BUG: scheduling while atomic: ifdown-eth/0x00000100/335

Issue

The following message was seen before a system panic:

bonding: bond2: making interface eth5 the new active one.
BUG: scheduling while atomic: ifdown-eth/0x00000100/335

A resultant vmcore provides something like the following call trace within the kernel ring buffer:

Call Trace:
 [<ffffffff8006243d>] __sched_text_start+0x7d/0xbd6
 [<ffffffff8001a674>] vsnprintf+0x400/0x62f
 [<ffffffff80047571>] sprintf+0x51/0x59
 [<ffffffff80063c6f>] __mutex_lock_slowpath+0x60/0x9b
 [<ffffffff80063cb9>] .text.lock.mutex+0xf/0x14
 [<ffffffff881d6fbd>] :mlx4_en:mlx4_en_vlan_rx_kill_vid+0xbb/0x1d0
 [<ffffffff8010c920>] sysfs_hash_and_remove+0xca/0x10e
 [<ffffffff8837f989>] :bonding:bond_del_vlans_from_slave+0x6a/0xb9
 [<ffffffff88381ce8>] :bonding:bond_release+0x2e2/0x3e8
 [<ffffffff800645ab>] __down_write_nested+0x12/0x92
 [<ffffffff8838a477>] :bonding:bonding_store_slaves+0x25c/0x2f7
 [<ffffffff8010cff3>] sysfs_write_file+0xb9/0xe8
 [<ffffffff80016aa6>] vfs_write+0xce/0x174
 [<ffffffff80017373>] sys_write+0x45/0x6e
 [<ffffffff8005d28d>] tracesys+0xd5/0xe0