linux-kernel - Full dynticks needs evtdesc set before marking cpu online.

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-next>] [day] [month] [year] [list]

Date:	Wed, 8 May 2013 18:57:36 -0500
From:	Robin Holt <holt@....com>
To:	Thomas Gleixner <tglx@...utronix.de>
Cc:	Frederic Weisbecker <fweisbec@...il.com>,
	linux-kernel@...r.kernel.org, Ingo Molnar <mingo@...nel.org>
Subject: Full dynticks needs evtdesc set before marking cpu online.

Thomas,

We are seeing failures booting medium sized machines which I think is
a change in expectations that dyntick put on x86's start_secondary.

During boot of cpus, we see an occassional panic in tick_do_broadcast at

195         if (!cpumask_empty(mask)) {
196                 /*
197                  * It might be necessary to actually check whether the devices
198                  * have different broadcast functions. For now, just use the
199                  * one of the first device. This works as long as we have this
200                  * misfeature only on x86 (lapic)
201                  */
202                 td = &per_cpu(tick_cpu_device, cpumask_first(mask));
203                 td->evtdev->broadcast(mask);
		        ^^^^^^
             NULL  --------+


This is called from:
211 static void tick_do_periodic_broadcast(void)
212 {
213         raw_spin_lock(&tick_broadcast_lock);
214 
215         cpumask_and(tmpmask, cpu_online_mask, tick_broadcast_mask);
216         tick_do_broadcast(tmpmask);


Now the problem.  In start_secondary, we have:
 272         lock_vector_lock();
 273         set_cpu_online(smp_processor_id(), true);
 274         unlock_vector_lock();
 275         per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
 276         x86_platform.nmi_init();
 277 
 278         /* enable local interrupts */
 279         local_irq_enable();
 280 
 281         /* to prevent fake stack check failure in clock setup */
 282         boot_init_stack_canary();
 283 
 284         x86_cpuinit.setup_percpu_clockev();

So we have the cpu marked online on line 273, but evtdesc is not set
until line 284.  This code has been in start_secondary for a considerable
period of time.  I think it is just being revealed now.

It does not show up with a normal config, but taking a 'make
x86_64_defconfig' kernel and changing CONFIG_MAXSMP seems to change boot
timing enouogh to make it reproducible on 4 socket and above machines.

The following makes it boot, but I am not sure if this is the right
thing to do.

$ git diff
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 9c73b51..8456432 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -264,6 +264,8 @@ notrace static void __cpuinit start_secondary(void *unused)
         */
        check_tsc_sync_target();
 
+       x86_cpuinit.setup_percpu_clockev();
+
        /*
         * We need to hold vector_lock so there the set of online cpus
         * does not change while we are assigning vectors to cpus.  Holding
@@ -281,8 +283,6 @@ notrace static void __cpuinit start_secondary(void *unused)
        /* to prevent fake stack check failure in clock setup */
        boot_init_stack_canary();
 
-       x86_cpuinit.setup_percpu_clockev();
-
        wmb();
        cpu_startup_entry(CPUHP_ONLINE);
 }


Thanks,
Robin Holt
--
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/