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, 107 insertions, 0 deletions

diff --git a/Documentation/driver-api/fpga/fpga-programming.rst b/Documentation/driver-api/fpga/fpga-programming.rst
new file mode 100644
index 000000000..fb4da4240
--- /dev/null
+++ b/Documentation/driver-api/fpga/fpga-programming.rst
@@ -0,0 +1,107 @@
+In-kernel API for FPGA Programming
+==================================
+
+Overview
+--------
+
+The in-kernel API for FPGA programming is a combination of APIs from
+FPGA manager, bridge, and regions.  The actual function used to
+trigger FPGA programming is fpga_region_program_fpga().
+
+fpga_region_program_fpga() uses functionality supplied by
+the FPGA manager and bridges.  It will:
+
+ * lock the region's mutex
+ * lock the mutex of the region's FPGA manager
+ * build a list of FPGA bridges if a method has been specified to do so
+ * disable the bridges
+ * program the FPGA using info passed in :c:expr:`fpga_region->info`.
+ * re-enable the bridges
+ * release the locks
+
+The struct fpga_image_info specifies what FPGA image to program.  It is
+allocated/freed by fpga_image_info_alloc() and freed with
+fpga_image_info_free()
+
+How to program an FPGA using a region
+-------------------------------------
+
+When the FPGA region driver probed, it was given a pointer to an FPGA manager
+driver so it knows which manager to use.  The region also either has a list of
+bridges to control during programming or it has a pointer to a function that
+will generate that list.  Here's some sample code of what to do next::
+
+	#include <linux/fpga/fpga-mgr.h>
+	#include <linux/fpga/fpga-region.h>
+
+	struct fpga_image_info *info;
+	int ret;
+
+	/*
+	 * First, alloc the struct with information about the FPGA image to
+	 * program.
+	 */
+	info = fpga_image_info_alloc(dev);
+	if (!info)
+		return -ENOMEM;
+
+	/* Set flags as needed, such as: */
+	info->flags = FPGA_MGR_PARTIAL_RECONFIG;
+
+	/*
+	 * Indicate where the FPGA image is. This is pseudo-code; you're
+	 * going to use one of these three.
+	 */
+	if (image is in a scatter gather table) {
+
+		info->sgt = [your scatter gather table]
+
+	} else if (image is in a buffer) {
+
+		info->buf = [your image buffer]
+		info->count = [image buffer size]
+
+	} else if (image is in a firmware file) {
+
+		info->firmware_name = devm_kstrdup(dev, firmware_name,
+						   GFP_KERNEL);
+
+	}
+
+	/* Add info to region and do the programming */
+	region->info = info;
+	ret = fpga_region_program_fpga(region);
+
+	/* Deallocate the image info if you're done with it */
+	region->info = NULL;
+	fpga_image_info_free(info);
+
+	if (ret)
+		return ret;
+
+	/* Now enumerate whatever hardware has appeared in the FPGA. */
+
+API for programming an FPGA
+---------------------------
+
+* fpga_region_program_fpga() -  Program an FPGA
+* fpga_image_info() -  Specifies what FPGA image to program
+* fpga_image_info_alloc() -  Allocate an FPGA image info struct
+* fpga_image_info_free() -  Free an FPGA image info struct
+
+.. kernel-doc:: drivers/fpga/fpga-region.c
+   :functions: fpga_region_program_fpga
+
+FPGA Manager flags
+
+.. kernel-doc:: include/linux/fpga/fpga-mgr.h
+   :doc: FPGA Manager flags
+
+.. kernel-doc:: include/linux/fpga/fpga-mgr.h
+   :functions: fpga_image_info
+
+.. kernel-doc:: drivers/fpga/fpga-mgr.c
+   :functions: fpga_image_info_alloc
+
+.. kernel-doc:: drivers/fpga/fpga-mgr.c
+   :functions: fpga_image_info_free