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, 116 insertions, 0 deletions

diff --git a/Documentation/admin-guide/device-mapper/era.rst b/Documentation/admin-guide/device-mapper/era.rst
new file mode 100644
index 000000000..90dd5c670
--- /dev/null
+++ b/Documentation/admin-guide/device-mapper/era.rst
@@ -0,0 +1,116 @@
+======
+dm-era
+======
+
+Introduction
+============
+
+dm-era is a target that behaves similar to the linear target.  In
+addition it keeps track of which blocks were written within a user
+defined period of time called an 'era'.  Each era target instance
+maintains the current era as a monotonically increasing 32-bit
+counter.
+
+Use cases include tracking changed blocks for backup software, and
+partially invalidating the contents of a cache to restore cache
+coherency after rolling back a vendor snapshot.
+
+Constructor
+===========
+
+era <metadata dev> <origin dev> <block size>
+
+ ================ ======================================================
+ metadata dev     fast device holding the persistent metadata
+ origin dev	  device holding data blocks that may change
+ block size       block size of origin data device, granularity that is
+		  tracked by the target
+ ================ ======================================================
+
+Messages
+========
+
+None of the dm messages take any arguments.
+
+checkpoint
+----------
+
+Possibly move to a new era.  You shouldn't assume the era has
+incremented.  After sending this message, you should check the
+current era via the status line.
+
+take_metadata_snap
+------------------
+
+Create a clone of the metadata, to allow a userland process to read it.
+
+drop_metadata_snap
+------------------
+
+Drop the metadata snapshot.
+
+Status
+======
+
+<metadata block size> <#used metadata blocks>/<#total metadata blocks>
+<current era> <held metadata root | '-'>
+
+========================= ==============================================
+metadata block size	  Fixed block size for each metadata block in
+			  sectors
+#used metadata blocks	  Number of metadata blocks used
+#total metadata blocks	  Total number of metadata blocks
+current era		  The current era
+held metadata root	  The location, in blocks, of the metadata root
+			  that has been 'held' for userspace read
+			  access. '-' indicates there is no held root
+========================= ==============================================
+
+Detailed use case
+=================
+
+The scenario of invalidating a cache when rolling back a vendor
+snapshot was the primary use case when developing this target:
+
+Taking a vendor snapshot
+------------------------
+
+- Send a checkpoint message to the era target
+- Make a note of the current era in its status line
+- Take vendor snapshot (the era and snapshot should be forever
+  associated now).
+
+Rolling back to an vendor snapshot
+----------------------------------
+
+- Cache enters passthrough mode (see: dm-cache's docs in cache.txt)
+- Rollback vendor storage
+- Take metadata snapshot
+- Ascertain which blocks have been written since the snapshot was taken
+  by checking each block's era
+- Invalidate those blocks in the caching software
+- Cache returns to writeback/writethrough mode
+
+Memory usage
+============
+
+The target uses a bitset to record writes in the current era.  It also
+has a spare bitset ready for switching over to a new era.  Other than
+that it uses a few 4k blocks for updating metadata::
+
+   (4 * nr_blocks) bytes + buffers
+
+Resilience
+==========
+
+Metadata is updated on disk before a write to a previously unwritten
+block is performed.  As such dm-era should not be effected by a hard
+crash such as power failure.
+
+Userland tools
+==============
+
+Userland tools are found in the increasingly poorly named
+thin-provisioning-tools project:
+
+    https://github.com/jthornber/thin-provisioning-tools