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, 328 insertions, 0 deletions

diff --git a/drivers/firmware/google/vpd.c b/drivers/firmware/google/vpd.c
new file mode 100644
index 000000000..fbb5a8d7c
--- /dev/null
+++ b/drivers/firmware/google/vpd.c
@@ -0,0 +1,328 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * vpd.c
+ *
+ * Driver for exporting VPD content to sysfs.
+ *
+ * Copyright 2017 Google Inc.
+ */
+
+#include <linux/ctype.h>
+#include <linux/init.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/kobject.h>
+#include <linux/list.h>
+#include <linux/device-id/coreboot.h>
+#include <linux/module.h>
+#include <linux/of_address.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/sysfs.h>
+
+#include "coreboot_table.h"
+#include "vpd_decode.h"
+
+#define CB_TAG_VPD      0x2c
+#define VPD_CBMEM_MAGIC 0x43524f53
+
+static struct kobject *vpd_kobj;
+
+struct vpd_cbmem {
+	u32 magic;
+	u32 version;
+	u32 ro_size;
+	u32 rw_size;
+	u8  blob[];
+};
+
+struct vpd_section {
+	bool enabled;
+	const char *name;
+	char *raw_name;                /* the string name_raw */
+	struct kobject *kobj;          /* vpd/name directory */
+	char *baseaddr;
+	struct bin_attribute bin_attr; /* vpd/name_raw bin_attribute */
+	struct list_head attribs;      /* key/value in vpd_attrib_info list */
+};
+
+struct vpd_attrib_info {
+	char *key;
+	const char *value;
+	struct bin_attribute bin_attr;
+	struct list_head list;
+};
+
+static struct vpd_section ro_vpd;
+static struct vpd_section rw_vpd;
+
+static ssize_t vpd_attrib_read(struct file *filp, struct kobject *kobp,
+			       const struct bin_attribute *bin_attr, char *buf,
+			       loff_t pos, size_t count)
+{
+	struct vpd_attrib_info *info = bin_attr->private;
+
+	return memory_read_from_buffer(buf, count, &pos, info->value,
+				       info->bin_attr.size);
+}
+
+/*
+ * vpd_section_check_key_name()
+ *
+ * The VPD specification supports only [a-zA-Z0-9_]+ characters in key names but
+ * old firmware versions may have entries like "S/N" which are problematic when
+ * exporting them as sysfs attributes. These keys present in old firmwares are
+ * ignored.
+ *
+ * Returns VPD_OK for a valid key name, VPD_FAIL otherwise.
+ *
+ * @key: The key name to check
+ * @key_len: key name length
+ */
+static int vpd_section_check_key_name(const u8 *key, s32 key_len)
+{
+	int c;
+
+	while (key_len-- > 0) {
+		c = *key++;
+
+		if (!isalnum(c) && c != '_')
+			return VPD_FAIL;
+	}
+
+	return VPD_OK;
+}
+
+static int vpd_section_attrib_add(const u8 *key, u32 key_len,
+				  const u8 *value, u32 value_len,
+				  void *arg)
+{
+	int ret;
+	struct vpd_section *sec = arg;
+	struct vpd_attrib_info *info;
+
+	/*
+	 * Return VPD_OK immediately to decode next entry if the current key
+	 * name contains invalid characters.
+	 */
+	if (vpd_section_check_key_name(key, key_len) != VPD_OK)
+		return VPD_OK;
+
+	info = kzalloc_obj(*info);
+	if (!info)
+		return -ENOMEM;
+
+	info->key = kstrndup(key, key_len, GFP_KERNEL);
+	if (!info->key) {
+		ret = -ENOMEM;
+		goto free_info;
+	}
+
+	sysfs_bin_attr_init(&info->bin_attr);
+	info->bin_attr.attr.name = info->key;
+	info->bin_attr.attr.mode = 0444;
+	info->bin_attr.size = value_len;
+	info->bin_attr.read = vpd_attrib_read;
+	info->bin_attr.private = info;
+
+	info->value = value;
+
+	INIT_LIST_HEAD(&info->list);
+
+	ret = sysfs_create_bin_file(sec->kobj, &info->bin_attr);
+	if (ret)
+		goto free_info_key;
+
+	list_add_tail(&info->list, &sec->attribs);
+	return 0;
+
+free_info_key:
+	kfree(info->key);
+free_info:
+	kfree(info);
+
+	return ret;
+}
+
+static void vpd_section_attrib_destroy(struct vpd_section *sec)
+{
+	struct vpd_attrib_info *info;
+	struct vpd_attrib_info *temp;
+
+	list_for_each_entry_safe(info, temp, &sec->attribs, list) {
+		sysfs_remove_bin_file(sec->kobj, &info->bin_attr);
+		kfree(info->key);
+		kfree(info);
+	}
+}
+
+static ssize_t vpd_section_read(struct file *filp, struct kobject *kobp,
+				const struct bin_attribute *bin_attr, char *buf,
+				loff_t pos, size_t count)
+{
+	struct vpd_section *sec = bin_attr->private;
+
+	return memory_read_from_buffer(buf, count, &pos, sec->baseaddr,
+				       sec->bin_attr.size);
+}
+
+static int vpd_section_create_attribs(struct vpd_section *sec)
+{
+	s32 consumed;
+	int ret;
+
+	consumed = 0;
+	do {
+		ret = vpd_decode_string(sec->bin_attr.size, sec->baseaddr,
+					&consumed, vpd_section_attrib_add, sec);
+	} while (ret == VPD_OK);
+
+	return 0;
+}
+
+static int vpd_section_init(const char *name, struct vpd_section *sec,
+			    phys_addr_t physaddr, size_t size)
+{
+	int err;
+
+	sec->baseaddr = memremap(physaddr, size, MEMREMAP_WB);
+	if (!sec->baseaddr)
+		return -ENOMEM;
+
+	sec->name = name;
+
+	/* We want to export the raw partition with name ${name}_raw */
+	sec->raw_name = kasprintf(GFP_KERNEL, "%s_raw", name);
+	if (!sec->raw_name) {
+		err = -ENOMEM;
+		goto err_memunmap;
+	}
+
+	sysfs_bin_attr_init(&sec->bin_attr);
+	sec->bin_attr.attr.name = sec->raw_name;
+	sec->bin_attr.attr.mode = 0444;
+	sec->bin_attr.size = size;
+	sec->bin_attr.read = vpd_section_read;
+	sec->bin_attr.private = sec;
+
+	err = sysfs_create_bin_file(vpd_kobj, &sec->bin_attr);
+	if (err)
+		goto err_free_raw_name;
+
+	sec->kobj = kobject_create_and_add(name, vpd_kobj);
+	if (!sec->kobj) {
+		err = -EINVAL;
+		goto err_sysfs_remove;
+	}
+
+	INIT_LIST_HEAD(&sec->attribs);
+	vpd_section_create_attribs(sec);
+
+	sec->enabled = true;
+
+	return 0;
+
+err_sysfs_remove:
+	sysfs_remove_bin_file(vpd_kobj, &sec->bin_attr);
+err_free_raw_name:
+	kfree(sec->raw_name);
+err_memunmap:
+	memunmap(sec->baseaddr);
+	return err;
+}
+
+static int vpd_section_destroy(struct vpd_section *sec)
+{
+	if (sec->enabled) {
+		vpd_section_attrib_destroy(sec);
+		kobject_put(sec->kobj);
+		sysfs_remove_bin_file(vpd_kobj, &sec->bin_attr);
+		kfree(sec->raw_name);
+		memunmap(sec->baseaddr);
+		sec->enabled = false;
+	}
+
+	return 0;
+}
+
+static int vpd_sections_init(phys_addr_t physaddr)
+{
+	struct vpd_cbmem *temp;
+	struct vpd_cbmem header;
+	int ret = 0;
+
+	temp = memremap(physaddr, sizeof(struct vpd_cbmem), MEMREMAP_WB);
+	if (!temp)
+		return -ENOMEM;
+
+	memcpy(&header, temp, sizeof(struct vpd_cbmem));
+	memunmap(temp);
+
+	if (header.magic != VPD_CBMEM_MAGIC)
+		return -ENODEV;
+
+	if (header.ro_size) {
+		ret = vpd_section_init("ro", &ro_vpd,
+				       physaddr + sizeof(struct vpd_cbmem),
+				       header.ro_size);
+		if (ret)
+			return ret;
+	}
+
+	if (header.rw_size) {
+		ret = vpd_section_init("rw", &rw_vpd,
+				       physaddr + sizeof(struct vpd_cbmem) +
+				       header.ro_size, header.rw_size);
+		if (ret) {
+			vpd_section_destroy(&ro_vpd);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static int vpd_probe(struct coreboot_device *dev)
+{
+	int ret;
+
+	vpd_kobj = kobject_create_and_add("vpd", firmware_kobj);
+	if (!vpd_kobj)
+		return -ENOMEM;
+
+	ret = vpd_sections_init(dev->cbmem_ref.cbmem_addr);
+	if (ret) {
+		kobject_put(vpd_kobj);
+		return ret;
+	}
+
+	return 0;
+}
+
+static void vpd_remove(struct coreboot_device *dev)
+{
+	vpd_section_destroy(&ro_vpd);
+	vpd_section_destroy(&rw_vpd);
+
+	kobject_put(vpd_kobj);
+}
+
+static const struct coreboot_device_id vpd_ids[] = {
+	{ .tag = CB_TAG_VPD },
+	{ /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(coreboot, vpd_ids);
+
+static struct coreboot_driver vpd_driver = {
+	.probe = vpd_probe,
+	.remove = vpd_remove,
+	.drv = {
+		.name = "vpd",
+	},
+	.id_table = vpd_ids,
+};
+module_coreboot_driver(vpd_driver);
+
+MODULE_AUTHOR("Google, Inc.");
+MODULE_DESCRIPTION("Driver for exporting Vital Product Data content to sysfs");
+MODULE_LICENSE("GPL");