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/m68k/virt/ints.c | 154 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 154 insertions(+) create mode 100644 arch/m68k/virt/ints.c (limited to 'arch/m68k/virt/ints.c') diff --git a/arch/m68k/virt/ints.c b/arch/m68k/virt/ints.c new file mode 100644 index 000000000..896aa6eb8 --- /dev/null +++ b/arch/m68k/virt/ints.c @@ -0,0 +1,154 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +#define GFPIC_REG_IRQ_PENDING 0x04 +#define GFPIC_REG_IRQ_DISABLE_ALL 0x08 +#define GFPIC_REG_IRQ_DISABLE 0x0c +#define GFPIC_REG_IRQ_ENABLE 0x10 + +static struct resource picres[6]; +static const char *picname[6] = { + "goldfish_pic.0", + "goldfish_pic.1", + "goldfish_pic.2", + "goldfish_pic.3", + "goldfish_pic.4", + "goldfish_pic.5" +}; + +/* + * 6 goldfish-pic for CPU IRQ #1 to IRQ #6 + * CPU IRQ #1 -> PIC #1 + * IRQ #1 to IRQ #31 -> unused + * IRQ #32 -> goldfish-tty + * CPU IRQ #2 -> PIC #2 + * IRQ #1 to IRQ #32 -> virtio-mmio from 1 to 32 + * CPU IRQ #3 -> PIC #3 + * IRQ #1 to IRQ #32 -> virtio-mmio from 33 to 64 + * CPU IRQ #4 -> PIC #4 + * IRQ #1 to IRQ #32 -> virtio-mmio from 65 to 96 + * CPU IRQ #5 -> PIC #5 + * IRQ #1 to IRQ #32 -> virtio-mmio from 97 to 128 + * CPU IRQ #6 -> PIC #6 + * IRQ #1 -> goldfish-timer + * IRQ #2 -> goldfish-rtc + * IRQ #3 to IRQ #32 -> unused + * CPU IRQ #7 -> NMI + */ + +static u32 gfpic_read(int pic, int reg) +{ + void __iomem *base = (void __iomem *)(virt_bi_data.pic.mmio + + pic * 0x1000); + + return ioread32be(base + reg); +} + +static void gfpic_write(u32 value, int pic, int reg) +{ + void __iomem *base = (void __iomem *)(virt_bi_data.pic.mmio + + pic * 0x1000); + + iowrite32be(value, base + reg); +} + +#define GF_PIC(irq) ((irq - IRQ_USER) / 32) +#define GF_IRQ(irq) ((irq - IRQ_USER) % 32) + +static void virt_irq_enable(struct irq_data *data) +{ + gfpic_write(BIT(GF_IRQ(data->irq)), GF_PIC(data->irq), + GFPIC_REG_IRQ_ENABLE); +} + +static void virt_irq_disable(struct irq_data *data) +{ + gfpic_write(BIT(GF_IRQ(data->irq)), GF_PIC(data->irq), + GFPIC_REG_IRQ_DISABLE); +} + +static unsigned int virt_irq_startup(struct irq_data *data) +{ + virt_irq_enable(data); + return 0; +} + +static irqreturn_t virt_nmi_handler(int irq, void *dev_id) +{ + static int in_nmi; + + if (READ_ONCE(in_nmi)) + return IRQ_HANDLED; + WRITE_ONCE(in_nmi, 1); + + pr_warn("Non-Maskable Interrupt\n"); + show_registers(get_irq_regs()); + + WRITE_ONCE(in_nmi, 0); + return IRQ_HANDLED; +} + +static struct irq_chip virt_irq_chip = { + .name = "virt", + .irq_enable = virt_irq_enable, + .irq_disable = virt_irq_disable, + .irq_startup = virt_irq_startup, + .irq_shutdown = virt_irq_disable, +}; + +static void goldfish_pic_irq(struct irq_desc *desc) +{ + u32 irq_pending; + unsigned int irq_num; + unsigned int pic = desc->irq_data.irq - 1; + + irq_pending = gfpic_read(pic, GFPIC_REG_IRQ_PENDING); + irq_num = IRQ_USER + pic * 32; + + do { + if (irq_pending & 1) + generic_handle_irq(irq_num); + ++irq_num; + irq_pending >>= 1; + } while (irq_pending); +} + +void __init virt_init_IRQ(void) +{ + unsigned int i; + + m68k_setup_irq_controller(&virt_irq_chip, handle_simple_irq, IRQ_USER, + NUM_VIRT_SOURCES - IRQ_USER); + + for (i = 0; i < 6; i++) { + + picres[i] = (struct resource) + DEFINE_RES_MEM_NAMED(virt_bi_data.pic.mmio + i * 0x1000, + 0x1000, picname[i]); + if (request_resource(&iomem_resource, &picres[i])) { + pr_err("Cannot allocate %s resource\n", picname[i]); + return; + } + + irq_set_chained_handler(virt_bi_data.pic.irq + i, + goldfish_pic_irq); + } + + if (request_irq(IRQ_AUTO_7, virt_nmi_handler, 0, "NMI", + virt_nmi_handler)) + pr_err("Couldn't register NMI\n"); +} -- cgit v1.3.1