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, 236 insertions, 0 deletions

diff --git a/crypto/md5.c b/crypto/md5.c
new file mode 100644
index 000000000..c167d203c
--- /dev/null
+++ b/crypto/md5.c
@@ -0,0 +1,236 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Crypto API support for MD5 and HMAC-MD5
+ *
+ * Copyright 2025 Google LLC
+ */
+#include <crypto/internal/hash.h>
+#include <crypto/md5.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+/*
+ * Export and import functions.  crypto_shash wants a particular format that
+ * matches that used by some legacy drivers.  It currently is the same as the
+ * library MD5 context, except the value in bytecount must be block-aligned and
+ * the remainder must be stored in an extra u8 appended to the struct.
+ */
+
+#define MD5_SHASH_STATE_SIZE (sizeof(struct md5_ctx) + 1)
+static_assert(sizeof(struct md5_ctx) == sizeof(struct md5_state));
+static_assert(offsetof(struct md5_ctx, state) == offsetof(struct md5_state, hash));
+static_assert(offsetof(struct md5_ctx, bytecount) == offsetof(struct md5_state, byte_count));
+static_assert(offsetof(struct md5_ctx, buf) == offsetof(struct md5_state, block));
+
+static int __crypto_md5_export(const struct md5_ctx *ctx0, void *out)
+{
+	struct md5_ctx ctx = *ctx0;
+	unsigned int partial;
+	u8 *p = out;
+
+	partial = ctx.bytecount % MD5_BLOCK_SIZE;
+	ctx.bytecount -= partial;
+	memcpy(p, &ctx, sizeof(ctx));
+	p += sizeof(ctx);
+	*p = partial;
+	return 0;
+}
+
+static int __crypto_md5_import(struct md5_ctx *ctx, const void *in)
+{
+	const u8 *p = in;
+
+	memcpy(ctx, p, sizeof(*ctx));
+	p += sizeof(*ctx);
+	ctx->bytecount += *p;
+	return 0;
+}
+
+static int __crypto_md5_export_core(const struct md5_ctx *ctx, void *out)
+{
+	memcpy(out, ctx, offsetof(struct md5_ctx, buf));
+	return 0;
+}
+
+static int __crypto_md5_import_core(struct md5_ctx *ctx, const void *in)
+{
+	memcpy(ctx, in, offsetof(struct md5_ctx, buf));
+	return 0;
+}
+
+const u8 md5_zero_message_hash[MD5_DIGEST_SIZE] = {
+	0xd4, 0x1d, 0x8c, 0xd9, 0x8f, 0x00, 0xb2, 0x04,
+	0xe9, 0x80, 0x09, 0x98, 0xec, 0xf8, 0x42, 0x7e,
+};
+EXPORT_SYMBOL_GPL(md5_zero_message_hash);
+
+#define MD5_CTX(desc) ((struct md5_ctx *)shash_desc_ctx(desc))
+
+static int crypto_md5_init(struct shash_desc *desc)
+{
+	md5_init(MD5_CTX(desc));
+	return 0;
+}
+
+static int crypto_md5_update(struct shash_desc *desc,
+			     const u8 *data, unsigned int len)
+{
+	md5_update(MD5_CTX(desc), data, len);
+	return 0;
+}
+
+static int crypto_md5_final(struct shash_desc *desc, u8 *out)
+{
+	md5_final(MD5_CTX(desc), out);
+	return 0;
+}
+
+static int crypto_md5_digest(struct shash_desc *desc,
+			     const u8 *data, unsigned int len, u8 *out)
+{
+	md5(data, len, out);
+	return 0;
+}
+
+static int crypto_md5_export(struct shash_desc *desc, void *out)
+{
+	return __crypto_md5_export(MD5_CTX(desc), out);
+}
+
+static int crypto_md5_import(struct shash_desc *desc, const void *in)
+{
+	return __crypto_md5_import(MD5_CTX(desc), in);
+}
+
+static int crypto_md5_export_core(struct shash_desc *desc, void *out)
+{
+	return __crypto_md5_export_core(MD5_CTX(desc), out);
+}
+
+static int crypto_md5_import_core(struct shash_desc *desc, const void *in)
+{
+	return __crypto_md5_import_core(MD5_CTX(desc), in);
+}
+
+#define HMAC_MD5_KEY(tfm) ((struct hmac_md5_key *)crypto_shash_ctx(tfm))
+#define HMAC_MD5_CTX(desc) ((struct hmac_md5_ctx *)shash_desc_ctx(desc))
+
+static int crypto_hmac_md5_setkey(struct crypto_shash *tfm,
+				  const u8 *raw_key, unsigned int keylen)
+{
+	hmac_md5_preparekey(HMAC_MD5_KEY(tfm), raw_key, keylen);
+	return 0;
+}
+
+static int crypto_hmac_md5_init(struct shash_desc *desc)
+{
+	hmac_md5_init(HMAC_MD5_CTX(desc), HMAC_MD5_KEY(desc->tfm));
+	return 0;
+}
+
+static int crypto_hmac_md5_update(struct shash_desc *desc,
+				  const u8 *data, unsigned int len)
+{
+	hmac_md5_update(HMAC_MD5_CTX(desc), data, len);
+	return 0;
+}
+
+static int crypto_hmac_md5_final(struct shash_desc *desc, u8 *out)
+{
+	hmac_md5_final(HMAC_MD5_CTX(desc), out);
+	return 0;
+}
+
+static int crypto_hmac_md5_digest(struct shash_desc *desc,
+				  const u8 *data, unsigned int len, u8 *out)
+{
+	hmac_md5(HMAC_MD5_KEY(desc->tfm), data, len, out);
+	return 0;
+}
+
+static int crypto_hmac_md5_export(struct shash_desc *desc, void *out)
+{
+	return __crypto_md5_export(&HMAC_MD5_CTX(desc)->hash_ctx, out);
+}
+
+static int crypto_hmac_md5_import(struct shash_desc *desc, const void *in)
+{
+	struct hmac_md5_ctx *ctx = HMAC_MD5_CTX(desc);
+
+	ctx->ostate = HMAC_MD5_KEY(desc->tfm)->ostate;
+	return __crypto_md5_import(&ctx->hash_ctx, in);
+}
+
+static int crypto_hmac_md5_export_core(struct shash_desc *desc, void *out)
+{
+	return __crypto_md5_export_core(&HMAC_MD5_CTX(desc)->hash_ctx, out);
+}
+
+static int crypto_hmac_md5_import_core(struct shash_desc *desc, const void *in)
+{
+	struct hmac_md5_ctx *ctx = HMAC_MD5_CTX(desc);
+
+	ctx->ostate = HMAC_MD5_KEY(desc->tfm)->ostate;
+	return __crypto_md5_import_core(&ctx->hash_ctx, in);
+}
+
+static struct shash_alg algs[] = {
+	{
+		.base.cra_name		= "md5",
+		.base.cra_driver_name	= "md5-lib",
+		.base.cra_priority	= 300,
+		.base.cra_blocksize	= MD5_BLOCK_SIZE,
+		.base.cra_module	= THIS_MODULE,
+		.digestsize		= MD5_DIGEST_SIZE,
+		.init			= crypto_md5_init,
+		.update			= crypto_md5_update,
+		.final			= crypto_md5_final,
+		.digest			= crypto_md5_digest,
+		.export			= crypto_md5_export,
+		.import			= crypto_md5_import,
+		.export_core		= crypto_md5_export_core,
+		.import_core		= crypto_md5_import_core,
+		.descsize		= sizeof(struct md5_ctx),
+		.statesize		= MD5_SHASH_STATE_SIZE,
+	},
+	{
+		.base.cra_name		= "hmac(md5)",
+		.base.cra_driver_name	= "hmac-md5-lib",
+		.base.cra_priority	= 300,
+		.base.cra_blocksize	= MD5_BLOCK_SIZE,
+		.base.cra_ctxsize	= sizeof(struct hmac_md5_key),
+		.base.cra_module	= THIS_MODULE,
+		.digestsize		= MD5_DIGEST_SIZE,
+		.setkey			= crypto_hmac_md5_setkey,
+		.init			= crypto_hmac_md5_init,
+		.update			= crypto_hmac_md5_update,
+		.final			= crypto_hmac_md5_final,
+		.digest			= crypto_hmac_md5_digest,
+		.export			= crypto_hmac_md5_export,
+		.import			= crypto_hmac_md5_import,
+		.export_core		= crypto_hmac_md5_export_core,
+		.import_core		= crypto_hmac_md5_import_core,
+		.descsize		= sizeof(struct hmac_md5_ctx),
+		.statesize		= MD5_SHASH_STATE_SIZE,
+	},
+};
+
+static int __init crypto_md5_mod_init(void)
+{
+	return crypto_register_shashes(algs, ARRAY_SIZE(algs));
+}
+module_init(crypto_md5_mod_init);
+
+static void __exit crypto_md5_mod_exit(void)
+{
+	crypto_unregister_shashes(algs, ARRAY_SIZE(algs));
+}
+module_exit(crypto_md5_mod_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("Crypto API support for MD5 and HMAC-MD5");
+
+MODULE_ALIAS_CRYPTO("md5");
+MODULE_ALIAS_CRYPTO("md5-lib");
+MODULE_ALIAS_CRYPTO("hmac(md5)");
+MODULE_ALIAS_CRYPTO("hmac-md5-lib");