- fixed issue where drift was being reset before saving on exit - added ptpengine:leap_seconds_notice_period for masters, with default of 12 hours (not 24 hours) - fixed issue where unicast node slots were not fully freed up - issue only when working at maximum capacity - fixed issue where denied grants when unicast capacity exceeded were showing incorrect message type
flag in TransportSpecific as an option ptpengine:dot2as - added master IP to status update - ensured master IP is updated as early as possible - fixed more dbg() recursions - added source MAC comparison so message counters are no longer updated for messages from self when running Ethernet transport
- use SIGNALING grantData entries for maintaining sequence numbers instead of incorrectly incrementing sequence numbers for other message types and using them for signaling - caused sequence numbers to start from 1 and discontinuities on renewals
- added FNV hashing function instead of naive sum+modulo - added sync destination hashing for faster lookups at high slave numbers. - fixed TX timestamp code issues where functions that should not b e there, got triggered. - fix for P2P bug where P2P was poluting unicast grant node index - increased UDP receive buffer size to allow processing of all messages.
- updated README and INSTALL - maximum unicast destination count changed to 2048 in configure.ac - RPM specs updated for rc5 - added hostname+pid info to status file - added "Status update:" line to periodic updates - fixed unused variable warnings for debug inet_ntoa() supporting struct in_addr - removed unicastAddr completely - ensured that initial timeLeft is set to requested +10 on first grant
unicast grant lookups - Added simple index table implementation for fast unicast Sync destination lookups (when no destination address available from Sync message, such as when force-looping packets), - Added re-requesting of messages if x not received, - Improved grant re-requesting on GM failures
- message counters (for rates) now exclude packets sent to self - ptpd now binds to the interface address when using unicast - needs tested on all OSes - updated INSTALL - added --disable-so-timestamping as last resort for cases where it's causing issues - increased upper boundary of unciast destination table to 1024. - minor refactoring - version bumped to 2.3.1-rc5 - fixed last minute showstopper where master would not send messages in hybrid mode - do not attempt retrieving IP addresses from Ethernet frames anymore - removed erroneous bind() to multicast address, which was breaking ptpengine:pid_as_clock_identity - added periodic warning when maximum unicast capacity reached
- Added debug warning when unicast grant table full, - Fixed condition trigering re-requesting announce: this applies to slaves only - was causing masters to cancel unicast grants for Announce.
master mode, based on the kernel's idea of whether one should or should not be inserted at the end of the day.
- increased robustness to network failures: * issue full network reset if multicast refresh failed (both master and slave) * issue full network reset after N slave resets (default is 5) - covers unicast cases - minor config file parser reload flag fixes - cleaned up multicast init failure error messages - version updated to 2.3.1-rc3