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, 121 insertions, 0 deletions

diff --git a/drivers/misc/sram-exec.c b/drivers/misc/sram-exec.c
new file mode 100644
index 000000000..b71dbbd73
--- /dev/null
+++ b/drivers/misc/sram-exec.c
@@ -0,0 +1,121 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * SRAM protect-exec region helper functions
+ *
+ * Copyright (C) 2017 Texas Instruments Incorporated - https://www.ti.com/
+ *	Dave Gerlach
+ */
+
+#include <linux/device.h>
+#include <linux/genalloc.h>
+#include <linux/mm.h>
+#include <linux/sram.h>
+#include <linux/set_memory.h>
+
+#include <asm/fncpy.h>
+
+#include "sram.h"
+
+static DEFINE_MUTEX(exec_pool_list_mutex);
+static LIST_HEAD(exec_pool_list);
+
+int sram_check_protect_exec(struct sram_dev *sram, struct sram_reserve *block,
+			    struct sram_partition *part)
+{
+	unsigned long base = (unsigned long)part->base;
+	unsigned long end = base + block->size;
+
+	if (!PAGE_ALIGNED(base) || !PAGE_ALIGNED(end)) {
+		dev_err(sram->dev,
+			"SRAM pool marked with 'protect-exec' is not page aligned and will not be created.\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+int sram_add_protect_exec(struct sram_partition *part)
+{
+	mutex_lock(&exec_pool_list_mutex);
+	list_add_tail(&part->list, &exec_pool_list);
+	mutex_unlock(&exec_pool_list_mutex);
+
+	return 0;
+}
+
+/**
+ * sram_exec_copy - copy data to a protected executable region of sram
+ *
+ * @pool: struct gen_pool retrieved that is part of this sram
+ * @dst: Destination address for the copy, that must be inside pool
+ * @src: Source address for the data to copy
+ * @size: Size of copy to perform, which starting from dst, must reside in pool
+ *
+ * Return: Address for copied data that can safely be called through function
+ *	   pointer, or NULL if problem.
+ *
+ * This helper function allows sram driver to act as central control location
+ * of 'protect-exec' pools which are normal sram pools but are always set
+ * read-only and executable except when copying data to them, at which point
+ * they are set to read-write non-executable, to make sure no memory is
+ * writeable and executable at the same time. This region must be page-aligned
+ * and is checked during probe, otherwise page attribute manipulation would
+ * not be possible. Care must be taken to only call the returned address as
+ * dst address is not guaranteed to be safely callable.
+ *
+ * NOTE: This function uses the fncpy macro to move code to the executable
+ * region. Some architectures have strict requirements for relocating
+ * executable code, so fncpy is a macro that must be defined by any arch
+ * making use of this functionality that guarantees a safe copy of exec
+ * data and returns a safe address that can be called as a C function
+ * pointer.
+ */
+void *sram_exec_copy(struct gen_pool *pool, void *dst, void *src,
+		     size_t size)
+{
+	struct sram_partition *part = NULL, *p;
+	unsigned long base;
+	int pages;
+	void *dst_cpy;
+	int ret;
+
+	mutex_lock(&exec_pool_list_mutex);
+	list_for_each_entry(p, &exec_pool_list, list) {
+		if (p->pool == pool)
+			part = p;
+	}
+	mutex_unlock(&exec_pool_list_mutex);
+
+	if (!part)
+		return NULL;
+
+	if (!gen_pool_has_addr(pool, (unsigned long)dst, size))
+		return NULL;
+
+	base = (unsigned long)part->base;
+	pages = PAGE_ALIGN(size) / PAGE_SIZE;
+
+	mutex_lock(&part->lock);
+
+	ret = set_memory_nx((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+	ret = set_memory_rw((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+
+	dst_cpy = fncpy(dst, src, size);
+
+	ret = set_memory_rox((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+
+	mutex_unlock(&part->lock);
+
+	return dst_cpy;
+
+error_out:
+	mutex_unlock(&part->lock);
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(sram_exec_copy);