]> pilppa.com Git - linux-2.6-omap-h63xx.git/commit
bonding: fix panic when taking bond interface down before removing module
authorAndy Gospodarek <andy@greyhouse.net>
Fri, 31 Oct 2008 00:41:16 +0000 (17:41 -0700)
committerJeff Garzik <jgarzik@redhat.com>
Fri, 31 Oct 2008 04:46:12 +0000 (00:46 -0400)
commitce39a800ea87c655de49af021c8b20ee323cb40d
treef461c679d9c03639256501a221bbd8542fe65858
parenta434e43f3d844192bc23bd7b408bac979c40efe7
bonding: fix panic when taking bond interface down before removing module

A panic was discovered with bonding when using mode 5 or 6 and trying to
remove the slaves from the bond after the interface was taken down.
When calling 'ifconfig bond0 down' the following happens:

    bond_close()
        bond_alb_deinitialize()
            tlb_deinitialize()
kfree(bond_info->tx_hashtbl)
                bond_info->tx_hashtbl = NULL

Unfortunately if there are still slaves in the bond, when removing the
module the following happens:

    bonding_exit()
        bond_free_all()
            bond_release_all()
                bond_alb_deinit_slave()
                    tlb_clear_slave()
                        tx_hash_table = BOND_ALB_INFO(bond).tx_hashtbl
u32 next_index = tx_hash_table[index].next

As you might guess we panic when trying to access a few entries into the
table that no longer exists.

I experimented with several options (like moving the calls to
tlb_deinitialize somewhere else), but it really makes the most sense to
be part of the bond_close routine.  It also didn't seem logical move
tlb_clear_slave around too much, so the simplest option seems to add a
check in tlb_clear_slave to make sure we haven't already wiped the
tx_hashtbl away before searching for all the non-existent hash-table
entries that used to point to the slave as the output interface.

Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
drivers/net/bonding/bond_alb.c