Fix /etc/hostid on root pool deadlock

Accidentally introduced by dc04a8c which now takes the SCL_VDEV lock
as a reader in zfs_blkptr_verify().  A deadlock can occur if the
/etc/hostid file resides on a dataset in the same pool.  This is
because reading the /etc/hostid file may occur while the caller is
holding the SCL_VDEV lock as a writer.  For example, to perform a
`zpool attach` as shown in the abbreviated stack below.

To resolve the issue we cache the system's hostid when initializing
the spa_t, or when modifying the multihost property.  The cached
value is then relied upon for subsequent accesses.

Call Trace:
    spa_config_enter+0x1e8/0x350 [zfs]
    zfs_blkptr_verify+0x33c/0x4f0 [zfs] <--- trying read lock
    zio_read+0x6c/0x140 [zfs]
    ...
    vfs_read+0xfc/0x1e0
    kernel_read+0x50/0x90
    ...
    spa_get_hostid+0x1c/0x38 [zfs]
    spa_config_generate+0x1a0/0x610 [zfs]
    vdev_label_init+0xa0/0xc80 [zfs]
    vdev_create+0x98/0xe0 [zfs]
    spa_vdev_attach+0x14c/0xb40 [zfs] <--- grabbed write lock

Reviewed-by: loli10K <ezomori.nozomu@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #9256
Closes #9285

include/sys/spa.h

@@ -1104,7 +1104,7 @@ extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
 extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
 extern boolean_t spa_multihost(spa_t *spa);
-extern unsigned long spa_get_hostid(void);
+extern uint32_t spa_get_hostid(spa_t *spa);
 extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
 
 extern int spa_mode(spa_t *spa);

include/sys/spa_impl.h

@@ -395,6 +395,7 @@ struct spa {
 	mmp_thread_t	spa_mmp;		/* multihost mmp thread */
 	list_t		spa_leaf_list;		/* list of leaf vdevs */
 	uint64_t	spa_leaf_list_gen;	/* track leaf_list changes */
+	uint32_t	spa_hostid;		/* cached system hostid */
 
 	/*
 	 * spa_refcount & spa_config_lock must be the last elements