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, 160 insertions, 0 deletions

diff --git a/include/crypto/blake2s.h b/include/crypto/blake2s.h
new file mode 100644
index 000000000..648cb7824
--- /dev/null
+++ b/include/crypto/blake2s.h
@@ -0,0 +1,160 @@
+/* SPDX-License-Identifier: GPL-2.0 OR MIT */
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#ifndef _CRYPTO_BLAKE2S_H
+#define _CRYPTO_BLAKE2S_H
+
+#include <linux/bug.h>
+#include <linux/kconfig.h>
+#include <linux/types.h>
+#include <linux/string.h>
+
+enum blake2s_lengths {
+	BLAKE2S_BLOCK_SIZE = 64,
+	BLAKE2S_HASH_SIZE = 32,
+	BLAKE2S_KEY_SIZE = 32,
+
+	BLAKE2S_128_HASH_SIZE = 16,
+	BLAKE2S_160_HASH_SIZE = 20,
+	BLAKE2S_224_HASH_SIZE = 28,
+	BLAKE2S_256_HASH_SIZE = 32,
+};
+
+/**
+ * struct blake2s_ctx - Context for hashing a message with BLAKE2s
+ * @h: compression function state
+ * @t: block counter
+ * @f: finalization indicator
+ * @buf: partial block buffer; 'buflen' bytes are valid
+ * @buflen: number of bytes buffered in @buf
+ * @outlen: length of output hash value in bytes, at most BLAKE2S_HASH_SIZE
+ */
+struct blake2s_ctx {
+	/* 'h', 't', and 'f' are used in assembly code, so keep them as-is. */
+	u32 h[8];
+	u32 t[2];
+	u32 f[2];
+	u8 buf[BLAKE2S_BLOCK_SIZE];
+	unsigned int buflen;
+	unsigned int outlen;
+};
+
+enum blake2s_iv {
+	BLAKE2S_IV0 = 0x6A09E667UL,
+	BLAKE2S_IV1 = 0xBB67AE85UL,
+	BLAKE2S_IV2 = 0x3C6EF372UL,
+	BLAKE2S_IV3 = 0xA54FF53AUL,
+	BLAKE2S_IV4 = 0x510E527FUL,
+	BLAKE2S_IV5 = 0x9B05688CUL,
+	BLAKE2S_IV6 = 0x1F83D9ABUL,
+	BLAKE2S_IV7 = 0x5BE0CD19UL,
+};
+
+static inline void __blake2s_init(struct blake2s_ctx *ctx, size_t outlen,
+				  const void *key, size_t keylen)
+{
+	ctx->h[0] = BLAKE2S_IV0 ^ (0x01010000 | keylen << 8 | outlen);
+	ctx->h[1] = BLAKE2S_IV1;
+	ctx->h[2] = BLAKE2S_IV2;
+	ctx->h[3] = BLAKE2S_IV3;
+	ctx->h[4] = BLAKE2S_IV4;
+	ctx->h[5] = BLAKE2S_IV5;
+	ctx->h[6] = BLAKE2S_IV6;
+	ctx->h[7] = BLAKE2S_IV7;
+	ctx->t[0] = 0;
+	ctx->t[1] = 0;
+	ctx->f[0] = 0;
+	ctx->f[1] = 0;
+	ctx->buflen = 0;
+	ctx->outlen = outlen;
+	if (keylen) {
+		memcpy(ctx->buf, key, keylen);
+		memset(&ctx->buf[keylen], 0, BLAKE2S_BLOCK_SIZE - keylen);
+		ctx->buflen = BLAKE2S_BLOCK_SIZE;
+	}
+}
+
+/**
+ * blake2s_init() - Initialize a BLAKE2s context for a new message (unkeyed)
+ * @ctx: the context to initialize
+ * @outlen: length of output hash value in bytes, at most BLAKE2S_HASH_SIZE
+ *
+ * Context: Any context.
+ */
+static inline void blake2s_init(struct blake2s_ctx *ctx, size_t outlen)
+{
+	__blake2s_init(ctx, outlen, NULL, 0);
+}
+
+/**
+ * blake2s_init_key() - Initialize a BLAKE2s context for a new message (keyed)
+ * @ctx: the context to initialize
+ * @outlen: length of output hash value in bytes, at most BLAKE2S_HASH_SIZE
+ * @key: the key
+ * @keylen: the key length in bytes, at most BLAKE2S_KEY_SIZE
+ *
+ * Context: Any context.
+ */
+static inline void blake2s_init_key(struct blake2s_ctx *ctx, size_t outlen,
+				    const void *key, size_t keylen)
+{
+	WARN_ON(IS_ENABLED(DEBUG) && (!outlen || outlen > BLAKE2S_HASH_SIZE ||
+		!key || !keylen || keylen > BLAKE2S_KEY_SIZE));
+
+	__blake2s_init(ctx, outlen, key, keylen);
+}
+
+/**
+ * blake2s_update() - Update a BLAKE2s context with message data
+ * @ctx: the context to update; must have been initialized
+ * @in: the message data
+ * @inlen: the data length in bytes
+ *
+ * This can be called any number of times.
+ *
+ * Context: Any context.
+ */
+void blake2s_update(struct blake2s_ctx *ctx, const u8 *in, size_t inlen);
+
+/**
+ * blake2s_final() - Finish computing a BLAKE2s hash
+ * @ctx: the context to finalize; must have been initialized
+ * @out: (output) the resulting BLAKE2s hash.  Its length will be equal to the
+ *	 @outlen that was passed to blake2s_init() or blake2s_init_key().
+ *
+ * After finishing, this zeroizes @ctx.  So the caller does not need to do it.
+ *
+ * Context: Any context.
+ */
+void blake2s_final(struct blake2s_ctx *ctx, u8 *out);
+
+/**
+ * blake2s() - Compute BLAKE2s hash in one shot
+ * @key: the key, or NULL for an unkeyed hash
+ * @keylen: the key length in bytes (at most BLAKE2S_KEY_SIZE), or 0 for an
+ *	    unkeyed hash
+ * @in: the message data
+ * @inlen: the data length in bytes
+ * @out: (output) the resulting BLAKE2s hash, with length @outlen
+ * @outlen: length of output hash value in bytes, at most BLAKE2S_HASH_SIZE
+ *
+ * Context: Any context.
+ */
+static inline void blake2s(const u8 *key, size_t keylen,
+			   const u8 *in, size_t inlen,
+			   u8 *out, size_t outlen)
+{
+	struct blake2s_ctx ctx;
+
+	WARN_ON(IS_ENABLED(DEBUG) && ((!in && inlen > 0) || !out || !outlen ||
+		outlen > BLAKE2S_HASH_SIZE || keylen > BLAKE2S_KEY_SIZE ||
+		(!key && keylen)));
+
+	__blake2s_init(&ctx, outlen, key, keylen);
+	blake2s_update(&ctx, in, inlen);
+	blake2s_final(&ctx, out);
+}
+
+#endif /* _CRYPTO_BLAKE2S_H */