sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted

alb_break_llc() decides whether to break LLC preference during active
load balance. It does so by testing that every runnable fair task on the
source rq prefers its LLC:

	env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable

But the two counters cover different sets. nr_pref_llc_running is updated
in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued,
so it follows queued tasks. h_nr_runnable is updated in set_delayed()/
clear_delayed() and drops delay-dequeued tasks.

So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted
in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks,
alb_break_llc() returns false, and active balance is free to pull a task
off its preferred LLC. Active balance only moves runnable tasks, and this
is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB
skips the per-task test in can_migrate_task(). The runnable set is the one
we want.

Fix it on the counter side. A task should be counted in
nr_pref_llc_running exactly while it is both queued on its preferred LLC
(pref_llc_queued) and runnable (!sched_delayed). Define that membership
once in task_pref_llc_runnable(), and adjust the counter only through
pref_llc_running_inc()/pref_llc_running_dec() from the four sites that
change either input: account_llc_enqueue(), account_llc_dequeue(),
set_delayed() and clear_delayed(). Gating every update on the same
predicate keeps the delay, wake and dequeue paths from double-counting
or underflowing; see the comments at those sites for the ordering.

nr_llc_running and sd->llc_counts are not touched and stay on queued
semantics.

Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC")
Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/
Reported-by: Zhan Xusheng <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com

1 files changed, 186 insertions, 0 deletions

diff --git a/drivers/misc/ocxl/sysfs.c b/drivers/misc/ocxl/sysfs.c
new file mode 100644
index 000000000..1b6a86f17
--- /dev/null
+++ b/drivers/misc/ocxl/sysfs.c
@@ -0,0 +1,186 @@
+// SPDX-License-Identifier: GPL-2.0+
+// Copyright 2017 IBM Corp.
+#include <linux/sysfs.h>
+#include "ocxl_internal.h"
+
+static inline struct ocxl_afu *to_afu(struct device *device)
+{
+	struct ocxl_file_info *info = container_of(device, struct ocxl_file_info, dev);
+
+	return info->afu;
+}
+
+static ssize_t global_mmio_size_show(struct device *device,
+				struct device_attribute *attr,
+				char *buf)
+{
+	struct ocxl_afu *afu = to_afu(device);
+
+	return sysfs_emit(buf, "%d\n",
+			afu->config.global_mmio_size);
+}
+
+static ssize_t pp_mmio_size_show(struct device *device,
+				struct device_attribute *attr,
+				char *buf)
+{
+	struct ocxl_afu *afu = to_afu(device);
+
+	return sysfs_emit(buf, "%d\n",
+			afu->config.pp_mmio_stride);
+}
+
+static ssize_t afu_version_show(struct device *device,
+				struct device_attribute *attr,
+				char *buf)
+{
+	struct ocxl_afu *afu = to_afu(device);
+
+	return sysfs_emit(buf, "%hhu:%hhu\n",
+			afu->config.version_major,
+			afu->config.version_minor);
+}
+
+static ssize_t contexts_show(struct device *device,
+		struct device_attribute *attr,
+		char *buf)
+{
+	struct ocxl_afu *afu = to_afu(device);
+
+	return sysfs_emit(buf, "%d/%d\n",
+			afu->pasid_count, afu->pasid_max);
+}
+
+static ssize_t reload_on_reset_show(struct device *device,
+				    struct device_attribute *attr,
+				    char *buf)
+{
+	struct ocxl_afu *afu = to_afu(device);
+	struct ocxl_fn *fn = afu->fn;
+	struct pci_dev *pci_dev = to_pci_dev(fn->dev.parent);
+	int val;
+
+	if (ocxl_config_get_reset_reload(pci_dev, &val))
+		return sysfs_emit(buf, "unavailable\n");
+
+	return sysfs_emit(buf, "%d\n", val);
+}
+
+static ssize_t reload_on_reset_store(struct device *device,
+				     struct device_attribute *attr,
+				     const char *buf, size_t count)
+{
+	struct ocxl_afu *afu = to_afu(device);
+	struct ocxl_fn *fn = afu->fn;
+	struct pci_dev *pci_dev = to_pci_dev(fn->dev.parent);
+	int rc, val;
+
+	rc = kstrtoint(buf, 0, &val);
+	if (rc || (val != 0 && val != 1))
+		return -EINVAL;
+
+	if (ocxl_config_set_reset_reload(pci_dev, val))
+		return -ENODEV;
+
+	return count;
+}
+
+static struct device_attribute afu_attrs[] = {
+	__ATTR_RO(global_mmio_size),
+	__ATTR_RO(pp_mmio_size),
+	__ATTR_RO(afu_version),
+	__ATTR_RO(contexts),
+	__ATTR_RW(reload_on_reset),
+};
+
+static ssize_t global_mmio_read(struct file *filp, struct kobject *kobj,
+				const struct bin_attribute *bin_attr, char *buf,
+				loff_t off, size_t count)
+{
+	struct ocxl_afu *afu = to_afu(kobj_to_dev(kobj));
+
+	if (count == 0 || off < 0 ||
+		off >= afu->config.global_mmio_size)
+		return 0;
+	memcpy_fromio(buf, afu->global_mmio_ptr + off, count);
+	return count;
+}
+
+static vm_fault_t global_mmio_fault(struct vm_fault *vmf)
+{
+	struct vm_area_struct *vma = vmf->vma;
+	struct ocxl_afu *afu = vma->vm_private_data;
+	unsigned long offset;
+
+	if (vmf->pgoff >= (afu->config.global_mmio_size >> PAGE_SHIFT))
+		return VM_FAULT_SIGBUS;
+
+	offset = vmf->pgoff;
+	offset += (afu->global_mmio_start >> PAGE_SHIFT);
+	return vmf_insert_pfn(vma, vmf->address, offset);
+}
+
+static const struct vm_operations_struct global_mmio_vmops = {
+	.fault = global_mmio_fault,
+};
+
+static int global_mmio_mmap(struct file *filp, struct kobject *kobj,
+			const struct bin_attribute *bin_attr,
+			struct vm_area_struct *vma)
+{
+	struct ocxl_afu *afu = to_afu(kobj_to_dev(kobj));
+
+	if ((vma_pages(vma) + vma->vm_pgoff) >
+		(afu->config.global_mmio_size >> PAGE_SHIFT))
+		return -EINVAL;
+
+	vm_flags_set(vma, VM_IO | VM_PFNMAP);
+	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+	vma->vm_ops = &global_mmio_vmops;
+	vma->vm_private_data = afu;
+	return 0;
+}
+
+int ocxl_sysfs_register_afu(struct ocxl_file_info *info)
+{
+	int i, rc;
+
+	for (i = 0; i < ARRAY_SIZE(afu_attrs); i++) {
+		rc = device_create_file(&info->dev, &afu_attrs[i]);
+		if (rc)
+			goto err;
+	}
+
+	sysfs_attr_init(&info->attr_global_mmio.attr);
+	info->attr_global_mmio.attr.name = "global_mmio_area";
+	info->attr_global_mmio.attr.mode = 0600;
+	info->attr_global_mmio.size = info->afu->config.global_mmio_size;
+	info->attr_global_mmio.read = global_mmio_read;
+	info->attr_global_mmio.mmap = global_mmio_mmap;
+	rc = device_create_bin_file(&info->dev, &info->attr_global_mmio);
+	if (rc) {
+		dev_err(&info->dev, "Unable to create global mmio attr for afu: %d\n", rc);
+		goto err;
+	}
+
+	return 0;
+
+err:
+	for (i--; i >= 0; i--)
+		device_remove_file(&info->dev, &afu_attrs[i]);
+
+	return rc;
+}
+
+void ocxl_sysfs_unregister_afu(struct ocxl_file_info *info)
+{
+	int i;
+
+	/*
+	 * device_remove_bin_file is safe to call if the file is not added as
+	 * the files are removed by name, and early exit if not found
+	 */
+	for (i = 0; i < ARRAY_SIZE(afu_attrs); i++)
+		device_remove_file(&info->dev, &afu_attrs[i]);
+	device_remove_bin_file(&info->dev, &info->attr_global_mmio);
+}