INIT_LIST_HEAD(&newstripes);
+	for (i = conf->max_nr_stripes; i; i--) {
+		nsh = kmem_cache_alloc(sc, GFP_NOIO);
+		if (!nsh)
+			break;
+
+		memset(nsh, 0, sizeof(*nsh) + (newsize-1)*sizeof(struct r5dev));
+
+		nsh->raid_conf = conf;
+		spin_lock_init(&nsh->lock);
+
+		list_add(&nsh->lru, &newstripes);
+	}
+	if (i) {
+		/* didn't get enough, give up */
+		while (!list_empty(&newstripes)) {
+			nsh = list_entry(newstripes.next, struct stripe_head, lru);
+			list_del(&nsh->lru);
+			kmem_cache_free(sc, nsh);
+		}
+		kmem_cache_destroy(sc);
+		return -ENOMEM;
+	}
+	/* OK, we have enough stripes, start collecting inactive
+	 * stripes and copying them over
+	 */
+	INIT_LIST_HEAD(&oldstripes);
+	list_for_each_entry(nsh, &newstripes, lru) {
+		spin_lock_irq(&conf->device_lock);
+		wait_event_lock_irq(conf->wait_for_stripe,
+				    !list_empty(&conf->inactive_list),
+				    conf->device_lock,
+				    unplug_slaves(conf->mddev);
+			);
+		osh = get_free_stripe(conf);
+		spin_unlock_irq(&conf->device_lock);
+		atomic_set(&nsh->count, 1);
+		for(i=0; i<conf->pool_size; i++)
+			nsh->dev[i].page = osh->dev[i].page;
+		for( ; i<newsize; i++)
+			nsh->dev[i].page = NULL;
+		list_add(&osh->lru, &oldstripes);
+	}
+	/* Got them all.
+	 * Return the new ones and free the old ones.
+	 * At this point, we are holding all the stripes so the array
+	 * is completely stalled, so now is a good time to resize
+	 * conf->disks.
+	 */
+	ndisks = kzalloc(newsize * sizeof(struct disk_info), GFP_NOIO);
+	if (ndisks) {
+		for (i=0; i<conf->raid_disks; i++)
+			ndisks[i] = conf->disks[i];
+		kfree(conf->disks);
+		conf->disks = ndisks;
+	} else
+		err = -ENOMEM;
+	while(!list_empty(&newstripes)) {
+		nsh = list_entry(newstripes.next, struct stripe_head, lru);
+		list_del_init(&nsh->lru);
+		for (i=conf->raid_disks; i < newsize; i++)
+			if (nsh->dev[i].page == NULL) {
+				struct page *p = alloc_page(GFP_NOIO);
+				nsh->dev[i].page = p;
+				if (!p)
+					err = -ENOMEM;
+			}
+		release_stripe(nsh);
+	}
+	while(!list_empty(&oldstripes)) {
+		osh = list_entry(oldstripes.next, struct stripe_head, lru);
+		list_del(&osh->lru);
+		kmem_cache_free(conf->slab_cache, osh);
+	}
+	kmem_cache_destroy(conf->slab_cache);
+	conf->slab_cache = sc;
+	conf->active_name = 1-conf->active_name;
+	conf->pool_size = newsize;
+	return err;
+}
+
 
 static int drop_one_stripe(raid5_conf_t *conf)
 {
@@ -339,7 +449,7 @@ static int drop_one_stripe(raid5_conf_t 
 		return 0;
 	if (atomic_read(&sh->count))
 		BUG();
-	shrink_buffers(sh, conf->raid_disks);
+	shrink_buffers(sh, conf->pool_size);
 	kmem_cache_free(conf->slab_cache, sh);
 	atomic_dec(&conf->active_stripes);
 	return 1;
drivers/md/raid6main.c renamed from drivers/md/raid6main.c~current~ +2 -2
diff ./drivers/md/raid6main.c~current~ ./drivers/md/raid6main.c
@@ -331,9 +331,9 @@ static int grow_stripes(raid6_conf_t *co
 	kmem_cache_t *sc;
 	int devs = conf->raid_disks;
 
-	sprintf(conf->cache_name, "raid6/%s", mdname(conf->mddev));
+	sprintf(conf->cache_name[0], "raid6/%s", mdname(conf->mddev));
 
-	sc = kmem_cache_create(conf->cache_name,
+	sc = kmem_cache_create(conf->cache_name[0],
 			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
 			       0, 0, NULL, NULL);
 	if (!sc)
include/linux/raid/raid5.h renamed from include/linux/raid/raid5.h~current~ +7 -2
diff ./include/linux/raid/raid5.h~current~ ./include/linux/raid/raid5.h
@@ -216,7 +216,11 @@ struct raid5_private_data {
 	struct list_head	bitmap_list; /* stripes delaying awaiting bitmap update */
 	atomic_t		preread_active_stripes; /* stripes with scheduled io */
 
-	char			cache_name[20];
+	/* unfortunately we need two cache names as we temporarily have
+	 * two caches.
+	 */
+	int			active_name;
+	char			cache_name[2][20];
 	kmem_cache_t		*slab_cache; /* for allocating stripes */
 
 	int			seq_flush, seq_write;
@@ -238,7 +242,8 @@ struct raid5_private_data {
 	wait_queue_head_t	wait_for_overlap;
 	int			inactive_blocked;	/* release of inactive stripes blocked,
 							 * waiting for 25% to be free
-							 */        
+							 */
+	int			pool_size; /* number of disks in stripeheads in pool */
 	spinlock_t		device_lock;
 	struct disk_info	*disks;
 };
#6 lore raw

Re: [PATCH 005 of 13] md: Allow stripes to be expanded in preparation for expanding an array.

Quoted 15 lines · NeilBrown

+static int resize_stripes(raid5_conf_t *conf, int newsize)
+{
+ /* make all the stripes able to hold 'newsize' devices.
+ * New slots in each stripe get 'page' set to a new page.
+ * We allocate all the new stripes first, then if that succeeds,
+ * copy everything across.
+ * Finally we add new pages. This could fail, but we leave
+ * the stripe cache at it's new size, just with some pages empty.
+ *
+ * We use GFP_NOIO allocations as IO to the raid5 is blocked
+ * at some points in this operation.
+ */
+ struct stripe_head *osh, *nsh;
+ struct list_head newstripes, oldstripes;

You can use LIST_HEAD() here, avoid the separate INIT_LIST_HEAD().

Quoted 11 lines

+ struct disk_info *ndisks;
+ int err = 0;
+ kmem_cache_t *sc;
+ int i;
+
+ if (newsize <= conf->pool_size)
+ return 0; /* never bother to shrink */
+
+ sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
+ sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
+ 0, 0, NULL, NULL);

kmem_cache_create() internally does a GFP_KERNEL allocation.

Quoted 5 lines

+ if (!sc)
+ return -ENOMEM;
+ INIT_LIST_HEAD(&newstripes);
+ for (i = conf->max_nr_stripes; i; i--) {
+ nsh = kmem_cache_alloc(sc, GFP_NOIO);

So either this can use GFP_KERNEL, or we have a problem.

author
#7 lore raw

Re: [PATCH 005 of 13] md: Allow stripes to be expanded in preparation for expanding an array.

Quoted 17 lines · On Thursday March 16, [email protected] wrote:
Quoted 15 lines · NeilBrown

+static int resize_stripes(raid5_conf_t *conf, int newsize)
+{
+ /* make all the stripes able to hold 'newsize' devices.
+ * New slots in each stripe get 'page' set to a new page.
+ * We allocate all the new stripes first, then if that succeeds,
+ * copy everything across.
+ * Finally we add new pages. This could fail, but we leave
+ * the stripe cache at it's new size, just with some pages empty.
+ *
+ * We use GFP_NOIO allocations as IO to the raid5 is blocked
+ * at some points in this operation.
+ */
+ struct stripe_head *osh, *nsh;
+ struct list_head newstripes, oldstripes;

Quoted 2 lines

You can use LIST_HEAD() here, avoid the separate INIT_LIST_HEAD().

I guess.....
I have to have the declaration "miles" from where I use the variable. Do I have to have the initialisation equally far? Ok, I'll do that..

Quoted 15 lines
Quoted 12 lines

+ struct disk_info *ndisks;
+ int err = 0;
+ kmem_cache_t *sc;
+ int i;
+
+ if (newsize <= conf->pool_size)
+ return 0; /* never bother to shrink */
+
+ sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
+ sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
+ 0, 0, NULL, NULL);

kmem_cache_create() internally does a GFP_KERNEL allocation.

Quoted 7 lines
Quoted 5 lines

+ if (!sc)
+ return -ENOMEM;
+ INIT_LIST_HEAD(&newstripes);
+ for (i = conf->max_nr_stripes; i; i--) {
+ nsh = kmem_cache_alloc(sc, GFP_NOIO);

So either this can use GFP_KERNEL, or we have a problem.

Good point.... Maybe the comment about GFP_NOIO just needs to be improved.
We cannot risk waiting on IO after the

	/* OK, we have enough stripes, start collecting inactive
	 * stripes and copying them over
	 */

comment, up to the second last while loop, that starts

	while(!list_empty(&newstripes)) {

Before the comment, which where the kmem_cache_create is, is OK.

Thanks!

NeilBrown

#8 lore raw

Re: [PATCH 005 of 13] md: Allow stripes to be expanded in preparation for expanding an array.

In reply to NeilBrown

Review

Quoted 6 lines · NeilBrown
(quoted diff) +1 -0
+		wait_event_lock_irq(conf->wait_for_stripe,
 +				    !list_empty(&conf->inactive_list),
 +				    conf->device_lock,
 +				    unplug_slaves(conf->mddev);
 +			);

Boy, that's an ugly-looking thing, isn't it?

__wait_event_lock_irq() already puts a semicolon after `cmd' so I think the one here isn't needed, which would make it a bit less of a surprise to look at.

#9 lore raw

Re: [PATCH 005 of 13] md: Allow stripes to be expanded in preparation for expanding an array.

In reply to NeilBrown
Quoted 28 lines · NeilBrown
(quoted diff) +1 -0
+	/* Got them all.
 +	 * Return the new ones and free the old ones.
 +	 * At this point, we are holding all the stripes so the array
 +	 * is completely stalled, so now is a good time to resize
 +	 * conf->disks.
 +	 */
 +	ndisks = kzalloc(newsize * sizeof(struct disk_info), GFP_NOIO);
 +	if (ndisks) {
 +		for (i=0; i<conf->raid_disks; i++)
 +			ndisks[i] = conf->disks[i];
 +		kfree(conf->disks);
 +		conf->disks = ndisks;
 +	} else
 +		err = -ENOMEM;
 +	while(!list_empty(&newstripes)) {
 +		nsh = list_entry(newstripes.next, struct stripe_head, lru);
 +		list_del_init(&nsh->lru);
 +		for (i=conf->raid_disks; i < newsize; i++)
 +			if (nsh->dev[i].page == NULL) {
 +				struct page *p = alloc_page(GFP_NOIO);
 +				nsh->dev[i].page = p;
 +				if (!p)
 +					err = -ENOMEM;
 +			}
 +		release_stripe(nsh);
 +	}
 +	while(!list_empty(&oldstripes)) {

Review

Quoted 10 lines
(quoted diff) +0 -0
 +		osh = list_entry(oldstripes.next, struct stripe_head, lru);
 +		list_del(&osh->lru);
 +		kmem_cache_free(conf->slab_cache, osh);
 +	}
 +	kmem_cache_destroy(conf->slab_cache);
 +	conf->slab_cache = sc;
 +	conf->active_name = 1-conf->active_name;
 +	conf->pool_size = newsize;
 +	return err;
 +}

Are you sure the -ENOMEM handling here is solid? It looks.... strange.

There are a few more GFP_NOIOs in this function, which can possibly become GFP_KERNEL.

author
#10 lore raw

Re: [PATCH 005 of 13] md: Allow stripes to be expanded in preparation for expanding an array.

Quoted 34 lines · On Thursday March 16, [email protected] wrote:
Quoted 33 lines · NeilBrown
(quoted diff) +1 -0
+	/* Got them all.
 +	 * Return the new ones and free the old ones.
 +	 * At this point, we are holding all the stripes so the array
 +	 * is completely stalled, so now is a good time to resize
 +	 * conf->disks.
 +	 */
 +	ndisks = kzalloc(newsize * sizeof(struct disk_info), GFP_NOIO);
 +	if (ndisks) {
 +		for (i=0; i<conf->raid_disks; i++)
 +			ndisks[i] = conf->disks[i];
 +		kfree(conf->disks);
 +		conf->disks = ndisks;
 +	} else
 +		err = -ENOMEM;
 +	while(!list_empty(&newstripes)) {
 +		nsh = list_entry(newstripes.next, struct stripe_head, lru);
 +		list_del_init(&nsh->lru);
 +		for (i=conf->raid_disks; i < newsize; i++)
 +			if (nsh->dev[i].page == NULL) {
 +				struct page *p = alloc_page(GFP_NOIO);
 +				nsh->dev[i].page = p;
 +				if (!p)
 +					err = -ENOMEM;
 +			}
 +		release_stripe(nsh);
 +	}
 +	while(!list_empty(&oldstripes)) {
 +		osh = list_entry(oldstripes.next, struct stripe_head, lru);
 +		list_del(&osh->lru);
 +		kmem_cache_free(conf->slab_cache, osh);
 +	}
 +	kmem_cache_destroy(conf->slab_cache);
Quoted 8 lines
Quoted 5 lines

+ conf->slab_cache = sc;
+ conf->active_name = 1-conf->active_name;
+ conf->pool_size = newsize;
+ return err;
+}

Are you sure the -ENOMEM handling here is solid? It looks.... strange.

The philosophy of the -ENOMEM handling is (awkwardly?) embodied in the comment
* Finally we add new pages. This could fail, but we leave
* the stripe cache at it's new size, just with some pages empty.

at the top of the function. The core function here is making some data structures bigger. In each case, having a bigger data structure than required is no big deal. So we try to increase the size of each of them (the stripe_head cache, the 'disks' array, and the pages allocated to each stripe.
If any of there fail we return -ENOMEM, but may allow others to succeed.

Does that help?

NeilBrown

author
#11 lore raw

PATCH 1/1 md: Infrastructure to allow normal IO to continue while array is expanding.

In reply to NeilBrown

We need to allow that different stripes are of different effective sizes, and use the appropriate size.
Also, when a stripe is being expanded, we must block any IO attempts until the stripe is stable again.

Key elements in this change are:
- each stripe_head gets a 'disk' field which is part of the key, thus there can sometimes be two stripe heads of the same area of the array, but covering different numbers of devices. One of these will be marked STRIPE_EXPANDING and so won't accept new requests.
- conf->expand_progress tracks how the expansion is progressing and is used to determine whether the target part of the array has been expanded yet or not.

### Diffstat output

 ./drivers/md/raid5.c         |   88 ++++++++++++++++++++++++++++---------------
 ./include/linux/raid/raid5.h |    6 ++
 2 files changed, 64 insertions(+), 30 deletions(-)
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +58 -30
diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
@@ -178,10 +178,10 @@ static int grow_buffers(struct stripe_he
 
 static void raid5_build_block (struct stripe_head *sh, int i);
 
-static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx)
+static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx, int disks)
 {
 	raid5_conf_t *conf = sh->raid_conf;
-	int disks = conf->raid_disks, i;
+	int i;
 
 	if (atomic_read(&sh->count) != 0)
 		BUG();
@@ -198,7 +198,9 @@ static void init_stripe(struct stripe_he
 	sh->pd_idx = pd_idx;
 	sh->state = 0;
 
-	for (i=disks; i--; ) {
+	sh->disks = disks;
+
+	for (i = sh->disks; i--; ) {
 		struct r5dev *dev = &sh->dev[i];
 
 		if (dev->toread || dev->towrite || dev->written ||
@@ -215,7 +217,7 @@ static void init_stripe(struct stripe_he
 	insert_hash(conf, sh);
 }
 
-static struct stripe_head *__find_stripe(raid5_conf_t *conf, sector_t sector)
+static struct stripe_head *__find_stripe(raid5_conf_t *conf, sector_t sector, int disks)
 {
 	struct stripe_head *sh;
 	struct hlist_node *hn;
@@ -223,7 +225,7 @@ static struct stripe_head *__find_stripe
 	CHECK_DEVLOCK();
 	PRINTK("__find_stripe, sector %llu\n", (unsigned long long)sector);
 	hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash)
-		if (sh->sector == sector)
+		if (sh->sector == sector && sh->disks == disks)
 			return sh;
 	PRINTK("__stripe %llu not in cache\n", (unsigned long long)sector);
 	return NULL;
@@ -232,8 +234,8 @@ static struct stripe_head *__find_stripe
 static void unplug_slaves(mddev_t *mddev);
 static void raid5_unplug_device(request_queue_t *q);
 
-static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector,
-					     int pd_idx, int noblock) 
+static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector, int disks,
+					     int pd_idx, int noblock)
 {
 	struct stripe_head *sh;
 
@@ -245,7 +247,7 @@ static struct stripe_head *get_active_st
 		wait_event_lock_irq(conf->wait_for_stripe,
 				    conf->quiesce == 0,
 				    conf->device_lock, /* nothing */);
-		sh = __find_stripe(conf, sector);
+		sh = __find_stripe(conf, sector, disks);
 		if (!sh) {
 			if (!conf->inactive_blocked)
 				sh = get_free_stripe(conf);
@@ -263,7 +265,7 @@ static struct stripe_head *get_active_st
 					);
 				conf->inactive_blocked = 0;
 			} else
-				init_stripe(sh, sector, pd_idx);
+				init_stripe(sh, sector, pd_idx, disks);
 		} else {
 			if (atomic_read(&sh->count)) {
 				if (!list_empty(&sh->lru))
@@ -300,6 +302,7 @@ static int grow_one_stripe(raid5_conf_t 
 		kmem_cache_free(conf->slab_cache, sh);
 		return 0;
 	}
+	sh->disks = conf->raid_disks;
 	/* we just created an active stripe so... */
 	atomic_set(&sh->count, 1);
 	atomic_inc(&conf->active_stripes);
@@ -470,7 +473,7 @@ static int raid5_end_read_request(struct
 {
  	struct stripe_head *sh = bi->bi_private;
 	raid5_conf_t *conf = sh->raid_conf;
-	int disks = conf->raid_disks, i;
+	int disks = sh->disks, i;
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
 
 	if (bi->bi_size)
@@ -568,7 +571,7 @@ static int raid5_end_write_request (stru
 {
  	struct stripe_head *sh = bi->bi_private;
 	raid5_conf_t *conf = sh->raid_conf;
-	int disks = conf->raid_disks, i;
+	int disks = sh->disks, i;
 	unsigned long flags;
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
 
@@ -722,7 +725,7 @@ static sector_t raid5_compute_sector(sec
 static sector_t compute_blocknr(struct stripe_head *sh, int i)
 {
 	raid5_conf_t *conf = sh->raid_conf;
-	int raid_disks = conf->raid_disks, data_disks = raid_disks - 1;
+	int raid_disks = sh->disks, data_disks = raid_disks - 1;
 	sector_t new_sector = sh->sector, check;
 	int sectors_per_chunk = conf->chunk_size >> 9;
 	sector_t stripe;
@@ -823,8 +826,7 @@ static void copy_data(int frombio, struc
 
 static void compute_block(struct stripe_head *sh, int dd_idx)
 {
-	raid5_conf_t *conf = sh->raid_conf;
-	int i, count, disks = conf->raid_disks;
+	int i, count, disks = sh->disks;
 	void *ptr[MAX_XOR_BLOCKS], *p;
 
 	PRINTK("compute_block, stripe %llu, idx %d\n", 
@@ -854,7 +856,7 @@ static void compute_block(struct stripe_
 static void compute_parity(struct stripe_head *sh, int method)
 {
 	raid5_conf_t *conf = sh->raid_conf;
-	int i, pd_idx = sh->pd_idx, disks = conf->raid_disks, count;
+	int i, pd_idx = sh->pd_idx, disks = sh->disks, count;
 	void *ptr[MAX_XOR_BLOCKS];
 	struct bio *chosen;
 
@@ -1042,7 +1044,7 @@ static int add_stripe_bio(struct stripe_
 static void handle_stripe(struct stripe_head *sh)
 {
 	raid5_conf_t *conf = sh->raid_conf;
-	int disks = conf->raid_disks;
+	int disks = sh->disks;
 	struct bio *return_bi= NULL;
 	struct bio *bi;
 	int i;
@@ -1636,12 +1638,10 @@ static inline void raid5_plug_device(rai
 	spin_unlock_irq(&conf->device_lock);
 }
 
-static int make_request (request_queue_t *q, struct bio * bi)
+static int make_request(request_queue_t *q, struct bio * bi)
 {
 	mddev_t *mddev = q->queuedata;
 	raid5_conf_t *conf = mddev_to_conf(mddev);
-	const unsigned int raid_disks = conf->raid_disks;
-	const unsigned int data_disks = raid_disks - 1;
 	unsigned int dd_idx, pd_idx;
 	sector_t new_sector;
 	sector_t logical_sector, last_sector;
@@ -1665,20 +1665,48 @@ static int make_request (request_queue_t
 
 	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
 		DEFINE_WAIT(w);
+		int disks;
 		
-		new_sector = raid5_compute_sector(logical_sector,
-						  raid_disks, data_disks, &dd_idx, &pd_idx, conf);
-
+	retry:
+		if (likely(conf->expand_progress == MaxSector))
+			disks = conf->raid_disks;
+		else {
+			spin_lock_irq(&conf->device_lock);
+			disks = conf->raid_disks;
+			if (logical_sector >= conf->expand_progress)
+				disks = conf->previous_raid_disks;
+			spin_unlock_irq(&conf->device_lock);
+		}
+ 		new_sector = raid5_compute_sector(logical_sector, disks, disks - 1,
+						  &dd_idx, &pd_idx, conf);
 		PRINTK("raid5: make_request, sector %llu logical %llu\n",
 			(unsigned long long)new_sector, 
 			(unsigned long long)logical_sector);
 
-	retry:
 		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
-		sh = get_active_stripe(conf, new_sector, pd_idx, (bi->bi_rw&RWA_MASK));
+		sh = get_active_stripe(conf, new_sector, disks, pd_idx, (bi->bi_rw&RWA_MASK));
 		if (sh) {
-			if (!add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
-				/* Add failed due to overlap.  Flush everything
+			if (unlikely(conf->expand_progress != MaxSector)) {
+				/* expansion might have moved on while waiting for a
+				 * stripe, so we much do the range check again.
+				 */
+				int must_retry = 0;
+				spin_lock_irq(&conf->device_lock);
+				if (logical_sector <  conf->expand_progress &&
+				    disks == conf->previous_raid_disks)
+					/* mismatch, need to try again */
+					must_retry = 1;
+				spin_unlock_irq(&conf->device_lock);
+				if (must_retry) {
+					release_stripe(sh);
+					goto retry;
+				}
+			}
+
+			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
+			    !add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
+				/* Stripe is busy expanding or
+				 * add failed due to overlap.  Flush everything
 				 * and wait a while
 				 */
 				raid5_unplug_device(mddev->queue);
@@ -1690,7 +1718,6 @@ static int make_request (request_queue_t
 			raid5_plug_device(conf);
 			handle_stripe(sh);
 			release_stripe(sh);
-
 		} else {
 			/* cannot get stripe for read-ahead, just give-up */
 			clear_bit(BIO_UPTODATE, &bi->bi_flags);
@@ -1766,9 +1793,9 @@ static sector_t sync_request(mddev_t *md
 
 	first_sector = raid5_compute_sector((sector_t)stripe*data_disks*sectors_per_chunk
 		+ chunk_offset, raid_disks, data_disks, &dd_idx, &pd_idx, conf);
-	sh = get_active_stripe(conf, sector_nr, pd_idx, 1);
+	sh = get_active_stripe(conf, sector_nr, raid_disks, pd_idx, 1);
 	if (sh == NULL) {
-		sh = get_active_stripe(conf, sector_nr, pd_idx, 0);
+		sh = get_active_stripe(conf, sector_nr, raid_disks, pd_idx, 0);
 		/* make sure we don't swamp the stripe cache if someone else
 		 * is trying to get access 
 		 */
@@ -1985,6 +2012,7 @@ static int run(mddev_t *mddev)
 	conf->level = mddev->level;
 	conf->algorithm = mddev->layout;
 	conf->max_nr_stripes = NR_STRIPES;
+	conf->expand_progress = MaxSector;
 
 	/* device size must be a multiple of chunk size */
 	mddev->size &= ~(mddev->chunk_size/1024 -1);
@@ -2115,7 +2143,7 @@ static void print_sh (struct stripe_head
 	printk("sh %llu,  count %d.\n",
 		(unsigned long long)sh->sector, atomic_read(&sh->count));
 	printk("sh %llu, ", (unsigned long long)sh->sector);
-	for (i = 0; i < sh->raid_conf->raid_disks; i++) {
+	for (i = 0; i < sh->disks; i++) {
 		printk("(cache%d: %p %ld) ", 
 			i, sh->dev[i].page, sh->dev[i].flags);
 	}
include/linux/raid/raid5.h renamed from include/linux/raid/raid5.h~current~ +6 -0
diff ./include/linux/raid/raid5.h~current~ ./include/linux/raid/raid5.h
@@ -135,6 +135,7 @@ struct stripe_head {
 	atomic_t		count;			/* nr of active thread/requests */
 	spinlock_t		lock;
 	int			bm_seq;	/* sequence number for bitmap flushes */
+	int			disks;			/* disks in stripe */
 	struct r5dev {
 		struct bio	req;
 		struct bio_vec	vec;
@@ -174,6 +175,7 @@ struct stripe_head {
 #define	STRIPE_DELAYED		6
 #define	STRIPE_DEGRADED		7
 #define	STRIPE_BIT_DELAY	8
+#define	STRIPE_EXPANDING	9
 
 /*
  * Plugging:
@@ -211,6 +213,10 @@ struct raid5_private_data {
 	int			raid_disks, working_disks, failed_disks;
 	int			max_nr_stripes;
 
+	/* used during an expand */
+	sector_t		expand_progress;	/* MaxSector when no expand happening */
+	int			previous_raid_disks;
+
 	struct list_head	handle_list; /* stripes needing handling */
 	struct list_head	delayed_list; /* stripes that have plugged requests */
 	struct list_head	bitmap_list; /* stripes delaying awaiting bitmap update */
#12 lore raw

Re: [PATCH 006 of 13] md: Infrastructure to allow normal IO to continue while array is expanding.

Quoted 16 lines · NeilBrown

- retry:
prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
- sh = get_active_stripe(conf, new_sector, pd_idx, (bi->bi_rw&RWA_MASK));
+ sh = get_active_stripe(conf, new_sector, disks, pd_idx, (bi->bi_rw&RWA_MASK));
if (sh) {
- if (!add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
- /* Add failed due to overlap. Flush everything
+ if (unlikely(conf->expand_progress != MaxSector)) {
+ /* expansion might have moved on while waiting for a
+ * stripe, so we much do the range check again.
+ */
+ int must_retry = 0;
+ spin_lock_irq(&conf->device_lock);
+ if (logical_sector < conf->expand_progress &&
+ disks == conf->previous_raid_disks)

Quoted 8 lines

+ /* mismatch, need to try again */
+ must_retry = 1;
+ spin_unlock_irq(&conf->device_lock);
+ if (must_retry) {
+ release_stripe(sh);
+ goto retry;
+ }
+ }

The locking in here looks strange. We take the lock, do some arithmetic and some tests and then drop the lock again. Is it not possible that the result of those tests now becomes invalid?

author
#13 lore raw

Re: [PATCH 006 of 13] md: Infrastructure to allow normal IO to continue while array is expanding.

Quoted 26 lines · On Thursday March 16, [email protected] wrote:
Quoted 16 lines · NeilBrown

- retry:
prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
- sh = get_active_stripe(conf, new_sector, pd_idx, (bi->bi_rw&RWA_MASK));
+ sh = get_active_stripe(conf, new_sector, disks, pd_idx, (bi->bi_rw&RWA_MASK));
if (sh) {
- if (!add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
- /* Add failed due to overlap. Flush everything
+ if (unlikely(conf->expand_progress != MaxSector)) {
+ /* expansion might have moved on while waiting for a
+ * stripe, so we much do the range check again.
+ */
+ int must_retry = 0;
+ spin_lock_irq(&conf->device_lock);
+ if (logical_sector < conf->expand_progress &&
+ disks == conf->previous_raid_disks)

Quoted 8 lines

+ /* mismatch, need to try again */
+ must_retry = 1;
+ spin_unlock_irq(&conf->device_lock);
+ if (must_retry) {
+ release_stripe(sh);
+ goto retry;
+ }
+ }

Quoted 3 lines

The locking in here looks strange. We take the lock, do some arithmetic and some tests and then drop the lock again. Is it not possible that the result of those tests now becomes invalid?

Obviously another comment missing.
conf->expand_progress is sector_t and so could be 64bits on a 32 bit platform, and so I cannot be sure it is updated atomically. So I always access it within a lock (unless I am comparing for equality with ~0).

Yes, the result can become invalid, but only in one direction: As expand_progress always increases, it is possible that it will pass logical_sector. When that happens, STRIPE_EXPANDING gets set on the stripe_head at logical_sector.
So because we took a reference to logical_sector *before* this test, and check for stripe_expanding *after* the test, we can easily catch that transition.

Putting it another way, this test is to catch cases where logical_sector is a long way from expand_progress, the subsequent test of STRIPE_EXPANDING catches cases where they are close together, and the ordering wrt get_stripe_active ensures there are no holes.

Now to put that into a few short-but-clear comments.

Thanks again!

NeilBrown

author
#14 lore raw

PATCH 1/1 md: Core of raid5 resize process

In reply to NeilBrown

This patch provides the core of the resize/expand process.

sync_request notices if a 'reshape' is happening and acts accordingly.

It allocated new stripe_heads for the next chunk-wide-stripe in the target geometry, marking them STRIPE_EXPANDING.
Then it finds which stripe heads in the old geometry can provide data needed by these and marks them STRIPE_EXPAND_SOURCE. This causes stripe_handle to read all blocks on those stripes.
Once all blocks on a STRIPE_EXPAND_SOURCE stripe_head are read, any that are needed are copied into the corresponding STRIPE_EXPANDING stripe_head. Once a STRIPE_EXPANDING stripe_head is full, it is marks STRIPE_EXPAND_READY and then is written out and released.

### Diffstat output

 ./drivers/md/md.c            |   14 ++-
 ./drivers/md/raid5.c         |  185 +++++++++++++++++++++++++++++++++++++------
 ./include/linux/raid/md_k.h  |    4 
 ./include/linux/raid/raid5.h |    4 
 4 files changed, 181 insertions(+), 26 deletions(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +11 -3
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -2161,7 +2161,9 @@ action_show(mddev_t *mddev, char *page)
 	char *type = "idle";
 	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
 	    test_bit(MD_RECOVERY_NEEDED, &mddev->recovery)) {
-		if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
+		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+			type = "reshape";
+		else if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
 			if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
 				type = "resync";
 			else if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
@@ -4084,8 +4086,10 @@ static void status_resync(struct seq_fil
 		seq_printf(seq, "] ");
 	}
 	seq_printf(seq, " %s =%3u.%u%% (%llu/%llu)",
+		   (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)?
+		    "reshape" :
 		      (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ?
-		       "resync" : "recovery"),
+		       "resync" : "recovery")),
 		      per_milli/10, per_milli % 10,
 		   (unsigned long long) resync,
 		   (unsigned long long) max_blocks);
@@ -4539,7 +4543,9 @@ static void md_do_sync(mddev_t *mddev)
 		 */
 		max_sectors = mddev->resync_max_sectors;
 		mddev->resync_mismatches = 0;
-	} else
+	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+		max_sectors = mddev->size << 1;
+	else
 		/* recovery follows the physical size of devices */
 		max_sectors = mddev->size << 1;
 
@@ -4675,6 +4681,8 @@ static void md_do_sync(mddev_t *mddev)
 	mddev->pers->sync_request(mddev, max_sectors, &skipped, 1);
 
 	if (!test_bit(MD_RECOVERY_ERR, &mddev->recovery) &&
+	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
+	    !test_bit(MD_RECOVERY_CHECK, &mddev->recovery) &&
 	    mddev->curr_resync > 2 &&
 	    mddev->curr_resync >= mddev->recovery_cp) {
 		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +163 -22

266 lines not on this page · raw message

include/linux/raid/md_k.h renamed from include/linux/raid/md_k.h~current~ +4 -0
diff ./include/linux/raid/md_k.h~current~ ./include/linux/raid/md_k.h
@@ -157,6 +157,9 @@ struct mddev_s
 	 * DONE:     thread is done and is waiting to be reaped
 	 * REQUEST:  user-space has requested a sync (used with SYNC)
 	 * CHECK:    user-space request for for check-only, no repair
+	 * RESHAPE:  A reshape is happening
+	 *
+	 * If neither SYNC or RESHAPE are set, then it is a recovery.
 	 */
 #define	MD_RECOVERY_RUNNING	0
 #define	MD_RECOVERY_SYNC	1
@@ -166,6 +169,7 @@ struct mddev_s
 #define	MD_RECOVERY_NEEDED	5
 #define	MD_RECOVERY_REQUESTED	6
 #define	MD_RECOVERY_CHECK	7
+#define MD_RECOVERY_RESHAPE	8
 	unsigned long			recovery;
 
 	int				in_sync;	/* know to not need resync */
include/linux/raid/raid5.h renamed from include/linux/raid/raid5.h~current~ +3 -1
diff ./include/linux/raid/raid5.h~current~ ./include/linux/raid/raid5.h
@@ -157,6 +157,7 @@ struct stripe_head {
 #define	R5_ReadError	8	/* seen a read error here recently */
 #define	R5_ReWrite	9	/* have tried to over-write the readerror */
 
+#define	R5_Expanded	10	/* This block now has post-expand data */
 /*
  * Write method
  */
@@ -176,7 +177,8 @@ struct stripe_head {
 #define	STRIPE_DEGRADED		7
 #define	STRIPE_BIT_DELAY	8
 #define	STRIPE_EXPANDING	9
-
+#define	STRIPE_EXPAND_SOURCE	10
+#define	STRIPE_EXPAND_READY	11
 /*
  * Plugging:
  *
#15 lore raw

Re: [PATCH 007 of 13] md: Core of raid5 resize process

drivers/md/md.c static void md_do_sync(mddev_t *mddev) · line 4550

Quoted 12 lines · NeilBrown
drivers/md/md.c renamed from drivers/md/md.c~current~ +0 -0
@@ -4539,7 +4543,9 @@ static void md_do_sync(mddev_t *mddev)
  		 */
  		max_sectors = mddev->resync_max_sectors;
  		mddev->resync_mismatches = 0;
 -	} else
 +	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 +		max_sectors = mddev->size << 1;
 +	else
  		/* recovery follows the physical size of devices */
  		max_sectors = mddev->size << 1;

This change is a no-op. Intentional?

author
#16 lore raw

Re: [PATCH 007 of 13] md: Core of raid5 resize process

Quoted 15 lines · On Thursday March 16, [email protected] wrote:

drivers/md/md.c static void md_do_sync(mddev_t *mddev) · line 4550

Quoted 12 lines · NeilBrown
drivers/md/md.c renamed from drivers/md/md.c~current~ +0 -0
@@ -4539,7 +4543,9 @@ static void md_do_sync(mddev_t *mddev)
  		 */
  		max_sectors = mddev->resync_max_sectors;
  		mddev->resync_mismatches = 0;
 -	} else
 +	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 +		max_sectors = mddev->size << 1;
 +	else
  		/* recovery follows the physical size of devices */
  		max_sectors = mddev->size << 1;

This change is a no-op. Intentional?

Uhmm... sort of.
A later patch adds stuff to the later branch but not the middle one. This comes from creating a patch to fix a bug, then merging it back into the wrong original patch...

NeilBrown

author
#17 lore raw

PATCH 1/1 md: Final stages of raid5 expand code.

In reply to NeilBrown

This patch adds raid5_reshape and end_reshape which will start and finish the reshape processes.

raid5_reshape is only enabled in CONFIG_MD_RAID5_RESHAPE is set, to discourage accidental use.

Read the 'help' for the CONFIG_MD_RAID5_RESHAPE entry.

and Make sure that you have backups, just in case.

### Diffstat output

 ./drivers/md/Kconfig      |   26 ++++++++++
 ./drivers/md/md.c         |    7 +-
 ./drivers/md/raid5.c      |  117 ++++++++++++++++++++++++++++++++++++++++++++++
 ./include/linux/raid/md.h |    3 -
 4 files changed, 149 insertions(+), 4 deletions(-)
drivers/md/Kconfig renamed from drivers/md/Kconfig~current~ +26 -0
diff ./drivers/md/Kconfig~current~ ./drivers/md/Kconfig
@@ -127,6 +127,32 @@ config MD_RAID5
 
 	  If unsure, say Y.
 
+config MD_RAID5_RESHAPE
+	bool "Support adding drives to a raid-5 array (experimental)"
+	depends on MD_RAID5 && EXPERIMENTAL
+	---help---
+	  A RAID-5 set can be expanded by adding extra drives. This
+	  requires "restriping" the array which means (almost) every
+	  block must be written to a different place.
+
+          This option allows such restriping to be done while the array
+	  is online.  However it is still EXPERIMENTAL code.  It should
+	  work, but please be sure that you have backups.
+
+	  You will need a version of mdadm newer than 2.3.1.   During the
+	  early stage of reshape there is a critical section where live data
+	  is being over-written.  A crash during this time needs extra care
+	  for recovery.  The newer mdadm takes a copy of the data in the
+	  critical section and will restore it, if necessary, after a crash.
+
+	  The mdadm usage is e.g.
+	       mdadm --grow /dev/md1 --raid-disks=6
+	  to grow '/dev/md1' to having 6 disks.
+
+	  Note: The array can only be expanded, not contracted.
+	  There should be enough spares already present to make the new
+	  array workable.
+
 config MD_RAID6
 	tristate "RAID-6 mode"
 	depends on BLK_DEV_MD
drivers/md/md.c renamed from drivers/md/md.c~current~ +4 -3
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -159,12 +159,12 @@ static int start_readonly;
  */
 static DECLARE_WAIT_QUEUE_HEAD(md_event_waiters);
 static atomic_t md_event_count;
-static void md_new_event(mddev_t *mddev)
+void md_new_event(mddev_t *mddev)
 {
 	atomic_inc(&md_event_count);
 	wake_up(&md_event_waiters);
 }
-
+EXPORT_SYMBOL_GPL(md_new_event);
 /*
  * Enables to iterate over all existing md arrays
  * all_mddevs_lock protects this list.
@@ -4463,7 +4463,7 @@ static DECLARE_WAIT_QUEUE_HEAD(resync_wa
 
 #define SYNC_MARKS	10
 #define	SYNC_MARK_STEP	(3*HZ)
-static void md_do_sync(mddev_t *mddev)
+void md_do_sync(mddev_t *mddev)
 {
 	mddev_t *mddev2;
 	unsigned int currspeed = 0,
@@ -4700,6 +4700,7 @@ static void md_do_sync(mddev_t *mddev)
 	set_bit(MD_RECOVERY_DONE, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
 }
+EXPORT_SYMBOL_GPL(md_do_sync);
 
 
 /*
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +117 -0

141 lines not on this page · raw message

include/linux/raid/md.h renamed from include/linux/raid/md.h~current~ +2 -1
diff ./include/linux/raid/md.h~current~ ./include/linux/raid/md.h
@@ -92,7 +92,8 @@ extern void md_super_write(mddev_t *mdde
 extern void md_super_wait(mddev_t *mddev);
 extern int sync_page_io(struct block_device *bdev, sector_t sector, int size,
 			struct page *page, int rw);
-
+extern void md_do_sync(mddev_t *mddev);
+extern void md_new_event(mddev_t *mddev);
 
 #define MD_BUG(x...) { printk("md: bug in file %s, line %d\n", __FILE__, __LINE__); md_print_devices(); }
 
author
#18 lore raw

PATCH 1/1 md: Checkpoint and allow restart of raid5 reshape

In reply to NeilBrown

We allow the superblock to record an 'old' and a 'new' geometry, and a position where any conversion is up to. The geometry allows for changing chunksize, layout and level as well as number of devices.

When using verion-0.90 superblock, we convert the version to 0.91 while the conversion is happening so that an old kernel will refuse the assemble the array.
For version-1, we use a feature bit for the same effect.

When starting an array we check for an incomplete reshape and restart the reshape process if needed.
If the reshape stopped at an awkward time (like when updating the first stripe) we refuse to assemble the array, and let user-space worry about it.

### Diffstat output

 ./drivers/md/md.c            |   69 ++++++++++++++++++++-
 ./drivers/md/raid1.c         |    5 +
 ./drivers/md/raid5.c         |  140 ++++++++++++++++++++++++++++++++++++-------
 ./include/linux/raid/md.h    |    2 
 ./include/linux/raid/md_k.h  |    8 ++
 ./include/linux/raid/md_p.h  |   32 ++++++++-
 ./include/linux/raid/raid5.h |    1 
 7 files changed, 231 insertions(+), 26 deletions(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +66 -3
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -659,7 +659,8 @@ static int super_90_load(mdk_rdev_t *rde
 	}
 
 	if (sb->major_version != 0 ||
-	    sb->minor_version != 90) {
+	    sb->minor_version < 90 ||
+	    sb->minor_version > 91) {
 		printk(KERN_WARNING "Bad version number %d.%d on %s\n",
 			sb->major_version, sb->minor_version,
 			b);
@@ -744,6 +745,20 @@ static int super_90_validate(mddev_t *md
 		mddev->bitmap_offset = 0;
 		mddev->default_bitmap_offset = MD_SB_BYTES >> 9;
 
+		if (mddev->minor_version >= 91) {
+			mddev->reshape_position = sb->reshape_position;
+			mddev->delta_disks = sb->delta_disks;
+			mddev->new_level = sb->new_level;
+			mddev->new_layout = sb->new_layout;
+			mddev->new_chunk = sb->new_chunk;
+		} else {
+			mddev->reshape_position = MaxSector;
+			mddev->delta_disks = 0;
+			mddev->new_level = mddev->level;
+			mddev->new_layout = mddev->layout;
+			mddev->new_chunk = mddev->chunk_size;
+		}
+
 		if (sb->state & (1<<MD_SB_CLEAN))
 			mddev->recovery_cp = MaxSector;
 		else {
@@ -838,7 +853,6 @@ static void super_90_sync(mddev_t *mddev
 
 	sb->md_magic = MD_SB_MAGIC;
 	sb->major_version = mddev->major_version;
-	sb->minor_version = mddev->minor_version;
 	sb->patch_version = mddev->patch_version;
 	sb->gvalid_words  = 0; /* ignored */
 	memcpy(&sb->set_uuid0, mddev->uuid+0, 4);
@@ -857,6 +871,17 @@ static void super_90_sync(mddev_t *mddev
 	sb->events_hi = (mddev->events>>32);
 	sb->events_lo = (u32)mddev->events;
 
+	if (mddev->reshape_position == MaxSector)
+		sb->minor_version = 90;
+	else {
+		sb->minor_version = 91;
+		sb->reshape_position = mddev->reshape_position;
+		sb->new_level = mddev->new_level;
+		sb->delta_disks = mddev->delta_disks;
+		sb->new_layout = mddev->new_layout;
+		sb->new_chunk = mddev->new_chunk;
+	}
+	mddev->minor_version = sb->minor_version;
 	if (mddev->in_sync)
 	{
 		sb->recovery_cp = mddev->recovery_cp;
@@ -1101,6 +1126,20 @@ static int super_1_validate(mddev_t *mdd
 			}
 			mddev->bitmap_offset = (__s32)le32_to_cpu(sb->bitmap_offset);
 		}
+		if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
+			mddev->reshape_position = le64_to_cpu(sb->reshape_position);
+			mddev->delta_disks = le32_to_cpu(sb->delta_disks);
+			mddev->new_level = le32_to_cpu(sb->new_level);
+			mddev->new_layout = le32_to_cpu(sb->new_layout);
+			mddev->new_chunk = le32_to_cpu(sb->new_chunk)<<9;
+		} else {
+			mddev->reshape_position = MaxSector;
+			mddev->delta_disks = 0;
+			mddev->new_level = mddev->level;
+			mddev->new_layout = mddev->layout;
+			mddev->new_chunk = mddev->chunk_size;
+		}
+
 	} else if (mddev->pers == NULL) {
 		/* Insist of good event counter while assembling */
 		__u64 ev1 = le64_to_cpu(sb->events);
@@ -1172,6 +1211,14 @@ static void super_1_sync(mddev_t *mddev,
 		sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_offset);
 		sb->feature_map = cpu_to_le32(MD_FEATURE_BITMAP_OFFSET);
 	}
+	if (mddev->reshape_position != MaxSector) {
+		sb->feature_map |= cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE);
+		sb->reshape_position = cpu_to_le64(mddev->reshape_position);
+		sb->new_layout = cpu_to_le32(mddev->new_layout);
+		sb->delta_disks = cpu_to_le32(mddev->delta_disks);
+		sb->new_level = cpu_to_le32(mddev->new_level);
+		sb->new_chunk = cpu_to_le32(mddev->new_chunk>>9);
+	}
 
 	max_dev = 0;
 	ITERATE_RDEV(mddev,rdev2,tmp)
@@ -1492,7 +1539,7 @@ static void sync_sbs(mddev_t * mddev)
 	}
 }
 
-static void md_update_sb(mddev_t * mddev)
+void md_update_sb(mddev_t * mddev)
 {
 	int err;
 	struct list_head *tmp;
@@ -1569,6 +1616,7 @@ repeat:
 	wake_up(&mddev->sb_wait);
 
 }
+EXPORT_SYMBOL_GPL(md_update_sb);
 
 /* words written to sysfs files may, or my not, be \n terminated.
  * We want to accept with case. For this we use cmd_match.
@@ -2540,6 +2588,14 @@ static int do_md_run(mddev_t * mddev)
 	mddev->level = pers->level;
 	strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
 
+	if (mddev->reshape_position != MaxSector &&
+	    pers->reshape == NULL) {
+		/* This personality cannot handle reshaping... */
+		mddev->pers = NULL;
+		module_put(pers->owner);
+		return -EINVAL;
+	}
+
 	mddev->recovery = 0;
 	mddev->resync_max_sectors = mddev->size << 1; /* may be over-ridden by personality */
 	mddev->barriers_work = 1;
@@ -3428,11 +3484,18 @@ static int set_array_info(mddev_t * mdde
 	mddev->default_bitmap_offset = MD_SB_BYTES >> 9;
 	mddev->bitmap_offset = 0;
 
+	mddev->reshape_position = MaxSector;
+
 	/*
 	 * Generate a 128 bit UUID
 	 */
 	get_random_bytes(mddev->uuid, 16);
 
+	mddev->new_level = mddev->level;
+	mddev->new_chunk = mddev->chunk_size;
+	mddev->new_layout = mddev->layout;
+	mddev->delta_disks = 0;
+
 	return 0;
 }
 
drivers/md/raid1.c renamed from drivers/md/raid1.c~current~ +5 -0
diff ./drivers/md/raid1.c~current~ ./drivers/md/raid1.c
@@ -1789,6 +1789,11 @@ static int run(mddev_t *mddev)
 		       mdname(mddev), mddev->level);
 		goto out;
 	}
+	if (mddev->reshape_position != MaxSector) {
+		printk("raid1: %s: reshape_position set but not supported\n",
+		       mdname(mddev));
+		goto out;
+	}
 	/*
 	 * copy the already verified devices into our private RAID1
 	 * bookkeeping area. [whatever we allocate in run(),
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +120 -20

237 lines not on this page · raw message

include/linux/raid/md.h renamed from include/linux/raid/md.h~current~ +2 -0
diff ./include/linux/raid/md.h~current~ ./include/linux/raid/md.h
@@ -95,6 +95,8 @@ extern int sync_page_io(struct block_dev
 extern void md_do_sync(mddev_t *mddev);
 extern void md_new_event(mddev_t *mddev);
 
+extern void md_update_sb(mddev_t * mddev);
+
 #define MD_BUG(x...) { printk("md: bug in file %s, line %d\n", __FILE__, __LINE__); md_print_devices(); }
 
 #endif 
include/linux/raid/md_k.h renamed from include/linux/raid/md_k.h~current~ +8 -0
diff ./include/linux/raid/md_k.h~current~ ./include/linux/raid/md_k.h
@@ -132,6 +132,14 @@ struct mddev_s
 
 	char				uuid[16];
 
+	/* If the array is being reshaped, we need to record the
+	 * new shape and an indication of where we are up to.
+	 * This is written to the superblock.
+	 * If reshape_position is MaxSector, then no reshape is happening (yet).
+	 */
+	sector_t			reshape_position;
+	int				delta_disks, new_level, new_layout, new_chunk;
+
 	struct mdk_thread_s		*thread;	/* management thread */
 	struct mdk_thread_s		*sync_thread;	/* doing resync or reconstruct */
 	sector_t			curr_resync;	/* blocks scheduled */
include/linux/raid/md_p.h renamed from include/linux/raid/md_p.h~current~ +29 -3
diff ./include/linux/raid/md_p.h~current~ ./include/linux/raid/md_p.h
@@ -102,6 +102,18 @@ typedef struct mdp_device_descriptor_s {
 #define MD_SB_ERRORS		1
 
 #define	MD_SB_BITMAP_PRESENT	8 /* bitmap may be present nearby */
+
+/*
+ * Notes:
+ * - if an array is being reshaped (restriped) in order to change the
+ *   the number of active devices in the array, 'raid_disks' will be
+ *   the larger of the old and new numbers.  'delta_disks' will
+ *   be the "new - old".  So if +ve, raid_disks is the new value, and
+ *   "raid_disks-delta_disks" is the old.  If -ve, raid_disks is the
+ *   old value and "raid_disks+delta_disks" is the new (smaller) value.
+ */
+
+
 typedef struct mdp_superblock_s {
 	/*
 	 * Constant generic information
@@ -146,7 +158,13 @@ typedef struct mdp_superblock_s {
 	__u32 cp_events_hi;	/* 10 high-order of checkpoint update count   */
 #endif
 	__u32 recovery_cp;	/* 11 recovery checkpoint sector count	      */
-	__u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 12];
+	/* There are only valid for minor_version > 90 */
+	__u64 reshape_position;	/* 12,13 next address in array-space for reshape */
+	__u32 new_level;	/* 14 new level we are reshaping to	      */
+	__u32 delta_disks;	/* 15 change in number of raid_disks	      */
+	__u32 new_layout;	/* 16 new layout			      */
+	__u32 new_chunk;	/* 17 new chunk size (bytes)		      */
+	__u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18];
 
 	/*
 	 * Personality information
@@ -207,7 +225,14 @@ struct mdp_superblock_1 {
 				 * NOTE: signed, so bitmap can be before superblock
 				 * only meaningful of feature_map[0] is set.
 				 */
-	__u8	pad1[128-100];	/* set to 0 when written */
+
+	/* These are only valid with feature bit '4' */
+	__u64	reshape_position;	/* next address in array-space for reshape */
+	__u32	new_level;	/* new level we are reshaping to		*/
+	__u32	delta_disks;	/* change in number of raid_disks		*/
+	__u32	new_layout;	/* new layout					*/
+	__u32	new_chunk;	/* new chunk size (bytes)			*/
+	__u8	pad1[128-124];	/* set to 0 when written */
 
 	/* constant this-device information - 64 bytes */
 	__u64	data_offset;	/* sector start of data, often 0 */
@@ -240,8 +265,9 @@ struct mdp_superblock_1 {
 
 /* feature_map bits */
 #define MD_FEATURE_BITMAP_OFFSET	1
+#define	MD_FEATURE_RESHAPE_ACTIVE	4
 
-#define	MD_FEATURE_ALL			1
+#define	MD_FEATURE_ALL			5
 
 #endif 
 
include/linux/raid/raid5.h renamed from include/linux/raid/raid5.h~current~ +1 -0
diff ./include/linux/raid/raid5.h~current~ ./include/linux/raid/raid5.h
@@ -224,6 +224,7 @@ struct raid5_private_data {
 	struct list_head	bitmap_list; /* stripes delaying awaiting bitmap update */
 	atomic_t		preread_active_stripes; /* stripes with scheduled io */
 
+	atomic_t		reshape_stripes; /* stripes with pending writes for reshape */
 	/* unfortunately we need two cache names as we temporarily have
 	 * two caches.
 	 */
author
#19 lore raw

PATCH 1/1 md: Only checkpoint expansion progress occasionally.

In reply to NeilBrown

Instead of checkpointing at each stripe, only checkpoint when a new write would overwrite uncheckpointed data. Block any write to the uncheckpointed area. Arbitrarily checkpoint at least every 3Meg.

### Diffstat output

 ./drivers/md/raid5.c         |   53 ++++++++++++++++++++++++++++++++++---------
 ./include/linux/raid/raid5.h |    3 ++
 2 files changed, 45 insertions(+), 11 deletions(-)
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +42 -11
diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
@@ -1747,8 +1747,9 @@ static int make_request(request_queue_t 
 	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
 		DEFINE_WAIT(w);
 		int disks;
-		
+
 	retry:
+		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
 		if (likely(conf->expand_progress == MaxSector))
 			disks = conf->raid_disks;
 		else {
@@ -1756,6 +1757,13 @@ static int make_request(request_queue_t 
 			disks = conf->raid_disks;
 			if (logical_sector >= conf->expand_progress)
 				disks = conf->previous_raid_disks;
+			else {
+				if (logical_sector >= conf->expand_lo) {
+					spin_unlock_irq(&conf->device_lock);
+					schedule();
+					goto retry;
+				}
+			}
 			spin_unlock_irq(&conf->device_lock);
 		}
  		new_sector = raid5_compute_sector(logical_sector, disks, disks - 1,
@@ -1764,7 +1772,6 @@ static int make_request(request_queue_t 
 			(unsigned long long)new_sector, 
 			(unsigned long long)logical_sector);
 
-		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
 		sh = get_active_stripe(conf, new_sector, disks, pd_idx, (bi->bi_rw&RWA_MASK));
 		if (sh) {
 			if (unlikely(conf->expand_progress != MaxSector)) {
@@ -1862,6 +1869,7 @@ static sector_t sync_request(mddev_t *md
 		 */
 		int i;
 		int dd_idx;
+		sector_t writepos, safepos, gap;
 
 		if (sector_nr == 0 &&
 		    conf->expand_progress != 0) {
@@ -1872,15 +1880,36 @@ static sector_t sync_request(mddev_t *md
 			return sector_nr;
 		}
 
-		/* Cannot proceed until we've updated the superblock... */
-		wait_event(conf->wait_for_overlap,
-			   atomic_read(&conf->reshape_stripes)==0);
-		mddev->reshape_position = conf->expand_progress;
-
-		mddev->sb_dirty = 1;
-		md_wakeup_thread(mddev->thread);
-		wait_event(mddev->sb_wait, mddev->sb_dirty == 0 ||
-			kthread_should_stop());
+		/* we update the metadata when there is more than 3Meg
+		 * in the block range (that is rather arbitrary, should
+		 * probably be time based) or when the data about to be
+		 * copied would over-write the source of the data at
+		 * the front of the range.
+		 * i.e. one new_stripe forward from expand_progress new_maps
+		 * to after where expand_lo old_maps to
+		 */
+		writepos = conf->expand_progress +
+			conf->chunk_size/512*(conf->raid_disks-1);
+		sector_div(writepos, conf->raid_disks-1);
+		safepos = conf->expand_lo;
+		sector_div(safepos, conf->previous_raid_disks-1);
+		gap = conf->expand_progress - conf->expand_lo;
+
+		if (writepos >= safepos ||
+		    gap > (conf->raid_disks-1)*3000*2 /*3Meg*/) {
+			/* Cannot proceed until we've updated the superblock... */
+			wait_event(conf->wait_for_overlap,
+				   atomic_read(&conf->reshape_stripes)==0);
+			mddev->reshape_position = conf->expand_progress;
+			mddev->sb_dirty = 1;
+			md_wakeup_thread(mddev->thread);
+			wait_event(mddev->sb_wait, mddev->sb_dirty == 0 ||
+				   kthread_should_stop());
+			spin_lock_irq(&conf->device_lock);
+			conf->expand_lo = mddev->reshape_position;
+			spin_unlock_irq(&conf->device_lock);
+			wake_up(&conf->wait_for_overlap);
+		}
 
 		for (i=0; i < conf->chunk_size/512; i+= STRIPE_SECTORS) {
 			int j;
@@ -2307,6 +2336,7 @@ static int run(mddev_t *mddev)
 
 	if (conf->expand_progress != MaxSector) {
 		printk("...ok start reshape thread\n");
+		conf->expand_lo = conf->expand_progress;
 		atomic_set(&conf->reshape_stripes, 0);
 		clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 		clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
@@ -2594,6 +2624,7 @@ static int raid5_reshape(mddev_t *mddev,
 	conf->previous_raid_disks = conf->raid_disks;
 	conf->raid_disks = raid_disks;
 	conf->expand_progress = 0;
+	conf->expand_lo = 0;
 	spin_unlock_irq(&conf->device_lock);
 
 	/* Add some new drives, as many as will fit.
include/linux/raid/raid5.h renamed from include/linux/raid/raid5.h~current~ +3 -0
diff ./include/linux/raid/raid5.h~current~ ./include/linux/raid/raid5.h
@@ -217,6 +217,9 @@ struct raid5_private_data {
 
 	/* used during an expand */
 	sector_t		expand_progress;	/* MaxSector when no expand happening */
+	sector_t		expand_lo; /* from here up to expand_progress it out-of-bounds
+					    * as we haven't flushed the metadata yet
+					    */
 	int			previous_raid_disks;
 
 	struct list_head	handle_list; /* stripes needing handling */
#20 lore raw

Re: [PATCH 010 of 13] md: Only checkpoint expansion progress occasionally.

Quoted 5 lines · NeilBrown

diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
--- ./drivers/md/raid5.c~current~ 2006-03-17 11:48:58.000000000 +1100
+++ ./drivers/md/raid5.c 2006-03-17 11:48:58.000000000 +1100
@@ -1747,8 +1747,9 @@ static int make_request(request_queue_t

That's a fairly complex function..

I wonder about this:

	spin_lock_irq(&conf->device_lock);
	if (--bi->bi_phys_segments == 0) {
		int bytes = bi->bi_size;

		if ( bio_data_dir(bi) == WRITE )
			md_write_end(mddev);
		bi->bi_size = 0;
		bi->bi_end_io(bi, bytes, 0);
	}
	spin_unlock_irq(&conf->device_lock);

bi_end_io() can be somewhat expensive. Does it need to happen under the lock?

author
#21 lore raw

PATCH 1/1 md: Split reshape handler in check_reshape and start_reshape.

In reply to NeilBrown

check_reshape checks validity and does things that can be done instantly - like adding devices to raid1.
start_reshape initiates a restriping process to convert the whole array.

### Diffstat output

 ./drivers/md/md.c           |   10 ++++---
 ./drivers/md/raid1.c        |   19 +++++++++++--
 ./drivers/md/raid5.c        |   60 ++++++++++++++++++++++++--------------------
 ./include/linux/raid/md_k.h |    3 +-
 4 files changed, 58 insertions(+), 34 deletions(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +6 -4
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -2589,7 +2589,7 @@ static int do_md_run(mddev_t * mddev)
 	strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
 
 	if (mddev->reshape_position != MaxSector &&
-	    pers->reshape == NULL) {
+	    pers->start_reshape == NULL) {
 		/* This personality cannot handle reshaping... */
 		mddev->pers = NULL;
 		module_put(pers->owner);
@@ -3551,14 +3551,16 @@ static int update_raid_disks(mddev_t *md
 {
 	int rv;
 	/* change the number of raid disks */
-	if (mddev->pers->reshape == NULL)
+	if (mddev->pers->check_reshape == NULL)
 		return -EINVAL;
 	if (raid_disks <= 0 ||
 	    raid_disks >= mddev->max_disks)
 		return -EINVAL;
-	if (mddev->sync_thread)
+	if (mddev->sync_thread || mddev->reshape_position != MaxSector)
 		return -EBUSY;
-	rv = mddev->pers->reshape(mddev, raid_disks);
+	mddev->delta_disks = raid_disks - mddev->raid_disks;
+
+	rv = mddev->pers->check_reshape(mddev);
 	return rv;
 }
 
drivers/md/raid1.c renamed from drivers/md/raid1.c~current~ +16 -3
diff ./drivers/md/raid1.c~current~ ./drivers/md/raid1.c
@@ -1976,7 +1976,7 @@ static int raid1_resize(mddev_t *mddev, 
 	return 0;
 }
 
-static int raid1_reshape(mddev_t *mddev, int raid_disks)
+static int raid1_reshape(mddev_t *mddev)
 {
 	/* We need to:
 	 * 1/ resize the r1bio_pool
@@ -1993,10 +1993,22 @@ static int raid1_reshape(mddev_t *mddev,
 	struct pool_info *newpoolinfo;
 	mirror_info_t *newmirrors;
 	conf_t *conf = mddev_to_conf(mddev);
-	int cnt;
+	int cnt, raid_disks;
 
 	int d, d2;
 
+	/* Cannot change chunk_size, layout, or level */
+	if (mddev->chunk_size != mddev->new_chunk ||
+	    mddev->layout != mddev->new_layout ||
+	    mddev->level != mddev->new_level) {
+		mddev->new_chunk = mddev->chunk_size;
+		mddev->new_layout = mddev->layout;
+		mddev->new_level = mddev->level;
+		return -EINVAL;
+	}
+
+	raid_disks = mddev->raid_disks + mddev->delta_disks;
+
 	if (raid_disks < conf->raid_disks) {
 		cnt=0;
 		for (d= 0; d < conf->raid_disks; d++)
@@ -2043,6 +2055,7 @@ static int raid1_reshape(mddev_t *mddev,
 
 	mddev->degraded += (raid_disks - conf->raid_disks);
 	conf->raid_disks = mddev->raid_disks = raid_disks;
+	mddev->delta_disks = 0;
 
 	conf->last_used = 0; /* just make sure it is in-range */
 	lower_barrier(conf);
@@ -2084,7 +2097,7 @@ static struct mdk_personality raid1_pers
 	.spare_active	= raid1_spare_active,
 	.sync_request	= sync_request,
 	.resize		= raid1_resize,
-	.reshape	= raid1_reshape,
+	.check_reshape	= raid1_reshape,
 	.quiesce	= raid1_quiesce,
 };
 
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +34 -26

111 lines not on this page · raw message

include/linux/raid/md_k.h renamed from include/linux/raid/md_k.h~current~ +2 -1
diff ./include/linux/raid/md_k.h~current~ ./include/linux/raid/md_k.h
@@ -261,7 +261,8 @@ struct mdk_personality
 	int (*spare_active) (mddev_t *mddev);
 	sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
 	int (*resize) (mddev_t *mddev, sector_t sectors);
-	int (*reshape) (mddev_t *mddev, int raid_disks);
+	int (*check_reshape) (mddev_t *mddev);
+	int (*start_reshape) (mddev_t *mddev);
 	int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
 	/* quiesce moves between quiescence states
 	 * 0 - fully active
author
#22 lore raw

PATCH 1/1 md: Make 'reshape' a possible sync_action action.

In reply to NeilBrown

This allows reshape to be triggerred via sysfs (which is the only way to start it happening).

### Diffstat output

 ./drivers/md/md.c |    9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/md/md.c renamed from drivers/md/md.c~current~ +8 -1
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -2242,7 +2242,14 @@ action_store(mddev_t *mddev, const char 
 		return -EBUSY;
 	else if (cmd_match(page, "resync") || cmd_match(page, "recover"))
 		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
-	else {
+	else if (cmd_match(page, "reshape")) {
+		int err;
+		if (mddev->pers->start_reshape == NULL)
+			return -EINVAL;
+		err = mddev->pers->start_reshape(mddev);
+		if (err)
+			return err;
+	} else {
 		if (cmd_match(page, "check"))
 			set_bit(MD_RECOVERY_CHECK, &mddev->recovery);
 		else if (cmd_match(page, "repair"))
author
#23 lore raw

PATCH 1/1 md: Support suspending of IO to regions of an md array.

In reply to NeilBrown

This allows user-space to access data safely.
This is needed for raid5 reshape as user-space needs to take a backup of the first few stripes before allowing reshape to commense.
It will also be useful in cluster-aware raid1 configurations so that all cluster members can leave a section of the array untouched while a resync/recovery happens.

A 'start' and 'end' of the suspended range are written to 2 sysfs attributes. Note that only one range can be suspended at a time.

### Diffstat output

 ./drivers/md/md.c           |   59 ++++++++++++++++++++++++++++++++++++++++++++
 ./drivers/md/raid5.c        |   14 ++++++++++
 ./include/linux/raid/md_k.h |    4 ++
 3 files changed, 77 insertions(+)
drivers/md/md.c renamed from drivers/md/md.c~current~ +59 -0
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
@@ -2360,6 +2360,63 @@ sync_completed_show(mddev_t *mddev, char
 static struct md_sysfs_entry
 md_sync_completed = __ATTR_RO(sync_completed);
 
+static ssize_t
+suspend_lo_show(mddev_t *mddev, char *page)
+{
+	return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_lo);
+}
+
+static ssize_t
+suspend_lo_store(mddev_t *mddev, const char *buf, size_t len)
+{
+	char *e;
+	unsigned long long new = simple_strtoull(buf, &e, 10);
+
+	if (mddev->pers->quiesce == NULL)
+		return -EINVAL;
+	if (buf == e || (*e && *e != '\n'))
+		return -EINVAL;
+	if (new >= mddev->suspend_hi ||
+	    (new > mddev->suspend_lo && new < mddev->suspend_hi)) {
+		mddev->suspend_lo = new;
+		mddev->pers->quiesce(mddev, 2);
+		return len;
+	} else
+		return -EINVAL;
+}
+static struct md_sysfs_entry md_suspend_lo =
+__ATTR(suspend_lo, S_IRUGO|S_IWUSR, suspend_lo_show, suspend_lo_store);
+
+
+static ssize_t
+suspend_hi_show(mddev_t *mddev, char *page)
+{
+	return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_hi);
+}
+
+static ssize_t
+suspend_hi_store(mddev_t *mddev, const char *buf, size_t len)
+{
+	char *e;
+	unsigned long long new = simple_strtoull(buf, &e, 10);
+
+	if (mddev->pers->quiesce == NULL)
+		return -EINVAL;
+	if (buf == e || (*e && *e != '\n'))
+		return -EINVAL;
+	if ((new <= mddev->suspend_lo && mddev->suspend_lo >= mddev->suspend_hi) ||
+	    (new > mddev->suspend_lo && new > mddev->suspend_hi)) {
+		mddev->suspend_hi = new;
+		mddev->pers->quiesce(mddev, 1);
+		mddev->pers->quiesce(mddev, 0);
+		return len;
+	} else
+		return -EINVAL;
+}
+static struct md_sysfs_entry md_suspend_hi =
+__ATTR(suspend_hi, S_IRUGO|S_IWUSR, suspend_hi_show, suspend_hi_store);
+
+
 static struct attribute *md_default_attrs[] = {
 	&md_level.attr,
 	&md_raid_disks.attr,
@@ -2377,6 +2434,8 @@ static struct attribute *md_redundancy_a
 	&md_sync_max.attr,
 	&md_sync_speed.attr,
 	&md_sync_completed.attr,
+	&md_suspend_lo.attr,
+	&md_suspend_hi.attr,
 	NULL,
 };
 static struct attribute_group md_redundancy_group = {
drivers/md/raid5.c renamed from drivers/md/raid5.c~current~ +14 -0
diff ./drivers/md/raid5.c~current~ ./drivers/md/raid5.c
@@ -1790,6 +1790,15 @@ static int make_request(request_queue_t 
 					goto retry;
 				}
 			}
+			/* FIXME what if we get a false positive because these
+			 * are being updated.
+			 */
+			if (logical_sector >= mddev->suspend_lo &&
+			    logical_sector < mddev->suspend_hi) {
+				release_stripe(sh);
+				schedule();
+				goto retry;
+			}
 
 			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
 			    !add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
@@ -2708,6 +2717,10 @@ static void raid5_quiesce(mddev_t *mddev
 	raid5_conf_t *conf = mddev_to_conf(mddev);
 
 	switch(state) {
+	case 2: /* resume for a suspend */
+		wake_up(&conf->wait_for_overlap);
+		break;
+
 	case 1: /* stop all writes */
 		spin_lock_irq(&conf->device_lock);
 		conf->quiesce = 1;
@@ -2721,6 +2734,7 @@ static void raid5_quiesce(mddev_t *mddev
 		spin_lock_irq(&conf->device_lock);
 		conf->quiesce = 0;
 		wake_up(&conf->wait_for_stripe);
+		wake_up(&conf->wait_for_overlap);
 		spin_unlock_irq(&conf->device_lock);
 		break;
 	}
include/linux/raid/md_k.h renamed from include/linux/raid/md_k.h~current~ +4 -0
diff ./include/linux/raid/md_k.h~current~ ./include/linux/raid/md_k.h
@@ -151,6 +151,10 @@ struct mddev_s
 	sector_t			resync_mismatches; /* count of sectors where
 							    * parity/replica mismatch found
 							    */
+
+	/* allow user-space to request suspension of IO to regions of the array */
+	sector_t			suspend_lo;
+	sector_t			suspend_hi;
 	/* if zero, use the system-wide default */
 	int				sync_speed_min;
 	int				sync_speed_max;