1. 22 May, 2014 20 commits
    • Florian Fainelli's avatar
      Documentation: devicetree: add old and deprecated 'fixed-link' · 91c1d980
      Florian Fainelli authored
      Update the fixed-link Device Tree binding documentation to contain
      information about the old and deprecated 5-digit 'fixed-link' property.
      Signed-off-by: default avatarFlorian Fainelli <f.fainelli@gmail.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      91c1d980
    • David S. Miller's avatar
      Merge branch 'vlan_features' · a33fa47e
      David S. Miller authored
      Michal Kubecek says:
      
      ====================
      net: device features handling fixes
      
      > I think we need to think more about what exact behavior is desired in
      > mixed csum feature set cases.  Probably whatever csum offload type is
      > the most prevelant should be the one advertised by the master.  It
      > means we will do that software checksum fixup for the oddball slaves,
      > but it seems the best we can do.
      
      I've been thinking about it some more and I believe what I proposed is
      correct: features that are or-ed (ONE_FOR_ALL) should start with 0,
      those which are and-ed (ALL_FOR_ALL) with 1.
      
      But once we do that, we have to fix another problem in vlan module:
      features of a vlan device are mostly computed as bitwise AND of features
      and vlan_features of its underlying device. The problem is checksumming
      features don't really behave like independent flags as their main
      purpose is to be used in can_checksum_protocol() whose logic is that
      HW_CSUM (GEN_CSUM) means "can checksum everything" so that it kind of
      contains IP(V6)_CSUM functionality. Therefore intersection of HW_CSUM
      and IP(V6)_CSUM should result in the latter.
      
      An alternative approach would be to always set IP(V6)_CSUM once HW_CSUM
      (any of GEN_CSUM) is set but as for long time people were taught not to
      do that and we even have a warning for it, switching to the exact
      opposite could cause a lot of problems.
      
      > Also, like Jay, I agree that whatever we do in bonding needs to be
      > done in team too and I'd like the bug fixed at the same time in both
      > places.
      
      Agreed. Added a similar patch for teaming driver.
      ====================
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      a33fa47e
    • Michal Kubeček's avatar
      teaming: fix vlan_features computing · 3625920b
      Michal Kubeček authored
      __team_compute_features() uses netdev_increment_features() to
      combine vlan_features of slaves into vlan_features of the team.
      As netdev_increment_features() only adds most features and we
      start with TEAM_VLAN_FEATURES, we can end up with features none
      of the slaves provided.
      
      Initialize vlan_features only with the flags which are both in
      TEAM_VLAN_FEATURES and NETIF_F_ALL_FOR_ALL. Right now there is
      no such feature so that we actually initialize vlan_features
      with zero but stating it explicitely will make the code more
      future proof.
      Signed-off-by: default avatarMichal Kubecek <mkubecek@suse.cz>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      3625920b
    • Michal Kubeček's avatar
      bonding: fix vlan_features computing · a9b3ace4
      Michal Kubeček authored
      bond_compute_features() uses netdev_increment_features() to
      combine vlan_features of slaves into vlan_features of the bond.
      As netdev_increment_features() only adds most features and we
      start with BOND_VLAN_FEATURES, we can end up with features none
      of the slaves provided.
      
      If there is at least one slave, initialize vlan_features only
      with the flags in NETIF_F_ALL_FOR_ALL. Right now there is none
      in BOND_VLAN_FEATURES but stating it explicitely will make the
      code more future proof.
      Signed-off-by: default avatarMichal Kubecek <mkubecek@suse.cz>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      a9b3ace4
    • Michal Kubeček's avatar
      vlan: more careful checksum features handling · da08143b
      Michal Kubeček authored
      When combining real_dev's features and vlan_features, simple
      bitwise AND is used. This doesn't work well for checksum
      offloading features as if one set has NETIF_F_HW_CSUM and the
      other NETIF_F_IP_CSUM and/or NETIF_F_IPV6_CSUM, we end up with
      no checksum offloading. However, from the logical point of view
      (how can_checksum_protocol() works), NETIF_F_HW_CSUM contains
      the functionality of NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM so
      that the result should be IP/IPV6.
      
      Add helper function netdev_intersect_features() implementing
      this logic and use it in vlan_dev_fix_features().
      Signed-off-by: default avatarMichal Kubecek <mkubecek@suse.cz>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      da08143b
    • Nimrod Andy's avatar
      net: fec: correct the MDIO clock source · 98a6eeb8
      Nimrod Andy authored
      Since imx serials FEC/ENET MDIO clock source is internal ipg clock,
      and "ahb" clock is defined as FEC/ENET bus clock, so the patch just
      correct the fec driver MDIO clock source.
      Signed-off-by: default avatarFugang Duan <B38611@freescale.com>
      Acked-by: default avatarFrank Li <frank.li@freescale.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      98a6eeb8
    • Nimrod Andy's avatar
      net: fec: optimize the clock management to save power · e8fcfcd5
      Nimrod Andy authored
      Add below clock management to save fec power:
      - After probe, disable all clocks incluing ipg, ahb, enet_out, ptp clock.
      - Open ethx interface enable necessary clocks.
        Close ethx interface disable all clocks.
      
      The patch also encapsulates the all enet clocks enable/disable to
      .fec_enet_clk_enable(), which can reduce the repetitional code in
      driver.
      Signed-off-by: default avatarFugang Duan <B38611@freescale.com>
      Acked-by: default avatarFrank Li <Frank.li@freescale.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      e8fcfcd5
    • David S. Miller's avatar
      Merge branch 'sw_tso' · fddb8872
      David S. Miller authored
      Ezequiel Garcia says:
      
      ====================
      net: Introduce a software TSO helper API
      
      Here's a first proposal for a generic software TSO helper API, following
      David's suggestion.
      
      There are at least two drivers that currently implement some form of software
      TSO: Solarflare network driver (drivers/net/ethernet/sfc) and Tilera GX
      network driver (drivers/net/ethernet/tile/tilegx.c).
      
      The rationale behind adding a generic API is to provide a boiler plate with the
      segmentation and other common tasks, making this support easier to add in other
      drivers.
      
      When designing the API, I've considered mainly two design choices:
      
        1. Implement a series of callbacks that each driver would implement
           and the net core code would call to fill in descriptors and egress
           that data.
      
        2. Implement an API for drivers to use in a driver's specific tx_tso
           function. This functions would exhaust a sk_buff payload, and use the
           API as helper for building the headers and segmented data.
      
      I've chosen (2), to avoid function pointers (which was Willy's concern) and
      because it seemed less fragile. Of course, this is argueable.
      
      The API is by no means complete, and lacks some features, however it allows
      to support TSO in mv643xx_eth and mvneta network drivers with some very
      good performance results. I've added this support as an example of the API
      in action.
      
      In particular the following needs some revisiting:
      
        1. IPv6 support is lacking.
      
        2. The required descriptor counting needs some verification. The current
           implementation might be too "sketchy". The tilegx one can be a good
           starting point.
      
        3. The implemenation assumes the hardware can compute the TCP and IP
           checksums for the built headers. However, some controllers may need
           some initial calculation (such as tilegx, for instance).
      
      Despite this, I hope this proposal is good enough to trigger some discussion
      and to check if I'm on the right track. Feedback is much appreciated!
      ====================
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      fddb8872
    • Ezequiel Garcia's avatar
      net: mv643xx_eth: Implement software TSO · 3ae8f4e0
      Ezequiel Garcia authored
      Now that the TSO helper API has been introduced, this commit makes use
      of it to add support for software TSO in this driver.
      
      This feature allows to improve outbound throughput performance significantly.
      Running iperf tests shows a 30% improvement, tested on a Kirkwood Openblocks
      A6 board.
      
      $ ethtool -K eth0 tso off
      $ iperf -c 192.168.0.45 -t 3
      ------------------------------------------------------------
      Client connecting to 192.168.0.45, TCP port 5001
      TCP window size: 43.8 KByte (default)
      ------------------------------------------------------------
      [  3] local 192.168.0.159 port 46389 connected with 192.168.0.45 port 5001
      [ ID] Interval       Transfer     Bandwidth
      [  3]  0.0- 3.0 sec   217 MBytes   607 Mbits/sec
      
      $ ethtool -K eth0 tso on
      $ iperf -c 192.168.0.45 -t 3
      ------------------------------------------------------------
      Client connecting to 192.168.0.45, TCP port 5001
      TCP window size: 43.8 KByte (default)
      ------------------------------------------------------------
      [  3] local 192.168.0.159 port 46390 connected with 192.168.0.45 port 5001
      [ ID] Interval       Transfer     Bandwidth
      [  3]  0.0- 3.0 sec   336 MBytes   938 Mbits/sec
      
      This commit is just an example of the usage of the TSO API, it works fine
      but needs some more work. In particular, the descriptor unmapping path must
      avoid unmapping the TSO headers.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      3ae8f4e0
    • Ezequiel Garcia's avatar
      net: mv643xx_eth: Use dma_map_single() to map the skb fragments · 69ad0dd7
      Ezequiel Garcia authored
      Using dma_map_single() instead of skb_frag_dma_map() allows to unmap
      all the descriptors using dma_unmap_single(). This change allows
      to introduce software TSO in a less intrusive way.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      69ad0dd7
    • Ezequiel Garcia's avatar
      net: mv643xx_eth: Factorize feature setting · 4d48d589
      Ezequiel Garcia authored
      In order to ease the addition of new features, let's factorize the
      feature list.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      4d48d589
    • Ezequiel Garcia's avatar
      net: mv643xx_eth: Avoid setting the initial TCP checksum · 84411f73
      Ezequiel Garcia authored
      As specified in the datasheet, the driver can set the "L4Chk_Mode" flag
      (bit 10) in the Tx descriptor command/status to specify that a frame is not
      IP fragmented and that the controller is in charge of generating the TCP/IP
      checksum. This must be used together with the "GL4chk" flag (bit 17).
      
      These two flags allow to avoid setting the initial TCP checksum in the l4i_chk
      field of the Tx descriptor, which is needed to support software TSO.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      84411f73
    • Ezequiel Garcia's avatar
      net: mv643xx_eth: Factorize initial checksum and command preparation · 0a8fa933
      Ezequiel Garcia authored
      Make the code more readable by moving the initial checksum setup
      and the command/status preparation to its own function.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      0a8fa933
    • Ezequiel Garcia's avatar
      net: mvneta: Implement software TSO · 2adb719d
      Ezequiel Garcia authored
      Now that the TSO helper API has been introduced, this commit makes use
      of it to implement the TSO in this driver.
      
      Using iperf to test and vmstat to check the CPU usage, shows a substantial
      CPU usage drop when TSO is on (~15% vs. ~25%). HTTP-based tests performed
      by Willy Tarreau have shown performance improvements.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      2adb719d
    • Ezequiel Garcia's avatar
      net: mvneta: Clean mvneta_tx() sk_buff handling · e19d2dda
      Ezequiel Garcia authored
      Rework mvneta_tx() so that the code that performs the final handling
      before a sk_buff is transmitted is done only if the numbers of fragments
      processed if positive.
      
      This is preparation work to add the support for software TSO.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      e19d2dda
    • Ezequiel Garcia's avatar
      net: mvneta: Factorize feature setting · 01ef26ca
      Ezequiel Garcia authored
      In order to ease the addition of new features, let's factorize the
      feature list.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      01ef26ca
    • Ezequiel Garcia's avatar
      net: Add a software TSO helper API · e876f208
      Ezequiel Garcia authored
      Although the implementation probably needs a lot of work, this initial API
      allows to implement software TSO in mvneta and mv643xx_eth drivers in a not
      so intrusive way.
      Signed-off-by: default avatarEzequiel Garcia <ezequiel.garcia@free-electrons.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      e876f208
    • David S. Miller's avatar
      Merge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nftables · 8af750d7
      David S. Miller authored
      Pablo Neira Ayuso says:
      
      ====================
      Netfilter/nftables updates for net-next
      
      The following patchset contains Netfilter/nftables updates for net-next,
      most relevantly they are:
      
      1) Add set element update notification via netlink, from Arturo Borrero.
      
      2) Put all object updates in one single message batch that is sent to
         kernel-space. Before this patch only rules where included in the batch.
         This series also introduces the generic transaction infrastructure so
         updates to all objects (tables, chains, rules and sets) are applied in
         an all-or-nothing fashion, these series from me.
      
      3) Defer release of objects via call_rcu to reduce the time required to
         commit changes. The assumption is that all objects are destroyed in
         reverse order to ensure that dependencies betweem them are fulfilled
         (ie. rules and sets are destroyed first, then chains, and finally
         tables).
      
      4) Allow to match by bridge port name, from Tomasz Bursztyka. This series
         include two patches to prepare this new feature.
      
      5) Implement the proper set selection based on the characteristics of the
         data. The new infrastructure also allows you to specify your preferences
         in terms of memory and computational complexity so the underlying set
         type is also selected according to your needs, from Patrick McHardy.
      
      6) Several cleanup patches for nft expressions, including one minor possible
         compilation breakage due to missing mark support, also from Patrick.
      ====================
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      8af750d7
    • David S. Miller's avatar
      Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-next · 758bd61a
      David S. Miller authored
      Jeff Kirsher says:
      
      ====================
      Intel Wired LAN Driver Updates
      
      This series contains updates to i40e and i40evf.
      
      Shannon makes minor changes to the AdminQ interface to bring it up to
      date.  Removes the hard coding of stats struct size in ethtool, in prep
      for adding data fields which are configuration dependent.
      
      Catherine removes some unused and unneeded PCI bus defines.
      
      Jesse fixes the copyright headers and finishes up the removal of the PTP
      Tx work functionality which allows us to rely on the Tx timesync interrupt.
      
      Mitch provides a number of fixes and cleanups for i40e/i40evf based on
      suggestions from Ben Hutchings.  First is to use a macro parameter for
      ethtool stats instead of just assuming that a valid netdev variable
      exists.  Second is not to tell ethtool that the VF can do 10GbaseT, when
      it really has no idea what its link speed is, so set the supported value
      to 0 instead.  Make the ethtool_ops structure constant since it is
      extremely unlikely to change at runtime.  Ethtool consistently reports
      0 values for our ITR settings because we never actually use them, so
      fix this by setting the default values to the specified default values.
      
      Greg avoids a compile error by wrapping the call to i40e_alloc_vfs() in
      CONFIG_PCI_IOV because the function itself is wrapped in the same
      conditional compile block.
      
      Alexander Gordeev updates the driver to use the new pci_enable_msi_range()
      and pci_enable_msix_range() or pci_enable_msi_exact() and
      pci_enable_msix_exact().
      
      Jean Sacren provides a fix where the wrong error code was being passed to
      i40e_open().
      ====================
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      758bd61a
    • Neal Cardwell's avatar
      tcp: make cwnd-limited checks measurement-based, and gentler · ca8a2263
      Neal Cardwell authored
      Experience with the recent e114a710 ("tcp: fix cwnd limited
      checking to improve congestion control") has shown that there are
      common cases where that commit can cause cwnd to be much larger than
      necessary. This leads to TSO autosizing cooking skbs that are too
      large, among other things.
      
      The main problems seemed to be:
      
      (1) That commit attempted to predict the future behavior of the
      connection by looking at the write queue (if TSO or TSQ limit
      sending). That prediction sometimes overestimated future outstanding
      packets.
      
      (2) That commit always allowed cwnd to grow to twice the number of
      outstanding packets (even in congestion avoidance, where this is not
      needed).
      
      This commit improves both of these, by:
      
      (1) Switching to a measurement-based approach where we explicitly
      track the largest number of packets in flight during the past window
      ("max_packets_out"), and remember whether we were cwnd-limited at the
      moment we finished sending that flight.
      
      (2) Only allowing cwnd to grow to twice the number of outstanding
      packets ("max_packets_out") in slow start. In congestion avoidance
      mode we now only allow cwnd to grow if it was fully utilized.
      Signed-off-by: default avatarNeal Cardwell <ncardwell@google.com>
      Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      ca8a2263
  2. 21 May, 2014 20 commits