clang was issuing some warnings related to unused return in function.
This patch adds some needed error treatment and ignore the rest of the
warnings by adding a cast to void.
Major redesign of the flow engine. Remove the flow queues that turned
out to be major choke points when using many threads. Flow manager now
walks the hash table directly. Simplify the way we get a new flow in
case of emergency.
Checksum of local traffic is often offloaded to the network device.
This causes some problems on parsing of this traffic. This patch
introduces a PKT_INCOMPLETE_CHECKSUM flag which can be used to
indicate that the checksum is not computed/correct for good reason.
Add counter to the stream reassembly engine to count stream gaps. Stream gaps
are the result of missing packets (usually due to packet loss). This missing
data stops the reassembly for the app layer.
Removal of per flow 'aldata' array. It contained a ptr for each ALPROTO. Instead now we have 2 ptrs in the flow: alparser and alstate.
Various cleanups and dead code removal from the app layer API.
Should safe 100+ bytes memory per flow on 64 bit.
Updated lots of unittests to reflect these changes.
Added stream counters:
- tcp.reassembly_memuse -- current memory use by reassembly in bytes
- tcp.memuse -- current memory use by stream tracking in bytes
- tcp.reused_ssn -- ssn reused by new session with identical tuple
- tcp.no_flow -- TCP packets with no flow - indicating flow engine memory at its limits
This patch introduces a function called StreamMsgForEach which
can be used to run a callback on all segments of a stream. This
is currently only supported for TCP as this is the only streaming
aware protocol.
This patch renames DECODER_SET_EVENT, DECODER_ISSET_EVENT and some
other structures to ENGINE equivalent to take into account the fact
the event list is now related to all engines and not only to decoder.
Per packet per app layer parser profiling. Example summary output:
Per App layer parser stats:
App Layer IP ver Proto cnt min max avg
-------------------- ------ ----- ------ ------ ---------- -------
ALPROTO_HTTP IPv4 6 163394 126 38560320 42814
ALPROTO_FTP IPv4 6 644 117 26100 2566
ALPROTO_TLS IPv4 6 670 117 7137 799
ALPROTO_SMB IPv4 6 114794 126 225270 957
ALPROTO_DCERPC IPv4 6 5207 126 25596 1266
Also added to the csv out.
In the csv out there is a new column "stream (no app)" that removes the
app layer parsers from the stream tracking. So raw stream engine performance
becomes visible.
Per packet profiling uses tick based accounting. It has 2 outputs, a summary
and a csv file that contains per packet stats.
Stats per packet include:
1) total ticks spent
2) ticks spent per individual thread module
3) "threading overhead" which is simply calculated by subtracting (2) of (1).
A number of changes were made to integrate the new code in a clean way:
a number of generic enums are now placed in tm-threads-common.h so we can
include them from any part of the engine.
Code depends on --enable-profiling just like the rule profiling code.
New yaml parameters:
profiling:
# packet profiling
packets:
# Profiling can be disabled here, but it will still have a
# performance impact if compiled in.
enabled: yes
filename: packet_stats.log
append: yes
# per packet csv output
csv:
# Output can be disabled here, but it will still have a
# performance impact if compiled in.
enabled: no
filename: packet_stats.csv
Example output of summary stats:
IP ver Proto cnt min max avg
------ ----- ------ ------ ---------- -------
IPv4 6 19436 11448 5404365 32993
IPv4 256 4 11511 49968 30575
Per Thread module stats:
Thread Module IP ver Proto cnt min max avg
------------------------ ------ ----- ------ ------ ---------- -------
TMM_DECODEPCAPFILE IPv4 6 19434 1242 47889 1770
TMM_DETECT IPv4 6 19436 1107 137241 1504
TMM_ALERTFASTLOG IPv4 6 19436 90 1323 155
TMM_ALERTUNIFIED2ALERT IPv4 6 19436 108 1359 138
TMM_ALERTDEBUGLOG IPv4 6 19436 90 1134 154
TMM_LOGHTTPLOG IPv4 6 19436 414 5392089 7944
TMM_STREAMTCP IPv4 6 19434 828 1299159 19438
The proto 256 is a counter for handling of pseudo/tunnel packets.
Example output of csv:
pcap_cnt,ipver,ipproto,total,TMM_DECODENFQ,TMM_VERDICTNFQ,TMM_RECEIVENFQ,TMM_RECEIVEPCAP,TMM_RECEIVEPCAPFILE,TMM_DECODEPCAP,TMM_DECODEPCAPFILE,TMM_RECEIVEPFRING,TMM_DECODEPFRING,TMM_DETECT,TMM_ALERTFASTLOG,TMM_ALERTFASTLOG4,TMM_ALERTFASTLOG6,TMM_ALERTUNIFIEDLOG,TMM_ALERTUNIFIEDALERT,TMM_ALERTUNIFIED2ALERT,TMM_ALERTPRELUDE,TMM_ALERTDEBUGLOG,TMM_ALERTSYSLOG,TMM_LOGDROPLOG,TMM_ALERTSYSLOG4,TMM_ALERTSYSLOG6,TMM_RESPONDREJECT,TMM_LOGHTTPLOG,TMM_LOGHTTPLOG4,TMM_LOGHTTPLOG6,TMM_PCAPLOG,TMM_STREAMTCP,TMM_DECODEIPFW,TMM_VERDICTIPFW,TMM_RECEIVEIPFW,TMM_RECEIVEERFFILE,TMM_DECODEERFFILE,TMM_RECEIVEERFDAG,TMM_DECODEERFDAG,threading
1,4,6,172008,0,0,0,0,0,0,47889,0,0,48582,1323,0,0,0,0,1359,0,1134,0,0,0,0,0,8028,0,0,0,49356,0,0,0,0,0,0,0,14337
First line of the file contains labels.
2 example gnuplot scripts added to plot the data.
Split stream reassembly in 2 parts: a part that sends ack'd data to the app
layer parsers as soon as it's available, and another part that queues up
data into larger chunks for raw inspection.
For convenience, a massive usage of 'Packet p;' declaration has
been done in the tests function. Although this was completely
legal, this is not possible anymore because of the new Packet
allocation structure. This massive patch modifies all suricata
files to use a SCMalloc allocated pointer to Packet instead.
This patch has been done using coccinelle (http://coccinelle.lip6.fr)
which is a semantic patching tool. This ensures that things like call
to SCFree() should have not been forget because the semantic patch
explicitly forces the call to SCFree(p) before each return. With this
patch all unittests are running fine with a small and a big default
packet size.
This patch implements the needed modification of payload access
in a Packet structure to support the abstraction introduced by
the extended data system.
The stream engine memory handling needed updating as it didn't scale. Changes:
- pools can now be initialized to size 0, meaning unlimited
- stream engine uses a memcap setting. Sessions, segments and aldata is part
of this, app layer state isn't.
- memory is accounted using a global int that is spinlocked.
- a counter for sessions that have not been picked up because of memcap was
added.
- all reassembly errors are converted to debug msgs.
- improve locking of application layer handling, making sure that the flow cannot be freed/cleared when the detection engine is still working with it.
- add a check to the app layer detection to make sure that a match function will only inspect an app layer state if it's of the right type.