Sync hangs in WFBitMapT and protocol error receiving P_SIZES

[johannesthoma]

Primary (Windows 7) doing I/O Secondary (Linux) doing disonnect-connect-wait-connected loop. After about 10 iterations Primary hangs in ProtocolError loop withblocked: upper (no I/O possible on the resource). Alsodrbdadm status does not work any more.

Secondary (Linux) shows that we make it into WFBitMapT but then disconnect. Relevant log lines:

Oct 14 16:21:25 johannes-VirtualBox kernel: [ 6027.367265] drbd w0 linbit-wdrbd: conn( Connecting -> Connected ) peer( Unknown -> Primary )
Oct 14 16:21:25 johannes-VirtualBox kernel: [ 6027.367268] drbd w0/17 drbd26 linbit-wdrbd: pdsk( DUnknown -> UpToDate ) repl( Off -> WFBitMapT )
Oct 14 16:21:33 192.168.56.103  U14:21:25.603|03441960(drbd_s_w0) #1249 print_state_change <6>drbd w0 pnode-id:3, cs(Connecting), prole(Unknown), cflag(0x240e), scf(0xa1a): conn( Connecting -> Connected ) peer( Unknown -> Secondary )
Oct 14 16:21:33 192.168.56.103  U14:21:25.603|03441960(drbd_s_w0) #1250 print_state_change <6>drbd w0/17 minor 5 pnode-id:3, pdsk(Inconsistent), prpl(Off), pdvflag(0x81400): repl( Off -> WFBitMapS )
Oct 14 16:21:33 192.168.56.103  [last 2 messages were in IRQ context or recursive]
Oct 14 16:21:33 192.168.56.103  U14:21:33.494|03441550(drbd_r_w0) #1251 drbdd <3>drbd w0 pnode-id:3, cs(Connected), prole(Secondary), cflag(0x200e), scf(0x0): error receiving P_SIZES, e: -5 l: 0!
Oct 14 16:21:33 192.168.56.103  U14:21:33.494|03441550(drbd_r_w0) #1252 print_state_change <6>drbd w0 pnode-id:3, cs(Connected), prole(Secondary), cflag(0x200e), scf(0x21): conn( Connected -> ProtocolError ) peer( Secondary -> Unknown )
Oct 14 16:21:33 192.168.56.103  U14:21:33.494|03441550(drbd_r_w0) #1253 print_state_change <6>drbd w0/17 minor 5 pnode-id:3, pdsk(Inconsistent), prpl(WFBitMapS), pdvflag(0x81400): repl( WFBitMapS -> Off )
Oct 14 16:21:33 192.168.56.103  [last 2 messages were in IRQ context or recursive]
Oct 14 16:21:33 192.168.56.103  U14:21:33.494|0352b2a0(drbd_a_w0) #1254 drbd_ack_receiver <6>drbd w0 pnode-id:3, cs(ProtocolError), prole(Unknown), cflag(0x200e), scf(0x0): ack_receiver terminated

It looks like the same error HenryKellner reported already via Slack.

[johannesthoma]

Seems to be fixed with the wake_up() waking all sleepers patch. About 100 iterations and no error receiving P_SIZES.