At every fuzz iteration (i.e for every trace file):
* keep the same ndpi context (`ndpi_init_detection_module` is very
slow);
* reset the flow table, otherwise it grows indefinitely.
This change should fix the "out-of-memory" errors reported by oss-fuzz.
Remove the last uses of `struct ndpi_id_struct`.
That code is not really used and it has not been updated for a very long
time: see #1279 for details.
Correlation among flows is achieved via LRU caches.
This change allows to further reduce memory consumption (see also
91bb77a8).
At nDPI 4.0 (more precisly, at a6b10cf, because memory stats
were wrong until that commit):
```
nDPI Memory statistics:
nDPI Memory (once): 221.15 KB
Flow Memory (per flow): 2.94 KB
```
Now:
```
nDPI Memory statistics:
nDPI Memory (once): 235.27 KB
Flow Memory (per flow): 688 B <--------
```
i.e. memory usage per flow has been reduced by 77%.
Close#1279
This allows to avoid triggering alerts for trusted albeit private certificate issuers.
Extended the example/protos.txt with the new syntax for specifying trusted issueDN.
Example:
trusted_issuer_dn:"CN=813845657003339838, O=Code42, OU=TEST, ST=MN, C=US"
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.
Remove some unused fields and re-organize other ones.
In particular:
* Update the parameters of `ndpi_ssl_version2str()` function
* Zattoo, Thunder: these timestamps aren't really used.
* Ftp/mail: these protocols are dissected only over TCP.
* Attention must be paid to TLS.Bittorrent flows to avoid invalid
read/write to `flow->protos.bittorrent.hash` field.
This is the last(?) commit of a long series (see 22241a1d, 227e586e,
730c2360, a8ffcd8b) aiming to reduce library memory consumption.
Before, at nDPI 4.0 (more precisly, at a6b10cf7, because memory stats
were wrong until that commit):
```
nDPI Memory statistics:
nDPI Memory (once): 221.15 KB
Flow Memory (per flow): 2.94 KB
```
Now:
```
nDPI Memory statistics:
nDPI Memory (once): 231.71 KB
Flow Memory (per flow): 1008 B <---------
```
i.e. memory usage per flow has been reduced by 66%, dropping below the
psychological threshold of 1 KB.
To further reduce this value, we probably need to look into #1279:
let's fight this battle another day.
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
`ndpiReader` is only an example, aiming to show nDPI capabilities
and integration, without any claim about performances.
Nonetheless its memory usage per flow is *huge*, limiting the kinds
of traces that we can test on a "normal" hardware (example: scan
attacks).
The key reason of that behaviour is that we preallocate all the memory
needed for *all* the available features.
Try to reduce memory usage simply allocating some structures only
when they are really needed. Most significant example: JOY algorithms.
This way we should use a lot less memory in the two most common
user-cases:
* `ndpiReader` invoked without any particular flag (i.e `ndpiReader -i
$FILENAME_OR_IFACE`)
* internal unit tests
Before (on x86_64):
```
struct ndpi_flow_info {
[...]
/* size: 7320, cachelines: 115, members: 72 */
```
After:
```
struct ndpi_flow_info {
[...]
/* size: 2128, cachelines: 34, members: 75 */
```
This protocol is detected via HTTP Content-Type header.
Until 89d548f9, nDPI had a dedicated automa (`content_automa`) to
classify a HTTP flow according to this header. Since then, this automa has
been useless because it is always empty.
Re-enable it to match only a string seems overkilling.
Remove all `content_automa` leftovers.
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
* Added ARM build and unit test run for SonarCloud-CI.
Signed-off-by: Toni Uhlig <matzeton@googlemail.com>
* Fixed Mingw64 build.
* adapted to SonarCloud-CI workflow
* removed broken and incomplete Windows example (tested on VS2017/VS2019)
* removed unnecessary include (e.g. pthread.h for the library which does not make use of it)
Signed-off-by: Toni Uhlig <matzeton@googlemail.com>
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.
Simplified the process of adding lines to AC_AUTOMATA_t.
Use the ndpi_string_to_automa() function to add patterns with domain names.
For other cases can use ndpi_add_string_value_to_automa().
ac_automata_feature(ac_automa, AC_FEATURE_LC) allows adding
and compare data in a case insensitive manner. For mandatory pattern comparison
from the end of the line, the "ac_pattern.rep.at_end=1" flag is used.
This eliminated unnecessary conversions to lowercase and adding "$" for
end-of-line matching in domain name patterns.
ac_match_handler() has been renamed ac_domain_match_handler() and has been greatly simplified.
ac_domain_match_handler() looks for the template with the highest domain level.
For special cases it is possible to manually specify the domain level.
Added test for checking ambiguous domain names like:
- short.weixin.qq.com is QQ, not Wechat
- instagram.faae1-1.fna.fbcdn.net is Instagram, not Facebook
If you specify a NULL handler when creating the AC_AUTOMATA_t structure,
then a pattern with the maximum length that satisfies the search conditions will be found
(exact match, from the beginning of the string, from the end of the string, or a substring).
Added debugging for ac_automata_search.
To do this, you need to enable debugging globally using ac_automata_enable_debug(1) and
enable debugging in the AC_AUTOMATA_t structure using ac_automata_name("name", AC_FEATURE_DEBUG).
The search will display "name" and a list of matching patterns.
Running "AHO_DEBUG=1 ndpiReader ..." will show the lines that were searched for templates
and which templates were found.
The ac_automata_dump() prototype has been changed. Now it outputs data to a file.
If it is specified as NULL, then the output will be directed to stdout.
If you need to get data as a string, then use open_memstream().
Added the ability to run individual tests via the do.sh script
Non-critical bugs.
If a file list is used, then all files except the last are not closed.
Opening the next file loses the memory allocated via pcap_open_offline() for the previous file.
If a bpf filter is used, then no memory is freed after pcap_compile.
