While TCP_OVERSIZE works only when tcp_write() is used with
TCP_WRITE_FLAG_COPY, this new code achieves
similar benefits for the use case that the caller manages their own
send buffers and passes successive chunks of those to tcp_write()
without TCP_WRITE_FLAG_COPY.
In particular, if a buffer is passed to
tcp_write() that is adjacent in memory to the previously passed
buffer, it will be combined into the previous ROM pbuf reference
whenever possible, thus extending that ROM pbuf rather than allocating
a new ROM pbuf.
For the aforementioned use case, the advantages of this code are
twofold:
1) fewer ROM pbufs need to be allocated to send the same data, and,
2) the MAC layer gets outgoing TCP packets with shorter pbuf chains.
Original patch by Ambroz Bizjak <ambrop7@gmail.com>
Edited by David van Moolenbroek <david@minix3.org>
Signed-off-by: goldsimon <goldsimon@gmx.de>
This corrects a case in lwip_netconn_do_writemore() where if a
non-blocking socket receives ERR_MEM in a call to tcp_write(), it would
return ERR_MEM, which would result in ENOMEM coming out of the socket
layer
This case can be gracefully handled by returning ERR_WOULDBLOCK since the
socket is already marked as no longer writable and sent_tcp/poll_tcp will
mark the socket as writable again based on available buffer space
This is very similiar to how ERR_MEM is resolved for blocking sockets
Change lwIP UDP API to match socket behavior. Multicast traffic is now only received on a UDP PCB (and therefore on a UDP socket/netconn) when the PCB is bound to IP_ADDR_ANY.
Generally speaking, packets with a loopback destination address -
127.0.0.1 for IPv4 and ::1 for IPv6 - should not be accepted on
non-loopback interfaces. For IPv4, this is implied by RFC 1122
Sec. 3.2.1.3. For IPv6, it is mandated by RFC 4291 Sec. 2.5.3.
Failure to perform this filtering may have security implications, as
applications that bind sockets to loopback addresses may not expect
that nodes on the local external network be able to produce traffic
that will arrive at such sockets.
With this patch, lwIP drops packets that are sent to a loopback
address but do not originate from the interface that has the loopback
address assigned to it. This approach works regardless of whether it
is lwIP or the system using it that implements a loopback netif. The
only exception that must be made is for configurations that enable
netif packet loopback but disable the lwIP loopback netif: in that
case, loopback packets are routed across non-loopback netifs and would
thus be lost by the new filter as well.
For IPv6, loopback-destined packets are also no longer forwarded; the
IPv4 forwarding code already had a check for that.
As a small performance improvement, the IPv6 link-local/loopback
address check is now performed only once per packet rather than
repeatedly for every candidate netif.
In general, netif_default may be NULL, and various places in the code
already check for this case before attempting to dereference the
netif_default pointer. Some places do not perform this check though,
and may cause null pointer dereferences if netif_default is not set.
This patch adds NULL checks to those places as well.
It is better to present correct IP types in netconn API.
Netconn API now accepts IPv6 mapped IPv4 addresses as well as IPv6 and IPv4 in send(), bind() and connect(), but does NOT map IPv4 to IPv6 mapped IPv4 in getaddr() and receive() functions.
bind() may change IP type when previous type is IPADDR_TYPE_ANY
connect() IP type must exactly match bind IP type
Use correct IPADDRx_ANY type when calling ip_route()
IPv6 netconns are created as IPADDR_TYPE_ANY raw/udp/tcp PCBs internally
bind, connect and sendto now accept IPv6 mapped IPv4 addresses or IPv4 addresses as argument
getaddr and receive functions now return IPv6 mapped IPv4 addresses instead of IPv4 addresses
This behavior is close to BSD socket API
This commit returns LwIP to previous behavior where if the next unsent
segment can't be sent due to the current send window, we start the
persist timer. This is done to engage window probing in the case that
the subsequent window update from the receiver is dropped, thus
preventing connection deadlock
This commit refines the previous logic to only target the following case:
1) Next unsent segment doesn't fit within the send window (not
congestion) and there is some room in the window
2) Unacked queue is empty (otherwise data is inflight and the RTO timer
will take care of any dropped window updates)
See commit d8f090a759 (which removed this
behavior) to reference the old logic. The old logic falsely started the
persit timer when the RTO timer was already running.
This commit cleans up a duplicate #if check for LWIP_WND_SCALE in init.c
which was already under #if LWIP_WND_SCALE
This commit also improves documentation for TCP_WND in the window scaling
case to communicate TCP_WND is always the calculated (scaled) window value,
not the value reported in the TCP header
Our developers were confused by having to set both the window and scaling
factor and only after studying the usage of TCP_WND throughout the code, was
it determined to be the calculated (scaled) window
This is done in the pppd upstream and was disabled because we don't have
the allowed addresses list required for the auth_ip_addr function.
This is mostly necessary for PPP in server mode to prevent the peer to
use the IP address it wants instead of the one we want, which is
currently allowed.
Rewrite auth_ip_addr in a simple way where we forbid PPP peer to use
loopback net, a multicast address or a reserved class address. Added
to that we consider that PPP in server mode with peer required to
authenticate must provide the peer IP address, reject any IP address
wanted by peer different than the one we wanted. This is actually
an allowed addresses "list" of one entry that follows what is done
in the unused auth_ip_addr function.
Commit 7df5496e7b revealed a regression introduced in commit 5a71509353
which broke IPCP reset state.
ask_for_local was set to 0 if ouraddr initial value is 0, if
ask_for_local was false go->ouraddr was cleared in reset callback,
commit 5a71509353 breaks it by removing this clearing. This regression
was silent because the whole ppp pcb runtime data was cleared before
reconnecting until commit 7df5496e7b which removed this giant clearing.
Fix it by reintroducing ask_for_local boolean value, with proper initial
value following what unused function ip_check_options do.
Fixes: 7df5496e7b ("PPP, rework initial/reconnect cleanup")
Fixes: 5a71509353 ("PPP, CORE, IPCP: removed useless ask_for_local boolean")
p needs to point to LWIP_MEM_ALIGN(memp_pools[i]->base) otherwise it will cause
assertion in overflow checking.
Fixes: c838e1ed5b ("Implement possibility to declare private memory pools")
Signed-off-by: Axel Lin <axel.lin@ingics.com>
