This series fixes some memory leaks and data races caused in the
same scenario where one thread converts an IPv6 socket into IPv4
with IPV6_ADDRFORM and another accesses the socket concurrently.

Note patch 1 and 5 conflict with these commits in net-next, respectively.

  * 24426654ed3a ("bpf: net: Avoid sk_setsockopt() taking sk lock when called from bpf")
  * 34704ef024ae ("bpf: net: Change do_tcp_getsockopt() to take the sockptr_t argument")


Changes:
  v3 (Resend):
    * Patch 2 & 3:
      * CC blamed commits' -EHOSTUNREACH authors to make patchwork happy

  v3: https://lore.kernel.org/netdev/20220929012542.55424-1-kuniyu@amazon.com/
    * Patch 2:
      * Add comment for np->rxopt.all = 0
      * Add inet6_cleanup_sock()
    * Patch 3:
      * Call inet6_cleanup_sock() instead of inet6_destroy_sock()

  v2: https://lore.kernel.org/netdev/20220928002741.64237-1-kuniyu@amazon.com/
    * Patch 3:
      * Move inet6_destroy_sock() from sk_prot->destroy()
        to sk->sk_destruct() and fix CONFIG_IPV6=m build failure
    * Patch 5:
      * Add WRITE_ONCE()s in tcp_v6_connect()
      * Add Reported-by tags and KCSAN log in changelog

  v1: https://lore.kernel.org/netdev/20220927161209.32939-1-kuniyu@amazon.com/


Kuniyuki Iwashima (5):
  tcp/udp: Fix memory leak in ipv6_renew_options().
  udp: Call inet6_destroy_sock() in setsockopt(IPV6_ADDRFORM).
  tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
  ipv6: Fix data races around sk->sk_prot.
  tcp: Fix data races around icsk->icsk_af_ops.

 include/net/ipv6.h       |  2 ++
 include/net/udp.h        |  2 +-
 net/core/sock.c          |  6 ++++--
 net/ipv4/af_inet.c       | 23 ++++++++++++++++-------
 net/ipv4/tcp.c           | 10 ++++++----
 net/ipv4/udp.c           |  8 ++++++--
 net/ipv6/af_inet6.c      | 15 ++++++++++++++-
 net/ipv6/ipv6_sockglue.c | 34 +++++++++++++++++++---------------
 net/ipv6/tcp_ipv6.c      |  6 ++++--
 net/ipv6/udp.c           | 15 ++++++++++++++-
 10 files changed, 86 insertions(+), 35 deletions(-)

-- 
2.30.2