diff --git a/MAINTAINERS b/MAINTAINERS index 1b40446c73..dec252f38b 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1096,6 +1096,12 @@ F: hw/isa/vt82c686.c F: hw/pci-host/bonito.c F: include/hw/isa/vt82c686.h +Loongson-3 virtual platforms +M: Huacai Chen +R: Jiaxun Yang +S: Maintained +F: hw/intc/loongson_liointc.c + Boston M: Paul Burton R: Aleksandar Rikalo @@ -3019,3 +3025,10 @@ M: Peter Maydell S: Maintained F: docs/conf.py F: docs/*/conf.py + +Miscellaneous +------------- +Performance Tools and Tests +M: Ahmed Karaman +S: Maintained +F: scripts/performance/ diff --git a/hw/intc/Kconfig b/hw/intc/Kconfig index f562342bab..2ae1e89497 100644 --- a/hw/intc/Kconfig +++ b/hw/intc/Kconfig @@ -64,3 +64,6 @@ config OMPIC config RX_ICU bool + +config LOONGSON_LIOINTC + bool diff --git a/hw/intc/Makefile.objs b/hw/intc/Makefile.objs index a4202636d4..3ac2b40fbb 100644 --- a/hw/intc/Makefile.objs +++ b/hw/intc/Makefile.objs @@ -51,3 +51,4 @@ obj-$(CONFIG_MIPS_CPS) += mips_gic.o obj-$(CONFIG_NIOS2) += nios2_iic.o obj-$(CONFIG_OMPIC) += ompic.o obj-$(CONFIG_IBEX) += ibex_plic.o +obj-$(CONFIG_LOONGSON_LIOINTC) += loongson_liointc.o diff --git a/hw/intc/loongson_liointc.c b/hw/intc/loongson_liointc.c new file mode 100644 index 0000000000..23ca51cc2e --- /dev/null +++ b/hw/intc/loongson_liointc.c @@ -0,0 +1,242 @@ +/* + * QEMU Loongson Local I/O interrupt controler. + * + * Copyright (c) 2020 Jiaxun Yang + * + * This program is free software: you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation, either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program. If not, see . + * + */ + +#include "qemu/osdep.h" +#include "hw/sysbus.h" +#include "qemu/module.h" +#include "hw/irq.h" +#include "hw/qdev-properties.h" + +#define D(x) + +#define NUM_IRQS 32 + +#define NUM_CORES 4 +#define NUM_IPS 4 +#define NUM_PARENTS (NUM_CORES * NUM_IPS) +#define PARENT_COREx_IPy(x, y) (NUM_IPS * x + y) + +#define R_MAPPER_START 0x0 +#define R_MAPPER_END 0x20 +#define R_ISR R_MAPPER_END +#define R_IEN 0x24 +#define R_IEN_SET 0x28 +#define R_IEN_CLR 0x2c +#define R_PERCORE_ISR(x) (0x40 + 0x8 * x) +#define R_END 0x64 + +#define TYPE_LOONGSON_LIOINTC "loongson.liointc" +#define LOONGSON_LIOINTC(obj) \ + OBJECT_CHECK(struct loongson_liointc, (obj), TYPE_LOONGSON_LIOINTC) + +struct loongson_liointc { + SysBusDevice parent_obj; + + MemoryRegion mmio; + qemu_irq parent_irq[NUM_PARENTS]; + + uint8_t mapper[NUM_IRQS]; /* 0:3 for core, 4:7 for IP */ + uint32_t isr; + uint32_t ien; + uint32_t per_core_isr[NUM_CORES]; + + /* state of the interrupt input pins */ + uint32_t pin_state; + bool parent_state[NUM_PARENTS]; +}; + +static void update_irq(struct loongson_liointc *p) +{ + uint32_t irq, core, ip; + uint32_t per_ip_isr[NUM_IPS] = {0}; + + /* level triggered interrupt */ + p->isr = p->pin_state; + + /* Clear disabled IRQs */ + p->isr &= p->ien; + + /* Clear per_core_isr */ + for (core = 0; core < NUM_CORES; core++) { + p->per_core_isr[core] = 0; + } + + /* Update per_core_isr and per_ip_isr */ + for (irq = 0; irq < NUM_IRQS; irq++) { + if (!(p->isr & (1 << irq))) { + continue; + } + + for (core = 0; core < NUM_CORES; core++) { + if ((p->mapper[irq] & (1 << core))) { + p->per_core_isr[core] |= (1 << irq); + } + } + + for (ip = 0; ip < NUM_IPS; ip++) { + if ((p->mapper[irq] & (1 << (ip + 4)))) { + per_ip_isr[ip] |= (1 << irq); + } + } + } + + /* Emit IRQ to parent! */ + for (core = 0; core < NUM_CORES; core++) { + for (ip = 0; ip < NUM_IPS; ip++) { + int parent = PARENT_COREx_IPy(core, ip); + if (p->parent_state[parent] != + (!!p->per_core_isr[core] && !!per_ip_isr[ip])) { + p->parent_state[parent] = !p->parent_state[parent]; + qemu_set_irq(p->parent_irq[parent], p->parent_state[parent]); + } + } + } +} + +static uint64_t +liointc_read(void *opaque, hwaddr addr, unsigned int size) +{ + struct loongson_liointc *p = opaque; + uint32_t r = 0; + + /* Mapper is 1 byte */ + if (size == 1 && addr < R_MAPPER_END) { + r = p->mapper[addr]; + goto out; + } + + /* Rest is 4 byte */ + if (size != 4 || (addr % 4)) { + goto out; + } + + if (addr >= R_PERCORE_ISR(0) && + addr < R_PERCORE_ISR(NUM_CORES)) { + int core = (addr - R_PERCORE_ISR(0)) / 4; + r = p->per_core_isr[core]; + goto out; + } + + switch (addr) { + case R_ISR: + r = p->isr; + break; + case R_IEN: + r = p->ien; + break; + default: + break; + } + +out: + D(qemu_log("%s: size=%d addr=%lx val=%x\n", __func__, size, addr, r)); + return r; +} + +static void +liointc_write(void *opaque, hwaddr addr, + uint64_t val64, unsigned int size) +{ + struct loongson_liointc *p = opaque; + uint32_t value = val64; + + D(qemu_log("%s: size=%d, addr=%lx val=%x\n", __func__, size, addr, value)); + + /* Mapper is 1 byte */ + if (size == 1 && addr < R_MAPPER_END) { + p->mapper[addr] = value; + goto out; + } + + /* Rest is 4 byte */ + if (size != 4 || (addr % 4)) { + goto out; + } + + if (addr >= R_PERCORE_ISR(0) && + addr < R_PERCORE_ISR(NUM_CORES)) { + int core = (addr - R_PERCORE_ISR(0)) / 4; + p->per_core_isr[core] = value; + goto out; + } + + switch (addr) { + case R_IEN_SET: + p->ien |= value; + break; + case R_IEN_CLR: + p->ien &= ~value; + break; + default: + break; + } + +out: + update_irq(p); +} + +static const MemoryRegionOps pic_ops = { + .read = liointc_read, + .write = liointc_write, + .endianness = DEVICE_NATIVE_ENDIAN, + .valid = { + .min_access_size = 1, + .max_access_size = 4 + } +}; + +static void irq_handler(void *opaque, int irq, int level) +{ + struct loongson_liointc *p = opaque; + + p->pin_state &= ~(1 << irq); + p->pin_state |= level << irq; + update_irq(p); +} + +static void loongson_liointc_init(Object *obj) +{ + struct loongson_liointc *p = LOONGSON_LIOINTC(obj); + int i; + + qdev_init_gpio_in(DEVICE(obj), irq_handler, 32); + + for (i = 0; i < NUM_PARENTS; i++) { + sysbus_init_irq(SYS_BUS_DEVICE(obj), &p->parent_irq[i]); + } + + memory_region_init_io(&p->mmio, obj, &pic_ops, p, + "loongson.liointc", R_END); + sysbus_init_mmio(SYS_BUS_DEVICE(obj), &p->mmio); +} + +static const TypeInfo loongson_liointc_info = { + .name = TYPE_LOONGSON_LIOINTC, + .parent = TYPE_SYS_BUS_DEVICE, + .instance_size = sizeof(struct loongson_liointc), + .instance_init = loongson_liointc_init, +}; + +static void loongson_liointc_register_types(void) +{ + type_register_static(&loongson_liointc_info); +} + +type_init(loongson_liointc_register_types) diff --git a/scripts/performance/topN_callgrind.py b/scripts/performance/topN_callgrind.py new file mode 100755 index 0000000000..67c59197af --- /dev/null +++ b/scripts/performance/topN_callgrind.py @@ -0,0 +1,140 @@ +#!/usr/bin/env python3 + +# Print the top N most executed functions in QEMU using callgrind. +# Syntax: +# topN_callgrind.py [-h] [-n] -- \ +# [] \ +# [] +# +# [-h] - Print the script arguments help message. +# [-n] - Specify the number of top functions to print. +# - If this flag is not specified, the tool defaults to 25. +# +# Example of usage: +# topN_callgrind.py -n 20 -- qemu-arm coulomb_double-arm +# +# This file is a part of the project "TCG Continuous Benchmarking". +# +# Copyright (C) 2020 Ahmed Karaman +# Copyright (C) 2020 Aleksandar Markovic +# +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation, either version 2 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see . + +import argparse +import os +import subprocess +import sys + + +# Parse the command line arguments +parser = argparse.ArgumentParser( + usage='topN_callgrind.py [-h] [-n] -- ' + ' [] ' + ' []') + +parser.add_argument('-n', dest='top', type=int, default=25, + help='Specify the number of top functions to print.') + +parser.add_argument('command', type=str, nargs='+', help=argparse.SUPPRESS) + +args = parser.parse_args() + +# Extract the needed variables from the args +command = args.command +top = args.top + +# Insure that valgrind is installed +check_valgrind_presence = subprocess.run(["which", "valgrind"], + stdout=subprocess.DEVNULL) +if check_valgrind_presence.returncode: + sys.exit("Please install valgrind before running the script!") + +# Run callgrind +callgrind = subprocess.run(( + ["valgrind", "--tool=callgrind", "--callgrind-out-file=/tmp/callgrind.data"] + + command), + stdout=subprocess.DEVNULL, + stderr=subprocess.PIPE) +if callgrind.returncode: + sys.exit(callgrind.stderr.decode("utf-8")) + +# Save callgrind_annotate output to /tmp/callgrind_annotate.out +with open("/tmp/callgrind_annotate.out", "w") as output: + callgrind_annotate = subprocess.run(["callgrind_annotate", + "/tmp/callgrind.data"], + stdout=output, + stderr=subprocess.PIPE) + if callgrind_annotate.returncode: + os.unlink('/tmp/callgrind.data') + output.close() + os.unlink('/tmp/callgrind_annotate.out') + sys.exit(callgrind_annotate.stderr.decode("utf-8")) + +# Read the callgrind_annotate output to callgrind_data[] +callgrind_data = [] +with open('/tmp/callgrind_annotate.out', 'r') as data: + callgrind_data = data.readlines() + +# Line number with the total number of instructions +total_instructions_line_number = 20 + +# Get the total number of instructions +total_instructions_line_data = callgrind_data[total_instructions_line_number] +total_number_of_instructions = total_instructions_line_data.split(' ')[0] +total_number_of_instructions = int( + total_number_of_instructions.replace(',', '')) + +# Line number with the top function +first_func_line = 25 + +# Number of functions recorded by callgrind, last two lines are always empty +number_of_functions = len(callgrind_data) - first_func_line - 2 + +# Limit the number of top functions to "top" +number_of_top_functions = (top if number_of_functions > + top else number_of_functions) + +# Store the data of the top functions in top_functions[] +top_functions = callgrind_data[first_func_line: + first_func_line + number_of_top_functions] + +# Print table header +print('{:>4} {:>10} {:<30} {}\n{} {} {} {}'.format('No.', + 'Percentage', + 'Function Name', + 'Source File', + '-' * 4, + '-' * 10, + '-' * 30, + '-' * 30, + )) + +# Print top N functions +for (index, function) in enumerate(top_functions, start=1): + function_data = function.split() + # Calculate function percentage + function_instructions = float(function_data[0].replace(',', '')) + function_percentage = (function_instructions / + total_number_of_instructions)*100 + # Get function name and source files path + function_source_file, function_name = function_data[1].split(':') + # Print extracted data + print('{:>4} {:>9.3f}% {:<30} {}'.format(index, + round(function_percentage, 3), + function_name, + function_source_file)) + +# Remove intermediate files +os.unlink('/tmp/callgrind.data') +os.unlink('/tmp/callgrind_annotate.out') diff --git a/scripts/performance/topN_perf.py b/scripts/performance/topN_perf.py new file mode 100755 index 0000000000..07be195fc8 --- /dev/null +++ b/scripts/performance/topN_perf.py @@ -0,0 +1,149 @@ +#!/usr/bin/env python3 + +# Print the top N most executed functions in QEMU using perf. +# Syntax: +# topN_perf.py [-h] [-n] -- \ +# [] \ +# [] +# +# [-h] - Print the script arguments help message. +# [-n] - Specify the number of top functions to print. +# - If this flag is not specified, the tool defaults to 25. +# +# Example of usage: +# topN_perf.py -n 20 -- qemu-arm coulomb_double-arm +# +# This file is a part of the project "TCG Continuous Benchmarking". +# +# Copyright (C) 2020 Ahmed Karaman +# Copyright (C) 2020 Aleksandar Markovic +# +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation, either version 2 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see . + +import argparse +import os +import subprocess +import sys + + +# Parse the command line arguments +parser = argparse.ArgumentParser( + usage='topN_perf.py [-h] [-n] -- ' + ' [] ' + ' []') + +parser.add_argument('-n', dest='top', type=int, default=25, + help='Specify the number of top functions to print.') + +parser.add_argument('command', type=str, nargs='+', help=argparse.SUPPRESS) + +args = parser.parse_args() + +# Extract the needed variables from the args +command = args.command +top = args.top + +# Insure that perf is installed +check_perf_presence = subprocess.run(["which", "perf"], + stdout=subprocess.DEVNULL) +if check_perf_presence.returncode: + sys.exit("Please install perf before running the script!") + +# Insure user has previllage to run perf +check_perf_executability = subprocess.run(["perf", "stat", "ls", "/"], + stdout=subprocess.DEVNULL, + stderr=subprocess.DEVNULL) +if check_perf_executability.returncode: + sys.exit( +""" +Error: +You may not have permission to collect stats. + +Consider tweaking /proc/sys/kernel/perf_event_paranoid, +which controls use of the performance events system by +unprivileged users (without CAP_SYS_ADMIN). + + -1: Allow use of (almost) all events by all users + Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK + 0: Disallow ftrace function tracepoint by users without CAP_SYS_ADMIN + Disallow raw tracepoint access by users without CAP_SYS_ADMIN + 1: Disallow CPU event access by users without CAP_SYS_ADMIN + 2: Disallow kernel profiling by users without CAP_SYS_ADMIN + +To make this setting permanent, edit /etc/sysctl.conf too, e.g.: + kernel.perf_event_paranoid = -1 + +* Alternatively, you can run this script under sudo privileges. +""" +) + +# Run perf record +perf_record = subprocess.run((["perf", "record", "--output=/tmp/perf.data"] + + command), + stdout=subprocess.DEVNULL, + stderr=subprocess.PIPE) +if perf_record.returncode: + os.unlink('/tmp/perf.data') + sys.exit(perf_record.stderr.decode("utf-8")) + +# Save perf report output to /tmp/perf_report.out +with open("/tmp/perf_report.out", "w") as output: + perf_report = subprocess.run( + ["perf", "report", "--input=/tmp/perf.data", "--stdio"], + stdout=output, + stderr=subprocess.PIPE) + if perf_report.returncode: + os.unlink('/tmp/perf.data') + output.close() + os.unlink('/tmp/perf_report.out') + sys.exit(perf_report.stderr.decode("utf-8")) + +# Read the reported data to functions[] +functions = [] +with open("/tmp/perf_report.out", "r") as data: + # Only read lines that are not comments (comments start with #) + # Only read lines that are not empty + functions = [line for line in data.readlines() if line and line[0] + != '#' and line[0] != "\n"] + +# Limit the number of top functions to "top" +number_of_top_functions = top if len(functions) > top else len(functions) + +# Store the data of the top functions in top_functions[] +top_functions = functions[:number_of_top_functions] + +# Print table header +print('{:>4} {:>10} {:<30} {}\n{} {} {} {}'.format('No.', + 'Percentage', + 'Name', + 'Invoked by', + '-' * 4, + '-' * 10, + '-' * 30, + '-' * 25)) + +# Print top N functions +for (index, function) in enumerate(top_functions, start=1): + function_data = function.split() + function_percentage = function_data[0] + function_name = function_data[-1] + function_invoker = ' '.join(function_data[2:-2]) + print('{:>4} {:>10} {:<30} {}'.format(index, + function_percentage, + function_name, + function_invoker)) + +# Remove intermediate files +os.unlink('/tmp/perf.data') +os.unlink('/tmp/perf_report.out') diff --git a/target/mips/kvm.c b/target/mips/kvm.c index 96cfa10cf2..72637a1e02 100644 --- a/target/mips/kvm.c +++ b/target/mips/kvm.c @@ -21,10 +21,12 @@ #include "qemu/main-loop.h" #include "qemu/timer.h" #include "sysemu/kvm.h" +#include "sysemu/kvm_int.h" #include "sysemu/runstate.h" #include "sysemu/cpus.h" #include "kvm_mips.h" #include "exec/memattrs.h" +#include "hw/boards.h" #define DEBUG_KVM 0 @@ -1270,3 +1272,27 @@ int kvm_arch_msi_data_to_gsi(uint32_t data) { abort(); } + +int mips_kvm_type(MachineState *machine, const char *vm_type) +{ +#if defined(KVM_CAP_MIPS_VZ) || defined(KVM_CAP_MIPS_TE) + int r; + KVMState *s = KVM_STATE(machine->accelerator); +#endif + +#if defined(KVM_CAP_MIPS_VZ) + r = kvm_check_extension(s, KVM_CAP_MIPS_VZ); + if (r > 0) { + return KVM_VM_MIPS_VZ; + } +#endif + +#if defined(KVM_CAP_MIPS_TE) + r = kvm_check_extension(s, KVM_CAP_MIPS_TE); + if (r > 0) { + return KVM_VM_MIPS_TE; + } +#endif + + return -1; +} diff --git a/target/mips/kvm_mips.h b/target/mips/kvm_mips.h index 1e4014792d..171d53dbe1 100644 --- a/target/mips/kvm_mips.h +++ b/target/mips/kvm_mips.h @@ -12,6 +12,8 @@ #ifndef KVM_MIPS_H #define KVM_MIPS_H +#include "cpu.h" + /** * kvm_mips_reset_vcpu: * @cpu: MIPSCPU @@ -23,4 +25,13 @@ void kvm_mips_reset_vcpu(MIPSCPU *cpu); int kvm_mips_set_interrupt(MIPSCPU *cpu, int irq, int level); int kvm_mips_set_ipi_interrupt(MIPSCPU *cpu, int irq, int level); +#ifdef CONFIG_KVM +int mips_kvm_type(MachineState *machine, const char *vm_type); +#else +static inline int mips_kvm_type(MachineState *machine, const char *vm_type) +{ + return 0; +} +#endif + #endif /* KVM_MIPS_H */