Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 9 May 2015 18:32:14 +0000 (UTC)
From:      Konstantin Belousov <kib@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r282679 - head/sys/kern
Message-ID:  <201505091832.t49IWEhZ047552@svn.freebsd.org>

next in thread | raw e-mail | index | archive | help
Author: kib
Date: Sat May  9 18:32:13 2015
New Revision: 282679
URL: https://svnweb.freebsd.org/changeset/base/282679

Log:
  Do not return from thread_single(SINGLE_BOUNDARY) until all stopped
  thread are guarenteed to be removed from the processors.
  
  Tested by:	pho
  Sponsored by:	The FreeBSD Foundation
  MFC after:	1 week

Modified:
  head/sys/kern/kern_thread.c

Modified: head/sys/kern/kern_thread.c
==============================================================================
--- head/sys/kern/kern_thread.c	Sat May  9 12:28:48 2015	(r282678)
+++ head/sys/kern/kern_thread.c	Sat May  9 18:32:13 2015	(r282679)
@@ -759,6 +759,29 @@ stopme:
 			PROC_LOCK(p);
 			PROC_SLOCK(p);
 		}
+	} else if (mode == SINGLE_BOUNDARY) {
+		/*
+		 * Wait until all suspended threads are removed from
+		 * the processors.  The thread_suspend_check()
+		 * increments p_boundary_count while it is still
+		 * running, which makes it possible for the execve()
+		 * to destroy vmspace while our other threads are
+		 * still using the address space.
+		 *
+		 * We lock the thread, which is only allowed to
+		 * succeed after context switch code finished using
+		 * the address space.
+		 */
+		FOREACH_THREAD_IN_PROC(p, td2) {
+			if (td2 == td)
+				continue;
+			thread_lock(td2);
+			KASSERT((td2->td_flags & TDF_BOUNDARY) != 0,
+			    ("td %p not on boundary", td2));
+			KASSERT(TD_IS_SUSPENDED(td2),
+			    ("td %p is not suspended", td2));
+			thread_unlock(td2);
+		}
 	}
 	PROC_SUNLOCK(p);
 	return (0);



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201505091832.t49IWEhZ047552>