Parent Directory
| 
Revision Log
| 
Revision Graph
| Links to HEAD: | (view) (download) (as text) (annotate) |
| Sticky Tag: |
Sources WoCified: see http://www.freaknet.org/alpt/src/utils/woc/readme
* rem.c:
rtt32_t rem_rtt_8to32(rtt8_t x)
rtt8_t rem_rtt_32to8(rtt32_t y)
bw32_t rem_bw_8to32(bw8_t x)
bw8_t rem_bw_32to8(bw32_t y)
* bandwidth_in_8bit() deprecated
* CITE.h added
* The QSPN v2 article has been completed. * The ANDNA_EXPIRATION_TIME has been incremented from 3days to 30days * The RTT_DELTA has been lowered from 1000ms to 258ms * Minor bugs fixed in andns.c, request.c * The main document has been updated. It now refers to the QSPN v2 too * The Ntk_bandwidth_measurement is being rewritten
pkts.c and all the related coed has been converted to the new API.
pkts.c now uses the new modular API of request.c. Almost all the sources have been updated.
andna.c now uses rq_del_request()
* the Request API and its documentation is now completed, we just need to convert pkts.c and some other sources with this new API. * accept.[ch] was removed because it has been never used, and now, with the API change it is obsolete.
* TODO updated * continuing in request.c
rand.c written
module.c completed: the NTK API for modules is ready, now it's time to modularize the rest of the code.
Yeahah. It works \o/ It was a stupid bug in pkt_compress(). -_' 0.0.9b coming soon
parse_options() fixed: it didn't handle well "-r 1"
ghgrgrh, this little socket optimization is taking more time than required...
cleanings
It almost works, but it hangs on get_ext_maps()
* An active tcp socket is kept for each rnode, in this way, it is possible to send them tcp packets without the need of estabilishing a new tcp connection each time. * The qspn packets are now sent using the tcp sockets described above. * Since the UDP isn't used anymore for ntk packets, the udp daemon has been removed. The udp port of the radar daemon is now 269/udp. The 271/udp port has been deprecated.
All gcc4 compilation warnings removed!
releasing...
small things
hash_md5 description in andna.h
first draft of netsukuku.conf.pod
* The adding of snsd records in the rhc is now handled in a better and optimized way. * snsd_*_merge() functions written
* it seems it is enough stable now. * global queries still don't work. Dunno why.
memory is cleaner and shines under the Sun.
added references to http://bugs.dyne.org
* If no SNSD record is present, fallback to the mainip * RSA memory leaks fixed in andna.c
Yeah, rh_cache bug fixed
* finally the gcc bug workaround on put_qspn_round() gccbug: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27945 * new macros: bufget(), bufput()
*** empty log message ***
Bugs you dieeeeeeeeeee!
* -l logfile option added * "rc.ntk cclean" cleans /usr/share/netsukuku * conf.c cleans the environment now
Once upon a time there was the Bug, the Coder came with his mighty Vim. With a quick movement of his fingers, '<=' became '<' and the head of the Bug was severed. All the gdbies, lived happy and segfree.
The amount of time passed debugging to find a specific bug is inversely proportional to the lines of code that need to be modified to fix it. l = 1/t; This means that if you pass an infinite amount of time debugging, you'll just need to change one bit in your code.
debugging...
WTH did I disabled andna_resolvconf_modify() in !restricted mode ? oiasdjhasod
resolv.conf is modified if -R is not specified AND if we are in restricted mode
- resolv.conf is modified if -R is not specified
remember not what changed
- load_maps and andna_load_caches use the file_exist() check, this prevents the stupid error message "file %s doesn't exist". - code cleaning: return (EXPR) ? 1 : 0; is the same thing of return (EXPR); while return (EXPR) ? 0 : 1; is the same thing of return !(EXPR);
* Check for active ntkd processes. Use a pid file in /var/run/ * various fixes
- functions added: * str_to_snsd_service() * snsd_proto_to_str() * snsd_service_to_str() - PACKET compression now works - code cleaned - test results: * snsd service registration WORKS * andna reverse ip resolution WORKS * andna normal resolution WORKS * andna snsd resolution UNTESTED
Compressed pkts coded ;)
snsd_cache.c has been completely tested. Time to test the real stuff in andna.c
Debugging snsd_cache.c: * stest.c has been written. It tests all the functions of snsd_cache.c * In this way, when it'll work completely, we are sure that * snsd_cache.c is ok. * Current status is 40-50%.
Now we can distinguish if a andna_cache is empty or malformed
load_snsd() tested and working
added ntk-dig.8 in man/SConstruct and man/makefile.am
ntkd compiled! Time to test the goodies
include errors solved
Building andns_snsd. andns_init() now takes a family argument. andna.c updated.
andna_resolve_hash() ready
andna_recv_rev_resolve_rq() and andna_reverse_resolve() converted to SNSD. Yay! Is really SNSD completed? ^_^
andna_resolve_hname() and andna_recv_resolve() are now completely converted to SNSD. llist.c: welcome the new list_copy_some() function ;)
Converting andna_resolve_hname() to SNSD
andna_register_hname() and andna_recv_reg_rq() converted to SNSD.
The code of the snsd requests has begun!
the snsd coding continues ;)
rh_cache and lcl_cache converted to SNSD
andna_cache converted to SNSD
snsd struct stuff finished
The MX code previously written has been removed. The coding of the SNSD has begun.
Andns library rewritten. andnslib works on andns protocolo (NTK_RFC 0006). dnslib works on dns packet andns works on dns_wrapper. MX now works for inet resolutions.
andna_mx_resolve() <-- this is the function which the ANDNS wrapper shall use
andna_cache.c: * the integration of the MX stuff in the rh_cache is complete * the calls to *_del_expired() have been optimized (expecially for the rh_cache)
llist.c: * it has been almost completely rewritten * list_add has now a different meaning: it adds the struct at the start of the llist, in this way it hasn't to traverse the entire llist. * a lot of new functions have been added: * list_append, list_swap, list_moveback, list_movefwd, list_moveontop, clist_append * LLIST_HDR() macro written andna: * The ANDNA keyring is now saved separately from the lcl_cache. It is saved in andna_lcl_keyring. * The `andna_' prefix is now used for the filename of the andna caches. * The struct of the resolved_hostnames_cache is changed, we just save the 32bit hash of the hname. still working on rh_cache....
0.0.7b Released
set_tos_sk();
tunnels
Branch "Released" updated to the current 0.0.6b version.
YEahaahahaha!!!! Now it works!!! Yuppi, yeahah
Nooooooooo! Another stupid ugly damned bug.
Finally I've found this damned bug
little fixes
ANDNS_SKEY_LEN was a fixed value but a packed rsa privkey isn't always of the same size. This has been fixed. Does the same happen for the pubkey?
Endian fixes in andna: it isn't necessary to convert the MD5 hash to network order
endianness fix in get_single_acache()
debug
cleaned some signedness warning
The NTK_RFC 0008 has been implemented (not tested).
- The NTK_RFC 0007 has been implemented. It fixes a bug in the ANDNA protocol. - Added a timeout of 60 seconds for the reverse resolution request.
* Fixed an heap overflow in andna.c * Consider also LOOPBACK address to reverse resolve in andns_rslv.c
All the possible situation regarding the manual editing of /etc/resolv.conf are now considered. The user cannot mess anymore it ;)
* Fixed a bug in the parser of netsukuku.conf * loginfo -> debug in andns*.c * When restoring resolv.conf.bak don't include the "nameserver 127.0.0.1" string in resolv.conf. * Code cleaned
andna_reverse_resolve returns only currently registered hnames. andna_reverse_resolve handles the request to resolve the same IP of the node (which is doing the request).
Comment the other nameservers when we put "nameserver 127.0.0.1" in /etc/resolv.conf
The ANDNS/ANDNA reverse resolution works!
Little errors fixed in recv_reverse_resolve()
little improvements
Is andns ok? Test, test, test
clean the opaque cristal.
* andna.c: do not receive a flood we started * andns_rslv.c: another endianness fix, andna_resolve_hname() returns host order ip. * andns_pkt.c: use xstrndup() instead of str=xmalloc(len); strncpy(str, from, len);
Pointer bug in andns fixed
Fixed a double xfree() bug in andns_rslv.c Andns integrated in dns_wrapper.c
Andns compiles.
andna bug fixed: there were too many op_set_filter() for the andna requests.
* igw_order() optimized. * #define FAMILY_LVLS GET_LEVELS(rpkt.from.family)
The 0001 still doesn't work perfectly. - join_rate system improved - op_filter() written: it is used to filter/drop incoming requests or replies. - Invalid groups, i.e 234 in level 3, now are not chosen: is_group_invalid(). - code cleaned
merge_ext_maps() has been written.
The test of the coded part of the 0001 has begun.
Main part of the NTK_RFC 0001 finished (gotta test it now).
In some andna pkts there is a `time_t timestamp' var, before this commit the time was stored in the time() format but this was wrong since the time can differ from one node to another, so the difference of seconds between the current time and time_t must be used, moreover for the andna pkts it isn't necessary to count the rtt between the two nodes since it is very small compared to the ANDNA_EXPIRATION_TIME.
Coding of the RFC 0001 continues: - Now the number of nodes present in a gnode is counted. At hooking this number is given to the hooking node which updates its counter. For this reason the qspn_round_pkt is changed, so the 0.0.5 is no more compatible with the previous versions. - The rehook() function has been written. It will be used to rehook. - Bug 0001 solved (see //lab.dyne.org/Netsukuku_bugs) - The command line options override the ones specified in the config file.
TODO updated
accept.c was disabled. There was a really stupid error in dns_wrapper.c, which maybe is the cause of all those random crashes!
hname_update_too_early fix
-R option added: it prevents the modification of /etc/resolv.conf. It can be set from netsukuku.conf too.
/etc/resolv.conf is now modified automatically. When the daemon is launched /etc/resolv.conf is modified automatically, so the "nameserver 127.0.0.1" is added in the first line of resolv.conf. The old resolv.conf is at /etc/resolv.conf.bak. When the daemon is closed /etc/resolv.conf.bak is moved to /etc/resolv.conf.
Compilation warnings annihilated
bug in andna_resolve_hname()
Small changes: - Fixed andna_check_counter() in andna.c. - Now when a node dies a qspn is sent. This is useful to keep the maps updated. Until now it was disabled.
things are getting more stable. Bug fixed: - the multi-if code didn't set right all the routes. krnl_route.c needed a change in add_nexthop() - get_gw_ips() overwrote the stack if single_gw was used. - some crashes in the andna.c code. - the initialization of the int_info struct for the get_qspn_round request was malformed. - qspn_b_add did a wrong memset. - the /usr/share/netsukuku directory wasn't created during the installation. - Other stuff...
The multi-interfaces code seems to work. Now you are welcomed to test the CVS version. Before the release of the 0.0.4b other tests must be made, so one or two weeks may pass.
Yeah! The multipath support has been enabled for the external map, this means that more than one route will be used to reach a gnode. In other words the load balacing has been fully enabled. Some tests has been done, it seems to work: 207.0.0.0/8 proto 15 nexthop via 209.71.250.30 dev br0 weight 255 nexthop via 124.137.206.193 dev br0 weight 254 Enjoy ^_^ ( The multi network interface is next! hehehe )
The code for the endianness support is now complete, but not all the tests has been made. All the the packet are now sent without any alignment. The free_nodes struct has been finally optimised: to keep the list of free nodes it uses single bits instead of single bytes, this means that the list is now of 32 bytes instead of 256. The bug in the load_rh_cache code is now fixed. The installation path of the man page is now fixed. _THIS CVS VERSION HAS STILL PROBLEMS, DO NOT USE IT_
From _NOW_ the source code of the CVS is not stable anymore. The code to handle the endianness in the int/short variables of the packets is being written. The main code to handle this is in src/endianness.[ch].
Netsukuku site (re)added in the CVS. Some GPL legal stuff fixed.
The DNS wrapper was completed, tested and it works. YEA. Thanks to Cyberz.
Getting ready for publication. Count to 3 ;)
All the ANDNA tests finished and working! The next commit ANDNA will become beta.
Hostname resolve request and reverse resolve request testend and working. Andna routing tested and worked. Hash_gnode mutation almost finished. 71% of the english traslation done.
Counter_node tested and working. English translation continued. Code cleaned.
Andna hostname registration tested and working (without counter_node for now ;)
The code for the configuration file was written. Other little fixes.
Andna improvements. Document cleaned. English traslation continued.
The ANDNA code is now complete! Joy my friends o/ Yes, and it compiles too. Time to become made testing it ^_.
Correzioni varie del documento. Aggiunta di andna.[ch]
This form allows you to request diffs between any two revisions of this file. For each of the two "sides" of the diff, select a symbolic revision name using the selection box, or choose 'Use Text Field' and enter a numeric revision.
| alpt (at) freaknet (dot) org | ViewVC Help |
| Powered by ViewVC 1.1-dev |