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, 151 insertions, 0 deletions

diff --git a/arch/powerpc/boot/mktree.c b/arch/powerpc/boot/mktree.c
new file mode 100644
index 000000000..dc603f3c1
--- /dev/null
+++ b/arch/powerpc/boot/mktree.c
@@ -0,0 +1,151 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Makes a tree bootable image for IBM Evaluation boards.
+ * Basically, just take a zImage, skip the ELF header, and stuff
+ * a 32 byte header on the front.
+ *
+ * We use htonl, which is a network macro, to make sure we're doing
+ * The Right Thing on an LE machine.  It's non-obvious, but it should
+ * work on anything BSD'ish.
+ */
+
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <netinet/in.h>
+#ifdef __sun__
+#include <inttypes.h>
+#else
+#include <stdint.h>
+#endif
+
+/* This gets tacked on the front of the image.  There are also a few
+ * bytes allocated after the _start label used by the boot rom (see
+ * head.S for details).
+ */
+typedef struct boot_block {
+	uint32_t bb_magic;		/* 0x0052504F */
+	uint32_t bb_dest;		/* Target address of the image */
+	uint32_t bb_num_512blocks;	/* Size, rounded-up, in 512 byte blks */
+	uint32_t bb_debug_flag;	/* Run debugger or image after load */
+	uint32_t bb_entry_point;	/* The image address to start */
+	uint32_t bb_checksum;	/* 32 bit checksum including header */
+	uint32_t reserved[2];
+} boot_block_t;
+
+#define IMGBLK	512
+unsigned int	tmpbuf[IMGBLK / sizeof(unsigned int)];
+
+int main(int argc, char *argv[])
+{
+	int	in_fd, out_fd;
+	int	nblks, i;
+	unsigned int	cksum, *cp;
+	struct	stat	st;
+	boot_block_t	bt;
+
+	if (argc < 5) {
+		fprintf(stderr, "usage: %s <zImage-file> <boot-image> <load address> <entry point>\n",argv[0]);
+		exit(1);
+	}
+
+	if (stat(argv[1], &st) < 0) {
+		perror("stat");
+		exit(2);
+	}
+
+	nblks = (st.st_size + IMGBLK) / IMGBLK;
+
+	bt.bb_magic = htonl(0x0052504F);
+
+	/* If we have the optional entry point parameter, use it */
+	bt.bb_dest = htonl(strtoul(argv[3], NULL, 0));
+	bt.bb_entry_point = htonl(strtoul(argv[4], NULL, 0));
+
+	/* We know these from the linker command.
+	 * ...and then move it up into memory a little more so the
+	 * relocation can happen.
+	 */
+	bt.bb_num_512blocks = htonl(nblks);
+	bt.bb_debug_flag = 0;
+
+	bt.bb_checksum = 0;
+
+	/* To be neat and tidy :-).
+	*/
+	bt.reserved[0] = 0;
+	bt.reserved[1] = 0;
+
+	if ((in_fd = open(argv[1], O_RDONLY)) < 0) {
+		perror("zImage open");
+		exit(3);
+	}
+
+	if ((out_fd = open(argv[2], (O_RDWR | O_CREAT | O_TRUNC), 0666)) < 0) {
+		perror("bootfile open");
+		exit(3);
+	}
+
+	cksum = 0;
+	cp = (void *)&bt;
+	for (i = 0; i < sizeof(bt) / sizeof(unsigned int); i++)
+		cksum += *cp++;
+
+	/* Assume zImage is an ELF file, and skip the 64K header.
+	*/
+	if (read(in_fd, tmpbuf, sizeof(tmpbuf)) != sizeof(tmpbuf)) {
+		fprintf(stderr, "%s is too small to be an ELF image\n",
+				argv[1]);
+		exit(4);
+	}
+
+	if (tmpbuf[0] != htonl(0x7f454c46)) {
+		fprintf(stderr, "%s is not an ELF image\n", argv[1]);
+		exit(4);
+	}
+
+	if (lseek(in_fd, (64 * 1024), SEEK_SET) < 0) {
+		fprintf(stderr, "%s failed to seek in ELF image\n", argv[1]);
+		exit(4);
+	}
+
+	nblks -= (64 * 1024) / IMGBLK;
+
+	/* And away we go......
+	*/
+	if (write(out_fd, &bt, sizeof(bt)) != sizeof(bt)) {
+		perror("boot-image write");
+		exit(5);
+	}
+
+	while (nblks-- > 0) {
+		if (read(in_fd, tmpbuf, sizeof(tmpbuf)) < 0) {
+			perror("zImage read");
+			exit(5);
+		}
+		cp = tmpbuf;
+		for (i = 0; i < sizeof(tmpbuf) / sizeof(unsigned int); i++)
+			cksum += *cp++;
+		if (write(out_fd, tmpbuf, sizeof(tmpbuf)) != sizeof(tmpbuf)) {
+			perror("boot-image write");
+			exit(5);
+		}
+	}
+
+	/* rewrite the header with the computed checksum.
+	*/
+	bt.bb_checksum = htonl(cksum);
+	if (lseek(out_fd, 0, SEEK_SET) < 0) {
+		perror("rewrite seek");
+		exit(1);
+	}
+	if (write(out_fd, &bt, sizeof(bt)) != sizeof(bt)) {
+		perror("boot-image rewrite");
+		exit(1);
+	}
+
+	exit(0);
+}