From 1690d8bb91e370ab772062b79bd434ce815c4729 Mon Sep 17 00:00:00 2001 From: Viresh Kumar Date: Fri, 4 Jan 2019 15:14:33 +0530 Subject: cpufreq: scpi/scmi: Fix freeing of dynamic OPPs Since the commit 2a4eb7358aba "OPP: Don't remove dynamic OPPs from _dev_pm_opp_remove_table()", dynamically created OPP aren't automatically removed anymore by dev_pm_opp_cpumask_remove_table(). This affects the scpi and scmi cpufreq drivers which no longer free OPPs on failures or on invocations of the policy->exit() callback. Create a generic OPP helper dev_pm_opp_remove_all_dynamic() which can be called from these drivers instead of dev_pm_opp_cpumask_remove_table(). In dev_pm_opp_remove_all_dynamic(), we need to make sure that the opp_list isn't getting accessed simultaneously from other parts of the OPP core while the helper is freeing dynamic OPPs, i.e. we can't drop the opp_table->lock while traversing through the OPP list. And to accomplish that, this patch also creates _opp_kref_release_unlocked() which can be called from this new helper with the opp_table lock already held. Cc: 4.20 # v4.20 Reported-by: Valentin Schneider Fixes: 2a4eb7358aba "OPP: Don't remove dynamic OPPs from _dev_pm_opp_remove_table()" Signed-off-by: Viresh Kumar Tested-by: Valentin Schneider Reviewed-by: Sudeep Holla Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/scmi-cpufreq.c | 4 ++-- drivers/cpufreq/scpi-cpufreq.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/cpufreq') diff --git a/drivers/cpufreq/scmi-cpufreq.c b/drivers/cpufreq/scmi-cpufreq.c index 50b1551ba894..c2e66528f5ee 100644 --- a/drivers/cpufreq/scmi-cpufreq.c +++ b/drivers/cpufreq/scmi-cpufreq.c @@ -176,7 +176,7 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy) out_free_priv: kfree(priv); out_free_opp: - dev_pm_opp_cpumask_remove_table(policy->cpus); + dev_pm_opp_remove_all_dynamic(cpu_dev); return ret; } @@ -188,7 +188,7 @@ static int scmi_cpufreq_exit(struct cpufreq_policy *policy) cpufreq_cooling_unregister(priv->cdev); dev_pm_opp_free_cpufreq_table(priv->cpu_dev, &policy->freq_table); kfree(priv); - dev_pm_opp_cpumask_remove_table(policy->related_cpus); + dev_pm_opp_remove_all_dynamic(priv->cpu_dev); return 0; } diff --git a/drivers/cpufreq/scpi-cpufreq.c b/drivers/cpufreq/scpi-cpufreq.c index 87a98ec77773..99449738faa4 100644 --- a/drivers/cpufreq/scpi-cpufreq.c +++ b/drivers/cpufreq/scpi-cpufreq.c @@ -177,7 +177,7 @@ out_free_cpufreq_table: out_free_priv: kfree(priv); out_free_opp: - dev_pm_opp_cpumask_remove_table(policy->cpus); + dev_pm_opp_remove_all_dynamic(cpu_dev); return ret; } @@ -190,7 +190,7 @@ static int scpi_cpufreq_exit(struct cpufreq_policy *policy) clk_put(priv->clk); dev_pm_opp_free_cpufreq_table(priv->cpu_dev, &policy->freq_table); kfree(priv); - dev_pm_opp_cpumask_remove_table(policy->related_cpus); + dev_pm_opp_remove_all_dynamic(priv->cpu_dev); return 0; } -- cgit v1.2.3 From 2f66196208c98b3d1b4294edffb2c5a8197be899 Mon Sep 17 00:00:00 2001 From: Sudeep Holla Date: Mon, 7 Jan 2019 18:51:53 +0000 Subject: cpufreq: check if policy is inactive early in __cpufreq_get() cpuinfo_cur_freq gets current CPU frequency as detected by hardware while scaling_cur_freq last known CPU frequency. Some platforms may not allow checking the CPU frequency of an offline CPU or the associated resources may have been released via cpufreq_exit when the CPU gets offlined, in which case the policy would have been invalidated already. If we attempt to get current frequency from the hardware, it may result in hang or crash. For example on Juno, I see: Unable to handle kernel NULL pointer dereference at virtual address 0000000000000188 [0000000000000188] pgd=0000000000000000 Internal error: Oops: 96000004 [#1] PREEMPT SMP Modules linked in: CPU: 5 PID: 4202 Comm: cat Not tainted 4.20.0-08251-ga0f2c0318a15-dirty #87 Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform pstate: 40000005 (nZcv daif -PAN -UAO) pc : scmi_cpufreq_get_rate+0x34/0xb0 lr : scmi_cpufreq_get_rate+0x34/0xb0 Call trace: scmi_cpufreq_get_rate+0x34/0xb0 __cpufreq_get+0x34/0xc0 show_cpuinfo_cur_freq+0x24/0x78 show+0x40/0x60 sysfs_kf_seq_show+0xc0/0x148 kernfs_seq_show+0x44/0x50 seq_read+0xd4/0x480 kernfs_fop_read+0x15c/0x208 __vfs_read+0x60/0x188 vfs_read+0x94/0x150 ksys_read+0x6c/0xd8 __arm64_sys_read+0x24/0x30 el0_svc_common+0x78/0x100 el0_svc_handler+0x38/0x78 el0_svc+0x8/0xc ---[ end trace 3d1024e58f77f6b2 ]--- So fix the issue by checking if the policy is invalid early in __cpufreq_get before attempting to get the current frequency. Signed-off-by: Sudeep Holla Acked-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/cpufreq.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) (limited to 'drivers/cpufreq') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 7aa3dcad2175..df34a12a388f 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1530,17 +1530,16 @@ static unsigned int __cpufreq_get(struct cpufreq_policy *policy) { unsigned int ret_freq = 0; - if (!cpufreq_driver->get) + if (unlikely(policy_is_inactive(policy)) || !cpufreq_driver->get) return ret_freq; ret_freq = cpufreq_driver->get(policy->cpu); /* - * Updating inactive policies is invalid, so avoid doing that. Also - * if fast frequency switching is used with the given policy, the check + * If fast frequency switching is used with the given policy, the check * against policy->cur is pointless, so skip it in that case too. */ - if (unlikely(policy_is_inactive(policy)) || policy->fast_switch_enabled) + if (policy->fast_switch_enabled) return ret_freq; if (ret_freq && policy->cur && @@ -1569,10 +1568,7 @@ unsigned int cpufreq_get(unsigned int cpu) if (policy) { down_read(&policy->rwsem); - - if (!policy_is_inactive(policy)) - ret_freq = __cpufreq_get(policy); - + ret_freq = __cpufreq_get(policy); up_read(&policy->rwsem); cpufreq_cpu_put(policy); -- cgit v1.2.3 From 0e141d1c65c1dd31c914eb2e11651adcc1a15912 Mon Sep 17 00:00:00 2001 From: Quentin Perret Date: Wed, 9 Jan 2019 10:42:36 +0000 Subject: cpufreq: scmi: Fix frequency invariance in slow path The scmi-cpufreq driver calls the arch_set_freq_scale() callback on frequency changes to provide scale-invariant load-tracking signals to the scheduler. However, in the slow path, it does so while specifying the current and max frequencies in different units, hence resulting in a broken freq_scale factor. Fix this by passing all frequencies in KHz, as stored in the CPUFreq frequency table. Fixes: 99d6bdf33877 (cpufreq: add support for CPU DVFS based on SCMI message protocol) Signed-off-by: Quentin Perret Acked-by: Viresh Kumar Acked-by: Sudeep Holla Cc: 4.17+ # v4.17+ Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/scmi-cpufreq.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/cpufreq') diff --git a/drivers/cpufreq/scmi-cpufreq.c b/drivers/cpufreq/scmi-cpufreq.c index c2e66528f5ee..242c3370544e 100644 --- a/drivers/cpufreq/scmi-cpufreq.c +++ b/drivers/cpufreq/scmi-cpufreq.c @@ -52,9 +52,9 @@ scmi_cpufreq_set_target(struct cpufreq_policy *policy, unsigned int index) int ret; struct scmi_data *priv = policy->driver_data; struct scmi_perf_ops *perf_ops = handle->perf_ops; - u64 freq = policy->freq_table[index].frequency * 1000; + u64 freq = policy->freq_table[index].frequency; - ret = perf_ops->freq_set(handle, priv->domain_id, freq, false); + ret = perf_ops->freq_set(handle, priv->domain_id, freq * 1000, false); if (!ret) arch_set_freq_scale(policy->related_cpus, freq, policy->cpuinfo.max_freq); -- cgit v1.2.3 From f896d06665ec8ae09f6412a47b97863d9ef8ff64 Mon Sep 17 00:00:00 2001 From: Amit Kucheria Date: Thu, 10 Jan 2019 05:30:53 +0530 Subject: cpufreq: qcom-hw: Move to device_initcall subsys_initcall causes problems registering the driver as a thermal cooling device. If "faster boot" is the main reason for doing subsys_initcall, this should be handled in the bootloader or another boot constraint framework. Signed-off-by: Amit Kucheria Signed-off-by: Viresh Kumar --- drivers/cpufreq/qcom-cpufreq-hw.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/cpufreq') diff --git a/drivers/cpufreq/qcom-cpufreq-hw.c b/drivers/cpufreq/qcom-cpufreq-hw.c index d83939a1b3d4..649dddd72749 100644 --- a/drivers/cpufreq/qcom-cpufreq-hw.c +++ b/drivers/cpufreq/qcom-cpufreq-hw.c @@ -296,7 +296,7 @@ static int __init qcom_cpufreq_hw_init(void) { return platform_driver_register(&qcom_cpufreq_hw_driver); } -subsys_initcall(qcom_cpufreq_hw_init); +device_initcall(qcom_cpufreq_hw_init); static void __exit qcom_cpufreq_hw_exit(void) { -- cgit v1.2.3 From 55538fbc79e926f159a5abcc571ba1c02529a3d1 Mon Sep 17 00:00:00 2001 From: Taniya Das Date: Thu, 31 Jan 2019 23:02:50 +0530 Subject: cpufreq: qcom: Read voltage LUT and populate OPP Add support to read the voltage look up table and populate OPP for all corresponding CPUS for consumers like the energy model could use the frequency and voltage from the OPP tables. Also update the logic to not add duplicate OPPs. Tested-by: Matthias Kaehlcke Reviewed-by: Matthias Kaehlcke Signed-off-by: Matthias Kaehlcke Signed-off-by: Taniya Das Signed-off-by: Viresh Kumar --- drivers/cpufreq/qcom-cpufreq-hw.c | 46 +++++++++++++++++++++++++++++++-------- 1 file changed, 37 insertions(+), 9 deletions(-) (limited to 'drivers/cpufreq') diff --git a/drivers/cpufreq/qcom-cpufreq-hw.c b/drivers/cpufreq/qcom-cpufreq-hw.c index 649dddd72749..72223536ca4e 100644 --- a/drivers/cpufreq/qcom-cpufreq-hw.c +++ b/drivers/cpufreq/qcom-cpufreq-hw.c @@ -10,18 +10,21 @@ #include #include #include +#include #include #define LUT_MAX_ENTRIES 40U #define LUT_SRC GENMASK(31, 30) #define LUT_L_VAL GENMASK(7, 0) #define LUT_CORE_COUNT GENMASK(18, 16) +#define LUT_VOLT GENMASK(11, 0) #define LUT_ROW_SIZE 32 #define CLK_HW_DIV 2 /* Register offsets */ #define REG_ENABLE 0x0 -#define REG_LUT_TABLE 0x110 +#define REG_FREQ_LUT 0x110 +#define REG_VOLT_LUT 0x114 #define REG_PERF_STATE 0x920 static unsigned long cpu_hw_rate, xo_rate; @@ -70,11 +73,12 @@ static unsigned int qcom_cpufreq_hw_fast_switch(struct cpufreq_policy *policy, return policy->freq_table[index].frequency; } -static int qcom_cpufreq_hw_read_lut(struct device *dev, +static int qcom_cpufreq_hw_read_lut(struct device *cpu_dev, struct cpufreq_policy *policy, void __iomem *base) { u32 data, src, lval, i, core_count, prev_cc = 0, prev_freq = 0, freq; + u32 volt; unsigned int max_cores = cpumask_weight(policy->cpus); struct cpufreq_frequency_table *table; @@ -83,23 +87,28 @@ static int qcom_cpufreq_hw_read_lut(struct device *dev, return -ENOMEM; for (i = 0; i < LUT_MAX_ENTRIES; i++) { - data = readl_relaxed(base + REG_LUT_TABLE + i * LUT_ROW_SIZE); + data = readl_relaxed(base + REG_FREQ_LUT + + i * LUT_ROW_SIZE); src = FIELD_GET(LUT_SRC, data); lval = FIELD_GET(LUT_L_VAL, data); core_count = FIELD_GET(LUT_CORE_COUNT, data); + data = readl_relaxed(base + REG_VOLT_LUT + + i * LUT_ROW_SIZE); + volt = FIELD_GET(LUT_VOLT, data) * 1000; + if (src) freq = xo_rate * lval / 1000; else freq = cpu_hw_rate / 1000; - /* Ignore boosts in the middle of the table */ - if (core_count != max_cores) { - table[i].frequency = CPUFREQ_ENTRY_INVALID; - } else { + if (freq != prev_freq && core_count == max_cores) { table[i].frequency = freq; - dev_dbg(dev, "index=%d freq=%d, core_count %d\n", i, + dev_pm_opp_add(cpu_dev, freq * 1000, volt); + dev_dbg(cpu_dev, "index=%d freq=%d, core_count %d\n", i, freq, core_count); + } else { + table[i].frequency = CPUFREQ_ENTRY_INVALID; } /* @@ -116,6 +125,7 @@ static int qcom_cpufreq_hw_read_lut(struct device *dev, if (prev_cc != max_cores) { prev->frequency = prev_freq; prev->flags = CPUFREQ_BOOST_FREQ; + dev_pm_opp_add(cpu_dev, prev_freq * 1000, volt); } break; @@ -127,6 +137,7 @@ static int qcom_cpufreq_hw_read_lut(struct device *dev, table[i].frequency = CPUFREQ_TABLE_END; policy->freq_table = table; + dev_pm_opp_set_sharing_cpus(cpu_dev, policy->cpus); return 0; } @@ -159,10 +170,18 @@ static int qcom_cpufreq_hw_cpu_init(struct cpufreq_policy *policy) struct device *dev = &global_pdev->dev; struct of_phandle_args args; struct device_node *cpu_np; + struct device *cpu_dev; struct resource *res; void __iomem *base; int ret, index; + cpu_dev = get_cpu_device(policy->cpu); + if (!cpu_dev) { + pr_err("%s: failed to get cpu%d device\n", __func__, + policy->cpu); + return -ENODEV; + } + cpu_np = of_cpu_device_node_get(policy->cpu); if (!cpu_np) return -EINVAL; @@ -199,12 +218,19 @@ static int qcom_cpufreq_hw_cpu_init(struct cpufreq_policy *policy) policy->driver_data = base + REG_PERF_STATE; - ret = qcom_cpufreq_hw_read_lut(dev, policy, base); + ret = qcom_cpufreq_hw_read_lut(cpu_dev, policy, base); if (ret) { dev_err(dev, "Domain-%d failed to read LUT\n", index); goto error; } + ret = dev_pm_opp_get_opp_count(cpu_dev); + if (ret <= 0) { + dev_err(cpu_dev, "Failed to add OPPs\n"); + ret = -ENODEV; + goto error; + } + policy->fast_switch_possible = true; return 0; @@ -215,8 +241,10 @@ error: static int qcom_cpufreq_hw_cpu_exit(struct cpufreq_policy *policy) { + struct device *cpu_dev = get_cpu_device(policy->cpu); void __iomem *base = policy->driver_data - REG_PERF_STATE; + dev_pm_opp_remove_all_dynamic(cpu_dev); kfree(policy->freq_table); devm_iounmap(&global_pdev->dev, base); -- cgit v1.2.3 From dab535052f67db0ff48b1b23e714b58650d1a787 Mon Sep 17 00:00:00 2001 From: Matthias Kaehlcke Date: Tue, 5 Feb 2019 09:52:24 -0800 Subject: cpufreq: qcom-hw: Register an Energy Model Try and register an Energy Model from qcom-cpufreq-hw to allow interested sub-systems like the task scheduler to use the provided information. Signed-off-by: Matthias Kaehlcke [ Viresh: Rebased over cpufreq related changes ] Signed-off-by: Viresh Kumar --- drivers/cpufreq/qcom-cpufreq-hw.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/cpufreq') diff --git a/drivers/cpufreq/qcom-cpufreq-hw.c b/drivers/cpufreq/qcom-cpufreq-hw.c index 72223536ca4e..c647cfb1f5e6 100644 --- a/drivers/cpufreq/qcom-cpufreq-hw.c +++ b/drivers/cpufreq/qcom-cpufreq-hw.c @@ -231,6 +231,8 @@ static int qcom_cpufreq_hw_cpu_init(struct cpufreq_policy *policy) goto error; } + dev_pm_opp_of_register_em(policy->cpus); + policy->fast_switch_possible = true; return 0; -- cgit v1.2.3