Commit Graph

222 Commits

Author SHA1 Message Date
gardners
a23612428b make rhizome over mdp block size configurable.
1K seems too big to be reliable on our prototype UHF radio links,
as CRCs fail almost all the time. #48
2013-02-20 16:18:57 +10:30
gardners
2252fdcaa7 created tool for automatically finding IN()s without matching
OUT()s or where return() is used instead of RETURN().
Added OUT() to end of all functions using IN() that lacked it to
make it easier to statically analyse this invariant.
Fixed several return instead of RETURNs detected through use of
this tool. #49
2013-02-20 16:18:56 +10:30
gardners
641d749ab4 Create files for large rhizome bundles
- configurable size threashold
2013-02-20 15:36:32 +10:30
gardners
0ac403717d improve rhizome mdp re-request timeout handling so that it doesn't
do amazingly bad things on low-bandwidth packet radio links.
2013-02-20 15:36:31 +10:30
gardners
34bf9127a1 make rhizome transfer timeouts consistent. 2013-02-20 15:36:31 +10:30
gardners
f465f3a9ee Add lots of profiling to assist in debugging
- made some debug output conditional.
- make it possible to debug slip decoding without getting swamped by other output.
- Improve reporting of unnamed schedule() calls.
- always collect performance statistics, and report on any excess
  use (>1sec in every 3), even if debug.timing is not enabled.
- include file size information in Rhizome transfer messages.
- reduce output when debug.rhizome_rx is set.
2013-02-20 15:36:30 +10:30
gardners
e551017896 push profiling down deeper into rhizome_fetch_poll()'s children. 2013-02-20 15:36:29 +10:30
gardners
17ea1b69b3 don't look at all BARs if we are too slow. 2013-02-20 15:36:29 +10:30
gardners
f114e6d19e add progress data for each rhizome transfer. 2013-02-20 15:36:28 +10:30
gardners
3bd4c49af5 add more info to /rssi http page. 2013-02-20 15:36:27 +10:30
Jeremy Lakeman
ca5eb7d8bb Reinstate simple slip code, improve debug messages 2013-02-20 15:36:27 +10:30
Jeremy Lakeman
e4d6520c46 Don't request manifests that are already in the fetch queue 2013-02-20 15:36:25 +10:30
Jeremy Lakeman
6d9bbe2e2c Refactor interface handling to separate encapsulation from stream type 2013-02-20 15:36:23 +10:30
Jeremy Lakeman
ba1800012e Only advertise BAR's on an alarm
- we no longer hit the database for every outgoing packet, attempting to announce bundles
- we no longer advertise manifests periodically
- when an interesting bar arrives, we ask for the manifest to be announced, which uses the existing packet format
2013-02-20 15:36:23 +10:30
Jeremy Lakeman
84ad4debfa Throttle serial radio devices when tx buffer is in use 2013-02-20 15:36:22 +10:30
Jeremy Lakeman
2967d1e00f Don't check for duplicates if an existing id has been supplied 2013-01-03 11:18:30 +10:30
Jeremy Lakeman
a492c05d64 Re-add database cleanup 2012-12-31 15:09:12 +10:30
Jeremy Lakeman
3f45623d04 New stress test, rhizome add while transferring 2012-12-27 16:51:31 +10:30
Jeremy Lakeman
0f81c1f641 Minor code cleanup 2012-12-14 17:07:28 +10:30
Jeremy Lakeman
4264dec978 Reduce logging 2012-12-12 13:18:19 +10:30
Jeremy Lakeman
1a26a3e452 Reduce log spam 2012-12-12 10:02:43 +10:30
Andrew Bettison
5985df751d Overhaul debug flags
Replace debugflags_t and DEBUG_XXX bit masks with config schema "debug.xxx"
entries.

No more support for "debug.all".
2012-12-11 15:59:46 +10:30
Jeremy Lakeman
4b73e42bed Merge remote-tracking branch 'origin/master' into rhizomeovermdp
Conflicts:
	rhizome_fetch.c
2012-12-06 13:27:04 +10:30
Jeremy Lakeman
59babd75a2 Fix SEGV if connect fails immediately 2012-12-06 10:35:20 +10:30
gardners
e8cfac1462 don't try mdp fall back when performing an http rhizome direct transfer. 2012-12-06 09:42:27 +10:30
gardners
e2cf0ff506 fixed bug with rhizome direct sync manifest fetching. 2012-12-06 09:31:41 +10:30
gardners
5416e4baca fixed bugs that were preventing writing to database blob from
buffer.  Increased maximum buffer size to useful size.
2012-12-04 17:26:47 +10:30
gardners
39f99da3f2 fixing various bugs with rhizome streaming to database. 2012-12-04 17:18:31 +10:30
Jeremy Lakeman
6a5efcc445 Remove buggy debug code 2012-12-04 16:50:27 +10:30
gardners
f827c7c1c8 fixed various problems following split of files into fileblobs 2012-12-04 16:47:45 +10:30
Jeremy Lakeman
33e44ed6be Don't unwatch if the file descriptor is already closed 2012-12-04 16:30:01 +10:30
gardners
6dfe654ab0 adjusted tuning of Rhizome over MDP to use 32x1KB packets for
better throughput.  Still sub-optimal for wired ethernet type
interfaces, but should be ok for WiFi where max broadcast throughput
is 100kb/sec.
2012-12-04 16:19:22 +10:30
Jeremy Lakeman
bd83c192bb Use consistent length constant 2012-12-04 16:18:51 +10:30
gardners
fc0c134cbf added speed indication on completion of rhizome transfer.
split files table into files and fileblobs to avoid pathological
delays when updating datavalid flag in files table (sqlite copies
the whole table row, including possibly large blob).
2012-12-04 16:03:56 +10:30
gardners
23cace8455 removed writing temporary file for debugging. 2012-12-04 13:49:47 +10:30
gardners
2e4e111764 fixed bug that prevented remaining rhizome tests from passing. 2012-12-04 08:50:00 +10:30
gardners
a7e09aac2c fixed scheduling bug that stopped rhizomeprotocol tests 1 and 3. 2012-12-04 08:30:58 +10:30
gardners
254b0a4fd6 fixed problem with rhizome over http (alarm without callback
specified).
2012-12-04 07:12:03 +10:30
gardners
fcd5142aaf fixed bug with streaming to database of rhizome fetches. 2012-12-04 06:56:10 +10:30
gardners
abcac03981 work on tracking down hash calculation bug for files > one
window long.
2012-12-04 06:26:44 +10:30
gardners
43316ade3a added buffer when streaming to database so that we don't waste
most of our time opening and closing blobs, and don't have to keep
the blob open all the time.
2012-12-03 21:47:02 +10:30
gardners
de7a4ce2e3 fixed bugs with rhizome streaming direct to database.
that now works, but rhizome direct pull now fails.
2012-12-03 20:51:14 +10:30
gardners
e108fe0822 fixed bug in handling of sqlite3_blob_write(). Still more to fix. 2012-12-03 17:10:39 +10:30
gardners
5d24ba7db9 first cut at progressive hashing and checking of hash when streaming
file into database via rhizome.
2012-12-03 17:05:12 +10:30
gardners
e08593b466 stream manifest to buffer, and files to blob in database instead
of via files.
NOT COMPLETE - needs to calculate progressive hash and decide
if file is ok at end of import.
2012-12-03 16:52:08 +10:30
gardners
7ec695c940 feedback from code review by Jeremy.
Some preparation for out-of-order handling.
Consolidated redundantly redundant fields in fetch slot structure.
Probably fixed spurious alarm bug.
2012-12-03 11:50:35 +10:30
gardners
115736f0a4 cleaned out excess debugging messages and upped block size from
200 bytes to 1KB to greatly improve throughput.  Will eventually
want to make block size selection be based on network interface.
2012-12-01 10:52:08 +10:30
gardners
b7ae55c143 receiving multiple bundles via rhizome over mdp works now, although
there is a bug with alarms being called on fetch slots after they
complete, even though they apparently get unscheduled.
2012-12-01 10:47:19 +10:30
gardners
55df66f6d6 culled some debug output.
mostly works, but fetch slots can get muddled.
2012-12-01 07:43:18 +10:30
gardners
ac734ff00d fixed bugs with rhizome over mdp receive for files >32*200 bytes
long.  also we request next block of data immediately after
receiving all of the previous block.
2012-12-01 07:37:22 +10:30
gardners
7e6d836627 remember to update timeout when storing data via rhizome mdp. 2012-11-30 22:55:10 +10:30
gardners
41261bf122 gadzooks, rhizome over mdp appears to work. 2012-11-30 22:45:27 +10:30
gardners
dbb1fe8d1e rhizome over mdp now writes content for in-order packets, and
import gets triggered, but file hash currently doesn't match.
2012-11-30 22:42:28 +10:30
gardners
79c90dde6e now rhizome mdp replies get parsed and the appropriate slot, if
any, is located.
2012-11-30 22:26:31 +10:30
gardners
1be28246e4 we now pass bid+version in Rhizome over MDP block request. 2012-11-30 21:13:29 +10:30
gardners
0c7419da5d changed bar to bid in rhizome mdp fetch elements, because that
is probably the better way to address the content.  Should add
version to make sure.
2012-11-30 21:08:06 +10:30
gardners
e5d5f3f750 fixed bugs with Rhizome over MDP requests. Now seems to output
a series of MDP block requests as it should.
2012-11-30 20:20:33 +10:30
gardners
11cce162dc fixed bugs with preparing Rhizome MDP packets (source sid was not
being set properly).
2012-11-30 20:06:47 +10:30
gardners
05745d9b9a same fix as previous commit, but in another place. 2012-11-30 20:03:53 +10:30
gardners
5becae0136 fixed segfault bug with obtaining own SID. 2012-11-30 20:03:18 +10:30
gardners
7a4e3d20f8 Actually send Rhizome over MDP request packets.
Actually call appropriate callback when receiving Rhizome over
MDP packets.
2012-11-30 14:17:05 +10:30
gardners
eb7524e068 factored out internal MDP services into a separate file, and made
core case statement in that brief and clear.
2012-11-30 11:48:00 +10:30
gardners
a665750f4c prepare and dispatch mdp frames for rhizome over mdp.
still need to set bodies of frames, and schedule call backs,
and handle the requests when received.
2012-11-29 21:08:11 +10:30
gardners
c05ebada8c more work on rhizome over MDP. 2012-11-29 20:39:30 +10:30
gardners
629a36f041 call switch to mdp fetch if http request setup fails. 2012-11-29 16:43:19 +10:30
gardners
a07e5761c4 added code to try switching to mdp if http fetching fails for
any reason.
2012-11-29 16:15:51 +10:30
gardners
faad1f26b1 Pass SID as well as ip:port through rhizome fetch request logic.
This is in preparation for rhizome over MDP.
2012-11-29 15:38:04 +10:30
Andrew Bettison
c84b7e5db4 Improve str.h functions
Add strn_startswith() and strncase_startswith().

Make all str*_startswith() functions take const char * arguments, to make it
possible to do safe programming with consts.
2012-11-20 18:10:08 +10:30
Andrew Bettison
0e435683f7 Move toprint functions from log.h to str.h
Move alloca_sockaddr() to strbuf_helpers.h

So that stand-alone executables can use them without pulling in the entire
logging framework.
2012-11-20 18:10:08 +10:30
Andrew Bettison
b66d8d6c5b Fix bug in 'rhizome direct pull'
Was not fetching payload of remote bundle, just manifest.  The problem was
caused by a change of logic recently to not activate any queued fetch
candidates immediately, but wait until the next fd_poll(), so that parsing a
single packetful of rhizome advertisments would start fetching the most
important one first, instead of the first one parsed.
2012-11-12 16:24:38 +10:30
Jeremy Lakeman
2993ee46c7 Move rhizome fetch info message to end of transfer 2012-11-12 15:35:53 +10:30
Jeremy Lakeman
c79c5d8968 Fix fetch process for manifests with the same payload, make sure failures don't get stuck in a loop 2012-11-09 15:57:06 +10:30
Andrew Bettison
a29f685bdd Fix assertion failure in new Rhizome fetch queue logic
Revealed by improved Rhizome stress test that uses wide range of file sizes.
2012-11-05 14:46:16 +10:30
Andrew Bettison
39fc4ce6de Issue #30, remove periodic Rhizome fetch alarm
Replace the main-loop scheduled periodic alarm with an "activate" alarm that is
scheduled whenever a fetch candidate is added to any queue, unless the alarm is
already scheduled.

Replace the "rhizome.fetch_interval_ms" config item with
"rhizome.fetch_delay_ms" [default 50], which is the number of milliseconds
between adding a fetch candidate and firing the "activate" alarm.  This allows
time for a few more Rhizome advertisment packets to arrive after the first one,
before deciding which fetches to start first.

Add new `is_scheduled()` alarm primitive.
2012-10-24 15:13:50 +10:30
Andrew Bettison
935a545ac7 Issue #30, fix Rhizome fetch slot allocation logic
Also improve the block comment documentation of many functions.
2012-10-24 12:08:41 +10:30
Andrew Bettison
9ee9b63aad Issue #30, fox SEGV bug in new Rhizome fetch queue code
The rhizomestress test now passes, also all rhizomeprotocol tests.
2012-10-24 09:59:36 +10:30
Andrew Bettison
a9d3a1387c Issue #30, rewrite Rhizome fetch queues and slots
Causes SEGV in rhizomestress test.
2012-10-23 18:10:20 +10:30
Andrew Bettison
65ea612e27 Fix bugs revealed by Rhizome stress test
Overhauled the file fetch queue logic in rhizome_fetch.c.

Now the 'rhizomeprotocol' stress test passes in approximately 5 minutes on my
2009-vintage Dell laptop.

Added a call to rhizome_enqueue_suggestions() in rhizome_fetch_close() so that
a new Rhizome GET request is sent as soon as a fetch slot becomes free, instead
of waiting for the (default 5 second) timer to trigger the next GET.
2012-10-18 17:30:22 +10:30
Andrew Bettison
5ee82689d2 Start fixing Rhizome stress bugs
This commit does not compile.
2012-10-18 11:56:02 +10:30
Jeremy Lakeman
8b9505670d Fix handling of socket closure on OSX 2012-10-08 16:53:52 +10:30
Andrew Bettison
89343c69cd Merge branch 'rhizomedirect' into 'master' 2012-10-05 17:45:30 +09:30
Andrew Bettison
2e7d19a3f7 Issue #9, fix rhizome direct push and sync tests
The "rhizome direct push" command (and also sync) was not waiting for the
server's HTTP response, so it was exiting before the server had finished
storing the bundle, which led to a race with the subsequent "assert
bundle_received_by" test.  Fixed by adding the missing code to receive the HTTP
response.

Refactored the code used for parsing HTTP responses in rhizome_fetch.c, and
used it in rhizome_direct_http.c.
2012-10-04 17:38:33 +09:30
gardners
d43db3335c Looks like rhizome direct http pull works now.
But tests don't yet pass ... #9
2012-10-03 15:50:43 +02:00
gardners
25a64b5485 fixed some more bugs in rhizome direct http pull.
Now actually triggers the http request to obtain the file, but
the peer address is 0.0.0.0:0.  Investigating. #9
2012-10-03 15:50:43 +02:00
gardners
81cc9ddb45 more work on rhizome direct http pull. Now passing all the right
data we need to allow the prefix2manifest reception to launch
the file body request via normal rhizome http fetch process. #9
2012-10-03 15:50:43 +02:00
gardners
17de56a04c fixed some problems with the rhizome direct http pull setup code.
still working on remaining issues. #9
2012-10-03 15:50:43 +02:00
gardners
a81dafa180 work towards making rhizome direct http pull work. #9 2012-10-03 15:49:40 +02:00
Andrew Bettison
5fed0460e1 Issue #9, 'rhizomeprotocol' HttpImport test now passes
Any HTTP client can import a bundle into Rhizome using a PUSH request.
2012-10-02 17:15:14 +09:30
Andrew Bettison
cc9cb8d827 Issue #9, fix old 'rhizomeprotocol' test failures
Now only the five new Rhizome Direct tests fail.
2012-10-02 16:32:48 +09:30
Andrew Bettison
17ef2249e4 Issue #9, rewrite rhizome_import_bundle()
Objective is to avoid having to call system("servald rhizome import ...") to
handle a Rhizome direct POST /rhizome/bundle request.  Antiquated code in and
around rhizome_import_bundle() needs much cleaning up, as indicated by some
TODO comments.  Invocations must unnecessarily write the manifest into a file,
when they already have it in memory, ready to pass to the function.

All the 'rhizomeops' tests pass, but two 'rhizomeprotocol' tests are broken
by the changes in this commit.
2012-09-28 17:55:43 +09:30
Andrew Bettison
76401bd31b Improve some Rhizome debug logging code 2012-09-27 15:34:13 +09:30
Jeremy Lakeman
ac3864ff20 Reuse command line parsing framework with monitor interface 2012-09-25 13:31:34 +09:30
Andrew Bettison
45e3d1d078 Add more debug to Rhizome fetch code
To help investigate issue #16, 'rhizomeprotocol' test case 1 failing on
Solaris.
2012-09-13 14:43:47 +09:30
Jeremy Lakeman
df0e20408c Close sockets on poll errors 2012-08-31 15:04:06 +09:30
Andrew Bettison
385e2bdb4e Improve Rhizome INFO messages 2012-08-17 19:42:11 +09:30
Andrew Bettison
12d8bc0c12 Use new typedef time_ms_t everywhere 2012-08-09 12:14:32 +09:30
Andrew Bettison
240be25793 Replace overlay_gettime_ms() with gettime_ms() 2012-07-30 17:22:38 +09:30
Andrew Bettison
04b95d2590 Merge branch 'andrew' into 'master' 2012-07-25 18:04:16 +09:30
Andrew Bettison
0e5c5e0e98 More work on dnahelper, almost done
Servald starts DNA helper, receives startup ACK, sends requests, receives
responses, handles malformed helper responses, echoes dnahelper stderr lines to
log, sends MDP reply packet, waits for dead helper process, all asynchronously.
Shuts down helper process during servald shutdown.

Remaining issues:
 - Does not impose a timeout on helper responses.
 - Only the first URI is reported by the "dna lookup" command.
2012-07-20 18:17:43 +09:30
Jeremy Lakeman
1d53726b7b Merge remote-tracking branch 'origin/master'
Conflicts:
	rhizome_fetch.c
	rhizome_http.c
	testframework.sh
2012-07-17 15:39:55 +09:30