1. 25 Nov, 2009 1 commit
  2. 18 Oct, 2009 1 commit
    • Eric Dumazet's avatar
      inet: rename some inet_sock fields · c720c7e8
      Eric Dumazet authored
      In order to have better cache layouts of struct sock (separate zones
      for rx/tx paths), we need this preliminary patch.
      
      Goal is to transfert fields used at lookup time in the first
      read-mostly cache line (inside struct sock_common) and move sk_refcnt
      to a separate cache line (only written by rx path)
      
      This patch adds inet_ prefix to daddr, rcv_saddr, dport, num, saddr,
      sport and id fields. This allows a future patch to define these
      fields as macros, like sk_refcnt, without name clashes.
      Signed-off-by: default avatarEric Dumazet <eric.dumazet@gmail.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      c720c7e8
  3. 13 Oct, 2009 1 commit
  4. 01 Feb, 2009 1 commit
  5. 21 Jan, 2009 1 commit
    • Evgeniy Polyakov's avatar
      inet: Allowing more than 64k connections and heavily optimize bind(0) time. · a9d8f911
      Evgeniy Polyakov authored
      With simple extension to the binding mechanism, which allows to bind more
      than 64k sockets (or smaller amount, depending on sysctl parameters),
      we have to traverse the whole bind hash table to find out empty bucket.
      And while it is not a problem for example for 32k connections, bind()
      completion time grows exponentially (since after each successful binding
      we have to traverse one bucket more to find empty one) even if we start
      each time from random offset inside the hash table.
      
      So, when hash table is full, and we want to add another socket, we have
      to traverse the whole table no matter what, so effectivelly this will be
      the worst case performance and it will be constant.
      
      Attached picture shows bind() time depending on number of already bound
      sockets.
      
      Green area corresponds to the usual binding to zero port process, which
      turns on kernel port selection as described above. Red area is the bind
      process, when number of reuse-bound sockets is not limited by 64k (or
      sysctl parameters). The same exponential growth (hidden by the green
      area) before number of ports reaches sysctl limit.
      
      At this time bind hash table has exactly one reuse-enbaled socket in a
      bucket, but it is possible that they have different addresses. Actually
      kernel selects the first port to try randomly, so at the beginning bind
      will take roughly constant time, but with time number of port to check
      after random start will increase. And that will have exponential growth,
      but because of above random selection, not every next port selection
      will necessary take longer time than previous. So we have to consider
      the area below in the graph (if you could zoom it, you could find, that
      there are many different times placed there), so area can hide another.
      
      Blue area corresponds to the port selection optimization.
      
      This is rather simple design approach: hashtable now maintains (unprecise
      and racely updated) number of currently bound sockets, and when number
      of such sockets becomes greater than predefined value (I use maximum
      port range defined by sysctls), we stop traversing the whole bind hash
      table and just stop at first matching bucket after random start. Above
      limit roughly corresponds to the case, when bind hash table is full and
      we turned on mechanism of allowing to bind more reuse-enabled sockets,
      so it does not change behaviour of other sockets.
      Signed-off-by: default avatarEvgeniy Polyakov <zbr@ioremap.net>
      Tested-by: default avatarDenys Fedoryschenko <denys@visp.net.lb>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      a9d8f911
  6. 24 Nov, 2008 1 commit
  7. 23 Nov, 2008 1 commit
    • Eric Dumazet's avatar
      net: Convert TCP/DCCP listening hash tables to use RCU · c25eb3bf
      Eric Dumazet authored
      This is the last step to be able to perform full RCU lookups
      in __inet_lookup() : After established/timewait tables, we
      add RCU lookups to listening hash table.
      
      The only trick here is that a socket of a given type (TCP ipv4,
      TCP ipv6, ...) can now flight between two different tables
      (established and listening) during a RCU grace period, so we
      must use different 'nulls' end-of-chain values for two tables.
      
      We define a large value :
      
      #define LISTENING_NULLS_BASE (1U << 29)
      
      So that slots in listening table are guaranteed to have different
      end-of-chain values than slots in established table. A reader can
      still detect it finished its lookup in the right chain.
      Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      c25eb3bf
  8. 20 Nov, 2008 2 commits
  9. 16 Nov, 2008 1 commit
    • Eric Dumazet's avatar
      net: Convert TCP & DCCP hash tables to use RCU / hlist_nulls · 3ab5aee7
      Eric Dumazet authored
      RCU was added to UDP lookups, using a fast infrastructure :
      - sockets kmem_cache use SLAB_DESTROY_BY_RCU and dont pay the
        price of call_rcu() at freeing time.
      - hlist_nulls permits to use few memory barriers.
      
      This patch uses same infrastructure for TCP/DCCP established
      and timewait sockets.
      
      Thanks to SLAB_DESTROY_BY_RCU, no slowdown for applications
      using short lived TCP connections. A followup patch, converting
      rwlocks to spinlocks will even speedup this case.
      
      __inet_lookup_established() is pretty fast now we dont have to
      dirty a contended cache line (read_lock/read_unlock)
      
      Only established and timewait hashtable are converted to RCU
      (bind table and listen table are still using traditional locking)
      Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      3ab5aee7
  10. 12 Nov, 2008 1 commit
  11. 25 Jul, 2008 1 commit
  12. 16 Jul, 2008 1 commit
  13. 16 Jun, 2008 3 commits
  14. 18 Apr, 2008 1 commit
  15. 16 Apr, 2008 1 commit
  16. 31 Mar, 2008 1 commit
  17. 25 Mar, 2008 2 commits
  18. 22 Mar, 2008 1 commit
    • Pavel Emelyanov's avatar
      [SOCK]: Add udp_hash member to struct proto. · 39d8cda7
      Pavel Emelyanov authored
      Inspired by the commit ab1e0a13 ([SOCK] proto: Add hashinfo member to 
      struct proto) from Arnaldo, I made similar thing for UDP/-Lite IPv4 
      and -v6 protocols.
      
      The result is not that exciting, but it removes some levels of
      indirection in udpxxx_get_port and saves some space in code and text.
      
      The first step is to union existing hashinfo and new udp_hash on the
      struct proto and give a name to this union, since future initialization 
      of tcpxxx_prot, dccp_vx_protinfo and udpxxx_protinfo will cause gcc 
      warning about inability to initialize anonymous member this way.
      Signed-off-by: default avatarPavel Emelyanov <xemul@openvz.org>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      39d8cda7
  19. 13 Feb, 2008 2 commits
  20. 05 Feb, 2008 1 commit
  21. 03 Feb, 2008 1 commit
    • Arnaldo Carvalho de Melo's avatar
      [SOCK] proto: Add hashinfo member to struct proto · ab1e0a13
      Arnaldo Carvalho de Melo authored
      This way we can remove TCP and DCCP specific versions of
      
      sk->sk_prot->get_port: both v4 and v6 use inet_csk_get_port
      sk->sk_prot->hash:     inet_hash is directly used, only v6 need
                             a specific version to deal with mapped sockets
      sk->sk_prot->unhash:   both v4 and v6 use inet_hash directly
      
      struct inet_connection_sock_af_ops also gets a new member, bind_conflict, so
      that inet_csk_get_port can find the per family routine.
      
      Now only the lookup routines receive as a parameter a struct inet_hashtable.
      
      With this we further reuse code, reducing the difference among INET transport
      protocols.
      
      Eventually work has to be done on UDP and SCTP to make them share this
      infrastructure and get as a bonus inet_diag interfaces so that iproute can be
      used with these protocols.
      
      net-2.6/net/ipv4/inet_hashtables.c:
        struct proto			     |   +8
        struct inet_connection_sock_af_ops |   +8
       2 structs changed
        __inet_hash_nolisten               |  +18
        __inet_hash                        | -210
        inet_put_port                      |   +8
        inet_bind_bucket_create            |   +1
        __inet_hash_connect                |   -8
       5 functions changed, 27 bytes added, 218 bytes removed, diff: -191
      
      net-2.6/net/core/sock.c:
        proto_seq_show                     |   +3
       1 function changed, 3 bytes added, diff: +3
      
      net-2.6/net/ipv4/inet_connection_sock.c:
        inet_csk_get_port                  |  +15
       1 function changed, 15 bytes added, diff: +15
      
      net-2.6/net/ipv4/tcp.c:
        tcp_set_state                      |   -7
       1 function changed, 7 bytes removed, diff: -7
      
      net-2.6/net/ipv4/tcp_ipv4.c:
        tcp_v4_get_port                    |  -31
        tcp_v4_hash                        |  -48
        tcp_v4_destroy_sock                |   -7
        tcp_v4_syn_recv_sock               |   -2
        tcp_unhash                         | -179
       5 functions changed, 267 bytes removed, diff: -267
      
      net-2.6/net/ipv6/inet6_hashtables.c:
        __inet6_hash |   +8
       1 function changed, 8 bytes added, diff: +8
      
      net-2.6/net/ipv4/inet_hashtables.c:
        inet_unhash                        | +190
        inet_hash                          | +242
       2 functions changed, 432 bytes added, diff: +432
      
      vmlinux:
       16 functions changed, 485 bytes added, 492 bytes removed, diff: -7
      
      /home/acme/git/net-2.6/net/ipv6/tcp_ipv6.c:
        tcp_v6_get_port                    |  -31
        tcp_v6_hash                        |   -7
        tcp_v6_syn_recv_sock               |   -9
       3 functions changed, 47 bytes removed, diff: -47
      
      /home/acme/git/net-2.6/net/dccp/proto.c:
        dccp_destroy_sock                  |   -7
        dccp_unhash                        | -179
        dccp_hash                          |  -49
        dccp_set_state                     |   -7
        dccp_done                          |   +1
       5 functions changed, 1 bytes added, 242 bytes removed, diff: -241
      
      /home/acme/git/net-2.6/net/dccp/ipv4.c:
        dccp_v4_get_port                   |  -31
        dccp_v4_request_recv_sock          |   -2
       2 functions changed, 33 bytes removed, diff: -33
      
      /home/acme/git/net-2.6/net/dccp/ipv6.c:
        dccp_v6_get_port                   |  -31
        dccp_v6_hash                       |   -7
        dccp_v6_request_recv_sock          |   +5
       3 functions changed, 5 bytes added, 38 bytes removed, diff: -33
      Signed-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      ab1e0a13
  22. 31 Jan, 2008 3 commits
  23. 28 Jan, 2008 4 commits
  24. 07 Nov, 2007 1 commit
    • Eric Dumazet's avatar
      [INET]: Remove per bucket rwlock in tcp/dccp ehash table. · 230140cf
      Eric Dumazet authored
      As done two years ago on IP route cache table (commit
      22c047cc) , we can avoid using one
      lock per hash bucket for the huge TCP/DCCP hash tables.
      
      On a typical x86_64 platform, this saves about 2MB or 4MB of ram, for
      litle performance differences. (we hit a different cache line for the
      rwlock, but then the bucket cache line have a better sharing factor
      among cpus, since we dirty it less often). For netstat or ss commands
      that want a full scan of hash table, we perform fewer memory accesses.
      
      Using a 'small' table of hashed rwlocks should be more than enough to
      provide correct SMP concurrency between different buckets, without
      using too much memory. Sizing of this table depends on
      num_possible_cpus() and various CONFIG settings.
      
      This patch provides some locking abstraction that may ease a future
      work using a different model for TCP/DCCP table.
      Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
      Acked-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      230140cf
  25. 18 Oct, 2007 1 commit
  26. 10 Oct, 2007 1 commit
  27. 11 Feb, 2007 1 commit
  28. 08 Feb, 2007 1 commit
    • Eric Dumazet's avatar
      [NET]: change layout of ehash table · dbca9b27
      Eric Dumazet authored
      ehash table layout is currently this one :
      
      First half of this table is used by sockets not in TIME_WAIT state
      Second half of it is used by sockets in TIME_WAIT state.
      
      This is non optimal because of for a given hash or socket, the two chain heads 
      are located in separate cache lines.
      Moreover the locks of the second half are never used.
      
      If instead of this halving, we use two list heads in inet_ehash_bucket instead 
      of only one, we probably can avoid one cache miss, and reduce ram usage, 
      particularly if sizeof(rwlock_t) is big (various CONFIG_DEBUG_SPINLOCK, 
      CONFIG_DEBUG_LOCK_ALLOC settings). So we still halves the table but we keep 
      together related chains to speedup lookups and socket state change.
      
      In this patch I did not try to align struct inet_ehash_bucket, but a future 
      patch could try to make this structure have a convenient size (a power of two 
      or a multiple of L1_CACHE_SIZE).
      I guess rwlock will just vanish as soon as RCU is plugged into ehash :) , so 
      maybe we dont need to scratch our heads to align the bucket...
      
      Note : In case struct inet_ehash_bucket is not a power of two, we could 
      probably change alloc_large_system_hash() (in case it use __get_free_pages()) 
      to free the unused space. It currently allocates a big zone, but the last 
      quarter of it could be freed. Again, this should be a temporary 'problem'.
      
      Patch tested on ipv4 tcp only, but should be OK for IPV6 and DCCP.
      Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      dbca9b27
  29. 07 Dec, 2006 2 commits