Christian Herdtweck [Fri, 20 Mar 2015 11:02:01 +0000]
update Readme, TODO and example pingcheck.conf
Christian Herdtweck [Fri, 20 Mar 2015 10:12:15 +0000]
fixed trouble in unittests: when I changed/added options I forgot to adjust corresponding unittests
Christian Herdtweck [Fri, 20 Mar 2015 08:38:19 +0000]
fixed same bug as for DestinationUnreachable in TimeExceeded messages: accept reply from mid-way
altered unit test accordingly
removed unnecessary implementations of match_xy functions that also just return false
Christian Herdtweck [Thu, 19 Mar 2015 17:30:00 +0000]
fixed bug in match_destination_unreachable: message does not have to come from original destination
that means we have to interpret the ip header within the icmp data, which is done by converting
IcmpData::raw_data to a stream which is fed to IpvXHeader.read
Also means that address to test for has to be forwarded to IcmpDestinationUnreachableData
and cannot be simply tested in IcmpData already --> change function prototype in all IcmpData classes
Christian Herdtweck [Thu, 19 Mar 2015 17:26:17 +0000]
added more tests, fixed trouble with position of test data relative to build dir
Christian Herdtweck [Thu, 19 Mar 2015 17:25:18 +0000]
can now interpret pcaps from tcpdump; created own pcap.h and pcap.cpp because is needed in 3 places now:
* icmp/icmppacketfactory
* tools/feed_packet_data
* tests/test_icmppacket
pcaps from tcpdump have network_type = 1, so include a interface struct in packet data before actual data.
Our own dumps were network_type=101 and did not include that interface info
Christian Herdtweck [Thu, 19 Mar 2015 17:16:05 +0000]
added more testing pcap data
Christian Herdtweck [Thu, 19 Mar 2015 09:13:11 +0000]
created data directory in test directory for pcap files for IcmpPacket tests
Christian Herdtweck [Thu, 19 Mar 2015 09:06:36 +0000]
added revision to version string (0.3 --> 0.3.1) to more easily see if testing the correct version (e.g. if patching tarballs)
also added build date to output of pingcheck --version
Christian Herdtweck [Wed, 18 Mar 2015 18:06:32 +0000]
added more getter to IcmpPacket so can test more
Christian Herdtweck [Wed, 18 Mar 2015 18:05:48 +0000]
added new unittest for IcmpPacket
Christian Herdtweck [Wed, 18 Mar 2015 15:01:32 +0000]
fixed bugs that caused DestinationUnreachable / TimeExceeded messages to NEVER match anything
also added function to get ttl of original request IP packet but unfortunately that always returns 1 if time exceeded :-(
Also improved error reporting in ICMP packet parsing process
Christian Herdtweck [Wed, 18 Mar 2015 14:57:22 +0000]
added option to print raw message data in text form for debugging
Christian Herdtweck [Wed, 18 Mar 2015 10:46:13 +0000]
added code to parse ICMP type TimeExceeded
* format is identical to DestinationUnreachable and other ICMP replies
--> created a common half-abstract base class IcmpData_PingFailReply
* added a match_time_exceeded to IcmpPacket which is called by IcmpPinger
Christian Herdtweck [Wed, 18 Mar 2015 10:45:07 +0000]
in case of unknown ICMP reply type, consume data instead of returning failure in IcmpPacket::read
Christian Herdtweck [Wed, 18 Mar 2015 10:38:22 +0000]
added ip source and destination address to IpHeader::to_string
Christian Herdtweck [Fri, 13 Mar 2015 14:18:52 +0000]
increased version to 0.3 since have now new ICMP packet structure and
packet feeder tool and will create new tar from this
Christian Herdtweck [Fri, 13 Mar 2015 14:15:17 +0000]
added a simple text output (to_string) for DestinationUnreachable ICMP data
Christian Herdtweck [Fri, 13 Mar 2015 14:13:03 +0000]
added a count of good/bad packets to feed_packet_data tool
Christian Herdtweck [Fri, 13 Mar 2015 13:32:15 +0000]
added rule to create feed_packet tool to CMakeLists.txt
Christian Herdtweck [Thu, 5 Mar 2015 10:15:38 +0000]
check if file is really open after opening; changed return codes and log levels
Christian Herdtweck [Wed, 4 Mar 2015 16:29:07 +0000]
got the tool feed_packet_data to compile and run and do reasonable stuff
Christian Herdtweck [Wed, 4 Mar 2015 16:28:20 +0000]
added a dump_option to IcmpPacketFactory creation function; better error message if temp file creation fails
Christian Herdtweck [Fri, 27 Feb 2015 17:06:37 +0000]
started creating a utility program in new folder tools that feeds packet data directly into ICMP/TCP factories
Christian Herdtweck [Fri, 13 Mar 2015 10:49:26 +0000]
moved remaining implementations in ipv6header.h to cpp; avoid conversion warnings
Christian Herdtweck [Fri, 13 Mar 2015 09:09:36 +0000]
give credit to Christopher Kohlhoff (boost asio) and Guilherme Ferreira
Christian Herdtweck [Fri, 13 Mar 2015 09:02:19 +0000]
consequently moved all implementation from h to cpp files
Christian Herdtweck [Thu, 12 Mar 2015 17:02:30 +0000]
fixed bug in reading of checksum (which was found using the unittests -- great thing!)
Christian Herdtweck [Thu, 12 Mar 2015 16:44:45 +0000]
updated icmp unittests
Christian Herdtweck [Thu, 12 Mar 2015 16:44:08 +0000]
moved some implementations from h to cpp files because of linker trouble
Christian Herdtweck [Thu, 12 Mar 2015 14:51:58 +0000]
now working again: creating echo requests and parsing replies
also:
* created a to_string method for easier output
* in constructors of IP headers set version right away
* compare that version to the one received in IcmpPacket::read
Christian Herdtweck [Thu, 12 Mar 2015 10:04:43 +0000]
re-ordered class for IcmpPacket
old: IcmpPacket has and IpvXHeader, an IcmpvXHeader which varies according to ICMP message type and also contains part of Icmp message data, and an IcmpData which is a string;
new: IcmpPacket has an IpHeader, an IcmpHeader which represents only the common first 4 byte of ICMP v4/v6, and IcmpData which is either a IcmpEchoData for echo request/reply or a IcmpDestinationUnreachableData
tried to abstract away as much as possible (for a lay programmer like me) the difference between v4 and v6
This code now compiles, need to see whether it does anything usefull
Tests not compiled/run yet
Christian Herdtweck [Thu, 12 Mar 2015 09:55:24 +0000]
added encode_string to MessagePayload; removed checksum and icmpchecksum since functionality moved into MessagePayload, IcmpHeader and IcmpData
Christian Herdtweck [Thu, 12 Mar 2015 09:53:22 +0000]
created abstract base class IpHeader of Ipv4Header and Ipv6Header
Christian Herdtweck [Thu, 26 Feb 2015 15:48:08 +0000]
added debug option enforcing dump of ICMP packets even if not broken
Christian Herdtweck [Wed, 25 Feb 2015 15:00:21 +0000]
fixed bug: dumping packets leaked char array (dump file name) on heap
Christian Herdtweck [Wed, 25 Feb 2015 14:34:16 +0000]
dumping now into a .pcap file with nicer filename, so can analyze with wireshark
Christian Herdtweck [Wed, 25 Feb 2015 12:54:08 +0000]
dumping package data works
Christian Herdtweck [Fri, 20 Feb 2015 17:36:50 +0000]
started copying icmp data before parsing it and dumping it into temp files if parsing fails
does compile but causes problems in data parsing
Christian Herdtweck [Thu, 19 Feb 2015 08:32:44 +0000]
make check for end of loop in icmppacketdistributor more understandable (|= --> =)
Thomas Jarosch [Wed, 11 Feb 2015 16:55:28 +0000]
Check weak_ptr before using it
Thomas Jarosch [Wed, 11 Feb 2015 16:48:16 +0000]
No need for intermediate weak_ptr
Thomas Jarosch [Wed, 11 Feb 2015 16:47:08 +0000]
Use const reference for smart_ptr in BOOST_FOREACH
Thomas Jarosch [Wed, 11 Feb 2015 16:23:45 +0000]
Move graph documentation to doc/
Christian Herdtweck [Wed, 11 Feb 2015 15:59:22 +0000]
fix bug that caused crash for broken packets; change order of vars to avoid compiler warnings; minor change in debug output
Christian Herdtweck [Wed, 11 Feb 2015 14:13:08 +0000]
updated TODO
Christian Herdtweck [Wed, 11 Feb 2015 13:58:37 +0000]
removed another debugging test and corrected one debuggin output
Christian Herdtweck [Thu, 5 Feb 2015 16:28:21 +0000]
created a graph to show how IcmpPinger, IcmpPacketDistributor and other important classes interact
created using yEd 3.12.2
Christian Herdtweck [Thu, 5 Feb 2015 16:17:54 +0000]
give up trying to define IcmpPacketDistributor in own files, just use icmppinger.h / .cpp for it
The two classes depend on each other, so none compiles without the other
if they are defined in separate files. In the same file, the compiler
manages.
Christian Herdtweck [Thu, 5 Feb 2015 15:22:14 +0000]
got packetdistributor to work, but compiles/links only with packetdistributor inside icmppinger.h/cpp
Christian Herdtweck [Wed, 4 Feb 2015 11:02:11 +0000]
limit line length of new files to 80 characters
Christian Herdtweck [Wed, 4 Feb 2015 11:01:43 +0000]
use new file name (icmppaKetdistributor --> icmppacCKetdistributor) in includes and macros
Christian Herdtweck [Mon, 26 Jan 2015 16:37:38 +0000]
also rename file in CMakeLists.txt
Christian Herdtweck [Mon, 26 Jan 2015 16:34:11 +0000]
rename IcmpPaketDistributor files to IcmpPacketDistributor (k --> ck)
Christian Herdtweck [Mon, 26 Jan 2015 16:32:01 +0000]
start removing socket use from IcmpPinger to only have IcmpPacketDistributor deal with raw data
Christian Herdtweck [Mon, 26 Jan 2015 14:20:26 +0000]
clean up IcmpPaketDistributors after stopping pingers
Christian Herdtweck [Mon, 26 Jan 2015 14:17:32 +0000]
Pinger holds weak_ptr to self, Pingers created through static create functions; IcmpPinger::create registers with IcmpPaketDistributor
Christian Herdtweck [Fri, 23 Jan 2015 17:07:37 +0000]
started to implement an icmp paket distributor; compiles and works at first, but gives segmentation fault at exit
Christian Herdtweck [Thu, 22 Jan 2015 17:17:13 +0000]
added a test for pingcheck version to avoid trouble with IP parsing in 'older' pingcheckers
Christian Herdtweck [Thu, 22 Jan 2015 16:43:36 +0000]
another amendment to signal handler: make signaled_flags_XY volatile sig_atomic_t (I was ignorant)
Christian Herdtweck [Thu, 22 Jan 2015 16:21:24 +0000]
added saving of history and results and function to load and print stats about them
Christian Herdtweck [Thu, 22 Jan 2015 16:20:39 +0000]
added a little failsave test into tester; simplified logging of exceptions; recognize unexpected online states and warn; removed pre-logging Tester._print
Christian Herdtweck [Thu, 22 Jan 2015 16:15:00 +0000]
had forgotten to reset default pingcheck binary from testing to production
Christian Herdtweck [Thu, 22 Jan 2015 14:27:57 +0000]
amendment to last commit: only makes sense if use different flags per signal type
Christian Herdtweck [Thu, 22 Jan 2015 14:09:46 +0000]
removed logging from signal handler; create one handler per signal type
Christian Herdtweck [Thu, 22 Jan 2015 14:03:19 +0000]
changed return type of get_configuration in main: return configuration and bool flag
this way, always get a valid configuration object with default values and some values set
even if parsing other options failed. Important for --version where usually fail to read
config file but do not need it anyway
Christian Herdtweck [Thu, 22 Jan 2015 14:00:12 +0000]
changed treatment of --version arg; remember to print it in Configuration like --daemon option
Christian Herdtweck [Thu, 22 Jan 2015 13:58:43 +0000]
corrected type of TTL: is now uint32_t everywhere (except for uint8_t in Payload data)
Christian Herdtweck [Thu, 22 Jan 2015 09:12:47 +0000]
removed Readme.code because was neither up to date nor informative; updated TODO
Christian Herdtweck [Thu, 22 Jan 2015 08:33:26 +0000]
raised version number from 0.1 to 0.2
Christian Herdtweck [Wed, 21 Jan 2015 17:04:06 +0000]
remember pings based on host, not IP; correct logging logic in main loop
for host-ip binding, needed to slightly change output from dnsresolver
Christian Herdtweck [Wed, 21 Jan 2015 12:52:36 +0000]
treat output as "left-over" only after terminatin process; limit number of lines
Christian Herdtweck [Wed, 21 Jan 2015 12:51:16 +0000]
removed assert that does not make sense and is often violated
(IPs are not updated very often, so can easily happen that an IP's TTL was reached.
It is easier to ignore these cases than to avoid them, IPs will with next test be updated.)
Christian Herdtweck [Wed, 21 Jan 2015 10:56:05 +0000]
had forgotten to disable testing setting in pingcheck main: limited the max number of exceptions ignore
Christian Herdtweck [Wed, 21 Jan 2015 10:43:35 +0000]
done with long_term_test.py
* set timeout for ping reply -- very important variable!
* try to get more output when pingcheck process is stopped/dying
* added and commented out sigchld / child signal handler
* different logic how state changes are set
* use ConndState
* lots of other changes...
Christian Herdtweck [Wed, 21 Jan 2015 10:37:48 +0000]
improved tell-connd state representation
* use constants for online states and subsystems
* add function to output complete state
* deal with exceptions from running binary
Christian Herdtweck [Wed, 21 Jan 2015 10:33:29 +0000]
Raised importance of log output in signal handler: info for most, notice for term/kill
Christian Herdtweck [Fri, 16 Jan 2015 17:01:13 +0000]
use logging for all output, improve understanding of connd, improved process termination, simplify code
Christian Herdtweck [Fri, 16 Jan 2015 08:58:09 +0000]
some minor doc change I had overlooked
Christian Herdtweck [Thu, 15 Jan 2015 17:37:20 +0000]
added a test python script that goes online/offline and checks if pingers pinged enough
Christian Herdtweck [Mon, 12 Jan 2015 15:54:44 +0000]
renamed IcmpPinger::handle_ping_done to handle_timeout; handle error_codes from timer async_wait
Christian Herdtweck [Mon, 12 Jan 2015 14:45:16 +0000]
some whitespace change I had forgotten
Christian Herdtweck [Mon, 12 Jan 2015 14:44:52 +0000]
added case for unexpected signals in signal handler
Christian Herdtweck [Mon, 12 Jan 2015 14:44:15 +0000]
make error return codes from main unique again
Christian Herdtweck [Mon, 12 Jan 2015 14:43:38 +0000]
introduce a max number of exceptions ignored in io_service main loop
mainly for testing and debugging
Christian Herdtweck [Mon, 12 Jan 2015 14:42:33 +0000]
properly initialize variables remembering original signal handlers; re-set only if were set
Christian Herdtweck [Mon, 12 Jan 2015 14:41:04 +0000]
continue waiting for data if receive handler received an error code
2 reasons:
* in other error cases do the same
* there is a timer running in the background which will send the next ping after some time
Christian Herdtweck [Mon, 12 Jan 2015 14:37:13 +0000]
do not wait for IcmpEchoReply if sending of IcmpEchoRequest has failed completely
Christian Herdtweck [Mon, 12 Jan 2015 14:35:48 +0000]
update docu messages for io_service
Christian Herdtweck [Mon, 12 Jan 2015 14:33:48 +0000]
Clean up output, whitespace,
* there were several cases of GlobalLogger.error() << "Error: ..." which is redundant
* there was one instance of streaming to cerr instead of GlobalLogger.error
* remove trailing whitespace
Christian Herdtweck [Tue, 30 Dec 2014 17:14:41 +0000]
minor change in logging: avoid duplicated "Error: "
Christian Herdtweck [Tue, 30 Dec 2014 17:14:06 +0000]
new signal handling in main; removed all thread-related vars, use same IoServiceItem instead
Christian Herdtweck [Tue, 23 Dec 2014 17:42:23 +0000]
temp commit before leaving, started to change signal handling and use singel io_service
Christian Herdtweck [Tue, 23 Dec 2014 13:29:42 +0000]
hotfix for problem with 0-packages
* added flag ReceiveHandlerInPlace to icmppinger
* to be sure, also check if reply was received already in handle_receive_icmp_packet
* in handle_receive_icmp_packet also check error code
* in icmp package read, print strings for failure codes
* prepend IP to pinger log output so can trace which thread did what
Christian Herdtweck [Mon, 15 Dec 2014 14:30:51 +0000]
changed logic of icmp package read/write and operator>> / operator<<
* read and write are implemented on their own, >> and << call them
(not the other way round as before)
* read returns not only a bool but a more informative flag
* moved operator>> and << from Icmpv4/6packet to base class IcmpPacket
since they are identical
Christian Herdtweck [Wed, 10 Dec 2014 14:58:52 +0000]
fixed parsing of icmp data: only parses right amount of data
Christian Herdtweck [Tue, 9 Dec 2014 17:52:00 +0000]
fixed possible reason for broken icmp packages: restricted read of icmp payload data; also moved def of buffer size from h to cpp
Christian Herdtweck [Tue, 9 Dec 2014 17:50:58 +0000]
fixed possible reason for broken icmp packages: restricted read of icmp payload data; also moved def of buffer size from h to cpp
Christian Herdtweck [Tue, 9 Dec 2014 15:02:36 +0000]
replace magic number 64k with constant SOCKET_BUFFER_SIZE
Christian Herdtweck [Tue, 9 Dec 2014 15:01:09 +0000]
improved logging output (once again)