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/mips/ralink/timer.c | 154 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 154 insertions(+) create mode 100644 arch/mips/ralink/timer.c (limited to 'arch/mips/ralink/timer.c') diff --git a/arch/mips/ralink/timer.c b/arch/mips/ralink/timer.c new file mode 100644 index 000000000..06bc2c014 --- /dev/null +++ b/arch/mips/ralink/timer.c @@ -0,0 +1,154 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Ralink RT2880 timer + * Author: John Crispin + * + * Copyright (C) 2013 John Crispin +*/ + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define TIMER_REG_TMRSTAT 0x00 +#define TIMER_REG_TMR0LOAD 0x10 +#define TIMER_REG_TMR0CTL 0x18 + +#define TMRSTAT_TMR0INT BIT(0) + +#define TMR0CTL_ENABLE BIT(7) +#define TMR0CTL_MODE_PERIODIC BIT(4) +#define TMR0CTL_PRESCALER 1 +#define TMR0CTL_PRESCALE_VAL (0xf - TMR0CTL_PRESCALER) +#define TMR0CTL_PRESCALE_DIV (65536 / BIT(TMR0CTL_PRESCALER)) + +struct rt_timer { + struct device *dev; + void __iomem *membase; + int irq; + unsigned long timer_freq; + unsigned long timer_div; +}; + +static inline void rt_timer_w32(struct rt_timer *rt, u8 reg, u32 val) +{ + __raw_writel(val, rt->membase + reg); +} + +static inline u32 rt_timer_r32(struct rt_timer *rt, u8 reg) +{ + return __raw_readl(rt->membase + reg); +} + +static irqreturn_t rt_timer_irq(int irq, void *_rt) +{ + struct rt_timer *rt = (struct rt_timer *) _rt; + + rt_timer_w32(rt, TIMER_REG_TMR0LOAD, rt->timer_freq / rt->timer_div); + rt_timer_w32(rt, TIMER_REG_TMRSTAT, TMRSTAT_TMR0INT); + + return IRQ_HANDLED; +} + + +static int rt_timer_request(struct rt_timer *rt) +{ + int err = request_irq(rt->irq, rt_timer_irq, 0, + dev_name(rt->dev), rt); + if (err) { + dev_err(rt->dev, "failed to request irq\n"); + } else { + u32 t = TMR0CTL_MODE_PERIODIC | TMR0CTL_PRESCALE_VAL; + rt_timer_w32(rt, TIMER_REG_TMR0CTL, t); + } + return err; +} + +static int rt_timer_config(struct rt_timer *rt, unsigned long divisor) +{ + if (rt->timer_freq < divisor) + rt->timer_div = rt->timer_freq; + else + rt->timer_div = divisor; + + rt_timer_w32(rt, TIMER_REG_TMR0LOAD, rt->timer_freq / rt->timer_div); + + return 0; +} + +static int rt_timer_enable(struct rt_timer *rt) +{ + u32 t; + + rt_timer_w32(rt, TIMER_REG_TMR0LOAD, rt->timer_freq / rt->timer_div); + + t = rt_timer_r32(rt, TIMER_REG_TMR0CTL); + t |= TMR0CTL_ENABLE; + rt_timer_w32(rt, TIMER_REG_TMR0CTL, t); + + return 0; +} + +static int rt_timer_probe(struct platform_device *pdev) +{ + struct rt_timer *rt; + struct clk *clk; + + rt = devm_kzalloc(&pdev->dev, sizeof(*rt), GFP_KERNEL); + if (!rt) { + dev_err(&pdev->dev, "failed to allocate memory\n"); + return -ENOMEM; + } + + rt->irq = platform_get_irq(pdev, 0); + if (rt->irq < 0) + return rt->irq; + + rt->membase = devm_platform_get_and_ioremap_resource(pdev, 0, NULL); + if (IS_ERR(rt->membase)) + return PTR_ERR(rt->membase); + + clk = devm_clk_get(&pdev->dev, NULL); + if (IS_ERR(clk)) { + dev_err(&pdev->dev, "failed get clock rate\n"); + return PTR_ERR(clk); + } + + rt->timer_freq = clk_get_rate(clk) / TMR0CTL_PRESCALE_DIV; + if (!rt->timer_freq) + return -EINVAL; + + rt->dev = &pdev->dev; + platform_set_drvdata(pdev, rt); + + rt_timer_request(rt); + rt_timer_config(rt, 2); + rt_timer_enable(rt); + + dev_info(&pdev->dev, "maximum frequency is %luHz\n", rt->timer_freq); + + return 0; +} + +static const struct of_device_id rt_timer_match[] = { + { .compatible = "ralink,rt2880-timer" }, + {}, +}; + +static struct platform_driver rt_timer_driver = { + .probe = rt_timer_probe, + .driver = { + .name = "rt-timer", + .of_match_table = rt_timer_match, + .suppress_bind_attrs = true, + }, +}; +builtin_platform_driver(rt_timer_driver); -- cgit v1.3.1