| Message ID | 20210727152512.1098329-4-thara.gopinath@linaro.org (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
| Series | Introduce LMh driver for Qualcomm SoCs | expand |
On 27-07-21, 11:25, Thara Gopinath wrote: > +static void qcom_lmh_dcvs_notify(struct qcom_cpufreq_data *data) > +{ > + /* In the unlikely case cpufreq is de-registered do not enable polling or h/w interrupt */ > + > + spin_lock(&data->throttle_lock); > + if (data->cancel_throttle) { > + spin_unlock(&data->throttle_lock); > + return; > + } > + spin_unlock(&data->throttle_lock); > + > + /* > + * If h/w throttled frequency is higher than what cpufreq has requested for, stop > + * polling and switch back to interrupt mechanism > + */ > + > + if (throttled_freq >= qcom_cpufreq_hw_get(cpumask_first(policy->cpus))) > + /* Clear the existing interrupts and enable it back */ > + enable_irq(data->throttle_irq); > + else > + mod_delayed_work(system_highpri_wq, &data->throttle_work, > + msecs_to_jiffies(10)); > +} > +static void qcom_cpufreq_hw_lmh_exit(struct qcom_cpufreq_data *data) > +{ > + if (data->throttle_irq <= 0) > + return; > + > + spin_lock(&data->throttle_lock); > + data->cancel_throttle = true; > + spin_unlock(&data->throttle_lock); > + cancel_delayed_work_sync(&data->throttle_work); > + free_irq(data->throttle_irq, data); > +} Lets see if we can still make it break :) CPU0 CPU1 qcom_lmh_dcvs_notify() qcom_cpufreq_hw_lmh_exit() spin_unlock() spin_lock(), cancel_throttle = true spin_unlock() cancel_delayed_work_sync() mod_delayed_work() free_irq() kfree(data) qcom_lmh_dcvs_poll() Uses data. Sorry, locking is fun :)
On 7/27/21 11:50 PM, Viresh Kumar wrote: > On 27-07-21, 11:25, Thara Gopinath wrote: >> +static void qcom_lmh_dcvs_notify(struct qcom_cpufreq_data *data) >> +{ > >> + /* In the unlikely case cpufreq is de-registered do not enable polling or h/w interrupt */ >> + >> + spin_lock(&data->throttle_lock); >> + if (data->cancel_throttle) { >> + spin_unlock(&data->throttle_lock); >> + return; >> + } >> + spin_unlock(&data->throttle_lock); >> + >> + /* >> + * If h/w throttled frequency is higher than what cpufreq has requested for, stop >> + * polling and switch back to interrupt mechanism >> + */ >> + >> + if (throttled_freq >= qcom_cpufreq_hw_get(cpumask_first(policy->cpus))) >> + /* Clear the existing interrupts and enable it back */ >> + enable_irq(data->throttle_irq); >> + else >> + mod_delayed_work(system_highpri_wq, &data->throttle_work, >> + msecs_to_jiffies(10)); >> +} > >> +static void qcom_cpufreq_hw_lmh_exit(struct qcom_cpufreq_data *data) >> +{ >> + if (data->throttle_irq <= 0) >> + return; >> + >> + spin_lock(&data->throttle_lock); >> + data->cancel_throttle = true; >> + spin_unlock(&data->throttle_lock); >> + cancel_delayed_work_sync(&data->throttle_work); >> + free_irq(data->throttle_irq, data); >> +} > > Lets see if we can still make it break :) > > CPU0 CPU1 > > qcom_lmh_dcvs_notify() qcom_cpufreq_hw_lmh_exit() > > spin_unlock() > spin_lock(), > cancel_throttle = true > spin_unlock() > > cancel_delayed_work_sync() > mod_delayed_work() > free_irq() > kfree(data) > qcom_lmh_dcvs_poll() > Uses data. > > > Sorry, locking is fun :) Ha! I was too lazy to write this down! So how about I make this a mutex and put mod_delayed_work() inside the lock. So it will be something like below qcom_lmh_dcvs_notify() qcom_cpufreq_hw_lmh_exit() mutex_lock() mutex_lock() if (data->cancel_throttle) { cancel_throttle = true mutex_unlock() mutex_unlock() return cancel_delayed_work_sync() } free_irq() enable_irq() / mod_delayed_work() mutex_unlock() I will let you break it! >
On 28-07-21, 18:19, Thara Gopinath wrote: > Ha! I was too lazy to write this down! So how about I make this a mutex and mutex may not work as you come here from irq. > put mod_delayed_work() inside the lock. So it will be something like below > > qcom_lmh_dcvs_notify() qcom_cpufreq_hw_lmh_exit() > > mutex_lock() mutex_lock() > if (data->cancel_throttle) { cancel_throttle = true > mutex_unlock() mutex_unlock() > return cancel_delayed_work_sync() > } free_irq() > enable_irq() / mod_delayed_work() > mutex_unlock() > > I will let you break it! I can't any further :) Consider merging below to this patch, it fixes sever other minor issues I see in the code.
On 7/29/21 2:17 AM, Viresh Kumar wrote: > On 28-07-21, 18:19, Thara Gopinath wrote: >> Ha! I was too lazy to write this down! So how about I make this a mutex and > > mutex may not work as you come here from irq. Hi! So the interrupt handler is a threaded handler. I moved it in v4 since one of the "_opp" api has an underlying mutex and was causing issues. So using a mutex should be pretty safe in this case. > >> put mod_delayed_work() inside the lock. So it will be something like below >> >> qcom_lmh_dcvs_notify() qcom_cpufreq_hw_lmh_exit() >> >> mutex_lock() mutex_lock() >> if (data->cancel_throttle) { cancel_throttle = true >> mutex_unlock() mutex_unlock() >> return cancel_delayed_work_sync() >> } free_irq() >> enable_irq() / mod_delayed_work() >> mutex_unlock() >> >> I will let you break it! > > I can't any further :) > > Consider merging below to this patch, it fixes sever other minor > issues I see in the code. IIUC, the main change you are suggesting below is to include enable_irq() / mod_delayed_work() under the spin_lock as well. Is that right ? In which case isn't a mutex better than spinlock? >
On 29-07-21, 07:13, Thara Gopinath wrote: > So the interrupt handler is a threaded handler. I moved it in v4 since one > of the "_opp" api has an underlying mutex and was causing issues. So using a > mutex should be pretty safe in this case. Ahh I see. > IIUC, the main change you are suggesting below is to include enable_irq() / > mod_delayed_work() under the spin_lock as well. Is that right ? In which > case isn't a mutex better than spinlock? Yeah, sure.
diff --git a/drivers/cpufreq/qcom-cpufreq-hw.c b/drivers/cpufreq/qcom-cpufreq-hw.c index f86859bf76f1..3cc27d9e2ed1 100644 --- a/drivers/cpufreq/qcom-cpufreq-hw.c +++ b/drivers/cpufreq/qcom-cpufreq-hw.c @@ -7,12 +7,14 @@ #include <linux/cpufreq.h> #include <linux/init.h> #include <linux/interconnect.h> +#include <linux/interrupt.h> #include <linux/kernel.h> #include <linux/module.h> #include <linux/of_address.h> #include <linux/of_platform.h> #include <linux/pm_opp.h> #include <linux/slab.h> +#include <linux/spinlock.h> #define LUT_MAX_ENTRIES 40U #define LUT_SRC GENMASK(31, 30) @@ -22,10 +24,13 @@ #define CLK_HW_DIV 2 #define LUT_TURBO_IND 1 +#define HZ_PER_KHZ 1000 + struct qcom_cpufreq_soc_data { u32 reg_enable; u32 reg_freq_lut; u32 reg_volt_lut; + u32 reg_current_vote; u32 reg_perf_state; u8 lut_row_size; }; @@ -33,7 +38,13 @@ struct qcom_cpufreq_soc_data { struct qcom_cpufreq_data { void __iomem *base; struct resource *res; + struct delayed_work throttle_work; const struct qcom_cpufreq_soc_data *soc_data; + struct cpufreq_policy *policy; + /* Lock to synchronize between de-init sequence and re-starting LMh polling/interrupts */ + spinlock_t throttle_lock; + int throttle_irq; + bool cancel_throttle; }; static unsigned long cpu_hw_rate, xo_rate; @@ -251,10 +262,92 @@ static void qcom_get_related_cpus(int index, struct cpumask *m) } } +static unsigned int qcom_lmh_get_throttle_freq(struct qcom_cpufreq_data *data) +{ + unsigned int val = readl_relaxed(data->base + data->soc_data->reg_current_vote); + + return (val & 0x3FF) * 19200; +} + +static void qcom_lmh_dcvs_notify(struct qcom_cpufreq_data *data) +{ + struct cpufreq_policy *policy = data->policy; + struct dev_pm_opp *opp; + struct device *dev; + unsigned long max_capacity, capacity, freq_hz, throttled_freq; + unsigned int freq; + + /* + * Get the h/w throttled frequency, normalize it using the + * registered opp table and use it to calculate thermal pressure. + */ + freq = qcom_lmh_get_throttle_freq(data); + freq_hz = freq * HZ_PER_KHZ; + + dev = get_cpu_device(cpumask_first(policy->cpus)); + opp = dev_pm_opp_find_freq_floor(dev, &freq_hz); + if (IS_ERR(opp) && PTR_ERR(opp) == -ERANGE) + opp = dev_pm_opp_find_freq_ceil(dev, &freq_hz); + + throttled_freq = freq_hz / HZ_PER_KHZ; + + /* Update thermal pressure */ + + max_capacity = arch_scale_cpu_capacity(cpumask_first(policy->cpus)); + capacity = mult_frac(max_capacity, throttled_freq, policy->cpuinfo.max_freq); + /* Don't pass boost capacity to scheduler */ + if (capacity > max_capacity) + capacity = max_capacity; + + arch_set_thermal_pressure(policy->cpus, max_capacity - capacity); + + /* In the unlikely case cpufreq is de-registered do not enable polling or h/w interrupt */ + + spin_lock(&data->throttle_lock); + if (data->cancel_throttle) { + spin_unlock(&data->throttle_lock); + return; + } + spin_unlock(&data->throttle_lock); + + /* + * If h/w throttled frequency is higher than what cpufreq has requested for, stop + * polling and switch back to interrupt mechanism + */ + + if (throttled_freq >= qcom_cpufreq_hw_get(cpumask_first(policy->cpus))) + /* Clear the existing interrupts and enable it back */ + enable_irq(data->throttle_irq); + else + mod_delayed_work(system_highpri_wq, &data->throttle_work, + msecs_to_jiffies(10)); +} + +static void qcom_lmh_dcvs_poll(struct work_struct *work) +{ + struct qcom_cpufreq_data *data; + + data = container_of(work, struct qcom_cpufreq_data, throttle_work.work); + + qcom_lmh_dcvs_notify(data); +} + +static irqreturn_t qcom_lmh_dcvs_handle_irq(int irq, void *data) +{ + struct qcom_cpufreq_data *c_data = data; + + /* Disable interrupt and enable polling */ + disable_irq_nosync(c_data->throttle_irq); + qcom_lmh_dcvs_notify(c_data); + + return 0; +} + static const struct qcom_cpufreq_soc_data qcom_soc_data = { .reg_enable = 0x0, .reg_freq_lut = 0x110, .reg_volt_lut = 0x114, + .reg_current_vote = 0x704, .reg_perf_state = 0x920, .lut_row_size = 32, }; @@ -274,6 +367,50 @@ static const struct of_device_id qcom_cpufreq_hw_match[] = { }; MODULE_DEVICE_TABLE(of, qcom_cpufreq_hw_match); +static int qcom_cpufreq_hw_lmh_init(struct cpufreq_policy *policy, int index) +{ + struct qcom_cpufreq_data *data = policy->driver_data; + struct platform_device *pdev = cpufreq_get_driver_data(); + char irq_name[15]; + int ret; + + /* + * Look for LMh interrupt. If no interrupt line is specified / + * if there is an error, allow cpufreq to be enabled as usual. + */ + data->throttle_irq = platform_get_irq(pdev, index); + if (data->throttle_irq <= 0) + return data->throttle_irq == -EPROBE_DEFER ? -EPROBE_DEFER : 0; + + data->cancel_throttle = false; + data->policy = policy; + + spin_lock_init(&data->throttle_lock); + INIT_DEFERRABLE_WORK(&data->throttle_work, qcom_lmh_dcvs_poll); + + snprintf(irq_name, sizeof(irq_name), "dcvsh-irq-%u", policy->cpu); + ret = request_threaded_irq(data->throttle_irq, NULL, qcom_lmh_dcvs_handle_irq, + IRQF_ONESHOT, irq_name, data); + if (ret) { + dev_err(&pdev->dev, "Error registering %s: %d\n", irq_name, ret); + return 0; + } + + return 0; +} + +static void qcom_cpufreq_hw_lmh_exit(struct qcom_cpufreq_data *data) +{ + if (data->throttle_irq <= 0) + return; + + spin_lock(&data->throttle_lock); + data->cancel_throttle = true; + spin_unlock(&data->throttle_lock); + cancel_delayed_work_sync(&data->throttle_work); + free_irq(data->throttle_irq, data); +} + static int qcom_cpufreq_hw_cpu_init(struct cpufreq_policy *policy) { struct platform_device *pdev = cpufreq_get_driver_data(); @@ -370,6 +507,10 @@ static int qcom_cpufreq_hw_cpu_init(struct cpufreq_policy *policy) dev_warn(cpu_dev, "failed to enable boost: %d\n", ret); } + ret = qcom_cpufreq_hw_lmh_init(policy, index); + if (ret) + goto error; + return 0; error: kfree(data); @@ -389,6 +530,7 @@ static int qcom_cpufreq_hw_cpu_exit(struct cpufreq_policy *policy) dev_pm_opp_remove_all_dynamic(cpu_dev); dev_pm_opp_of_cpumask_remove_table(policy->related_cpus); + qcom_cpufreq_hw_lmh_exit(data); kfree(policy->freq_table); kfree(data); iounmap(base);
Add interrupt support to notify the kernel of h/w initiated frequency throttling by LMh. Convey this to scheduler via thermal presssure interface. Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org> --- v3->v4: - Dropped devm_ versions of request_irq and free_irq as per comments on mailing list. - Introduced locking to prevent race between LMh de-init sequence and re-enabling of interrupts/polling after a thermal throttle event. - Moved the LMh de-init sequence to qcom_cpufreq_hw_lmh_exit as per Viresh's review comments - Code rearrangement as per Bjorn's review comments. - Moved the interrupt handling to threaded interrupt handling since Steev reported some scheduling while atomic bug on the mailing list. v2->v3: - Cosmetic fixes from review comments on the list. - Moved all LMh initializations to qcom_cpufreq_hw_lmh_init. - Added freeing of LMh interrupt and cancelling the polling worker to qcom_cpufreq_hw_cpu_exit as per Viresh's suggestion. - LMh interrupts are now tied to cpu dev and not cpufreq dev. This will be useful for further generation of SoCs where the same interrupt signals multiple cpu clusters. v1->v2: - Introduced qcom_cpufreq_hw_lmh_init to consolidate LMh related initializations as per Viresh's review comment. - Moved the piece of code restarting polling/re-enabling LMh interrupt to qcom_lmh_dcvs_notify therby simplifying isr and timer callback as per Viresh's suggestion. - Droped cpus from qcom_cpufreq_data and instead using cpus from cpufreq_policy in qcom_lmh_dcvs_notify as per Viresh's review comment. - Dropped dt property qcom,support-lmh as per Bjorn's suggestion. - Other minor/cosmetic fixes drivers/cpufreq/qcom-cpufreq-hw.c | 142 ++++++++++++++++++++++++++++++ 1 file changed, 142 insertions(+)