linux-kernel - [PATCH -next] kernel/watchdog_hld.c: Complete the flush of the hard deadlock log messages to the serial port

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>] [day] [month] [year] [list]

Date:   Mon, 2 Nov 2020 06:38:03 +0000
From:   Xu Qiang <xuqiang36@...wei.com>
To:     <sfr@...b.auug.org.au>, <akpm@...ux-foundation.org>,
        <koct9i@...il.com>, <xuqiang36@...wei.com>,
        <linux-kernel@...r.kernel.org>
CC:     <rui.xiang@...wei.com>
Subject: [PATCH -next] kernel/watchdog_hld.c: Complete the flush of the hard deadlock log messages to the serial port

when hardlockup_all_cpu_backtrace is on, and there are
a large number of cores in the system, it takes
a long time to output the hard deadlock logs of all cores
to the serial port. When the console_flush_on_panic function
in panic is executed, console_locked is still held.
As a result, garbled characters are displayed in the serial port log.

To solve this problem, wait for a maximum of 10s for the serial port
to be released before entering the panic mode.

Signed-off-by: Xu Qiang <xuqiang36@...wei.com>
---
 kernel/watchdog_hld.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/kernel/watchdog_hld.c b/kernel/watchdog_hld.c
index a546bc54f6ff..d3410b9fd3c3 100644
--- a/kernel/watchdog_hld.c
+++ b/kernel/watchdog_hld.c
@@ -13,8 +13,10 @@
 #define pr_fmt(fmt) "NMI watchdog: " fmt
 
 #include <linux/nmi.h>
+#include <linux/delay.h>
 #include <linux/atomic.h>
 #include <linux/module.h>
+#include <linux/console.h>
 #include <linux/sched/debug.h>
 
 #include <asm/irq_regs.h>
@@ -129,6 +131,7 @@ static void watchdog_overflow_callback(struct perf_event *event,
 	 * then this is a good indication the cpu is stuck
 	 */
 	if (is_hardlockup()) {
+		int i;
 		int this_cpu = smp_processor_id();
 
 		/* only print hardlockups once */
@@ -151,9 +154,15 @@ static void watchdog_overflow_callback(struct perf_event *event,
 		if (sysctl_hardlockup_all_cpu_backtrace &&
 				!test_and_set_bit(0, &hardlockup_allcpu_dumped))
 			trigger_allbutself_cpu_backtrace();
-
-		if (hardlockup_panic)
+		if (hardlockup_panic) {
+			/* Wait for all CPUs to complete wake_up_klogd_work_func */
+			for (i = 0; i < 10 * 1000; i++) {
+				if (console_trylock())
+					break;
+				mdelay(1);
+			}
 			nmi_panic(regs, "Hard LOCKUP");
+		}
 		atomic_inc(&hardlockup_detected);
 
 		__this_cpu_write(hard_watchdog_warn, true);
-- 
2.25.0