linux-pm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jonas Meurer <jonas@freesources.org>
To: linux-pm@vger.kernel.org
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>,
	Pavel Machek <pavel@ucw.cz>, Len Brown <len.brown@intel.com>,
	Tim Dittler <tim.dittler@systemli.org>
Subject: [RFC PATCH] PM: Add a switch for disabling/enabling sync() before suspend
Date: Mon, 7 Oct 2019 12:50:14 +0200	[thread overview]
Message-ID: <56b2db6a-2f76-a6d3-662a-819cfb18d424@freesources.org> (raw)


[-- Attachment #1.1: Type: text/plain, Size: 5823 bytes --]

[Sorry, resending with the correct mailinglist address as recipient]

Hello,

This patch adds a run-time switch at `/sys/power/suspend_sync`.

The switch allows to enable or disable the final sync() from the suspend.c
Linux Kernel system suspend implementation. This is useful to avoid race
conditions if block devices have been suspended before. Be aware that you
have to take care of sync() yourself before suspending the system if you
disable it here.

Since this is my first patch against the Linux kernel and I don't
consider it ready for inclusion yet, I decided to send it to pm-linux
and the PM subsystem maintainers only first. Would be very glad if you
could take a look and comment on it :)

Some questions:

* There already is a build-time config flag[2] for en- or disabling the
  sync() in suspend.c. Is it acceptable to have both a build-time *and*
  a *run-time* switch? Or would a run-time switch have to replace the
  build-time switch? If so, a direct question to Rafael, as you added
  the build-time flag: Would that be ok for you?
* I'm unsure about the naming: since the default is to have the sync
  enabled, would `suspend_disable_sync` be a better name for the switch,
  obviously defaulting to 0 then and skipping the sync at value 1?

To give a bit more contect: In Debian, we're currently working[3] on
support to suspend unlocked dm-crypt devices before system suspend.
During that work, we realized that the final sync() from Linux Kernel
system suspend implementation can lead to a dead lock.

I wrote a simple reproducer[4] to cause the dead lock in a reliable way.


[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/kernel/power/suspend.c?id=54ecb8f#n569
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=2fd77f
[3] https://salsa.debian.org/mejo/cryptsetup-suspend
[4] https://salsa.debian.org/mejo/cryptsetup-suspend/snippets/334


Signed-off-by: Jonas Meurer <jonas@freesources.org>
---
 Documentation/ABI/testing/sysfs-power |   16 ++++++++++++++-
 include/linux/suspend.h               |    2 +
 kernel/power/main.c                   |   35 ++++++++++++++++++++++++++++++++++
 kernel/power/suspend.c                |    2 -
 4 files changed, 53 insertions(+), 2 deletions(-)

--- a/kernel/power/suspend.c
+++ b/kernel/power/suspend.c
@@ -575,7 +575,7 @@ static int enter_state(suspend_state_t s
 	if (state == PM_SUSPEND_TO_IDLE)
 		s2idle_begin();
 
-	if (!IS_ENABLED(CONFIG_SUSPEND_SKIP_SYNC)) {
+	if (!IS_ENABLED(CONFIG_SUSPEND_SKIP_SYNC) && suspend_sync_enabled) {
 		trace_suspend_resume(TPS("sync_filesystems"), 0, true);
 		ksys_sync_helper();
 		trace_suspend_resume(TPS("sync_filesystems"), 0, false);
--- a/include/linux/suspend.h
+++ b/include/linux/suspend.h
@@ -328,6 +328,7 @@ extern void arch_suspend_disable_irqs(vo
 extern void arch_suspend_enable_irqs(void);
 
 extern int pm_suspend(suspend_state_t state);
+extern bool suspend_sync_enabled;
 #else /* !CONFIG_SUSPEND */
 #define suspend_valid_only_mem	NULL
 
@@ -340,6 +341,7 @@ static inline bool pm_suspend_via_s2idle
 
 static inline void suspend_set_ops(const struct platform_suspend_ops *ops) {}
 static inline int pm_suspend(suspend_state_t state) { return -ENOSYS; }
+static inline bool suspend_sync_enabled(void) { return true; }
 static inline bool idle_should_enter_s2idle(void) { return false; }
 static inline void __init pm_states_init(void) {}
 static inline void s2idle_set_ops(const struct platform_s2idle_ops *ops) {}
--- a/kernel/power/main.c
+++ b/kernel/power/main.c
@@ -191,6 +191,40 @@ static ssize_t mem_sleep_store(struct ko
 power_attr(mem_sleep);
 #endif /* CONFIG_SUSPEND */
 
+#ifdef CONFIG_SUSPEND
+/*
+ * suspend_sync: invoke ksys_sync_helper() before suspend.
+ *
+ * show() returns whether ksys_sync_helper() is invoked before suspend.
+ * store() accepts 0 or 1.  0 disables ksys_sync_helper() and 1 enables it.
+ */
+bool suspend_sync_enabled = true;
+
+static ssize_t suspend_sync_show(struct kobject *kobj,
+				   struct kobj_attribute *attr, char *buf)
+{
+	return sprintf(buf, "%d\n", suspend_sync_enabled);
+}
+
+static ssize_t suspend_sync_store(struct kobject *kobj,
+				    struct kobj_attribute *attr,
+				    const char *buf, size_t n)
+{
+	unsigned long val;
+
+	if (kstrtoul(buf, 10, &val))
+		return -EINVAL;
+
+	if (val > 1)
+		return -EINVAL;
+
+	suspend_sync_enabled = !!val;
+	return n;
+}
+
+power_attr(suspend_sync);
+#endif /* CONFIG_SUSPEND */
+
 #ifdef CONFIG_PM_SLEEP_DEBUG
 int pm_test_level = TEST_NONE;
 
@@ -769,6 +803,7 @@ static struct attribute * g[] = {
 	&wakeup_count_attr.attr,
 #ifdef CONFIG_SUSPEND
 	&mem_sleep_attr.attr,
+	&suspend_sync_attr.attr,
 #endif
 #ifdef CONFIG_PM_AUTOSLEEP
 	&autosleep_attr.attr,
--- a/Documentation/ABI/testing/sysfs-power
+++ b/Documentation/ABI/testing/sysfs-power
@@ -300,4 +300,18 @@ Description:
 		attempt.
 
 		Using this sysfs file will override any values that were
-		set using the kernel command line for disk offset.
\ No newline at end of file
+		set using the kernel command line for disk offset.
+
+What:		/sys/power/suspend_sync
+Date:		October 2019
+Contact:	Jonas Meurer <jonas@freesources.org>
+Description:
+		This file controls the switch to enable or disable the final
+		sync() before system suspend. This is useful to avoid race
+		conditions if block devices have been suspended before. Be
+		aware that you have to take care of sync() yourself before
+		suspending the system if you disable it here.
+
+		Writing a "1" (default) to this file enables the sync() and
+		writing a "0" disables it. Reads from the file return the
+		current value.
-- 
Jonas Meurer






[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

             reply	other threads:[~2019-10-07 11:13 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-07 10:50 Jonas Meurer [this message]
2019-10-10 15:00 ` [RFC PATCH] PM: Add a switch for disabling/enabling sync() before suspend Jonas Meurer
2019-10-11 10:22 ` Rafael J. Wysocki
2019-10-14 17:46   ` Jonas Meurer
2019-10-14 17:48     ` [PATCH v2 1/2] " Jonas Meurer
2019-10-14 17:49     ` [PATCH v2 2/2] PM: Change CONFIG_SUSPEND_SKIP_SYNC to CONFIG_SKIP_SYNC_ON_SUSPEND Jonas Meurer
2019-11-04 10:51       ` [PATCH v3 2/2] PM: CONFIG_SUSPEND_SKIP_SYNC sets default for '/sys/power/sync_on_suspend' Jonas Meurer
2019-10-21 10:47     ` [RFC PATCH] PM: Add a switch for disabling/enabling sync() before suspend Jonas Meurer
2019-10-21 21:47       ` Rafael J. Wysocki
2019-10-22  8:54         ` Jonas Meurer
2019-11-04 10:57           ` Jonas Meurer
2019-11-12 11:00             ` Jonas Meurer
2019-12-02 14:12               ` Yannik Sembritzki
2019-12-02 17:05                 ` Jonas Meurer
2019-10-22 10:39       ` Pavel Machek
2019-10-31 15:56         ` Jonas Meurer

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=56b2db6a-2f76-a6d3-662a-819cfb18d424@freesources.org \
    --to=jonas@freesources.org \
    --cc=len.brown@intel.com \
    --cc=linux-pm@vger.kernel.org \
    --cc=pavel@ucw.cz \
    --cc=rjw@rjwysocki.net \
    --cc=tim.dittler@systemli.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).