1. 07 Mar, 2016 13 commits
  2. 05 Mar, 2016 5 commits
  3. 04 Mar, 2016 20 commits
  4. 03 Mar, 2016 2 commits
    • Eric Dumazet's avatar
      net: sched: use pfifo_fast for non real queues · 1f27cde3
      Eric Dumazet authored
      Some devices declare a high number of TX queues, then set a much
      lower real_num_tx_queues
      
      This cause setups using fq_codel, sfq or fq as the default qdisc to consume
      more memory than really needed.
      Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      1f27cde3
    • David Ahern's avatar
      net: ipv6: Fix refcnt on host routes · 799977d9
      David Ahern authored
      Andrew and Ying Huang's test robot both reported usage count problems that
      trace back to the 'keep address on ifdown' patch.
      
      >From Andrew:
      We execute CRIU test on linux-next. On the current linux-next kernel
      they hangs on creating a network namespace.
      
      The kernel log contains many massages like this:
      [ 1036.122108] unregister_netdevice: waiting for lo to become free.
      Usage count = 2
      [ 1046.165156] unregister_netdevice: waiting for lo to become free.
      Usage count = 2
      [ 1056.210287] unregister_netdevice: waiting for lo to become free.
      Usage count = 2
      
      I tried to revert this patch and the bug disappeared.
      
      Here is a set of commands to reproduce this bug:
      
      [root@linux-next-test linux-next]# uname -a
      Linux linux-next-test 4.5.0-rc6-next-20160301+ #3 SMP Wed Mar 2
      17:32:18 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux
      
      [root@linux-next-test ~]# unshare -n
      [root@linux-next-test ~]# ip link set up dev lo
      [root@linux-next-test ~]# ip a
      1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
      group default qlen 1
          link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
          inet 127.0.0.1/8 scope host lo
             valid_lft forever preferred_lft forever
          inet6 ::1/128 scope host
             valid_lft forever preferred_lft forever
      [root@linux-next-test ~]# logout
      [root@linux-next-test ~]# unshare -n
      
       -----
      
      The problem is a change made to RTM_DELADDR case in __ipv6_ifa_notify that
      was added in an early version of the offending patch and is no longer
      needed.
      
      Fixes: f1705ec1 ("net: ipv6: Make address flushing on ifdown optional")
      Cc: Andrey Wagin <avagin@gmail.com>
      Cc: Ying Huang <ying.huang@linux.intel.com>
      Signed-off-by: default avatarDavid Ahern <dsa@cumulusnetworks.com>
      Tested-by: default avatarJeremiah Mahler <jmmahler@gmail.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      799977d9