This log was autogenerated from our repository on Thu 29 Jul 2010 01:08:34 AM CEST ------------------------------------------------------------------------ r713 | wdb | 2010-07-28 21:25:00 +0200 (Wed, 28 Jul 2010) | 3 lines Selftests all pass on Ubuntu 9.04: kernel 2.6.28-11-server Updated selftest documentation in minor ways ------------------------------------------------------------------------ r712 | wdb | 2010-03-28 22:08:47 +0200 (Sun, 28 Mar 2010) | 4 lines Access control: updated fiilter and stream permissions Compilation: implemented trivial xcompile/xload functions ------------------------------------------------------------------------ r711 | cristea | 2010-03-24 14:14:43 +0100 (Wed, 24 Mar 2010) | 2 lines Fixed skb_transmit (for slowtransmit): more stable when using _fetch_out at high-rates, and automatic detection of hw-supported checksum NICs ------------------------------------------------------------------------ r710 | cristea | 2010-01-04 23:26:20 +0100 (Mon, 04 Jan 2010) | 3 lines added debugging functions for sourceip_set and destip_set ------------------------------------------------------------------------ r709 | wdb | 2010-01-03 23:42:31 +0100 (Sun, 03 Jan 2010) | 3 lines New destip_set function that rewrites IP packets' destination address field ------------------------------------------------------------------------ r708 | cristea | 2009-11-04 18:18:43 +0100 (Wed, 04 Nov 2009) | 1 line A new function in kernel-space added: sock_hijack. Includes an example user-tool program in apps/socket_hijack ------------------------------------------------------------------------ r707 | wdb | 2009-11-03 19:47:36 +0100 (Tue, 03 Nov 2009) | 10 lines update to compile on Linux kernel 2.6.31 (Ubuntu 2.6.31-14, standard for Ubuntu 9.10) BUGS: regression testing discovered these bugs - make test FAILS. infinite loop (?) at bufstack.bin - make test_kernel FAILS. same issue at mkfifo.bin BUG: programming.bin crashes kernel ------------------------------------------------------------------------ r706 | cristea | 2009-11-01 01:12:01 +0100 (Sun, 01 Nov 2009) | 1 line a missing file ------------------------------------------------------------------------ r705 | cristea | 2009-11-01 00:58:51 +0100 (Sun, 01 Nov 2009) | 1 line New FPL functions added, bugfix on multiple _in hooks (Reggie), ratelimiter function tested in both user/kernel spaces (Reggie). ------------------------------------------------------------------------ r704 | cristea | 2009-10-25 00:27:15 +0200 (Sun, 25 Oct 2009) | 3 lines minor patch at transmission to work with/without hardware checksum offload ------------------------------------------------------------------------ r703 | cristea | 2009-10-20 15:50:59 +0200 (Tue, 20 Oct 2009) | 3 lines commented out the DEBUG:netfilter_toggle_process2 message due to slow kprintf in our machines. ------------------------------------------------------------------------ r702 | wdb | 2009-10-19 03:01:37 +0200 (Mon, 19 Oct 2009) | 2 lines minor ------------------------------------------------------------------------ r701 | wdb | 2009-10-19 02:50:40 +0200 (Mon, 19 Oct 2009) | 2 lines minor updates to slquant, to recreate two figures in my dissertation ------------------------------------------------------------------------ r700 | cristea | 2009-10-12 15:49:27 +0200 (Mon, 12 Oct 2009) | 3 lines commented try_module_get/module_put calls in FPL ------------------------------------------------------------------------ r699 | wdb | 2009-10-05 03:36:58 +0200 (Mon, 05 Oct 2009) | 2 lines minor ------------------------------------------------------------------------ r698 | wdb | 2009-10-05 03:33:21 +0200 (Mon, 05 Oct 2009) | 2 lines minor ------------------------------------------------------------------------ r697 | wdb | 2009-10-05 03:11:56 +0200 (Mon, 05 Oct 2009) | 2 lines relaxed module refcounting and purge cleanup to reenable kernel module reloading ------------------------------------------------------------------------ r696 | wdb | 2009-08-30 00:24:27 +0200 (Sun, 30 Aug 2009) | 1 line ------------------------------------------------------------------------ r695 | wdb | 2009-08-29 23:53:47 +0200 (Sat, 29 Aug 2009) | 2 lines fixed bugs in new bufstack implementation that caused make test_kernel to fail ------------------------------------------------------------------------ r694 | wdb | 2009-08-29 23:22:50 +0200 (Sat, 29 Aug 2009) | 9 lines fixed minor bugs: - graphs that clean up automatically when done (e.g., those that parse a file) caused double cleanup - illegal read after free in __instance_share - other minor ones added a big red warning when using streamline lib w/o kernel module ------------------------------------------------------------------------ r693 | cristea | 2009-08-25 15:57:21 +0200 (Tue, 25 Aug 2009) | 2 lines minor defs update for compilation on older kernels ------------------------------------------------------------------------ r692 | cristea | 2009-08-25 15:28:44 +0200 (Tue, 25 Aug 2009) | 2 lines License update for ratelimit module. ------------------------------------------------------------------------ r691 | cristea | 2009-08-25 15:27:48 +0200 (Tue, 25 Aug 2009) | 1 line minor defs/include adjustments for compatibility with older kernels (to be compatible on 2.6.18-clustervision-136.1_cvos of DAS3) ------------------------------------------------------------------------ r690 | cristea | 2009-08-25 14:54:22 +0200 (Tue, 25 Aug 2009) | 2 lines introduced ratelimit function for userspace and kernelspace. Still in testing (kernel module). ------------------------------------------------------------------------ r689 | wdb | 2009-08-23 22:49:05 +0200 (Sun, 23 Aug 2009) | 10 lines updated grammar to support "a | b + !c" expressions updated check_stream to its 3 arg variant renamed and expanded peek_ and poke_options added access control to porttable added 'port' filter that supports port-based filtering and copying to ensure data is in a user-readible buffer finished buffer implementation stacking LEFT A BIG FAT BUG ------------------------------------------------------------------------ r688 | wdb | 2009-08-23 03:19:53 +0200 (Sun, 23 Aug 2009) | 3 lines added split_rr: a round robin traffic splitter filter ------------------------------------------------------------------------ r687 | wdb | 2009-08-23 02:36:09 +0200 (Sun, 23 Aug 2009) | 11 lines many minor fixes and updates: updated boolean and/or syntax fixed boolean and/or implementation added segment-backed state added session handling interface added noring buffer cleaned up prefix signature generation (was prone to failure, now robust: uses MD5) cleaned up slquantconnect.py a bit and added README ------------------------------------------------------------------------ r686 | wdb | 2009-08-17 02:01:23 +0200 (Mon, 17 Aug 2009) | 2 lines fixed bug in slquant to construction API link ------------------------------------------------------------------------ r685 | wdb | 2009-08-17 01:19:09 +0200 (Mon, 17 Aug 2009) | 3 lines hooked up slquant to the construction API ------------------------------------------------------------------------ r684 | wdb | 2009-08-16 16:50:16 +0200 (Sun, 16 Aug 2009) | 3 lines got slquant resolving large problems ------------------------------------------------------------------------ r683 | wdb | 2009-08-10 12:39:35 +0200 (Mon, 10 Aug 2009) | 2 lines halfway (?) in connecting slquant to streamline ------------------------------------------------------------------------ r682 | wdb | 2009-08-09 17:53:11 +0200 (Sun, 09 Aug 2009) | 2 lines started work on slquant-libstreamline connection ------------------------------------------------------------------------ r681 | wdb | 2009-08-05 13:04:54 +0200 (Wed, 05 Aug 2009) | 2 lines improved solver test ------------------------------------------------------------------------ r680 | wdb | 2009-08-04 17:26:10 +0200 (Tue, 04 Aug 2009) | 2 lines minor ------------------------------------------------------------------------ r679 | wdb | 2009-08-03 19:44:44 +0200 (Mon, 03 Aug 2009) | 2 lines minor ------------------------------------------------------------------------ r678 | wdb | 2009-08-03 09:29:47 +0200 (Mon, 03 Aug 2009) | 2 lines minor ------------------------------------------------------------------------ r677 | wdb | 2009-08-03 00:36:15 +0200 (Mon, 03 Aug 2009) | 2 lines minor ------------------------------------------------------------------------ r676 | wdb | 2009-08-03 00:03:24 +0200 (Mon, 03 Aug 2009) | 3 lines more work on slquant ------------------------------------------------------------------------ r675 | wdb | 2009-08-02 20:55:38 +0200 (Sun, 02 Aug 2009) | 2 lines updated slquant for FFPF test ------------------------------------------------------------------------ r674 | wdb | 2009-08-02 17:32:40 +0200 (Sun, 02 Aug 2009) | 6 lines fixed bugs in slquant: * solver gives correct results for tasks thrown at it so far * toG6 works clearly marked one open bug: mixing of paths with saved and unsaved data ------------------------------------------------------------------------ r673 | wdb | 2009-07-29 23:44:53 +0200 (Wed, 29 Jul 2009) | 2 lines further worked on slquant. not done ------------------------------------------------------------------------ r672 | wdb | 2009-07-29 18:27:17 +0200 (Wed, 29 Jul 2009) | 5 lines improved slquant: * simplified * fixed bug in slqtemplate * fixed cost calculation ------------------------------------------------------------------------ r671 | wdb | 2009-07-25 20:53:55 +0200 (Sat, 25 Jul 2009) | 2 lines added the dissertation todos to the list ------------------------------------------------------------------------ r670 | wdb | 2009-07-24 16:16:35 +0200 (Fri, 24 Jul 2009) | 3 lines finished (hopefully) pipeline bench ------------------------------------------------------------------------ r669 | wdb | 2009-07-23 21:14:59 +0200 (Thu, 23 Jul 2009) | 2 lines minor ------------------------------------------------------------------------ r668 | wdb | 2009-07-23 18:17:00 +0200 (Thu, 23 Jul 2009) | 7 lines fixed a number of smaller bugs and undocumented maxima: e.g., maximal request length maximal number of mmapped buffers version used to run pipeline benchmark ------------------------------------------------------------------------ r667 | wdb | 2009-07-22 21:19:56 +0200 (Wed, 22 Jul 2009) | 4 lines got the pipeline test working (it seems) removed debug messages ------------------------------------------------------------------------ r666 | wdb | 2009-07-21 22:22:44 +0200 (Tue, 21 Jul 2009) | 2 lines updated pipeline test ------------------------------------------------------------------------ r665 | wdb | 2009-07-20 21:41:41 +0200 (Mon, 20 Jul 2009) | 2 lines updated for pipeline test. ------------------------------------------------------------------------ r664 | wdb | 2009-07-20 17:21:22 +0200 (Mon, 20 Jul 2009) | 9 lines modified streamline to 1) require the kernel module loaded for ALL applications because this avoids different executions between with and without cases 2) by default create buffers in the kernel, unless OB_LOCAL is passed ------------------------------------------------------------------------ r663 | wdb | 2009-07-20 14:28:02 +0200 (Mon, 20 Jul 2009) | 3 lines updated to compile with Ubuntu Jaunty (9.04) : gcc 4.3.3 and Linux kernel 2.6.29 ------------------------------------------------------------------------ r662 | wdb | 2009-07-16 16:23:45 +0200 (Thu, 16 Jul 2009) | 2 lines minor. templates appear broken ------------------------------------------------------------------------ r661 | wdb | 2009-07-15 21:18:14 +0200 (Wed, 15 Jul 2009) | 3 lines warning:broken implemented part of sysconstraint and memaccess model ------------------------------------------------------------------------ r660 | wdb | 2009-07-15 18:13:01 +0200 (Wed, 15 Jul 2009) | 4 lines discovered two issues in template writing. fixed one (and marked other) started work on correct data access cost. not done yet ------------------------------------------------------------------------ r659 | wdb | 2009-07-15 12:50:12 +0200 (Wed, 15 Jul 2009) | 2 lines added profiling ------------------------------------------------------------------------ r658 | wdb | 2009-07-14 23:59:23 +0200 (Tue, 14 Jul 2009) | 5 lines integrated GLPK solver into slquant using excellent pyglpk package improved imageviewer to have reposition and zoom-to-fit broke up slquant code into smaller, more maintainable packages ------------------------------------------------------------------------ r657 | wdb | 2009-07-14 18:09:32 +0200 (Tue, 14 Jul 2009) | 3 lines slquant: wrote cyclecost calculation (not done) slquant: finished imageviewer ------------------------------------------------------------------------ r656 | wdb | 2009-07-14 10:09:08 +0200 (Tue, 14 Jul 2009) | 2 lines minor ------------------------------------------------------------------------ r655 | wdb | 2009-07-13 19:27:42 +0200 (Mon, 13 Jul 2009) | 3 lines updated slquant: now works up to G6. lacks G4 (parallelize) and cost calculation ------------------------------------------------------------------------ r654 | wdb | 2009-07-12 21:53:34 +0200 (Sun, 12 Jul 2009) | 3 lines update slquant ------------------------------------------------------------------------ r653 | wdb | 2009-07-12 19:10:45 +0200 (Sun, 12 Jul 2009) | 2 lines added new slquant ------------------------------------------------------------------------ r652 | wdb | 2009-07-09 20:04:02 +0200 (Thu, 09 Jul 2009) | 2 lines minor ------------------------------------------------------------------------ r651 | wdb | 2009-07-08 21:13:23 +0200 (Wed, 08 Jul 2009) | 2 lines minor ------------------------------------------------------------------------ r650 | wdb | 2009-07-06 10:00:30 +0200 (Mon, 06 Jul 2009) | 2 lines pipeline streamline TODO added ------------------------------------------------------------------------ r649 | wdb | 2009-07-06 09:57:49 +0200 (Mon, 06 Jul 2009) | 2 lines debugged pipeline benchmark ------------------------------------------------------------------------ r648 | wdb | 2009-07-05 22:46:41 +0200 (Sun, 05 Jul 2009) | 3 lines added container task prototype updated pipeline test ------------------------------------------------------------------------ r647 | wdb | 2009-07-05 13:47:21 +0200 (Sun, 05 Jul 2009) | 2 lines incorporated prototype template code ------------------------------------------------------------------------ r646 | wdb | 2009-07-03 18:34:33 +0200 (Fri, 03 Jul 2009) | 3 lines imported pipeline benchmark ------------------------------------------------------------------------ r645 | wdb | 2009-06-23 19:25:38 +0200 (Tue, 23 Jun 2009) | 2 lines 'started' session handling interface ------------------------------------------------------------------------ r644 | wdb | 2009-05-11 15:00:18 +0200 (Mon, 11 May 2009) | 2 lines minor todo update ------------------------------------------------------------------------ r643 | wdb | 2009-05-11 14:49:46 +0200 (Mon, 11 May 2009) | 2 lines minor TODO update ------------------------------------------------------------------------ r642 | wdb | 2009-05-05 22:55:59 +0200 (Tue, 05 May 2009) | 2 lines import of multicore multilayer cache streamline simulator ------------------------------------------------------------------------ r641 | wdb | 2009-05-05 22:52:28 +0200 (Tue, 05 May 2009) | 2 lines imported slquant solver ------------------------------------------------------------------------ r640 | wdb | 2009-05-04 14:30:50 +0200 (Mon, 04 May 2009) | 2 lines imported some bug TODOs from snippets ------------------------------------------------------------------------ r639 | cristea | 2009-03-22 23:12:35 +0100 (Sun, 22 Mar 2009) | 2 lines corrected the checksum condition. Fragmented case not tested yet. ------------------------------------------------------------------------ r638 | wdb | 2009-03-22 19:52:27 +0100 (Sun, 22 Mar 2009) | 3 lines bugfix: no checksum on fragmented ip beyond first fragment bugfix: fixed expected tcp checksum location on transmit ------------------------------------------------------------------------ r637 | cristea | 2009-03-20 20:31:11 +0100 (Fri, 20 Mar 2009) | 2 lines a typo ------------------------------------------------------------------------ r636 | wdb | 2009-03-20 18:23:17 +0100 (Fri, 20 Mar 2009) | 3 lines minor fix in transmit.c ------------------------------------------------------------------------ r635 | wdb | 2009-03-03 11:43:19 +0100 (Tue, 03 Mar 2009) | 2 lines modified skb transmit to perform IP csum recalculation for all IP packets, not just TCP/UDP ------------------------------------------------------------------------ r634 | cristea | 2009-02-23 09:57:58 +0100 (Mon, 23 Feb 2009) | 2 lines Solved the size_t error when compiling into Fedora10: /opt/streamline/include/streamline/block.h:21: error: expected ')' before 'size' ------------------------------------------------------------------------ r633 | wdb | 2009-02-17 21:34:23 +0100 (Tue, 17 Feb 2009) | 2 lines modified skbtransmit to only recalculate checksums on TCP and UDP (it also did erroneously on ICMP) ------------------------------------------------------------------------ r632 | wdb | 2009-02-17 13:28:54 +0100 (Tue, 17 Feb 2009) | 3 lines new: metadata_processed unittest and example for finding out how many packets were processed control/metadata support functions to make this work ------------------------------------------------------------------------ r631 | wdb | 2009-02-16 18:44:20 +0100 (Mon, 16 Feb 2009) | 2 lines minor changes ------------------------------------------------------------------------ r630 | wdb | 2009-02-13 12:21:56 +0100 (Fri, 13 Feb 2009) | 2 lines minor pcap compilation issue: cast const char to char ------------------------------------------------------------------------ r629 | wdb | 2009-02-13 10:26:09 +0100 (Fri, 13 Feb 2009) | 2 lines removed config.status ------------------------------------------------------------------------ r628 | wdb | 2009-02-12 20:54:12 +0100 (Thu, 12 Feb 2009) | 12 lines NEW: automatic request removal when all input has been processed. graphs that have only fixed-length sources (such as files) will now automatically detect when all sources are exhausted and then remove all applicable requests. CLEANUP: as a result, I was able to remove the nasty and racey timer-based callback to read a file once we guessed that the whole graph incl. all descendents was active. Now, file read occurs in the new ->activate() method, which is certain to be called after all descendents have been activated ------------------------------------------------------------------------ r627 | wdb | 2009-02-12 17:18:19 +0100 (Thu, 12 Feb 2009) | 3 lines modified filein and pcapin filters to use the new activate() method to avoid race conditions on data arrival and graph constructions ------------------------------------------------------------------------ r626 | wdb | 2009-02-12 16:13:20 +0100 (Thu, 12 Feb 2009) | 4 lines added activate() method to the filter API to allow sources to start only when their children are known to be active converted the shell filter to use activate() for spawning a subprocess ------------------------------------------------------------------------ r625 | wdb | 2009-02-12 00:47:32 +0100 (Thu, 12 Feb 2009) | 2 lines fixed some problems encountered on Debian Etch ------------------------------------------------------------------------ r623 | wdb | 2009-02-12 00:08:27 +0100 (Thu, 12 Feb 2009) | 5 lines milestone: examples on netstreamline.org/doc/examples.php now all work minor bugfix: filesout tried to close unopened files. ------------------------------------------------------------------------ r622 | wdb | 2009-02-12 00:00:09 +0100 (Thu, 12 Feb 2009) | 2 lines minor bugfix: untcp crashed due to expression substatements being in the wrong order ------------------------------------------------------------------------ r621 | wdb | 2009-02-11 23:41:53 +0100 (Wed, 11 Feb 2009) | 2 lines bugfix in new parser: expression in quotes would contain first quote ------------------------------------------------------------------------ r620 | wdb | 2009-02-11 23:14:11 +0100 (Wed, 11 Feb 2009) | 3 lines minor changes to documentation ------------------------------------------------------------------------ r619 | wdb | 2009-02-11 11:12:08 +0100 (Wed, 11 Feb 2009) | 2 lines cleaned up inflate/deflate ------------------------------------------------------------------------ r618 | wdb | 2009-02-11 10:50:15 +0100 (Wed, 11 Feb 2009) | 2 lines modularized kernel deflate/inflate, to enable streamline loading without kernel zlib support ------------------------------------------------------------------------ r617 | wdb | 2009-02-11 01:30:13 +0100 (Wed, 11 Feb 2009) | 2 lines minor ------------------------------------------------------------------------ r616 | wdb | 2009-02-11 01:04:39 +0100 (Wed, 11 Feb 2009) | 2 lines bugfix ------------------------------------------------------------------------ r615 | wdb | 2009-02-10 22:55:58 +0100 (Tue, 10 Feb 2009) | 2 lines minor ------------------------------------------------------------------------ r614 | wdb | 2009-02-10 21:27:09 +0100 (Tue, 10 Feb 2009) | 2 lines zlib inflate/deflate expansion ------------------------------------------------------------------------ r613 | wdb | 2009-02-10 17:35:57 +0100 (Tue, 10 Feb 2009) | 2 lines debug version ------------------------------------------------------------------------ r612 | wdb | 2009-02-10 17:25:11 +0100 (Tue, 10 Feb 2009) | 2 lines in debug ------------------------------------------------------------------------ r611 | wdb | 2009-02-09 19:06:33 +0100 (Mon, 09 Feb 2009) | 3 lines first import of new deflate compressor. needs work (and corresponding inflate) ------------------------------------------------------------------------ r610 | wdb | 2009-02-03 21:29:54 +0100 (Tue, 03 Feb 2009) | 6 lines new feature: requests no longer need explicit --expression=$VALUE parameters, VALUE is enough for instance netfilter_in --expression=eth0 cannot be written as netfilter_in eth0 minor bugfix: the pass filter was registered twice, causing a dictionary collision ------------------------------------------------------------------------ r609 | wdb | 2009-02-03 19:37:17 +0100 (Tue, 03 Feb 2009) | 7 lines introduced quantitative optimization logic, from filters up to core/process.c values are computed, but not yet used. push processing down heuristic is still active fixed issue in pcapout. wouldn't generate correct data on 64bit platforms due to different sizes of struct timeval on 32 and 64bit systems ------------------------------------------------------------------------ r608 | wdb | 2009-02-03 01:03:40 +0100 (Tue, 03 Feb 2009) | 2 lines minor fix in boolean or ------------------------------------------------------------------------ r607 | wdb | 2009-02-02 19:14:03 +0100 (Mon, 02 Feb 2009) | 4 lines updated zlib compress/uncompress filters to use the new slblock_writable API they now both work correctly also, this allowed me to remove two stale hacky interfaces in core/file.c ------------------------------------------------------------------------ r605 | wdb | 2009-02-01 20:58:36 +0100 (Sun, 01 Feb 2009) | 2 lines minor fixes to shell and grammar ------------------------------------------------------------------------ r604 | wdb | 2009-02-01 19:25:01 +0100 (Sun, 01 Feb 2009) | 2 lines resolved some parsing issues and fixed up the shell implementations. parameters are now supported ------------------------------------------------------------------------ r603 | wdb | 2009-01-31 13:38:24 +0100 (Sat, 31 Jan 2009) | 2 lines minor fix ------------------------------------------------------------------------ r602 | wdb | 2009-01-31 13:37:22 +0100 (Sat, 31 Jan 2009) | 5 lines changed file interface from CALL_stream to slfile_CALL to avoid symbol collisions with apps added slcli option to switch between languages, made 1.8 the default builds on 2.6.27 and passes all tests ------------------------------------------------------------------------ r601 | wdb | 2009-01-23 20:07:12 +0100 (Fri, 23 Jan 2009) | 3 lines fixed bugs in v1.8 language. most seems gone, but @ and ! still have problems ------------------------------------------------------------------------ r600 | wdb | 2009-01-20 10:19:28 +0100 (Tue, 20 Jan 2009) | 2 lines introduction of parser for the new, more unixy, 1.8 language ------------------------------------------------------------------------ r599 | wdb | 2008-11-23 19:56:06 +0100 (Sun, 23 Nov 2008) | 2 lines minor comment ------------------------------------------------------------------------ r598 | wdb | 2008-11-12 16:59:09 +0100 (Wed, 12 Nov 2008) | 3 lines updated `update instance` to not reuse state across updates ------------------------------------------------------------------------ r597 | cristea | 2008-11-11 12:38:29 +0100 (Tue, 11 Nov 2008) | 2 lines Improved the "make fpl" option ------------------------------------------------------------------------ r596 | cristea | 2008-11-11 11:22:32 +0100 (Tue, 11 Nov 2008) | 3 lines Changed the ifdef check of (2.6.19) into (2.6.18) in order to allow compiling without problems on CentOS 5 kernels (e.g., 2.6.18-92.1.17.el5-i686) ------------------------------------------------------------------------ r595 | wdb | 2008-11-11 10:23:53 +0100 (Tue, 11 Nov 2008) | 3 lines bugfix: crash on info instance of unknown request bugfix: netfilter hook define on 2.6.24 should be on 2.6.25 ------------------------------------------------------------------------ r594 | wdb | 2008-11-07 10:23:09 +0100 (Fri, 07 Nov 2008) | 3 lines slightly modified the stack to not spew underflow warnings in stack_foreach commented out a line in mapi that caused dual definitions of socket stuff ------------------------------------------------------------------------ r593 | wdb | 2008-11-07 10:03:44 +0100 (Fri, 07 Nov 2008) | 2 lines added name= field to info instance ------------------------------------------------------------------------ r592 | wdb | 2008-11-04 19:59:28 +0100 (Tue, 04 Nov 2008) | 2 lines changed CLI handling of SIGINT and SIGTERM ------------------------------------------------------------------------ r591 | wdb | 2008-11-03 18:56:43 +0100 (Mon, 03 Nov 2008) | 3 lines changed the stack to rely less on a magic value in stack_pop and not at all in stack_foreach fixed parser's stack accordingly ------------------------------------------------------------------------ r590 | wdb | 2008-11-03 18:05:11 +0100 (Mon, 03 Nov 2008) | 4 lines minor stack update: aprintf to dprintf and unittest extended with stack_foreach test ------------------------------------------------------------------------ r589 | wdb | 2008-11-03 18:03:05 +0100 (Mon, 03 Nov 2008) | 6 lines updated the generic stack. now each stack has a private name(space) stack_pop works simpler no default stack length is defined unittest has been added ------------------------------------------------------------------------ r588 | wdb | 2008-11-02 11:12:25 +0100 (Sun, 02 Nov 2008) | 2 lines added prettyprinted expression field to `info instance` output ------------------------------------------------------------------------ r587 | cristea | 2008-11-01 03:01:57 +0100 (Sat, 01 Nov 2008) | 3 lines Added tokenless fpl component ------------------------------------------------------------------------ r586 | wdb | 2008-10-27 19:39:48 +0100 (Mon, 27 Oct 2008) | 3 lines added expression=$ipv4 address for netfilter filters ------------------------------------------------------------------------ r585 | wdb | 2008-10-23 09:08:22 +0200 (Thu, 23 Oct 2008) | 2 lines minor bugfix: NULL pointer dereference on info request ------------------------------------------------------------------------ r584 | wdb | 2008-10-17 17:24:21 +0200 (Fri, 17 Oct 2008) | 6 lines BUILD: changed version numbering to show the subversion revision for repository builds BUILD: modified the process to automatically detect changes to Makefile.in and call ./configure NEW: started implementing stackable buffer implementations ------------------------------------------------------------------------ r583 | wdb | 2008-10-16 10:15:32 +0200 (Thu, 16 Oct 2008) | 5 lines added 'info request' command that shows edges and filters added the underlying control_info_request() function changed slrequest_add to number requests with increasing natural numbers from 1, instead of random ------------------------------------------------------------------------ r582 | cristea | 2008-10-15 18:19:57 +0200 (Wed, 15 Oct 2008) | 2 lines Bugfix on fpl_tr component. ------------------------------------------------------------------------ r580 | cristea | 2008-10-14 18:56:06 +0200 (Tue, 14 Oct 2008) | 3 lines fpl_tr bugfix1 ------------------------------------------------------------------------ r579 | cristea | 2008-10-14 16:54:29 +0200 (Tue, 14 Oct 2008) | 2 lines Added the new fpl_tr example (tr.kef) ------------------------------------------------------------------------ r578 | cristea | 2008-10-14 16:48:58 +0200 (Tue, 14 Oct 2008) | 3 lines Added TokenRemove (fpl_tr) component ------------------------------------------------------------------------ r577 | wdb | 2008-10-12 23:11:22 +0200 (Sun, 12 Oct 2008) | 3 lines added "update instance" to the CLI: the interactive equivalent of slnode_update ------------------------------------------------------------------------ r576 | cristea | 2008-10-10 16:37:29 +0200 (Fri, 10 Oct 2008) | 3 lines Added one more fpl component: udpdest.kef ------------------------------------------------------------------------ r575 | cristea | 2008-10-10 16:36:50 +0200 (Fri, 10 Oct 2008) | 3 lines Added one more fpl: udpdest ------------------------------------------------------------------------ r574 | cristea | 2008-10-09 19:12:28 +0200 (Thu, 09 Oct 2008) | 3 lines Improved the FPL_TB: it is able to update an existing token (if the packet has IPoptions already) ------------------------------------------------------------------------ r573 | cristea | 2008-10-09 18:03:50 +0200 (Thu, 09 Oct 2008) | 3 lines Fixed the SL_PARAM handling for fpl components ------------------------------------------------------------------------ r572 | wdb | 2008-10-09 11:53:17 +0200 (Thu, 09 Oct 2008) | 2 lines extended netfilter_toggle to also affect netfilter_out ------------------------------------------------------------------------ r571 | wdb | 2008-10-09 11:50:03 +0200 (Thu, 09 Oct 2008) | 5 lines updated netfilter_in to listen to return value override when a global variable is set added (netfilter_toggle) that modifies this global variable added -all-> edge to the parser as a simplification of -0:65535-> ------------------------------------------------------------------------ r570 | cristea | 2008-10-08 20:20:13 +0200 (Wed, 08 Oct 2008) | 2 lines Working version of fpl components. ------------------------------------------------------------------------ r569 | wdb | 2008-10-08 09:28:57 +0200 (Wed, 08 Oct 2008) | 2 lines added debug statement ------------------------------------------------------------------------ r568 | wdb | 2008-10-08 09:17:12 +0200 (Wed, 08 Oct 2008) | 4 lines changed NF_DROP to NF_STOLEN on netfilter_fetch_out fixed non-atomic kmalloc in interrupt context ------------------------------------------------------------------------ r567 | cristea | 2008-10-08 00:21:57 +0200 (Wed, 08 Oct 2008) | 3 lines Fixed packet writing code with slblock_writable() ------------------------------------------------------------------------ r566 | wdb | 2008-10-07 22:31:41 +0200 (Tue, 07 Oct 2008) | 3 lines made sure netfilter_out > tx could not cause an infinite loop ------------------------------------------------------------------------ r565 | cristea | 2008-10-07 17:53:33 +0200 (Tue, 07 Oct 2008) | 4 lines ToTest packet writing with insert (netfilter_fetch_out) > (udp) > (fpl_tb,expression="1234") > (skb_transmit) ------------------------------------------------------------------------ r564 | cristea | 2008-10-07 05:21:51 +0200 (Tue, 07 Oct 2008) | 1 line Added fpl_tryex for parser, added first tb (tokenbuild) fpl component. ------------------------------------------------------------------------ r563 | wdb | 2008-10-03 09:12:38 +0200 (Fri, 03 Oct 2008) | 2 lines exported slblock_writable symbol in kernel (for third-party modules) ------------------------------------------------------------------------ r562 | wdb | 2008-10-02 15:03:32 +0200 (Thu, 02 Oct 2008) | 3 lines added slblock_writable to fake_fpl2 ------------------------------------------------------------------------ r561 | wdb | 2008-10-02 13:26:42 +0200 (Thu, 02 Oct 2008) | 2 lines fixed bug that blocked running of multiple different pcap clients at the same time ------------------------------------------------------------------------ r560 | wdb | 2008-10-01 10:22:27 +0200 (Wed, 01 Oct 2008) | 2 lines created netfilter_fetch_out ------------------------------------------------------------------------ r559 | cristea | 2008-09-27 00:23:28 +0200 (Sat, 27 Sep 2008) | 1 line Added write SL_PARAM (or other values) into Token. ------------------------------------------------------------------------ r558 | cristea | 2008-09-24 18:37:01 +0200 (Wed, 24 Sep 2008) | 3 lines bugfix due to the trying of tryex feature in streamline ------------------------------------------------------------------------ r557 | cristea | 2008-09-24 00:50:45 +0200 (Wed, 24 Sep 2008) | 2 lines one more fpl component: replaces IP_DEST and transmit packet ------------------------------------------------------------------------ r556 | cristea | 2008-09-24 00:49:57 +0200 (Wed, 24 Sep 2008) | 2 lines Tried .tryex for expression parsing of fpl ------------------------------------------------------------------------ r555 | cristea | 2008-09-23 23:52:24 +0200 (Tue, 23 Sep 2008) | 1 line Extended the fpl_compiler with SL_PARAM syntax for handling the streamline expression provided as a parameter to the fpl_module such as: insert (rx) > (fpl, expression="192.168.0.1") > (tx) ------------------------------------------------------------------------ r554 | cristea | 2008-09-17 13:37:43 +0200 (Wed, 17 Sep 2008) | 1 line Updated fpl_module template ------------------------------------------------------------------------ r553 | cristea | 2008-09-17 13:31:11 +0200 (Wed, 17 Sep 2008) | 1 line Updated tbs.kef in the fpl examples ------------------------------------------------------------------------ r552 | cristea | 2008-09-16 11:49:59 +0200 (Tue, 16 Sep 2008) | 1 line Implemented write into PKT.[] by using assignment expressions (e.g., PKT.DW[4] = M[1]. or IP_DEST = M[1]) ------------------------------------------------------------------------ r551 | wdb | 2008-09-12 04:03:49 +0200 (Fri, 12 Sep 2008) | 6 lines DEBUG make netfilter copy all data before passing it on to streamline create a fake version of Mihai's FPL2 code ------------------------------------------------------------------------ r550 | cristea | 2008-09-11 22:15:07 +0200 (Thu, 11 Sep 2008) | 1 line Added new user-application (tbs) and kernel fpl_tbs component. ------------------------------------------------------------------------ r549 | cristea | 2008-09-11 22:05:25 +0200 (Thu, 11 Sep 2008) | 1 line Updated the fpl_compiler to support PKT.TOKEN syntax. ------------------------------------------------------------------------ r548 | wdb | 2008-09-08 11:49:47 +0200 (Mon, 08 Sep 2008) | 3 lines updated skbtransmit to work with ip packets that contain options (not tested) ------------------------------------------------------------------------ r547 | wdb | 2008-09-08 11:41:42 +0200 (Mon, 08 Sep 2008) | 3 lines bugfix: the same signal could be added multiple time to the timeout queue, because the pending=0 test was set after each batch-induced signal ------------------------------------------------------------------------ r546 | wdb | 2008-09-08 09:25:50 +0200 (Mon, 08 Sep 2008) | 3 lines Made Streamline compile on Gentoo Base System release 1.12.11.1 (2008.1) Had to fix some include directives for sockets ------------------------------------------------------------------------ r545 | wdb | 2008-08-28 16:37:37 +0200 (Thu, 28 Aug 2008) | 6 lines bugfix: slsighandler list overflowed when space ID exceeded 15, leading to nasty inconsistent kernel state new: a generic stack-based list implementation for (relatively) fast lookup. new: unit-test for the above : ------------------------------------------------------------------------ r544 | wdb | 2008-08-28 10:51:18 +0200 (Thu, 28 Aug 2008) | 2 lines unstable, testing ------------------------------------------------------------------------ r543 | wdb | 2008-08-27 11:00:39 +0200 (Wed, 27 Aug 2008) | 2 lines minor: removed debug statement in pcap ------------------------------------------------------------------------ r542 | wdb | 2008-08-27 10:27:21 +0200 (Wed, 27 Aug 2008) | 2 lines Fixed regression in pcap ------------------------------------------------------------------------ r541 | wdb | 2008-08-25 00:21:36 +0200 (Mon, 25 Aug 2008) | 2 lines fixed regression ------------------------------------------------------------------------ r540 | wdb | 2008-08-24 22:50:52 +0200 (Sun, 24 Aug 2008) | 10 lines Changes for the UPVN application New: slnode_update() allows changing running filters (e.g., replacing filter implementation) New: in SRL requests, filters may be passed 'reuse' and 'share' options to manipulate behavior of filter replacement and prefix sharing with regards to state New: unittest for the above features Mod: slrequest_(gs)etnodedata are now called slnode_(gs)etdata ------------------------------------------------------------------------ r539 | wdb | 2008-08-24 16:32:01 +0200 (Sun, 24 Aug 2008) | 3 lines minor: pipesfs update ------------------------------------------------------------------------ r538 | wdb | 2008-08-24 16:22:21 +0200 (Sun, 24 Aug 2008) | 3 lines minor bugfix: set access control flags on buffer open correctly. Now the second OSR test works decently ------------------------------------------------------------------------ r537 | wdb | 2008-08-22 20:05:56 +0200 (Fri, 22 Aug 2008) | 2 lines blocking semantics in datapath resolved. at first glance pipesfs seems fixed. ------------------------------------------------------------------------ r536 | wdb | 2008-08-22 17:44:19 +0200 (Fri, 22 Aug 2008) | 5 lines Fixed minor bug in support/timer exception handling (when timers are unavailable) All unittests and all test_kernel tests pass! ------------------------------------------------------------------------ r535 | wdb | 2008-08-22 16:37:58 +0200 (Fri, 22 Aug 2008) | 12 lines BUGFIX: Writing into a pipesfs pipe would not call its children. Other fixes: * Writing into a pipe no longer requires 2 memcpys and a dynalloc/free, because we switched to the new asynchronous write buffer interface. * Producer locking is fixed to work in userspace, kernel process context and kernel int. contenxt * Producer locking is applied to acquire/complete * IBuf now has an asynchronous (acquire/complete) interface * On write through, offset was 1 too high * In write, space was always the space default, which is incorrect after an ID update ------------------------------------------------------------------------ r534 | wdb | 2008-08-21 19:01:09 +0200 (Thu, 21 Aug 2008) | 8 lines FIX: allow multiple concurrent users of pipes in PipesFS, each with their own offset (see also the previous commit) and smaller items: MOD: generator now returns a simple long containing an increasing number. FIX: task_stop and task_stop_all in userspace could interfere and cause unstable state. o more FIX: task_stop would reference data that had been previously released. No more. ------------------------------------------------------------------------ r533 | wdb | 2008-08-21 17:48:49 +0200 (Thu, 21 Aug 2008) | 14 lines BUGFIX: could no longer read from pipesfs pipes Details: *FIX: buf->shared->origbd was 8-bit, needs to be 16-bit *FIX: copy_(to|from)_user returnvalue was used incorrectly (taken as copied instead of NOT copied bytes) *MOD: Waking up polling processes no longer takes the form of a signal callback. The old arrangement could lead to userspace readers signalling kernel writers. In the current version, PipesFS works again, but I noticed that multiple users of the same inode (e.g., pipe) share a buffer interface, which results in only one of the clients receiving any piece of data. I will now try to resolve this ------------------------------------------------------------------------ r532 | wdb | 2008-08-21 11:39:22 +0200 (Thu, 21 Aug 2008) | 2 lines hacky ugly debug version, only committed to switch to different environment (virtualized) ------------------------------------------------------------------------ r531 | cristea | 2008-08-17 21:50:17 +0200 (Sun, 17 Aug 2008) | 1 line minor update in fpl ------------------------------------------------------------------------ r530 | wdb | 2008-08-14 19:51:56 +0200 (Thu, 14 Aug 2008) | 9 lines new: PIPE_BUF variable as defined by Posix pipes new: Added locking to serialize buffer writing in the kernel. This was also required for the atomicity requirement of PIPE_BUF new: mpipe() "multiway pipe" call. BROKEN (too permissive access control) new: mpipe selftest. Fails, as the mpipe is still broken update: expanded the 'make test' autorun set with some other preexisting unittests new: 'make test_kernel' target to automatically verify behavior requiring the kernel module ------------------------------------------------------------------------ r529 | wdb | 2008-08-14 17:43:25 +0200 (Thu, 14 Aug 2008) | 2 lines introduced unittest for the new storage access api ------------------------------------------------------------------------ r528 | wdb | 2008-08-14 11:35:16 +0200 (Thu, 14 Aug 2008) | 3 lines added functionality to write to a filter's persistent memory. added locks to the persistent memory region for safe concurrent access ------------------------------------------------------------------------ r526 | wdb | 2008-08-13 14:12:13 +0200 (Wed, 13 Aug 2008) | 3 lines updated doc/TESTED to include latest machines. This will be distribution 1.7.4.4 ------------------------------------------------------------------------ r525 | wdb | 2008-08-13 13:53:51 +0200 (Wed, 13 Aug 2008) | 7 lines Fixed compilation on Fedora Core 6 (2.6.19....). Details: minor BUGFIX: beltwayfs:use of work_struct in 2.6.19 or older was incorrect (used delayed_work somewhere) minor BUGFIX: pipesfs:file->f_path was used somewhere in 2.6.19 or older, while it did not yet exist. minor BUGFIX: pipesfs:copy_(to|from)_user disregarded return vale. caused warnings minor BUGFIX: pipesfs:link_iops was declared const. caused warning ------------------------------------------------------------------------ r524 | wdb | 2008-08-13 13:24:42 +0200 (Wed, 13 Aug 2008) | 2 lines minor bugfix in block test ------------------------------------------------------------------------ r523 | cristea | 2008-08-12 14:36:12 +0200 (Tue, 12 Aug 2008) | 1 line Updated fpl interface regards to the fdata->mem. and fptr->data instead of pdu ------------------------------------------------------------------------ r522 | wdb | 2008-08-11 19:05:56 +0200 (Mon, 11 Aug 2008) | 7 lines (minor) bugfix: dbuf updated read pointer when using read_indexed (minor) bugfix: dbuf has an unsigned index where it should be signed. caused unnecessary overflow add: KNOWN_ISSUES file, discussing the concurrency issues updated: BUGS, TESTED and pipesfs documentation ------------------------------------------------------------------------ r521 | wdb | 2008-08-08 21:54:13 +0200 (Fri, 08 Aug 2008) | 2 lines minor bugfix: clear outstanding signals, to rearm timeout signal ------------------------------------------------------------------------ r520 | wdb | 2008-08-08 14:43:25 +0200 (Fri, 08 Aug 2008) | 15 lines Version used to run bind and mplayer benchmarks. New * asynchronous buffer write interface through optional acquire()/complete() members Bugfix * the index generated by application splice was only valid for the first loop through the ring Known issues * timeout value doesn't work: timeout-based signalling is nonexistent/too slow * obscure kernel paging error (panic) in mhash_get<-slsighandler_del when closing bind when slcli had also been active * HW issue: tg3 (or at least the broadcom in my HP6710b) observes hardware hang when transmitting non-aligned packets. ------------------------------------------------------------------------ r519 | wdb | 2008-08-06 20:35:21 +0200 (Wed, 06 Aug 2008) | 2 lines minor bugfix to previous commit: libpcap, sockets and pipes seem to work. need to try fbufvbufdbuf, fork and mkfifo ------------------------------------------------------------------------ r518 | wdb | 2008-08-06 20:19:21 +0200 (Wed, 06 Aug 2008) | 23 lines MAJOR cleanup of buffer synchronisation (blocking, Posix semantics, signalling). This version compiles and the default tests + obshared_(reader|writer) succeeds, but it needs more testing. If necessary, a next commit fixes minor issues with pcap/socket/pipe/slcli handling. summary: * SRP and MRP are now fully implemented and enabled. FRP remains the default * BUGFIX: the signal moderation background task ran for each signal, instead of the 1st of an epoch * BUGFIX: Buffers now only return EOF on reading of a buffer w/o writers !that used to have them! (POSIX compliance) * signals now differentiate between read and write updates * simplified buffer signaling path: removed duplicate code and fixed omissions in wake_up calling (BUGS!) * reduced the amount of wake_up to a minimum, to maximize efficiency minor: * struct buf_impl has been extended with a sync() member to allow impl. specific shared readpointer updating * replaced the _GSPACE and _GID macros with clearer global ID structure * replaced the explicit epoch counter ------------------------------------------------------------------------ r517 | wdb | 2008-08-05 12:15:14 +0200 (Tue, 05 Aug 2008) | 4 lines BUGFIX: deadlock on both rx/tx traffic in kernel due to preemption with lock held in slsig_handle BUGFIX: transmission stopped due to new blocking semantics without continuous buffer cleaning (temporary?) solution is to revert to non-blocking behavior in this case ------------------------------------------------------------------------ r516 | wdb | 2008-08-01 00:08:50 +0200 (Fri, 01 Aug 2008) | 3 lines fixed issues and verified that older (2.6.24.2) kernel still compile with new 2.6.26 changes ------------------------------------------------------------------------ r515 | wdb | 2008-07-31 19:59:53 +0200 (Thu, 31 Jul 2008) | 13 lines MAJOR: updated streamline to work with Linux kernel 2.6.26 specific changes for 2.6.26: * sysfs class_device was replaced by device * had to change from vm_operations_struct->nopage to ->fault * for the same: had to drop NOPAGE_SIGBUS specific changes for 2.6.25: * use struct nameidata.path instead of .mnt and .dentry members * ip_route_output_key takes another (default) parameter * netfilter NF_IP_(PRE|POST)_ROUTING labels changes ------------------------------------------------------------------------ r514 | wdb | 2008-07-31 15:49:03 +0200 (Thu, 31 Jul 2008) | 4 lines minor bugfix: backing_del only freed one of the four backing buffers this commit petrifies the revision of the july 2008 bufscale tests (src/apps/benchmark/buffer.bin) ------------------------------------------------------------------------ r513 | wdb | 2008-07-30 18:59:00 +0200 (Wed, 30 Jul 2008) | 2 lines retested pcap; fixed minor issue for this ------------------------------------------------------------------------ r512 | wdb | 2008-07-10 16:45:51 +0200 (Thu, 10 Jul 2008) | 2 lines mino ------------------------------------------------------------------------ r511 | wdb | 2008-07-10 15:31:14 +0200 (Thu, 10 Jul 2008) | 5 lines bugfix: pipesfs: delayed filter_update may refer to freed() inode bugfix: netfilter register/deregister could sleep in atomic context bugfix: filter_update may have been called with NULL due to slrun verifying parent instead of real node ------------------------------------------------------------------------ r509 | wdb | 2008-07-09 20:33:35 +0200 (Wed, 09 Jul 2008) | 3 lines minor update to documentation ------------------------------------------------------------------------ r508 | wdb | 2008-07-09 11:37:24 +0200 (Wed, 09 Jul 2008) | 2 lines minor ------------------------------------------------------------------------ r507 | wdb | 2008-07-09 11:06:59 +0200 (Wed, 09 Jul 2008) | 2 lines mino ------------------------------------------------------------------------ r506 | wdb | 2008-07-09 10:34:01 +0200 (Wed, 09 Jul 2008) | 3 lines CRITICAL fix: uninstall no longer wipes your drive with the wrong settings ------------------------------------------------------------------------ r504 | wdb | 2008-07-08 10:56:26 +0200 (Tue, 08 Jul 2008) | 3 lines minor ------------------------------------------------------------------------ r503 | wdb | 2008-07-08 10:39:51 +0200 (Tue, 08 Jul 2008) | 3 lines minor fixes to 1.7.4. will become 1.7.4.1 ------------------------------------------------------------------------ r501 | wdb | 2008-07-07 18:13:35 +0200 (Mon, 07 Jul 2008) | 2 lines release ready ------------------------------------------------------------------------ r500 | wdb | 2008-07-07 18:08:26 +0200 (Mon, 07 Jul 2008) | 3 lines last bugs out ------------------------------------------------------------------------ r499 | wdb | 2008-07-07 16:19:07 +0200 (Mon, 07 Jul 2008) | 2 lines minor ------------------------------------------------------------------------ r498 | wdb | 2008-07-07 14:06:53 +0200 (Mon, 07 Jul 2008) | 4 lines BUGFIX: in pipe_stream, readers were not correctly identified (O_RDWR) BUGFIX: in beltwayfs:poll, waiting on both reading and writing caused deadlock ------------------------------------------------------------------------ r497 | wdb | 2008-07-07 10:34:17 +0200 (Mon, 07 Jul 2008) | 2 lines minor ------------------------------------------------------------------------ r496 | wdb | 2008-07-07 08:32:45 +0200 (Mon, 07 Jul 2008) | 3 lines minor ------------------------------------------------------------------------ r495 | wdb | 2008-07-06 12:21:49 +0200 (Sun, 06 Jul 2008) | 2 lines minor ------------------------------------------------------------------------ r494 | wdb | 2008-07-06 12:13:37 +0200 (Sun, 06 Jul 2008) | 5 lines BUGFIX fixed test/mkfifo.bin regression : permission flags set incorrectly (hex vs. octal) BUGFIX fixed test/write_after_read.bin regression : write beyond 2KB failed, ... ------------------------------------------------------------------------ r493 | wdb | 2008-07-04 19:00:17 +0200 (Fri, 04 Jul 2008) | 2 lines fixed pipes over beltway issue. read/write now works, select still has a bug ------------------------------------------------------------------------ r492 | wdb | 2008-07-04 18:30:15 +0200 (Fri, 04 Jul 2008) | 3 lines fixed multiple locking/contention/sleeping issues in pipesfs and slsignal ------------------------------------------------------------------------ r491 | wdb | 2008-07-04 17:17:16 +0200 (Fri, 04 Jul 2008) | 2 lines Fixed various pipesfs bugs. ------------------------------------------------------------------------ r490 | wdb | 2008-07-03 21:33:43 +0200 (Thu, 03 Jul 2008) | 3 lines verified that http and get filters work ------------------------------------------------------------------------ r489 | wdb | 2008-07-03 21:24:28 +0200 (Thu, 03 Jul 2008) | 3 lines fixed many locking/concurrency/other crappy issues. not done yet ------------------------------------------------------------------------ r488 | wdb | 2008-07-03 10:23:41 +0200 (Thu, 03 Jul 2008) | 2 lines minor ------------------------------------------------------------------------ r487 | wdb | 2008-07-03 00:46:16 +0200 (Thu, 03 Jul 2008) | 3 lines minor ------------------------------------------------------------------------ r486 | wdb | 2008-07-02 23:30:06 +0200 (Wed, 02 Jul 2008) | 2 lines fixed stripheader, http and httpplus filters ------------------------------------------------------------------------ r485 | wdb | 2008-07-02 18:03:20 +0200 (Wed, 02 Jul 2008) | 2 lines ran test/*, fixed regressions (including bad ones) ------------------------------------------------------------------------ r484 | wdb | 2008-07-02 17:00:24 +0200 (Wed, 02 Jul 2008) | 2 lines fixed CPU soft lockup in task code. ------------------------------------------------------------------------ r483 | wdb | 2008-07-01 17:18:14 +0200 (Tue, 01 Jul 2008) | 2 lines minor update ------------------------------------------------------------------------ r482 | wdb | 2008-07-01 15:53:02 +0200 (Tue, 01 Jul 2008) | 2 lines minor ------------------------------------------------------------------------ r481 | wdb | 2008-06-30 16:42:05 +0200 (Mon, 30 Jun 2008) | 2 lines some work on example functions for pipesfs: idebug,http,httpplus ------------------------------------------------------------------------ r480 | wdb | 2008-06-30 15:44:13 +0200 (Mon, 30 Jun 2008) | 3 lines inserted a README for pipesfs ------------------------------------------------------------------------ r479 | wdb | 2008-06-30 14:59:21 +0200 (Mon, 30 Jun 2008) | 3 lines debugging pipes over beltway (not done yet) ------------------------------------------------------------------------ r478 | wdb | 2008-06-30 11:43:33 +0200 (Mon, 30 Jun 2008) | 3 lines introduced 'ccat': a continuous cat tool for reading from pipesfs pipes ------------------------------------------------------------------------ r477 | wdb | 2008-06-30 11:36:09 +0200 (Mon, 30 Jun 2008) | 2 lines updated final part of pipes over beltway ------------------------------------------------------------------------ r476 | wdb | 2008-06-29 22:43:40 +0200 (Sun, 29 Jun 2008) | 3 lines updated first half of 'pipes over beltway' mapped pipe-access ------------------------------------------------------------------------ r475 | wdb | 2008-06-29 19:39:29 +0200 (Sun, 29 Jun 2008) | 3 lines minor bugfix: removed error messages in bufmain_exit on common (and harmless) exception clause ------------------------------------------------------------------------ r474 | wdb | 2008-06-29 19:20:37 +0200 (Sun, 29 Jun 2008) | 12 lines added splicing benchmark and optimized the new implementation somewhat. speedup is about 3x, slowdown in case of miss ~22% for this NEW_WAR implementation. speedup is about 4x, slowdown in case of miss ~5% for the old OLD_WAR implementation. In other words: the other version is faster, but this version has a wider scope: -- includes splicing from any DBuf -- includes splicing from more N of the last-used application buffers OLD_WAR only allows splicing from one last-used DBuf element as observed in __peek_stream(); additionally, fixed a longstanding, but minor bug: on registration of a NULL buffer, name registration was tried and "dict NULL key insert thwarted" was shown. ------------------------------------------------------------------------ r473 | wdb | 2008-06-29 17:13:29 +0200 (Sun, 29 Jun 2008) | 3 lines committed new type splicing together with unittest ------------------------------------------------------------------------ r472 | wdb | 2008-06-28 19:50:45 +0200 (Sat, 28 Jun 2008) | 2 lines minor ------------------------------------------------------------------------ r471 | wdb | 2008-06-28 19:12:45 +0200 (Sat, 28 Jun 2008) | 2 lines pipesfs: all works besides beltway link ------------------------------------------------------------------------ r470 | wdb | 2008-06-28 16:58:26 +0200 (Sat, 28 Jun 2008) | 3 lines minor updates to pipesfs ------------------------------------------------------------------------ r469 | wdb | 2008-06-27 19:00:11 +0200 (Fri, 27 Jun 2008) | 3 lines worked on pipesfs:substreams fixed bug in netmonfs ------------------------------------------------------------------------ r468 | wdb | 2008-06-27 16:43:36 +0200 (Fri, 27 Jun 2008) | 2 lines minor update to pipes ------------------------------------------------------------------------ r467 | wdb | 2008-06-26 21:55:29 +0200 (Thu, 26 Jun 2008) | 4 lines updated pipesfs: files and pipes temporarily disabled special files (all, .available, .expression) ------------------------------------------------------------------------ r466 | wdb | 2008-06-26 12:58:23 +0200 (Thu, 26 Jun 2008) | 2 lines changes to pipesfs:file ------------------------------------------------------------------------ r465 | wdb | 2008-06-25 16:29:17 +0200 (Wed, 25 Jun 2008) | 2 lines minor fix on pipesfs .available ------------------------------------------------------------------------ r464 | wdb | 2008-06-25 16:19:37 +0200 (Wed, 25 Jun 2008) | 2 lines update to pipefs: symlink and pipe inode ------------------------------------------------------------------------ r463 | wdb | 2008-06-20 20:44:31 +0200 (Fri, 20 Jun 2008) | 3 lines updated the tg3 patch to use the new device driver interface ------------------------------------------------------------------------ r462 | wdb | 2008-06-19 18:02:04 +0200 (Thu, 19 Jun 2008) | 6 lines iconintued on pipesfs wq y ------------------------------------------------------------------------ r461 | wdb | 2008-06-19 15:02:07 +0200 (Thu, 19 Jun 2008) | 2 lines minor ------------------------------------------------------------------------ r460 | wdb | 2008-06-19 14:50:32 +0200 (Thu, 19 Jun 2008) | 3 lines heavily annotated pipesfs for debugging ------------------------------------------------------------------------ r459 | wdb | 2008-06-18 10:16:01 +0200 (Wed, 18 Jun 2008) | 2 lines minor ------------------------------------------------------------------------ r458 | wdb | 2008-06-17 14:11:18 +0200 (Tue, 17 Jun 2008) | 2 lines finished initial import of pipesfs. needs debugging ------------------------------------------------------------------------ r457 | wdb | 2008-06-16 16:42:24 +0200 (Mon, 16 Jun 2008) | 2 lines updated pipesfs ------------------------------------------------------------------------ r456 | wdb | 2008-06-16 01:03:34 +0200 (Mon, 16 Jun 2008) | 2 lines more pipesfs work ------------------------------------------------------------------------ r455 | wdb | 2008-06-16 00:54:12 +0200 (Mon, 16 Jun 2008) | 2 lines more work on pipesfs ------------------------------------------------------------------------ r454 | wdb | 2008-06-14 22:42:56 +0200 (Sat, 14 Jun 2008) | 3 lines imported pipesfs skeleton ------------------------------------------------------------------------ r453 | wdb | 2008-06-13 18:51:54 +0200 (Fri, 13 Jun 2008) | 2 lines fixed write-after-read regression: didn't check whether the iptr matched the block ------------------------------------------------------------------------ r452 | wdb | 2008-06-06 01:06:00 +0200 (Fri, 06 Jun 2008) | 2 lines still not done ------------------------------------------------------------------------ r451 | wdb | 2008-06-05 19:57:34 +0200 (Thu, 05 Jun 2008) | 2 lines worked on pipe (again... sigh) ------------------------------------------------------------------------ r450 | wdb | 2008-06-04 18:57:03 +0200 (Wed, 04 Jun 2008) | 2 lines minor ------------------------------------------------------------------------ r449 | wdb | 2008-06-04 00:17:13 +0200 (Wed, 04 Jun 2008) | 6 lines New: polling through fs node now also supports blocking write renamed the polling fs beltwayfs (was pollsockfs), as it now supports generic buffers Still working on getting the pipe benchmark to work, not done yet ------------------------------------------------------------------------ r448 | wdb | 2008-06-03 15:42:27 +0200 (Tue, 03 Jun 2008) | 2 lines hacking away, probably broken for now ------------------------------------------------------------------------ r447 | wdb | 2008-06-01 23:52:38 +0200 (Sun, 01 Jun 2008) | 3 lines minor: updated write-after-read test to use in-kernel IBuf for transmission ------------------------------------------------------------------------ r446 | wdb | 2008-06-01 23:35:23 +0200 (Sun, 01 Jun 2008) | 2 lines fixed up the copy avoidance test (src/test/write_after_read.c) ------------------------------------------------------------------------ r445 | wdb | 2008-05-30 19:36:00 +0200 (Fri, 30 May 2008) | 3 lines configuration used during final tcpdump benchmarks ------------------------------------------------------------------------ r444 | wdb | 2008-05-29 22:45:50 +0200 (Thu, 29 May 2008) | 2 lines minor ------------------------------------------------------------------------ r443 | wdb | 2008-05-29 22:31:57 +0200 (Thu, 29 May 2008) | 2 lines bugfix: pcap_read using read_stream ------------------------------------------------------------------------ r442 | wdb | 2008-05-29 14:22:06 +0200 (Thu, 29 May 2008) | 2 lines minor changes, addition of blocking read. unfortunately, read_stream now seems broken ------------------------------------------------------------------------ r441 | wdb | 2008-05-29 00:33:45 +0200 (Thu, 29 May 2008) | 3 lines fixed one bug in pipe_test: streamline stuff was never closed ------------------------------------------------------------------------ r440 | wdb | 2008-05-28 22:25:30 +0200 (Wed, 28 May 2008) | 5 lines version with which tcpdump benchmarks were run + additional changes: - fixed huge race in fork_stream ------------------------------------------------------------------------ r439 | wdb | 2008-05-28 18:12:04 +0200 (Wed, 28 May 2008) | 2 lines minor ------------------------------------------------------------------------ r438 | wdb | 2008-05-27 15:54:58 +0200 (Tue, 27 May 2008) | 3 lines (minor) reduced unnecessary occurrence of the "warning: diverging inputspaces" warning ------------------------------------------------------------------------ r437 | wdb | 2008-05-26 20:50:20 +0200 (Mon, 26 May 2008) | 8 lines (minor) update to get pcap parallel execution perfect * modified buf_dup to set new buffer offsets to that of the current write pointer * modified cli to be able to handle this new situation (by seeking to historical data) fixed apparently longstanding, but hidden, bug in test/file ------------------------------------------------------------------------ r436 | wdb | 2008-05-26 19:29:09 +0200 (Mon, 26 May 2008) | 5 lines PCAP: improved backend for new tests -- switched to real polling on pollsockfs, instead of usleep -- moved all previously global state into per-pcap session structure ------------------------------------------------------------------------ r435 | wdb | 2008-05-21 20:05:11 +0200 (Wed, 21 May 2008) | 1 line ------------------------------------------------------------------------ r434 | wdb | 2008-05-21 20:03:24 +0200 (Wed, 21 May 2008) | 3 lines minor update ------------------------------------------------------------------------ r433 | wdb | 2008-05-21 19:43:18 +0200 (Wed, 21 May 2008) | 5 lines fork_stream finally works! one caveat: some state has to be cleaned up by the garbage collector when a connection closes ------------------------------------------------------------------------ r432 | wdb | 2008-05-20 23:54:58 +0200 (Tue, 20 May 2008) | 2 lines fork_stream is still not done, but getting there... ------------------------------------------------------------------------ r431 | wdb | 2008-05-20 21:43:13 +0200 (Tue, 20 May 2008) | 3 lines fork_stream still fails, but passes test 2 (of 3). ------------------------------------------------------------------------ r430 | wdb | 2008-05-20 17:05:09 +0200 (Tue, 20 May 2008) | 2 lines fork_stream still broken. halfway commit ------------------------------------------------------------------------ r429 | wdb | 2008-05-09 15:17:58 +0200 (Fri, 09 May 2008) | 4 lines tested fork_stream. doesn't hang as badly anymore (fix in previous commit), but fork with kernelspace ring still fails ------------------------------------------------------------------------ r428 | wdb | 2008-05-09 14:51:17 +0200 (Fri, 09 May 2008) | 2 lines bugfix: cbuf_write used buf->offset instead of buf->shared->offset somewhere ------------------------------------------------------------------------ r427 | wdb | 2008-05-09 10:47:24 +0200 (Fri, 09 May 2008) | 5 lines fork_stream fixes: added reference counts to sysfs_up and sysfs_up_msg pipes. disabled sysfs_up waitqueues while debugging found a bug: waitqueue was never used! ------------------------------------------------------------------------ r426 | wdb | 2008-05-08 10:38:25 +0200 (Thu, 08 May 2008) | 7 lines implemented extensive fork_stream testcase in test/fork. userspace is fine. now about to test kernelspace MINOR: extended dictionary selftest with support for new dict_copy and dict_clear fixed some dictionary bugs along the way ------------------------------------------------------------------------ r425 | wdb | 2008-05-07 23:24:35 +0200 (Wed, 07 May 2008) | 3 lines NEW: fork_stream: forking so that streamline won't crash ------------------------------------------------------------------------ r424 | wdb | 2008-05-03 12:13:57 +0200 (Sat, 03 May 2008) | 7 lines updated pipe tests: pipe_posix now works correctly: fixed many race conditions pipe_pthread now works correctly (?): fixed memory access errors and deadlocks pipe is still not ready: correct behaviour requires fork_stream ------------------------------------------------------------------------ r423 | wdb | 2008-05-01 13:55:19 +0200 (Thu, 01 May 2008) | 6 lines MAJOR: reinserted the resizable buffer test MINOR: separated out benchmark and 'extra' applications in build process fixed resizable buffer test bitrot fixed resizable buffer bug (read in freed() region) ------------------------------------------------------------------------ r422 | wdb | 2008-04-29 16:55:01 +0200 (Tue, 29 Apr 2008) | 2 lines modified buffer size defaults ------------------------------------------------------------------------ r421 | wdb | 2008-04-25 22:25:12 +0200 (Fri, 25 Apr 2008) | 6 lines Socket fixes: - now uses linux support to generate destination MAC address - correct behaviour of ip id field - transmit bug fixed in select that made all slsockets return ready ------------------------------------------------------------------------ r420 | wdb | 2008-04-14 16:56:59 +0200 (Mon, 14 Apr 2008) | 2 lines minor fixes (fptr->pdu to fptr->data) ------------------------------------------------------------------------ r419 | wdb | 2008-04-14 07:43:19 +0200 (Mon, 14 Apr 2008) | 2 lines forcing make links in scripts/kernelbuild ------------------------------------------------------------------------ r418 | wdb | 2008-04-14 07:39:27 +0200 (Mon, 14 Apr 2008) | 2 lines bugfix in pcapout ------------------------------------------------------------------------ r417 | wdb | 2008-04-14 06:20:19 +0200 (Mon, 14 Apr 2008) | 2 lines expanded (inspect) capture length ------------------------------------------------------------------------ r416 | wdb | 2008-04-14 06:12:32 +0200 (Mon, 14 Apr 2008) | 2 lines bugfix and removal of debug statement. thanks to xingang ------------------------------------------------------------------------ r415 | wdb | 2008-04-14 01:26:34 +0200 (Mon, 14 Apr 2008) | 2 lines more changes due to xingang's bug reports ------------------------------------------------------------------------ r414 | wdb | 2008-04-11 22:33:57 +0200 (Fri, 11 Apr 2008) | 2 lines fix ------------------------------------------------------------------------ r413 | wdb | 2008-04-11 22:18:57 +0200 (Fri, 11 Apr 2008) | 3 lines minor fix ------------------------------------------------------------------------ r412 | wdb | 2008-04-11 19:54:18 +0200 (Fri, 11 Apr 2008) | 3 lines pcap bugfix. bug reported by xingang ------------------------------------------------------------------------ r411 | wdb | 2008-03-29 12:50:48 +0100 (Sat, 29 Mar 2008) | 4 lines updated select: __fast should be faster, __slow has been dropped updated socket vfs: now this is loaded/unloaded automatically in the scripts ------------------------------------------------------------------------ r410 | wdb | 2008-03-28 22:21:49 +0100 (Fri, 28 Mar 2008) | 6 lines New: socket slow path. without streamline-specific device drivers sockets no longer fail, but use a slow-path based on netfilter and skb_transmit Fixed: had to fix (transmit) ------------------------------------------------------------------------ r409 | wdb | 2008-03-28 17:47:21 +0100 (Fri, 28 Mar 2008) | 2 lines minor ------------------------------------------------------------------------ r408 | wdb | 2008-03-28 16:49:56 +0100 (Fri, 28 Mar 2008) | 2 lines minor fix for i386 ------------------------------------------------------------------------ r407 | wdb | 2008-03-27 21:19:05 +0100 (Thu, 27 Mar 2008) | 10 lines NEW: Unix commands are now integrated as Streamline filters thus (ps, expression="ax") will call the ps command. Fixed: the (shell) command required some fixes on SRL request failure the feedback is a bit more verbose: names instead of numbers all CLI commands are now added to the log, to make them more legible columns are now nicely aligned on `list ` in the CLI ------------------------------------------------------------------------ r406 | wdb | 2008-03-26 20:34:15 +0100 (Wed, 26 Mar 2008) | 6 lines NEW: function aliasing allows specific functions to pretend to be generic names (e.g., (e1000.rx.1) is (rx)) UPDATE: shell function asynchronous programming this is in preparation of transparent shell integration (e.g., (grep,expression="-nrI hello *")) ------------------------------------------------------------------------ r405 | wdb | 2008-03-25 19:10:39 +0100 (Tue, 25 Mar 2008) | 4 lines inserted "slpreload" script that makes it easier to call a binary through streamline minor modification to wrap/socket.c: removed debug statement ------------------------------------------------------------------------ r404 | wdb | 2008-03-25 18:41:14 +0100 (Tue, 25 Mar 2008) | 3 lines added support for optional components (Sun RPC) ------------------------------------------------------------------------ r403 | wdb | 2008-03-25 12:57:09 +0100 (Tue, 25 Mar 2008) | 2 lines merging changesets, compiles cleanly ------------------------------------------------------------------------ r402 | wdb | 2008-03-25 12:53:50 +0100 (Tue, 25 Mar 2008) | 3 lines dunno ------------------------------------------------------------------------ r401 | wdb | 2008-03-18 10:22:34 +0100 (Tue, 18 Mar 2008) | 9 lines NEW: * copybreak in the device driver, transparent to devices CLEANUP: * reduced code in headerfiles * core/datapath * functions/tcpreassembly (some, a long way to go) ------------------------------------------------------------------------ r400 | wdb | 2008-03-07 10:14:43 +0100 (Fri, 07 Mar 2008) | 2 lines minor: made dns benchmark client fire requests at a more stable rate ------------------------------------------------------------------------ r399 | wdb | 2008-03-06 19:18:24 +0100 (Thu, 06 Mar 2008) | 2 lines minor ------------------------------------------------------------------------ r398 | wdb | 2008-03-04 20:46:24 +0100 (Tue, 04 Mar 2008) | 2 lines minor ------------------------------------------------------------------------ r397 | wdb | 2008-03-04 18:54:00 +0100 (Tue, 04 Mar 2008) | 7 lines (belongs to previous) - added clean IP_CSUM, TCP_SUM and TSO interface on sldriver's tx path - 'fixed' tg3 tx UDP checksumming bug. UDP now zeroes the checksum - fixed tg3 tx IP checksumming bug ------------------------------------------------------------------------ r396 | wdb | 2008-03-04 18:51:26 +0100 (Tue, 04 Mar 2008) | 10 lines bugfixes: * application hung on close in some instances - incorrect route removal. - the bug mentioned earlier that I hacked around. this is the proper fix - fuckin' hard to track down. * logfiles has mode 000 (inaccessible). - longstanding - turned out I didn't read mode_t with O_CREAT, but with !O_CREAT in orig_open wq ------------------------------------------------------------------------ r395 | wdb | 2008-03-03 20:19:41 +0100 (Mon, 03 Mar 2008) | 3 lines (belongs to previous: tg3 patch update) ------------------------------------------------------------------------ r394 | wdb | 2008-03-03 20:09:45 +0100 (Mon, 03 Mar 2008) | 12 lines socket bitrot fix (and other bugs) * could not read from pbuf, because - comparison against unsigned in fbuf_readindexed should be signed - buf_rx was used, buf_main id written in iptr * sysf_area unmap failed for one buffer. disabled - come to think of it, might be double map because of above. will then reenable * `list buffers` crashed because __close_mapped didn't remove an unmapped item from the list * out of memory condition occurred with tg3 driver ------------------------------------------------------------------------ r393 | wdb | 2008-03-02 19:17:24 +0100 (Sun, 02 Mar 2008) | 11 lines MAJOR: - Fixed tg3 device driver. all tests (alloc, linux rx, linux tx, streamline rx, streamline tx, free, insmod, rmmod) succeed (should add locking on ifconfig down w/ simultaneous processing) - Modified device driver and block APIs to handle real traffic MINOR - added (generator_ip) filter, that sends valid ICMP echo requests - reran `make tests`. passed all after some fixes ------------------------------------------------------------------------ r392 | wdb | 2008-03-01 21:04:16 +0100 (Sat, 01 Mar 2008) | 7 lines updated tg3 device driver: - skblock_alloc, _getcontainer, and _free work - linux rx and tx work - streamline tx fails - streamline rx is untested ------------------------------------------------------------------------ r391 | wdb | 2008-02-29 17:19:37 +0100 (Fri, 29 Feb 2008) | 2 lines updated testset ------------------------------------------------------------------------ r390 | wdb | 2008-02-29 17:13:08 +0100 (Fri, 29 Feb 2008) | 3 lines minor: added some commentary to the loopsl device driver ------------------------------------------------------------------------ r389 | wdb | 2008-02-29 17:02:33 +0100 (Fri, 29 Feb 2008) | 3 lines updated tg3 driver: added DMA hw bug detection ------------------------------------------------------------------------ r388 | wdb | 2008-02-29 16:57:26 +0100 (Fri, 29 Feb 2008) | 3 lines committed updated tg3 device driveer ------------------------------------------------------------------------ r387 | wdb | 2008-02-28 22:59:05 +0100 (Thu, 28 Feb 2008) | 3 lines updated e1000 driver ------------------------------------------------------------------------ r386 | wdb | 2008-02-26 12:03:50 +0100 (Tue, 26 Feb 2008) | 3 lines kernel-close bug on x86_64 disappeared. mystery != good! ------------------------------------------------------------------------ r385 | wdb | 2008-02-26 11:31:17 +0100 (Tue, 26 Feb 2008) | 2 lines caching libc pointers ------------------------------------------------------------------------ r384 | wdb | 2008-02-25 22:54:08 +0100 (Mon, 25 Feb 2008) | 3 lines removed stale file NB: today I also fixed the LIBTOOL issue ------------------------------------------------------------------------ r383 | wdb | 2008-02-25 22:46:23 +0100 (Mon, 25 Feb 2008) | 3 lines hacked around kernel-close bug. found location of error, but don't have a proper fix yet ------------------------------------------------------------------------ r382 | wdb | 2008-02-25 20:08:54 +0100 (Mon, 25 Feb 2008) | 5 lines New e1000 driver * Uses the new sldevice API * Is untested. Compiles, but will probably break or hang ------------------------------------------------------------------------ r381 | wdb | 2008-02-25 19:45:33 +0100 (Mon, 25 Feb 2008) | 2 lines optional elements: openssl, pcap, readline dependencies compile on x86_64 ------------------------------------------------------------------------ r380 | wdb | 2008-02-25 19:25:10 +0100 (Mon, 25 Feb 2008) | 6 lines made x86_64 ready (compiles, tests succeed, but applications may still fail) made 2.6.24.2 ready removed the disk and cdisk buffers in kernelspace because they depend on sys_read, which is deprecated ------------------------------------------------------------------------ r379 | wdb | 2008-02-25 17:34:28 +0100 (Mon, 25 Feb 2008) | 2 lines minor ------------------------------------------------------------------------ r378 | wdb | 2008-02-25 17:33:51 +0100 (Mon, 25 Feb 2008) | 2 lines minor ------------------------------------------------------------------------ r377 | wdb | 2008-02-25 17:30:55 +0100 (Mon, 25 Feb 2008) | 2 lines minor ------------------------------------------------------------------------ r376 | wdb | 2008-02-25 17:30:19 +0100 (Mon, 25 Feb 2008) | 2 lines libtool updates ------------------------------------------------------------------------ r375 | wdb | 2008-02-25 17:21:29 +0100 (Mon, 25 Feb 2008) | 2 lines minor libtool changes ------------------------------------------------------------------------ r374 | wdb | 2008-02-19 19:19:36 +0100 (Tue, 19 Feb 2008) | 17 lines MAJOR inserted sldevice: stream device interface inserted slblock: stream allocation and modification interface MINOR added unittests for both modified configure.in to hopefully squash the libtool issue simplified fptr structure: * now similar to an iovec, but machine independent * used wherever it is useful, incl. expressions and static mem * this is the reason I have touched so many files in this revision fixed userspace buffer->data allocation to always be page_aligned merged slotted and bytestream buffer logic ------------------------------------------------------------------------ r373 | wdb | 2008-02-19 10:17:00 +0100 (Tue, 19 Feb 2008) | 2 lines new api ------------------------------------------------------------------------ r372 | wdb | 2008-02-12 18:57:39 +0100 (Tue, 12 Feb 2008) | 7 lines new: first pass modification API new: dummy compiler function new: bpf compiler function WONT COMPILE!! halfway through implementation of the driver API ------------------------------------------------------------------------ r371 | wdb | 2008-02-11 23:04:43 +0100 (Mon, 11 Feb 2008) | 3 lines tried to fix libtool issue that occurs on some distros ------------------------------------------------------------------------ r370 | cristea | 2008-01-09 14:17:13 +0100 (Wed, 09 Jan 2008) | 1 line Bugfix on boundchecking of variables: check1, check2, checkLoops ------------------------------------------------------------------------ r369 | wdb | 2007-12-04 08:08:32 +0100 (Tue, 04 Dec 2007) | 3 lines preliminary add-header hack to untcp ------------------------------------------------------------------------ r368 | wdb | 2007-12-03 01:47:41 +0100 (Mon, 03 Dec 2007) | 2 lines fixed recently introduced build error ------------------------------------------------------------------------ r367 | wdb | 2007-12-02 00:47:51 +0100 (Sun, 02 Dec 2007) | 3 lines updated CLI help and 'list' command output according to Mihai's feedback ------------------------------------------------------------------------ r366 | wdb | 2007-11-27 05:02:48 +0100 (Tue, 27 Nov 2007) | 2 lines added DEBUG warning ------------------------------------------------------------------------ r365 | wdb | 2007-11-27 04:46:31 +0100 (Tue, 27 Nov 2007) | 3 lines reintegrated userspace and kernel buildpaths (through hidden .src.kernel symlinks) ------------------------------------------------------------------------ r364 | wdb | 2007-11-26 18:08:59 +0100 (Mon, 26 Nov 2007) | 3 lines updated make install and make uninstall targets ------------------------------------------------------------------------ r363 | wdb | 2007-11-23 20:20:05 +0100 (Fri, 23 Nov 2007) | 2 lines BUGFIX: ring allocation in the kernel using __get_free_pages was non-mappable ------------------------------------------------------------------------ r362 | wdb | 2007-11-22 08:58:25 +0100 (Thu, 22 Nov 2007) | 3 lines bug'fix': small ring allocations fucked up the memory map ------------------------------------------------------------------------ r361 | wdb | 2007-11-15 23:55:13 +0100 (Thu, 15 Nov 2007) | 2 lines buggy? but best BIND score ------------------------------------------------------------------------ r360 | wdb | 2007-11-15 21:21:42 +0100 (Thu, 15 Nov 2007) | 3 lines minor change ------------------------------------------------------------------------ r359 | wdb | 2007-11-14 17:25:11 +0100 (Wed, 14 Nov 2007) | 5 lines FIX: file reading works again FIX: lseek caused infinite look (cbuf->lseek -> default_lseek -> ...) thanks to Heriberto Castillo's emails ------------------------------------------------------------------------ r358 | wdb | 2007-11-13 22:05:02 +0100 (Tue, 13 Nov 2007) | 3 lines just ran bind.local benchmarks ------------------------------------------------------------------------ r357 | wdb | 2007-11-13 10:28:40 +0100 (Tue, 13 Nov 2007) | 3 lines add: runtime modifiable "safety" variable (to tweak batching factor) add: reception stresstest ------------------------------------------------------------------------ r356 | wdb | 2007-11-11 21:34:00 +0100 (Sun, 11 Nov 2007) | 2 lines bind tests okay ------------------------------------------------------------------------ r355 | wdb | 2007-11-11 17:27:48 +0100 (Sun, 11 Nov 2007) | 2 lines bind performance is good ------------------------------------------------------------------------ r354 | wdb | 2007-11-09 18:14:25 +0100 (Fri, 09 Nov 2007) | 3 lines applied Sven Anderson's bugfix ------------------------------------------------------------------------ r353 | wdb | 2007-11-06 18:13:31 +0100 (Tue, 06 Nov 2007) | 3 lines mplayer tests shine! ------------------------------------------------------------------------ r352 | wdb | 2007-11-05 18:23:11 +0100 (Mon, 05 Nov 2007) | 4 lines tested, fixed and enabled gettimeofday() syscall replacement (wrap/other.c) fixed new batching mechanism ------------------------------------------------------------------------ r351 | wdb | 2007-11-05 12:42:13 +0100 (Mon, 05 Nov 2007) | 6 lines NEW: apps/nonsl/getrusage, a benchmarking shell application changed signal throttling method fixed races that only occurred at high rates ------------------------------------------------------------------------ r350 | wdb | 2007-11-04 00:41:37 +0100 (Sun, 04 Nov 2007) | 3 lines fixed race in interrupt-context/user-context signalling (kernel used to hang at high rates) minor fix-ups while benchmarking ------------------------------------------------------------------------ r349 | wdb | 2007-11-03 20:49:34 +0100 (Sat, 03 Nov 2007) | 2 lines fixed dns benchmarking client ------------------------------------------------------------------------ r348 | wdb | 2007-11-02 23:43:28 +0100 (Fri, 02 Nov 2007) | 2 lines minor ------------------------------------------------------------------------ r347 | wdb | 2007-11-02 23:28:57 +0100 (Fri, 02 Nov 2007) | 2 lines made pollsockfs and netmonfs hold a lock on the module: no rmmod while mounted ------------------------------------------------------------------------ r346 | wdb | 2007-11-02 21:48:49 +0100 (Fri, 02 Nov 2007) | 5 lines new: half streamline, half loopback network connection anything written to linux device sl0 ends up in streamline function loopsl_rx anything written to streamline function loopsl_tx ends up in sl0 ------------------------------------------------------------------------ r345 | wdb | 2007-11-02 14:56:59 +0100 (Fri, 02 Nov 2007) | 2 lines added a massively parallel DNS client to stress-test bind ------------------------------------------------------------------------ r344 | wdb | 2007-11-02 12:14:13 +0100 (Fri, 02 Nov 2007) | 3 lines made select_stream_fast return without a single context switch if data is ready add a test ------------------------------------------------------------------------ r343 | wdb | 2007-11-01 20:46:45 +0100 (Thu, 01 Nov 2007) | 3 lines make exportdata=yes filter data (class != 0) ------------------------------------------------------------------------ r342 | wdb | 2007-11-01 19:46:33 +0100 (Thu, 01 Nov 2007) | 2 lines minor update of pollsock ------------------------------------------------------------------------ r341 | wdb | 2007-11-01 17:02:56 +0100 (Thu, 01 Nov 2007) | 2 lines minor update ------------------------------------------------------------------------ r340 | wdb | 2007-11-01 15:44:31 +0100 (Thu, 01 Nov 2007) | 3 lines new: pollable streams (1/2) ------------------------------------------------------------------------ r339 | wdb | 2007-10-31 16:41:25 +0100 (Wed, 31 Oct 2007) | 2 lines tested with 2.6.15 again. had to fix support/timer delayed_func ------------------------------------------------------------------------ r338 | wdb | 2007-10-31 10:50:18 +0100 (Wed, 31 Oct 2007) | 3 lines minor fix ------------------------------------------------------------------------ r337 | wdb | 2007-10-31 10:46:01 +0100 (Wed, 31 Oct 2007) | 3 lines minor update ------------------------------------------------------------------------ r336 | wdb | 2007-10-31 09:28:55 +0100 (Wed, 31 Oct 2007) | 2 lines minor build fix ------------------------------------------------------------------------ r335 | wdb | 2007-10-27 00:13:02 +0200 (Sat, 27 Oct 2007) | 2 lines fixed usleep in socket ------------------------------------------------------------------------ r334 | wdb | 2007-10-26 21:28:12 +0200 (Fri, 26 Oct 2007) | 2 lines (skip) ------------------------------------------------------------------------ r333 | wdb | 2007-10-25 11:43:11 +0200 (Thu, 25 Oct 2007) | 4 lines bugfix: a signal triggering a signal registration led to signaltable corruptions bugfix: (minor) cli read error on file input ------------------------------------------------------------------------ r332 | wdb | 2007-10-24 14:48:40 +0200 (Wed, 24 Oct 2007) | 5 lines mplayer works new: incomplete sunrpc wrapper fixed: poll and poll_stream ------------------------------------------------------------------------ r331 | wdb | 2007-10-23 17:53:56 +0200 (Tue, 23 Oct 2007) | 4 lines new: getsockname() implementation new: gettimeofday test ------------------------------------------------------------------------ r330 | wdb | 2007-10-22 17:15:31 +0200 (Mon, 22 Oct 2007) | 4 lines BIND works! dummync works ------------------------------------------------------------------------ r329 | wdb | 2007-10-21 16:49:50 +0200 (Sun, 21 Oct 2007) | 4 lines bugfix socket recvmsg O_TSTAMP end-of-queue sign missing bind almost works now ------------------------------------------------------------------------ r328 | wdb | 2007-10-19 23:41:08 +0200 (Fri, 19 Oct 2007) | 2 lines progress. some. ------------------------------------------------------------------------ r327 | wdb | 2007-10-19 18:33:11 +0200 (Fri, 19 Oct 2007) | 4 lines BUGFIX: localspace_destroy didn't clean up all state BUGFIX: select_stream ------------------------------------------------------------------------ r326 | wdb | 2007-10-17 23:48:08 +0200 (Wed, 17 Oct 2007) | 6 lines new: dummync, a udp-only server-only netcat clone tested: select(), recvmsg(), recvfrom() breaks: cornercases in dummync send or process fixup: select(), poll() ------------------------------------------------------------------------ r325 | wdb | 2007-10-17 15:19:51 +0200 (Wed, 17 Oct 2007) | 6 lines simplified buffer import: removed src/core/remotebuf.[ch] reduced runtime signalhandler lookup codesize ibuf's readthrough function now always has a private bufferview ------------------------------------------------------------------------ r324 | wdb | 2007-10-16 22:13:23 +0200 (Tue, 16 Oct 2007) | 2 lines added dummy DNS server ------------------------------------------------------------------------ r323 | cristea | 2007-10-16 19:06:24 +0200 (Tue, 16 Oct 2007) | 2 lines Fixed 2 warnings treated as errors about uninitialsed data (variable sock) ------------------------------------------------------------------------ r322 | wdb | 2007-10-15 17:15:10 +0200 (Mon, 15 Oct 2007) | 3 lines fixed exportdata= and size= use ------------------------------------------------------------------------ r321 | wdb | 2007-10-13 13:45:59 +0200 (Sat, 13 Oct 2007) | 2 lines unstable ------------------------------------------------------------------------ r320 | wdb | 2007-10-13 01:34:49 +0200 (Sat, 13 Oct 2007) | 4 lines new: exportdata= flag to export directly to a databuffer (as opposed to index) new: size= flag to set the size of the buffer exported to ------------------------------------------------------------------------ r319 | wdb | 2007-10-12 22:02:34 +0200 (Fri, 12 Oct 2007) | 5 lines NEW: multithreaded operation -> removed sigsetjmp/siglongjmp -> replaced src/support/timer.[ch] and src/core/signal.[ch] api -> added new tests ------------------------------------------------------------------------ r318 | cristea | 2007-10-06 12:36:52 +0200 (Sat, 06 Oct 2007) | 1 line bugfix in memsearch function. ------------------------------------------------------------------------ r317 | wdb | 2007-10-03 15:49:08 +0200 (Wed, 03 Oct 2007) | 2 lines fixed race condition where the timer thread would not listen to an early kill signal ------------------------------------------------------------------------ r316 | cristea | 2007-10-03 13:00:38 +0200 (Wed, 03 Oct 2007) | 2 lines fpl_compiler update/bugfix added new fpl applications as examples ------------------------------------------------------------------------ r315 | wdb | 2007-10-03 11:49:22 +0200 (Wed, 03 Oct 2007) | 3 lines updated tg3 driver ------------------------------------------------------------------------ r314 | wdb | 2007-10-02 17:15:09 +0200 (Tue, 02 Oct 2007) | 3 lines add multithreading/multiprocessing test ------------------------------------------------------------------------ r313 | wdb | 2007-10-02 15:58:15 +0200 (Tue, 02 Oct 2007) | 2 lines minor: md5 kernel warnings fixed ------------------------------------------------------------------------ r312 | wdb | 2007-10-02 15:50:54 +0200 (Tue, 02 Oct 2007) | 6 lines new: TCP passthrough to non-streamline sockets new: clone() support (subject to flags) new: fork, vfork, execve hard errors new: setsockopt SOL_IP, IP_PKTINFO support ------------------------------------------------------------------------ r311 | cristea | 2007-10-02 11:05:02 +0200 (Tue, 02 Oct 2007) | 1 line Missing updates done for the FPL support made a day before. ------------------------------------------------------------------------ r310 | cristea | 2007-10-02 10:33:49 +0200 (Tue, 02 Oct 2007) | 1 line Missing updates done for the FPL support made a day before. ------------------------------------------------------------------------ r309 | wdb | 2007-10-02 09:57:27 +0200 (Tue, 02 Oct 2007) | 3 lines #define fix ------------------------------------------------------------------------ r308 | cristea | 2007-10-01 17:07:05 +0200 (Mon, 01 Oct 2007) | 1 line Put back the FPL compiler with support for streamline FPL functions and IXP1200, IXP2400 FIX framework. Check doc/README.fpl for more details. ------------------------------------------------------------------------ r307 | wdb | 2007-09-28 18:00:51 +0200 (Fri, 28 Sep 2007) | 2 lines fixing timer issue. not done yet ------------------------------------------------------------------------ r306 | wdb | 2007-09-28 17:00:04 +0200 (Fri, 28 Sep 2007) | 8 lines bind seems to be working new: setsockopt implemented with support for SO_BSDCOMPAT, SO_LINGER, SO_REUSEADDR, SO_TIMESTAMP new: getsockopt implemented with support for SO_RCVBUF new: ioctl support for SIOCGIFNETMASK new: recvmsg support for SO_TIMESTAMP fix: timer_del unnecessarily reset the timer when it was already fine-grained enough ------------------------------------------------------------------------ r305 | wdb | 2007-09-27 20:08:51 +0200 (Thu, 27 Sep 2007) | 6 lines new: timer using sysfs file to replace POSIX SIGALRM cleanup: signalling code new: kernel purge gets called by ./scripts/streamline stop ------------------------------------------------------------------------ r304 | wdb | 2007-09-27 14:22:39 +0200 (Thu, 27 Sep 2007) | 3 lines fix kernel compile on 2.6.19-fedora fix Makefile.in 'kernel' target ------------------------------------------------------------------------ r303 | wdb | 2007-09-27 14:11:59 +0200 (Thu, 27 Sep 2007) | 2 lines added signal testfile ------------------------------------------------------------------------ r302 | wdb | 2007-09-27 13:48:20 +0200 (Thu, 27 Sep 2007) | 4 lines added fcntl for sockets expanded ioctl for sockets ------------------------------------------------------------------------ r301 | wdb | 2007-09-26 13:11:52 +0200 (Wed, 26 Sep 2007) | 6 lines fix: first block after pbuf map-in was not handled in userspace fix: fbuf_write got broken when fbuf_acquired was used fix: remove spurious signals after unmap through new stream_del_signala fix: sysfs signalling still suffered from the buffering in sysfs attributes ------------------------------------------------------------------------ r300 | wdb | 2007-09-25 17:40:57 +0200 (Tue, 25 Sep 2007) | 5 lines removed bottlenecks in channel_sysfs and ..._up and ..._upcall fixed parallel upcall pipes (they illegally shared individual thread members) renamed ethernet to netfilter ------------------------------------------------------------------------ r299 | wdb | 2007-09-24 21:47:11 +0200 (Mon, 24 Sep 2007) | 12 lines trying to get (GNU) netcat working. not there yet new: context-switch free gettimeofday() (but is not accurate) new: ioctl() support for sockets new: stream_available estimates amount of waiting data new: setsockopt() supported fix: labelled edges were not used for prefix generation this could lead to incorrectly merged graphs fix: test/msg.bin hardcoded its space id ------------------------------------------------------------------------ r298 | wdb | 2007-09-24 10:51:20 +0200 (Mon, 24 Sep 2007) | 2 lines minor ------------------------------------------------------------------------ r297 | wdb | 2007-09-15 22:44:58 +0200 (Sat, 15 Sep 2007) | 2 lines made progress on socket reception. not completely okay yet ------------------------------------------------------------------------ r296 | wdb | 2007-09-15 20:46:10 +0200 (Sat, 15 Sep 2007) | 5 lines NEW: expanded socket api, with sendmsg/recvmsg, unconnected sendto, socket read/write support tested all functions for transmission on tg3 reception fails so far for lack of blocking behaviour ------------------------------------------------------------------------ r295 | wdb | 2007-09-15 12:46:55 +0200 (Sat, 15 Sep 2007) | 3 lines skip ------------------------------------------------------------------------ r294 | wdb | 2007-09-14 14:36:15 +0200 (Fri, 14 Sep 2007) | 2 lines remove stale dir ------------------------------------------------------------------------ r293 | wdb | 2007-09-14 14:35:52 +0200 (Fri, 14 Sep 2007) | 5 lines NEW broadcom ('tg3') driver with streamline support for kernel 2.6.22.4 updated the older intel pro/1000 driver as well. both now use the standard linux patch format ------------------------------------------------------------------------ r292 | wdb | 2007-09-10 17:51:19 +0200 (Mon, 10 Sep 2007) | 2 lines fixed up export for low-level driver support ------------------------------------------------------------------------ r291 | wdb | 2007-09-07 09:29:10 +0200 (Fri, 07 Sep 2007) | 2 lines tested compile on ubuntu feisty clean (2.6.20-?) ------------------------------------------------------------------------ r290 | wdb | 2007-09-06 16:57:42 +0200 (Thu, 06 Sep 2007) | 2 lines renamed transport.[ch] file.[ch] -- this better reflects the contents IMHO ------------------------------------------------------------------------ r289 | wdb | 2007-09-06 15:49:34 +0200 (Thu, 06 Sep 2007) | 8 lines NEW: LD_PRELOAD support for sockets and files moved to using LIBTOOL for linking added core/wrap.[ch] for redirection moved all unit_tests into separate files under src/test added __close_socket to cleanly use close() on sockets fixed username detection in support/log.c ------------------------------------------------------------------------ r288 | wdb | 2007-09-05 18:02:24 +0200 (Wed, 05 Sep 2007) | 2 lines fixed ntohs/ntohl functions ------------------------------------------------------------------------ r287 | wdb | 2007-09-05 17:40:23 +0200 (Wed, 05 Sep 2007) | 9 lines NEW: added libslsocket and libslfile wrappers for LD_PRELOAD this forced me to expand include/streamline/socket.h to include more definitions and no longer depend on glibc equivalents also, in many files the include paths had to be changed, hence the long changelist ------------------------------------------------------------------------ r286 | wdb | 2007-09-04 17:14:33 +0200 (Tue, 04 Sep 2007) | 2 lines added e1000 hack ------------------------------------------------------------------------ r285 | wdb | 2007-09-04 17:08:40 +0200 (Tue, 04 Sep 2007) | 2 lines minor ------------------------------------------------------------------------ r284 | wdb | 2007-09-03 16:06:54 +0200 (Mon, 03 Sep 2007) | 2 lines modified fbuf to work with hacked e1000 ------------------------------------------------------------------------ r283 | wdb | 2007-08-23 04:14:30 +0200 (Thu, 23 Aug 2007) | 1 line ------------------------------------------------------------------------ r281 | wdb | 2007-08-23 04:09:11 +0200 (Thu, 23 Aug 2007) | 3 lines updated Changelog. this will be 1.7.2 ------------------------------------------------------------------------ r280 | wdb | 2007-08-23 03:49:19 +0200 (Thu, 23 Aug 2007) | 3 lines cleaned out non-src directories ------------------------------------------------------------------------ r279 | wdb | 2007-08-23 03:41:53 +0200 (Thu, 23 Aug 2007) | 3 lines removed some more proprietary code for which we have no distribution license ------------------------------------------------------------------------ r278 | wdb | 2007-08-23 03:23:44 +0200 (Thu, 23 Aug 2007) | 3 lines removed proprietary code that I didn't even know was there ------------------------------------------------------------------------ r277 | wdb | 2007-08-23 02:49:08 +0200 (Thu, 23 Aug 2007) | 3 lines replaced the read-after-write hack with a radix-tree based lookup to support transparent splicing when writing to an ibuf ------------------------------------------------------------------------ r276 | wdb | 2007-08-23 00:34:41 +0200 (Thu, 23 Aug 2007) | 2 lines fixed 2.6.17 build issue in netmonfs ------------------------------------------------------------------------ r275 | wdb | 2007-08-23 00:21:08 +0200 (Thu, 23 Aug 2007) | 1 line ------------------------------------------------------------------------ r274 | wdb | 2007-08-22 20:50:49 +0200 (Wed, 22 Aug 2007) | 3 lines updated cli help screen removed stale cli options ------------------------------------------------------------------------ r273 | wdb | 2007-08-22 20:12:57 +0200 (Wed, 22 Aug 2007) | 3 lines added support for kernel 2.6.22 ------------------------------------------------------------------------ r272 | wdb | 2007-08-22 16:55:45 +0200 (Wed, 22 Aug 2007) | 2 lines added radix tree implementation ------------------------------------------------------------------------ r271 | wdb | 2007-08-22 00:15:53 +0200 (Wed, 22 Aug 2007) | 2 lines minor bugfix in new netmonfs application ------------------------------------------------------------------------ r270 | wdb | 2007-08-21 23:52:44 +0200 (Tue, 21 Aug 2007) | 5 lines (MINOR) updated `make install' removed stale script ------------------------------------------------------------------------ r269 | wdb | 2007-08-21 23:31:52 +0200 (Tue, 21 Aug 2007) | 2 lines ? ------------------------------------------------------------------------ r268 | wdb | 2007-08-17 12:27:03 +0200 (Fri, 17 Aug 2007) | 2 lines temporarily disabled in-kernel disk_stream support to be able to build on FC6 ------------------------------------------------------------------------ r267 | wdb | 2007-08-11 01:54:48 +0200 (Sat, 11 Aug 2007) | 5 lines finalized the fbufdbufvbuf test. saved state shows pathological case for malloc bugfix: on buffer removal, its signals are stopped (was a read-after-free bug) bugfix: some illegal accesses valgrind found in vbuf ------------------------------------------------------------------------ r266 | wdb | 2007-08-10 19:25:57 +0200 (Fri, 10 Aug 2007) | 1 line ------------------------------------------------------------------------ r265 | wdb | 2007-08-10 19:23:44 +0200 (Fri, 10 Aug 2007) | 2 lines fixup read/write to special files ------------------------------------------------------------------------ r264 | wdb | 2007-08-10 19:09:23 +0200 (Fri, 10 Aug 2007) | 4 lines NEW APP: netmonfs, a filesystem interface to streamline. see src/apps/netmonfs.c for more info BUGFIX: buffer access control has finally been implemented. Note that remapping still requires root perms. ------------------------------------------------------------------------ r263 | wdb | 2007-08-09 18:26:39 +0200 (Thu, 09 Aug 2007) | 5 lines ran more fbuf/vbuf/dbuf tests BUGFIX fixed a few cornercase bugs in vbuf and dbuf NEW added a "malloc" buffer that does not use contiguous regions, mostly for benchmarking ------------------------------------------------------------------------ r262 | wdb | 2007-08-03 17:59:43 +0200 (Fri, 03 Aug 2007) | 3 lines BUGFIX: insertion of request with +1 rootnodes would cause segfault ------------------------------------------------------------------------ r261 | mtaneja | 2007-07-24 20:57:34 +0200 (Tue, 24 Jul 2007) | 6 lines intermediate checkin added track_protos.c acsmx.h ------------------------------------------------------------------------ r260 | wdb | 2007-07-20 17:58:16 +0200 (Fri, 20 Jul 2007) | 2 lines updated vbuf and dbuf. started writing fbuf.vs.vbuf.vs.dbuf test ------------------------------------------------------------------------ r259 | wdb | 2007-07-19 16:40:19 +0200 (Thu, 19 Jul 2007) | 2 lines moved back to development build. new labels are 1.7.2-DATETIME ------------------------------------------------------------------------ r258 | wdb | 2007-07-19 16:39:03 +0200 (Thu, 19 Jul 2007) | 3 lines created 1.7.1 tag ------------------------------------------------------------------------ r257 | wdb | 2007-07-19 16:34:41 +0200 (Thu, 19 Jul 2007) | 6 lines updated the Changelog updated the TODO BUGFIX: got "read" and "data" operations in the cli working again this will be 1.7.1 ------------------------------------------------------------------------ r256 | wdb | 2007-07-19 16:00:52 +0200 (Thu, 19 Jul 2007) | 3 lines minor fix for RELEASE build ------------------------------------------------------------------------ r255 | wdb | 2007-07-19 15:55:03 +0200 (Thu, 19 Jul 2007) | 4 lines enabled ipfix purging fixed some last (?) issues in ipfix ------------------------------------------------------------------------ r254 | mtaneja | 2007-07-19 01:55:22 +0200 (Thu, 19 Jul 2007) | 3 lines added readme file to tracker app added partial support for bit torrent ------------------------------------------------------------------------ r253 | wdb | 2007-07-18 18:34:37 +0200 (Wed, 18 Jul 2007) | 2 lines updated the TODO ------------------------------------------------------------------------ r252 | wdb | 2007-07-18 17:13:38 +0200 (Wed, 18 Jul 2007) | 4 lines BUGFIX pcap now handles input and output BUGFIX removed warning on double unmap. probably just a band-aid ------------------------------------------------------------------------ r251 | wdb | 2007-07-18 16:00:13 +0200 (Wed, 18 Jul 2007) | 4 lines ipfix.bin also works with function in userspace minor bugfix: [enter] in cli no longer gives an error ------------------------------------------------------------------------ r250 | wdb | 2007-07-18 15:25:03 +0200 (Wed, 18 Jul 2007) | 4 lines BUGFIX: all remaining timers are killed on module onload ipfix works (with the function in the kernel. must still try userspace) ------------------------------------------------------------------------ r249 | wdb | 2007-07-17 18:35:50 +0200 (Tue, 17 Jul 2007) | 2 lines fixed up IPFIX ------------------------------------------------------------------------ r248 | wdb | 2007-07-16 17:50:04 +0200 (Mon, 16 Jul 2007) | 2 lines ipfix updated. unstable ------------------------------------------------------------------------ r247 | wdb | 2007-07-16 14:59:27 +0200 (Mon, 16 Jul 2007) | 2 lines (skip) ------------------------------------------------------------------------ r246 | mtaneja | 2007-07-16 12:43:49 +0200 (Mon, 16 Jul 2007) | 2 lines tracker implementation complete ------------------------------------------------------------------------ r245 | wdb | 2007-07-16 11:34:49 +0200 (Mon, 16 Jul 2007) | 2 lines special cased func.newex to allow nested operation ------------------------------------------------------------------------ r244 | wdb | 2007-07-12 14:19:20 +0200 (Thu, 12 Jul 2007) | 2 lines updated TODO ------------------------------------------------------------------------ r243 | wdb | 2007-07-12 01:06:56 +0200 (Thu, 12 Jul 2007) | 1 line ------------------------------------------------------------------------ r242 | wdb | 2007-07-11 14:51:32 +0200 (Wed, 11 Jul 2007) | 2 lines updated fauxresbuf dynamic test to show expected behaviour ------------------------------------------------------------------------ r241 | wdb | 2007-07-10 19:49:47 +0200 (Tue, 10 Jul 2007) | 4 lines added fauxresbuf gnuplot output and fixed some bugs along the way (and tweaked a bit) ------------------------------------------------------------------------ r240 | wdb | 2007-07-10 14:39:29 +0200 (Tue, 10 Jul 2007) | 2 lines removed localspace test that failed because of changes in startup/shutdown ------------------------------------------------------------------------ r239 | wdb | 2007-07-10 14:20:50 +0200 (Tue, 10 Jul 2007) | 2 lines bugfix: recently introduced issue (skip) ------------------------------------------------------------------------ r238 | wdb | 2007-07-10 13:33:27 +0200 (Tue, 10 Jul 2007) | 2 lines bugfix: (bug introduced this morning) close_stream with NULL name ------------------------------------------------------------------------ r237 | wdb | 2007-07-10 13:15:08 +0200 (Tue, 10 Jul 2007) | 4 lines bugfix: userspace shutdown was not completed due to uncaught alarm. fixed cleanup: startup/shutdown has been simplified. only 1 kernelmodule left cleanup: messagehandlers now use the default startup/shutdown code ------------------------------------------------------------------------ r236 | wdb | 2007-07-09 20:16:41 +0200 (Mon, 09 Jul 2007) | 2 lines (skip) ------------------------------------------------------------------------ r235 | wdb | 2007-07-09 20:15:56 +0200 (Mon, 09 Jul 2007) | 2 lines bugfix: serialization bug in create_prefix ------------------------------------------------------------------------ r234 | wdb | 2007-07-09 20:10:39 +0200 (Mon, 09 Jul 2007) | 5 lines updated pseudo-resizable buffer (PSB) added PSB test bugfix: bufmain::cleanup was not called on close. is now for some apps ------------------------------------------------------------------------ r233 | wdb | 2007-07-09 18:15:35 +0200 (Mon, 09 Jul 2007) | 5 lines bugfix: collisions on prefix will now be extremely rare. using MD5 for expressions added Ron Rivest's default MD5 implementation bugfix: discover always returned success, so errors were caught only on insert ------------------------------------------------------------------------ r232 | wdb | 2007-07-07 18:01:48 +0200 (Sat, 07 Jul 2007) | 3 lines cleaned up testftp ------------------------------------------------------------------------ r231 | wdb | 2007-07-07 17:39:03 +0200 (Sat, 07 Jul 2007) | 4 lines BUGFIX: ethernet_out now has mac->h_proto set correctly NEW: ethernet_out now supports device filtering, like ethernet_in ------------------------------------------------------------------------ r230 | wdb | 2007-06-26 19:33:25 +0200 (Tue, 26 Jun 2007) | 2 lines BUGFIX: default size was never selected correctly ------------------------------------------------------------------------ r229 | mtaneja | 2007-06-26 18:06:52 +0200 (Tue, 26 Jun 2007) | 1 line ------------------------------------------------------------------------ r228 | wdb | 2007-06-25 13:47:34 +0200 (Mon, 25 Jun 2007) | 3 lines further optimized signal path ------------------------------------------------------------------------ r227 | wdb | 2007-06-25 02:32:20 +0200 (Mon, 25 Jun 2007) | 3 lines tweaked ringsize to better fit (my) tcpdump tests ------------------------------------------------------------------------ r226 | wdb | 2007-06-25 02:19:40 +0200 (Mon, 25 Jun 2007) | 5 lines BUGFIX: signalling code entered unlimited loop (in kernel!) updated signalling to scale as expected. tcpdump results are more or less comparable to original (for sl:debug) ------------------------------------------------------------------------ r225 | wdb | 2007-06-23 19:54:24 +0200 (Sat, 23 Jun 2007) | 2 lines (skip) ------------------------------------------------------------------------ r224 | wdb | 2007-06-23 19:31:36 +0200 (Sat, 23 Jun 2007) | 2 lines (skip) ------------------------------------------------------------------------ r223 | wdb | 2007-06-23 19:29:51 +0200 (Sat, 23 Jun 2007) | 3 lines BUGFIX: minor issues that broke compilation of RELEASE build ------------------------------------------------------------------------ r222 | wdb | 2007-06-23 18:30:25 +0200 (Sat, 23 Jun 2007) | 2 lines BUGFIX: modified cross-space buffer refcounting to allow correct cleanup of OB_SHARED buffers ------------------------------------------------------------------------ r221 | wdb | 2007-06-23 15:48:54 +0200 (Sat, 23 Jun 2007) | 5 lines BUGFIX: softcpu lockup un userspace->kernel signal due to incorrectly initialized var (0 instead of -1) CLEANUP: all buffer-related signal handlers now use the "global buffer id" un buf->shared CLEANUP: transport path has been simplified further, as has remotebuf ------------------------------------------------------------------------ r220 | wdb | 2007-06-23 10:31:09 +0200 (Sat, 23 Jun 2007) | 2 lines bugfix: default ringsize was not correctly detected ------------------------------------------------------------------------ r219 | wdb | 2007-06-22 10:09:12 +0200 (Fri, 22 Jun 2007) | 2 lines added ftptest in relation to an email to the mailing list by Manik ------------------------------------------------------------------------ r218 | mtaneja | 2007-06-21 16:24:39 +0200 (Thu, 21 Jun 2007) | 2 lines ntermediate checkin ------------------------------------------------------------------------ r217 | wdb | 2007-06-21 02:11:33 +0200 (Thu, 21 Jun 2007) | 2 lines stable? ------------------------------------------------------------------------ r216 | wdb | 2007-06-21 00:33:32 +0200 (Thu, 21 Jun 2007) | 3 lines BUGFIX: signalling from userspace to kernel caused soft lockup. I had never run userspace to kernel with the new signalling code before ------------------------------------------------------------------------ r215 | wdb | 2007-06-20 22:47:22 +0200 (Wed, 20 Jun 2007) | 2 lines put bandaid around timer_del. hopefully this stops the kernel panics ------------------------------------------------------------------------ r214 | mtaneja | 2007-06-19 23:27:56 +0200 (Tue, 19 Jun 2007) | 2 lines more bug fixes . ------------------------------------------------------------------------ r213 | wdb | 2007-06-15 20:08:01 +0200 (Fri, 15 Jun 2007) | 2 lines fixed bug in per-buffer signalling ------------------------------------------------------------------------ r212 | wdb | 2007-06-15 19:45:40 +0200 (Fri, 15 Jun 2007) | 3 lines added per-buffer/per-space max_batch to circumvent packet drops ------------------------------------------------------------------------ r211 | wdb | 2007-06-15 16:41:19 +0200 (Fri, 15 Jun 2007) | 2 lines BUGFIX: timeout signals from kernel would not reach userspace ------------------------------------------------------------------------ r210 | wdb | 2007-06-15 16:17:28 +0200 (Fri, 15 Jun 2007) | 2 lines BUGFIX: tcpdump reopen buffer reused stale entry ------------------------------------------------------------------------ r209 | wdb | 2007-06-15 13:16:14 +0200 (Fri, 15 Jun 2007) | 2 lines skip ------------------------------------------------------------------------ r208 | mtaneja | 2007-06-15 01:12:16 +0200 (Fri, 15 Jun 2007) | 2 lines kernel version of ipfix (not working as of now ) ------------------------------------------------------------------------ r207 | wdb | 2007-06-15 00:43:44 +0200 (Fri, 15 Jun 2007) | 4 lines BUGFIX: infinite loop of open_kernelstream BUGFIX: hash_del in ipfix_del ------------------------------------------------------------------------ r206 | wdb | 2007-06-15 00:26:33 +0200 (Fri, 15 Jun 2007) | 2 lines BUGFIX: control_getdata crashed on NULL reply ------------------------------------------------------------------------ r205 | wdb | 2007-06-14 00:35:44 +0200 (Thu, 14 Jun 2007) | 2 lines rewrote hashtable functions to show where overflow occurs ------------------------------------------------------------------------ r204 | wdb | 2007-06-13 23:04:06 +0200 (Wed, 13 Jun 2007) | 5 lines added support for libeditline as a replacement for libreadline removed the userspace 'datapath' polling thread, which is no longer needed ------------------------------------------------------------------------ r203 | wdb | 2007-06-12 10:32:49 +0200 (Tue, 12 Jun 2007) | 3 lines BUGFIX: OB_SHARED ------------------------------------------------------------------------ r202 | wdb | 2007-06-11 14:47:12 +0200 (Mon, 11 Jun 2007) | 5 lines BUGFIX: soft lockup on fast sending in kernel BUGFIX: timer and signal bugs CLEANUP: transport.c had become cluttered over the years seriously simplified its controlflow (less duplicate code) ------------------------------------------------------------------------ r201 | wdb | 2007-06-09 18:22:35 +0200 (Sat, 09 Jun 2007) | 2 lines BUGFIX ethernet out now seems to work ------------------------------------------------------------------------ r200 | wdb | 2007-06-07 19:38:33 +0200 (Thu, 07 Jun 2007) | 2 lines kernel timer won't listen ------------------------------------------------------------------------ r199 | wdb | 2007-06-07 18:27:10 +0200 (Thu, 07 Jun 2007) | 11 lines final touches. many smaller bugfixes BUGFIX: duplicate buffer mappings from within the same app caused contention NEW: logfiles are now separated by user. NEW: logfile 'streamline.$USERNAME.latest.log' always points to the newest entry CLEANUP: support/macros.h had become unwieldy. I spun off some pieces BUGFIX: kernel timer could cause panic because it wasn't cleaned up correctly BUGFIX: the first instance in a space (e.g, (userspace)) would not get reused because it appeared to have no input, while the request entry had inputs. ------------------------------------------------------------------------ r198 | wdb | 2007-06-07 14:16:44 +0200 (Thu, 07 Jun 2007) | 2 lines updated timer kernel-implementation to recur ------------------------------------------------------------------------ r197 | wdb | 2007-06-05 11:44:44 +0200 (Tue, 05 Jun 2007) | 11 lines MILESTONE: pcap works *) tcpdump works *) multiple parallel instances work and share resources *) many bugfixes along the way WARNING: signal coalescing may be iffy I've been hacking while at holiday, so internet access was very intermittent. To be precise, this is the first commit after 3 days of work. ------------------------------------------------------------------------ r196 | wdb | 2007-06-02 00:53:52 +0200 (Sat, 02 Jun 2007) | 2 lines it seems there's one bug left in tcpdump-style processing ------------------------------------------------------------------------ r195 | wdb | 2007-06-01 23:05:19 +0200 (Fri, 01 Jun 2007) | 2 lines resolved deadlock in *other* access method (used by slcli for example) ------------------------------------------------------------------------ r194 | wdb | 2007-06-01 22:24:10 +0200 (Fri, 01 Jun 2007) | 2 lines removed deadlock in tcpdump and related peek_stream() applications ------------------------------------------------------------------------ r193 | wdb | 2007-05-31 17:33:45 +0200 (Thu, 31 May 2007) | 3 lines (skip) ------------------------------------------------------------------------ r192 | wdb | 2007-05-31 17:15:17 +0200 (Thu, 31 May 2007) | 3 lines MILESTONE: cli app is working again, with signalling enabled ------------------------------------------------------------------------ r191 | wdb | 2007-05-31 00:21:45 +0200 (Thu, 31 May 2007) | 2 lines bugfixes ------------------------------------------------------------------------ r190 | wdb | 2007-05-30 23:39:01 +0200 (Wed, 30 May 2007) | 3 lines bugfix: channel cleanup bugs ------------------------------------------------------------------------ r189 | wdb | 2007-05-30 21:01:46 +0200 (Wed, 30 May 2007) | 5 lines fixed memory issues on destruction tcpdump works, albeit without signalling ------------------------------------------------------------------------ r188 | wdb | 2007-05-30 10:13:08 +0200 (Wed, 30 May 2007) | 2 lines temporarily disabled signals ------------------------------------------------------------------------ r187 | wdb | 2007-05-29 18:50:46 +0200 (Tue, 29 May 2007) | 2 lines fixed some bugs in the new signalling infrastructure. not sure it is okay now ------------------------------------------------------------------------ r186 | wdb | 2007-05-29 17:38:17 +0200 (Tue, 29 May 2007) | 9 lines added signalling channel types sig_new and sig_dup split the active channel_instance list into per-(type,space) lists added a new sysfs_upcall channel that should be markedly faster than sending MSG_SIG messages through sysfs_up added delayed signalling to circumvent scheduling in atomic context issues warning: going to VMWare, this new stuff has to be debugged first ------------------------------------------------------------------------ r185 | wdb | 2007-05-29 00:21:04 +0200 (Tue, 29 May 2007) | 2 lines (skip) ------------------------------------------------------------------------ r184 | wdb | 2007-05-29 00:09:17 +0200 (Tue, 29 May 2007) | 6 lines MILESTONE: kernel->userspace channel works. currently we run in interrupt mode (no batching), so signalling overhead is prohibitive along the way many unrelated bugs have also been found and fixed ------------------------------------------------------------------------ r183 | wdb | 2007-05-28 20:54:45 +0200 (Mon, 28 May 2007) | 2 lines we have a version of the pipe that works ------------------------------------------------------------------------ r182 | wdb | 2007-05-28 19:46:26 +0200 (Mon, 28 May 2007) | 3 lines new pipe still unstable ------------------------------------------------------------------------ r181 | wdb | 2007-05-28 19:12:59 +0200 (Mon, 28 May 2007) | 3 lines bugfix: new kernel->userspace channel wasn't being cleaned up correctly ------------------------------------------------------------------------ r180 | wdb | 2007-05-28 18:23:55 +0200 (Mon, 28 May 2007) | 4 lines bugfix: destructors were not all called correctly temporarily disabled the kernel->uspace channel ------------------------------------------------------------------------ r179 | wdb | 2007-05-28 15:23:20 +0200 (Mon, 28 May 2007) | 2 lines (skip) ------------------------------------------------------------------------ r178 | wdb | 2007-05-28 15:13:55 +0200 (Mon, 28 May 2007) | 2 lines going to vmware (so you know it ain't good) ------------------------------------------------------------------------ r177 | wdb | 2007-05-28 14:16:03 +0200 (Mon, 28 May 2007) | 3 lines new: kernel->userspace pipe; also demanded some support code (message handlers et al) ------------------------------------------------------------------------ r176 | wdb | 2007-05-28 04:23:28 +0200 (Mon, 28 May 2007) | 8 lines fixed some minor bugs. what's left for pcap/sockets is 1) setting up a messaging channel from the kernel to uspace (easy) 2) making sure no signals get lost by taking into account ringsize (easy too) I'll (try to) do both tomorrow ------------------------------------------------------------------------ r175 | wdb | 2007-05-28 03:42:47 +0200 (Mon, 28 May 2007) | 2 lines made it compile in the kernel ------------------------------------------------------------------------ r174 | wdb | 2007-05-28 03:26:16 +0200 (Mon, 28 May 2007) | 2 lines (skip) ------------------------------------------------------------------------ r173 | wdb | 2007-05-28 03:23:08 +0200 (Mon, 28 May 2007) | 11 lines wide-ranging changes to allow kernel->userspace signals cleanup: datapath_import now uses a signalling callback instead of list_walk cleanup: ibuf remote_lookup now uses a hashtable instead of list_walk extended the timer interface with timer_pause_expire and timer_pause: cheaper versions of timer_usleep when used within a loop moved ibuf:remotebuf to its own set of files moved datapath:import to its own set of files fixed smaller bugs, e.g., in generator ------------------------------------------------------------------------ r172 | wdb | 2007-05-27 17:32:38 +0200 (Sun, 27 May 2007) | 3 lines bugfix: small but important refcount issues in (new code in) functions/ethernet.c ------------------------------------------------------------------------ r171 | wdb | 2007-05-27 16:01:30 +0200 (Sun, 27 May 2007) | 6 lines bugfix: __gnode_hashify prefix was faulty; renamed to __node_create_prefix in src/Makefile, made all apps dependent on libstreamline.so, so that we don't have to rebuild them everytime ------------------------------------------------------------------------ r170 | wdb | 2007-05-27 15:43:01 +0200 (Sun, 27 May 2007) | 7 lines refined gnode_hashify with expression support added per-device support to pcap bugfix: request-name dictionary lead to collisions, while multiple identical requests are allowed. now the string is attached directly to the struct gnode ------------------------------------------------------------------------ r169 | wdb | 2007-05-27 15:02:55 +0200 (Sun, 27 May 2007) | 4 lines new: added (ethernet) device filtering (e.g., (ethernet,expression="eth0")) contrary to older versions we now use a fast hashtable lookup at runtime ------------------------------------------------------------------------ r168 | wdb | 2007-05-25 20:05:48 +0200 (Fri, 25 May 2007) | 18 lines new: made pcap_read_op follow both pcap_dispatch and pcap_loop semantics new: added checkstream() calls around pcap's data access new: slsleep(), slusleep() and slpause() functions (our equivalents of POSIX) new: timer_wait(), timer_interrupt() for above moved: streamline_request_... to slrequest_ new: slrequest_add now accepts format strings moved: slsignalhandler_... to slsig_... new: created slrequest_setbinary to replace the previous precompiled function hacks (PRECOMPILED and bin_expr) ------------------------------------------------------------------------ r167 | wdb | 2007-05-25 13:38:47 +0200 (Fri, 25 May 2007) | 2 lines (skip) ------------------------------------------------------------------------ r166 | wdb | 2007-05-24 18:41:49 +0200 (Thu, 24 May 2007) | 3 lines cleaned up libpcap support; not done yet ------------------------------------------------------------------------ r165 | wdb | 2007-05-24 16:57:37 +0200 (Thu, 24 May 2007) | 3 lines bugfix: on reopening a sysfs_area the returned number was not ci->key but key. ------------------------------------------------------------------------ r164 | wdb | 2007-05-24 15:41:45 +0200 (Thu, 24 May 2007) | 2 lines bugfix: MSG_TRANS_CREATE works again ------------------------------------------------------------------------ r163 | wdb | 2007-05-23 00:07:06 +0200 (Wed, 23 May 2007) | 4 lines new: format string support in sl_log bugfix: minor memleak in route_del ------------------------------------------------------------------------ r162 | wdb | 2007-05-21 20:52:28 +0200 (Mon, 21 May 2007) | 8 lines bugfix: ethernet_out generated IP, now it generates ETH new: transport:__create_remote_stream: open a stream in kernelspace if OB_SHARED is passed (BROKEN, awaits fix) bugfix: RPC request could get connection ID 0, which is reserved for non-RPC bugfix: make doc failed (thanks Jaap) new: CLI shutdown nicely on Ctrl^D (thanks Jaap) ------------------------------------------------------------------------ r161 | wdb | 2007-05-21 14:55:36 +0200 (Mon, 21 May 2007) | 5 lines added tailroom to socket write_stream to allow for skb_shared_info in skbuff-based transmit remove fake skbuff. replaced it with upper limit on necessary tailroom ------------------------------------------------------------------------ r160 | wdb | 2007-05-21 13:10:51 +0200 (Mon, 21 May 2007) | 5 lines bugfix in timer: alarm_stop killed application (SIGALRM) bugfix in socket: shutdown before recv()/send() caused crash modification of adaptive signalling: grow/shrink max_batch instead of rate ------------------------------------------------------------------------ r159 | mtaneja | 2007-05-21 03:40:05 +0200 (Mon, 21 May 2007) | 6 lines added ipfix implementation fixed some bugs in tracker fixed compilation errors in timer.c and signal.c ------------------------------------------------------------------------ r158 | wdb | 2007-05-20 22:58:00 +0200 (Sun, 20 May 2007) | 2 lines disabled transmit ------------------------------------------------------------------------ r157 | wdb | 2007-05-20 21:14:47 +0200 (Sun, 20 May 2007) | 5 lines sockets serious bugfix transmission faster alternative kernel panic (?) -- going to vmware ------------------------------------------------------------------------ r156 | wdb | 2007-05-16 20:19:21 +0200 (Wed, 16 May 2007) | 7 lines - rewrote signals implementation. "sigbatched" is now default, also in kernel - instrumented signals to supply profiling information - expanded timer to allow multiple parallel timers atop a single SIGALRM realtime clock ------------------------------------------------------------------------ r155 | wdb | 2007-05-15 17:08:39 +0200 (Tue, 15 May 2007) | 2 lines no kpanic in vmware, switching back to host ------------------------------------------------------------------------ r154 | wdb | 2007-05-15 16:03:46 +0200 (Tue, 15 May 2007) | 2 lines working on cross-space signalling; experiencing a kernel panic ------------------------------------------------------------------------ r153 | wdb | 2007-05-14 22:11:34 +0200 (Mon, 14 May 2007) | 3 lines minor changes to benchmarky ------------------------------------------------------------------------ r152 | wdb | 2007-05-10 16:00:13 +0200 (Thu, 10 May 2007) | 3 lines updated update strategy of benchmark ------------------------------------------------------------------------ r151 | wdb | 2007-05-10 15:30:41 +0200 (Thu, 10 May 2007) | 2 lines made the socket benchmark adaptive: the client no longer overflows the sender ------------------------------------------------------------------------ r150 | wdb | 2007-05-09 20:20:01 +0200 (Wed, 09 May 2007) | 2 lines updated socket_ratelimit benchmark ------------------------------------------------------------------------ r149 | wdb | 2007-05-08 20:48:31 +0200 (Tue, 08 May 2007) | 3 lines bugfix: more functions that used '2' to talk to the kernel (porttable_...) ------------------------------------------------------------------------ r148 | wdb | 2007-05-08 20:03:48 +0200 (Tue, 08 May 2007) | 4 lines first compile on atlas. fixed some compilation errors (such as 'const const' declarations) bugfix: the kernelspace was directly referred to as '2' in core/transport.c. this is no longer true; I replaced the constant with a define ------------------------------------------------------------------------ r147 | wdb | 2007-05-07 13:06:45 +0200 (Mon, 07 May 2007) | 2 lines fixed ethernet_out ptr ------------------------------------------------------------------------ r146 | wdb | 2007-05-07 09:51:44 +0200 (Mon, 07 May 2007) | 2 lines bugfix: ethernet_out pktlen ------------------------------------------------------------------------ r145 | mtaneja | 2007-05-05 20:59:17 +0200 (Sat, 05 May 2007) | 2 lines bug fixes to mapi.c ------------------------------------------------------------------------ r144 | wdb | 2007-05-04 09:55:35 +0200 (Fri, 04 May 2007) | 2 lines bugfix ------------------------------------------------------------------------ r143 | mtaneja | 2007-05-01 21:46:42 +0200 (Tue, 01 May 2007) | 4 lines fix build errors in ask_trans.c fixes to mapi.c changed ethernet-out to ethernet_out ------------------------------------------------------------------------ r142 | wdb | 2007-04-27 18:33:36 +0200 (Fri, 27 Apr 2007) | 2 lines completed UDP channel. debugging last issues ------------------------------------------------------------------------ r141 | wdb | 2007-04-25 23:04:37 +0200 (Wed, 25 Apr 2007) | 2 lines further integrated distrib processing ------------------------------------------------------------------------ r140 | mtaneja | 2007-04-24 00:55:24 +0200 (Tue, 24 Apr 2007) | 2 lines bug fixes ------------------------------------------------------------------------ r137 | wdb | 2007-04-23 22:20:12 +0200 (Mon, 23 Apr 2007) | 2 lines fixed issues manik brought up ------------------------------------------------------------------------ r136 | wdb | 2007-04-17 18:18:48 +0200 (Tue, 17 Apr 2007) | 4 lines updated TODO list verified socket implementation ------------------------------------------------------------------------ r135 | wdb | 2007-04-17 16:38:59 +0200 (Tue, 17 Apr 2007) | 2 lines minor ------------------------------------------------------------------------ r134 | wdb | 2007-04-16 15:12:05 +0200 (Mon, 16 Apr 2007) | 2 lines all works. merging of areas will need merging of buffer_export requests it seems ------------------------------------------------------------------------ r133 | wdb | 2007-04-15 16:14:08 +0200 (Sun, 15 Apr 2007) | 2 lines most works. only cleanup of sysfs channel needs to be added ------------------------------------------------------------------------ r132 | wdb | 2007-04-15 15:44:58 +0200 (Sun, 15 Apr 2007) | 2 lines everthing works. only deallocation of sysfs channels seems to fail ------------------------------------------------------------------------ r131 | wdb | 2007-04-15 02:03:52 +0200 (Sun, 15 Apr 2007) | 3 lines sysfs_area works (opens). still causes problem on close also, metadata fails after a couple of times ------------------------------------------------------------------------ r130 | wdb | 2007-04-15 01:06:41 +0200 (Sun, 15 Apr 2007) | 3 lines sysfs channel works! sysfs_area channels goes into the worst kind of kernel panic ------------------------------------------------------------------------ r129 | wdb | 2007-04-13 15:19:53 +0200 (Fri, 13 Apr 2007) | 2 lines debugging sysfs. still crashes ------------------------------------------------------------------------ r128 | wdb | 2007-04-13 13:40:43 +0200 (Fri, 13 Apr 2007) | 2 lines minor bugfix ------------------------------------------------------------------------ r127 | wdb | 2007-04-13 13:38:52 +0200 (Fri, 13 Apr 2007) | 2 lines fixed some bugs in sysfs channel ------------------------------------------------------------------------ r126 | wdb | 2007-04-13 11:31:19 +0200 (Fri, 13 Apr 2007) | 2 lines all compiles ------------------------------------------------------------------------ r125 | wdb | 2007-04-11 19:01:01 +0200 (Wed, 11 Apr 2007) | 3 lines first implementations of channels sysfs, sysfs_area and udp ------------------------------------------------------------------------ r124 | wdb | 2007-04-11 10:20:07 +0200 (Wed, 11 Apr 2007) | 6 lines added 'make install' option added 'make modules_install' option renamed cli to slcli to allow installation added udp channel skeleton ------------------------------------------------------------------------ r123 | mtaneja | 2007-04-11 00:52:46 +0200 (Wed, 11 Apr 2007) | 2 lines moved tracker to apps/tracker directory ------------------------------------------------------------------------ r122 | mtaneja | 2007-04-11 00:52:09 +0200 (Wed, 11 Apr 2007) | 2 lines delting my_tracker directory and code ------------------------------------------------------------------------ r121 | mtaneja | 2007-04-11 00:51:25 +0200 (Wed, 11 Apr 2007) | 2 lines deleting tracker directory ------------------------------------------------------------------------ r120 | wdb | 2007-04-10 23:11:08 +0200 (Tue, 10 Apr 2007) | 2 lines started work on the new sysfs channel ------------------------------------------------------------------------ r119 | wdb | 2007-04-10 20:56:52 +0200 (Tue, 10 Apr 2007) | 1 line ------------------------------------------------------------------------ r118 | wdb | 2007-04-10 19:10:24 +0200 (Tue, 10 Apr 2007) | 2 lines removed some stale files ------------------------------------------------------------------------ r117 | wdb | 2007-04-10 19:00:20 +0200 (Tue, 10 Apr 2007) | 3 lines (disregard) ------------------------------------------------------------------------ r116 | wdb | 2007-04-10 18:24:15 +0200 (Tue, 10 Apr 2007) | 4 lines MILESTONE: kernel compiles CLEANUP: incorporated /include/streamline ------------------------------------------------------------------------ r115 | wdb | 2007-04-10 17:54:23 +0200 (Tue, 10 Apr 2007) | 2 lines (disregard) ------------------------------------------------------------------------ r114 | wdb | 2007-04-10 17:21:55 +0200 (Tue, 10 Apr 2007) | 3 lines (disregard) ------------------------------------------------------------------------ r113 | wdb | 2007-04-10 17:19:46 +0200 (Tue, 10 Apr 2007) | 2 lines (disregard) ------------------------------------------------------------------------ r112 | wdb | 2007-04-10 17:12:44 +0200 (Tue, 10 Apr 2007) | 2 lines MILESTONE: basic userspace workload succeeds again ------------------------------------------------------------------------ r111 | wdb | 2007-04-10 16:59:55 +0200 (Tue, 10 Apr 2007) | 2 lines make test works (with only local space) ------------------------------------------------------------------------ r110 | wdb | 2007-04-10 10:13:43 +0200 (Tue, 10 Apr 2007) | 4 lines finalized move from 1 buf_data area to 2 areas. FIX: many buf->shared-size bugs that didn't account for sizeof(buf_data) ------------------------------------------------------------------------ r109 | wdb | 2007-04-10 10:00:04 +0200 (Tue, 10 Apr 2007) | 6 lines moved buffer layout from 1 area (buf->shared | data) to two -> this allows better use of & operator and less memory waste when remapping split vbuf and dbuf sourcecode off from buffers/basic ------------------------------------------------------------------------ r108 | wdb | 2007-04-08 09:38:44 +0200 (Sun, 08 Apr 2007) | 2 lines halfway down the move of src/control to src/msg and src/core ------------------------------------------------------------------------ r107 | wdb | 2007-04-07 22:50:15 +0200 (Sat, 07 Apr 2007) | 5 lines NEW: finalized the new channel structure and necessary control messageswq -left one minor compiler error to start with next time NEW: added a multihash structure: a hashtable with multiple values per key ------------------------------------------------------------------------ r106 | wdb | 2007-04-04 20:09:33 +0200 (Wed, 04 Apr 2007) | 7 lines WARNING: will not compile in the middle of rewriting channels NEW: forwarding through pipes (as opposed to through remapping) ------------------------------------------------------------------------ r105 | wdb | 2007-04-02 14:15:12 +0200 (Mon, 02 Apr 2007) | 8 lines WARNING: I'm still cleaning up channels. this build is broken CLEANUP: replaced all manual lists with src/support/list implementations CLEANUP: replaced all manual tables with src/support/hashtable where possible FIX: found bugs in these manual list implementations CLEANUP: reduced no. of channel and channel impl functions ------------------------------------------------------------------------ r104 | wdb | 2007-03-28 16:37:36 +0200 (Wed, 28 Mar 2007) | 3 lines kernel still iffy ------------------------------------------------------------------------ r103 | wdb | 2007-03-28 16:15:17 +0200 (Wed, 28 Mar 2007) | 4 lines fixed some bugs. some arose during cleanup, hopefully others were older (like the infamous mmap_simple issue (?)) ------------------------------------------------------------------------ r102 | wdb | 2007-03-28 13:46:08 +0200 (Wed, 28 Mar 2007) | 8 lines theme: major cleanup. * removed minor and/or stale files * simplified messaging * removed struct id new: ethernet-out, a badly named function that fetches outgoing packets warning: userspace/kernelspace communication is buggy. will now fix ------------------------------------------------------------------------ r101 | wdb | 2007-03-22 15:22:05 +0100 (Thu, 22 Mar 2007) | 2 lines added pingpong server ------------------------------------------------------------------------ r100 | wdb | 2007-03-20 10:54:06 +0100 (Tue, 20 Mar 2007) | 4 lines file compression works network encrypt&compress works ------------------------------------------------------------------------ r99 | wdb | 2007-03-20 10:36:06 +0100 (Tue, 20 Mar 2007) | 3 lines network compression ------------------------------------------------------------------------ r98 | wdb | 2007-03-20 02:36:54 +0100 (Tue, 20 Mar 2007) | 2 lines disregard. changed ibuf length ------------------------------------------------------------------------ r97 | wdb | 2007-03-20 02:24:04 +0100 (Tue, 20 Mar 2007) | 2 lines added socketfree version of ratelimit (recv only) ------------------------------------------------------------------------ r96 | wdb | 2007-03-20 01:33:24 +0100 (Tue, 20 Mar 2007) | 2 lines minor changes ------------------------------------------------------------------------ r95 | wdb | 2007-03-20 00:58:59 +0100 (Tue, 20 Mar 2007) | 2 lines fixed bugs for ratelimit ------------------------------------------------------------------------ r94 | wdb | 2007-03-20 00:37:49 +0100 (Tue, 20 Mar 2007) | 2 lines fixed apparently longstanding cbuf bug ------------------------------------------------------------------------ r93 | wdb | 2007-03-19 23:34:43 +0100 (Mon, 19 Mar 2007) | 2 lines test fails ------------------------------------------------------------------------ r92 | wdb | 2007-03-19 22:57:34 +0100 (Mon, 19 Mar 2007) | 2 lines warnings no more ------------------------------------------------------------------------ r91 | wdb | 2007-03-19 20:36:06 +0100 (Mon, 19 Mar 2007) | 3 lines both httpwikis work ADD: append_stream ------------------------------------------------------------------------ r90 | wdb | 2007-03-19 19:14:29 +0100 (Mon, 19 Mar 2007) | 2 lines fixed output bug in httpwiki_pcap ------------------------------------------------------------------------ r89 | wdb | 2007-03-19 18:10:28 +0100 (Mon, 19 Mar 2007) | 2 lines updated httpwiki to fix a bug found with the PCAP version ------------------------------------------------------------------------ r88 | wdb | 2007-03-19 18:06:26 +0100 (Mon, 19 Mar 2007) | 2 lines added pcap httpwiki, forgot to do that in previous commit, contrary to msg ------------------------------------------------------------------------ r87 | wdb | 2007-03-19 18:04:27 +0100 (Mon, 19 Mar 2007) | 3 lines added pcap-based httpwiki started 'nonsl' directory for alternative implementations (that we use for comparison) ------------------------------------------------------------------------ r86 | wdb | 2007-03-19 16:29:52 +0100 (Mon, 19 Mar 2007) | 2 lines added ratelimiter ------------------------------------------------------------------------ r85 | mtaneja | 2007-03-19 12:12:50 +0100 (Mon, 19 Mar 2007) | 1 line ------------------------------------------------------------------------ r84 | mtaneja | 2007-03-19 01:37:45 +0100 (Mon, 19 Mar 2007) | 10 lines ibug fixes --This line, and those below, will be ignored-- M my_tracker/tracker_rpc.h M my_tracker/tracker_rpc.x M my_tracker/tracker_get_stats.c M my_tracker/tracker_file M my_tracker/tracker_rpc_svc.c M my_tracker/tracker.c M my_tracker/tracker_rpc_xdr.c ------------------------------------------------------------------------ r83 | mtaneja | 2007-03-18 22:33:27 +0100 (Sun, 18 Mar 2007) | 12 lines irp--This line, and those below, will be ignored-- A my_tracker/tracker_rpc_clnt.c A my_tracker/tracker_rpc.h A my_tracker/tracker_rpc.x A my_tracker/tracker_get_stats.c A my_tracker/tracker_file A my_tracker/tracker_rpc_svc.c M my_tracker/tracker.c M my_tracker/Makefile A my_tracker/tracker_rpc_xdr.c ------------------------------------------------------------------------ r82 | wdb | 2007-03-18 21:56:08 +0100 (Sun, 18 Mar 2007) | 2 lines it all seems to work fine! ------------------------------------------------------------------------ r81 | wdb | 2007-03-18 20:22:54 +0100 (Sun, 18 Mar 2007) | 2 lines back to host ------------------------------------------------------------------------ r80 | wdb | 2007-03-18 18:54:04 +0100 (Sun, 18 Mar 2007) | 2 lines back to vmware ------------------------------------------------------------------------ r79 | wdb | 2007-03-18 17:11:41 +0100 (Sun, 18 Mar 2007) | 2 lines reverted FBUF sizes. ------------------------------------------------------------------------ r78 | wdb | 2007-03-18 17:04:49 +0100 (Sun, 18 Mar 2007) | 4 lines reduced standard ring size to increase cache hitrate (hopefully) made FBUFs allocate (by default) pow2 + header may break ------------------------------------------------------------------------ r77 | wdb | 2007-03-18 16:39:34 +0100 (Sun, 18 Mar 2007) | 5 lines MAJOR BUGFIX: replace ibuf backend with slotted buffer, to solve longstanding wrap issue (implementation can still be a bit buggy) PERFFIX: decreased ibuf size to exactly 16 bytes. nice alignment CLEANUP: replaced id_local* with id_local. faster member lookup can mildy enhance performance ------------------------------------------------------------------------ r76 | wdb | 2007-03-18 11:59:10 +0100 (Sun, 18 Mar 2007) | 2 lines integrated second speedup ------------------------------------------------------------------------ r75 | wdb | 2007-03-18 11:44:01 +0100 (Sun, 18 Mar 2007) | 2 lines back to host ------------------------------------------------------------------------ r74 | wdb | 2007-03-18 10:58:29 +0100 (Sun, 18 Mar 2007) | 2 lines implemented first optimalisation in sockets->send ------------------------------------------------------------------------ r73 | wdb | 2007-03-18 09:49:25 +0100 (Sun, 18 Mar 2007) | 2 lines kpanic. to vmware ------------------------------------------------------------------------ r72 | wdb | 2007-03-18 01:14:05 +0100 (Sun, 18 Mar 2007) | 3 lines fastpath still broken. for tomorrow ------------------------------------------------------------------------ r71 | wdb | 2007-03-18 01:07:38 +0100 (Sun, 18 Mar 2007) | 2 lines minor update ------------------------------------------------------------------------ r70 | wdb | 2007-03-18 01:01:51 +0100 (Sun, 18 Mar 2007) | 2 lines new (untested) fast sockets ------------------------------------------------------------------------ r69 | wdb | 2007-03-18 00:02:12 +0100 (Sun, 18 Mar 2007) | 2 lines tested slow version of sockets. okay except for overflow in rx ibuf ------------------------------------------------------------------------ r68 | wdb | 2007-03-17 19:34:18 +0100 (Sat, 17 Mar 2007) | 2 lines cleaned up sockets ------------------------------------------------------------------------ r67 | wdb | 2007-03-17 16:32:17 +0100 (Sat, 17 Mar 2007) | 3 lines sendto/recv loop works, but data read seems garbled. lunchtime ------------------------------------------------------------------------ r66 | wdb | 2007-03-17 15:39:40 +0100 (Sat, 17 Mar 2007) | 2 lines FIX double slow send works ------------------------------------------------------------------------ r65 | wdb | 2007-03-17 15:20:06 +0100 (Sat, 17 Mar 2007) | 3 lines merged collisions between two branches ------------------------------------------------------------------------ r64 | wdb | 2007-03-17 00:18:48 +0100 (Sat, 17 Mar 2007) | 2 lines nightnight ------------------------------------------------------------------------ r63 | wdb | 2007-03-16 22:26:42 +0100 (Fri, 16 Mar 2007) | 2 lines broken ------------------------------------------------------------------------ r62 | wdb | 2007-03-16 21:51:38 +0100 (Fri, 16 Mar 2007) | 2 lines broken (transmit) ------------------------------------------------------------------------ r61 | wdb | 2007-03-16 20:31:10 +0100 (Fri, 16 Mar 2007) | 2 lines fix: fbuf count. ------------------------------------------------------------------------ r60 | wdb | 2007-03-16 15:50:50 +0100 (Fri, 16 Mar 2007) | 2 lines moving to vmplayer ------------------------------------------------------------------------ r59 | wdb | 2007-03-16 11:11:26 +0100 (Fri, 16 Mar 2007) | 2 lines one step forward... ------------------------------------------------------------------------ r58 | wdb | 2007-03-16 09:52:24 +0100 (Fri, 16 Mar 2007) | 2 lines dunno ------------------------------------------------------------------------ r57 | mtaneja | 2007-03-16 04:57:55 +0100 (Fri, 16 Mar 2007) | 1 line ------------------------------------------------------------------------ r56 | mtaneja | 2007-03-16 03:06:15 +0100 (Fri, 16 Mar 2007) | 2 lines bug fixes and added global lock o to mapi and i hate emacs !!! ------------------------------------------------------------------------ r55 | mtaneja | 2007-03-15 21:41:09 +0100 (Thu, 15 Mar 2007) | 2 lines fixes ------------------------------------------------------------------------ r54 | mtaneja | 2007-03-15 21:40:50 +0100 (Thu, 15 Mar 2007) | 2 lines changed log_** calls to simple echo calls ------------------------------------------------------------------------ r53 | wdb | 2007-03-15 20:06:23 +0100 (Thu, 15 Mar 2007) | 2 lines updated makefile ------------------------------------------------------------------------ r52 | mtaneja | 2007-03-15 19:51:10 +0100 (Thu, 15 Mar 2007) | 1 line ------------------------------------------------------------------------ r51 | wdb | 2007-03-15 19:49:31 +0100 (Thu, 15 Mar 2007) | 2 lines forget this ------------------------------------------------------------------------ r50 | wdb | 2007-03-15 16:39:43 +0100 (Thu, 15 Mar 2007) | 3 lines sockets controlplane succeeds for UDP. next stop testing actual transmission ------------------------------------------------------------------------ r49 | wdb | 2007-03-15 00:22:51 +0100 (Thu, 15 Mar 2007) | 3 lines debugged sockets a bit. still a deadlock incorporated (buffer) functions ------------------------------------------------------------------------ r48 | wdb | 2007-03-14 22:59:19 +0100 (Wed, 14 Mar 2007) | 1 line ------------------------------------------------------------------------ r47 | wdb | 2007-03-14 22:39:07 +0100 (Wed, 14 Mar 2007) | 3 lines integrated sockets into the mainline. could be buggy ------------------------------------------------------------------------ r46 | wdb | 2007-03-14 09:40:03 +0100 (Wed, 14 Mar 2007) | 2 lines fix: bpf results ------------------------------------------------------------------------ r45 | mtaneja | 2007-03-13 21:11:47 +0100 (Tue, 13 Mar 2007) | 3 lines added tracker files to streamline source -Manik ------------------------------------------------------------------------ r44 | mtaneja | 2007-03-13 21:10:57 +0100 (Tue, 13 Mar 2007) | 2 lines remove stray print comments ------------------------------------------------------------------------ r43 | wdb | 2007-03-13 00:47:11 +0100 (Tue, 13 Mar 2007) | 2 lines FIX: used to allow exported buffer to be re-exported in instance_reuse ------------------------------------------------------------------------ r42 | mtaneja | 2007-03-12 20:21:46 +0100 (Mon, 12 Mar 2007) | 2 lines changed export call.. ------------------------------------------------------------------------ r41 | wdb | 2007-03-12 15:11:59 +0100 (Mon, 12 Mar 2007) | 3 lines FIX: datapath export cond didn't set iptr->index correctly when writing to main backing store ------------------------------------------------------------------------ r40 | wdb | 2007-03-12 14:27:02 +0100 (Mon, 12 Mar 2007) | 2 lines FIX: bug in "export" ------------------------------------------------------------------------ r39 | wdb | 2007-03-10 22:19:24 +0100 (Sat, 10 Mar 2007) | 3 lines FIX: dict_lookup failed due to missing \0 ------------------------------------------------------------------------ r38 | mtaneja | 2007-03-10 12:16:36 +0100 (Sat, 10 Mar 2007) | 2 lines bug fixes ------------------------------------------------------------------------ r37 | mtaneja | 2007-03-10 12:16:27 +0100 (Sat, 10 Mar 2007) | 2 lines bug fixes ------------------------------------------------------------------------ r36 | mtaneja | 2007-03-10 12:15:58 +0100 (Sat, 10 Mar 2007) | 2 lines mapi test routines ------------------------------------------------------------------------ r35 | mtaneja | 2007-03-10 12:15:31 +0100 (Sat, 10 Mar 2007) | 2 lines added lines for mapi interface ------------------------------------------------------------------------ r34 | wdb | 2007-03-09 16:51:45 +0100 (Fri, 09 Mar 2007) | 12 lines FIX: made (broken) transmission/sockets build conditional FIX: removed hack to work around no-longer-existent translation issue CLEANUP: cleaned out the dictionary CLEANUP: cleaned the build: 'make' no longer compiles tests and thus 'make kernel' no longer forces a test recompile FIX: reenabled test goals, apparently tests had been disabled for a while FIX: some tests failed for various reasons. all pass again REMOVE: moved Asia's prospector code to an optional build path ADD: created parser/translator.[ch], which takes the code from parser/parser.c concerning translation ------------------------------------------------------------------------ r33 | mtaneja | 2007-03-09 00:15:29 +0100 (Fri, 09 Mar 2007) | 3 lines added mapi_interface.* not tested yet :) ------------------------------------------------------------------------ r32 | vupiggy | 2007-03-07 21:07:36 +0100 (Wed, 07 Mar 2007) | 50 lines Add the component of transmission. To make the code can pass the compilation, please save the contents below into /usr/include/sys/sl_transmit.h #ifndef _SL_TRANSMIT_H_ #define _SL_TRANSMIT_H_ #include #include ssize_t streamline_transmit (unsigned int plen) { return syscall (__NR_sl_transmit, plen); } int streamline_connect (int sockfd) { return syscall (__NR_sl_connect, sockfd); } #endif /* _SL_TRANSMIT_H_ */ And to try to run the transmission, please add two system call into the kernel manually. Sorry for the inconvenience. The system call file should look like this(and suggested put this file into net/streamline dir under kernel source code): #include #include #include long (*do_transmit)(int size) = NULL; EXPORT_SYMBOL(do_transmit); asmlinkage long sys_sl_transmit (int size) { return do_transmit ? do_transmit (size) : -ENOSYS; } int (*do_connect)(pid_t pid) = NULL; EXPORT_SYMBOL(do_connect); asmlinkage int sys_sl_connect (pid_t pid) { return do_connect ? do_connect (pid) : -ENOSYS; } == Luke ------------------------------------------------------------------------ r31 | wdb | 2007-03-07 17:33:18 +0100 (Wed, 07 Mar 2007) | 4 lines added slrequest_getnodedata and slrequest_getnode to get metadata about active nodes fixed bug in packetcounter. packetcounter didn't work! ------------------------------------------------------------------------ r30 | wdb | 2007-03-05 18:12:32 +0100 (Mon, 05 Mar 2007) | 6 lines added a function template added a (buffer) function, that can be used to explicitly read from a buffer added shell signal handler registration routines (have yet to be filled in) added check_stream, a function to check whether peek_stream was used correctly some other minor changes that I forgot, this is a multi-day commit (?) ------------------------------------------------------------------------ r28 | wdb | 2007-02-27 22:19:05 +0100 (Tue, 27 Feb 2007) | 9 lines fixed SIGIO signal catching in userspace fixed (inspect) output in release build fixed (httpwiki) Host: searching replaced strstr in httpwiki with our (bounded) version added general option parsing to CLI using GNU argp_parse added CLI flag 'verbose', which redirects sl_log messages to stdout ------------------------------------------------------------------------ r27 | wdb | 2007-02-23 16:56:45 +0100 (Fri, 23 Feb 2007) | 2 lines fixed small signal.c in kernel issue ------------------------------------------------------------------------ r26 | wdb | 2007-02-23 13:32:31 +0100 (Fri, 23 Feb 2007) | 2 lines nothing ------------------------------------------------------------------------ r25 | wdb | 2007-02-22 20:35:06 +0100 (Thu, 22 Feb 2007) | 4 lines included signalling into the main code BUG: file descriptor SIGIO callback seems broken ------------------------------------------------------------------------ r24 | wdb | 2007-02-22 19:04:06 +0100 (Thu, 22 Feb 2007) | 6 lines moved core/signal to support/filedes. it deals only with posix filedescriptors anyway added another core/signal that implements streamline-internal signalling. cleaned up interfaces/id. this affected many different files added safety unload hook to ethernet's module_unload ------------------------------------------------------------------------ r23 | wdb | 2007-02-22 14:49:56 +0100 (Thu, 22 Feb 2007) | 2 lines removed stale debug code ------------------------------------------------------------------------ r22 | mtaneja | 2007-02-22 13:00:16 +0100 (Thu, 22 Feb 2007) | 1 line ------------------------------------------------------------------------ r21 | wdb | 2007-02-20 21:25:46 +0100 (Tue, 20 Feb 2007) | 5 lines added the optional .info field to functions, which explains what a function does changed the 'list functions' metadata request to show .info if it exists added .info to 4 functions ------------------------------------------------------------------------ r20 | wdb | 2007-02-20 18:19:48 +0100 (Tue, 20 Feb 2007) | 3 lines added support for external applications in function (shell) ------------------------------------------------------------------------ r19 | wdb | 2007-02-20 17:58:46 +0100 (Tue, 20 Feb 2007) | 4 lines finished httpwiki rewrite. verified that everything works okay fixed a recently introduced bug in apps/cli: datapath_threaded_start was never called ------------------------------------------------------------------------ r18 | wdb | 2007-02-20 17:20:25 +0100 (Tue, 20 Feb 2007) | 2 lines replaced filters with functions ------------------------------------------------------------------------ r17 | wdb | 2007-02-20 17:02:38 +0100 (Tue, 20 Feb 2007) | 2 lines moved src/filters/pcapout.[ch] to src/filters/pcap.[ch] and fixed the includes ------------------------------------------------------------------------ r16 | wdb | 2007-02-20 16:26:35 +0100 (Tue, 20 Feb 2007) | 6 lines removed the obsolete _process member and the necessary infrastructure to support it split off src/core/datapath.[ch] from src/core/instance.[ch] cleaned the datapath at the same time: removed datapath_hook updated ethernet hook to use the new hook method (using newex's fdata->self) ------------------------------------------------------------------------ r15 | wdb | 2007-02-18 19:45:19 +0100 (Sun, 18 Feb 2007) | 2 lines httpwiki is still buggy ------------------------------------------------------------------------ r14 | wdb | 2007-02-18 18:24:32 +0100 (Sun, 18 Feb 2007) | 5 lines changed function member definition from the mess we had to : 0 on success, !0 on failure edited httpwiki function and app (needs testing) moved functions with outdated APIs to src/filters/deprecated ------------------------------------------------------------------------ r13 | wdb | 2007-02-16 18:41:59 +0100 (Fri, 16 Feb 2007) | 11 lines added centralized signal handling/callback interface for hook functions: core/signal.[ch] changed (file), (stdio), (sockets), and (pcap) to use this simpler interface (fixed many bugs) added a non-stdout logging interface. will be extended to replace all printfs cleaned up interfaces/transport.c::create_stream for readability removed datapath_timer and friends removed datapath_timer code from the apps (cli and gui) ran a leak check on standard operations. found and removed some leaks ------------------------------------------------------------------------ r12 | wdb | 2007-02-14 20:41:18 +0100 (Wed, 14 Feb 2007) | 3 lines changed socket filter (see TODO) found some nasty bugs in interface. not sure they're really reproducible ------------------------------------------------------------------------ r11 | wdb | 2007-02-14 17:08:33 +0100 (Wed, 14 Feb 2007) | 6 lines replaced the 'restart' and 'stop' scripts with an /etc/init.d like script 'streamline' minor: removed slsockets and sl_wqueue kernel modules for now. these lead to kernel panics minor: renamed the mknod script minor: changed ethernet function's default backend from netrx to netfilter minor: fixed dict.c issue that made kernelspace compile fail ------------------------------------------------------------------------ r10 | wdb | 2007-02-12 20:39:37 +0100 (Mon, 12 Feb 2007) | 5 lines added (stdin) and (stdout) functions. these must be merged with (file) later on fixed issues between tryex/newex convention and (generator) and (inspect) fixed some other minor stuff ------------------------------------------------------------------------ r9 | wdb | 2007-02-12 14:38:29 +0100 (Mon, 12 Feb 2007) | 3 lines added (shell) function ------------------------------------------------------------------------ r8 | wdb | 2007-02-11 19:46:26 +0100 (Sun, 11 Feb 2007) | 4 lines added zlib compression. NB: it currently drops blocks that would grow instead of compress. this needs out-of-place rewriting. I have to write a clean interface to do that ------------------------------------------------------------------------ r7 | wdb | 2007-02-10 17:26:15 +0100 (Sat, 10 Feb 2007) | 2 lines verified that the blowfish implementation works ------------------------------------------------------------------------ r6 | wdb | 2007-02-10 16:39:05 +0100 (Sat, 10 Feb 2007) | 4 lines updated (tcp) filter to work as shown on the website minor: fixed '-e' parameter output in Makefile ------------------------------------------------------------------------ r5 | wdb | 2007-02-10 13:37:56 +0100 (Sat, 10 Feb 2007) | 3 lines added the OpenSSL blowfish symmetric cipher. it doesn't seem to work perfectly yet ------------------------------------------------------------------------ r4 | wdb | 2007-02-10 13:12:25 +0100 (Sat, 10 Feb 2007) | 3 lines fixed incorrect use of autoconf's library detect functions ------------------------------------------------------------------------ r3 | wdb | 2007-02-10 12:25:51 +0100 (Sat, 10 Feb 2007) | 4 lines added support for expressions read from files, using the file= keyword fixed include paths that were messed up when rearranging files ------------------------------------------------------------------------ r2 | wdb | 2007-02-08 15:43:28 +0100 (Thu, 08 Feb 2007) | 3 lines added all Changelogs from the sourceforge releases to the main Changelog file ------------------------------------------------------------------------ r1 | wdb | 2007-02-07 11:54:48 +0100 (Wed, 07 Feb 2007) | 2 lines import from cvs ------------------------------------------------------------------------