From a9e3760b0838299649c0d57cca44daaf40ba3c33 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 14 Sep 2026 22:12:34 -1000 Subject: sched_ext: Maintain an online cid mask in the scheduler arena Schedulers on the default cid mapping treat [0, nr_online_cids) as the online set and restart on hotplug. Schedulers that install their own mapping with scx_bpf_cid_override() have no way to learn which cids are online: the count no longer identifies members and the CPU-form cpumask is unusable from cid programs. This is an obvious hole in the cid API. Add scx_bpf_online_cmask(), a kernel-maintained cmask in the scheduler's arena, allocated alongside the per-CPU scratch masks and populated after the cid mapping is finalized and before ops.init(), for child schedulers too. The pointer stays valid through ops.exit() with no reference to take. It is the arena offset as a void pointer, the same form struct_ops arena arguments arrive in. The verifier types the void return as a scalar for the program's arena cast. The mask follows the SCX hotplug notifications: seeded from cpu_active_mask and updated before ops.cid_online/offline() runs, so it lags cpu_online_mask only inside a hotplug transition. Updates walk the scheduler list under the lock that also serializes unlinking. Reads are live, not atomic snapshots. Root initialization excludes hotplug. v2: Reworded the getter kerneldoc (Andrea Righi). Signed-off-by: Tejun Heo Reviewed-by: Andrea Righi --- drivers/clocksource/timer-gxp.c | 215 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 215 insertions(+) create mode 100644 drivers/clocksource/timer-gxp.c (limited to 'drivers/clocksource/timer-gxp.c') diff --git a/drivers/clocksource/timer-gxp.c b/drivers/clocksource/timer-gxp.c new file mode 100644 index 000000000..1becc9433 --- /dev/null +++ b/drivers/clocksource/timer-gxp.c @@ -0,0 +1,215 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (C) 2022 Hewlett-Packard Enterprise Development Company, L.P. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define TIMER0_FREQ 1000000 +#define GXP_TIMER_CNT_OFS 0x00 +#define GXP_TIMESTAMP_OFS 0x08 +#define GXP_TIMER_CTRL_OFS 0x14 + +/* TCS Stands for Timer Control/Status: these are masks to be used in */ +/* the Timer Count Registers */ +#define MASK_TCS_ENABLE 0x01 +#define MASK_TCS_PERIOD 0x02 +#define MASK_TCS_RELOAD 0x04 +#define MASK_TCS_TC 0x80 + +struct gxp_timer { + void __iomem *counter; + void __iomem *control; + struct clock_event_device evt; +}; + +static struct gxp_timer *gxp_timer; + +static void __iomem *system_clock __ro_after_init; + +static inline struct gxp_timer *to_gxp_timer(struct clock_event_device *evt_dev) +{ + return container_of(evt_dev, struct gxp_timer, evt); +} + +static u64 notrace gxp_sched_read(void) +{ + return readl_relaxed(system_clock); +} + +static int gxp_time_set_next_event(unsigned long event, struct clock_event_device *evt_dev) +{ + struct gxp_timer *timer = to_gxp_timer(evt_dev); + + /* Stop counting and disable interrupt before updating */ + writeb_relaxed(MASK_TCS_TC, timer->control); + writel_relaxed(event, timer->counter); + writeb_relaxed(MASK_TCS_TC | MASK_TCS_ENABLE, timer->control); + + return 0; +} + +static irqreturn_t gxp_timer_interrupt(int irq, void *dev_id) +{ + struct gxp_timer *timer = (struct gxp_timer *)dev_id; + + if (!(readb_relaxed(timer->control) & MASK_TCS_TC)) + return IRQ_NONE; + + writeb_relaxed(MASK_TCS_TC, timer->control); + + timer->evt.event_handler(&timer->evt); + + return IRQ_HANDLED; +} + +static int __init gxp_timer_init(struct device_node *node) +{ + void __iomem *base; + struct clk *clk; + u32 freq; + int ret, irq; + + gxp_timer = kzalloc_obj(*gxp_timer); + if (!gxp_timer) { + ret = -ENOMEM; + pr_err("Can't allocate gxp_timer"); + return ret; + } + + clk = of_clk_get(node, 0); + if (IS_ERR(clk)) { + ret = PTR_ERR(clk); + pr_err("%pOFn clock not found: %d\n", node, ret); + goto err_free; + } + + ret = clk_prepare_enable(clk); + if (ret) { + pr_err("%pOFn clock enable failed: %d\n", node, ret); + goto err_clk_enable; + } + + base = of_iomap(node, 0); + if (!base) { + ret = -ENXIO; + pr_err("Can't map timer base registers"); + goto err_iomap; + } + + /* Set the offsets to the clock register and timer registers */ + gxp_timer->counter = base + GXP_TIMER_CNT_OFS; + gxp_timer->control = base + GXP_TIMER_CTRL_OFS; + system_clock = base + GXP_TIMESTAMP_OFS; + + gxp_timer->evt.name = node->name; + gxp_timer->evt.rating = 300; + gxp_timer->evt.features = CLOCK_EVT_FEAT_ONESHOT; + gxp_timer->evt.set_next_event = gxp_time_set_next_event; + gxp_timer->evt.cpumask = cpumask_of(0); + + irq = irq_of_parse_and_map(node, 0); + if (irq <= 0) { + ret = -EINVAL; + pr_err("GXP Timer Can't parse IRQ %d", irq); + goto err_exit; + } + + freq = clk_get_rate(clk); + + ret = clocksource_mmio_init(system_clock, node->name, freq, + 300, 32, clocksource_mmio_readl_up); + if (ret) { + pr_err("%pOFn init clocksource failed: %d", node, ret); + goto err_exit; + } + + sched_clock_register(gxp_sched_read, 32, freq); + + irq = irq_of_parse_and_map(node, 0); + if (irq <= 0) { + ret = -EINVAL; + pr_err("%pOFn Can't parse IRQ %d", node, irq); + goto err_exit; + } + + clockevents_config_and_register(&gxp_timer->evt, TIMER0_FREQ, + 0xf, 0xffffffff); + + ret = request_irq(irq, gxp_timer_interrupt, IRQF_TIMER | IRQF_SHARED, + node->name, gxp_timer); + if (ret) { + pr_err("%pOFn request_irq() failed: %d", node, ret); + goto err_exit; + } + + pr_debug("gxp: system timer (irq = %d)\n", irq); + return 0; + +err_exit: + iounmap(base); +err_iomap: + clk_disable_unprepare(clk); +err_clk_enable: + clk_put(clk); +err_free: + kfree(gxp_timer); + return ret; +} + +/* + * This probe gets called after the timer is already up and running. This will create + * the watchdog device as a child since the registers are shared. + */ + +static int gxp_timer_probe(struct platform_device *pdev) +{ + struct platform_device *gxp_watchdog_device; + struct device *dev = &pdev->dev; + int ret; + + if (!gxp_timer) { + pr_err("Gxp Timer not initialized, cannot create watchdog"); + return -ENOMEM; + } + + gxp_watchdog_device = platform_device_alloc("gxp-wdt", -1); + if (!gxp_watchdog_device) { + pr_err("Timer failed to allocate gxp-wdt"); + return -ENOMEM; + } + + /* Pass the base address (counter) as platform data and nothing else */ + gxp_watchdog_device->dev.platform_data = gxp_timer->counter; + gxp_watchdog_device->dev.parent = dev; + + ret = platform_device_add(gxp_watchdog_device); + if (ret) + platform_device_put(gxp_watchdog_device); + + return ret; +} + +static const struct of_device_id gxp_timer_of_match[] = { + { .compatible = "hpe,gxp-timer", }, + {}, +}; + +static struct platform_driver gxp_timer_driver = { + .probe = gxp_timer_probe, + .driver = { + .name = "gxp-timer", + .of_match_table = gxp_timer_of_match, + .suppress_bind_attrs = true, + }, +}; + +builtin_platform_driver(gxp_timer_driver); + +TIMER_OF_DECLARE(gxp, "hpe,gxp-timer", gxp_timer_init); -- cgit v1.3.1