sched_ext: Maintain an online cid mask in the scheduler arenagrafted

Schedulers on the default cid mapping treat [0, nr_online_cids) as the
online set and restart on hotplug. Schedulers that install their own mapping
with scx_bpf_cid_override() have no way to learn which cids are online: the
count no longer identifies members and the CPU-form cpumask is unusable from
cid programs. This is an obvious hole in the cid API.

Add scx_bpf_online_cmask(), a kernel-maintained cmask in the scheduler's
arena, allocated alongside the per-CPU scratch masks and populated after the
cid mapping is finalized and before ops.init(), for child schedulers too.
The pointer stays valid through ops.exit() with no reference to take. It is
the arena offset as a void pointer, the same form struct_ops arena arguments
arrive in. The verifier types the void return as a scalar for the program's
arena cast.

The mask follows the SCX hotplug notifications: seeded from cpu_active_mask
and updated before ops.cid_online/offline() runs, so it lags cpu_online_mask
only inside a hotplug transition. Updates walk the scheduler list under the
lock that also serializes unlinking. Reads are live, not atomic snapshots.
Root initialization excludes hotplug.

v2: Reworded the getter kerneldoc (Andrea Righi).

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>

1 files changed, 179 insertions, 0 deletions

diff --git a/drivers/char/hangcheck-timer.c b/drivers/char/hangcheck-timer.c
new file mode 100644
index 000000000..231cbf7b3
--- /dev/null
+++ b/drivers/char/hangcheck-timer.c
@@ -0,0 +1,179 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * hangcheck-timer.c
+ *
+ * Driver for a little io fencing timer.
+ *
+ * Copyright (C) 2002, 2003 Oracle.  All rights reserved.
+ *
+ * Author: Joel Becker <joel.becker@oracle.com>
+ */
+
+/*
+ * The hangcheck-timer driver uses the TSC to catch delays that
+ * jiffies does not notice.  A timer is set.  When the timer fires, it
+ * checks whether it was delayed and if that delay exceeds a given
+ * margin of error.  The hangcheck_tick module parameter takes the timer
+ * duration in seconds.  The hangcheck_margin parameter defines the
+ * margin of error, in seconds.  The defaults are 60 seconds for the
+ * timer and 180 seconds for the margin of error.  IOW, a timer is set
+ * for 60 seconds.  When the timer fires, the callback checks the
+ * actual duration that the timer waited.  If the duration exceeds the
+ * allotted time and margin (here 60 + 180, or 240 seconds), the machine
+ * is restarted.  A healthy machine will have the duration match the
+ * expected timeout very closely.
+ */
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/reboot.h>
+#include <linux/init.h>
+#include <linux/delay.h>
+#include <linux/uaccess.h>
+#include <linux/sysrq.h>
+#include <linux/timer.h>
+#include <linux/hrtimer.h>
+
+#define VERSION_STR "0.9.1"
+
+#define DEFAULT_IOFENCE_MARGIN 60	/* Default fudge factor, in seconds */
+#define DEFAULT_IOFENCE_TICK 180	/* Default timer timeout, in seconds */
+
+static int hangcheck_tick = DEFAULT_IOFENCE_TICK;
+static int hangcheck_margin = DEFAULT_IOFENCE_MARGIN;
+static int hangcheck_reboot;  /* Defaults to not reboot */
+static int hangcheck_dump_tasks;  /* Defaults to not dumping SysRQ T */
+
+/* options - modular */
+module_param(hangcheck_tick, int, 0);
+MODULE_PARM_DESC(hangcheck_tick, "Timer delay.");
+module_param(hangcheck_margin, int, 0);
+MODULE_PARM_DESC(hangcheck_margin, "If the hangcheck timer has been delayed more than hangcheck_margin seconds, the driver will fire.");
+module_param(hangcheck_reboot, int, 0);
+MODULE_PARM_DESC(hangcheck_reboot, "If nonzero, the machine will reboot when the timer margin is exceeded.");
+module_param(hangcheck_dump_tasks, int, 0);
+MODULE_PARM_DESC(hangcheck_dump_tasks, "If nonzero, the machine will dump the system task state when the timer margin is exceeded.");
+
+MODULE_AUTHOR("Oracle");
+MODULE_DESCRIPTION("Hangcheck-timer detects when the system has gone out to lunch past a certain margin.");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(VERSION_STR);
+
+/* options - nonmodular */
+#ifndef MODULE
+
+static int __init hangcheck_parse_tick(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_tick = par;
+	return 1;
+}
+
+static int __init hangcheck_parse_margin(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_margin = par;
+	return 1;
+}
+
+static int __init hangcheck_parse_reboot(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_reboot = par;
+	return 1;
+}
+
+static int __init hangcheck_parse_dump_tasks(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_dump_tasks = par;
+	return 1;
+}
+
+__setup("hcheck_tick", hangcheck_parse_tick);
+__setup("hcheck_margin", hangcheck_parse_margin);
+__setup("hcheck_reboot", hangcheck_parse_reboot);
+__setup("hcheck_dump_tasks", hangcheck_parse_dump_tasks);
+#endif /* not MODULE */
+
+#define TIMER_FREQ 1000000000ULL
+
+/* Last time scheduled */
+static unsigned long long hangcheck_tsc, hangcheck_tsc_margin;
+
+static void hangcheck_fire(struct timer_list *);
+
+static DEFINE_TIMER(hangcheck_ticktock, hangcheck_fire);
+
+static void hangcheck_fire(struct timer_list *unused)
+{
+	unsigned long long cur_tsc, tsc_diff;
+
+	cur_tsc = ktime_get_ns();
+
+	if (cur_tsc > hangcheck_tsc)
+		tsc_diff = cur_tsc - hangcheck_tsc;
+	else
+		tsc_diff = (cur_tsc + (~0ULL - hangcheck_tsc)); /* or something */
+
+	if (tsc_diff > hangcheck_tsc_margin) {
+		if (hangcheck_dump_tasks) {
+			pr_crit("Hangcheck: Task state:\n");
+#ifdef CONFIG_MAGIC_SYSRQ
+			handle_sysrq('t');
+#endif  /* CONFIG_MAGIC_SYSRQ */
+		}
+		if (hangcheck_reboot) {
+			pr_crit("Hangcheck: hangcheck is restarting the machine.\n");
+			emergency_restart();
+		} else {
+			pr_crit("Hangcheck: hangcheck value past margin!\n");
+		}
+	}
+#if 0
+	/*
+	 * Enable to investigate delays in detail
+	 */
+	pr_debug("Hangcheck: called %lld ns since last time (%lld ns overshoot)\n",
+			tsc_diff, tsc_diff - hangcheck_tick*TIMER_FREQ);
+#endif
+	mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ));
+	hangcheck_tsc = ktime_get_ns();
+}
+
+
+static int __init hangcheck_init(void)
+{
+	pr_debug("Hangcheck: starting hangcheck timer %s (tick is %d seconds, margin is %d seconds).\n",
+	       VERSION_STR, hangcheck_tick, hangcheck_margin);
+	hangcheck_tsc_margin =
+		(unsigned long long)hangcheck_margin + hangcheck_tick;
+	hangcheck_tsc_margin *= TIMER_FREQ;
+
+	hangcheck_tsc = ktime_get_ns();
+	mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ));
+
+	return 0;
+}
+
+
+static void __exit hangcheck_exit(void)
+{
+	timer_delete_sync(&hangcheck_ticktock);
+	pr_debug("Hangcheck: Stopped hangcheck timer.\n");
+}
+
+module_init(hangcheck_init);
+module_exit(hangcheck_exit);