From 0dd450fe13da4aeacc69916ecfe39d3d0b531900 Mon Sep 17 00:00:00 2001 From: Mischa Jonker Date: Thu, 7 Nov 2013 14:55:11 +0100 Subject: [PATCH] ARC: Add perf support for ARC700 cores This adds basic perf support for ARC700 cores. Most PERF_COUNT_HW* events are supported now. Signed-off-by: Mischa Jonker Acked-by: Peter Zijlstra Signed-off-by: Vineet Gupta --- arch/arc/boot/dts/angel4.dts | 4 + arch/arc/include/asm/perf_event.h | 204 +++++++++++++++++++++++- arch/arc/kernel/Makefile | 1 + arch/arc/kernel/perf_event.c | 322 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 530 insertions(+), 1 deletion(-) create mode 100644 arch/arc/kernel/perf_event.c diff --git a/arch/arc/boot/dts/angel4.dts b/arch/arc/boot/dts/angel4.dts index 4fb2d6f..bcf662d 100644 --- a/arch/arc/boot/dts/angel4.dts +++ b/arch/arc/boot/dts/angel4.dts @@ -67,5 +67,9 @@ reg = <1>; }; }; + + arcpmu0: pmu { + compatible = "snps,arc700-pmu"; + }; }; }; diff --git a/arch/arc/include/asm/perf_event.h b/arch/arc/include/asm/perf_event.h index 115ad96..cbf755e 100644 --- a/arch/arc/include/asm/perf_event.h +++ b/arch/arc/include/asm/perf_event.h @@ -1,5 +1,7 @@ /* - * Copyright (C) 2011-2012 Synopsys, Inc. (www.synopsys.com) + * Linux performance counter support for ARC + * + * Copyright (C) 2011-2013 Synopsys, Inc. (www.synopsys.com) * * 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 @@ -10,4 +12,204 @@ #ifndef __ASM_PERF_EVENT_H #define __ASM_PERF_EVENT_H +/* real maximum varies per CPU, this is the maximum supported by the driver */ +#define ARC_PMU_MAX_HWEVENTS 64 + +#define ARC_REG_CC_BUILD 0xF6 +#define ARC_REG_CC_INDEX 0x240 +#define ARC_REG_CC_NAME0 0x241 +#define ARC_REG_CC_NAME1 0x242 + +#define ARC_REG_PCT_BUILD 0xF5 +#define ARC_REG_PCT_COUNTL 0x250 +#define ARC_REG_PCT_COUNTH 0x251 +#define ARC_REG_PCT_SNAPL 0x252 +#define ARC_REG_PCT_SNAPH 0x253 +#define ARC_REG_PCT_CONFIG 0x254 +#define ARC_REG_PCT_CONTROL 0x255 +#define ARC_REG_PCT_INDEX 0x256 + +#define ARC_REG_PCT_CONTROL_CC (1 << 16) /* clear counts */ +#define ARC_REG_PCT_CONTROL_SN (1 << 17) /* snapshot */ + +struct arc_reg_pct_build { +#ifdef CONFIG_CPU_BIG_ENDIAN + unsigned int m:8, c:8, r:6, s:2, v:8; +#else + unsigned int v:8, s:2, r:6, c:8, m:8; +#endif +}; + +struct arc_reg_cc_build { +#ifdef CONFIG_CPU_BIG_ENDIAN + unsigned int c:16, r:8, v:8; +#else + unsigned int v:8, r:8, c:16; +#endif +}; + +#define PERF_COUNT_ARC_DCLM (PERF_COUNT_HW_MAX + 0) +#define PERF_COUNT_ARC_DCSM (PERF_COUNT_HW_MAX + 1) +#define PERF_COUNT_ARC_ICM (PERF_COUNT_HW_MAX + 2) +#define PERF_COUNT_ARC_BPOK (PERF_COUNT_HW_MAX + 3) +#define PERF_COUNT_ARC_EDTLB (PERF_COUNT_HW_MAX + 4) +#define PERF_COUNT_ARC_EITLB (PERF_COUNT_HW_MAX + 5) +#define PERF_COUNT_ARC_HW_MAX (PERF_COUNT_HW_MAX + 6) + +/* + * The "generalized" performance events seem to really be a copy + * of the available events on x86 processors; the mapping to ARC + * events is not always possible 1-to-1. Fortunately, there doesn't + * seem to be an exact definition for these events, so we can cheat + * a bit where necessary. + * + * In particular, the following PERF events may behave a bit differently + * compared to other architectures: + * + * PERF_COUNT_HW_CPU_CYCLES + * Cycles not in halted state + * + * PERF_COUNT_HW_REF_CPU_CYCLES + * Reference cycles not in halted state, same as PERF_COUNT_HW_CPU_CYCLES + * for now as we don't do Dynamic Voltage/Frequency Scaling (yet) + * + * PERF_COUNT_HW_BUS_CYCLES + * Unclear what this means, Intel uses 0x013c, which according to + * their datasheet means "unhalted reference cycles". It sounds similar + * to PERF_COUNT_HW_REF_CPU_CYCLES, and we use the same counter for it. + * + * PERF_COUNT_HW_STALLED_CYCLES_BACKEND + * PERF_COUNT_HW_STALLED_CYCLES_FRONTEND + * The ARC 700 can either measure stalls per pipeline stage, or all stalls + * combined; for now we assign all stalls to STALLED_CYCLES_BACKEND + * and all pipeline flushes (e.g. caused by mispredicts, etc.) to + * STALLED_CYCLES_FRONTEND. + * + * We could start multiple performance counters and combine everything + * afterwards, but that makes it complicated. + * + * Note that I$ cache misses aren't counted by either of the two! + */ + +static const char * const arc_pmu_ev_hw_map[] = { + [PERF_COUNT_HW_CPU_CYCLES] = "crun", + [PERF_COUNT_HW_REF_CPU_CYCLES] = "crun", + [PERF_COUNT_HW_BUS_CYCLES] = "crun", + [PERF_COUNT_HW_INSTRUCTIONS] = "iall", + [PERF_COUNT_HW_BRANCH_MISSES] = "bpfail", + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = "ijmp", + [PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = "bflush", + [PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = "bstall", + [PERF_COUNT_ARC_DCLM] = "dclm", + [PERF_COUNT_ARC_DCSM] = "dcsm", + [PERF_COUNT_ARC_ICM] = "icm", + [PERF_COUNT_ARC_BPOK] = "bpok", + [PERF_COUNT_ARC_EDTLB] = "edtlb", + [PERF_COUNT_ARC_EITLB] = "eitlb", +}; + +#define C(_x) PERF_COUNT_HW_CACHE_##_x +#define CACHE_OP_UNSUPPORTED 0xffff + +static const unsigned arc_pmu_cache_map[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { + [C(L1D)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = PERF_COUNT_ARC_DCLM, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = PERF_COUNT_ARC_DCSM, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, + [C(L1I)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = PERF_COUNT_ARC_ICM, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, + [C(LL)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, + [C(DTLB)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = PERF_COUNT_ARC_EDTLB, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, + [C(ITLB)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = PERF_COUNT_ARC_EITLB, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, + [C(BPU)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PERF_COUNT_HW_BRANCH_INSTRUCTIONS, + [C(RESULT_MISS)] = PERF_COUNT_HW_BRANCH_MISSES, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, + [C(NODE)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED, + [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED, + }, + }, +}; + #endif /* __ASM_PERF_EVENT_H */ diff --git a/arch/arc/kernel/Makefile b/arch/arc/kernel/Makefile index c242ef0..8004b4f 100644 --- a/arch/arc/kernel/Makefile +++ b/arch/arc/kernel/Makefile @@ -19,6 +19,7 @@ obj-$(CONFIG_KPROBES) += kprobes.o obj-$(CONFIG_ARC_MISALIGN_ACCESS) += unaligned.o obj-$(CONFIG_KGDB) += kgdb.o obj-$(CONFIG_ARC_METAWARE_HLINK) += arc_hostlink.o +obj-$(CONFIG_PERF_EVENTS) += perf_event.o obj-$(CONFIG_ARC_FPU_SAVE_RESTORE) += fpu.o CFLAGS_fpu.o += -mdpfp diff --git a/arch/arc/kernel/perf_event.c b/arch/arc/kernel/perf_event.c new file mode 100644 index 0000000..759e5f3 --- /dev/null +++ b/arch/arc/kernel/perf_event.c @@ -0,0 +1,322 @@ +/* + * Linux performance counter support for ARC700 series + * + * Copyright (C) 2013 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. + * + */ +#include +#include +#include +#include +#include +#include + +struct arc_pmu { + struct pmu pmu; + int counter_size; /* in bits */ + int n_counters; + unsigned long used_mask[BITS_TO_LONGS(ARC_PMU_MAX_HWEVENTS)]; + int ev_hw_idx[PERF_COUNT_ARC_HW_MAX]; +}; + +/* read counter #idx; note that counter# != event# on ARC! */ +static uint64_t arc_pmu_read_counter(int idx) +{ + uint32_t tmp; + uint64_t result; + + /* + * ARC supports making 'snapshots' of the counters, so we don't + * need to care about counters wrapping to 0 underneath our feet + */ + 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 |= read_aux_reg(ARC_REG_PCT_SNAPL); + + return result; +} + +static void arc_perf_event_update(struct perf_event *event, + struct hw_perf_event *hwc, int idx) +{ + struct arc_pmu *arc_pmu = container_of(event->pmu, struct arc_pmu, pmu); + uint64_t prev_raw_count, new_raw_count; + int64_t delta; + + do { + prev_raw_count = local64_read(&hwc->prev_count); + new_raw_count = arc_pmu_read_counter(idx); + } while (local64_cmpxchg(&hwc->prev_count, prev_raw_count, + new_raw_count) != prev_raw_count); + + delta = (new_raw_count - prev_raw_count) & + ((1ULL << arc_pmu->counter_size) - 1ULL); + + local64_add(delta, &event->count); + local64_sub(delta, &hwc->period_left); +} + +static void arc_pmu_read(struct perf_event *event) +{ + arc_perf_event_update(event, &event->hw, event->hw.idx); +} + +static int arc_pmu_cache_event(u64 config) +{ + unsigned int cache_type, cache_op, cache_result; + int ret; + + cache_type = (config >> 0) & 0xff; + cache_op = (config >> 8) & 0xff; + cache_result = (config >> 16) & 0xff; + if (cache_type >= PERF_COUNT_HW_CACHE_MAX) + return -EINVAL; + if (cache_type >= PERF_COUNT_HW_CACHE_OP_MAX) + return -EINVAL; + if (cache_type >= PERF_COUNT_HW_CACHE_RESULT_MAX) + return -EINVAL; + + ret = arc_pmu_cache_map[cache_type][cache_op][cache_result]; + + if (ret == CACHE_OP_UNSUPPORTED) + return -ENOENT; + + return ret; +} + +/* initializes hw_perf_event structure if event is supported */ +static int arc_pmu_event_init(struct perf_event *event) +{ + struct arc_pmu *arc_pmu = container_of(event->pmu, struct arc_pmu, pmu); + struct hw_perf_event *hwc = &event->hw; + int ret; + + switch (event->attr.type) { + case PERF_TYPE_HARDWARE: + if (event->attr.config >= PERF_COUNT_HW_MAX) + return -ENOENT; + if (arc_pmu->ev_hw_idx[event->attr.config] < 0) + return -ENOENT; + hwc->config = arc_pmu->ev_hw_idx[event->attr.config]; + pr_debug("initializing event %d with cfg %d\n", + (int) event->attr.config, (int) hwc->config); + return 0; + case PERF_TYPE_HW_CACHE: + ret = arc_pmu_cache_event(event->attr.config); + if (ret < 0) + return ret; + hwc->config = arc_pmu->ev_hw_idx[ret]; + return 0; + default: + return -ENOENT; + } +} + +/* starts all counters */ +static void arc_pmu_enable(struct pmu *pmu) +{ + uint32_t tmp; + tmp = read_aux_reg(ARC_REG_PCT_CONTROL); + write_aux_reg(ARC_REG_PCT_CONTROL, (tmp & 0xffff0000) | 0x1); +} + +/* stops all counters */ +static void arc_pmu_disable(struct pmu *pmu) +{ + uint32_t tmp; + tmp = read_aux_reg(ARC_REG_PCT_CONTROL); + write_aux_reg(ARC_REG_PCT_CONTROL, (tmp & 0xffff0000) | 0x0); +} + +/* + * Assigns hardware counter to hardware condition. + * Note that there is no separate start/stop mechanism; + * stopping is achieved by assigning the 'never' condition + */ +static void arc_pmu_start(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + int idx = hwc->idx; + + if (WARN_ON_ONCE(idx == -1)) + return; + + if (flags & PERF_EF_RELOAD) + WARN_ON_ONCE(!(event->hw.state & PERF_HES_UPTODATE)); + + event->hw.state = 0; + + /* enable ARC pmu here */ + write_aux_reg(ARC_REG_PCT_INDEX, idx); + write_aux_reg(ARC_REG_PCT_CONFIG, hwc->config); +} + +static void arc_pmu_stop(struct perf_event *event, int flags) +{ + struct hw_perf_event *hwc = &event->hw; + int idx = hwc->idx; + + if (!(event->hw.state & PERF_HES_STOPPED)) { + /* stop ARC pmu here */ + write_aux_reg(ARC_REG_PCT_INDEX, idx); + + /* condition code #0 is always "never" */ + write_aux_reg(ARC_REG_PCT_CONFIG, 0); + + event->hw.state |= PERF_HES_STOPPED; + } + + if ((flags & PERF_EF_UPDATE) && + !(event->hw.state & PERF_HES_UPTODATE)) { + arc_perf_event_update(event, &event->hw, idx); + event->hw.state |= PERF_HES_UPTODATE; + } +} + +static void arc_pmu_del(struct perf_event *event, int flags) +{ + struct arc_pmu *arc_pmu = container_of(event->pmu, struct arc_pmu, pmu); + + arc_pmu_stop(event, PERF_EF_UPDATE); + __clear_bit(event->hw.idx, arc_pmu->used_mask); + + perf_event_update_userpage(event); +} + +/* allocate hardware counter and optionally start counting */ +static int arc_pmu_add(struct perf_event *event, int flags) +{ + struct arc_pmu *arc_pmu = container_of(event->pmu, struct arc_pmu, pmu); + struct hw_perf_event *hwc = &event->hw; + int idx = hwc->idx; + + if (__test_and_set_bit(idx, arc_pmu->used_mask)) { + idx = find_first_zero_bit(arc_pmu->used_mask, + arc_pmu->n_counters); + if (idx == arc_pmu->n_counters) + return -EAGAIN; + + __set_bit(idx, arc_pmu->used_mask); + hwc->idx = idx; + } + + write_aux_reg(ARC_REG_PCT_INDEX, idx); + write_aux_reg(ARC_REG_PCT_CONFIG, 0); + write_aux_reg(ARC_REG_PCT_COUNTL, 0); + write_aux_reg(ARC_REG_PCT_COUNTH, 0); + local64_set(&hwc->prev_count, 0); + + hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED; + if (flags & PERF_EF_START) + arc_pmu_start(event, PERF_EF_RELOAD); + + perf_event_update_userpage(event); + + return 0; +} + +static int arc_pmu_device_probe(struct platform_device *pdev) +{ + struct arc_pmu *arc_pmu; + struct arc_reg_pct_build pct_bcr; + struct arc_reg_cc_build cc_bcr; + int i, j, ret; + + union cc_name { + struct { + uint32_t word0, word1; + char sentinel; + } indiv; + char str[9]; + } cc_name; + + + READ_BCR(ARC_REG_PCT_BUILD, pct_bcr); + if (!pct_bcr.v) { + pr_err("This core does not have performance counters!\n"); + return -ENODEV; + } + + arc_pmu = devm_kzalloc(&pdev->dev, sizeof(struct arc_pmu), + GFP_KERNEL); + if (!arc_pmu) + return -ENOMEM; + + arc_pmu->n_counters = pct_bcr.c; + BUG_ON(arc_pmu->n_counters > ARC_PMU_MAX_HWEVENTS); + + arc_pmu->counter_size = 32 + (pct_bcr.s << 4); + pr_info("ARC PMU found with %d counters of size %d bits\n", + arc_pmu->n_counters, arc_pmu->counter_size); + + READ_BCR(ARC_REG_CC_BUILD, cc_bcr); + + if (!cc_bcr.v) + pr_err("Strange! Performance counters exist, but no countable conditions?\n"); + + pr_info("ARC PMU has %d countable conditions\n", cc_bcr.c); + + cc_name.str[8] = 0; + for (i = 0; i < PERF_COUNT_HW_MAX; i++) + arc_pmu->ev_hw_idx[i] = -1; + + for (j = 0; j < cc_bcr.c; j++) { + write_aux_reg(ARC_REG_CC_INDEX, j); + cc_name.indiv.word0 = read_aux_reg(ARC_REG_CC_NAME0); + cc_name.indiv.word1 = read_aux_reg(ARC_REG_CC_NAME1); + 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 %d to idx %d with name %s\n", + i, j, cc_name.str); + arc_pmu->ev_hw_idx[i] = j; + } + } + } + + arc_pmu->pmu = (struct pmu) { + .pmu_enable = arc_pmu_enable, + .pmu_disable = arc_pmu_disable, + .event_init = arc_pmu_event_init, + .add = arc_pmu_add, + .del = arc_pmu_del, + .start = arc_pmu_start, + .stop = arc_pmu_stop, + .read = arc_pmu_read, + }; + + ret = perf_pmu_register(&arc_pmu->pmu, pdev->name, PERF_TYPE_RAW); + + return ret; +} + +#ifdef CONFIG_OF +static const struct of_device_id arc_pmu_match[] = { + { .compatible = "snps,arc700-pmu" }, + {}, +}; +MODULE_DEVICE_TABLE(of, arc_pmu_match); +#endif + +static struct platform_driver arc_pmu_driver = { + .driver = { + .name = "arc700-pmu", + .of_match_table = of_match_ptr(arc_pmu_match), + }, + .probe = arc_pmu_device_probe, +}; + +module_platform_driver(arc_pmu_driver); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Mischa Jonker "); +MODULE_DESCRIPTION("ARC PMU driver"); -- 2.7.4