Age | Commit message (Collapse) | Author | |
---|---|---|---|
2020-10-21 | Fix #ifdef DROPBEAR_FUZZ (caught by linter) | Matt Johnston | |
2020-10-20 | Load password and key for client fuzzer. | Matt Johnston | |
Add fuzz_dump() | |||
2020-06-15 | Disallow leading lines before the ident for server (#102) | Vladislav Grishenko | |
Per RFC4253 4.2 clients must be able to process other lines of data before the version string, server behavior is not defined neither with MUST/SHOULD nor with MAY. If server process up to 50 lines too - it may cause too long hanging session with invalid/evil client that consume host resources and potentially may lead to DDoS on poor embedded boxes. Let's require first line from client to be version string and fail early if it's not - matches both RFC and real OpenSSH behavior. | |||
2019-05-15 | Rename EPKA -> Plugin | Matt Johnston | |
2019-05-15 | External Public-Key Authentication API (#72) | fabriziobertocci | |
* Implemented dynamic loading of an external plug-in shared library to delegate public key authentication * Moved conditional compilation of the plugin infrastructure into the configure.ac script to be able to add -ldl to dropbear build only when the flag is enabled * Added tags file to the ignore list * Updated API to have the constructor to return function pointers in the pliugin instance. Added support for passing user name to the checkpubkey function. Added options to the session returned by the plugin and have dropbear to parse and process them * Added -rdynamic to the linker flags when EPKA is enabled * Changed the API to pass a previously created session to the checkPubKey function (created during preauth) * Added documentation to the API * Added parameter addrstring to plugin creation function * Modified the API to retrieve the auth options. Instead of having them as field of the EPKASession struct, they are stored internally (plugin-dependent) in the plugin/session and retrieved through a pointer to a function (in the session) * Changed option string to be a simple char * instead of unsigned char * | |||
2019-03-20 | Add a sanity check for DROPBEAR_SVR_MULTIUSER==0 mode | Matt Johnston | |
2018-03-06 | workaround memory sanitizer FD_ZERO false positives | Matt Johnston | |
2018-03-04 | get rid of unused packet_type in encrypted write queue | Matt Johnston | |
2018-02-28 | - #if not #ifdef for DROPBEAR_FUZZ | Matt Johnston | |
- fix some unused variables --HG-- branch : fuzz | |||
2018-02-26 | merge from main | Matt Johnston | |
--HG-- branch : fuzz | |||
2018-02-20 | use a full prototype (#56) | François Perrad | |
2018-02-17 | merge from main | Matt Johnston | |
--HG-- branch : fuzz | |||
2018-02-14 | make signal flags volatile, simplify handling | Matt Johnston | |
2017-05-20 | glaring wrapfd problems fixed | Matt Johnston | |
--HG-- branch : fuzz | |||
2017-05-20 | closer to working | Matt Johnston | |
--HG-- branch : fuzz | |||
2016-07-21 | merge 2016.74 | Matt Johnston | |
2016-05-04 | Convert #ifdef to #if, other build changes | Matt Johnston | |
2016-04-29 | Avoid busy loop while waiting for rekey response | Matt Johnston | |
2016-07-11 | better TRACE of failed remote ident | Matt Johnston | |
2016-04-29 | Avoid busy loop while waiting for rekey response | Matt Johnston | |
2016-03-16 | fix empty C prototypes | Francois Perrad | |
2015-11-27 | Fix ses.channel_signal_pending race | Matt Johnston | |
2015-09-29 | make sure that the test for queued packets to write occurs after | Matt Johnston | |
those packets might have been enqueued by set_connect_fds() | |||
2015-08-07 | only update keepalive timeout post-auth (when keepalives are sent) | Matt Johnston | |
2015-08-03 | set timeouts to time remaining rather than timeout duration | Matt Johnston | |
2015-08-03 | Fix problem where auth timeout wasn't checked when waiting for ident | Matt Johnston | |
2015-06-04 | buf_getstring and buf_putstring now use non-unsigned char* | Matt Johnston | |
2015-06-04 | Merge pull request #13 from gazoo74/fix-warnings | Matt Johnston | |
Fix warnings | |||
2015-05-05 | Fix pointer differ in signess warnings [-Werror=pointer-sign] | Gaël PORTAY | |
2015-05-03 | Make sure kexfirstinitialise is called early enough | Matt Johnston | |
2015-05-02 | Fix no-writev fallback | Matt Johnston | |
2015-03-20 | Make main socket nonblocking. Limit writequeue size. | Matt Johnston | |
2015-03-01 | reword comment for clarity | Matt Johnston | |
--HG-- branch : nocircbuffer | |||
2015-02-28 | Add cleanup | Matt Johnston | |
--HG-- branch : fastopen | |||
2015-02-28 | merge from default | Matt Johnston | |
--HG-- branch : fastopen | |||
2015-02-24 | Some additional cleanup functions | Matt Johnston | |
2015-02-24 | Free memory before exiting. Based on patch from Thorsten Horstmann. | Matt Johnston | |
Client side is not complete. | |||
2015-02-24 | Fix for old compilers, variable declarations at beginning of functions | Thorsten Horstmann | |
and /**/ comments | |||
2015-02-20 | Move generic network routines to netio.c | Matt Johnston | |
--HG-- branch : fastopen | |||
2015-02-20 | Update priority once the socket is open | Matt Johnston | |
--HG-- branch : fastopen | |||
2015-02-18 | async connections working | Matt Johnston | |
--HG-- branch : fastopen | |||
2015-02-13 | Add envirnonment variable for debug timestamps to roughly match | Matt Johnston | |
network timestamps (in tshark) | |||
2015-02-13 | Fix print that no longer works since we're not using fourCCs | Matt Johnston | |
2014-08-19 | Make keepalive handling more robust, this should now match what OpenSSH does | Matt Johnston | |
2014-08-13 | Don't send SSH_MSG_UNIMPLEMENTED for keepalive responses | Matt Johnston | |
2014-07-16 | Set tcp priority as follows: | Matt Johnston | |
if (connecting || ptys || x11) tos = LOWDELAY; else if (tcp_forwards) tos = 0; else tos = BULK; TCP forwards could be either lowdelay or bulk, hence the default priority. | |||
2014-07-09 | Fix auth timeout regression | Matt Johnston | |
2014-07-09 | Make -K keepalive behave like OpenSSH's ServerAliveInterval | Matt Johnston | |
2014-03-15 | Experiment of always writing data if available. Might waste a writev() with | Matt Johnston | |
EAGAIN but always saves a select() - needs testing with bandwidth-limited and CPU-limited situations. | |||
2014-03-13 | Add new monotonic_now() wrapper so that timeouts are unaffected by | Matt Johnston | |
system clock changes |