From: Hoeun Ryu <hoeun....@lge.com>
Make printk_safe_flush() safe in NMI context.
nmi_trigger_cpumask_backtrace() can be called in NMI context. For example the
function is called in watchdog_overflow_callback() if the flag of hardlockup
backtrace (sysctl_hardlockup_all_cpu_backtrace) is true and
watchdog_overflow_callback() function is called in NMI context on some
architectures.
Calling printk_safe_flush() in nmi_trigger_cpumask_backtrace() eventually tries
to lock logbuf_lock in vprintk_emit() that might be already be part
of another non-nmi context on the same CPU or a soft- or hard-lockup on another
CPU. The example of deadlock can be
CPU0
local_irq_save();
for (;;)
req = blk_peek_request(q);
if (unlikely(!scsi_device_online(sdev)))
printk()
vprintk_emit()
console_unlock()
logbuf_lock_irqsave()
slow-serial-console-write() // close to watchdog threshold
watchdog_overflow_callback()
trigger_allbutself_cpu_backtrace()
printk_safe_flush()
vprintk_emit()
logbuf_lock_irqsave()
^^^^ deadlock
and some other cases.
This patch prevents a deadlock in printk_safe_flush() in NMI context. It makes
sure that we continue and eventually call printk_safe_flush_on_panic() from
panic()
that has better chances to succeed.
There is a risk that logbuf_lock was not part of a soft- or dead-lockup and we
might just loose the messages. But then there is a high chance that irq_work
will
get called and the messages will get flushed the normal way.
Signed-off-by: Hoeun Ryu <hoeun....@lge.com>
Suggested-by: Petr Mladek <pmla...@suse.com>
Suggested-by: Sergey Senozhatsky <sergey.senozhatsky.w...@gmail.com>
---
v2: fix comments in commit message and code. no change in code itself.
kernel/printk/printk_safe.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/kernel/printk/printk_safe.c b/kernel/printk/printk_safe.c
index 3e3c200..3b5c660 100644
--- a/kernel/printk/printk_safe.c
+++ b/kernel/printk/printk_safe.c
@@ -254,6 +254,17 @@ void printk_safe_flush(void)
{
int cpu;
+ /*
+ * Just avoid a deadlock here.
+ * It makes sure that we continue and eventually call
+ * printk_safe_flush_on_panic() from panic() that has better chances to
succeed.
+ * There is a risk that logbuf_lock was not part of a soft- or
dead-lockup and
+ * we might just loose the messages. But then there is a high chance
that
+ * irq_work will get called and the messages will get flushed the
normal way.
+ */
+ if (this_cpu_read(printk_context) & PRINTK_NMI_CONTEXT_MASK)
+ return;
+
for_each_possible_cpu(cpu) {
#ifdef CONFIG_PRINTK_NMI
__printk_safe_flush(&per_cpu(nmi_print_seq, cpu).work);
--
2.1.4
