linux-kernel - Re: [PATCH 2/6] [watchdog] convert touch_softlockup_watchdog to touch

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20100421213142.GY15159@redhat.com>
Date:	Wed, 21 Apr 2010 17:31:42 -0400
From:	Don Zickus <dzickus@...hat.com>
To:	Frederic Weisbecker <fweisbec@...il.com>
Cc:	mingo@...e.hu, peterz@...radead.org, gorcunov@...il.com,
	aris@...hat.com, linux-kernel@...r.kernel.org,
	randy.dunlap@...cle.com
Subject: Re: [PATCH 2/6] [watchdog] convert touch_softlockup_watchdog to
 touch_watchdog

On Wed, Apr 21, 2010 at 10:46:01PM +0200, Frederic Weisbecker wrote:
> On Tue, Apr 20, 2010 at 11:23:59AM -0400, Don Zickus wrote:
> > Just a scripted conversion to remove touch_softlockup_watchdog.
> > 
> > Also converts the once case of touch_all_softlockup_watchdogs to
> > touch_all_watchdogs.
> > 
> > This is done as part of the removal of the old softlockup code and
> > transition to the new softlockup code.
> > 
> > Signed-off-by: Don Zickus <dzickus@...hat.com>
> 
> 
> In fact I worry a bit about this unification of watchdog touching.
> When we touch the softlockup watchdog, do we also want to touch
> the nmi watchdog?
> 
> Most of the time, I think we don't want to. We usually touch the
> softlockup detector because we know we are abnormally delaying
> the softlockup kthread from being scheduled, and if we are in such
> situation, it means we are doing something in a sensitive context:
> typically the kind of context favorable to create hardlockups...
> 
> But the opposite is right: if we touch the nmi watchdog: it means we
> are abnormally delaying irqs, which means we also are abnormally
> delaying the softlockup kthread from being scheduled, so if we
> touch the nmi watchdog, we also want to touch the softlockup
> detector.
> 
> Hence I guess we want to keep the current state:
> 
> - touch_nmi_watchdog() = touch softlockup and nmi watchdogs
> - touch_softlockup_watchdog() = only touch softlockup watchdog

Hmm ok I see what you are saying.  A little tweak and I have this
compiled-tested only patch that I think satisifies you.

I didn't really touch the touch_nmi_watchdog() code in the kernel, so it
still calls a stub function in kernel/watchdog.c.  Add a boolean to that
path and I think it accomplishes the logic you are looking for.

Cheers,
Don


diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 9898c7c..c1a89ac 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -31,6 +31,7 @@ int watchdog_enabled;
 int __read_mostly softlockup_thresh = 60;
 
 static DEFINE_PER_CPU(unsigned long, watchdog_touch_ts);
+static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
 static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog);
 static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer);
 static DEFINE_PER_CPU(bool, watchdog_touch_sync);
@@ -147,6 +148,7 @@ void touch_watchdog_sync(void)
 
 void touch_nmi_watchdog(void)
 {
+	__get_cpu_var(watchdog_nmi_touch) = true;
 	touch_watchdog();
 }
 EXPORT_SYMBOL(touch_nmi_watchdog);
@@ -203,11 +205,10 @@ void watchdog_overflow_callback(struct perf_event *event, int nmi,
 		 struct pt_regs *regs)
 {
 	int this_cpu = smp_processor_id();
-	unsigned long touch_ts = per_cpu(watchdog_touch_ts, this_cpu);
 	char warn = per_cpu(watchdog_warn, this_cpu);
 
-	if (touch_ts == 0) {
-		__touch_watchdog();
+	if (__get_cpu_var(watchdog_nmi_touch) == true) {
+		__get_cpu_var(watchdog_nmi_touch) = false;
 		return;
 	}
 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/