1. 26 Apr, 2006 1 commit
  2. 20 Apr, 2006 1 commit
    • David S. Miller's avatar
      [NET]: Add skb->truesize assertion checking. · dc6de336
      David S. Miller authored
      
      
      Add some sanity checking.  truesize should be at least sizeof(struct
      sk_buff) plus the current packet length.  If not, then truesize is
      seriously mangled and deserves a kernel log message.
      
      Currently we'll do the check for release of stream socket buffers.
      
      But we can add checks to more spots over time.
      
      Incorporating ideas from Herbert Xu.
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      dc6de336
  3. 28 Mar, 2006 1 commit
    • Denis Vlasenko's avatar
      [NET]: deinline 200+ byte inlines in sock.h · f0088a50
      Denis Vlasenko authored
      
      
      Sizes in bytes (allyesconfig, i386) and files where those inlines
      are used:
      
      238 sock_queue_rcv_skb 2.6.16/net/x25/x25_in.o
      238 sock_queue_rcv_skb 2.6.16/net/rose/rose_in.o
      238 sock_queue_rcv_skb 2.6.16/net/packet/af_packet.o
      238 sock_queue_rcv_skb 2.6.16/net/netrom/nr_in.o
      238 sock_queue_rcv_skb 2.6.16/net/llc/llc_sap.o
      238 sock_queue_rcv_skb 2.6.16/net/llc/llc_conn.o
      238 sock_queue_rcv_skb 2.6.16/net/irda/af_irda.o
      238 sock_queue_rcv_skb 2.6.16/net/ipx/af_ipx.o
      238 sock_queue_rcv_skb 2.6.16/net/ipv6/udp.o
      238 sock_queue_rcv_skb 2.6.16/net/ipv6/raw.o
      238 sock_queue_rcv_skb 2.6.16/net/ipv4/udp.o
      238 sock_queue_rcv_skb 2.6.16/net/ipv4/raw.o
      238 sock_queue_rcv_skb 2.6.16/net/ipv4/ipmr.o
      238 sock_queue_rcv_skb 2.6.16/net/econet/econet.o
      238 sock_queue_rcv_skb 2.6.16/net/econet/af_econet.o
      238 sock_queue_rcv_skb 2.6.16/net/bluetooth/sco.o
      238 sock_queue_rcv_skb 2.6.16/net/bluetooth/l2cap.o
      238 sock_queue_rcv_skb 2.6.16/net/bluetooth/hci_sock.o
      238 sock_queue_rcv_skb 2.6.16/net/ax25/ax25_in.o
      238 sock_queue_rcv_skb 2.6.16/net/ax25/af_ax25.o
      238 sock_queue_rcv_skb 2.6.16/net/appletalk/ddp.o
      238 sock_queue_rcv_skb 2.6.16/drivers/net/pppoe.o
      
      276 sk_receive_skb 2.6.16/net/decnet/dn_nsp_in.o
      276 sk_receive_skb 2.6.16/net/dccp/ipv6.o
      276 sk_receive_skb 2.6.16/net/dccp/ipv4.o
      276 sk_receive_skb 2.6.16/net/dccp/dccp_ipv6.o
      276 sk_receive_skb 2.6.16/drivers/net/pppoe.o
      
      209 sk_dst_check 2.6.16/net/ipv6/ip6_output.o
      209 sk_dst_check 2.6.16/net/ipv4/udp.o
      209 sk_dst_check 2.6.16/net/decnet/dn_nsp_out.o
      
      Large inlines with multiple callers:
      Size  Uses Wasted Name and definition
      ===== ==== ====== ================================================
        238   21   4360 sock_queue_rcv_skb    include/net/sock.h
        109   10    801 sock_recv_timestamp   include/net/sock.h
        276    4    768 sk_receive_skb        include/net/sock.h
         94    8    518 __sk_dst_check        include/net/sock.h
        209    3    378 sk_dst_check  include/net/sock.h
        131    4    333 sk_setup_caps include/net/sock.h
        152    2    132 sk_stream_alloc_pskb  include/net/sock.h
        125    2    105 sk_stream_writequeue_purge    include/net/sock.h
      Signed-off-by: default avatarAndrew Morton <akpm@osdl.org>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      f0088a50
  4. 24 Mar, 2006 1 commit
  5. 20 Mar, 2006 1 commit
  6. 17 Mar, 2006 1 commit
  7. 02 Feb, 2006 1 commit
  8. 06 Jan, 2006 1 commit
  9. 03 Jan, 2006 3 commits
  10. 10 Nov, 2005 1 commit
  11. 08 Nov, 2005 1 commit
  12. 28 Oct, 2005 1 commit
  13. 08 Oct, 2005 1 commit
  14. 03 Oct, 2005 1 commit
    • Eric Dumazet's avatar
      [INET]: speedup inet (tcp/dccp) lookups · 81c3d547
      Eric Dumazet authored
      
      
      Arnaldo and I agreed it could be applied now, because I have other
      pending patches depending on this one (Thank you Arnaldo)
      
      (The other important patch moves skc_refcnt in a separate cache line,
      so that the SMP/NUMA performance doesnt suffer from cache line ping pongs)
      
      1) First some performance data :
      --------------------------------
      
      tcp_v4_rcv() wastes a *lot* of time in __inet_lookup_established()
      
      The most time critical code is :
      
      sk_for_each(sk, node, &head->chain) {
           if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
               goto hit; /* You sunk my battleship! */
      }
      
      The sk_for_each() does use prefetch() hints but only the begining of
      "struct sock" is prefetched.
      
      As INET_MATCH first comparison uses inet_sk(__sk)->daddr, wich is far
      away from the begining of "struct sock", it has to bring into CPU
      cache cold cache line. Each iteration has to use at least 2 cache
      lines.
      
      This can be problematic if some chains are very long.
      
      2) The goal
      -----------
      
      The idea I had is to change things so that INET_MATCH() may return
      FALSE in 99% of cases only using the data already in the CPU cache,
      using one cache line per iteration.
      
      3) Description of the patch
      ---------------------------
      
      Adds a new 'unsigned int skc_hash' field in 'struct sock_common',
      filling a 32 bits hole on 64 bits platform.
      
      struct sock_common {
      	unsigned short		skc_family;
      	volatile unsigned char	skc_state;
      	unsigned char		skc_reuse;
      	int			skc_bound_dev_if;
      	struct hlist_node	skc_node;
      	struct hlist_node	skc_bind_node;
      	atomic_t		skc_refcnt;
      +	unsigned int		skc_hash;
      	struct proto		*skc_prot;
      };
      
      Store in this 32 bits field the full hash, not masked by (ehash_size -
      1) Using this full hash as the first comparison done in INET_MATCH
      permits us immediatly skip the element without touching a second cache
      line in case of a miss.
      
      Suppress the sk_hashent/tw_hashent fields since skc_hash (aliased to
      sk_hash and tw_hash) already contains the slot number if we mask with
      (ehash_size - 1)
      
      File include/net/inet_hashtables.h
      
      64 bits platforms :
      #define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
           (((__sk)->sk_hash == (__hash))
           ((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie))   &&  \
           ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   &&  \
           (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
      
      32bits platforms:
      #define TCP_IPV4_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
           (((__sk)->sk_hash == (__hash))                 &&  \
           (inet_sk(__sk)->daddr          == (__saddr))   &&  \
           (inet_sk(__sk)->rcv_saddr      == (__daddr))   &&  \
           (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
      
      
      - Adds a prefetch(head->chain.first) in 
      __inet_lookup_established()/__tcp_v4_check_established() and 
      __inet6_lookup_established()/__tcp_v6_check_established() and 
      __dccp_v4_check_established() to bring into cache the first element of the 
      list, before the {read|write}_lock(&head->lock);
      Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
      Acked-by: default avatarArnaldo Carvalho de Melo <acme@ghostprotocols.net>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      81c3d547
  15. 05 Sep, 2005 1 commit
  16. 01 Sep, 2005 2 commits
  17. 29 Aug, 2005 14 commits
  18. 23 Aug, 2005 1 commit
    • Ralf Baechle's avatar
      [NET]: Fix socket bitop damage · 53b924b3
      Ralf Baechle authored
      
      
      The socket flag cleanups that went into 2.6.12-rc1 are basically oring
      the flags of an old socket into the socket just being created.
      Unfortunately that one was just initialized by sock_init_data(), so already
      has SOCK_ZAPPED set.  As the result zapped sockets are created and all
      incoming connection will fail due to this bug which again was carefully
      replicated to at least AX.25, NET/ROM or ROSE.
      
      In order to keep the abstraction alive I've introduced sock_copy_flags()
      to copy the socket flags from one sockets to another and used that
      instead of the bitwise copy thing.  Anyway, the idea here has probably
      been to copy all flags, so sock_copy_flags() should be the right thing.
      With this the ham radio protocols are usable again, so I hope this will
      make it into 2.6.13.
      Signed-off-by: default avatarRalf Baechle DL5RB <ralf@linux-mips.org>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      53b924b3
  19. 08 Jul, 2005 1 commit
  20. 05 Jul, 2005 1 commit
    • David S. Miller's avatar
      [TCP]: Simplify SKB data portion allocation with NETIF_F_SG. · c65f7f00
      David S. Miller authored
      
      
      The ideal and most optimal layout for an SKB when doing
      scatter-gather is to put all the headers at skb->data, and
      all the user data in the page array.
      
      This makes SKB splitting and combining extremely simple,
      especially before a packet goes onto the wire the first
      time.
      
      So, when sk_stream_alloc_pskb() is given a zero size, make
      sure there is no skb_tailroom().  This is achieved by applying
      SKB_DATA_ALIGN() to the header length used here.
      
      Next, make select_size() in TCP output segmentation use a
      length of zero when NETIF_F_SG is true on the outgoing
      interface.
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      c65f7f00
  21. 18 Jun, 2005 2 commits
    • Arnaldo Carvalho de Melo's avatar
      [NET] Rename open_request to request_sock · 60236fdd
      Arnaldo Carvalho de Melo authored
      
      
      Ok, this one just renames some stuff to have a better namespace and to
      dissassociate it from TCP:
      
      struct open_request  -> struct request_sock
      tcp_openreq_alloc    -> reqsk_alloc
      tcp_openreq_free     -> reqsk_free
      tcp_openreq_fastfree -> __reqsk_free
      
      With this most of the infrastructure closely resembles a struct
      sock methods subset.
      Signed-off-by: default avatarArnaldo Carvalho de Melo <acme@ghostprotocols.net>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      60236fdd
    • Arnaldo Carvalho de Melo's avatar
      [NET] Generalise TCP's struct open_request minisock infrastructure · 2e6599cb
      Arnaldo Carvalho de Melo authored
      
      
      Kept this first changeset minimal, without changing existing names to
      ease peer review.
      
      Basicaly tcp_openreq_alloc now receives the or_calltable, that in turn
      has two new members:
      
      ->slab, that replaces tcp_openreq_cachep
      ->obj_size, to inform the size of the openreq descendant for
        a specific protocol
      
      The protocol specific fields in struct open_request were moved to a
      class hierarchy, with the things that are common to all connection
      oriented PF_INET protocols in struct inet_request_sock, the TCP ones
      in tcp_request_sock, that is an inet_request_sock, that is an
      open_request.
      
      I.e. this uses the same approach used for the struct sock class
      hierarchy, with sk_prot indicating if the protocol wants to use the
      open_request infrastructure by filling in sk_prot->rsk_prot with an
      or_calltable.
      
      Results? Performance is improved and TCP v4 now uses only 64 bytes per
      open request minisock, down from 96 without this patch :-)
      
      Next changeset will rename some of the structs, fields and functions
      mentioned above, struct or_calltable is way unclear, better name it
      struct request_sock_ops, s/struct open_request/struct request_sock/g,
      etc.
      Signed-off-by: default avatarArnaldo Carvalho de Melo <acme@ghostprotocols.net>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      2e6599cb
  22. 05 May, 2005 2 commits