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, 51 insertions, 0 deletions

diff --git a/Documentation/admin-guide/device-mapper/dm-ebs.rst b/Documentation/admin-guide/device-mapper/dm-ebs.rst
new file mode 100644
index 000000000..c09f66db5
--- /dev/null
+++ b/Documentation/admin-guide/device-mapper/dm-ebs.rst
@@ -0,0 +1,51 @@
+======
+dm-ebs
+======
+
+
+This target is similar to the linear target except that it emulates
+a smaller logical block size on a device with a larger logical block
+size.  Its main purpose is to provide emulation of 512 byte sectors on
+devices that do not provide this emulation (i.e. 4K native disks).
+
+Supported emulated logical block sizes 512, 1024, 2048 and 4096.
+
+Underlying block size can be set to > 4K to test buffering larger units.
+
+
+Table parameters
+----------------
+  <dev path> <offset> <emulated sectors> [<underlying sectors>]
+
+Mandatory parameters:
+
+    <dev path>:
+        Full pathname to the underlying block-device,
+        or a "major:minor" device-number.
+    <offset>:
+        Starting sector within the device;
+        has to be a multiple of <emulated sectors>.
+    <emulated sectors>:
+        Number of sectors defining the logical block size to be emulated;
+        1, 2, 4, 8 sectors of 512 bytes supported.
+
+Optional parameter:
+
+    <underlying sectors>:
+        Number of sectors defining the logical block size of <dev path>.
+        2^N supported, e.g. 8 = emulate 8 sectors of 512 bytes = 4KiB.
+        If not provided, the logical block size of <dev path> will be used.
+
+
+Examples:
+
+Emulate 1 sector = 512 bytes logical block size on /dev/sda starting at
+offset 1024 sectors with underlying devices block size automatically set:
+
+ebs /dev/sda 1024 1
+
+Emulate 2 sector = 1KiB logical block size on /dev/sda starting at
+offset 128 sectors, enforce 2KiB underlying device block size.
+This presumes 2KiB logical blocksize on /dev/sda or less to work:
+
+ebs /dev/sda 128 2 4