summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* Add missing header files.Linus Nordberg2011-03-092-0/+15
|
* Use retransmisison timer in request object.Linus Nordberg2011-03-091-7/+10
|
* Add retransmission timer support (UDP).Linus Nordberg2011-03-0911-42/+129
|
* Add RSE_DISCO.Linus Nordberg2011-03-093-0/+3
|
* Rename and move around a few helper functions.Linus Nordberg2011-03-095-81/+88
|
* Don't make UDP write event persistant.Linus Nordberg2011-03-091-22/+46
| | | | | rs_packet_send() adds the event again when there's a packet to send and the write callback does the same if it doesn't drain the queue.
* Add (disabled) linkage to libefence.Linus Nordberg2011-03-081-1/+1
|
* Zap a double free.Linus Nordberg2011-03-081-3/+3
|
* Update rs_err_msg() prototype.Linus Nordberg2011-03-081-2/+2
| | | | | Also, rename enum rs_err_code to not collide with the function named the same.
* examples/client-blocking: pull out the error before freeing conn.Linus Nordberg2011-03-081-1/+3
|
* Timeout implemented in request objects, supported by TCP.Linus Nordberg2011-03-087-26/+80
| | | | TODO: UDP.
* Clean up struct rs_error somewhat.Linus Nordberg2011-03-083-16/+6
| | | | More to be done here!
* Don't return -1 to user but rather an error code.Linus Nordberg2011-03-082-2/+5
| | | | NOTE: Changes rs_conn_receive_packet() and rs_packet_send() only.
* examples/client-blocking.c: Clean up and improve error handling.Linus Nordberg2011-03-071-37/+31
|
* Requests are now high level, i.e. not for user doing the dispatching.Linus Nordberg2011-03-071-60/+7
|
* Move verification of response packets up to a level where it makes sense.Linus Nordberg2011-03-076-90/+125
| | | | | | | | | | | | | | | | | | | | | Replace the user_dispatch_flag on connections with conn_user_dispatch_p(). Remove the 'original' member from packet and instead have an upper layer verify. Rename packet valid_flag --> received_flag to reflect that we don't verify. Move _close_conn() --> conn_close(). Move packet flags into a single unsigned int, for portability. (_read_packet): Don't verify packet. (rs_conn_receive_packet): Don't touch PKT_OUT if there isn't a packet. (rs_conn_receive_packet): Verify packet using packet_verify_response().
* Formatting changes.Linus Nordberg2011-03-073-6/+8
|
* Remove lib/radsec.h.Linus Nordberg2011-03-071-9/+0
| | | | | 1. Two radsec.h is icky. 2. Its contents fit nicely in send.c.
* Move event_set_timeout --> tcp_set_connect_timeout.Linus Nordberg2011-03-064-38/+38
|
* request: Save callers user_data and do invoke callers callbacks.Linus Nordberg2011-03-062-9/+31
|
* Replace asserts with EINVAL checks.Linus Nordberg2011-03-061-9/+3
|
* Restructure code, moving most code out of packet.cLinus Nordberg2011-03-0626-789/+931
| | | | Also, move copyright notice out of COPYING and into every file.
* UDP w/o bufferevents, part 1.Linus Nordberg2011-03-067-86/+233
| | | | Sending, no retransmitting and no receiving.
* Cosmetic changes.Linus Nordberg2011-03-061-5/+5
|
* Don't free config object until we destroy the context.Linus Nordberg2011-03-063-10/+25
|
* Revert 578e3551 -- double free in second invocation of dict_free().Linus Nordberg2011-03-061-1/+0
|
* Do _copy_ the strings from the config object.Linus Nordberg2011-03-061-4/+4
|
* Do free the config object.Linus Nordberg2011-03-061-0/+1
|
* _init_evb: Don't crash on socket errors.Linus Nordberg2011-03-033-2/+13
|
* Correct an error code.Linus Nordberg2011-03-031-1/+3
|
* Correct an error string.Linus Nordberg2011-03-031-1/+1
|
* Be consistent with naming of formal arguments.Linus Nordberg2011-03-021-1/+1
|
* Error handling cleanup 3.Linus Nordberg2011-03-013-4/+6
| | | | Split timeout errors in connecting and I/O.
* Error handling cleanup 2.Linus Nordberg2011-03-013-12/+9
| | | | RSE_NOSYS doesn't need a string.
* Error handling cleanupLinus Nordberg2011-03-015-20/+18
| | | | | Remove RSE_SOME_ERROR. rs_tls_init: Push SSL error stack on libradsec error stack.
* rs_context_destroy: Free freeradius dictionary.Linus Nordberg2011-03-011-0/+1
|
* Don't have the connection freeing peers -- it's the contexts responsibility.Linus Nordberg2011-03-012-9/+3
| | | | Also, let _rs_peer_destroy() free secret.
* rs_request_send: Check arguments and that we have a packet to send.Linus Nordberg2011-03-011-0/+3
|
* Add RSE_INVAL.Linus Nordberg2011-03-012-1/+2
|
* Add asserts to request object.Linus Nordberg2011-02-281-1/+4
|
* Update status in HACKING.Linus Nordberg2011-02-281-15/+16
|
* Make tests run again.Linus Nordberg2011-02-282-5/+30
| | | | NOTE: 2 exceptions ATM.
* Remove last trace of server and client in API.Linus Nordberg2011-02-282-8/+8
|
* Rename relevant functions to reflect authn as opposed to authz.Linus Nordberg2011-02-285-16/+38
| | | | Also add a "bare" create request function.
* API changes: rs_client* --> rs_peer.Linus Nordberg2011-02-285-74/+83
| | | | | Move peer (former server) things to new file peer.c. Update examples.
* Cosmetic header file changes.Linus Nordberg2011-02-282-29/+43
|
* Time out on connect and break event loop on errors.Linus Nordberg2011-02-251-0/+48
|
* API change. Have the request object create its own request message.Linus Nordberg2011-02-254-18/+22
|
* Get rid of loopbreak as a mean for signalling successful packet handling.Linus Nordberg2011-02-252-29/+16
| | | | | | | | | | | | | In blocking mode, breaking the event loop with loopbreak will probably be needed for exceptions like timeout so we shouldn't use it for breaking the loop and signalling success. In the user callbaks used in blocking mode (_rcb and _wcb), disable read and write events on the event buffer. This will stop the loop. Also, set a flag in the packet struct to signal success. In the "low level callbacks" (_read_cb and _write_cb), enable read or write, accordingly.
* Config file changes and small API changes.Linus Nordberg2011-02-248-137/+252
| | | | | | | | 'timeout' and 'tries' move from 'server' stanza to top. 'tries' is now 'retries'. Moving around in internal data structs, making struct peer strictly config. Bug fixes in configuration code. Adding some more cleanup code, freeing allocated memory (still not done!).