Unkillable KSE threaded proc

Andrew Gallatin gallatin at cs.duke.edu
Tue Sep 14 09:06:18 PDT 2004

FWIW, for the case where there is one lingering thread, calling
thread_unsuspend_one() on it seems to get it to exit..

Maybe there is some sort of race while exiting which causes the wrong
number of threads to be either suspended, or unsuspended.  If too many
are suspended, one is left lingering.  If too few are suspended, the
system deadlocks because a thread never gets off the cpu.

Would it help at all to try with libthr and see what it does?
Let me know what more I can do to help get this fixed..


By "one lingering thread", I mean the case I first complained about.

540 c164e700 e52e1000 1387     1   538 000c482 (threaded)  mx_pingpong
   thread 0xc1fb8320 ksegrp 0xc15bb850 [SUSP]

db> tr 540
sched_switch(c1fb8320,0,0,15fc9814,e30bebc7) at sched_switch+0xd8
mi_switch(1,0,e881fc44,c051e6dd,c1fb8320) at mi_switch+0x1c7
thread_single(1,c06eaae0,e881fc64,c164e700,c1fb8320) at
exit1(c1fb8320,9,0,e881fce4,c051877e) at exit1+0x115
expand_name(c1fb8320,9,100,0,0) at expand_name
postsig(9,202,c06e5dd8,17f,8058f84) at postsig+0x204
ast(e881fd48) at ast+0x5e4
doreti_ast() at doreti_ast+0x17
db> call thread_unsuspend_one(0xc1fb8320)

More information about the freebsd-threads mailing list