1. 16 Mar, 2009 31 commits
  2. 15 Mar, 2009 9 commits
    • Ilpo Järvinen's avatar
      tcp: make sure xmit goal size never becomes zero · afece1c6
      Ilpo Järvinen authored
      
      
      It's not too likely to happen, would basically require crafted
      packets (must hit the max guard in tcp_bound_to_half_wnd()).
      It seems that nothing that bad would happen as there's tcp_mems
      and congestion window that prevent runaway at some point from
      hurting all too much (I'm not that sure what all those zero
      sized segments we would generate do though in write queue).
      Preventing it regardless is certainly the best way to go.
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Cc: Evgeniy Polyakov <zbr@ioremap.net>
      Cc: Ingo Molnar <mingo@elte.hu>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      afece1c6
    • Ilpo Järvinen's avatar
      tcp: cache result of earlier divides when mss-aligning things · 2a3a041c
      Ilpo Järvinen authored
      
      
      The results is very unlikely change every so often so we
      hardly need to divide again after doing that once for a
      connection. Yet, if divide still becomes necessary we
      detect that and do the right thing and again settle for
      non-divide state. Takes the u16 space which was previously
      taken by the plain xmit_size_goal.
      
      This should take care part of the tso vs non-tso difference
      we found earlier.
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      2a3a041c
    • Ilpo Järvinen's avatar
      tcp: simplify tcp_current_mss · 0c54b85f
      Ilpo Järvinen authored
      
      
      There's very little need for most of the callsites to get
      tp->xmit_goal_size updated. That will cost us divide as is,
      so slice the function in two. Also, the only users of the
      tp->xmit_goal_size are directly behind tcp_current_mss(),
      so there's no need to store that variable into tcp_sock
      at all! The drop of xmit_goal_size currently leaves 16-bit
      hole and some reorganization would again be necessary to
      change that (but I'm aiming to fill that hole with u16
      xmit_goal_size_segs to cache the results of the remaining
      divide to get that tso on regression).
      
      Bring xmit_goal_size parts into tcp.c
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Cc: Evgeniy Polyakov <zbr@ioremap.net>
      Cc: Ingo Molnar <mingo@elte.hu>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      0c54b85f
    • Ilpo Järvinen's avatar
      tcp: don't check mtu probe completion in the loop · 72211e90
      Ilpo Järvinen authored
      
      
      It seems that no variables clash such that we couldn't do
      the check just once later on. Therefore move it.
      
      Also kill dead obvious comment, dead argument and add
      unlikely since this mtu probe does not happen too often.
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      72211e90
    • Ilpo Järvinen's avatar
      tcp: consolidate paws check · c887e6d2
      Ilpo Järvinen authored
      
      
      Wow, it was quite tricky to merge that stream of negations
      but I think I finally got it right:
      
      check & replace_ts_recent:
      (s32)(rcv_tsval - ts_recent) >= 0                  => 0
      (s32)(ts_recent - rcv_tsval) <= 0                  => 0
      
      discard:
      (s32)(ts_recent - rcv_tsval)  > TCP_PAWS_WINDOW    => 1
      (s32)(ts_recent - rcv_tsval) <= TCP_PAWS_WINDOW    => 0
      
      I toggled the return values of tcp_paws_check around since
      the old encoding added yet-another negation making tracking
      of truth-values really complicated.
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      c887e6d2
    • Ilpo Järvinen's avatar
      tcp: kill dead end_seq variable in clean_rtx_queue · c43d558a
      Ilpo Järvinen authored
      
      
      I've already forgotten what for this was necessary, anyway
      it's no longer used (if it ever was).
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      c43d558a
    • Ilpo Järvinen's avatar
      tcp: remove pointless .dsack/.num_sacks code · 5861f8e5
      Ilpo Järvinen authored
      
      
      In the pure assignment case, the earlier zeroing is
      still in effect.
      
      David S. Miller raised concerns if the ifs are there to avoid
      dirtying cachelines. I came to these conclusions:
      
      > We'll be dirty it anyway (now that I check), the first "real" statement
      > in tcp_rcv_established is:
      >
      >       tp->rx_opt.saw_tstamp = 0;
      >
      > ...that'll land on the same dword. :-/
      >
      > I suppose the blocks are there just because they had more complexity
      > inside when they had to calculate the eff_sacks too (maybe it would
      > have been better to just remove them in that drop-patch so you would
      > have had less head-ache :-)).
      Signed-off-by: default avatarIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      5861f8e5
    • Jarek Poplawski's avatar
      pkt_sched: Change misleading code in class delete. · 7cd0a638
      Jarek Poplawski authored
      While looking for a possible reason of bugzilla report on HTB oops:
      http://bugzilla.kernel.org/show_bug.cgi?id=12858
      
      
      I found the code in htb_delete calling htb_destroy_class on zero
      refcount is very misleading: it can suggest this is a common path, and
      destroy is called under sch_tree_lock. Actually, this can never happen
      like this because before deletion cops->get() is done, and after
      delete a class is still used by tclass_notify. The class destroy is
      always called from cops->put(), so without sch_tree_lock.
      
      This doesn't mean much now (since 2.6.27) because all vulnerable calls
      were moved from htb_destroy_class to htb_delete, but there was a bug
      in older kernels. The same change is done for other classful scheds,
      which, it seems, didn't have similar locking problems here.
      Reported-by: default avatarm0sia <m0sia@m0sia.ru>
      Signed-off-by: default avatarJarek Poplawski <jarkao2@gmail.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      7cd0a638
    • Eric Dumazet's avatar
      net: reorder fields of struct socket · 8bdd663a
      Eric Dumazet authored
      
      
      On x86_64, its rather unfortunate that "wait_queue_head_t wait"
      field of "struct socket" spans two cache lines (assuming a 64
      bytes cache line in current cpus)
      
      offsetof(struct socket, wait)=0x30
      sizeof(wait_queue_head_t)=0x18
      
      This might explain why Kenny Chang noticed that his multicast workload
      was performing bad with 64 bit kernels, since more cache lines ping pongs
      were involved.
      
      This litle patch moves "wait" field next "fasync_list" so that both
      fields share a single cache line, to speedup sock_def_readable()
      Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      8bdd663a