6.2
 - list:set timeout variant fixes
 - References are protected by rwlock instead of mutex
 - Add explicit text message to detect patched kernel (netlink.patch)
 - Timeout can be modified for already added elements

6.1
 - The hash:*port* types ignored the address range with non TCP/UDP, fixed
 - Fix checking the revision number of the set type at create command
 - SCTP, UDPLITE support to hash:*port* types added
 - Fix revision reporting got broken by the revision checking patch

6.0
 - Reorganized kernel/ subdir
 - netfilter: ipset: fix linking with CONFIG_IPV6=n (Patrick McHardy)
 - netfilter: ipset: send error message manually
 - netfilter: ipset: add missing break statemtns in 
   ip_set_get_ip_port() (Patrick McHardy)
 - netfilter: ipset: add missing include to xt_set.h (Patrick McHardy)
 - netfilter: ipset: remove unnecessary includes (Patrick McHardy)
 - netfilter: ipset: use nla_parse_nested() (Patrick McHardy)
 - Separate ipset errnos completely from system ones and bump protocol
   version
 - Use better error codes in xt_set.c
 - Fix sparse warning about shadowed definition
 - bitmap:ip type: flavour specific adt functions (Patrick McHardy's review)
 - bitmap:port type: flavour specific adt functions (Patrick McHardy's
   review)
 - Move the type specifici attribute validation to the core
   (suggested by Patrick McHardy)
 - Use vzalloc() instead of __vmalloc() (Eric Dumazet, Patrick McHardy)
 - Use meaningful error messages in xt_set.c (Patrick McHardy's review)
 - Constified attribute cannot be written (Patrick McHardy's review)
 - Send (N)ACK at dumping only when NLM_F_ACK is set
   (Patrick McHardy's review)
 - Correct the error codes: use ENOENT and EMSGSIZE (Patrick McHardy's
   review)

5.4
 - Fixed broken ICMP and ICMPv6 handling
 - Fix trailing whitespaces and pr_* messages
 - Un-inline functions which are not small enough (suggested by Patrick McHardy)
 - Fix module loading at create/header command (reported by Patrick McHardy)
 - Fix wrong kzalloc flag in type_pf_expire
 - The get_ip*_port functions are too large to be inlined, move into the core
 - Add missing __GFP_HIGHMEM flag to __vmalloc (suggested by Eric Dumazet)
 - Enforce network-order data in the netlink protocol
 - Use annotated types and fix sparse warnings (suggested by Patrick McHardy)
 - Move ip_set_alloc, ip_set_free and ip_set_get_ipaddr* into core
   (suggested by Patrick McHardy)
 - NETMASK*, HOSTMASK* macros are too generic, use small inline functions
   (suggested by Patrick McHardy)
 - Use static LIST_HEAD() for ip_set_type_list (suggested by Patrick McHardy)
 - Move NLA_PUT_NET* macros to include/net/netlink.h (suggested by Patrick
   McHardy)
 - The module parameter max_sets should be unsigned int (reported by Patrick
   McHardy)
 - Get rid of ip_set_kernel.h (suggested by Patrick McHardy)
 - Fix the placement style of boolean operators at continued lines
   (suggested by Patrick McHardy)

5.3
 - There is no need to call synchronize_net() at swapping
 - Replace strncpy with strlcpy at creating a set
 - Update copyright date and some style changes
 - Use jhash.h accepted in kernel, with backward compatibility
 - Separate prefixlens from ip_set core
 - Remove unused ctnl parameter from call_ad (Jan Engelhardt)
 - Comment the possible return values of the add/del/test type-functions

5.2
 - Kernel version check at minimal supported version was mistyped, now fixed.

5.1
 - Kernel version compatibility: support bumped starting from 2.6.34
   (Supporting older kernel releases would mean too much burden for me,
   sorry.)
 - kernel: use EXPORT_SYMBOL_GPL (Jan Engelhardt)
 - kernel: const annotations (Jan Engelhardt)
 - kernel: use __read_mostly for registration-type structures
   (Jan Engelhardt)
 - kernel: do not mix const and __read_mostly (Jan Engelhardt)
 - xt_set: avoid user types in exported kernel headers (Jan Engelhardt)
 - build: enable parallel building (Jan Engelhardt)
 - Fix Kbuild for me to delete backup files

5.0
 - New main branch - ipset completely rewritten

4.2
  - nethash and ipportnethash types counted every entry twice
    which could produce bogus entries when listing/saving these types
    of sets (bug reported by Husnu Demir)

4.1
  - Do not use init_MUTEX either (Jan Engelhardt)
  - Improve listing/saving hash type of sets by not copying empty
    entries unnecessarily to userspace.

4.0
  - Compilation of ip_set_iptree.c fails with kernel 2.6.20 due to
    missing include of linux/jiffies.h (Jan Engelhardt)
  - Do not use DECLARE_MUTEX (compatibility fix on 2.6.31-rt, Jan
    Engelhardt)
  - Flushing iptreemap type of sets caused high ksoftirqd load due to
    zeroed out gc parameter (bug reported by Georg Chini)
  - New protocol is introduced to handle aligment issues properly
    (bug reported by Georg Chini)
  - Binding support is removed

3.2
  - Mixed up formats in ip_set_iptree.c (Rob Sterenborg)
  - Don't use 'bool' for backward compatibility reasons (Rob Sterenborg)

3.1
  - Nonexistent sets were reported as existing sets when testing
    from userspace in setlist type of sets (bug reported by Victor A.
    Safronov)
  - When saving sets, setlist type of sets must come last in order
    to satisfy the dependency from the elements (bug reported by Marty B.)
  - Sparse insists that the flags argument to kmalloc() is gfp_t
    (Stephen Hemminger)
  - Correct format specifiers and change %i to %d (Jan Engelhardt)
  - Fix the definition of 'bool' for kernels <= 2.6.18 (Jan Engelhardt)

3.0
  - New kernel-userspace protocol release
  - Bigendian and 64/32bit fixes (Stefan Gula, bugzilla id 593)
  - Support of 2.4.3[67].*  kernels fixed
  - Compiling with debugging enabled fixed

2.5.0
  - Use the spinlock initiator instead of setting the locks directly
    as it causes compilation errors with 2.6.29-rt (Jan Engelhardt).
2.4.9
  - References to the old include file replaced with new one in order to
    really use the new Jenkins' hash function.

2.4.8
  - The Jenkins' hash lookup2() replaced with Jenkins' faster/better lookup3()
    hash function.
  - Bug fixed: after elements are added and deleted from a hash, an element
    can successfully be added in spite it's already in the hash and thus
    duplicates can occur (Shih-Yi Chen). 
  - Compatibility with old gcc without 'bool' added.

2.4.7
  - Typo which broke compilation with kernels < 2.6.28
    fixed (reported by Richard Lucassen, Danny Rawlins)

2.4.6
  - Compatibility fix for kernels >= 2.6.28

2.4.5
  - setlist type does not work properly together with swapping
    sets, bug reported by Thomas Jacob.
  - Include linux/capability.h explicitly in ip_set.c (Jan Engelhardt)

2.4.4
  - Premature checking prevents to add valid elements to hash
    types, fixed (bug reported by JC Janos).

2.4.2
  - When flushing a nethash/ipportnethash type of set, it can
    lead to a kernel crash due to a wrong type declaration,
    bug reported by Krzysztof Oledzki.
  - iptree and iptreemap types require the header file linux/timer.h,
    also reported by Krzysztof Oledzki.
  
2.4.1
  - Zero-valued element are not accepted by hash type of sets
    because we cannot make a difference between a zero-valued
    element and not-set element.
2.4
  - ipportiphash, ipportnethash and setlist types added
  - set type modules reworked to avoid code duplication
    as much as possible, code unification macros 
  - expand_macros Makefile target added to help debugging
    code unification macros
  - ip_set_addip_kernel and ip_set_delip_kernel
    changed from void to int, __ip_set_get_byname and
    __ip_set_put_byid added for the sake of setlist type
  - unnecessary includes removed
  - compatibility fix for kernels >= 2.6.27:
    semaphore.h was moved from asm/ to linux/ (James King)
  - ChangeLog forked for kernel part
