From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug 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 Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- arch/arm/lib/delay.c | 103 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 103 insertions(+) create mode 100644 arch/arm/lib/delay.c (limited to 'arch/arm/lib/delay.c') diff --git a/arch/arm/lib/delay.c b/arch/arm/lib/delay.c new file mode 100644 index 000000000..7e9304012 --- /dev/null +++ b/arch/arm/lib/delay.c @@ -0,0 +1,103 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Delay loops based on the OpenRISC implementation. + * + * Copyright (C) 2012 ARM Limited + * + * Author: Will Deacon + */ + +#include +#include +#include +#include +#include + +/* + * Default to the loop-based delay implementation. + */ +struct arm_delay_ops arm_delay_ops __ro_after_init = { + .delay = __loop_delay, + .const_udelay = __loop_const_udelay, + .udelay = __loop_udelay, +}; + +static const struct delay_timer *delay_timer; +static bool delay_calibrated; +static u64 delay_res; + +bool delay_read_timer(unsigned long *timer_val) +{ + if (!delay_timer) + return false; + *timer_val = delay_timer->read_current_timer(); + return true; +} +EXPORT_SYMBOL_GPL(delay_read_timer); + +static inline u64 cyc_to_ns(u64 cyc, u32 mult, u32 shift) +{ + return (cyc * mult) >> shift; +} + +static void __timer_delay(unsigned long cycles) +{ + cycles_t start = get_cycles(); + + while ((get_cycles() - start) < cycles) + cpu_relax(); +} + +static void __timer_const_udelay(unsigned long xloops) +{ + unsigned long long loops = xloops; + loops *= arm_delay_ops.ticks_per_jiffy; + __timer_delay(loops >> UDELAY_SHIFT); +} + +static void __timer_udelay(unsigned long usecs) +{ + __timer_const_udelay(usecs * UDELAY_MULT); +} + +void __init register_current_timer_delay(const struct delay_timer *timer) +{ + u32 new_mult, new_shift; + u64 res; + + clocks_calc_mult_shift(&new_mult, &new_shift, timer->freq, + NSEC_PER_SEC, 3600); + res = cyc_to_ns(1ULL, new_mult, new_shift); + + if (res > 1000) { + pr_err("Ignoring delay timer %ps, which has insufficient resolution of %lluns\n", + timer, res); + return; + } + + if (!delay_calibrated && (!delay_res || (res < delay_res))) { + pr_info("Switching to timer-based delay loop, resolution %lluns\n", res); + delay_timer = timer; + lpj_fine = timer->freq / HZ; + delay_res = res; + + /* cpufreq may scale loops_per_jiffy, so keep a private copy */ + arm_delay_ops.ticks_per_jiffy = lpj_fine; + arm_delay_ops.delay = __timer_delay; + arm_delay_ops.const_udelay = __timer_const_udelay; + arm_delay_ops.udelay = __timer_udelay; + } else { + pr_info("Ignoring duplicate/late registration of read_current_timer delay\n"); + } +} + +unsigned long calibrate_delay_is_known(void) +{ + delay_calibrated = true; + return lpj_fine; +} + +void calibration_delay_done(void) +{ + delay_calibrated = true; +} -- cgit v1.3.1