linux/arch/x86/kernel/crash.c
Yoshihiro YUNOMAE 17405453f4 x86/ioapic/kcrash: Prevent crash_kexec() from deadlocking on ioapic_lock
Prevent crash_kexec() from deadlocking on ioapic_lock. When
crash_kexec() is executed on a CPU, the CPU will take ioapic_lock
in disable_IO_APIC(). So if the cpu gets an NMI while locking
ioapic_lock, a deadlock will happen.

In this patch, ioapic_lock is zapped/initialized before disable_IO_APIC().

You can reproduce this deadlock the following way:

1. Add mdelay(1000) after raw_spin_lock_irqsave() in
   native_ioapic_set_affinity()@arch/x86/kernel/apic/io_apic.c

   Although the deadlock can occur without this modification, it will increase
   the potential of the deadlock problem.

2. Build and install the kernel

3. Set up the OS which will run panic() and kexec when NMI is injected
    # echo "kernel.unknown_nmi_panic=1" >> /etc/sysctl.conf
    # vim /etc/default/grub
      add "nmi_watchdog=0 crashkernel=256M" in GRUB_CMDLINE_LINUX line
    # grub2-mkconfig

4. Reboot the OS

5. Run following command for each vcpu on the guest
    # while true; do echo <CPU num> > /proc/irq/<IO-APIC-edge or IO-APIC-fasteoi>/smp_affinitity; done;
   By running this command, cpus will get ioapic_lock for setting affinity.

6. Inject NMI (push a dump button or execute 'virsh inject-nmi <domain>' if you
   use VM). After injecting NMI, panic() is called in an nmi-handler context.
   Then, kexec will normally run in panic(), but the operation will be stopped
   by deadlock on ioapic_lock in crash_kexec()->machine_crash_shutdown()->
   native_machine_crash_shutdown()->disable_IO_APIC()->clear_IO_APIC()->
   clear_IO_APIC_pin()->ioapic_read_entry().

Signed-off-by: Yoshihiro YUNOMAE <yoshihiro.yunomae.ez@hitachi.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: yrl.pp-manager.tt@hitachi.com
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Seiji Aguchi <seiji.aguchi@hds.com>
Link: http://lkml.kernel.org/r/20130820070107.28245.83806.stgit@yunodevel
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2013-08-20 09:26:33 +02:00

141 lines
3.2 KiB
C

/*
* Architecture specific (i386/x86_64) functions for kexec based crash dumps.
*
* Created by: Hariprasad Nellitheertha (hari@in.ibm.com)
*
* Copyright (C) IBM Corporation, 2004. All rights reserved.
*
*/
#include <linux/init.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/smp.h>
#include <linux/reboot.h>
#include <linux/kexec.h>
#include <linux/delay.h>
#include <linux/elf.h>
#include <linux/elfcore.h>
#include <linux/module.h>
#include <asm/processor.h>
#include <asm/hardirq.h>
#include <asm/nmi.h>
#include <asm/hw_irq.h>
#include <asm/apic.h>
#include <asm/hpet.h>
#include <linux/kdebug.h>
#include <asm/cpu.h>
#include <asm/reboot.h>
#include <asm/virtext.h>
int in_crash_kexec;
/*
* This is used to VMCLEAR all VMCSs loaded on the
* processor. And when loading kvm_intel module, the
* callback function pointer will be assigned.
*
* protected by rcu.
*/
crash_vmclear_fn __rcu *crash_vmclear_loaded_vmcss = NULL;
EXPORT_SYMBOL_GPL(crash_vmclear_loaded_vmcss);
static inline void cpu_crash_vmclear_loaded_vmcss(void)
{
crash_vmclear_fn *do_vmclear_operation = NULL;
rcu_read_lock();
do_vmclear_operation = rcu_dereference(crash_vmclear_loaded_vmcss);
if (do_vmclear_operation)
do_vmclear_operation();
rcu_read_unlock();
}
#if defined(CONFIG_SMP) && defined(CONFIG_X86_LOCAL_APIC)
static void kdump_nmi_callback(int cpu, struct pt_regs *regs)
{
#ifdef CONFIG_X86_32
struct pt_regs fixed_regs;
#endif
#ifdef CONFIG_X86_32
if (!user_mode_vm(regs)) {
crash_fixup_ss_esp(&fixed_regs, regs);
regs = &fixed_regs;
}
#endif
crash_save_cpu(regs, cpu);
/*
* VMCLEAR VMCSs loaded on all cpus if needed.
*/
cpu_crash_vmclear_loaded_vmcss();
/* Disable VMX or SVM if needed.
*
* We need to disable virtualization on all CPUs.
* Having VMX or SVM enabled on any CPU may break rebooting
* after the kdump kernel has finished its task.
*/
cpu_emergency_vmxoff();
cpu_emergency_svm_disable();
disable_local_APIC();
}
static void kdump_nmi_shootdown_cpus(void)
{
in_crash_kexec = 1;
nmi_shootdown_cpus(kdump_nmi_callback);
disable_local_APIC();
}
#else
static void kdump_nmi_shootdown_cpus(void)
{
/* There are no cpus to shootdown */
}
#endif
void native_machine_crash_shutdown(struct pt_regs *regs)
{
/* This function is only called after the system
* has panicked or is otherwise in a critical state.
* The minimum amount of code to allow a kexec'd kernel
* to run successfully needs to happen here.
*
* In practice this means shooting down the other cpus in
* an SMP system.
*/
/* The kernel is broken so disable interrupts */
local_irq_disable();
kdump_nmi_shootdown_cpus();
/*
* VMCLEAR VMCSs loaded on this cpu if needed.
*/
cpu_crash_vmclear_loaded_vmcss();
/* Booting kdump kernel with VMX or SVM enabled won't work,
* because (among other limitations) we can't disable paging
* with the virt flags.
*/
cpu_emergency_vmxoff();
cpu_emergency_svm_disable();
lapic_shutdown();
#ifdef CONFIG_X86_IO_APIC
/* Prevent crash_kexec() from deadlocking on ioapic_lock. */
ioapic_zap_locks();
disable_IO_APIC();
#endif
#ifdef CONFIG_HPET_TIMER
hpet_disable();
#endif
crash_save_cpu(regs, safe_smp_processor_id());
}