mirror/lwip
1
0
Fork 0
mirror of https://github.com/lwip-tcpip/lwip.git synced 2024-11-19 05:10:40 +00:00
Code Issues Packages Projects Releases Wiki Activity
be7ae5e36b
lwip/UPGRADING
David van Moolenbroek 29ddfd1d71 Add support for IPv6 address scopes 
This patch adds full support for IPv6 address scopes, thereby aiming
to be compliant with IPv6 standards in general and RFC 4007 in
particular. The high-level summary is that link-local addresses are
now meaningful only in the context of their own link, guaranteeing
full isolation between links (and their addresses) in this respect.
This isolation even allows multiple interfaces to have the same
link-local addresses locally assigned.

The implementation achieves this by extending the lwIP IPv6 address
structure with a zone field that, for addresses that have a scope,
carries the scope's zone in which that address has meaning. The zone
maps to one or more interfaces. By default, lwIP uses a policy that
provides a 1:1 mapping between links and interfaces, and considers
all other addresses unscoped, corresponding to the default policy
sketched in RFC 4007 Sec. 6. The implementation allows for replacing
the default policy with a custom policy if desired, though.

The lwIP core implementation has been changed to provide somewhat of
a balance between correctness and efficiency on on side, and backward
compatibility on the other. In particular, while the application would
ideally always provide a zone for a scoped address, putting this in as
a requirement would likely break many applications. Instead, the API
accepts both "properly zoned" IPv6 addresses and addresses that, while
scoped, "lack" a zone. lwIP will try to add a zone as soon as possible
for efficiency reasons, in particular from TCP/UDP/RAW PCB bind and
connect calls, but this may fail, and sendto calls may bypass that
anyway. Ultimately, a zone is always added when an IP packet is sent
when needed, because the link-layer lwIP code (and ND6 in particualar)
requires that all addresses be properly zoned for correctness: for
example, to provide isolation between links in the ND6 destination
cache. All this applies to packet output only, because on packet
input, all scoped addresses will be given a zone automatically.

It is also worth remarking that on output, no attempt is made to stop
outgoing packets with addresses for a zone not matching the outgoing
interface. However, unless the application explicitly provides
addresses that will result in such zone violations, the core API
implementation (and the IPv6 routing algorithm in particular) itself
will never take decisions that result in zone violations itself.

This patch adds a new header file, ip6_zone.h, which contains comments
that explain several implementation aspects in a bit more detail.

For now, it is possible to disable scope support by changing the new
LWIP_IPV6_SCOPES configuration option. For users of the core API, it
is important to note that scoped addresses that are locally assigned
to a netif must always have a zone set; the standard netif address
assignment functions always do this on behalf of the caller, though.
Also, core API users will want to enable LWIP_IPV6_SCOPES_DEBUG at
least initially when upgrading, to ensure that all addresses are
properly initialized.
2017-02-03 22:29:57 +01:00

249 lines
11 KiB
Plaintext
Raw Blame History

This file lists major changes between release versions that require
ports or applications to be changed. Use it to update a port or an
application written for an older version of lwIP to correctly work
with newer versions.
(git master)
* [Enter new changes just after this line - do not remove this line]
++ Application changes:
* The IPv6 implementation now supports address scopes. All addresses that have a scope according
to the default policy (link-local unicast addresses, interface-local and link-local multicast
addresses) should now have a zone set on them before being passed to the core API, although
lwIP will currently attempt to select a zone on the caller's behalf when necessary.
Applications that directly assign IPv6 addresses to interfaces (which is NOT recommended)
must now ensure that link-local addresses carry the netif's zone. See the new ip6_zone.h header
file for more information and relevant macros. For now it is still possible to turn off scopes
support through the new LWIP_IPV6_SCOPES option. When upgrading an implementation that uses the
core API directly, it is highly recommended to enable LWIP_IPV6_SCOPES_DEBUG at least for
a while, to ensure e.g. proper address initialization.
(2.0.1)
++ Application changes:
* UDP does NOT receive multicast traffic from ALL netifs on an UDP PCB bound to a specific
netif any more. Users need to bind to IP_ADDR_ANY to receive multicast traffic and compare
ip_current_netif() to the desired netif for every packet.
See bug #49662 for an explanation.
(2.0.0)
++ Application changes:
* Changed netif "up" flag handling to be an administrative flag (as opposed to the previous meaning of
"ip4-address-valid", a netif will now not be used for transmission if not up) -> even a DHCP netif
has to be set "up" before starting the DHCP client
* Added IPv6 support (dual-stack or IPv4/IPv6 only)
* Changed ip_addr_t to be a union in dual-stack mode (use ip4_addr_t where referring to IPv4 only).
* Major rewrite of SNMP (added MIB parser that creates code stubs for custom MIBs);
supports SNMPv2c (experimental v3 support)
* Moved some core applications from contrib repository to src/apps (and include/lwip/apps)
+++ Raw API:
* Changed TCP listen backlog: removed tcp_accepted(), added the function pair tcp_backlog_delayed()/
tcp_backlog_accepted() to explicitly delay backlog handling on a connection pcb
+++ Socket API:
* Added an implementation for posix sendmsg()
* Added LWIP_FIONREAD_LINUXMODE that makes ioctl/FIONREAD return the size of the next pending datagram
++ Port changes
+++ new files:
* MANY new and moved files!
* Added src/Filelists.mk for use in Makefile projects
* Continued moving stack-internal parts from abc.h to abc_priv.h in sub-folder "priv"
to let abc.h only contain the actual application programmer's API
+++ sys layer:
* Made LWIP_TCPIP_CORE_LOCKING==1 the default as it usually performs better than
the traditional message passing (although with LWIP_COMPAT_MUTEX you are still
open to priority inversion, so this is not recommended any more)
* Added LWIP_NETCONN_SEM_PER_THREAD to use one "op_completed" semaphore per thread
instead of using one per netconn (these semaphores are used even with core locking
enabled as some longer lasting functions like big writes still need to delay)
* Added generalized abstraction for itoa(), strnicmp(), stricmp() and strnstr()
in def.h (to be overridden in cc.h) instead of config
options for netbiosns, httpd, dns, etc. ...
* New abstraction for hton* and ntoh* functions in def.h.
To override them, use the following in cc.h:
#define lwip_htons(x) <your_htons>
#define lwip_htonl(x) <your_htonl>
+++ new options:
* TODO
+++ new pools:
* Added LWIP_MEMPOOL_* (declare/init/alloc/free) to declare private memp pools
that share memp.c code but do not have to be made global via lwippools.h
* Added pools for IPv6, MPU_COMPATIBLE, dns-api, netif-api, etc.
* added hook LWIP_HOOK_MEMP_AVAILABLE() to get informed when a memp pool was empty and an item
is now available
* Signature of LWIP_HOOK_VLAN_SET macro was changed
* LWIP_DECLARE_MEMORY_ALIGNED() may be used to declare aligned memory buffers (mem/memp)
or to move buffers to dedicated memory using compiler attributes
* Standard C headers are used to define sized types and printf formatters
(disable by setting LWIP_NO_STDINT_H=1 or LWIP_NO_INTTYPES_H=1 if your compiler
does not support these)
++ Major bugfixes/improvements
* Added IPv6 support (dual-stack or IPv4/IPv6 only)
* Major rewrite of PPP (incl. keep-up with apache pppd)
see doc/ppp.txt for an upgrading how-to
* Major rewrite of SNMP (incl. MIB parser)
* Fixed timing issues that might have lead to losing a DHCP lease
* Made rx processing path more robust against crafted errors
* TCP window scaling support
* modification of api modules to support FreeRTOS-MPU (don't pass stack-pointers to other threads)
* made DNS client more robust
* support PBUF_REF for RX packets
* LWIP_NETCONN_FULLDUPLEX allows netconn/sockets to be used for reading/writing from separate
threads each (needs LWIP_NETCONN_SEM_PER_THREAD)
* Moved and reordered stats (mainly memp/mib2)
(1.4.0)
++ Application changes:
* Replaced struct ip_addr by typedef ip_addr_t (struct ip_addr is kept for
compatibility to old applications, but will be removed in the future).
* Renamed mem_realloc() to mem_trim() to prevent confusion with realloc()
+++ Raw API:
* Changed the semantics of tcp_close() (since it was rather a
shutdown before): Now the application does *NOT* get any calls to the recv
callback (aside from NULL/closed) after calling tcp_close()
* When calling tcp_abort() from a raw API TCP callback function,
make sure you return ERR_ABRT to prevent accessing unallocated memory.
(ERR_ABRT now means the applicaiton has called tcp_abort!)
+++ Netconn API:
* Changed netconn_receive() and netconn_accept() to return
err_t, not a pointer to new data/netconn.
+++ Socket API:
* LWIP_SO_RCVTIMEO: when accept() or recv() time out, they
now set errno to EWOULDBLOCK/EAGAIN, not ETIMEDOUT.
* Added a minimal version of posix fctl() to have a
standardised way to set O_NONBLOCK for nonblocking sockets.
+++ all APIs:
* correctly implemented SO(F)_REUSEADDR
++ Port changes
+++ new files:
* Added 4 new files: def.c, timers.c, timers.h, tcp_impl.h:
* Moved stack-internal parts of tcp.h to tcp_impl.h, tcp.h now only contains
the actual application programmer's API
* Separated timer implementation from sys.h/.c, moved to timers.h/.c;
Added timer implementation for NO_SYS==1, set NO_SYS_NO_TIMERS==1 if you
still want to use your own timer implementation for NO_SYS==0 (as before).
+++ sys layer:
* Converted mbox- and semaphore-functions to take pointers to sys_mbox_t/
sys_sem_t;
* Converted sys_mbox_new/sys_sem_new to take pointers and return err_t;
* Added Mutex concept in sys_arch (define LWIP_COMPAT_MUTEX to let sys.h use
binary semaphores instead of mutexes - as before)
+++ new options:
* Don't waste memory when chaining segments, added option TCP_OVERSIZE to
prevent creating many small pbufs when calling tcp_write with many small
blocks of data. Instead, pbufs are allocated larger than needed and the
space is used for later calls to tcp_write.
* Added LWIP_NETIF_TX_SINGLE_PBUF to always copy to try to create single pbufs
in tcp_write/udp_send.
* Added an additional option LWIP_ETHERNET to support ethernet without ARP
(necessary for pure PPPoE)
* Add MEMP_SEPARATE_POOLS to place memory pools in separate arrays. This may
be used to place these pools into user-defined memory by using external
declaration.
* Added TCP_SNDQUEUELOWAT corresponding to TCP_SNDLOWAT
+++ new pools:
* Netdb uses a memp pool for allocating memory when getaddrinfo() is called,
so MEMP_NUM_NETDB has to be set accordingly.
* DNS_LOCAL_HOSTLIST_IS_DYNAMIC uses a memp pool instead of the heap, so
MEMP_NUM_LOCALHOSTLIST has to be set accordingly.
* Snmp-agent uses a memp pools instead of the heap, so MEMP_NUM_SNMP_* have
to be set accordingly.
* PPPoE uses a MEMP pool instead of the heap, so MEMP_NUM_PPPOE_INTERFACES
has to be set accordingly
* Integrated loopif into netif.c - loopif does not have to be created by the
port any more, just define LWIP_HAVE_LOOPIF to 1.
* Added define LWIP_RAND() for lwip-wide randomization (needs to be defined
in cc.h, e.g. used by igmp)
* Added printf-formatter X8_F to printf u8_t as hex
* The heap now may be moved to user-defined memory by defining
LWIP_RAM_HEAP_POINTER as a void pointer to that memory's address
* added autoip_set_struct() and dhcp_set_struct() to let autoip and dhcp work
with user-allocated structs instead of calling mem_malloc
* Added const char* name to mem- and memp-stats for easier debugging.
* Calculate the TCP/UDP checksum while copying to only fetch data once:
Define LWIP_CHKSUM_COPY to a memcpy-like function that returns the checksum
* Added SO_REUSE_RXTOALL to pass received UDP broadcast/multicast packets to
more than one pcb.
* Changed the semantics of ARP_QUEUEING==0: ARP_QUEUEING now cannot be turned
off any more, if this is set to 0, only one packet (the most recent one) is
queued (like demanded by RFC 1122).
++ Major bugfixes/improvements
* Implemented tcp_shutdown() to only shut down one end of a connection
* Implemented shutdown() at socket- and netconn-level
* Added errorset support to select() + improved select speed overhead
* Merged pppd to v2.3.11 (including some backported bugfixes from 2.4.x)
* Added timer implementation for NO_SYS==1 (may be disabled with NO_SYS_NO_TIMERS==1
* Use macros defined in ip_addr.h to work with IP addresses
* Implemented many nonblocking socket/netconn functions
* Fixed ARP input processing: only add a new entry if a request was directed as us
* mem_realloc() to mem_trim() to prevent confusion with realloc()
* Some improvements for AutoIP (don't route/forward link-local addresses, don't break
existing connections when assigning a routable address)
* Correctly handle remote side overrunning our rcv_wnd in ooseq case
* Removed packing from ip_addr_t, the packed version is now only used in protocol headers
* Corrected PBUF_POOL_BUFSIZE for ports where ETH_PAD_SIZE > 0
* Added support for static ARP table entries
(STABLE-1.3.2)
* initial version of this file
Reference in New Issue View Git Blame Copy Permalink