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, 188 insertions, 0 deletions

diff --git a/drivers/crypto/caam/Kconfig b/drivers/crypto/caam/Kconfig
new file mode 100644
index 000000000..ec57bf6aa
--- /dev/null
+++ b/drivers/crypto/caam/Kconfig
@@ -0,0 +1,188 @@
+# SPDX-License-Identifier: GPL-2.0
+config CRYPTO_DEV_FSL_CAAM_COMMON
+	tristate
+
+config CRYPTO_DEV_FSL_CAAM_CRYPTO_API_DESC
+	tristate
+
+config CRYPTO_DEV_FSL_CAAM_AHASH_API_DESC
+	tristate
+
+config CRYPTO_DEV_FSL_CAAM
+	tristate "Freescale CAAM-Multicore platform driver backend"
+	depends on FSL_SOC || ARCH_MXC || ARCH_LAYERSCAPE || COMPILE_TEST
+	select SOC_BUS
+	select CRYPTO_DEV_FSL_CAAM_COMMON
+	imply FSL_MC_BUS
+	help
+	  Enables the driver module for Freescale's Cryptographic Accelerator
+	  and Assurance Module (CAAM), also known as the SEC version 4 (SEC4).
+	  This module creates job ring devices, and configures h/w
+	  to operate as a DPAA component automatically, depending
+	  on h/w feature availability.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called caam.
+
+if CRYPTO_DEV_FSL_CAAM
+
+config CRYPTO_DEV_FSL_CAAM_DEBUG
+	bool "Enable debug output in CAAM driver"
+	help
+	  Selecting this will enable printing of various debug
+	  information in the CAAM driver.
+
+menuconfig CRYPTO_DEV_FSL_CAAM_JR
+	tristate "Freescale CAAM Job Ring driver backend"
+	select CRYPTO_ENGINE
+	default y
+	help
+	  Enables the driver module for Job Rings which are part of
+	  Freescale's Cryptographic Accelerator
+	  and Assurance Module (CAAM). This module adds a job ring operation
+	  interface.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called caam_jr.
+
+if CRYPTO_DEV_FSL_CAAM_JR
+
+config CRYPTO_DEV_FSL_CAAM_RINGSIZE
+	int "Job Ring size"
+	range 2 9
+	default "9"
+	help
+	  Select size of Job Rings as a power of 2, within the
+	  range 2-9 (ring size 4-512).
+	  Examples:
+		2 => 4
+		3 => 8
+		4 => 16
+		5 => 32
+		6 => 64
+		7 => 128
+		8 => 256
+		9 => 512
+
+config CRYPTO_DEV_FSL_CAAM_INTC
+	bool "Job Ring interrupt coalescing"
+	help
+	  Enable the Job Ring's interrupt coalescing feature.
+
+	  Note: the driver already provides adequate
+	  interrupt coalescing in software.
+
+config CRYPTO_DEV_FSL_CAAM_INTC_COUNT_THLD
+	int "Job Ring interrupt coalescing count threshold"
+	depends on CRYPTO_DEV_FSL_CAAM_INTC
+	range 1 255
+	default 255
+	help
+	  Select number of descriptor completions to queue before
+	  raising an interrupt, in the range 1-255. Note that a selection
+	  of 1 functionally defeats the coalescing feature, and a selection
+	  equal or greater than the job ring size will force timeouts.
+
+config CRYPTO_DEV_FSL_CAAM_INTC_TIME_THLD
+	int "Job Ring interrupt coalescing timer threshold"
+	depends on CRYPTO_DEV_FSL_CAAM_INTC
+	range 1 65535
+	default 2048
+	help
+	  Select number of bus clocks/64 to timeout in the case that one or
+	  more descriptor completions are queued without reaching the count
+	  threshold. Range is 1-65535.
+
+config CRYPTO_DEV_FSL_CAAM_CRYPTO_API
+	bool "Register algorithm implementations with the Crypto API"
+	default y
+	select CRYPTO_DEV_FSL_CAAM_CRYPTO_API_DESC
+	select CRYPTO_AEAD
+	select CRYPTO_AUTHENC
+	select CRYPTO_SKCIPHER
+	select CRYPTO_LIB_DES
+	select CRYPTO_XTS
+	help
+	  Selecting this will offload crypto for users of the
+	  scatterlist crypto API (such as the linux native IPSec
+	  stack) to the SEC4 via job ring.
+
+config CRYPTO_DEV_FSL_CAAM_CRYPTO_API_QI
+	bool "Queue Interface as Crypto API backend"
+	depends on FSL_DPAA && NET
+	default y
+	select CRYPTO_DEV_FSL_CAAM_CRYPTO_API_DESC
+	select CRYPTO_AUTHENC
+	select CRYPTO_SKCIPHER
+	select CRYPTO_DES
+	select CRYPTO_XTS
+	help
+	  Selecting this will use CAAM Queue Interface (QI) for sending
+	  & receiving crypto jobs to/from CAAM. This gives better performance
+	  than job ring interface when the number of cores are more than the
+	  number of job rings assigned to the kernel. The number of portals
+	  assigned to the kernel should also be more than the number of
+	  job rings.
+
+config CRYPTO_DEV_FSL_CAAM_AHASH_API
+	bool "Register hash algorithm implementations with Crypto API"
+	default y
+	select CRYPTO_DEV_FSL_CAAM_AHASH_API_DESC
+	select CRYPTO_HASH
+	help
+	  Selecting this will offload ahash for users of the
+	  scatterlist crypto API to the SEC4 via job ring.
+
+config CRYPTO_DEV_FSL_CAAM_PKC_API
+	bool "Register public key cryptography implementations with Crypto API"
+	default y
+	select CRYPTO_RSA
+	help
+	  Selecting this will allow SEC Public key support for RSA.
+	  Supported cryptographic primitives: encryption, decryption,
+	  signature and verification.
+
+config CRYPTO_DEV_FSL_CAAM_RNG_API
+	bool "Register caam device for hwrng API"
+	default y
+	select HW_RANDOM
+	help
+	  Selecting this will register the SEC4 hardware rng to
+	  the hw_random API for supplying the kernel entropy pool.
+
+config CRYPTO_DEV_FSL_CAAM_BLOB_GEN
+	bool
+
+config CRYPTO_DEV_FSL_CAAM_RNG_TEST
+	bool "Test caam rng"
+	select CRYPTO_DEV_FSL_CAAM_RNG_API
+	help
+	  Selecting this will enable a self-test to run for the
+	  caam RNG.
+	  This test is several minutes long and executes
+	  just before the RNG is registered with the hw_random API.
+
+endif # CRYPTO_DEV_FSL_CAAM_JR
+
+endif # CRYPTO_DEV_FSL_CAAM
+
+config CRYPTO_DEV_FSL_DPAA2_CAAM
+	tristate "QorIQ DPAA2 CAAM (DPSECI) driver"
+	depends on FSL_MC_DPIO
+	depends on NETDEVICES
+	select CRYPTO_DEV_FSL_CAAM_COMMON
+	select CRYPTO_DEV_FSL_CAAM_CRYPTO_API_DESC
+	select CRYPTO_DEV_FSL_CAAM_AHASH_API_DESC
+	select CRYPTO_SKCIPHER
+	select CRYPTO_AUTHENC
+	select CRYPTO_AEAD
+	select CRYPTO_HASH
+	select CRYPTO_DES
+	select CRYPTO_XTS
+	help
+	  CAAM driver for QorIQ Data Path Acceleration Architecture 2.
+	  It handles DPSECI DPAA2 objects that sit on the Management Complex
+	  (MC) fsl-mc bus.
+
+	  To compile this as a module, choose M here: the module
+	  will be called dpaa2_caam.