linux-kernel - [RFC PATCH v2 2/5] kernel: allow to configure PREEMPT_NONE, PREEMPT

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-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20201009122926.29962-3-mhocko@kernel.org>
Date:   Fri,  9 Oct 2020 14:29:23 +0200
From:   Michal Hocko <mhocko@...nel.org>
To:     Peter Zijlstra <peterz@...radead.org>,
        Thomas Gleixner <tglx@...utronix.de>
Cc:     Mel Gorman <mgorman@...e.de>,
        Frederic Weisbecker <fweisbecker@...e.de>,
        Ingo Molnar <mingo@...hat.com>,
        LKML <linux-kernel@...r.kernel.org>,
        Michal Hocko <mhocko@...e.com>
Subject: [RFC PATCH v2 2/5] kernel: allow to configure PREEMPT_NONE, PREEMPT_VOLUNTARY on kernel command line

From: Michal Hocko <mhocko@...e.com>

Many people are still relying on pre built distribution kernels and so
distributions have to provide mutliple kernel flavors to offer different
preemption models. Most of them are providing PREEMPT_NONE for typical
server deployments and PREEMPT_VOLUNTARY for desktop users.

Having two different kernel binaries differing only by the preemption
mode seems rather wasteful and inflexible. Especially when the difference
between PREEMPT_NONE and PREEMPT_VOLUNTARY is really minimal. Both only
allow explicit scheduling points while running in the kernel and it is
only might_sleep which adds additional preemption points for
PREEMPT_VOLUNTARY.

Add a kernel command line parameter preempt_mode=[none, voluntary] which
allows to override the default compile time preemption mode
(CONFIG_PREEMPT_NONE resp. CONFIG_PREEMPT_VOLUTARY). Voluntary preempt
mode is guarded by a jump label to prevent any potential runtime overhead.

Add an explicit include of jump_label to gpio/consumer.h to make sure
all its consumers will get static_branch_likely) as kernel.h cannot
include it directly.

Signed-off-by: Michal Hocko <mhocko@...e.com>
---
 .../admin-guide/kernel-parameters.txt         |  5 ++++
 include/linux/gpio/consumer.h                 |  1 +
 include/linux/kernel.h                        | 13 ++++++--
 kernel/sched/core.c                           | 30 +++++++++++++++++++
 4 files changed, 47 insertions(+), 2 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index a1068742a6df..96bb74faeb50 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3873,6 +3873,11 @@
 			Format: {"off"}
 			Disable Hardware Transactional Memory
 
+	preempt_mode={none,voluntary}
+			Set the preemption mode.
+			none - equivalent to CONFIG_PREEMPT_NONE
+			voluntary - equivalent to CONFIG_PREEMPT_VOLUNTARY
+
 	print-fatal-signals=
 			[KNL] debug: print fatal signals
 
diff --git a/include/linux/gpio/consumer.h b/include/linux/gpio/consumer.h
index 901aab89d025..d64e6dda5755 100644
--- a/include/linux/gpio/consumer.h
+++ b/include/linux/gpio/consumer.h
@@ -3,6 +3,7 @@
 #define __LINUX_GPIO_CONSUMER_H
 
 #include <linux/bits.h>
+#include <linux/jump_label.h>
 #include <linux/bug.h>
 #include <linux/compiler_types.h>
 #include <linux/err.h>
diff --git a/include/linux/kernel.h b/include/linux/kernel.h
index c25b8e41c0ea..d2d37bd5ecd5 100644
--- a/include/linux/kernel.h
+++ b/include/linux/kernel.h
@@ -14,6 +14,7 @@
 #include <linux/typecheck.h>
 #include <linux/printk.h>
 #include <linux/build_bug.h>
+#include <linux/jump_label_type.h>
 #include <asm/byteorder.h>
 #include <asm/div64.h>
 #include <uapi/linux/kernel.h>
@@ -192,11 +193,19 @@ struct completion;
 struct pt_regs;
 struct user;
 
+#ifndef CONFIG_PREEMPTION
 #ifdef CONFIG_PREEMPT_VOLUNTARY
+DECLARE_STATIC_KEY_TRUE(preempt_voluntary_key);
+#else
+DECLARE_STATIC_KEY_FALSE(preempt_voluntary_key);
+#endif
+
 extern int _cond_resched(void);
-# define might_resched() _cond_resched()
+# define might_resched() \
+	do { if (static_branch_likely(&preempt_voluntary_key)) _cond_resched(); } while (0)
 #else
-# define might_resched() do { } while (0)
+# define might_resched() \
+	do { } while (0)
 #endif
 
 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 2d95dc3f4644..07d37d862637 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -43,6 +43,14 @@ EXPORT_TRACEPOINT_SYMBOL_GPL(sched_update_nr_running_tp);
 
 DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
 
+#ifdef CONFIG_PREEMPT_VOLUNTARY
+DEFINE_STATIC_KEY_TRUE(preempt_voluntary_key);
+#else
+/* PREEMPT_NONE vs PREEMPT_VOLUNTARY */
+DEFINE_STATIC_KEY_FALSE(preempt_voluntary_key);
+#endif
+EXPORT_SYMBOL(preempt_voluntary_key);
+
 #if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_JUMP_LABEL)
 /*
  * Debugging: various feature bits
@@ -8482,3 +8490,25 @@ void call_trace_sched_update_nr_running(struct rq *rq, int count)
 {
         trace_sched_update_nr_running_tp(rq, count);
 }
+
+#ifndef CONFIG_PREEMPTION
+static int __init setup_non_preempt_mode(char *str)
+{
+	if (!strcmp(str, "none")) {
+		if (IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY)) {
+			static_branch_disable(&preempt_voluntary_key);
+			pr_info("Switching to PREEMPT_NONE mode.");
+		}
+	} else if (!strcmp(str, "voluntary")) {
+		if (!IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY)) {
+			static_branch_enable(&preempt_voluntary_key);
+			pr_info("Switching to PREEMPT_VOLUNTARY mode.");
+		}
+	} else {
+		pr_warn("Unsupported preempt mode %s\n", str);
+		return 1;
+	}
+	return 0;
+}
+__setup("preempt_mode=", setup_non_preempt_mode);
+#endif
-- 
2.28.0