aoe: avoid races between device destruction and discovery

This change avoids a race that could result in a NULL pointer derference
following a WARNing from kobject_add_internal, "don't try to register
things with the same name in the same directory."

The problem was found with a test that forgets and discovers an
aoe device in a loop:

  while test ! -r /tmp/stop; do
	aoe-flush -a
	aoe-discover
  done

The race was between aoedev_flush taking aoedevs out of the devlist,
allowing a new discovery of the same AoE target to take place before the
driver gets around to calling sysfs_remove_group.  Fixing that one
revealed another race between do_open and add_disk, and this patch avoids
that, too.

The fix required some care, because for flushing (forgetting) an aoedev,
some of the steps must be performed under lock and some must be able to
sleep.  Also, for discovering a new aoedev, some steps might sleep.

The check for a bad aoedev pointer remains from a time when about half of
this patch was done, and it was possible for the
bdev->bd_disk->private_data to become corrupted.  The check should be
removed eventually, but it is not expected to add significant overhead,
occurring in the aoeblk_open routine.

Signed-off-by: Ed Cashin <ecashin@coraid.com>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

3 files changed, 146 insertions, 63 deletions

diff --git a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
index b6d2b16358b..d50e9455b93 100644
--- a/drivers/block/aoe/aoe.h
+++ b/drivers/block/aoe/aoe.h
@@ -74,8 +74,11 @@ enum {
 	DEVFL_TKILL = (1<<1),	/* flag for timer to know when to kill self */
 	DEVFL_EXT = (1<<2),	/* device accepts lba48 commands */
 	DEVFL_GDALLOC = (1<<3),	/* need to alloc gendisk */
-	DEVFL_KICKME = (1<<4),	/* slow polling network card catch */
-	DEVFL_NEWSIZE = (1<<5),	/* need to update dev size in block layer */
+	DEVFL_GD_NOW = (1<<4),	/* allocating gendisk */
+	DEVFL_KICKME = (1<<5),	/* slow polling network card catch */
+	DEVFL_NEWSIZE = (1<<6),	/* need to update dev size in block layer */
+	DEVFL_FREEING = (1<<7),	/* set when device is being cleaned up */
+	DEVFL_FREED = (1<<8),	/* device has been cleaned up */
 };
 
 enum {
diff --git a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
index 57ac72c1715..6b5b7876ecf 100644
--- a/drivers/block/aoe/aoeblk.c
+++ b/drivers/block/aoe/aoeblk.c
@@ -147,9 +147,18 @@ aoeblk_open(struct block_device *bdev, fmode_t mode)
 	struct aoedev *d = bdev->bd_disk->private_data;
 	ulong flags;
 
+	if (!virt_addr_valid(d)) {
+		pr_crit("aoe: invalid device pointer in %s\n",
+			__func__);
+		WARN_ON(1);
+		return -ENODEV;
+	}
+	if (!(d->flags & DEVFL_UP) || d->flags & DEVFL_TKILL)
+		return -ENODEV;
+
 	mutex_lock(&aoeblk_mutex);
 	spin_lock_irqsave(&d->lock, flags);
-	if (d->flags & DEVFL_UP) {
+	if (d->flags & DEVFL_UP && !(d->flags & DEVFL_TKILL)) {
 		d->nopen++;
 		spin_unlock_irqrestore(&d->lock, flags);
 		mutex_unlock(&aoeblk_mutex);
@@ -259,6 +268,18 @@ aoeblk_gdalloc(void *vp)
 	struct request_queue *q;
 	enum { KB = 1024, MB = KB * KB, READ_AHEAD = 2 * MB, };
 	ulong flags;
+	int late = 0;
+
+	spin_lock_irqsave(&d->lock, flags);
+	if (d->flags & DEVFL_GDALLOC
+	&& !(d->flags & DEVFL_TKILL)
+	&& !(d->flags & DEVFL_GD_NOW))
+		d->flags |= DEVFL_GD_NOW;
+	else
+		late = 1;
+	spin_unlock_irqrestore(&d->lock, flags);
+	if (late)
+		return;
 
 	gd = alloc_disk(AOE_PARTITIONS);
 	if (gd == NULL) {
@@ -282,6 +303,11 @@ aoeblk_gdalloc(void *vp)
 	}
 
 	spin_lock_irqsave(&d->lock, flags);
+	WARN_ON(!(d->flags & DEVFL_GD_NOW));
+	WARN_ON(!(d->flags & DEVFL_GDALLOC));
+	WARN_ON(d->flags & DEVFL_TKILL);
+	WARN_ON(d->gd);
+	WARN_ON(d->flags & DEVFL_UP);
 	blk_queue_max_hw_sectors(q, BLK_DEF_MAX_SECTORS);
 	q->backing_dev_info.name = "aoe";
 	q->backing_dev_info.ra_pages = READ_AHEAD / PAGE_CACHE_SIZE;
@@ -306,6 +332,11 @@ aoeblk_gdalloc(void *vp)
 
 	add_disk(gd);
 	aoedisk_add_sysfs(d);
+
+	spin_lock_irqsave(&d->lock, flags);
+	WARN_ON(!(d->flags & DEVFL_GD_NOW));
+	d->flags &= ~DEVFL_GD_NOW;
+	spin_unlock_irqrestore(&d->lock, flags);
 	return;
 
 err_mempool:
@@ -314,7 +345,8 @@ err_disk:
 	put_disk(gd);
 err:
 	spin_lock_irqsave(&d->lock, flags);
-	d->flags &= ~DEVFL_GDALLOC;
+	d->flags &= ~DEVFL_GD_NOW;
+	schedule_work(&d->work);
 	spin_unlock_irqrestore(&d->lock, flags);
 }
 
diff --git a/drivers/block/aoe/aoedev.c b/drivers/block/aoe/aoedev.c
index f0c0c7416ae..3776715eb25 100644
--- a/drivers/block/aoe/aoedev.c
+++ b/drivers/block/aoe/aoedev.c
@@ -15,7 +15,6 @@
 #include "aoe.h"
 
 static void dummy_timer(ulong);
-static void aoedev_freedev(struct aoedev *);
 static void freetgt(struct aoedev *d, struct aoetgt *t);
 static void skbpoolfree(struct aoedev *d);
 
@@ -236,29 +235,6 @@ aoedev_downdev(struct aoedev *d)
 		set_capacity(d->gd, 0);
 }
 
-static void
-aoedev_freedev(struct aoedev *d)
-{
-	struct aoetgt **t, **e;
-
-	cancel_work_sync(&d->work);
-	if (d->gd) {
-		aoedisk_rm_sysfs(d);
-		del_gendisk(d->gd);
-		put_disk(d->gd);
-		blk_cleanup_queue(d->blkq);
-	}
-	t = d->targets;
-	e = t + NTARGETS;
-	for (; t < e && *t; t++)
-		freetgt(d, *t);
-	if (d->bufpool)
-		mempool_destroy(d->bufpool);
-	skbpoolfree(d);
-	minor_free(d->sysminor);
-	kfree(d);
-}
-
 /* return whether the user asked for this particular
  * device to be flushed
  */
@@ -283,17 +259,62 @@ user_req(char *s, size_t slen, struct aoedev *d)
 	return !strncmp(s, p, lim);
 }
 
-int
-aoedev_flush(const char __user *str, size_t cnt)
+static void
+freedev(struct aoedev *d)
+{
+	struct aoetgt **t, **e;
+	int freeing = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d->lock, flags);
+	if (d->flags & DEVFL_TKILL
+	&& !(d->flags & DEVFL_FREEING)) {
+		d->flags |= DEVFL_FREEING;
+		freeing = 1;
+	}
+	spin_unlock_irqrestore(&d->lock, flags);
+	if (!freeing)
+		return;
+
+	del_timer_sync(&d->timer);
+	if (d->gd) {
+		aoedisk_rm_sysfs(d);
+		del_gendisk(d->gd);
+		put_disk(d->gd);
+		blk_cleanup_queue(d->blkq);
+	}
+	t = d->targets;
+	e = t + NTARGETS;
+	for (; t < e && *t; t++)
+		freetgt(d, *t);
+	if (d->bufpool)
+		mempool_destroy(d->bufpool);
+	skbpoolfree(d);
+	minor_free(d->sysminor);
+
+	spin_lock_irqsave(&d->lock, flags);
+	d->flags |= DEVFL_FREED;
+	spin_unlock_irqrestore(&d->lock, flags);
+}
+
+enum flush_parms {
+	NOT_EXITING = 0,
+	EXITING = 1,
+};
+
+static int
+flush(const char __user *str, size_t cnt, int exiting)
 {
 	ulong flags;
 	struct aoedev *d, **dd;
-	struct aoedev *rmd = NULL;
 	char buf[16];
 	int all = 0;
 	int specified = 0;	/* flush a specific device */
+	unsigned int skipflags;
+
+	skipflags = DEVFL_GDALLOC | DEVFL_NEWSIZE | DEVFL_TKILL;
 
-	if (cnt >= 3) {
+	if (!exiting && cnt >= 3) {
 		if (cnt > sizeof buf)
 			cnt = sizeof buf;
 		if (copy_from_user(buf, str, cnt))
@@ -303,39 +324,71 @@ aoedev_flush(const char __user *str, size_t cnt)
 			specified = 1;
 	}
 
+	flush_scheduled_work();
+	/* pass one: without sleeping, do aoedev_downdev */
 	spin_lock_irqsave(&devlist_lock, flags);
-	dd = &devlist;
-	while ((d = *dd)) {
+	for (d = devlist; d; d = d->next) {
 		spin_lock(&d->lock);
-		if (specified) {
+		if (exiting) {
+			/* unconditionally take each device down */
+		} else if (specified) {
 			if (!user_req(buf, cnt, d))
-				goto skip;
+				goto cont;
 		} else if ((!all && (d->flags & DEVFL_UP))
-		|| (d->flags & (DEVFL_GDALLOC|DEVFL_NEWSIZE))
+		|| d->flags & skipflags
 		|| d->nopen
 		|| d->ref)
-			goto skip;
+			goto cont;
 
-		*dd = d->next;
 		aoedev_downdev(d);
 		d->flags |= DEVFL_TKILL;
+cont:
 		spin_unlock(&d->lock);
-		d->next = rmd;
-		rmd = d;
-		continue;
-skip:
-		spin_unlock(&d->lock);
-		dd = &d->next;
 	}
 	spin_unlock_irqrestore(&devlist_lock, flags);
-	while ((d = rmd)) {
-		rmd = d->next;
-		del_timer_sync(&d->timer);
-		aoedev_freedev(d);	/* must be able to sleep */
+
+	/* pass two: call freedev, which might sleep,
+	 * for aoedevs marked with DEVFL_TKILL
+	 */
+restart:
+	spin_lock_irqsave(&devlist_lock, flags);
+	for (d = devlist; d; d = d->next) {
+		spin_lock(&d->lock);
+		if (d->flags & DEVFL_TKILL
+		&& !(d->flags & DEVFL_FREEING)) {
+			spin_unlock(&d->lock);
+			spin_unlock_irqrestore(&devlist_lock, flags);
+			freedev(d);
+			goto restart;
+		}
+		spin_unlock(&d->lock);
 	}
+
+	/* pass three: remove aoedevs marked with DEVFL_FREED */
+	for (dd = &devlist, d = *dd; d; d = *dd) {
+		struct aoedev *doomed = NULL;
+
+		spin_lock(&d->lock);
+		if (d->flags & DEVFL_FREED) {
+			*dd = d->next;
+			doomed = d;
+		} else {
+			dd = &d->next;
+		}
+		spin_unlock(&d->lock);
+		kfree(doomed);
+	}
+	spin_unlock_irqrestore(&devlist_lock, flags);
+
 	return 0;
 }
 
+int
+aoedev_flush(const char __user *str, size_t cnt)
+{
+	return flush(str, cnt, NOT_EXITING);
+}
+
 /* This has been confirmed to occur once with Tms=3*1000 due to the
  * driver changing link and not processing its transmit ring.  The
  * problem is hard enough to solve by returning an error that I'm
@@ -388,7 +441,14 @@ aoedev_by_aoeaddr(ulong maj, int min, int do_alloc)
 
 	for (d=devlist; d; d=d->next)
 		if (d->aoemajor == maj && d->aoeminor == min) {
+			spin_lock(&d->lock);
+			if (d->flags & DEVFL_TKILL) {
+				spin_unlock(&d->lock);
+				d = NULL;
+				goto out;
+			}
 			d->ref++;
+			spin_unlock(&d->lock);
 			break;
 		}
 	if (d || !do_alloc || minor_get(&sysminor, maj, min) < 0)
@@ -448,21 +508,9 @@ freetgt(struct aoedev *d, struct aoetgt *t)
 void
 aoedev_exit(void)
 {
-	struct aoedev *d;
-	ulong flags;
-
+	flush_scheduled_work();
 	aoe_flush_iocq();
-	while ((d = devlist)) {
-		devlist = d->next;
-
-		spin_lock_irqsave(&d->lock, flags);
-		aoedev_downdev(d);
-		d->flags |= DEVFL_TKILL;
-		spin_unlock_irqrestore(&d->lock, flags);
-
-		del_timer_sync(&d->timer);
-		aoedev_freedev(d);
-	}
+	flush(NULL, 0, EXITING);
 }
 
 int __init