kolan
/
CMake
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

KWSys: Restore SIGSTOP/SIGKILL to end process tree 

On UNIX systems we kill a tree of processes by performing a DFS walk of
the tree.  We send SIGSTOP to each process encountered, recursively
handle its children, and then send SIGKILL.

We once used the above approach in the past, but it was removed by the
commit "Do not send both SIGSTOP and SIGKILL when killing a process".
The commit was meant to work-around an OS X 10.3 bug in which the child
would not always honor SIGKILL after SIGSTOP.  At the time we wrongly
assumed that the process tree remains intact after SIGKILL and before
the child is reaped.  In fact the grandchildren may be re-parented to
ppid=1 even before the child is reaped, which causes the DFS walk to
miss them.
This commit is contained in:
Brad King 2009-11-30 13:14:02 -05:00
parent 4f43c1f276
commit b21f1eaeae
1 changed files with 15 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
Source/kwsys/ProcessUNIX.c
View File

@ -2392,13 +2392,8 @@ static void kwsysProcessKill(pid_t process_id)
DIR* procdir;
#endif
/* Kill the process now to make sure it does not create more
children. Do not reap it yet so we can identify its existing
children. There is a small race condition here. If the child
forks after we begin looking for children below but before it
receives this kill signal we might miss a child. Also we might
not be able to catch up to a fork bomb. */
kill(process_id, SIGKILL);
/* Suspend the process to be sure it will not create more children. */
kill(process_id, SIGSTOP);
/* Kill all children if we can find them. */
#if defined(__linux__) || defined(__CYGWIN__)
@ -2486,6 +2481,19 @@ static void kwsysProcessKill(pid_t process_id)
}
#endif
}
/* Kill the process. */
kill(process_id, SIGKILL);
#if defined(__APPLE__)
/* On OS X 10.3 the above SIGSTOP occasionally prevents the SIGKILL
from working. Just in case, we resume the child and kill it
again. There is a small race condition in this obscure case. If
the child manages to fork again between these two signals, we
will not catch its children. */
kill(process_id, SIGCONT);
kill(process_id, SIGKILL);
#endif
}
/*--------------------------------------------------------------------------*/