The tcp_timer_stop() is designed to be called when the pcb belongs solely
to the calling thread and can't be looked up via in_pcblookup(). The
function may release lock which creates a race if used on a pcb that can
be found by an other thread.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG FreeBSD src repository
- Lint
Lint Skipped - Unit
Tests Skipped - Build Status
Buildable 57975 Build 54863: arc lint + arc unit
Event Timeline
Comment Actions
If the summary is true, wouldn't it be good to KASSERT that (inp->inp_flags & INP_DROPPED) != 0 after we have removed the two calls of tcp_timer_stop() from the RACK code? Willing to work on it...