linux-kernel - [PATCH] watchdog: when watchdog_enabled is 0, let (soft,nmi)switch remain 1 after we read them in proc

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>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20240822070944.GA13716@didi-ThinkCentre-M930t-N000>
Date: Thu, 22 Aug 2024 15:09:44 +0800
From: Tio Zhang <tiozhang@...iglobal.com>
To: <akpm@...ux-foundation.org>, <dianders@...omium.org>
CC: <linux-kernel@...r.kernel.org>, <tglx@...utronix.de>,
	<zyhtheonly@...il.com>, <tiozhang@...iglobal.com>, <zyhtheonly@...h.net>
Subject: [PATCH] watchdog: when watchdog_enabled is 0, let (soft,nmi)switch
 remain 1 after we read them in proc

For users set "watchdog_user_enabled=0" but remaining
"(soft,nmi)watchdog_user_enabled=1". Watchdog threads(,nmi watchdog)
rework only if users reset "watchdog_user_enabled=1" without printing
those watchdog swicthes. Otherwise (soft,nmi)watchdog_user_enabled
will turn to 0 because of printing their values (It makes sense to print 0
since they do not work any more, but it does not make sense to let user's
swicthes change to 0 only by prints).

And after that, watchdog only should work again by doing:
(soft,nmi)watchdog_user_enabled=1
*** can't print, or everything go back to 0 again ***
watchdog_user_enabled=1

So this patch fixes this situation:

| name                       | value
|----------------------------|--------------------------
| watchdog_enabled           | 0
|----------------------------|--------------------------
| nmi_watchdog_user_enabled  | 1
|----------------------------|--------------------------
| soft_watchdog_user_enabled | 1
|----------------------------|--------------------------
| watchdog_user_enabled      | 0
|----------------------------|--------------------------
    
            cat /proc/sys/kernel/*watchdog
                          |
                          |
                          V
| name                       | value
|----------------------------|--------------------------
| watchdog_enabled           | 0
|----------------------------|--------------------------
| nmi_watchdog_user_enabled  | 0
|----------------------------|--------------------------
| soft_watchdog_user_enabled | 0
|----------------------------|--------------------------
| watchdog_user_enabled      | 0
|----------------------------|--------------------------


Signed-off-by: Tio Zhang <tiozhang@...iglobal.com>
---
 kernel/watchdog.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 51915b44ac73..42e69e83e76d 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -995,8 +995,18 @@ static int proc_watchdog_common(int which, struct ctl_table *table, int write,
 		 * On read synchronize the userspace interface. This is a
 		 * racy snapshot.
 		 */
+		old = READ_ONCE(*param);
 		*param = (watchdog_enabled & which) != 0;
 		err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
+		/*
+		 * When "old" is 1 and watchdog_enabled is 0,
+		 * it should not be change to 0 for printing
+		 * nmi_watchdog_user_enabled or soft_watchdog_user_enabled.
+		 * So after we print it as 0,
+		 * we should recover it to 1.
+		 */
+		if (old && !watchdog_enabled)
+			*param = old;
 	} else {
 		old = READ_ONCE(*param);
 		err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
-- 
2.17.1