linux-kernel - [PATCH] timers: Add warning if timer list is corrupted

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]

Message-Id: <1695610822-3242-1-git-send-email-yw85.kim@samsung.com>
Date:   Mon, 25 Sep 2023 12:00:22 +0900
From:   "yw85.kim" <yw85.kim@...sung.com>
To:     Thomas Gleixner <tglx@...utronix.de>,
        John Stultz <jstultz@...gle.com>,
        Stephen Boyd <sboyd@...nel.org>, linux-kernel@...r.kernel.org
Cc:     "yw85.kim" <yw85.kim@...sung.com>
Subject: [PATCH] timers: Add warning if timer list is corrupted

[Description]
If some drivers use timers wrong, it can cause list corruption
(LIST_POISON2) and oops.

        detach_timer(struct timer_list *timer, bool clear_pending)
          __hlist_del(entry);
            next->pprev = pprev;  // oops, next is LIST_POISON2(0x122)

However, it is difficult to find the cause through backtrace.

        [1-11.7841] Unable to handle kernel NULL pointer dereference at
          virtual address 00000126
        [1-11.7841] pgd = 0ac52d81
        [1-11.7841] [00000126] *pgd=00000000
        Die cpu info :
        [1-11.7841] Internal error: Oops: 805 [#1] PREEMPT SMP ARM
        [1-11.7842] PC is at run_timer_softirq+0x1f0/0x684
        [1-11.7842] LR is at __warn+0xb4/0xf8
        ...
        [1-11.7842] Backtrace:
        [1-11.7842] [<c00bd464>] (run_timer_softirq)
        [1-11.7842] [<c000a438>] (__do_softirq+0x1ac/0x50c)
        [1-11.7842] [<c00362d4>] (run_ksoftirqd+0x84/0x98)
        [1-11.7842] [<c005a564>] (smpboot_thread_fn+0x140/0x2c0)
        [1-11.7843] [<c00560c0>] (kthread+0x1cc/0x1d0)

Adding this warning will give us more detailed information.
In this case the hdmi_switch driver was the cause.

A problem occurred because multiple schedule delayed works
were created using same dw(struct delayed_work dw) structure.

        [1-11.7835] WARNING: CPU: 1 PID: 17 at kernel/time/timer.c:830
          run_timer_softirq+0x50c/0x684
        [1-11.7836] timer: list corruption, entry->next is LIST_POISON2,
          entry:dw+0x10/0xfffe1a3c [hdmi_switch],
          fn:delayed_work_timer_fn+0x0/0x28

Signed-off-by: yw85.kim <yw85.kim@...sung.com>
---
 kernel/time/timer.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/time/timer.c b/kernel/time/timer.c
index 63a8ce7..e1e3340 100644
--- a/kernel/time/timer.c
+++ b/kernel/time/timer.c
@@ -877,6 +877,10 @@ static inline void detach_timer(struct timer_list *timer, bool clear_pending)
 
 	debug_deactivate(timer);
 
+	if (unlikely(entry->next == LIST_POISON2))
+		WARN_ONCE(1, "timer: list corruption, next is LIST_POISON2, entry:%pS, fn:%pS\n",
+			entry, timer->function);
+
 	__hlist_del(entry);
 	if (clear_pending)
 		entry->pprev = NULL;
-- 
2.7.4