Basically:
* "classification by-ip" (i.e. `flow->guessed_protocol_id_by_ip` is
NEVER returned in the protocol stack (i.e.
`flow->detected_protocol_stack[]`);
* if the application is interested into such information, it can access
`ndpi_protocol->protocol_by_ip` itself.
There are mainly 4 points in the code that set the "classification
by-ip" in the protocol stack: the generic `ndpi_set_detected_protocol()`/
`ndpi_detection_giveup()` functions and the HTTP/STUN dissectors.
In the unit tests output, a print about `ndpi_protocol->protocol_by_ip`
has been added for each flow: the huge diff of this commit is mainly due
to that.
Strictly speaking, this change is NOT an API/ABI breakage, but there are
important differences in the classification results. For examples:
* TLS flows without the initial handshake (or without a matching
SNI/certificate) are simply classified as `TLS`;
* similar for HTTP or QUIC flows;
* DNS flows without a matching request domain are simply classified as
`DNS`; we don't have `DNS/Google` anymore just because the server is
8.8.8.8 (that was an outrageous behaviour...);
* flows previusoly classified only "by-ip" are now classified as
`NDPI_PROTOCOL_UNKNOWN`.
See #1425 for other examples of why adding the "classification by-ip" in
the protocol stack is a bad idea.
Please, note that IPV6 is not supported :( (long standing issue in nDPI) i.e.
`ndpi_protocol->protocol_by_ip` wil be always `NDPI_PROTOCOL_UNKNOWN` for
IPv6 flows.
Define `NDPI_CONFIDENCE_MATCH_BY_IP` has been removed.
Close#1687
Add (basic) internal stats to the main data structures used by the
library; they might be usefull to check how effective these structures
are.
Add an option to `ndpiReader` to dump them; enabled by default in the
unit tests.
This new option enables/disables dumping of "num dissectors calls"
values, too (see b4cb14ec).
Since e6b332aa, we have proper support for detecting client/server
direction. So Tinc dissector is now able to properly initialize the
cache entry only when needed and not anymore at the SYN time; initializing
that entry for **every** SYN packets was a complete waste of resources.
Since 4896dabb, the various `struct ndpi_call_function_struct`
structures are not more separate objects and therefore comparing them
using only their pointers is bogus: this bug was triggered by this
change because `ndpi_str->callback_buffer_size_tcp_no_payload` is now 0.
In a lot of places in ndPI we use *packet* source/dest info
(address/port/direction) when we are interested in *flow* client/server
info, instead.
Add basic logic to autodetect this kind of information.
nDPI doesn't perform any "flow management" itself but this task is
delegated to the external application. It is then likely that the
application might provide more reliable hints about flow
client/server direction and about the TCP handshake presence: in that case,
these information might be (optionally) passed to the library, disabling
the internal "autodetect" logic.
These new fields have been used in some LRU caches and in the "guessing"
algorithm.
It is quite likely that some other code needs to be updated.
As a general rule, the higher the confidence value, the higher the
"reliability/precision" of the classification.
In other words, this new field provides an hint about "how" the flow
classification has been obtained.
For example, the application may want to ignore classification "by-port"
(they are not real DPI classifications, after all) or give a second
glance at flows classified via LRU caches (because of false positives).
Setting only one value for the confidence field is a bit tricky: more
work is probably needed in the next future to tweak/fix/improve the logic.
The goal is to have a (roughly) idea about how many packets nDPI needs
to properly classify a flow.
Log this information (and guessed flows number too) during unit tests,
to keep track of improvements/regressions across commits.
