Based from pppd 2.4.5, released 2009-11-17, with huge changes to match
code size and memory requirements for embedded devices, including:
- Gluing together the previous low-level PPP code in lwIP to pppd 2.4.5, which
is more or less what pppd sys-* files are, so that we get something working
using the unix port.
- Merged some patchs from lwIP Git repository which add interesting features
or fix bugs.
- Merged some patchs from Debian pppd package which add interesting features
or fix bugs.
- Ported PPP timeout handling to the lwIP timers system
- Disabled all the PPP code using filesystem access, replaced in necessary cases
to configuration variables.
- Disabled all the PPP code forking processes.
- Removed IPX support, lwIP does not support IPX.
- Ported and improved random module from the previous PPP port.
- Removed samba TDB (file-driven database) usage, because it needs a filesystem.
- MS-CHAP required a DES implementation, we added the latest PolarSSL DES
implementation which is under a BSD-ish license.
- Also switched to PolarSSL MD4,MD5,SHA1 implementations, which are meant to be
used in embedded devices with reduced memory footprint.
- Removed PPP configuration file parsing support.
- Added macro definition EAP_SUPPORT to make EAP support optional.
- Added macro definition CHAP_SUPPORT to make CHAP support optional.
- Added macro definition MSCHAP_SUPPORT to make MSCHAP support optional.
- Added macro definition PAP_SUPPORT to make PAP support optional.
- Cleared all Linux syscall calls.
- Disabled demand support using a macro, so that it can be ported later.
- Disabled ECP support using a macro, so that it can be ported later.
- Disabled CCP support using a macro, so that it can be ported later.
- Disabled CBCP support using a macro, so that it can be ported later.
- Disabled LQR support using a macro, so that it can be ported later.
- Print packet debug feature optional, through PRINTPKT_SUPPORT
- Removed POSIX signal usage.
- Fully ported PPPoS code from the previous port.
- Fully ported PPPoE code from the previous port.
- Fully ported VJ compression protocol code from the previous port.
- Removed all malloc()/free() use from PPP, replaced by stack usage or PBUF.
- Disabled PPP server support using a macro, so that it can be ported later.
- Switched all PPP debug to lwIP debug system.
- Created PPP Control Block (PPP PCB), removed PPP unit integer everywhere,
removed all global variables everywhere, did everything necessary for
the PPP stack to support more than one PPP session (pppd only support
one session per process).
- Removed the statically allocated output buffer, now using PBUF.
- Improved structure size of all PPP modules, deep analyze of code to reduce
variables size to the bare minimum. Switched all boolean type (char type in
most architecture) to compiler generated bitfields.
- Added PPP IPv6 support, glued lwIP IPv6 support to PPP.
- Now using a persistent netif interface which can then be used in lwIP
functions requiring a netif.
- Now initializing PPP in lwip_init() function.
- Reworked completely the PPP state machine, so that we don't end up in
anymore in inconsistent state, especially with PPPoE.
- Improved the way we handle PPP reconnection after disconnect, cleaning
everything required so that we start the PPP connection again from a
clean state.
- Added PPP holdoff support, allow the lwIP user to wait a little bit before
reconnecting, prevents connection flood, especially when using PPPoL2TP.
- Added PPPoL2TP LAC support (a.k.a. UDP tunnels), adding a VPN client
feature to lwIP, L2TP being a widely used tunnel protocol.
- Switched all used PPP types to lwIP types (u8t, u16t, u32t, ...)
- Added PPP API "sequential" thread-safe API, based from NETIFAPI.
This patch adds support for RFC3542-style checksum computation on raw,
IPv6 sockets via the IPV6_CHECKSUM socket option.
This allows the development of application-layer utilities such as
ping6 which are unable to compute the raw packet checksum without a
prior knowledge of the source address selection.
The LWIP_MALLOC_MEMPOOL macro needs to use the aligned size of the
memp_malloc_helper structure, since mem_alloc() uses it to calculate
the required pool element size. If LWIP_MEM_ALIGN_SIZE(x) is redefined
to align to something larger than 4, then in some cases
the current code can lead to unexpected mem_alloc() failures.
For example:
#define LWIP_MEM_ALIGN_SIZE(size) (((size) + 31) & ~31)
and the largest MALLOC pool is of size 60 bytes, e.g.:
#define LWIP_MALLOC_MEMPOOL(256, 60)
then the following call:
mem_malloc(58)
will cause an assertion.
