Reported by oss-fuzz:
```
==685288==ERROR: AddressSanitizer: SEGV on unknown address 0x61a100000687 (pc 0x0000005aba64 bp 0x7ffe3f29f510 sp 0x7ffe3f29f400 T0)
==685288==The signal is caused by a READ memory access.
SCARINESS: 20 (wild-addr-read)
#0 0x5aba64 in quic_len ndpi/src/lib/protocols/quic.c:203:12
#1 0x5aba64 in decrypt_initial_packet ndpi/src/lib/protocols/quic.c:993:16
#2 0x5aba64 in get_clear_payload ndpi/src/lib/protocols/quic.c:1302:21
#3 0x5aba64 in ndpi_search_quic ndpi/src/lib/protocols/quic.c:1658:19
#4 0x579f00 in check_ndpi_detection_func ndpi/src/lib/ndpi_main.c:4683:6
#5 0x57abe6 in ndpi_check_flow_func ndpi/src/lib/ndpi_main.c:0
#6 0x583b2c in ndpi_detection_process_packet ndpi/src/lib/ndpi_main.c:5545:15
#7 0x55e75e in LLVMFuzzerTestOneInput ndpi/fuzz/fuzz_process_packet.c:30:3
[...]
```
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.
There are no valid reasons for a (generic) protocol to ignore IPv6
traffic.
Note that:
* I have not found the specifications of "CheckPoint High Availability
Protocol", so I don't know how/if it supports IPv6
* all LRU caches are still IPv4 only
Even if src_id/dst_id stuff is probably useless (see #1279), the right
way to update the protocol classification is via `ndpi_set_detected_protocol()`
Looking at `struct ndpi_flow_struct` the two bigger fields are
`host_server_name[240]` (mainly for HTTP hostnames and DNS domains) and
`protos.tls_quic.client_requested_server_name[256]`
(for TLS/QUIC SNIs).
This commit aims to reduce `struct ndpi_flow_struct` size, according to
two simple observations:
1) maximum one of these two fields is used for each flow. So it seems safe
to merge them;
2) even if hostnames/SNIs might be very long, in practice they are rarely
longer than a fews tens of bytes. So, using a (single) large buffer is a
waste of memory for all kinds of flows. If we need to truncate the name,
we keep the *last* characters, easing domain matching.
Analyzing some real traffic, it seems safe to assume that the vast
majority of hostnames/SNIs is shorter than 80 bytes.
Hostnames/SNIs are always converted to lowercase.
Attention was given so as to be sure that unit-tests outputs are not
affected by this change.
Because of a bug, TLS/QUIC SNI were always truncated to 64 bytes (the
*first* 64 ones): as a consequence, there were some "Suspicious DGA
domain name" and "TLS Certificate Mismatch" false positives.
We can write to `flow->protos` only after a proper classification.
This issue has been found in Kerberos, DHCP, HTTP, STUN, IMO, FTP,
SMTP, IMAP and POP code.
There are two kinds of fixes:
* write to `flow->protos` only if a final protocol has been detected
* move protocol state out of `flow->protos`
The hard part is to find, for each protocol, the right tradeoff between
memory usage and code complexity.
Handle Kerberos like DNS: if we find a request, we set the protocol
and an extra callback to further parsing the reply.
For all the other protocols, move the state out of `flow->protos`. This
is an issue only for the FTP/MAIL stuff.
Add DHCP Class Identification value to the output of ndpiReader and to
the Jason serialization.
Extend code coverage of fuzz tests.
Close#1343Close#1342
There are no real reasons to embed `struct ndpi_packet_struct` (i.e. "packet")
in `struct ndpi_flow_struct` (i.e. "flow"). In other words, we can avoid
saving dissection information of "current packet" into the "flow" state,
i.e. in the flow management table.
The nDPI detection module processes only one packet at the time, so it is
safe to save packet dissection information in `struct ndpi_detection_module_struct`,
reusing always the same "packet" instance and saving a huge amount of memory.
Bottom line: we need only one copy of "packet" (for detection module),
not one for each "flow".
It is not clear how/why "packet" ended up in "flow" in the first place.
It has been there since the beginning of the GIT history, but in the original
OpenDPI code `struct ipoque_packet_struct` was embedded in
`struct ipoque_detection_module_struct`, i.e. there was the same exact
situation this commit wants to achieve.
Most of the changes in this PR are some boilerplate to update something
like "flow->packet" into something like "module->packet" throughout the code.
Some attention has been paid to update `ndpi_init_packet()` since we need
to reset some "packet" fields before starting to process another packet.
There has been one important change, though, in ndpi_detection_giveup().
Nothing changed for the applications/users, but this function can't access
"packet" anymore.
The reason is that this function can be called "asynchronously" with respect
to the data processing, i.e in context where there is no valid notion of
"current packet"; for example ndpiReader calls it after having processed all
the traffic, iterating the entire session table.
Mining LRU stuff seems a bit odd (even before this patch): probably we need
to rethink it, as a follow-up.
Fix all the warnings.
Getting rid of "-Wno-unused-parameter" is quite complex because some
parameters usage depends on compilation variable (i.e.
`--enable-debug-messages`).
The "-Werror" flag has been added only in Travis builds to avoid
breaking the builds to users using uncommon/untested
OS/compiler/enviroment.
Tested on:
* x86_64; Ubuntu 20.04; gcc 7,8,9,10,11; clang 7,8,9,10,11,12
* x86_64; CentOS 7.7; gcc 4.8.5 (with "--disable-gcrypt" flag)
* Raspberry 4; Debian 10.10; gcc 8.3.0
* Add support for Snapchat voip calls
Snapchat multiplexes some of its audio/video real time traffic with QUIC
sessions. The peculiarity of these sessions is that they are Q046 and
don't have any SNI.
* Fix tests with libgcrypt disabled
* QUIC: fix return value on error path on quic_cipher_init()
* QUIC: allow dissection of sessions forcing version negotiation
Enhance heuristic to avoid false positives.
The "offset" field is a variable-length integer.
This bug hasn't any practical effects right now, since we are ignoring any
packet with "offset" != 0 (and the value 0 is always encoded in only one byte).
But extracting a correct "offset" is important if we are ever going to handle
fragmented Client Hello messages.
LGTM found a real issue on a boundary check
Fix unit tests: a pcap ha been uploaded twice (with different names)
Fix compilation when using DPDK (see #990)
Add QUIC payload and header decryption: most of the crypto code has been
"copied-and-incolled" from Wireshark. That code has been clearly marked
as such. All credits for that code should go to the original authors.
I tried to keep the Wireshark code as similar as possible to the original,
comments included, to ease future backporting of fixes.
Inevitably, glibc data types and data structures, tvbuff abstraction and
allocation functions have been converted.