x86: Avoid calling arch_trigger_all_cpu_backtrace() at the same time
authorDongdong Deng <dongdong.deng@windriver.com>
Wed, 5 Jan 2011 03:38:08 +0000 (22:38 -0500)
committerIngo Molnar <mingo@elte.hu>
Wed, 5 Jan 2011 13:22:57 +0000 (14:22 +0100)
commit554ec063982752e9a569ab9189eeffa3d96731b2
tree1f27aba0e34881c1d8b428d18e1c1fbaa866c612
parent9ab181fa9ff73a38fccd0a4f1c40a38dfe62b535
x86: Avoid calling arch_trigger_all_cpu_backtrace() at the same time

The spin_lock_debug/rcu_cpu_stall detector uses
trigger_all_cpu_backtrace() to dump cpu backtrace.
Therefore it is possible that trigger_all_cpu_backtrace()
could be called at the same time on different CPUs, which
triggers and 'unknown reason NMI' warning. The following case
illustrates the problem:

      CPU1                    CPU2                     ...   CPU N
                       trigger_all_cpu_backtrace()
                       set "backtrace_mask" to cpu mask
                               |
generate NMI interrupts  generate NMI interrupts       ...
    \                          |                               /
     \                         |                              /

The "backtrace_mask" will be cleaned by the first NMI interrupt
at nmi_watchdog_tick(), then the following NMI interrupts
generated by other cpus's arch_trigger_all_cpu_backtrace() will
be taken as unknown reason NMI interrupts.

This patch uses a test_and_set to avoid the problem, and stop
the arch_trigger_all_cpu_backtrace() from calling to avoid
dumping a double cpu backtrace info when there is already a
trigger_all_cpu_backtrace() in progress.

Signed-off-by: Dongdong Deng <dongdong.deng@windriver.com>
Reviewed-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Cc: fweisbec@gmail.com
LKML-Reference: <1294198689-15447-2-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Don Zickus <dzickus@redhat.com>
arch/x86/kernel/apic/hw_nmi.c