[uml-devel] [patch] fix uml slowness caused by ptrace preemption bug on host

[Miklos S. <mi...@sz...>]

From: Miklos Szeredi <msz...@su...>

This patch fixes bug #12208.

This turned out to be not a scheduler regression, but an already
existing problem in ptrace being triggered by subtle scheduler
changes.

The problem is this:

 - task A is ptracing task B
 - task B stops on a trace event
 - task A is woken up and preempts task B
 - task A calls ptrace on task B, which does ptrace_check_attach()
 - this calls wait_task_inactive(), which sees that task B is still on the runq
 - task A goes to sleep for a jiffy
 - ...

Since UML does lots of the above sequences, those jiffies quickly add
up to make it slow as hell.

This patch solves this by not scheduling on preempt_enable() after
ptrace_stop() has woken up the tracer.

Signed-off-by: Miklos Szeredi <msz...@su...>
---
 kernel/signal.c |    7 +++++++
 1 file changed, 7 insertions(+)

Index: linux.git/kernel/signal.c
===================================================================
--- linux.git.orig/kernel/signal.c	2009-03-19 22:48:59.000000000 +0100
+++ linux.git/kernel/signal.c	2009-03-19 22:54:22.000000000 +0100
@@ -1572,10 +1572,16 @@ static void ptrace_stop(int exit_code, i
 	/* Let the debugger run.  */
 	__set_current_state(TASK_TRACED);
 	spin_unlock_irq(&current->sighand->siglock);
+	preempt_disable();
 	read_lock(&tasklist_lock);
 	if (may_ptrace_stop()) {
 		do_notify_parent_cldstop(current, CLD_TRAPPED);
 		read_unlock(&tasklist_lock);
+		/*
+		 * Don't want to allow preemption here, because
+		 * sys_ptrace() needs this task to be inactive.
+		 */
+		preempt_enable_no_resched();
 		schedule();
 	} else {
 		/*
@@ -1586,6 +1592,7 @@ static void ptrace_stop(int exit_code, i
 		if (clear_code)
 			current->exit_code = 0;
 		read_unlock(&tasklist_lock);
+		preempt_enable();
 	}
 
 	/*