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, 77 insertions, 0 deletions

diff --git a/net/sunrpc/Kconfig b/net/sunrpc/Kconfig
new file mode 100644
index 000000000..e7808e571
--- /dev/null
+++ b/net/sunrpc/Kconfig
@@ -0,0 +1,77 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config SUNRPC
+	tristate
+	depends on MULTIUSER
+
+config SUNRPC_GSS
+	tristate
+	select OID_REGISTRY
+	depends on MULTIUSER
+
+config SUNRPC_BACKCHANNEL
+	bool
+	depends on SUNRPC
+
+config SUNRPC_SWAP
+	bool
+	depends on SUNRPC
+
+config RPCSEC_GSS_KRB5
+	tristate "Secure RPC: Kerberos V mechanism"
+	depends on SUNRPC && CRYPTO
+	default y
+	select SUNRPC_GSS
+	select CRYPTO_KRB5
+	help
+	  Choose Y here to enable Secure RPC using the Kerberos version 5
+	  GSS-API mechanism (RFC 1964).
+
+	  Secure RPC calls with Kerberos require an auxiliary user-space
+	  daemon which may be found in the Linux nfs-utils package
+	  available from http://linux-nfs.org/.  In addition, user-space
+	  Kerberos support should be installed.
+
+	  If unsure, say Y.
+
+config SUNRPC_DEBUG
+	bool "RPC: Enable dprintk debugging"
+	depends on SUNRPC && SYSCTL
+	select DEBUG_FS
+	help
+	  This option enables a sysctl-based debugging interface
+	  that is be used by the 'rpcdebug' utility to turn on or off
+	  logging of different aspects of the kernel RPC activity.
+
+	  Disabling this option will make your kernel slightly smaller,
+	  but makes troubleshooting NFS issues significantly harder.
+
+	  If unsure, say Y.
+
+config SUNRPC_DEBUG_TRACE
+	bool "RPC: Send dfprintk() output to the trace buffer"
+	depends on SUNRPC_DEBUG && TRACING
+	default n
+	help
+          dprintk() output can be voluminous, which can overwhelm the
+          kernel's logging facility as it must be sent to the console.
+          This option causes dprintk() output to go to the trace buffer
+          instead of the kernel log.
+
+          This will cause warnings about trace_printk() being used to be
+          logged at boot time, so say N unless you are debugging a problem
+          with sunrpc-based clients or services.
+
+config SUNRPC_XPRT_RDMA
+	tristate "RPC-over-RDMA transport"
+	depends on SUNRPC && INFINIBAND && INFINIBAND_ADDR_TRANS
+	default SUNRPC && INFINIBAND
+	select SG_POOL
+	help
+	  This option allows the NFS client and server to use RDMA
+	  transports (InfiniBand, iWARP, or RoCE).
+
+	  To compile this support as a module, choose M. The module
+	  will be called rpcrdma.ko.
+
+	  If unsure, or you know there is no RDMA capability on your
+	  hardware platform, say N.