Change Details

All missing pieces to claim full RFC6675 support are in place, but disabled by default. These algorithms are all fenced behind the preexisting sysctl "rfc6675_pipe" (disabled by default). It was noted earlier, that referencing RFCs directly in sysctls is not the optimal choice. To better reflect these mechanisms belonging with SACK loss recovery, shuffling the sysctl into the net.inet.tcp.sack branch, and rename it to "newsack". Further, to get some additional testing coverage, enabling by default in HEAD, by mid-march. Unless this change (or the change of defaults) has to be backed out, planning for an MFC to stable/13 mid-april.

All missing pieces to claim full RFC6675 support are in place, but disabled by default. These algorithms are all fenced behind the preexisting sysctl "rfc6675_pipe" (disabled by default). It was noted earlier, that referencing RFCs directly in sysctls is not the optimal choice. To better reflect these mechanisms belonging with SACK loss recovery, shuffling the sysctl into the net.inet.tcp.sack branch, and rename it to "revised". Further, to get some additional testing coverage, enabling by default in HEAD, by mid-march. Unless this change (or the change of defaults) has to be backed out, planning for an MFC to stable/13 mid-april.

All missing pieces to claim full RFC6675 support are in place, but disabled by default. These algorithms are all fenced behind the preexisting sysctl "rfc6675_pipe" (disabled by default). It was noted earlier, that referencing RFCs directly in sysctls is not the optimal choice. To better reflect these mechanisms belonging with SACK loss recovery, shuffling the sysctl into the net.inet.tcp.sack branch, and rename it to "newsack""revised". Further, to get some additional testing coverage, enabling by default in HEAD, by mid-march. Unless this change (or the change of defaults) has to be backed out, planning for an MFC to stable/13 mid-april.