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, 156 insertions, 0 deletions

diff --git a/arch/x86/boot/cmdline.c b/arch/x86/boot/cmdline.c
new file mode 100644
index 000000000..21d56ae83
--- /dev/null
+++ b/arch/x86/boot/cmdline.c
@@ -0,0 +1,156 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* -*- linux-c -*- ------------------------------------------------------- *
+ *
+ *   Copyright (C) 1991, 1992 Linus Torvalds
+ *   Copyright 2007 rPath, Inc. - All Rights Reserved
+ *
+ * ----------------------------------------------------------------------- */
+
+/*
+ * Simple command-line parser for early boot.
+ */
+
+#include "boot.h"
+
+static inline int myisspace(u8 c)
+{
+	return c <= ' ';	/* Close enough approximation */
+}
+
+/*
+ * Find a non-boolean option, that is, "option=argument".  In accordance
+ * with standard Linux practice, if this option is repeated, this returns
+ * the last instance on the command line.
+ *
+ * Returns the length of the argument (regardless of if it was
+ * truncated to fit in the buffer), or -1 on not found.
+ */
+int __cmdline_find_option(unsigned long cmdline_ptr, const char *option, char *buffer, int bufsize)
+{
+	addr_t cptr;
+	char c;
+	int len = -1;
+	const char *opptr = NULL;
+	char *bufptr = buffer;
+	enum {
+		st_wordstart,	/* Start of word/after whitespace */
+		st_wordcmp,	/* Comparing this word */
+		st_wordskip,	/* Miscompare, skip */
+		st_bufcpy	/* Copying this to buffer */
+	} state = st_wordstart;
+
+	if (!cmdline_ptr)
+		return -1;      /* No command line */
+
+	cptr = cmdline_ptr & 0xf;
+	set_fs(cmdline_ptr >> 4);
+
+	while (cptr < 0x10000 && (c = rdfs8(cptr++))) {
+		switch (state) {
+		case st_wordstart:
+			if (myisspace(c))
+				break;
+
+			/* else */
+			state = st_wordcmp;
+			opptr = option;
+			fallthrough;
+
+		case st_wordcmp:
+			if (c == '=' && !*opptr) {
+				len = 0;
+				bufptr = buffer;
+				state = st_bufcpy;
+			} else if (myisspace(c)) {
+				state = st_wordstart;
+			} else if (c != *opptr++) {
+				state = st_wordskip;
+			}
+			break;
+
+		case st_wordskip:
+			if (myisspace(c))
+				state = st_wordstart;
+			break;
+
+		case st_bufcpy:
+			if (myisspace(c)) {
+				state = st_wordstart;
+			} else {
+				if (len < bufsize-1)
+					*bufptr++ = c;
+				len++;
+			}
+			break;
+		}
+	}
+
+	if (bufsize)
+		*bufptr = '\0';
+
+	return len;
+}
+
+/*
+ * Find a boolean option (like quiet,noapic,nosmp....)
+ *
+ * Returns the position of that option (starts counting with 1)
+ * or 0 on not found
+ */
+int __cmdline_find_option_bool(unsigned long cmdline_ptr, const char *option)
+{
+	addr_t cptr;
+	char c;
+	int pos = 0, wstart = 0;
+	const char *opptr = NULL;
+	enum {
+		st_wordstart,	/* Start of word/after whitespace */
+		st_wordcmp,	/* Comparing this word */
+		st_wordskip,	/* Miscompare, skip */
+	} state = st_wordstart;
+
+	if (!cmdline_ptr)
+		return -1;      /* No command line */
+
+	cptr = cmdline_ptr & 0xf;
+	set_fs(cmdline_ptr >> 4);
+
+	while (cptr < 0x10000) {
+		c = rdfs8(cptr++);
+		pos++;
+
+		switch (state) {
+		case st_wordstart:
+			if (!c)
+				return 0;
+			else if (myisspace(c))
+				break;
+
+			state = st_wordcmp;
+			opptr = option;
+			wstart = pos;
+			fallthrough;
+
+		case st_wordcmp:
+			if (!*opptr)
+				if (!c || myisspace(c))
+					return wstart;
+				else
+					state = st_wordskip;
+			else if (!c)
+				return 0;
+			else if (c != *opptr++)
+				state = st_wordskip;
+			break;
+
+		case st_wordskip:
+			if (!c)
+				return 0;
+			else if (myisspace(c))
+				state = st_wordstart;
+			break;
+		}
+	}
+
+	return 0;	/* Buffer overrun */
+}