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, 88 insertions, 0 deletions

diff --git a/lib/syscall.c b/lib/syscall.c
new file mode 100644
index 000000000..006e256d2
--- /dev/null
+++ b/lib/syscall.c
@@ -0,0 +1,88 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/ptrace.h>
+#include <linux/sched.h>
+#include <linux/sched/task_stack.h>
+#include <linux/export.h>
+#include <asm/syscall.h>
+
+static int collect_syscall(struct task_struct *target, struct syscall_info *info)
+{
+	unsigned long args[6] = { };
+	struct pt_regs *regs;
+
+	if (!try_get_task_stack(target)) {
+		/* Task has no stack, so the task isn't in a syscall. */
+		memset(info, 0, sizeof(*info));
+		info->data.nr = -1;
+		return 0;
+	}
+
+	regs = task_pt_regs(target);
+	if (unlikely(!regs)) {
+		put_task_stack(target);
+		return -EAGAIN;
+	}
+
+	info->sp = user_stack_pointer(regs);
+	info->data.instruction_pointer = instruction_pointer(regs);
+
+	info->data.nr = syscall_get_nr(target, regs);
+	if (info->data.nr != -1L)
+		syscall_get_arguments(target, regs, args);
+
+	info->data.args[0] = args[0];
+	info->data.args[1] = args[1];
+	info->data.args[2] = args[2];
+	info->data.args[3] = args[3];
+	info->data.args[4] = args[4];
+	info->data.args[5] = args[5];
+
+	put_task_stack(target);
+	return 0;
+}
+
+/**
+ * task_current_syscall - Discover what a blocked task is doing.
+ * @target:		thread to examine
+ * @info:		structure with the following fields:
+ *			 .sp        - filled with user stack pointer
+ *			 .data.nr   - filled with system call number or -1
+ *			 .data.args - filled with @maxargs system call arguments
+ *			 .data.instruction_pointer - filled with user PC
+ *
+ * If @target is blocked in a system call, returns zero with @info.data.nr
+ * set to the call's number and @info.data.args filled in with its
+ * arguments. Registers not used for system call arguments may not be available
+ * and it is not kosher to use &struct user_regset calls while the system
+ * call is still in progress.  Note we may get this result if @target
+ * has finished its system call but not yet returned to user mode, such
+ * as when it's stopped for signal handling or syscall exit tracing.
+ *
+ * If @target is blocked in the kernel during a fault or exception,
+ * returns zero with *@info.data.nr set to -1 and does not fill in
+ * @info.data.args. If so, it's now safe to examine @target using
+ * &struct user_regset get() calls as long as we're sure @target won't return
+ * to user mode.
+ *
+ * Returns -%EAGAIN if @target does not remain blocked.
+ */
+int task_current_syscall(struct task_struct *target, struct syscall_info *info)
+{
+	unsigned long ncsw;
+	unsigned int state;
+
+	if (target == current)
+		return collect_syscall(target, info);
+
+	state = READ_ONCE(target->__state);
+	if (unlikely(!state))
+		return -EAGAIN;
+
+	ncsw = wait_task_inactive(target, state);
+	if (unlikely(!ncsw) ||
+	    unlikely(collect_syscall(target, info)) ||
+	    unlikely(wait_task_inactive(target, state) != ncsw))
+		return -EAGAIN;
+
+	return 0;
+}