Date: Thu, 29 Dec 2011 06:20:36 +0100
   From: [email protected] (Emmanuel Dreyfus)

   I can reliabily crash latest netbsd-5 kernel with an unprivilegied
   ktrace -p on an unexisting process...

What prevents ktrace_common from returning -- and sys_ktrace from
subsequently fd_aborting the file -- before the ktrace thread has
dropped its reference to the file and exited?

Perhaps ktrace_common should take responsibility for the file, and
sys_ktrace should have responsibility only for the descriptor table
slot.

Reply via email to