patch bonding: don't use stale speed and duplex information解释

来源:互联网 发布:mac 文件夹隐藏与显示 编辑:程序博客网 时间:2024/06/10 03:05
commit 266b495f11d6706018f66250cb02a788ff2490d7
Author: Jay Vosburgh <jay.vosburgh@canonical.com>
Date:   Mon Feb 8 12:10:02 2016 -0800

    bonding: don't use stale speed and duplex information
    
    There is presently a race condition between the bonding periodic
    link monitor and the updating of a slave's speed and duplex.  The former
    occurs on a periodic basis, and the latter in response to a driver's
    calling of netif_carrier_on.
    
        It is possible for the periodic monitor to run between the
    driver call of netif_carrier_on and the receipt of the NETDEV_CHANGE
    event that causes bonding to update the slave's speed and duplex.  This
    manifests most notably as a report that a slave is up and "0 Mbps full
    duplex" after enslavement, but in principle could report an incorrect
    speed and duplex after any link up event if the device comes up with a
    different speed or duplex.  This affects the 802.3ad aggregator
    selection, as the speed and duplex are selection criteria.
    
        This is fixed by updating the speed and duplex in the periodic
    monitor, prior to using that information.
    
        This was done historically in bonding, but the call to
    bond_update_speed_duplex was removed in commit 876254ae2758 ("bonding:
    don't call update_speed_duplex() under spinlocks"), as it might sleep
    under lock.  Later, the locking was changed to only hold RTNL, and so
    after commit 876254ae2758 ("bonding: don't call update_speed_duplex()
    under spinlocks") this call is again safe.
    
    Tested-by: "Tantilov, Emil S" <emil.s.tantilov@intel.com>
    Cc: Veaceslav Falico <vfalico@gmail.com>
    Cc: dingtianhong <dingtianhong@huawei.com>
    Fixes: 876254ae2758 ("bonding: don't call update_speed_duplex() under spinlocks")
    Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>
    Acked-by: Ding Tianhong <dingtianhong@huawei.com>
    Signed-off-by: David S. Miller <davem@davemloft.net>


主要原因是添加slave网卡时,在up网卡和网卡协商speed之间有一段时间。当时间足够大时,slave的结构体中的speed无法获得速度。
speed的值为-1. 在802.3ad模式时,就会出错“0 Mbps full duplex”。
0 0