According to rfc5681:
https://tools.ietf.org/html/rfc5681
Paragraph 3.2. Fast Retransmit/Fast Recovery
The TCP sender SHOULD use the "fast retransmit" algorithm to detect
and repair loss, based on incoming duplicate ACKs. The fast
retransmit algorithm uses the arrival of 3 duplicate ACKs (as defined
in section 2, without any intervening ACKs which move SND.UNA) as an
indication that a segment has been lost. After receiving 3 duplicate
ACKs, TCP performs a retransmission of what appears to be the missing
segment, without waiting for the retransmission timer to expire.
Now consider the following scenario:
Server sends packets to client P0, P1, P2 .. PK.
Client sends packets to server P`0 P`1 ... P`k.
I.e. it is a pipelined conversation. Now lets assume that P1 is lost, Client will
send an empty "duplicate" ack upon receive of P2, P3... In addition client will
also send a new packet with "Client Data", P`0 P`1 .. e.t.c. according to sever receive
window and client congestion window.
Current implementation resets "duplicate" ack count upon receive of packets from client
that holds new data. This in turn prevents server from fast recovery upon 3-duplicate acks
receive.
This is not required as in this case "sender unacknowledged window" is not moving.
Signed-off-by: Solganik Alexander <sashas@lightbitslabs.com>
Two new API:
err_t mdns_search_service(const char *name, const char *service, enum mdns_sd_proto proto,
struct netif *netif, search_result_fn_t result_fn, void *arg,
s8_t *request_id);
void mdns_search_stop(s8_t request_id);
One compilation flags:
LWIP_MDNS_SEARCH
One options flags:
MDNS_MAX_REQUESTS
Some structure declarations moved to allow use by callback result function.
Result domain names are early uncompress before calling application callback
because it cannot be made by application itself.
Allow search services with multiples labels included, like '_services._dns-sd'.
Search for `_services._dns-sd._udp.local.` is handled in a special way.
Only `PTR` answers are send back to the application.
The `mdns_search_service()` function won't assert if no more space in `mdns_request`
table, just return an error if too many simultanous requests.
Apparently the TFTP server now also invokes the error() function in the
tftp_context struct.
Some tftp clients (for example Windows 10 TFTP client) will open the
remote file before checking the local file can be opened - and will then
send an error indication to the server to indicate there was an error
opening the local file. When the happens, the LWIP tftp server will
invoke the error() member of the tftp_context.
This adds support for RFC4075 SNTP server configuration via DHCPv6.
The DHCPv6 options transmitted are now conditional on how LwIP is
configured.
A new SNTP application option SNTP_GET_SERVERS_FROM_DHCPV6 is used
to enable. For simplicity this is configured to use the global
LWIP_DHCP6_GET_NTP_SRV configuration setting.
Tests:
- Check the global options now control the DHCPv6 request sent
in Wireshark
- Check against 0, 1 and 3 SNTP servers configured on an odhcpd
server configured to support RFC 4075 SNTP server lists.
Verify that the SNTP server list is updated on connection
establishment on an ESP8266 WeMOS D1.
- Verify that SNTP packets are sent and recieved from a
configured server and that system time is updated.
Signed-off-by: David J. Fiddes <D.J@fiddes.net>
Use only one entropy/ctr_drbg context for all altcp_tls_config structure allocated.
(Small adjustments before committing: fix coding style, adapt to changes in master)
- client->msg_idx can be > MQTT_VAR_HEADER_BUFFER_LEN in long message splitted in multiple pbufs
- renamed fixed_hdr_idx to fixed_hdr_len because it is length of fixed header in rx_buffer, not an index to it
- removed the cpy_start as data always copied right after the fixed header
Make pppoe_create() actually store the passed service name and
concentrator name, so that they are passed in the PADI/PADR/PADS
packets.
Assume that the user application won't be freeing the strings and just
copy the string pointers, therefore remove the mem_free() in
pppoe_destroy().
Since only the pointers are copied now, make them 'const' in
pppoe_softc.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
lwip/src/netif/ppp/pppoe.c:768:24: error: pointer targets in passing argument 1 of ‘strlen’ differ in signedness [-Werror=poin$
l1 = (int)strlen(sc->sc_service_name);
lwip/src/netif/ppp/pppoe.c:772:24: error: pointer targets in passing argument 1 of ‘strlen’ differ in signedness [-Werror=poin$
l2 = (int)strlen(sc->sc_concentrator_name);
sc->sc_service_name and sc->sc_concentrator_name are best defined as
char* because there are passed to libc strings functions which expect
a char*.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
Rename PPPOE_TODO to PPPOE_SCNAME_SUPPORT because this is the only
feature enclosed by them. Prepare for proper service name and
concentrator name support by moving PPPOE_SCNAME_SUPPORT define to
ppp_opts.h.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
src/apps/mdns/mdns.c: In function 'mdns_debug_print_answer':
src/apps/mdns/mdns.c:796:24: warning: ', rdata = ' directive output may be truncated writing 10 bytes into a region of size between 8 and 15 [-Wformat-truncation=]
snprintf(string, 35, "Type = %2d, class = %1d, rdata = ", a->info.type, a->info.klass);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/apps/mdns/mdns.c:796:3: note: 'snprintf' output between 31 and 38 bytes into a destination of size 35
snprintf(string, 35, "Type = %2d, class = %1d, rdata = ", a->info.type, a->info.klass);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
lwip/lwip/src/apps/http/makefsdata/makefsdata.c:929:56: error: ‘snprintf’ output may be truncated before the last format character [-Werror=format-truncation=]
snprintf(qualifiedName, sizeof(qualifiedName), "%s/%s", curSubdir, filename);
/home/dziegel/lwip/lwip/src/apps/http/makefsdata/makefsdata.c:929:3: note: ‘snprintf’ output 2 or more bytes (assuming 257) into a destination of size 256
snprintf(qualifiedName, sizeof(qualifiedName), "%s/%s", curSubdir, filename);
Reduce subdir string length by 3 bytes to make the warning go away. The whole file path including directory AND filename is limited to MAX_PATH_LEN - so it is reasonable to reserve 3 bytes less for directory - the filename won't fit anyway in the remaining 3 bytes.
If the host observes a response (after probing) containing RR's
that he thought were unique to him, there is a conflict. If a host
observes such conflict, it resets back to probing and the probing
procedures will resolve the conflict. (RFC6762 section 9)
by replacing strnlen with strlen. It's a user-supplied string, so we can assume it is correctly \0 terminated (as done several times elsewhere in the code)
According to RFC6762 section 8.1:
If fifteen conflicts occur within any ten-second period, then the
host MUST wait at least five seconds before each successive
additional probe attempt.
If the host restarts mDNS, instead of a probe wait timeout randomly
chosen between 0 and 250ms, we wait 5s if more then 15 conflicts
occured in 10seconds. This flag is reset from the moment probing
succeeded.
mDNS: first version probe tiebreaking added
This first version works for our MCU boards but does only compare
the first question and answer. It does not sort the list of answers
and it does not search for answers or questions.
mDNS: improved probe question handling
For every probe question in our packet, we check the presence
in the incomming packet until we find a match. then we perform
the tiebreaking. At the moment we still only look at the first
answer.
mDNS: evaluate all answers in the authoritative section
The authoritative sections of both messages are evaluated.
The records that answer the probe question are sorted in to lists.
Then the lists are compared pairwise. A winner and a loser are
chosen and the host reacts on the outcome (ignore or stop probing).
mDNS: add support for name decompression in rdata when comparing.
For probe tiebreaking we need to compare RR's. It is possible that
the rdata of an RR contains compessed names (SRV record). We need
to decompress it first before we do the comparison.
Probe packet creation -> by splitting the definition and the sending
we can use the definition function for probe tiebreaking.
outpkt creation -> by splitting the creation and sending we can use
the creation function for probe tiebreaking.
The host only responded to a probe query via unicast because
according to the RFC, a probe should have the QU bit on.
This is a should and not a must so we need to be careful.
We added multicast probe answering support with the needed timouts.
Avahi for example probes with the QM queries. With this commit the
conflict is resolved.
The implementation was not fully to the standard. This commit adds
a nicer state machine implementation and multiple announce messages
with a minimum of 2.
See below commit messages for more information
mDNS: support for direct and delayed sends
There are two ways to send a response, directly and after a delay.
A probe or an announce msg are now send via the direct way and
all responses are send via the delayed way.
mDNS improved delay vs direct send behavior on questions
All multicast answers are delayed at the moment. While all unicast
answers are send out directly. A unicast answer is send when this
is requested by the QU bit, when a unicast question was send or
when the question originated from a legacy querier.
mDNS: add probe query detection.
If a probe query is detected a direct unicast respond is send.
Independent of the QU/QM bit.
mDNS split delayed multicast msgs into ipv4 and ipv6 buffers.
We are implementing a two resolvers in one (IPv6 and IPv4 together).
For directly send answers, this does not matter. But for delayed
answers, we need to make a separate buffer for both.
mDNS: addr bug, we should not clear full outmsg
memset deleted also the dest_addr and dest_port, which should remain
and is constant. This commit contains a function that resets only the
needed parts of the outmsg struct.
mDNS: do not multicast a rr within one second.
RFC6762 section 6: prevent network flooding. When a multicast packet
is send out, we start a timeout of 1s within this 1 second all
multicast requests are ignored. We do not make a difference between
the records, we set the delay for all records.
mDNS: improved split for unicast vs multicast and direct vs delayed
unicast delayed message are now possible and multicast direct msgs to.
MDNS: changed printfs to lwip debug messages
MDNS: change timeouts from max time to random time
mDNS: send multicast response on QU questions if not multicasted recently.
If a QU question is received, the responder should multicast the
answer if it did not multicast that record within 25% of it's ttl.
we implemented a stripped down version, meaning that we look at the
records as one set and use one timer for all records. So if the
responder multicasted a record within 30s of the QU question it
will respond with a unicast answer. if not, it will respond
multicast.
mDNS: timeouts -> create function for mdns timeout handling
mdns_set_timeout will check if the timer is running or not and will
update the flag to running after starting the timer.
Multicast timeouts were not set everywhere they needed to be. This
is solved.
mulit <-> multi typo fixed.
mDNS: solve commenting and style issues
mDNS: add #if LWIP_IPVx to new code
LWIP_IPV4/6 can be enabled or disabled, all combination should work.
it's more practical to use netif as a wrapper instead of wrapping
netif in the mdns packets. netif contains all information.
Netif is passed along were needed.
MDNS_TTL changed to MDNS_IP_TTL for clarity.
The mDNS RR ttl does not need to be settalbe, this is against the
RFC. RFC6762 states that for rr's with the hostname in them
somewhere, the ttl should be 120s. If it's not in their it should
be 75 minutes and if the answer is send to a legacy dns querier,
it should be no more then 10s.
This patch corrects the ttl behavior to the RFC.
see RFC 6762 section 8.2: for tiebreaking to work correctly in all
cases, the Authority section must contain all the records and
proposed rdata being probed for uniqueness.
mDNS.c needed to become cleaner. Domain related functions are
moved to the mdns_domain.c util module. The output related
functions are split off (clear separation between defining the
packet and generating the packet). The output functions can
now be found in mdns_out.c.
mDNS move probe question packet generation to mdns_send_outpacket
The probe any questions were added to the pbuf in the send_probe
routine. It is better if we move all pbuf generation to the output
function so later on packets can be delayed etc. keep it all in
one place.
mDNS: move legacy question generation to mdns_send_outpacket
It's better to do the pbuf generation in one place.
Especially important for message delaying etc.
mDNS take out domain related functionality and put in other file.
The mDNS file is getting very big and a lot still needs to be added.
For clarity reasons it's better to split these domain functions
from the main mDNS file.
mDNS split off output related functionality and put in other file
A lot of functions are only needed for the generation of the pbuf,
by separating them into another file we clean up the mdns file.
We only need the mdns_send_outpacket function as interface.
Packet definition is now completely separated from packet
generation.
This reverts commit 684adaca29.
It changes the behaviour to assert for applications running good so far.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
lwip_getsockopt_internal/lwip_setsockopt_internal were renamed to
lwip_getsockopt_impl/lwip_setsockopt_impl. Update the comment accordingly.
Signed-off-by: Axel Lin <axel.lin@ingics.com>
This is the default way for mbedTLS. Add entropy sources via defines (see
mbedtls_entropy_init).
This removes the use of ALTCP_MBEDTLS_RNG_FN
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
New etharp queries should restart the 5 second timeout on the ARP
table entry if it is still pending.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
The instance of the SNMP node was found, that's why the functions have been called. So if set_test() and/or set_value() do not exist, it means the SNMP node instance is not writable.
Add define LWIP_HTTPD_SSI_EXTENSIONS to make the extension list
overridable, clarify documentation of LWIP_HTTPD_SSI_BY_FILE_EXTENSION
a bit.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
By defining LWIP_PLATFORM_ASSERT to a function, we can prevent arch.h
from including stdio.h and stdlib.h
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
This reverts commit 413b26a7e1.
It breaks the build if LWIP_SOCKET is disabled.
Plus it breaks git history for inet.h and sockets.h for no real reason.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
New macro LWIP_SOCKET_HEADERS can be used to replace
LwIP sockets headers by OS's headers.
By default LWIP_SOCKET_HEADERS==LWIP_SOCKET, when LWIP_SOCKET==1
and LWIP_SOCKET_HEADERS==0, user must provide their custom headers
by setting LWIP_INCLUDE_SOCKETS and LWIP_INCLUDE_INET.
pppos_input() is safe to call from outside tcpip_thread when
PPP_INPROC_IRQ_SAFE == 1, so only check if PPP_INPROC_IRQ_SAFE == 0
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
This reverts commit 1570dd8ad1.
Buffering pcb->local_port is needed because TCP_PCB_REMOVE_ACTIVE()
sets it to 0 via tcp_pcb_remove() (comment: "reset the local port
to prevent the pcb from being 'bound'").
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
- implement server/client mode (API is nicer to use)
- Increase TFTP_MAX_MODE_LEN to be able to contain "netascii" transfer mode
- Adapt tftp_example to changes
Squashed commit of the following:
commit 2d98d8e2ef1941c3824ffb874f1e529d284667fc
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Sep 13 16:15:06 2018 +0200
AUTOIP: correct functionality autoip_supplied_address
It does not mean that if autoip is bound, it also supplied the
netif address. A check is added.
commit 2ca0a2183991ef73860c4207d95799b37acc64cc
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Sep 13 16:06:49 2018 +0200
AUTOIP: keep using the same link local address as much as possible
Only calculate a new link local address at start up or when a
conflict occured. On link up or down -> keep same address.
TODO: in the future a function for persistent storage should be
added.
commit aa70a693351e4c898aa28d8521308794614838f1
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Sep 13 14:43:14 2018 +0200
ACD, AUTOIP & DHCP: make link up & down functions where needed
When the link goes down or up the approriate functions should be
stopped or started again. To accomodate this, network_changed is
adjusted to network_changed_link_up and network_changed_link_down.
DHCP does not need to control AUTOIP. AUTOIP can take care of
itself. The only thing DHCP needs to do is starting it when
discovering is failing. The AUTOIP state variable is removed from
DHCP.
commit ad469eb006b47f8a8c37f7c0de0216f47a8c19c7
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Sep 13 11:25:58 2018 +0200
ACD: add address change listener + passive conflict detection mode
In the case their previously was a LL address on a netif that
is now configured with a routable address, we want the LL
address to be able to keep receiving packets.
for as long as the LL address is available on the interface it
should do ongoing conflict detection. But we cannot defend when
the LL address is not the netif address any more.
An address change listener is added to detect when an ACD module
needs to go from active ongoing conflict detection to passive.
When a conflict is detected autoip is stopped and will not be able
to receive packets any more. Because we have a valid routable
address on the netif, autoip is not restarted.
commit 07c4ec20cea78e2b4a6f5599569abaf075619c62
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Sep 11 16:25:37 2018 +0200
ACD: make module reusable within a netif
DHCP and AUTOIP both have ACD running simultaniously. The ACD
struct is added to the DHCP and AUTOIP structs. In the netif a list
of ACD modules is kept to loop over if functions need to perform
some action on all ACD modules (for example tmr function). With
acd_add a module can be added to the list. ACD_FOREACH loops over
the list similar to NETIF_FOREACH.
commit ee3b4585b7768f5353dd80190a2929bad45f7ff4
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Sep 10 16:36:18 2018 +0200
etharp & acd: add probe and announce functions to etharp.
The new probe and announce functions are independent of
netif->ip_addr. This means we can keep the LL address working
while we start to probe for the newly received routable address.
The netif->ip_addr does not need to be any for probing to work
with this patch.
commit 7d3032bae8f1b8081368a807682388eb642729e0
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Sep 3 13:35:51 2018 +0200
autoip -> unused random function removed + small comment update
commit be749ba4eb26ddc69233c85d532dc035741275c5
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Aug 31 16:37:36 2018 +0200
ACD: update comment in header
commit 4491842991c90b3a58fa327f70aa42f04174546b
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Aug 31 15:44:45 2018 +0200
ACD: subscribe to link down messages to stop the acd process
When the link goes down the acd process should be stopped
independent of the acd client. Otherwise the acd will keep probing
or announcing while their simply is no connection.
commit 71f668aa7583354e132c20b3b50ba2c86bf08738
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Aug 31 13:59:44 2018 +0200
DHCP coop autoip bug fix: wait after last discovery before starting autoip
After the last dhcp discovery we need to give the dhcp server the
time to respond. Currently the discovery message is send and autoip
is started simultaniously. This is changed. Autoip will now be started
after the next discovery timeout.
commit 0da16604ec079195533f2591f0d0f04bdf212a72
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Aug 31 11:23:31 2018 +0200
ACD: rate limit interval bug solved -> first decline then wait then restart.
According the the ACD RFC we need to limit the rate of acquiring
and probing addresses after MAX_CONFLICTS. It is important to first
decline the address and stop the netif from using the address before
the time is started. After this rate limit time, the address acquiring
process can be started again.
To make this possible we had to change the callback function and
the location in the process were the rate limiting is done.
commit a89a0601a251acb14abe270116f38c6d25c2d7a9
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Aug 30 17:27:10 2018 +0200
DHCP: after a succesful dhcp_reboot, the address should be probed.
Reboot means that the connection was gone for some amount of time.
This is seen as a new connection for the ACD module so should be
the address should be probed before use.
commit 853afb448ba35c6e2b35e8238c9c367c599dece7
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Aug 30 17:19:12 2018 +0200
DHCP_DOES_ARP_CHECK changed to DHCP_DOES_ACD_CHECK
small update in opt.h to add the correct dependencies.
commit e28b4766bdef69e76f6170c470c93f5b251c579a
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Thu Aug 30 17:05:43 2018 +0200
DHCP check code replaced by ACD module.
when DHCP_DOES_ARP_CHECK is enabled, the ACD module will take care
of address conflict detection. Via a call back function the DHCP
state machine will continue and bind to an address if no conflicts
are found.
dhcp_arp_reply is obsolete because the ACD module replaces its
function.
commit 52193a0f5d13e8786a4db2fff1f1a8f1367a4eba
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Aug 27 16:05:22 2018 +0200
Issue 2 + 3 from Comment #5, task #13508 solved
Removed C++ comments and // ----... marks as requested.
commit 7faaf61275d67ccfb88ea7e26c249428c3088536
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Aug 27 15:55:24 2018 +0200
Issue 1 from Comment #5, task #13508 solved
Comments added to clarify how arp messages are handled.
commit 9348aea22623b705759fd30b873f06a50a104d16
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:45:35 2018 +0200
acd: add comment for callback function
commit 49fdd1177c05b74d49fa179564dcaa5e650adbcc
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:42:59 2018 +0200
autoip: Add debugging output and complete comments
commit 591856b82c029687a657a1b1ccc674522e6f4be0
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:25:18 2018 +0200
Removed autoip from timeouts.c
We do not need a timer anymore. All timing is part of acd.
commit 3b3272fdd14015cfca0b3b6d149505b1cf0e36b6
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:22:32 2018 +0200
autoip_network_changed -> first bring down the netif and then acd
Make it do what it says in the comment
commit 77b0ccf96efd22774279c6f9b5bade18c5e42c59
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:21:23 2018 +0200
Simplified the autoIP states + update autoip_start()
commit 96e0581d36857f8b70c4b4cce4fb323fd3dd51ab
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:10:48 2018 +0200
Deleted unused variables, functions and defines from autoip module
The functionality is now embedded in the acd module so can be
removed from autoip.
commit 9296e2ebb4b51019aaccfc47e8b9f51b265d37cd
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 24 11:07:34 2018 +0200
Bug fix acd -> rate limiting needs to be after MAX conflicts
It was only after MAX conflicts + 1.
commit c55e16903c045d0ea84336b50eccbf24d3d097e9
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 23 17:52:45 2018 +0200
Small intruduction to module added
commit d1498a37293bd9f97f6b938b48e5980ab0a01bd1
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 23 17:49:58 2018 +0200
Copyright information added
I based the copyright text on the other files. I'm not sure this
is how its done. I kept Dominik in their because I copied quite
some code from the autoip module.
commit 359a845ef5e73061832069f364b370634ee0b071
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 23 17:08:44 2018 +0200
Fixing comments in acd.c
commit 100d72549d0ef44157143d031848a727f5dfbe69
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 23 15:43:40 2018 +0200
Probe wait time and Probe interval time made random via lwip_rand
According to the RFC a random amount of time needs to be waited
before probing can be started also a random amount of time
needs to be waited between the probes.
The random time is calculated via the LWIP_RAND function (see
lwipopts and sys_arch).
commit f7f037c32e9416f8b803c3c7af617b871b55ee35
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 16:56:48 2018 +0200
autoip will start probing again when the network has changed
commit 1f40f6274195f24aa1b05caf82b79285ad189c2a
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 16:55:47 2018 +0200
Duplicate code removed -> already present in autoip_start
commit be59431271da862a8ca330dbca638842c87765bc
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 16:52:49 2018 +0200
add rate limiting to acd when too many conflicts are detected.
When more then max conflicts are detected during device on time,
the rate at which probing for a new address starts is limited.
For clarity we combined the conflict counting and the callback into
a new function acd_restart.
commit 91448455e95edb24c1f418c341b6fb306391f4f1
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 15:50:01 2018 +0200
ARP packet conflict detection added.
If a conflict is detected between the begin until after announce
wait, a new address needs to be chosen immediatly.
This can occur in two situations:
1) another host already has this ip address
2) another host is also probing for the same address.
If a conflict is detected during announcing or during the ongoing
conflict detection, we defend our ip address once. If a second
conflict occurs during defend interval, we take another ip address.
If not we can keep our address and connections.
When a conflict occurs and it's decided we need a new address,
autoip is simply restarted. To do: test if a acd stop is needed.
This will become more important when DHCP is added I believe.
commit 65f47ba9444d8b9f767dc908319579323eeb8664
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 15:33:44 2018 +0200
Update ACD state machine with PROBE_WAIT and and ANNOUNCE_WAIT
By adding the ANNOUNCE_WAIT state, we could remove duplicate code
that was present in the original autoip state implementation.
But because we cannot directly go to ANNOUNCING we needed this
extra state during the wait period. It also makes the different
states clearer.
Their is no need to number the enum because the compiler takes
care of this standard numbering.
Also a indent issue on the state machine code is solved here.
commit 2d9f4414c7b1f2ed35c0b5cea78dabb9c9afee77
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 15:28:33 2018 +0200
Added an extra check for etharp responses.
Their is no need to answer a request that is initiated by ourselves.
In my test case, a conflicting request would be seen here as a
request needing a reply. Which off course isn't needed at all.
The acd module will let the requester know that it is using our
ip address.
I could not think of a reason not to add this extra check but please
check if this doesn't break other functionality.
commit f84cc1dba4061219bd1aadb97bd340278db07cd7
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 15:27:13 2018 +0200
Redirect all incoming ARP messages to the ACD module.
The ACD module will scan the packets and react on conflicts.
autoip does not need the arp packets any more.
commit 9faf266993cc2df0b9434720b59b5922f17d7d33
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 10:08:39 2018 +0200
Initiate acd from autoip implementation.
The state machine in autoip is now replaced by the acd module.
commit 40a5a40d911c98e6ee8566c0adce7716f26f20e0
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 10:05:59 2018 +0200
Add simple acd state machine to timer function.
This state machine will be started from acd_start and will go
through the probing and the announcing state. At this moment the
timings are made fixed for easy debugging. Will be made variable
afterwards.
commit 0a2629bf1f7942e80ec11bcc3e163ac1a2b9a580
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 09:44:19 2018 +0200
Disable gratuitous ARP send from netif_issue_report when acd is enabled
The acd module needs to be fully under control on what arp messages are
send. It wouldn't be a clean solution if we kept this announce as one
of the announce messages.
So when ACD is enabled, this message will not be send.
commit a7b9a4d5039168723c2043677176c5c8fc69f8df
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 09:42:46 2018 +0200
ARP probe and announce functions added.
Both use etharp_requests to send out an arp message.
commit e2f1ce3fa84e2ce01bc944d8049412e156cd7cc7
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Tue Jul 17 09:26:07 2018 +0200
Add conflict callback mechanism to acd module
When starting the acd with acd_start, a conflict callback has to be
provided. After probing when no conflict is detected, the callback
is called with the conflict variable on 0. From the moment a conflict
is detected, it will be called set to 1.
By not making a fixed link between acd, autoip and dhcp but working
with callback functions the user (caller) can easily use it for
fixed ip too. This keeps the acd module completely independent from
the other application layer protocols.
commit 21e7995888d2941601e99873ee129d1ea927e3f7
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 16 14:41:52 2018 +0200
acd_stop function added.
When calling this function the acd state machine will be put in
ACD_STATE_OFF. This will disable the acd functionality until
acd_start is called.
commit 503037fb462504e46a14ce7486bc763dbc6a8690
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 16 14:35:48 2018 +0200
acd_start function added
This function will allocate a struct for the acd module if one wasn't
added with acd_set_struct. It will then initialize the acd struct
and start the probe wait timer (for now fixed).
commit e439f6dffc44e93078a2976783bdebfe17304d8c
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 16 14:22:39 2018 +0200
ACD_DEBUG define added for enabling / disabling debug messages.
commit 80d33e1eaf092934ace1045fac096464cd5be5e9
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 16 11:39:37 2018 +0200
ACD struct added together with set struct function
ACD needs some variables: state, timing, probe counter etc.
We added the struct as netif data. For now one ACD module per netif
is enough. We found some unclarities about ACD with multiple IP
address on one interface. For now, ACD will only be implemented for
the IP address that is going to be used / is being used on the netif.
commit 2c4cca36744973318c3efe7cbae6384b52dc71a8
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Mon Jul 16 11:35:06 2018 +0200
Add acd timer to timeouts.c with interval of 100ms.
This timer will be used for time to wait counting etc.
Very similar to the auto ip timer.
commit e2ed447e00c4df790df21509acb4ab09b5b79e66
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Jul 13 16:56:39 2018 +0200
Add ACD protocol definitions
commit 445733214460eae18817556439bed979e9b3747c
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Jul 13 16:07:58 2018 +0200
Add c++ option (to use C code module in c++)
commit 63d78bc1f77ac6698c3d663a1c67b40a0c297125
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Jul 13 15:57:34 2018 +0200
Added on off option for ACD module
commit d5ec4b69eb4175d4dd569bcfc80ae0e192780015
Author: Jasper Verschueren <jasper.verschueren@apart-audio.com>
Date: Fri Jul 13 15:09:40 2018 +0200
ACD IPv4 module files added
The files for IPv4 ACD or Address Conflict Detection are added
to the lwip source.
Reverts a regression introduced in
3a8af612b3b818a89de5846cc9b046756af184cc:
Use hardware address fetched from neighbor cache *not* the hardware
address of the interface as destination address.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
See bug #54655:
"MQTT brokers such as Google Cloud IoT Core requires MQTT clients
to send JSON Web Token (JWT) as password field of the MQTT Connect
packet. JWT can be more than 255 bytes.
Currently, the MQTT library restricts password to be less than 256
bytes, thus it prevents connectivity to Google Cloud IoT Core."
Fix that by just converting the local variables for these from u8_t
to u16_t.
Suggested-by: Richmond Umagat <richmond.umagat@brtchip.com>
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
TLS clients that need 2-way authentication (e.g. Amazon AWS IoT cloud mqtt)
need to pass a certificate and private key when creating the tls altcp_pcb.
Added a new function altcp_tls_create_config_client_2wayauth() for this that
replaces altcp_tls_create_config_client() for such clients.
See bug #54601.
not needed any more since commit b0344518e8
"sockets.c: Don't rely on #defines in socket.h to be in sync with ip.h
Map SO_* to SOF_* #defines"
Reported by Joan Lledó on mailinglist 6.Sep.2018
magic_init() is first time called from ppp_init(), which is called from
lwip_init(). If system has no RTC, sys_jiffies() returns same value in
this moment after every power-up or system reset. This value used in LCP
magic number generation after ppp_connect(), which leads to same magic
number after every restart. Subsequent magic_randomize() calls takes
place in ppp_input(), after magic number generation.
Call magic_randomize() somewhere near start of ppp_connect() (and
ppp_listen()) as it might be called later at a random time.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
When the network is changed, dhcp is rebooted.
It will send a dhcp request again to verify it's lease.
DHCP requests are send out in selecting state, rebinding, renewing
and rebooting. But in the rebooting state the hostname option is
not included. This means that after reboot, the hostname will be
unknown to the DNS.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
This fixes an issue seen on MacOS with Clang 9.0.0:
../../../../lwip/src/core/ipv4/etharp.c:1069:142: error: format specifies type 'unsigned short' but the argument has type 'netif_addr_idx_t'
(aka 'unsigned char') [-Werror,-Wformat]
LWIP_DEBUGF(ETHARP_DEBUG | LWIP_DBG_TRACE, ("etharp_query: dropped previously queued packet %p for ARP entry %"U16_F"\n", (void *)q, i));
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~
../../../../lwip/src/core/ipv4/etharp.c:1074:120: error: format specifies type 'unsigned short' but the argument has type 'netif_addr_idx_t'
(aka 'unsigned char') [-Werror,-Wformat]
LWIP_DEBUGF(ETHARP_DEBUG | LWIP_DBG_TRACE, ("etharp_query: queued packet %p on ARP entry %"U16_F"\n", (void *)q, i));
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~
By default, netif_addr_idx_t is u8_t, so cast up to u16_t to match the format type
e.g. if the calling stack should not invoke the mdns functions due
to high stack usage, disable the option and trigger it yourself.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
Also remove additonal (cname == '\0' || cname == '.') check because
it's covered by !lwip_isupper(cname) test.
Signed-off-by: Axel Lin <axel.lin@ingics.com>
If a chained pbuf starts with DHCP_OPTION_PAD, an overflow check
triggers and the packet is ignored.
Fix this by changing the way the offset is increased for PAD.
Also ignore a packet that is missing the END option.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
Can be combined into a single check using LWIP_MIN.
See bug #54194
Reported-by: Andrey Vinogradov <andrey.vinogradov@teplomonitor.ru>
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
Fix below build error:
In file included from
../../../../lwip/src/../test/unit/tcp/test_tcp.c:3:0:
../../../../lwip/src/../test/unit/tcp/test_tcp.c: In function
‘test_tcp_rto_timeout_syn_sent_impl’:
../../../../lwip/src/../test/unit/tcp/test_tcp.c:1246:113: error: suggest parentheses around arithmetic in operand of ‘|’ [-Werror=parentheses]
const u16_t tcp_syn_opts_len = LWIP_TCP_OPT_LENGTH(TF_SEG_OPTS_MSS|TF_SEG_OPTS_WND_SCALE|TF_SEG_OPTS_SACK_PERM|TF_SEG_OPTS_TS);
^
../../../../lwip/src/include/lwip/priv/tcp_priv.h:305:4: note: in definition of macro ‘LWIP_TCP_OPT_LENGTH’
(flags & TF_SEG_OPTS_MSS ? LWIP_TCP_OPT_LEN_MSS : 0) + \
^~~~~
Signed-off-by: Axel Lin <axel.lin@ingics.com>
Fail to build if PPP_SUPPORT is disabled because PPP_NUM_TIMEOUTS
references defines that are unknown if PPP_SUPPORT == 0. Move
PPP_NUM_TIMEOUTS to PPP_SUPPORT block and set a default value to 0
if PPP_NUM_TIMEOUTS is still not defined at the end of ppp_opts.h.
Fixes: e4e0fc4193 ("PPP: set PPP_NUM_TIMEOUTS_PER_PCB value to computed number of simultaneous timers")
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
A deep analysis of simultaneously running timers showed we can use a
slightly smaller value for PPP_NUM_TIMEOUTS_PER_PCB, value which was
arbitrarily chosen to a safe value based on the number of enabled PPP
features. Add the boring and long analysis to the end of our internal
header file and use the result in ppp_opts.h.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
Only the transmission side works for now (todo: dual and tradeoff mode).
Also, only a 10 second timeout mode is implemented for now.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
Managed to find the spirit behind the RFC. Looks like we need to send
a ZLB packet with counters as is to the packet (ZLB or not) we
previously sent to ack the message. Luckily we don't need more than
received NS/NR counters to forge the resent ack.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
We never use the current peer NS value but always the next expected peer
NS value (current value plus one).
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
Conform a little bit more to the RFC by dropping duplicate received
packets. It might help discover Ns sequence issues.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
It can't be an issue since we only send and receive a few L2TP control
packets and we don't care about anything received next other than
sending Ack packet. For the sake of correctness properly handle Ns/Nr
counters wraparounds, it doesn't add more code anyway.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
L2TP tunnel is really UP only once we received the ICCN Ack packet,
generally with a ZLB packet. We actually drop all data packets until
we reach the "data" L2TP state, which is a sane thing to do, therefore
it makes no sense at all to start the PPP session a bit earlier.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
Our Ns counter is the current slot rather than the next to ease packet
retransmission. Therefore we increment the Ns counter before using the
next slot instead of after. The RFC is written with post-increment in
mind rather than pre-increment, thus when the RFC says that Ns in not
incremented for ZLB packets it actually means that ZLB packets are sent
with the next Ns without post-incrementing the Ns value, meaning the
ZLB packet does not take a slot. Since we are using a pre-incremented
value for real slots we need to send ZLB packets with our current Ns
value plus one.
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
lwip/src/netif/ppp/pppoe.c: In function ‘pppoe_send_padt’:
lwip/src/netif/ppp/pppoe.c:1048:108: error: ‘sc’ undeclared (first use in this function)
sc is not passed to ‘pppoe_send_padt’ function because it might be
called to terminate unknown sessions.
Fixes: d4047ea1d1 ("Try to fix issues reported by coverity")
Signed-off-by: Sylvain Rochet <gradator@gradator.net>
The new NULL checking against netif_input_fn make ppp stop working
because in ppp_new() it calls netif_add with NULL netif_input_fn.
Signed-off-by: Axel Lin <axel.lin@ingics.com>
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
This partly reverts 31bc2f9b which broke pppos tx as 'pppos_output_append'
implements a custom scheme of checking available pbuf size.
Added a comment why PBUF_POOL is ok for tx in this special case.
See bug #51908
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
Fix the case where nd6_send_rs() fails: send one solicitation here,
not LWIP_ND6_MAX_MULTICAST_SOLICIT.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
... and add a unit test for it.
Moved resetting netif->rs_count from all reports to link-up and netif-up only.
While at it, clean up the interface a bit so that netif->rs_count is touched
from nd6.c only.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
arp table functions, nd6 destination cache and struct netif_hint
can now be u16_t or u8_t depending on table size, so up to 32K
entries can be used in these tables (s16_t)
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
This should hide the internal type used for access to the ARP table
which currently is s8_t or u8_t, depending on the use case.
Signed-off-by: Simon Goldschmidt <goldsimon@gmx.de>
This optimizes the netif_loop_output to only schedule a call to poll when
the first packet is enqueued. This ensures netif_poll is ran once per
burst of packets that are sent (which is typical in a TCP transfer)
The old behavior scheduled a call to poll for every packet that was
enqueued and this lead to exhaustion of the MEMP_TCPIP_MSG_API memory pool
and tcpip_mbox (if port is using static mbox size). The extra callbacks are
wasted work because netif_poll drains the entire queue when ran
This issue presented itself when large TCP transfer go across the loopback
netif
This disable checksum generation and checking for the loopback netif
when LWIP_CHECKSUM_CTRL_PER_NETIF is enabled
Checksums are not needed for the loopback adapter and this will increase
performance for loopback communication
This timeout is used to measure TFTP_TIMEOUT_MSECS fine enough.
Calling tftp_tmr at a 50ms interval to handle a 1 seconds timeout
produces way too much cpu load (and prevents sleep).
Signed-off-by: goldsimon <goldsimon@gmx.de>
Gcc complains that an array index is a 'char' when passing a 'char' to
toupper(). Quieten this by coercing to an 'unsigned char'.
Signed-off-by: goldsimon <goldsimon@gmx.de>
