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, 262 insertions, 0 deletions

diff --git a/crypto/tea.c b/crypto/tea.c
new file mode 100644
index 000000000..7c66efcb5
--- /dev/null
+++ b/crypto/tea.c
@@ -0,0 +1,262 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* 
+ * Cryptographic API.
+ *
+ * TEA, XTEA, and XETA crypto algorithms
+ *
+ * The TEA and Xtended TEA algorithms were developed by David Wheeler 
+ * and Roger Needham at the Computer Laboratory of Cambridge University.
+ *
+ * Due to the order of evaluation in XTEA many people have incorrectly
+ * implemented it.  XETA (XTEA in the wrong order), exists for
+ * compatibility with these implementations.
+ *
+ * Copyright (c) 2004 Aaron Grothe ajgrothe@yahoo.com
+ */
+
+#include <crypto/algapi.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/mm.h>
+#include <linux/unaligned.h>
+#include <linux/types.h>
+
+#define TEA_KEY_SIZE		16
+#define TEA_BLOCK_SIZE		8
+#define TEA_ROUNDS		32
+#define TEA_DELTA		0x9e3779b9
+
+#define XTEA_KEY_SIZE		16
+#define XTEA_BLOCK_SIZE		8
+#define XTEA_ROUNDS		32
+#define XTEA_DELTA		0x9e3779b9
+
+struct tea_ctx {
+	u32 KEY[4];
+};
+
+struct xtea_ctx {
+	u32 KEY[4];
+};
+
+static int tea_setkey(struct crypto_tfm *tfm, const u8 *in_key,
+		      unsigned int key_len)
+{
+	struct tea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	ctx->KEY[0] = get_unaligned_le32(&in_key[0]);
+	ctx->KEY[1] = get_unaligned_le32(&in_key[4]);
+	ctx->KEY[2] = get_unaligned_le32(&in_key[8]);
+	ctx->KEY[3] = get_unaligned_le32(&in_key[12]);
+
+	return 0; 
+
+}
+
+static void tea_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	u32 y, z, n, sum = 0;
+	u32 k0, k1, k2, k3;
+	struct tea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	y = get_unaligned_le32(&src[0]);
+	z = get_unaligned_le32(&src[4]);
+
+	k0 = ctx->KEY[0];
+	k1 = ctx->KEY[1];
+	k2 = ctx->KEY[2];
+	k3 = ctx->KEY[3];
+
+	n = TEA_ROUNDS;
+
+	while (n-- > 0) {
+		sum += TEA_DELTA;
+		y += ((z << 4) + k0) ^ (z + sum) ^ ((z >> 5) + k1);
+		z += ((y << 4) + k2) ^ (y + sum) ^ ((y >> 5) + k3);
+	}
+	
+	put_unaligned_le32(y, &dst[0]);
+	put_unaligned_le32(z, &dst[4]);
+}
+
+static void tea_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	u32 y, z, n, sum;
+	u32 k0, k1, k2, k3;
+	struct tea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	y = get_unaligned_le32(&src[0]);
+	z = get_unaligned_le32(&src[4]);
+
+	k0 = ctx->KEY[0];
+	k1 = ctx->KEY[1];
+	k2 = ctx->KEY[2];
+	k3 = ctx->KEY[3];
+
+	sum = TEA_DELTA << 5;
+
+	n = TEA_ROUNDS;
+
+	while (n-- > 0) {
+		z -= ((y << 4) + k2) ^ (y + sum) ^ ((y >> 5) + k3);
+		y -= ((z << 4) + k0) ^ (z + sum) ^ ((z >> 5) + k1);
+		sum -= TEA_DELTA;
+	}
+	
+	put_unaligned_le32(y, &dst[0]);
+	put_unaligned_le32(z, &dst[4]);
+}
+
+static int xtea_setkey(struct crypto_tfm *tfm, const u8 *in_key,
+		       unsigned int key_len)
+{
+	struct xtea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	ctx->KEY[0] = get_unaligned_le32(&in_key[0]);
+	ctx->KEY[1] = get_unaligned_le32(&in_key[4]);
+	ctx->KEY[2] = get_unaligned_le32(&in_key[8]);
+	ctx->KEY[3] = get_unaligned_le32(&in_key[12]);
+
+	return 0; 
+
+}
+
+static void xtea_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	u32 y, z, sum = 0;
+	u32 limit = XTEA_DELTA * XTEA_ROUNDS;
+	struct xtea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	y = get_unaligned_le32(&src[0]);
+	z = get_unaligned_le32(&src[4]);
+
+	while (sum != limit) {
+		y += ((z << 4 ^ z >> 5) + z) ^ (sum + ctx->KEY[sum&3]); 
+		sum += XTEA_DELTA;
+		z += ((y << 4 ^ y >> 5) + y) ^ (sum + ctx->KEY[sum>>11 &3]); 
+	}
+	
+	put_unaligned_le32(y, &dst[0]);
+	put_unaligned_le32(z, &dst[4]);
+}
+
+static void xtea_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	u32 y, z, sum;
+	struct tea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	y = get_unaligned_le32(&src[0]);
+	z = get_unaligned_le32(&src[4]);
+
+	sum = XTEA_DELTA * XTEA_ROUNDS;
+
+	while (sum) {
+		z -= ((y << 4 ^ y >> 5) + y) ^ (sum + ctx->KEY[sum>>11 & 3]);
+		sum -= XTEA_DELTA;
+		y -= ((z << 4 ^ z >> 5) + z) ^ (sum + ctx->KEY[sum & 3]);
+	}
+	
+	put_unaligned_le32(y, &dst[0]);
+	put_unaligned_le32(z, &dst[4]);
+}
+
+
+static void xeta_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	u32 y, z, sum = 0;
+	u32 limit = XTEA_DELTA * XTEA_ROUNDS;
+	struct xtea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	y = get_unaligned_le32(&src[0]);
+	z = get_unaligned_le32(&src[4]);
+
+	while (sum != limit) {
+		y += (z << 4 ^ z >> 5) + (z ^ sum) + ctx->KEY[sum&3];
+		sum += XTEA_DELTA;
+		z += (y << 4 ^ y >> 5) + (y ^ sum) + ctx->KEY[sum>>11 &3];
+	}
+	
+	put_unaligned_le32(y, &dst[0]);
+	put_unaligned_le32(z, &dst[4]);
+}
+
+static void xeta_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	u32 y, z, sum;
+	struct tea_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	y = get_unaligned_le32(&src[0]);
+	z = get_unaligned_le32(&src[4]);
+
+	sum = XTEA_DELTA * XTEA_ROUNDS;
+
+	while (sum) {
+		z -= (y << 4 ^ y >> 5) + (y ^ sum) + ctx->KEY[sum>>11 & 3];
+		sum -= XTEA_DELTA;
+		y -= (z << 4 ^ z >> 5) + (z ^ sum) + ctx->KEY[sum & 3];
+	}
+	
+	put_unaligned_le32(y, &dst[0]);
+	put_unaligned_le32(z, &dst[4]);
+}
+
+static struct crypto_alg tea_algs[3] = { {
+	.cra_name		=	"tea",
+	.cra_driver_name	=	"tea-generic",
+	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize		=	TEA_BLOCK_SIZE,
+	.cra_ctxsize		=	sizeof (struct tea_ctx),
+	.cra_module		=	THIS_MODULE,
+	.cra_u			=	{ .cipher = {
+	.cia_min_keysize	=	TEA_KEY_SIZE,
+	.cia_max_keysize	=	TEA_KEY_SIZE,
+	.cia_setkey		= 	tea_setkey,
+	.cia_encrypt		=	tea_encrypt,
+	.cia_decrypt		=	tea_decrypt } }
+}, {
+	.cra_name		=	"xtea",
+	.cra_driver_name	=	"xtea-generic",
+	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize		=	XTEA_BLOCK_SIZE,
+	.cra_ctxsize		=	sizeof (struct xtea_ctx),
+	.cra_module		=	THIS_MODULE,
+	.cra_u			=	{ .cipher = {
+	.cia_min_keysize	=	XTEA_KEY_SIZE,
+	.cia_max_keysize	=	XTEA_KEY_SIZE,
+	.cia_setkey		= 	xtea_setkey,
+	.cia_encrypt		=	xtea_encrypt,
+	.cia_decrypt		=	xtea_decrypt } }
+}, {
+	.cra_name		=	"xeta",
+	.cra_driver_name	=	"xeta-generic",
+	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize		=	XTEA_BLOCK_SIZE,
+	.cra_ctxsize		=	sizeof (struct xtea_ctx),
+	.cra_module		=	THIS_MODULE,
+	.cra_u			=	{ .cipher = {
+	.cia_min_keysize	=	XTEA_KEY_SIZE,
+	.cia_max_keysize	=	XTEA_KEY_SIZE,
+	.cia_setkey		= 	xtea_setkey,
+	.cia_encrypt		=	xeta_encrypt,
+	.cia_decrypt		=	xeta_decrypt } }
+} };
+
+static int __init tea_mod_init(void)
+{
+	return crypto_register_algs(tea_algs, ARRAY_SIZE(tea_algs));
+}
+
+static void __exit tea_mod_fini(void)
+{
+	crypto_unregister_algs(tea_algs, ARRAY_SIZE(tea_algs));
+}
+
+MODULE_ALIAS_CRYPTO("tea");
+MODULE_ALIAS_CRYPTO("xtea");
+MODULE_ALIAS_CRYPTO("xeta");
+
+module_init(tea_mod_init);
+module_exit(tea_mod_fini);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("TEA, XTEA & XETA Cryptographic Algorithms");