From: "Steven Rostedt (VMware)" The variable printk_safe_irq_ready is set and never cleared at system boot up, when there's only one CPU active. It is set before other CPUs come on line. Also, it is extremely unlikely that an NMI would trigger this early in boot up (which I wonder why we even have this variable at all). Also mark the printk_safe_irq_ready as read mostly, as it is set at system boot up, and never touched again. Link: http://lkml.kernel.org/r/20171011124647.7781f98f@gandalf.local.home Reviewed-by: Petr Mladek Signed-off-by: Steven Rostedt (VMware) --- kernel/printk/printk_safe.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/kernel/printk/printk_safe.c b/kernel/printk/printk_safe.c index 3cdaeaef9ce1..724d9292d4b9 100644 --- a/kernel/printk/printk_safe.c +++ b/kernel/printk/printk_safe.c @@ -39,7 +39,7 @@ * There are situations when we want to make sure that all buffers * were handled or when IRQs are blocked. */ -static int printk_safe_irq_ready; +static int printk_safe_irq_ready __read_mostly; #define SAFE_LOG_BUF_LEN ((1 << CONFIG_PRINTK_SAFE_LOG_BUF_SHIFT) - \ sizeof(atomic_t) - \ @@ -63,11 +63,8 @@ static DEFINE_PER_CPU(struct printk_safe_seq_buf, nmi_print_seq); /* Get flushed in a more safe context. */ static void queue_flush_work(struct printk_safe_seq_buf *s) { - if (printk_safe_irq_ready) { - /* Make sure that IRQ work is really initialized. */ - smp_rmb(); + if (printk_safe_irq_ready) irq_work_queue(&s->work); - } } /* @@ -398,8 +395,12 @@ void __init printk_safe_init(void) #endif } - /* Make sure that IRQ works are initialized before enabling. */ - smp_wmb(); + /* + * In the highly unlikely event that a NMI were to trigger at + * this moment. Make sure IRQ work is set up before this + * variable is set. + */ + barrier(); printk_safe_irq_ready = 1; /* Flush pending messages that did not have scheduled IRQ works. */ -- 2.13.2