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/ABI/testing/securityfs-secrets-coco b/Documentation/ABI/testing/securityfs-secrets-coco
new file mode 100644
index 000000000..f2b690915
--- /dev/null
+++ b/Documentation/ABI/testing/securityfs-secrets-coco
@@ -0,0 +1,51 @@
+What:		security/secrets/coco
+Date:		February 2022
+Contact:	Dov Murik <dovmurik@linux.ibm.com>
+Description:
+		Exposes confidential computing (coco) EFI secrets to
+		userspace via securityfs.
+
+		EFI can declare memory area used by confidential computing
+		platforms (such as AMD SEV and SEV-ES) for secret injection by
+		the Guest Owner during VM's launch.  The secrets are encrypted
+		by the Guest Owner and decrypted inside the trusted enclave,
+		and therefore are not readable by the untrusted host.
+
+		The efi_secret module exposes the secrets to userspace.  Each
+		secret appears as a file under <securityfs>/secrets/coco,
+		where the filename is the GUID of the entry in the secrets
+		table.  This module is loaded automatically by the EFI driver
+		if the EFI secret area is populated.
+
+		Two operations are supported for the files: read and unlink.
+		Reading the file returns the content of secret entry.
+		Unlinking the file overwrites the secret data with zeroes and
+		removes the entry from the filesystem.  A secret cannot be read
+		after it has been unlinked.
+
+		For example, listing the available secrets::
+
+		  # modprobe efi_secret
+		  # ls -l /sys/kernel/security/secrets/coco
+		  -r--r----- 1 root root 0 Jun 28 11:54 736870e5-84f0-4973-92ec-06879ce3da0b
+		  -r--r----- 1 root root 0 Jun 28 11:54 83c83f7f-1356-4975-8b7e-d3a0b54312c6
+		  -r--r----- 1 root root 0 Jun 28 11:54 9553f55d-3da2-43ee-ab5d-ff17f78864d2
+		  -r--r----- 1 root root 0 Jun 28 11:54 e6f5a162-d67f-4750-a67c-5d065f2a9910
+
+		Reading the secret data by reading a file::
+
+		  # cat /sys/kernel/security/secrets/coco/e6f5a162-d67f-4750-a67c-5d065f2a9910
+		  the-content-of-the-secret-data
+
+		Wiping a secret by unlinking a file::
+
+		  # rm /sys/kernel/security/secrets/coco/e6f5a162-d67f-4750-a67c-5d065f2a9910
+		  # ls -l /sys/kernel/security/secrets/coco
+		  -r--r----- 1 root root 0 Jun 28 11:54 736870e5-84f0-4973-92ec-06879ce3da0b
+		  -r--r----- 1 root root 0 Jun 28 11:54 83c83f7f-1356-4975-8b7e-d3a0b54312c6
+		  -r--r----- 1 root root 0 Jun 28 11:54 9553f55d-3da2-43ee-ab5d-ff17f78864d2
+
+		Note: The binary format of the secrets table injected by the
+		Guest Owner is described in
+		drivers/virt/coco/efi_secret/efi_secret.c under "Structure of
+		the EFI secret area".