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>

4 files changed, 118 insertions, 0 deletions

diff --git a/tools/testing/selftests/fpu/.gitignore b/tools/testing/selftests/fpu/.gitignore
new file mode 100644
index 000000000..d6d12ac1d
--- /dev/null
+++ b/tools/testing/selftests/fpu/.gitignore
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0+
+test_fpu
diff --git a/tools/testing/selftests/fpu/Makefile b/tools/testing/selftests/fpu/Makefile
new file mode 100644
index 000000000..ea62c176e
--- /dev/null
+++ b/tools/testing/selftests/fpu/Makefile
@@ -0,0 +1,9 @@
+# SPDX-License-Identifier: GPL-2.0+
+
+LDLIBS := -lm
+
+TEST_GEN_PROGS := test_fpu
+
+TEST_PROGS := run_test_fpu.sh
+
+include ../lib.mk
diff --git a/tools/testing/selftests/fpu/run_test_fpu.sh b/tools/testing/selftests/fpu/run_test_fpu.sh
new file mode 100755
index 000000000..d77be93ec
--- /dev/null
+++ b/tools/testing/selftests/fpu/run_test_fpu.sh
@@ -0,0 +1,46 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+#
+# Load kernel module for FPU tests
+
+uid=$(id -u)
+if [ $uid -ne 0 ]; then
+	echo "$0: Must be run as root"
+	exit 1
+fi
+
+if ! which modprobe > /dev/null 2>&1; then
+	echo "$0: You need modprobe installed"
+        exit 4
+fi
+
+if ! modinfo test_fpu > /dev/null 2>&1; then
+	echo "$0: You must have the following enabled in your kernel:"
+	echo "CONFIG_TEST_FPU=m"
+	exit 4
+fi
+
+NR_CPUS=$(getconf _NPROCESSORS_ONLN)
+if [ ! $NR_CPUS ]; then
+	NR_CPUS=1
+fi
+
+modprobe test_fpu
+
+if [ ! -e /sys/kernel/debug/selftest_helpers/test_fpu ]; then
+	mount -t debugfs none /sys/kernel/debug
+
+	if [ ! -e /sys/kernel/debug/selftest_helpers/test_fpu ]; then
+		echo "$0: Error mounting debugfs"
+		exit 4
+	fi
+fi
+
+echo "Running 1000 iterations on all CPUs... "
+for i in $(seq 1 1000); do
+	for c in $(seq 1 $NR_CPUS); do
+		./test_fpu &
+	done
+done
+
+rmmod test_fpu
diff --git a/tools/testing/selftests/fpu/test_fpu.c b/tools/testing/selftests/fpu/test_fpu.c
new file mode 100644
index 000000000..200238522
--- /dev/null
+++ b/tools/testing/selftests/fpu/test_fpu.c
@@ -0,0 +1,61 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* This testcase operates with the test_fpu kernel driver.
+ * It modifies the FPU control register in user mode and calls the kernel
+ * module to perform floating point operations in the kernel. The control
+ * register value should be independent between kernel and user mode.
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <fenv.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+const char *test_fpu_path = "/sys/kernel/debug/selftest_helpers/test_fpu";
+
+int main(void)
+{
+	char dummy[1];
+	int fd = open(test_fpu_path, O_RDONLY);
+
+	if (fd < 0) {
+		printf("[SKIP]\tcan't access %s: %s\n",
+		       test_fpu_path, strerror(errno));
+		return 0;
+	}
+
+	if (read(fd, dummy, 1) < 0) {
+		printf("[FAIL]\taccess with default rounding mode failed\n");
+		return 1;
+	}
+
+	fesetround(FE_DOWNWARD);
+	if (read(fd, dummy, 1) < 0) {
+		printf("[FAIL]\taccess with downward rounding mode failed\n");
+		return 2;
+	}
+	if (fegetround() != FE_DOWNWARD) {
+		printf("[FAIL]\tusermode rounding mode clobbered\n");
+		return 3;
+	}
+
+	/* Note: the tests up to this point are quite safe and will only return
+	 * an error. But the exception mask setting can cause misbehaving kernel
+	 * to crash.
+	 */
+	feclearexcept(FE_ALL_EXCEPT);
+	feenableexcept(FE_ALL_EXCEPT);
+	if (read(fd, dummy, 1) < 0) {
+		printf("[FAIL]\taccess with fpu exceptions unmasked failed\n");
+		return 4;
+	}
+	if (fegetexcept() != FE_ALL_EXCEPT) {
+		printf("[FAIL]\tusermode fpu exception mask clobbered\n");
+		return 5;
+	}
+
+	printf("[OK]\ttest_fpu\n");
+	return 0;
+}