public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] execve TGID dethreading bug fix
@ 2002-03-07  9:36 David Howells
  0 siblings, 0 replies; only message in thread
From: David Howells @ 2002-03-07  9:36 UTC (permalink / raw)
  To: torvalds; +Cc: linux-kernel, dhowells


Hi Linus,

I've attached a patch to kill all subsidiary threads in a thread group when
the main thread exits. I've made it against 2.5.6-pre3.

Features:

 - It sends the subsidiary threads SIGKILL with SI_DETHREAD.

 - Subsidiary threads doing an execve() just leave the thread group (rather
   than forcing the master thread to do an execve() which would be more POSIX
   like).

David


diff -uNr linux-2.5.6-pre3/fs/exec.c linux-execve-256p3/fs/exec.c
--- linux-2.5.6-pre3/fs/exec.c	Thu Mar  7 09:08:50 2002
+++ linux-execve-256p3/fs/exec.c	Thu Mar  7 09:13:47 2002
@@ -509,23 +509,49 @@
 
 /*
  * An execve() will automatically "de-thread" the process.
- * Note: we don't have to hold the tasklist_lock to test
- * whether we migth need to do this. If we're not part of
- * a thread group, there is no way we can become one
- * dynamically. And if we are, we only need to protect the
- * unlink - even if we race with the last other thread exit,
- * at worst the list_del_init() might end up being a no-op.
+ * - if a master thread (PID==TGID) is doing this, then all subsidiary threads
+ *   will be killed (otherwise there will end up being two independent thread
+ *   groups with the same TGID).
+ * - if a subsidary thread is doing this, then it just leaves the thread group
  */
-static inline void de_thread(struct task_struct *tsk)
+static void de_thread(struct task_struct *tsk)
 {
-	if (!list_empty(&tsk->thread_group)) {
-		write_lock_irq(&tasklist_lock);
+	struct task_struct *sub;
+	struct list_head *head, *ptr;
+	struct siginfo info;
+	int pause;
+
+	write_lock_irq(&tasklist_lock);
+
+	if (tsk->tgid != tsk->pid) {
+		/* subsidiary thread - just escapes the group */
+		list_del_init(&tsk->thread_group);
+		tsk->tgid = tsk->pid;
+		pause = 0;
+	}
+	else {
+		/* master thread - kill all subsidiary threads */
+		info.si_signo = SIGKILL;
+		info.si_errno = 0;
+		info.si_code = SI_DETHREAD;
+		info.si_pid = current->pid;
+		info.si_uid = current->uid;
+
+		head = tsk->thread_group.next;
 		list_del_init(&tsk->thread_group);
-		write_unlock_irq(&tasklist_lock);
+
+		list_for_each(ptr,head) {
+			sub = list_entry(ptr,struct task_struct,thread_group);
+			send_sig_info(SIGKILL,&info,sub);
+		}
+
+		pause = 1;
 	}
 
-	/* Minor oddity: this might stay the same. */
-	tsk->tgid = tsk->pid;
+	write_unlock_irq(&tasklist_lock);
+
+	/* give the subsidiary threads a chance to clean themselves up */
+	if (pause) yield();
 }
 
 int flush_old_exec(struct linux_binprm * bprm)
@@ -566,7 +592,8 @@
 
 	flush_thread();
 
-	de_thread(current);
+	if (!list_empty(&current->thread_group))
+		de_thread(current);
 
 	if (bprm->e_uid != current->euid || bprm->e_gid != current->egid || 
 	    permission(bprm->file->f_dentry->d_inode,MAY_READ))
diff -uNr linux-2.5.6-pre3/include/asm-i386/siginfo.h linux-execve-256p3/include/asm-i386/siginfo.h
--- linux-2.5.6-pre3/include/asm-i386/siginfo.h	Thu Mar  7 09:07:06 2002
+++ linux-execve-256p3/include/asm-i386/siginfo.h	Thu Mar  7 09:17:53 2002
@@ -108,6 +108,7 @@
 #define SI_ASYNCIO	-4		/* sent by AIO completion */
 #define SI_SIGIO	-5		/* sent by queued SIGIO */
 #define SI_TKILL	-6		/* sent by tkill system call */
+#define SI_DETHREAD	-7		/* sent by execve() killing subsidiary threads */
 
 #define SI_FROMUSER(siptr)	((siptr)->si_code <= 0)
 #define SI_FROMKERNEL(siptr)	((siptr)->si_code > 0)

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2002-03-07  9:37 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-03-07  9:36 [PATCH] execve TGID dethreading bug fix David Howells

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox