Move partition scanning from userspace to module.

Currently, zpool online -e (dynamic vdev expansion) doesn't work on
whole disks because we're invoking ioctl(BLKRRPART) from userspace
while ZFS still has a partition open on the disk, which results in
EBUSY.

This patch moves the BLKRRPART invocation from the zpool utility to the
module. Specifically, this is done just before opening the device in
vdev_disk_open() which is called inside vdev_reopen(). This requires
jumping through some hoops to get to the disk device from the partition
device, and to make sure we can still open the partition after the
BLKRRPART call.

Note that this new code path is triggered on dynamic vdev expansion
only; other actions, like creating a new pool, are unchanged and still
call BLKRRPART from userspace.

This change also depends on API changes which are available in 2.6.37
and latter kernels.  The build system has been updated to detect this,
but there is no compatibility mode for older kernels.  This means that
online expansion will NOT be available in older kernels.  However, it
will still be possible to expand the vdev offline.

Signed-off-by: Brian Behlendorf <[email protected]>
Closes #808

1 files changed, 70 insertions, 2 deletions

diff --git a/module/zfs/vdev_disk.c b/module/zfs/vdev_disk.c
index 28a4861ab..eee03d080 100644
--- a/module/zfs/vdev_disk.c
+++ b/module/zfs/vdev_disk.c
@@ -158,10 +158,75 @@ vdev_elevator_switch(vdev_t *v, char *elevator)
 	return (error);
 }
 
+/*
+ * Expanding a whole disk vdev involves invoking BLKRRPART on the
+ * whole disk device. This poses a problem, because BLKRRPART will
+ * return EBUSY if one of the disk's partitions is open. That's why
+ * we have to do it here, just before opening the data partition.
+ * Unfortunately, BLKRRPART works by dropping all partitions and
+ * recreating them, which means that for a short time window, all
+ * /dev/sdxN device files disappear (until udev recreates them).
+ * This means two things:
+ *  - When we open the data partition just after a BLKRRPART, we
+ *    can't do it using the normal device file path because of the
+ *    obvious race condition with udev. Instead, we use reliable
+ *    kernel APIs to get a handle to the new partition device from
+ *    the whole disk device.
+ *  - Because vdev_disk_open() initially needs to find the device
+ *    using its path, multiple vdev_disk_open() invocations in
+ *    short succession on the same disk with BLKRRPARTs in the
+ *    middle have a high probability of failure (because of the
+ *    race condition with udev). A typical situation where this
+ *    might happen is when the zpool userspace tool does a
+ *    TRYIMPORT immediately followed by an IMPORT. For this
+ *    reason, we only invoke BLKRRPART in the module when strictly
+ *    necessary (zpool online -e case), and rely on userspace to
+ *    do it when possible.
+ */
+static struct block_device *
+vdev_disk_rrpart(const char *path, int mode, vdev_disk_t *vd)
+{
+#if defined(HAVE_3ARG_BLKDEV_GET) && defined(HAVE_GET_GENDISK)
+	struct block_device *bdev, *result = ERR_PTR(-ENXIO);
+	struct gendisk *disk;
+	int error, partno;
+
+	bdev = vdev_bdev_open(path, vdev_bdev_mode(mode), vd);
+	if (IS_ERR(bdev))
+		return bdev;
+
+	disk = get_gendisk(bdev->bd_dev, &partno);
+	vdev_bdev_close(bdev, vdev_bdev_mode(mode));
+
+	if (disk) {
+		bdev = bdget(disk_devt(disk));
+		if (bdev) {
+			error = blkdev_get(bdev, vdev_bdev_mode(mode), vd);
+			if (error == 0)
+				error = ioctl_by_bdev(bdev, BLKRRPART, 0);
+			vdev_bdev_close(bdev, vdev_bdev_mode(mode));
+		}
+
+		bdev = bdget_disk(disk, partno);
+		if (bdev) {
+			error = blkdev_get(bdev,
+			    vdev_bdev_mode(mode) | FMODE_EXCL, vd);
+			if (error == 0)
+				result = bdev;
+		}
+		put_disk(disk);
+	}
+
+	return result;
+#else
+	return ERR_PTR(-EOPNOTSUPP);
+#endif /* defined(HAVE_3ARG_BLKDEV_GET) && defined(HAVE_GET_GENDISK) */
+}
+
 static int
 vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *ashift)
 {
-	struct block_device *bdev;
+	struct block_device *bdev = ERR_PTR(-ENXIO);
 	vdev_disk_t *vd;
 	int mode, block_size;
 
@@ -190,7 +255,10 @@ vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *ashift)
 	 * level vdev validation.
 	 */
 	mode = spa_mode(v->vdev_spa);
-	bdev = vdev_bdev_open(v->vdev_path, vdev_bdev_mode(mode), vd);
+	if (v->vdev_wholedisk && v->vdev_expanding)
+		bdev = vdev_disk_rrpart(v->vdev_path, mode, vd);
+	if (IS_ERR(bdev))
+		bdev = vdev_bdev_open(v->vdev_path, vdev_bdev_mode(mode), vd);
 	if (IS_ERR(bdev)) {
 		kmem_free(vd, sizeof(vdev_disk_t));
 		return -PTR_ERR(bdev);