Scripts I run using the latest cvs code (as of 10/15/07 @ 11:15 AM PDT) will not exit. A Ctrl-C or explicit exit() displays the following: stapio:sigproc:22 sigproc 2 (Interrupt) stapio:stp_main_loop:293 got STP_EXIT, closed=0 stapio:cleanup_and_exit:216 CLEANUP AND EXIT closed=0 stapio:close_relayfs:220 closing stapio:close_relayfs:239 done stapio:cleanup_and_exit:229 closing control channel staprun:cleanup:179 rc=3, inserted_module=1 staprun:stop_symbol_thread:464 Stopping symbol thread. staprun:cleanup:194 removing module stap_14f450c36f752aed577949ccb1f8d21b_501 then it just hangs. "rmmod <stap_module>" indicates the module is still in use: ERROR: Module stap_14f450c36f752aed577949ccb1f8d21b_501 is in use "ps -e | grep stap" indicates stap and staprun are still running: 24153 pts/4 00:00:00 stap 24155 pts/4 00:00:00 staprun "kill -9 <staprun_pid>" causes staprun and stap to exit and the module to be removed. I'm running the 2.6.23-mm1 kernel on x86_64.
Does this always happen? Was there something unusual about the scripts you were running? I tested on x86_64 but not on the mm1 kernel you are using. It seems unlikely the problem would be kernel related. I'll widen my testing. Hanging on the module removal indicates some resource is still in use.
I tried several scripts in the examples directory and all hang on exit. I didn't think this would be kernel version related either, but I just tried 2.6.22.1-41.fc7 and 2.6.23-rc8-mm2 and both work fine. No hangs. I'm only seeing the problem on 2.6.23-mm1. I'm using this kernel because it has the latest marker patches.
FYI... I just tried 2.6.23-mm1 on i386 and scripts hang there too.
There seems to be no shortage of complaints about the bugs in that kernel. In particular threads seem messed up. http://lkml.org/lkml/2007/10/13/175
(In reply to comment #4) > There seems to be no shortage of complaints about the bugs in that kernel. In > particular threads seem messed up. > > http://lkml.org/lkml/2007/10/13/175 Should we close this for now?
You might try the current linus git tree instead. As of yesterday, it has markers.
Funny you should mention that. I just tried linus' git tree with markers and they work fine, although I had to add 'source "kernel/Kconfig.instrumentation"' to the x86_64 Kconfig file. Probably an x86/x86_64 merge issue, since all other archs besides i386 and x86_64 do source that file. I'm also not seeing the "hang on script exit" problem on linus' git tree.
Let's disregard illness in this -mm tree.