]> err.no Git - linux-2.6/commitdiff
freezer vs stopped or traced
authorRoland McGrath <roland@redhat.com>
Tue, 4 Mar 2008 04:22:05 +0000 (20:22 -0800)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Tue, 4 Mar 2008 15:59:54 +0000 (07:59 -0800)
This changes the "freezer" code used by suspend/hibernate in its treatment
of tasks in TASK_STOPPED (job control stop) and TASK_TRACED (ptrace) states.

As I understand it, the intent of the "freezer" is to hold all tasks
from doing anything significant.  For this purpose, TASK_STOPPED and
TASK_TRACED are "frozen enough".  It's possible the tasks might resume
from ptrace calls (if the tracer were unfrozen) or from signals
(including ones that could come via timer interrupts, etc).  But this
doesn't matter as long as they quickly block again while "freezing" is
in effect.  Some minor adjustments to the signal.c code make sure that
try_to_freeze() very shortly follows all wakeups from both kinds of
stop.  This lets the freezer code safely leave stopped tasks unmolested.

Changing this fixes the longstanding bug of seeing after resuming from
suspend/hibernate your shell report "[1] Stopped" and the like for all
your jobs stopped by ^Z et al, as if you had freshly fg'd and ^Z'd them.
It also removes from the freezer the arcane special case treatment for
ptrace'd tasks, which relied on intimate knowledge of ptrace internals.

Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/power/process.c
kernel/signal.c

index 7c2118f9597f3139cfef277233a739d5d717b159..f1d0b345c9ba86a24ac48ee29cc4cb6f800b1c94 100644 (file)
@@ -75,22 +75,15 @@ void refrigerator(void)
        __set_current_state(save);
 }
 
-static void fake_signal_wake_up(struct task_struct *p, int resume)
+static void fake_signal_wake_up(struct task_struct *p)
 {
        unsigned long flags;
 
        spin_lock_irqsave(&p->sighand->siglock, flags);
-       signal_wake_up(p, resume);
+       signal_wake_up(p, 0);
        spin_unlock_irqrestore(&p->sighand->siglock, flags);
 }
 
-static void send_fake_signal(struct task_struct *p)
-{
-       if (task_is_stopped(p))
-               force_sig_specific(SIGSTOP, p);
-       fake_signal_wake_up(p, task_is_stopped(p));
-}
-
 static int has_mm(struct task_struct *p)
 {
        return (p->mm && !(p->flags & PF_BORROWED_MM));
@@ -121,7 +114,7 @@ static int freeze_task(struct task_struct *p, int with_mm_only)
        if (freezing(p)) {
                if (has_mm(p)) {
                        if (!signal_pending(p))
-                               fake_signal_wake_up(p, 0);
+                               fake_signal_wake_up(p);
                } else {
                        if (with_mm_only)
                                ret = 0;
@@ -135,7 +128,7 @@ static int freeze_task(struct task_struct *p, int with_mm_only)
                } else {
                        if (has_mm(p)) {
                                set_freeze_flag(p);
-                               send_fake_signal(p);
+                               fake_signal_wake_up(p);
                        } else {
                                if (with_mm_only) {
                                        ret = 0;
@@ -182,15 +175,17 @@ static int try_to_freeze_tasks(int freeze_user_space)
                        if (frozen(p) || !freezeable(p))
                                continue;
 
-                       if (task_is_traced(p) && frozen(p->parent)) {
-                               cancel_freezing(p);
-                               continue;
-                       }
-
                        if (!freeze_task(p, freeze_user_space))
                                continue;
 
-                       if (!freezer_should_skip(p))
+                       /*
+                        * Now that we've done set_freeze_flag, don't
+                        * perturb a task in TASK_STOPPED or TASK_TRACED.
+                        * It is "frozen enough".  If the task does wake
+                        * up, it will immediately call try_to_freeze.
+                        */
+                       if (!task_is_stopped_or_traced(p) &&
+                           !freezer_should_skip(p))
                                todo++;
                } while_each_thread(g, p);
                read_unlock(&tasklist_lock);
index 84917fe507f77b8ff949af7f958e11ac2ed423b7..6af1210092c39a45db3552ecc9199728f8938410 100644 (file)
@@ -1623,7 +1623,6 @@ static void ptrace_stop(int exit_code, int clear_code, siginfo_t *info)
        /* Let the debugger run.  */
        __set_current_state(TASK_TRACED);
        spin_unlock_irq(&current->sighand->siglock);
-       try_to_freeze();
        read_lock(&tasklist_lock);
        if (!unlikely(killed) && may_ptrace_stop()) {
                do_notify_parent_cldstop(current, CLD_TRAPPED);
@@ -1640,6 +1639,13 @@ static void ptrace_stop(int exit_code, int clear_code, siginfo_t *info)
                read_unlock(&tasklist_lock);
        }
 
+       /*
+        * While in TASK_TRACED, we were considered "frozen enough".
+        * Now that we woke up, it's crucial if we're supposed to be
+        * frozen that we freeze now before running anything substantial.
+        */
+       try_to_freeze();
+
        /*
         * We are back.  Now reacquire the siglock before touching
         * last_siginfo, so that we are sure to have synchronized with
@@ -1757,9 +1763,15 @@ int get_signal_to_deliver(siginfo_t *info, struct k_sigaction *return_ka,
        sigset_t *mask = &current->blocked;
        int signr = 0;
 
+relock:
+       /*
+        * We'll jump back here after any time we were stopped in TASK_STOPPED.
+        * While in TASK_STOPPED, we were considered "frozen enough".
+        * Now that we woke up, it's crucial if we're supposed to be
+        * frozen that we freeze now before running anything substantial.
+        */
        try_to_freeze();
 
-relock:
        spin_lock_irq(&current->sighand->siglock);
        for (;;) {
                struct k_sigaction *ka;