From 97e981324d492340e33baa9680780046377b561f Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Thu, 6 Dec 2018 00:17:50 -0800 Subject: ARCv2: boot log: BPU return stack depth Signed-off-by: Vineet Gupta --- arch/arc/kernel/setup.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/setup.c b/arch/arc/kernel/setup.c index 2e018b8c2e19..33e232145d83 100644 --- a/arch/arc/kernel/setup.c +++ b/arch/arc/kernel/setup.c @@ -195,6 +195,7 @@ static void read_arc_build_cfg_regs(void) cpu->bpu.full = bpu.ft; cpu->bpu.num_cache = 256 << bpu.bce; cpu->bpu.num_pred = 2048 << bpu.pte; + cpu->bpu.ret_stk = 4 << bpu.rse; if (cpu->core.family >= 0x54) { unsigned int exec_ctrl; @@ -299,10 +300,10 @@ static char *arc_cpu_mumbojumbo(int cpu_id, char *buf, int len) if (cpu->bpu.ver) n += scnprintf(buf + n, len - n, - "BPU\t\t: %s%s match, cache:%d, Predict Table:%d", + "BPU\t\t: %s%s match, cache:%d, Predict Table:%d Return stk: %d", IS_AVAIL1(cpu->bpu.full, "full"), IS_AVAIL1(!cpu->bpu.full, "partial"), - cpu->bpu.num_cache, cpu->bpu.num_pred); + cpu->bpu.num_cache, cpu->bpu.num_pred, cpu->bpu.ret_stk); if (is_isa_arcv2()) { struct bcr_lpb lpb; -- cgit v1.2.3 From 7dd380c338f1ec20ed46607ccd03541a7683cd67 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Wed, 9 Jan 2019 09:36:00 -0800 Subject: ARC: boot log: print Action point details This now prints the number of action points {2,4,8} and {min,full} targets supported. Signed-off-by: Vineet Gupta --- arch/arc/include/asm/arcregs.h | 10 +++++++++- arch/arc/kernel/setup.c | 22 +++++++++++++++------- 2 files changed, 24 insertions(+), 8 deletions(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/include/asm/arcregs.h b/arch/arc/include/asm/arcregs.h index 9f10d32ee1bd..f1b86cef0905 100644 --- a/arch/arc/include/asm/arcregs.h +++ b/arch/arc/include/asm/arcregs.h @@ -216,6 +216,14 @@ struct bcr_fp_arcv2 { #endif }; +struct bcr_actionpoint { +#ifdef CONFIG_CPU_BIG_ENDIAN + unsigned int pad:21, min:1, num:2, ver:8; +#else + unsigned int ver:8, num:2, min:1, pad:21; +#endif +}; + #include struct bcr_bpu_arcompact { @@ -302,7 +310,7 @@ struct cpuinfo_arc { struct { unsigned int swap:1, norm:1, minmax:1, barrel:1, crc:1, swape:1, pad1:2, fpu_sp:1, fpu_dp:1, dual:1, dual_enb:1, pad2:4, - debug:1, ap:1, smart:1, rtt:1, pad3:4, + ap_num:4, ap_full:1, smart:1, rtt:1, pad3:1, timer0:1, timer1:1, rtc:1, gfrc:1, pad4:4; } extn; struct bcr_mpy extn_mpy; diff --git a/arch/arc/kernel/setup.c b/arch/arc/kernel/setup.c index 33e232145d83..feb90093e6b1 100644 --- a/arch/arc/kernel/setup.c +++ b/arch/arc/kernel/setup.c @@ -123,6 +123,7 @@ static void read_arc_build_cfg_regs(void) struct cpuinfo_arc *cpu = &cpuinfo_arc700[smp_processor_id()]; const struct id_to_str *tbl; struct bcr_isa_arcv2 isa; + struct bcr_actionpoint ap; FIX_PTR(cpu); @@ -208,8 +209,11 @@ static void read_arc_build_cfg_regs(void) } } - READ_BCR(ARC_REG_AP_BCR, bcr); - cpu->extn.ap = bcr.ver ? 1 : 0; + READ_BCR(ARC_REG_AP_BCR, ap); + if (ap.ver) { + cpu->extn.ap_num = 2 << ap.num; + cpu->extn.ap_full = !!ap.min; + } READ_BCR(ARC_REG_SMART_BCR, bcr); cpu->extn.smart = bcr.ver ? 1 : 0; @@ -217,8 +221,6 @@ static void read_arc_build_cfg_regs(void) READ_BCR(ARC_REG_RTT_BCR, bcr); cpu->extn.rtt = bcr.ver ? 1 : 0; - cpu->extn.debug = cpu->extn.ap | cpu->extn.smart | cpu->extn.rtt; - READ_BCR(ARC_REG_ISA_CFG_BCR, isa); /* some hacks for lack of feature BCR info in old ARC700 cores */ @@ -337,11 +339,17 @@ static char *arc_extn_mumbojumbo(int cpu_id, char *buf, int len) IS_AVAIL1(cpu->extn.fpu_sp, "SP "), IS_AVAIL1(cpu->extn.fpu_dp, "DP ")); - if (cpu->extn.debug) - n += scnprintf(buf + n, len - n, "DEBUG\t\t: %s%s%s\n", - IS_AVAIL1(cpu->extn.ap, "ActionPoint "), + if (cpu->extn.ap_num | cpu->extn.smart | cpu->extn.rtt) { + n += scnprintf(buf + n, len - n, "DEBUG\t\t: %s%s", IS_AVAIL1(cpu->extn.smart, "smaRT "), IS_AVAIL1(cpu->extn.rtt, "RTT ")); + if (cpu->extn.ap_num) { + n += scnprintf(buf + n, len - n, "ActionPoint %d/%s", + cpu->extn.ap_num, + cpu->extn.ap_full ? "full":"min"); + } + n += scnprintf(buf + n, len - n, "\n"); + } if (cpu->dccm.sz || cpu->iccm.sz) n += scnprintf(buf + n, len - n, "Extn [CCM]\t: DCCM @ %x, %d KB / ICCM: @ %x, %d KB\n", -- cgit v1.2.3 From 14f81a91ad29a57d6c3cf123ad9489f2ca9133fb Mon Sep 17 00:00:00 2001 From: Eugeniy Paltsev Date: Thu, 13 Dec 2018 19:56:18 +0300 Subject: ARC: perf: trivial code cleanup * Use BIT(), lower_32_bits(), upper_32_bits() macroses, fix code style violations. * Use u32, u64, s64 instead of uint32_t, uint64_t, int64_t * Fix description comment as this code doesn't belong only to ARC700 anymore. * Use SPDX License Identifier. * Remove useless ifdefs. ifdef around 'arc_pmu_match' structure declaration is useless as we refer to 'arc_pmu_match' in several places which aren't guarded with ifdef. Nevertheless 'ARC' option selects 'OF' unconditionally so we can simply get rid of this ifdef. Acked-by: Vineet Gupta Signed-off-by: Eugeniy Paltsev Signed-off-by: Vineet Gupta --- arch/arc/kernel/perf_event.c | 85 ++++++++++++++++++++++---------------------- 1 file changed, 42 insertions(+), 43 deletions(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/perf_event.c b/arch/arc/kernel/perf_event.c index 8aec462d90fb..693f32d60c35 100644 --- a/arch/arc/kernel/perf_event.c +++ b/arch/arc/kernel/perf_event.c @@ -1,15 +1,10 @@ -/* - * Linux performance counter support for ARC700 series - * - * Copyright (C) 2013-2015 Synopsys, Inc. (www.synopsys.com) - * - * This code is inspired by the perf support of various other architectures. - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License version 2 as - * published by the Free Software Foundation. - * - */ +// SPDX-License-Identifier: GPL-2.0+ +// +// Linux performance counter support for ARC CPUs. +// This code is inspired by the perf support of various other architectures. +// +// Copyright (C) 2013-2018 Synopsys, Inc. (www.synopsys.com) + #include #include #include @@ -19,6 +14,9 @@ #include #include +/* HW holds 8 symbols + one for null terminator */ +#define ARCPMU_EVENT_NAME_LEN 9 + struct arc_pmu { struct pmu pmu; unsigned int irq; @@ -49,6 +47,7 @@ static int callchain_trace(unsigned int addr, void *data) { struct arc_callchain_trace *ctrl = data; struct perf_callchain_entry_ctx *entry = ctrl->perf_stuff; + perf_callchain_store(entry, addr); if (ctrl->depth++ < 3) @@ -57,8 +56,8 @@ static int callchain_trace(unsigned int addr, void *data) return -1; } -void -perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs) +void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, + struct pt_regs *regs) { struct arc_callchain_trace ctrl = { .depth = 0, @@ -68,8 +67,8 @@ perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *re arc_unwind_core(NULL, regs, callchain_trace, &ctrl); } -void -perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs) +void perf_callchain_user(struct perf_callchain_entry_ctx *entry, + struct pt_regs *regs) { /* * User stack can't be unwound trivially with kernel dwarf unwinder @@ -82,10 +81,10 @@ static struct arc_pmu *arc_pmu; static DEFINE_PER_CPU(struct arc_pmu_cpu, arc_pmu_cpu); /* read counter #idx; note that counter# != event# on ARC! */ -static uint64_t arc_pmu_read_counter(int idx) +static u64 arc_pmu_read_counter(int idx) { - uint32_t tmp; - uint64_t result; + u32 tmp; + u64 result; /* * ARC supports making 'snapshots' of the counters, so we don't @@ -94,7 +93,7 @@ static uint64_t arc_pmu_read_counter(int idx) write_aux_reg(ARC_REG_PCT_INDEX, idx); tmp = read_aux_reg(ARC_REG_PCT_CONTROL); write_aux_reg(ARC_REG_PCT_CONTROL, tmp | ARC_REG_PCT_CONTROL_SN); - result = (uint64_t) (read_aux_reg(ARC_REG_PCT_SNAPH)) << 32; + result = (u64) (read_aux_reg(ARC_REG_PCT_SNAPH)) << 32; result |= read_aux_reg(ARC_REG_PCT_SNAPL); return result; @@ -103,9 +102,9 @@ static uint64_t arc_pmu_read_counter(int idx) static void arc_perf_event_update(struct perf_event *event, struct hw_perf_event *hwc, int idx) { - uint64_t prev_raw_count = local64_read(&hwc->prev_count); - uint64_t new_raw_count = arc_pmu_read_counter(idx); - int64_t delta = new_raw_count - prev_raw_count; + u64 prev_raw_count = local64_read(&hwc->prev_count); + u64 new_raw_count = arc_pmu_read_counter(idx); + s64 delta = new_raw_count - prev_raw_count; /* * We aren't afraid of hwc->prev_count changing beneath our feet @@ -155,7 +154,7 @@ static int arc_pmu_event_init(struct perf_event *event) int ret; if (!is_sampling_event(event)) { - hwc->sample_period = arc_pmu->max_period; + hwc->sample_period = arc_pmu->max_period; hwc->last_period = hwc->sample_period; local64_set(&hwc->period_left, hwc->sample_period); } @@ -192,6 +191,7 @@ static int arc_pmu_event_init(struct perf_event *event) pr_debug("init cache event with h/w %08x \'%s\'\n", (int)hwc->config, arc_pmu_ev_hw_map[ret]); return 0; + default: return -ENOENT; } @@ -200,7 +200,7 @@ static int arc_pmu_event_init(struct perf_event *event) /* starts all counters */ static void arc_pmu_enable(struct pmu *pmu) { - uint32_t tmp; + u32 tmp; tmp = read_aux_reg(ARC_REG_PCT_CONTROL); write_aux_reg(ARC_REG_PCT_CONTROL, (tmp & 0xffff0000) | 0x1); } @@ -208,7 +208,7 @@ static void arc_pmu_enable(struct pmu *pmu) /* stops all counters */ static void arc_pmu_disable(struct pmu *pmu) { - uint32_t tmp; + u32 tmp; tmp = read_aux_reg(ARC_REG_PCT_CONTROL); write_aux_reg(ARC_REG_PCT_CONTROL, (tmp & 0xffff0000) | 0x0); } @@ -228,7 +228,7 @@ static int arc_pmu_event_set_period(struct perf_event *event) local64_set(&hwc->period_left, left); hwc->last_period = period; overflow = 1; - } else if (unlikely(left <= 0)) { + } else if (unlikely(left <= 0)) { /* left underflowed by less than period. */ left += period; local64_set(&hwc->period_left, left); @@ -246,8 +246,8 @@ static int arc_pmu_event_set_period(struct perf_event *event) write_aux_reg(ARC_REG_PCT_INDEX, idx); /* Write value */ - write_aux_reg(ARC_REG_PCT_COUNTL, (u32)value); - write_aux_reg(ARC_REG_PCT_COUNTH, (value >> 32)); + write_aux_reg(ARC_REG_PCT_COUNTL, lower_32_bits(value)); + write_aux_reg(ARC_REG_PCT_COUNTH, upper_32_bits(value)); perf_event_update_userpage(event); @@ -277,7 +277,7 @@ static void arc_pmu_start(struct perf_event *event, int flags) /* Enable interrupt for this counter */ if (is_sampling_event(event)) write_aux_reg(ARC_REG_PCT_INT_CTRL, - read_aux_reg(ARC_REG_PCT_INT_CTRL) | (1 << idx)); + read_aux_reg(ARC_REG_PCT_INT_CTRL) | BIT(idx)); /* enable ARC pmu here */ write_aux_reg(ARC_REG_PCT_INDEX, idx); /* counter # */ @@ -295,9 +295,9 @@ static void arc_pmu_stop(struct perf_event *event, int flags) * Reset interrupt flag by writing of 1. This is required * to make sure pending interrupt was not left. */ - write_aux_reg(ARC_REG_PCT_INT_ACT, 1 << idx); + write_aux_reg(ARC_REG_PCT_INT_ACT, BIT(idx)); write_aux_reg(ARC_REG_PCT_INT_CTRL, - read_aux_reg(ARC_REG_PCT_INT_CTRL) & ~(1 << idx)); + read_aux_reg(ARC_REG_PCT_INT_CTRL) & ~BIT(idx)); } if (!(event->hw.state & PERF_HES_STOPPED)) { @@ -349,9 +349,10 @@ static int arc_pmu_add(struct perf_event *event, int flags) if (is_sampling_event(event)) { /* Mimic full counter overflow as other arches do */ - write_aux_reg(ARC_REG_PCT_INT_CNTL, (u32)arc_pmu->max_period); + write_aux_reg(ARC_REG_PCT_INT_CNTL, + lower_32_bits(arc_pmu->max_period)); write_aux_reg(ARC_REG_PCT_INT_CNTH, - (arc_pmu->max_period >> 32)); + upper_32_bits(arc_pmu->max_period)); } write_aux_reg(ARC_REG_PCT_CONFIG, 0); @@ -392,7 +393,7 @@ static irqreturn_t arc_pmu_intr(int irq, void *dev) idx = __ffs(active_ints); /* Reset interrupt flag by writing of 1 */ - write_aux_reg(ARC_REG_PCT_INT_ACT, 1 << idx); + write_aux_reg(ARC_REG_PCT_INT_ACT, BIT(idx)); /* * On reset of "interrupt active" bit corresponding @@ -400,7 +401,7 @@ static irqreturn_t arc_pmu_intr(int irq, void *dev) * Now we need to re-enable interrupt for the counter. */ write_aux_reg(ARC_REG_PCT_INT_CTRL, - read_aux_reg(ARC_REG_PCT_INT_CTRL) | (1 << idx)); + read_aux_reg(ARC_REG_PCT_INT_CTRL) | BIT(idx)); event = pmu_cpu->act_counter[idx]; hwc = &event->hw; @@ -414,7 +415,7 @@ static irqreturn_t arc_pmu_intr(int irq, void *dev) arc_pmu_stop(event, 0); } - active_ints &= ~(1U << idx); + active_ints &= ~BIT(idx); } while (active_ints); done: @@ -450,10 +451,10 @@ static int arc_pmu_device_probe(struct platform_device *pdev) union cc_name { struct { - uint32_t word0, word1; + u32 word0, word1; char sentinel; } indiv; - char str[9]; + char str[ARCPMU_EVENT_NAME_LEN]; } cc_name; @@ -481,9 +482,9 @@ static int arc_pmu_device_probe(struct platform_device *pdev) pr_info("ARC perf\t: %d counters (%d bits), %d conditions%s\n", arc_pmu->n_counters, counter_size, cc_bcr.c, - has_interrupts ? ", [overflow IRQ support]":""); + has_interrupts ? ", [overflow IRQ support]" : ""); - cc_name.str[8] = 0; + cc_name.str[ARCPMU_EVENT_NAME_LEN - 1] = 0; for (i = 0; i < PERF_COUNT_ARC_HW_MAX; i++) arc_pmu->ev_hw_idx[i] = -1; @@ -538,14 +539,12 @@ static int arc_pmu_device_probe(struct platform_device *pdev) return perf_pmu_register(&arc_pmu->pmu, pdev->name, PERF_TYPE_RAW); } -#ifdef CONFIG_OF static const struct of_device_id arc_pmu_match[] = { { .compatible = "snps,arc700-pct" }, { .compatible = "snps,archs-pct" }, {}, }; MODULE_DEVICE_TABLE(of, arc_pmu_match); -#endif static struct platform_driver arc_pmu_driver = { .driver = { -- cgit v1.2.3 From 0e956150fe09fa4430c42a9bbe48a72967fa0012 Mon Sep 17 00:00:00 2001 From: Eugeniy Paltsev Date: Thu, 13 Dec 2018 19:56:19 +0300 Subject: ARC: perf: introduce Kernel PMU events support Export all available ARC architected hardware events as kernel PMU events to make non-generic events accessible. ARC PMU HW allow us to read the list of all available events names. So we generate kernel PMU event list dynamically in arc_pmu_device_probe() using human-readable events names we got from HW instead of using pre-defined events list. -------------------------->8-------------------------- $ perf list [snip] arc_pmu/bdata64/ [Kernel PMU event] arc_pmu/bdcstall/ [Kernel PMU event] arc_pmu/bdslot/ [Kernel PMU event] arc_pmu/bfbmp/ [Kernel PMU event] arc_pmu/bfirqex/ [Kernel PMU event] arc_pmu/bflgstal/ [Kernel PMU event] arc_pmu/bflush/ [Kernel PMU event] -------------------------->8-------------------------- Signed-off-by: Eugeniy Paltsev Signed-off-by: Vineet Gupta --- arch/arc/kernel/perf_event.c | 106 ++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 105 insertions(+), 1 deletion(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/perf_event.c b/arch/arc/kernel/perf_event.c index 693f32d60c35..d60aaaead421 100644 --- a/arch/arc/kernel/perf_event.c +++ b/arch/arc/kernel/perf_event.c @@ -17,12 +17,28 @@ /* HW holds 8 symbols + one for null terminator */ #define ARCPMU_EVENT_NAME_LEN 9 +enum arc_pmu_attr_groups { + ARCPMU_ATTR_GR_EVENTS, + ARCPMU_ATTR_GR_FORMATS, + ARCPMU_NR_ATTR_GR +}; + +struct arc_pmu_raw_event_entry { + char name[ARCPMU_EVENT_NAME_LEN]; +}; + struct arc_pmu { struct pmu pmu; unsigned int irq; int n_counters; + int n_events; u64 max_period; int ev_hw_idx[PERF_COUNT_ARC_HW_MAX]; + + struct arc_pmu_raw_event_entry *raw_entry; + struct attribute **attrs; + struct perf_pmu_events_attr *attr; + const struct attribute_group *attr_groups[ARCPMU_NR_ATTR_GR + 1]; }; struct arc_pmu_cpu { @@ -192,6 +208,17 @@ static int arc_pmu_event_init(struct perf_event *event) (int)hwc->config, arc_pmu_ev_hw_map[ret]); return 0; + case PERF_TYPE_RAW: + if (event->attr.config >= arc_pmu->n_events) + return -ENOENT; + + hwc->config |= event->attr.config; + pr_debug("init raw event with idx %lld \'%s\'\n", + event->attr.config, + arc_pmu->raw_entry[event->attr.config].name); + + return 0; + default: return -ENOENT; } @@ -442,6 +469,67 @@ static void arc_cpu_pmu_irq_init(void *data) write_aux_reg(ARC_REG_PCT_INT_ACT, 0xffffffff); } +/* Event field occupies the bottom 15 bits of our config field */ +PMU_FORMAT_ATTR(event, "config:0-14"); +static struct attribute *arc_pmu_format_attrs[] = { + &format_attr_event.attr, + NULL, +}; + +static struct attribute_group arc_pmu_format_attr_gr = { + .name = "format", + .attrs = arc_pmu_format_attrs, +}; + +static ssize_t arc_pmu_events_sysfs_show(struct device *dev, + struct device_attribute *attr, + char *page) +{ + struct perf_pmu_events_attr *pmu_attr; + + pmu_attr = container_of(attr, struct perf_pmu_events_attr, attr); + return sprintf(page, "event=0x%04llx\n", pmu_attr->id); +} + +/* + * We don't add attrs here as we don't have pre-defined list of perf events. + * We will generate and add attrs dynamically in probe() after we read HW + * configuration. + */ +static struct attribute_group arc_pmu_events_attr_gr = { + .name = "events", +}; + +static void arc_pmu_add_raw_event_attr(int j, char *str) +{ + memmove(arc_pmu->raw_entry[j].name, str, ARCPMU_EVENT_NAME_LEN - 1); + arc_pmu->attr[j].attr.attr.name = arc_pmu->raw_entry[j].name; + arc_pmu->attr[j].attr.attr.mode = VERIFY_OCTAL_PERMISSIONS(0444); + arc_pmu->attr[j].attr.show = arc_pmu_events_sysfs_show; + arc_pmu->attr[j].id = j; + arc_pmu->attrs[j] = &(arc_pmu->attr[j].attr.attr); +} + +static int arc_pmu_raw_alloc(struct device *dev) +{ + arc_pmu->attr = devm_kmalloc_array(dev, arc_pmu->n_events + 1, + sizeof(*arc_pmu->attr), GFP_KERNEL | __GFP_ZERO); + if (!arc_pmu->attr) + return -ENOMEM; + + arc_pmu->attrs = devm_kmalloc_array(dev, arc_pmu->n_events + 1, + sizeof(*arc_pmu->attrs), GFP_KERNEL | __GFP_ZERO); + if (!arc_pmu->attrs) + return -ENOMEM; + + arc_pmu->raw_entry = devm_kmalloc_array(dev, arc_pmu->n_events, + sizeof(*arc_pmu->raw_entry), GFP_KERNEL | __GFP_ZERO); + if (!arc_pmu->raw_entry) + return -ENOMEM; + + return 0; +} + static int arc_pmu_device_probe(struct platform_device *pdev) { struct arc_reg_pct_build pct_bcr; @@ -473,6 +561,11 @@ static int arc_pmu_device_probe(struct platform_device *pdev) if (!arc_pmu) return -ENOMEM; + arc_pmu->n_events = cc_bcr.c; + + if (arc_pmu_raw_alloc(&pdev->dev)) + return -ENOMEM; + has_interrupts = is_isa_arcv2() ? pct_bcr.i : 0; arc_pmu->n_counters = pct_bcr.c; @@ -504,8 +597,14 @@ static int arc_pmu_device_probe(struct platform_device *pdev) arc_pmu->ev_hw_idx[i] = j; } } + + arc_pmu_add_raw_event_attr(j, cc_name.str); } + arc_pmu_events_attr_gr.attrs = arc_pmu->attrs; + arc_pmu->attr_groups[ARCPMU_ATTR_GR_EVENTS] = &arc_pmu_events_attr_gr; + arc_pmu->attr_groups[ARCPMU_ATTR_GR_FORMATS] = &arc_pmu_format_attr_gr; + arc_pmu->pmu = (struct pmu) { .pmu_enable = arc_pmu_enable, .pmu_disable = arc_pmu_disable, @@ -515,6 +614,7 @@ static int arc_pmu_device_probe(struct platform_device *pdev) .start = arc_pmu_start, .stop = arc_pmu_stop, .read = arc_pmu_read, + .attr_groups = arc_pmu->attr_groups, }; if (has_interrupts) { @@ -536,7 +636,11 @@ static int arc_pmu_device_probe(struct platform_device *pdev) } else arc_pmu->pmu.capabilities |= PERF_PMU_CAP_NO_INTERRUPT; - return perf_pmu_register(&arc_pmu->pmu, pdev->name, PERF_TYPE_RAW); + /* + * perf parser doesn't really like '-' symbol in events name, so let's + * use '_' in arc pct name as it goes to kernel PMU event prefix. + */ + return perf_pmu_register(&arc_pmu->pmu, "arc_pct", PERF_TYPE_RAW); } static const struct of_device_id arc_pmu_match[] = { -- cgit v1.2.3 From baf9cc85ba01f32cf2ee79042a64b874a58cfb92 Mon Sep 17 00:00:00 2001 From: Eugeniy Paltsev Date: Thu, 13 Dec 2018 19:56:20 +0300 Subject: ARC: perf: move HW events mapping to separate function Move HW events mapping to separate function to make code more readable. Signed-off-by: Eugeniy Paltsev Signed-off-by: Vineet Gupta --- arch/arc/kernel/perf_event.c | 48 ++++++++++++++++++++++++++++++-------------- 1 file changed, 33 insertions(+), 15 deletions(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/perf_event.c b/arch/arc/kernel/perf_event.c index d60aaaead421..248c7b61690a 100644 --- a/arch/arc/kernel/perf_event.c +++ b/arch/arc/kernel/perf_event.c @@ -530,11 +530,39 @@ static int arc_pmu_raw_alloc(struct device *dev) return 0; } +static inline bool event_in_hw_event_map(int i, char *name) +{ + if (!arc_pmu_ev_hw_map[i]) + return false; + + if (!strlen(arc_pmu_ev_hw_map[i])) + return false; + + if (strcmp(arc_pmu_ev_hw_map[i], name)) + return false; + + return true; +} + +static void arc_pmu_map_hw_event(int j, char *str) +{ + int i; + + /* See if HW condition has been mapped to a perf event_id */ + for (i = 0; i < ARRAY_SIZE(arc_pmu_ev_hw_map); i++) { + if (event_in_hw_event_map(i, str)) { + pr_debug("mapping perf event %2d to h/w event \'%8s\' (idx %d)\n", + i, str, j); + arc_pmu->ev_hw_idx[i] = j; + } + } +} + static int arc_pmu_device_probe(struct platform_device *pdev) { struct arc_reg_pct_build pct_bcr; struct arc_reg_cc_build cc_bcr; - int i, j, has_interrupts; + int i, has_interrupts; int counter_size; /* in bits */ union cc_name { @@ -582,23 +610,13 @@ static int arc_pmu_device_probe(struct platform_device *pdev) arc_pmu->ev_hw_idx[i] = -1; /* loop thru all available h/w condition indexes */ - for (j = 0; j < cc_bcr.c; j++) { - write_aux_reg(ARC_REG_CC_INDEX, j); + for (i = 0; i < cc_bcr.c; i++) { + write_aux_reg(ARC_REG_CC_INDEX, i); cc_name.indiv.word0 = read_aux_reg(ARC_REG_CC_NAME0); cc_name.indiv.word1 = read_aux_reg(ARC_REG_CC_NAME1); - /* See if it has been mapped to a perf event_id */ - for (i = 0; i < ARRAY_SIZE(arc_pmu_ev_hw_map); i++) { - if (arc_pmu_ev_hw_map[i] && - !strcmp(arc_pmu_ev_hw_map[i], cc_name.str) && - strlen(arc_pmu_ev_hw_map[i])) { - pr_debug("mapping perf event %2d to h/w event \'%8s\' (idx %d)\n", - i, cc_name.str, j); - arc_pmu->ev_hw_idx[i] = j; - } - } - - arc_pmu_add_raw_event_attr(j, cc_name.str); + arc_pmu_map_hw_event(i, cc_name.str); + arc_pmu_add_raw_event_attr(i, cc_name.str); } arc_pmu_events_attr_gr.attrs = arc_pmu->attrs; -- cgit v1.2.3 From 29133260f7c2e4ce50b4da6bf0674331bc0a4da5 Mon Sep 17 00:00:00 2001 From: Eugeniy Paltsev Date: Thu, 13 Dec 2018 19:56:21 +0300 Subject: ARC: perf: avoid kernel killing where it is possible No, not gonna die tonight. Signed-off-by: Eugeniy Paltsev Signed-off-by: Vineet Gupta --- arch/arc/kernel/perf_event.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/perf_event.c b/arch/arc/kernel/perf_event.c index 248c7b61690a..861a8aea51f9 100644 --- a/arch/arc/kernel/perf_event.c +++ b/arch/arc/kernel/perf_event.c @@ -580,10 +580,12 @@ static int arc_pmu_device_probe(struct platform_device *pdev) return -ENODEV; } BUILD_BUG_ON(ARC_PERF_MAX_COUNTERS > 32); - BUG_ON(pct_bcr.c > ARC_PERF_MAX_COUNTERS); + if (WARN_ON(pct_bcr.c > ARC_PERF_MAX_COUNTERS)) + return -EINVAL; READ_BCR(ARC_REG_CC_BUILD, cc_bcr); - BUG_ON(!cc_bcr.v); /* Counters exist but No countable conditions ? */ + if (WARN(!cc_bcr.v, "Counters exist but No countable conditions?")) + return -EINVAL; arc_pmu = devm_kzalloc(&pdev->dev, sizeof(struct arc_pmu), GFP_KERNEL); if (!arc_pmu) -- cgit v1.2.3 From ab6c03676cb190156603cf4c5ecf97aa406c9c53 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Mon, 17 Dec 2018 14:11:19 -0800 Subject: ARC: show_regs: lockdep: avoid page allocator... and use smaller/on-stack buffer instead The motivation for this change was lockdep splat like below. | potentially unexpected fatal signal 11. | BUG: sleeping function called from invalid context at ../mm/page_alloc.c:4317 | in_atomic(): 1, irqs_disabled(): 0, pid: 57, name: segv | no locks held by segv/57. | Preemption disabled at: | [<8182f17e>] get_signal+0x4a6/0x7c4 | CPU: 0 PID: 57 Comm: segv Not tainted 4.17.0+ #23 | | Stack Trace: | arc_unwind_core.constprop.1+0xd0/0xf4 | __might_sleep+0x1f6/0x234 | __get_free_pages+0x174/0xca0 | show_regs+0x22/0x330 | get_signal+0x4ac/0x7c4 # print_fatal_signals() -> preempt_disable() | do_signal+0x30/0x224 | resume_user_mode_begin+0x90/0xd8 So signal handling core calls show_regs() with preemption disabled but an ensuing GFP_KERNEL page allocator call is flagged by lockdep. We could have switched to GFP_NOWAIT, but turns out that is not enough anways and eliding page allocator call leads to less code and instruction traces to sift thru when debugging pesky crashes. FWIW, this patch doesn't cure the lockdep splat (which next patch does). Reviewed-by: William Kucharski Signed-off-by: Vineet Gupta --- arch/arc/kernel/troubleshoot.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/troubleshoot.c b/arch/arc/kernel/troubleshoot.c index e8d9fb452346..5c6663321e87 100644 --- a/arch/arc/kernel/troubleshoot.c +++ b/arch/arc/kernel/troubleshoot.c @@ -18,6 +18,8 @@ #include #include +#define ARC_PATH_MAX 256 + /* * Common routine to print scratch regs (r0-r12) or callee regs (r13-r25) * -Prints 3 regs per line and a CR. @@ -58,11 +60,12 @@ static void show_callee_regs(struct callee_regs *cregs) print_reg_file(&(cregs->r13), 13); } -static void print_task_path_n_nm(struct task_struct *tsk, char *buf) +static void print_task_path_n_nm(struct task_struct *tsk) { char *path_nm = NULL; struct mm_struct *mm; struct file *exe_file; + char buf[ARC_PATH_MAX]; mm = get_task_mm(tsk); if (!mm) @@ -72,7 +75,7 @@ static void print_task_path_n_nm(struct task_struct *tsk, char *buf) mmput(mm); if (exe_file) { - path_nm = file_path(exe_file, buf, 255); + path_nm = file_path(exe_file, buf, ARC_PATH_MAX-1); fput(exe_file); } @@ -80,10 +83,9 @@ done: pr_info("Path: %s\n", !IS_ERR(path_nm) ? path_nm : "?"); } -static void show_faulting_vma(unsigned long address, char *buf) +static void show_faulting_vma(unsigned long address) { struct vm_area_struct *vma; - char *nm = buf; struct mm_struct *active_mm = current->active_mm; /* can't use print_vma_addr() yet as it doesn't check for @@ -96,8 +98,11 @@ static void show_faulting_vma(unsigned long address, char *buf) * if the container VMA is not found */ if (vma && (vma->vm_start <= address)) { + char buf[ARC_PATH_MAX]; + char *nm = "?"; + if (vma->vm_file) { - nm = file_path(vma->vm_file, buf, PAGE_SIZE - 1); + nm = file_path(vma->vm_file, buf, ARC_PATH_MAX-1); if (IS_ERR(nm)) nm = "?"; } @@ -173,13 +178,8 @@ void show_regs(struct pt_regs *regs) { struct task_struct *tsk = current; struct callee_regs *cregs; - char *buf; - buf = (char *)__get_free_page(GFP_KERNEL); - if (!buf) - return; - - print_task_path_n_nm(tsk, buf); + print_task_path_n_nm(tsk); show_regs_print_info(KERN_INFO); show_ecr_verbose(regs); @@ -189,7 +189,7 @@ void show_regs(struct pt_regs *regs) (void *)regs->blink, (void *)regs->ret); if (user_mode(regs)) - show_faulting_vma(regs->ret, buf); /* faulting code, not data */ + show_faulting_vma(regs->ret); /* faulting code, not data */ pr_info("[STAT32]: 0x%08lx", regs->status32); @@ -221,8 +221,6 @@ void show_regs(struct pt_regs *regs) cregs = (struct callee_regs *)current->thread.callee_reg; if (cregs) show_callee_regs(cregs); - - free_page((unsigned long)buf); } void show_kernel_fault_diag(const char *str, struct pt_regs *regs, -- cgit v1.2.3 From f731a8e89f8c78985707c626680f3e24c7a60772 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Tue, 18 Dec 2018 10:39:58 -0800 Subject: ARC: show_regs: lockdep: re-enable preemption signal handling core calls show_regs() with preemption disabled which on ARC takes mmap_sem for mm/vma access, causing lockdep splat. | [ARCLinux]# ./segv-null-ptr | potentially unexpected fatal signal 11. | BUG: sleeping function called from invalid context at kernel/fork.c:1011 | in_atomic(): 1, irqs_disabled(): 0, pid: 70, name: segv-null-ptr | no locks held by segv-null-ptr/70. | CPU: 0 PID: 70 Comm: segv-null-ptr Not tainted 4.18.0+ #69 | | Stack Trace: | arc_unwind_core+0xcc/0x100 | ___might_sleep+0x17a/0x190 | mmput+0x16/0xb8 | show_regs+0x52/0x310 | get_signal+0x5ee/0x610 | do_signal+0x2c/0x218 | resume_user_mode_begin+0x90/0xd8 Workaround by re-enabling preemption temporarily. Note that the preemption disabling in core code around show_regs() was introduced by commit 3a9f84d354ce ("signals, debug: fix BUG: using smp_processor_id() in preemptible code in print_fatal_signal()") to silence a differnt lockdep seen on x86 bakc in 2009. Cc: Signed-off-by: Vineet Gupta --- arch/arc/kernel/troubleshoot.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'arch/arc/kernel') diff --git a/arch/arc/kernel/troubleshoot.c b/arch/arc/kernel/troubleshoot.c index 5c6663321e87..215f515442e0 100644 --- a/arch/arc/kernel/troubleshoot.c +++ b/arch/arc/kernel/troubleshoot.c @@ -179,6 +179,12 @@ void show_regs(struct pt_regs *regs) struct task_struct *tsk = current; struct callee_regs *cregs; + /* + * generic code calls us with preemption disabled, but some calls + * here could sleep, so re-enable to avoid lockdep splat + */ + preempt_enable(); + print_task_path_n_nm(tsk); show_regs_print_info(KERN_INFO); @@ -221,6 +227,8 @@ void show_regs(struct pt_regs *regs) cregs = (struct callee_regs *)current->thread.callee_reg; if (cregs) show_callee_regs(cregs); + + preempt_disable(); } void show_kernel_fault_diag(const char *str, struct pt_regs *regs, -- cgit v1.2.3