Re: [patch] exclude sync signals from signalfd sets

[Oleg Nesterov <oleg@tv-sign.ru>]

On 06/20, Linus Torvalds wrote:
> 
> 
> On Wed, 20 Jun 2007, Davide Libenzi wrote:
> > 
> > The following patch excludes synchronous signals from being dequeued
> > by a signalfd.
> 
> I really prefer the current code. Listign special cases is just ugly. Just 
> make it so that thread-local signals stay thread-local, and it's all good.

OK. How about the patch below then? I think this is what Nicholas suggested.

(Sorry for persistance! I can't explain why I think the current behaviour is
 not good. Just a personal feeling).

Oleg.

--- signalfd.c~	2007-06-21 13:00:21.000000000 +0400
+++ signalfd.c	2007-06-21 13:04:45.000000000 +0400
@@ -208,6 +208,15 @@ static int signalfd_copyinfo(struct sign
 	return err ? -EFAULT: sizeof(*uinfo);
 }
 
+static int
+__signalfd_dequeue(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
+{
+	if (tsk->tgid == current->tgid)
+		tsk = current;
+
+	return dequeue_signal(tsk, mask, info);
+}
+
 static ssize_t signalfd_dequeue(struct signalfd_ctx *ctx, siginfo_t *info,
 				int nonblock)
 {
@@ -218,7 +227,7 @@ static ssize_t signalfd_dequeue(struct s
 	if (!signalfd_lock(ctx, &lk))
 		return 0;
 
-	ret = dequeue_signal(lk.tsk, &ctx->sigmask, info);
+	ret = __signalfd_dequeue(lk.tsk, &ctx->sigmask, info);
 	switch (ret) {
 	case 0:
 		if (!nonblock)
@@ -232,7 +241,7 @@ static ssize_t signalfd_dequeue(struct s
 	add_wait_queue(&ctx->wqh, &wait);
 	for (;;) {
 		set_current_state(TASK_INTERRUPTIBLE);
-		ret = dequeue_signal(lk.tsk, &ctx->sigmask, info);
+		ret = __signalfd_dequeue(lk.tsk, &ctx->sigmask, info);
 		signalfd_unlock(&lk);
 		if (ret != 0)
 			break;
@@ -330,7 +339,7 @@ asmlinkage long sys_signalfd(int ufd, si
 
 		init_waitqueue_head(&ctx->wqh);
 		ctx->sigmask = sigmask;
-		ctx->tsk = current;
+		ctx->tsk = current->group_leader;
 
 		sighand = current->sighand;
 		/*