- Do not reply with a reset message to a reset or an ACK message
- Reply with a reset message when not able to process a CON/NON message
(not even a suitable error reply)
This adds the new `nanocoap_server_observe` module that implements the
server side of the CoAP Observe option. It does require cooperation
from the resource handler to work, though.
Co-Authored-By: mguetschow <mikolai.guetschow@tu-dresden.de>
Co-authored-by: benpicco <benpicco@googlemail.com>
This allows sending a separate response with CoAP Options and adds a
helper to detect duplicate requests, so that resource handlers can
repeat their empty ACK on duplicates.
Calling `coap_get_token()` and `coap_get_token_length()` on an
(mostly) uninitialized `coap_pkt_t` did work so far due to
implementation details matching the expectations, but this is not
backed up by any API contract.
This fixes the API abuse by introducing and using a new API that does
read a token and token length from a CoAP over UDP packet out of a
buffer. This now provides the behavior expected by the caller and
commits to it via API contract.
Co-authored-by: mguetschow <mikolai.guetschow@tu-dresden.de>
Co-authored-by: benpicco <benpicco@googlemail.com>
This reverts commit e3d00682bcad2e0f26ee15b1c73d5da022b18786, which
added a work around for two bugs:
- ztimer triggering too early (fixed in
https://github.com/RIOT-OS/RIOT/pull/20924)
- gnrc_sock_recv() returning when an old "timeout" message is still
in the message queue (fixed in
https://github.com/RIOT-OS/RIOT/pull/21113)
With those bugs fixed, the work around should not longer be needed.
When module `nanocoap_server_separate` is not used, the functions to
send separate responses are still provided, just in a broken version:
They will send the separate replies from a different endpoint than the
request was received at (even on machines with only one IP address, as
also the source port is randomized).
This changes the behavior to only provide the functions for separate
response when the do work, so that others will detect an invalid
configuration at compile time rather than at run time.
The documentation is duly updated.
An RST message has no token, so don't reply with a token when sending
RST.
This also adds unit tests to ensure this this exact bug does not sneak
back in.
When RFC 8974 support (module `nanocoap_token_ext`) is in use, the
request token may be longer than the buffer in the separate response
context is large. This adds a check to not overflow the buffer.
Sadly, this is an API change: Preparing the separate response context
can actually fail, so we need to report this with a return value.
The example application has been adapted to only proceed if the separate
reply context could have been prepared, and rather directly emit a
reset message if the token exceeds the static buffer.
Co-authored-by: benpicco <benpicco@googlemail.com>
gnrc_tcp_open() previously would eventually fail with a timeout without
sending any data when no netif was specified and a link-local target
address was used. This fixes the behavior:
- If there is only one netif, we just pick that
- If there are multiple netifs, fail directly with `-EINVAL` rather than
sending out nothing and waiting for a timeout.
Co-authored-by: benpicco <benpicco@googlemail.com>
Before, handlers writing blockwise transfer assumed that the response
header length will match the request header length. This is true for
UDP, but not for TCP: The CoAP over TCP header contains a Len field,
that gets extended for larger messages. Since the reply often is indeed
larger than the request, this is indeed often the case for CoAP over
TCP.
Note: Right now, no CoAP over TCP implementation is upstream. However,
getting rid of incorrect assumptions now will make life easier
later on.
In case no payload is added, `coap_build_reply_header()` would return
`sizeof(coap_hdr_t) + token_length` regardless of the actual header
length returned by `coap_build_hdr()`. These can be different if
RFC 8974 extended tokens are enabled (module `nanocoap_token_ext`
used): If an extended token length field is used, its size is not
considered.
Co-authored-by: benpicco <benpicco@googlemail.com>
gcoap contains a hack where a `coap_pkt_t` is pulled out of thin air,
parts of the members are left uninitialized and a function is called on
that mostly uninitialized data while crossing fingers hard that the
result will be correct. (With the current implementation of the used
function this hack does actually work.)
Estimated level of insanity: 😱😱😱😱😱
This adds to insane functions to get the length of a token and the
length of a header of a CoAP packet while crossing fingers hard that
the packet is valid and that the functions do not overread.
Estimated level of insanity: 😱😱😱
The newly introduced insane functions are used to replace the old
insane hack, resulting in an estimated reduction of insanity of 😱😱.
Side note: This actually does fix a bug, as the old code did not take
into account the length of the extended TKL field in case of
RFC 8974 being used. But that is a bug in the abused API,
and not in the caller abusing the API.
