stress-ng工具
一、stress-ng 工具简介
stress-ng 是一款功能强大的 Linux 系统压力测试工具，被誉为系统稳定性验证的“瑞士军刀”。它是传统 stress 工具的升级版，由 Colin Ian King 开发维护，目前已成为 Linux 系统压力测试的事实标准。
1. 核心特点
stress-ng 支持超过 200 种压力测试场景（ stressors），可以对计算机系统的几乎所有物理子系统和操作系统内核接口进行全面压力测试。它的主要优势包括：
(1) 全面的压力类型：支持 CPU 浮点运算、内存分配/泄漏、磁盘随机读写、网络套接字压力等.
(2) 细粒度控制：可精确调整压力强度（如 CPU 负载百分比）、并发数、测试时长和资源限制.
(3) 丰富的统计输出：支持输出测试过程中的资源使用率、延迟、错误数等指标
(4) 跨平台支持：支持 Linux 主流发行版、macOS、BSD 等系统
(5) 多架构适配：原生支持 x86、ARM（ARMv7/ARMv8）、RISC-V 等主流架构
2. 安装方式
根据不同系统，有多种安装方法：
(1) Ubuntu
# Ubuntu/Debian
sudo apt install stress-ng
源码编译安装：
wget https://github.com/ColinIanKing/stress-ng/archive/refs/tags/V0.17.04.tar.gz tar -xzf V0.17.04.tar.gz cd stress-ng-0.17.04 make && sudo make install
(2) 交叉编译
对于 ARM 开发板等嵌入式场景，可以通过配置交叉编译器进行编译
export CC=aarch64-linux-gnu-gcc
make
(3) 交叉静态编译
//法1: make clean make CC=aarch64-linux-gnu-gcc LDFLAGS=-static -j$(nproc) //法2: //部分版本的 Makefile 支持 STATIC=1 选项： make clean STATIC=1 make CC=aarch64-linux-gnu-gcc -j$(nproc)
3. 常用压测命令
(1) CPU 压力测试
# 使用 4 个 CPU 核心进行压力测试，持续 60 秒 stress-ng --cpu 4 --timeout 60s # 指定 CPU 负载为 100% stress-ng --cpu 4 --cpu-load 100 --timeout 60s # 启2个worker做圆周率算法进行 CPU 压力测试 stress-ng -c 2 --cpu-method pi # 启2个worker从迭代使用30多种不同的压力算法，包括pi, crc16, fft等等。 stress-ng -c 2 --cpu-method all # 启2个worker调用socket相关函数产生压力 stress-ng --sock 2
(2) 内存压力测试
# 创建 4 个虚拟机进程，每个分配 1GB 内存 stress-ng --vm 4 --vm-bytes 1G --timeout 60s
(3) 磁盘 I/O 压力测试
# 测试磁盘 I/O 和文件系统性能 stress-ng --hdd 1 --hdd-bytes 1G
(4) 综合压力测试
# 同时进行 CPU、I/O 和内存压力测试 stress-ng --cpu 4 --io 2 --vm 2 --timeout 60s
(5) 高级用法
# 将压力指定到特定 CPU 核心 stress-ng --sock 4 --taskset 0,2-3,6 # 实时监控 CPU 温度 stress-ng --cpu 4 --tz -t 60 # 输出详细性能指标 stress-ng --matrix 0 -t 60s --metrics-brief
4. 典型应用场景
(1) 系统稳定性验证：通过模拟高负载场景，检测系统在极端条件下的稳定性。例如，某电商平台在 618 大促前使用 stress-ng 进行全链路压测，成功发现了 CPU 瓶颈和内存分配问题。
(2) 硬件故障排查：stress-ng 能有效发现硬件设计缺陷，如散热问题导致的 thermal throttling。通过 --tz 参数实时监测传感器温度，可以快速定位硬件问题。
(3) 嵌入式开发验证：在 ARM 开发板上模拟实车测试环境，验证模型在资源受限时的性能表现。有案例显示，通过 stress-ng 加压测试，成功将工业级 EtherCAT 主站方案的周期抖动从 50μs 降低到 2.
(4) 性能基线建立：通过压测确定系统性能上限，为后续优化提供数据支撑。
5. 注意事项
某些测试可能导致系统温度过高，在硬件设计不良的系统上需谨慎使用。
以 root 权限运行时，系统会调整内存设置，使压力测试器在低内存情况下无法被杀死。
建议配合 htop、vmstat、mpstat 等监控工具实时观察系统状态。
测试时应关注 bogo ops（完成的操作次数）、用户时间、系统时间等关键指标。
stress-ng 不仅能帮助运维和测试人员全面评估系统稳定性、发现硬件设计缺陷，还能为性能调优提供可靠的数据支撑，是 Linux 系统压测不可或缺的工具。
6. 参考
【Stress-NG】Linux压测工具-编译和测试高通平台芯片
Linux系统优化与调试之二：压力测试神器stress-ng
二、stress-ng 帮助信息
/data/local/tmp # ./stress-ng --help stress-ng, version 0.19.02 (gcc 11.4.0, aarch64 Linux 6.1.115-android14-11-9a12e4e1f24f65) Usage: stress-ng [OPTION [ARG]] General control options: --abort abort all stressors if any stressor fails --aggressive enable all aggressive options -a N, --all N start N workers of each stress test -b N, --backoff N wait of N microseconds before work starts --change-cpu force child processes to use different CPU to that of parent --class name specify a class of stressors, use with --sequential -n, --dry-run do not run --ftrace enable kernel function call tracing -h, --help show help --ignite-cpu alter kernel controls to make CPU run hot --interrupts check for error interrupts --ionice-class C specify ionice class (idle, besteffort, realtime) --ionice-level L specify ionice level (0 max, 7 min) --iostat S show I/O statistics every S seconds -j, --job jobfile run the named jobfile --keep-files do not remove files or directories -k, --keep-name keep stress worker names to be 'stress-ng' -K, --klog-check check kernel message log for errors --ksm enable kernel samepage merging --log-brief less verbose log messages --log-file filename log messages to a log file --log-lockless log messages without message locking --maximize enable maximum stress options --max-fd N set maximum file descriptor limit --mbind set NUMA memory binding to specific nodes -M, --metrics print pseudo metrics of activity --metrics-brief enable metrics and only show non-zero results --minimize enable minimal stress options --no-madvise don't use random madvise options for each mmap --no-oom-adjust disable all forms of out-of-memory score adjustments --no-rand-seed seed random numbers with the same constant --oom-avoid Try to avoid stressors from being OOM'd --oom-avoid-bytes N Number of bytes free to stop further memory allocations --oomable Do not respawn a stressor if it gets OOM'd --page-in touch allocated pages that are not in core --parallel N synonym for 'all N' --pathological enable stressors that are known to hang a machine --perf display perf statistics --permute N run permutations of stressors with N stressors per permutation -q, --quiet quiet output -r, --random N start N random workers --rapl report RAPL power domain measurements over entire run (Linux x86 only) --raplstat S show RAPL power domain stats every S seconds (Linux x86 only) --sched type set scheduler type --sched-prio N set scheduler priority level N --sched-period N set period for SCHED_DEADLINE to N nanosecs (Linux only) --sched-runtime N set runtime for SCHED_DEADLINE to N nanosecs (Linux only) --sched-deadline N set deadline for SCHED_DEADLINE to N nanosecs (Linux only) --sched-reclaim set reclaim cpu bandwidth for deadline scheduler (Linux only) --seed N set the random number generator seed with a 64 bit value --sequential N run all stressors one by one, invoking N of them --skip-silent silently skip unimplemented stressors --smart show changes in S.M.A.R.T. data --sn use scientific notation for metrics --status S show stress-ng progress status every S seconds --stderr all output to stderr --stdout all output to stdout (now the default) --stressor-time log start and end run times of each stressor --stressors show available stress tests --syslog log messages to the syslog --taskset use specific CPUs (set CPU affinity) --temp-path path specify path for temporary directories and files --thermalstat S show CPU and thermal load stats every S seconds --thrash force all pages in causing swap thrashing -t N, --timeout T timeout after T seconds --timer-slack N set slack slack to N nanoseconds, 0 for default --times show run time summary at end of the run --timestamp timestamp log output --tz collect temperatures from thermal zones (Linux only) -v, --verbose verbose output --verify verify results (not available on all tests) --verifiable show stressors that enable verification via --verify -V, --version show version --vmstat S show memory and process statistics every S seconds --vmstat-units U vmstat memory units, one of k | m | g | t | p | e -x, --exclude list list of stressors to exclude (not run) -w, --with list list of stressors to invoke (use with --seq or --all) -Y, --yaml file output results to YAML formatted file Stressor specific options: --access N start N workers that stress file access permissions --access-ops N stop after N file access bogo operations --acl N start N workers exercising valid ACL file mode bits --acl-rand randomize ordering of ACL file mode tests --acl-ops N stop acl workers after N bogo operations --af-alg N start N workers that stress AF_ALG socket domain --af-alg-dump dump internal list from /proc/crypto to stdout --af-alg-ops N stop after N af-alg bogo operations --affinity N start N workers that rapidly change CPU affinity --affinity-delay D delay in nanoseconds between affinity changes --affinity-ops N stop after N affinity bogo operations --affinity-pin keep per stressor threads pinned to same CPU --affinity-rand change affinity randomly rather than sequentially --affinity-sleep N sleep in nanoseconds between affinity changes --aio N start N workers that issue async I/O requests --aio-ops N stop after N bogo async I/O requests --aio-requests N number of async I/O requests per worker --aiol N start N workers that exercise Linux async I/O --aiol-ops N stop after N bogo Linux aio async I/O requests --aiol-requests N number of Linux aio async I/O requests per worker --alarm N start N workers exercising alarm timers --alarm-ops N stop after N alarm bogo operations --apparmor start N workers exercising AppArmor interfaces --apparmor-ops N stop after N bogo AppArmor worker bogo operations --atomic start N workers exercising GCC atomic operations --atomic-ops stop after N bogo atomic bogo operations --bad-altstack N start N workers exercising bad signal stacks --bad-altstack-ops N stop after N bogo signal stack SIGSEGVs --bad-ioctl N start N stressors that perform illegal ioctls on devices --bad-ioctl-ops N stop after N bad ioctl bogo operations --bad-ioctl-method M method of selecting ioctl command [ random | inc | random-inc | stride ] --besselmath N start N workers exercising bessel math functions --besselmath-ops N stop after N besselmath bogo bessel math operations --besselmath-method M select bessel math function to exercise -B N, --bigheap N start N workers that grow the heap using realloc() --bigheap-bytes N grow heap up to N bytes in total --bigheap-growth N grow heap by N bytes per iteration --bigheap-mlock attempt to mlock newly mapped pages --bigheap-ops N stop after N bogo bigheap operations --binderfs N start N workers exercising binderfs --binderfs-ops N stop after N bogo binderfs operations --bind-mount N start N workers exercising bind mounts --bind-mount-ops N stop after N bogo bind mount operations --bitonicsort N start N workers bitonic sorting 32 bit random integers --bitonicsort-ops N stop after N bitonic sort bogo operations --bitonicsort-size N number of 32 bit integers to sort --bitops N start N workers that perform CPU only loading --bitops-method M specify stress bitops method M, default is all --bitops-ops N stop after N bitops bogo operations --branch N start N workers that force branch misprediction --branch-ops N stop after N branch misprediction branches --brk N start N workers performing rapid brk calls --brk-bytes N grow brk region up to N bytes in total --brk-mlock attempt to mlock newly mapped brk pages --brk-notouch don't touch (page in) new data segment page --brk-ops N stop after N brk bogo operations --bsearch N start N workers that exercise a binary search --bsearch-method M select bsearch method [ bsearch-libc | bsearch-nonlibc ] --bsearch-ops N stop after N binary search bogo operations --bsearch-size N number of 32 bit integers to bsearch --bubblesort N start N workers heap sorting 32 bit random integers --bubblesort-method M select sort method [ bubblesort-libc | bubblesort-nonlibc --bubblesort-ops N stop after N heap sort bogo operations --bubblesort-size N number of 32 bit integers to sort -C N, --cache N start N CPU cache thrashing workers --cache-size N override the default cache size setting to N bytes --cache-enable-all enable all cache options (fence,flush,sfence,etc..) --cache-fence serialize stores --cache-level N only exercise specified cache --cache-no-affinity do not change CPU affinity --cache-ops N stop after N cache bogo operations --cache-permute permute (mix) cache operations --cache-prefetch prefetch for memory reads/writes --cache-ways N only fill specified number of cache ways --cache-clwb cache line writeback (x86 only) --cachehammer N start N CPU cache thrashing workers --cachehammer-numa move pages to randomly chosen NUMA nodes --cachehammer-ops N stop after N cache bogo operations --cacheline N start N workers that exercise cachelines --cacheline-affinity modify CPU affinity --cacheline-method M use cacheline stressing method M --cacheline-ops N stop after N cacheline bogo operations --cap N start N workers exercising capget --cap-ops N stop cap workers after N bogo capget operations --cgroup N start N workers exercising cgroup mount/read/write/umounts --cgroup-ops N stop after N iterations of cgroup actions --chattr N start N workers thrashing chattr file mode bits --chattr-ops N stop chattr workers after N bogo operations --chdir N start N workers thrashing chdir on many paths --chdir-dirs N select number of directories to exercise chdir on --chdir-ops N stop chdir workers after N bogo chdir operations --chmod N start N workers thrashing chmod file mode bits --chmod-ops N stop chmod workers after N bogo operations --chown N start N workers thrashing chown file ownership --chown-ops N stop chown workers after N bogo operations --chroot N start N workers thrashing chroot --chroot-ops N stop chroot workers after N bogo operations --clock N start N workers thrashing clocks and POSIX timers --clock-ops N stop clock workers after N bogo operations --clone N start N workers that rapidly create and reap clones --clone-max N set upper limit of N clones per worker --clone-ops N stop after N bogo clone operations --close N start N workers that exercise races on close --close-ops N stop after N bogo close operations --context N start N workers exercising user context --context-ops N stop context workers after N bogo operations --copy-file N start N workers that copy file data --copy-file-bytes N specify size of file to be copied --copy-file-ops N stop after N copy bogo operations -c N, --cpu N start N workers that perform CPU only loading -l P, --cpu-load P load CPU by P %, 0=sleep, 100=full load (see -c) --cpu-load-slice S specify time slice during busy load --cpu-method M specify stress cpu method M, default is all --cpu-old-metrics use old CPU metrics instead of normalized metrics --cpu-ops N stop after N cpu bogo operations --cpu-online N start N workers offlining/onlining the CPUs --cpu-online-affinity set CPU affinity to the CPU to be offlined --cpu-online-all attempt to exercise all CPUs include CPU 0 --cpu-online-ops N stop after N offline/online operations --cpu-sched N start N workers that exercise cpu scheduling --cpu-sched-ops N stop after N bogo cpu scheduling operations --crypt N start N workers performing password encryption --crypt-method M select encryption method [ all | MD5 | NT | SHA-1 | SHA-256 | SHA-512 | scrypt | SunMD5 | yescrypt] --crypt-ops N stop after N bogo crypt operations --cyclic N start N cyclic real time benchmark stressors --cyclic-dist N calculate distribution of interval N nanosecs --cyclic-method M specify cyclic method M, default is clock_ns --cyclic-ops N stop after N cyclic timing cycles --cyclic-policy P used rr or fifo scheduling policy --cyclic-prio N real time scheduling priority 1..100 --cyclic-samples N number of latency samples to take --cyclic-sleep N sleep time of real time timer in nanosecs --daemon N start N workers creating multiple daemons --daemon-ops N stop when N daemons have been created --daemon-wait stressor wait for daemon to exit and not init --dccp N start N workers exercising network DCCP I/O --dccp-domain D specify DCCP domain, default is ipv4 --dccp-if I use network interface I, e.g. lo, eth0, etc. --dccp-ops N stop after N DCCP bogo operations --dccp-opts option DCCP data send options [send|sendmsg|sendmmsg] --dccp-port P use DCCP ports P to P + number of workers - 1 --dccp-msgs N number of DCCP messages to send per connection --dekker N start N workers that exercise the Dekker algorithm --dekker-ops N stop after N dekker mutex bogo operations -D N, --dentry N start N dentry thrashing stressors --dentry-ops N stop after N dentry bogo operations --dentry-order O specify unlink order (reverse, forward, stride) --dentries N create N dentries per iteration --dev N start N device entry thrashing stressors --dev-file name specify the /dev/ file to exercise --dev-ops N stop after N device thrashing bogo ops --dev-shm N start N /dev/shm file and mmap stressors --dev-shm-ops N stop after N /dev/shm bogo ops --dfp N start N workers performing decimal floating point math ops --dfp-method M select the decimal floating point method to operate with --dfp-ops N stop after N decimal floating point math bogo operations --dir N start N directory thrashing stressors --dir-dirs N select number of directories to exercise dir on --dir-ops N stop after N directory bogo operations --dirdeep N start N directory depth stressors --dirdeep-bytes N size of files to create per level (see --dirdeep-files) --dirdeep-dirs N create N directories per level --dirdeep-files N create N files per level (see --dirdeep-bytes) --dirdeep-inodes N create a maximum N inodes (N can also be %) --dirdeep-ops N stop after N directory depth bogo operations --dirmany N start N directory file populating stressors --dirmany-bytes specify size of files (default 0) --dirmany-ops N stop after N directory file bogo operations --dnotify N start N workers exercising dnotify events --dnotify-ops N stop dnotify workers after N bogo operations --dup N start N workers exercising dup/close --dup-ops N stop after N dup/close bogo operations --dynlib N start N workers exercising dlopen/dlclose --dynlib-ops N stop after N dlopen/dlclose bogo operations --easy-opcode N start N workers exercising random easy opcodes --easy-opcode-ops N stop after N easy opcode bogo operations --efivar N start N workers that read EFI variables --efivar-ops N stop after N EFI variable bogo read operations --eigen N start N workers exercising eigen operations --eigen-method M specify eigen stress method M, default is all --eigen-ops N stop after N matrix bogo operations --eigen-size N specify the size of the N x N eigen --enosys N start N workers that call non-existent system calls --enosys-ops N stop after N enosys bogo operations --env N start N workers setting environment vars --env-ops N stop after N env bogo operations --epoll N start N workers doing epoll handled socket activity --epoll-domain D specify socket domain, default is unix --epoll-ops N stop after N epoll bogo operations --epoll-port P use socket ports P upwards --epoll-sockets N specify maximum number of open sockets --eventfd N start N workers stressing eventfd read/writes --eventfs-nonblock poll with non-blocking I/O on eventfd fd --eventfd-ops N stop eventfd workers after N bogo operations --exec N start N workers spinning on fork() and exec() --exec-fork-method M select exec fork method: clone fork spawn vfork --exec-max P create P workers per iteration, default is 4096 --exec-method M select exec method: all, execve, execveat --exec-no-pthread do not use pthread_create --exec-ops N stop after N exec bogo operations --exit-group N start N workers that exercise exit_group --exit-group-ops N stop exit_group workers after N bogo exit_group loops --expmath N start N workers exercising exponential math functions --expmath-ops N stop after N expmath bogo exponential math operations --expmath-method M select exponential math function to exercise --factor N start N workers performing large integer factorization --factor-digits N specific number of digits of number to factor --factor-ops N stop after N factorisation operations --fallocate N start N workers fallocating 16MB files --fallocate-bytes N specify size of file to allocate --fallocate-ops N stop after N fallocate bogo operations --fanotify N start N workers exercising fanotify events --fanotify-ops N stop fanotify workers after N bogo operations --far-branch N start N far branching workers --far-branch-flush periodically flush instruction cache --far-branch-ops N stop after N far branching bogo operations --far-branch-pageout soft offline and/or pageout object code pages --far-branch-pages N number of pages to populate with functions --fault N start N workers producing page faults --fault-ops N stop after N page fault bogo operations --fcntl N start N workers exercising fcntl commands --fcntl-ops N stop after N fcntl bogo operations --fd-abuse N start N workers abusing file descriptors --fd-abuse N stop fd-abuse after bogo operations --fd-fork N start N workers exercising dup/fork/close --fd-fork-file file select file to dup [ null, random, stdin, stdout, zero ] --fd-fork-fds N set maximum number of file descriptors to use --fd-fork-ops N stop after N dup/fork/close bogo operations --fd-race N start N workers sending file descriptors over sockets --fd-race-ops N stop after N fd_race bogo operations --fd-race-dev race on /dev/* files --fd-race-proc race on /proc/* files --fibsearch N start N workers that exercise a Fibonacci search --fibsearch-ops N stop after N Fibonacci search bogo operations --fibsearch-size N number of 32 bit integers to Fibonacci search --fiemap N start N workers exercising the FIEMAP ioctl --fiemap-bytes N specify size of file to fiemap --fiemap-ops N stop after N FIEMAP ioctl bogo operations --fifo N start N workers exercising fifo I/O --fifo-data-size N set fifo read/write size in bytes (default 8) --fifo-ops N stop after N fifo bogo operations --fifo-readers N number of fifo reader stressors to start --file-ioctl N start N workers exercising file specific ioctls --file-ioctl-ops N stop after N file ioctl bogo operations --filename N start N workers exercising filenames --filename-ops N stop after N filename bogo operations --filename-opts opt specify allowed filename options --filerace N start N workers that attempt to race file system calls --filerace-ops N stop after N filerace bogo operations --flipflop N start N workers exercising flipflop --flipflop-bits N number of bits to be exercised by 2 x N x pthreads --flipflop-taskset1 S1 list of CPUs to pin N ping threads to --flipflop-taskset2 S2 list of CPUs to pin N pong threads to --flipflop-ops N stop after N flipflop bogo operations --flock N start N workers locking a single file --flock-ops N stop after N flock bogo operations --flushcache N start N CPU instruction + data cache flush workers --flushcache-ops N stop after N flush cache bogo operations --fma N start N workers performing floating point multiply-add ops --fma-ops N stop after N floating point multiply-add bogo operations --fma-libc use fma libc fused multiply-add helpers -f N, --fork N start N workers spinning on fork() and exit() --fork-max P create P forked processes per iteration, default is 1 --fork-ops N stop after N fork bogo operations --fork-pageout force pageout memory resident pages --fork-unmap forcibly unmap unused shared library pages (dangerous) --fork-vm enable extra virtual memory pressure --forkheavy N start N workers that rapidly fork and reap resource heavy processes --forkheavy-allocs N attempt to allocate N x resources --forkheavy-mlock attempt to mlock newly mapped pages --forkheavy-ops N stop after N bogo fork operations --forkheavy-procs N attempt to fork N processes --fp N start N workers performing floating point math ops --fp-method M select the floating point method to operate with --fp-ops N stop after N floating point math bogo operations --fp-error N start N workers exercising floating point errors --fp-error-ops N stop after N fp-error bogo operations --fpunch N start N workers punching holes in a 16MB file --fpunch-bytes N size of file being punched --fpunch-ops N stop after N punch bogo operations --fractal N start N workers performing large integer fractalization --fractal-iterations N number of iterations --fractal-method M fractal method [ mandelbrot | julia ] --fractal-ops N stop after N fractalisation operations --fractal-xsize N width of fractal --fractal-ysize N height of fractal --fsize N start N workers exercising file size limits --fsize-ops N stop after N fsize bogo operations --fstat N start N workers exercising fstat on files --fstat-dir path fstat files in the specified directory --fstat-ops N stop after N fstat bogo operations --full N start N workers exercising /dev/full --full-ops N stop after N /dev/full bogo I/O operations --funccall N start N workers exercising 1 to 9 arg functions --funccall-method M select function call method M --funccall-ops N stop after N function call bogo operations --funcret N start N workers exercising function return copying --funcret-method M select method of exercising a function return type --funcret-ops N stop after N function return bogo operations --futex N start N workers exercising a fast mutex --futex-ops N stop after N fast mutex bogo operations --get N start N workers exercising the get*() system calls --get-ops N stop after N get bogo operations --get-slow-sync synchronize get*() system amongst N workers --getdent N start N workers reading directories using getdents --getdent-ops N stop after N getdents bogo operations --getrandom N start N workers fetching random data via getrandom() --getrandom-ops N stop after N getrandom bogo operations --goto N start N workers that exercise heavy branching --goto-direction D select goto direction forward, backward, random --goto-ops N stop after 1024 x N goto bogo operations --gpu N start N GPU worker --gpu-devnode name specify CPU device node name --gpu-frag N specify shader core usage per pixel --gpu-ops N stop after N gpu render bogo operations --gpu-tex-size N specify upload texture NxN --gpu-upload N specify upload texture N times per frame --gpu-xsize X specify framebuffer size x --gpu-ysize Y specify framebuffer size y --handle N start N workers exercising name_to_handle_at --handle-ops N stop after N handle bogo operations --hash N start N workers that exercise various hash functions --hash-method M specify stress hash method M, default is all --hash-ops N stop after N hash bogo operations -d N, --hdd N start N workers spinning on write()/unlink() --hdd-bytes N write N bytes per hdd worker (default is 1GB) --hdd-ops N stop after N hdd bogo operations --hdd-opts list specify list of various stressor options --hdd-write-size N set the default write size to N bytes --heapsort N start N workers heap sorting 32 bit random integers --heapsort-method M select sort method [ heapsort-libc | heapsort-nonlibc --heapsort-ops N stop after N heap sort bogo operations --heapsort-size N number of 32 bit integers to sort --hrtimers N start N workers that exercise high resolution timers --hrtimers-adjust adjust rate to try and maximum timer rate --hrtimers-ops N stop after N bogo high-res timer bogo operations --hsearch N start N workers that exercise a hash table search --hsearch-ops N stop after N hash search bogo operations --hsearch-size N number of integers to insert into hash table --hyperbolic N start N workers exercisinhg hyberbolic functions --hyperbolic-ops N stop after N hyperbolic bogo hyperbolic function operations --hyperbolic-method M select hyperbolic function to exercise --icache N start N CPU instruction cache thrashing workers --icache-ops N stop after N icache bogo operations --icmp-flood N start N ICMP packet flood workers --icmp-flood-ops N stop after N ICMP bogo operations (ICMP packets) --icmp-flood-max-size use randomized packet sizes up to 65535 bytes --idle-page N start N idle page scanning workers --idle-page-ops N stop after N idle page scan bogo operations --inode-flags N start N workers exercising various inode flags --inode-flags-ops N stop inode-flags workers after N bogo operations --inotify N start N workers exercising inotify events --inotify-ops N stop inotify workers after N bogo operations --insertionsort N start N workers heap sorting 32 bit random integers --insertionsort-ops N stop after N heap sort bogo operations --insertionsort-size N number of 32 bit integers to sort --intmath N start N workers that exercising signed integer math operations --intmath-fast use int_fast*_t fast integers --intmath-method select the method of integer math operation --intmath-ops N stop after N bogo signed integer math operations -i N, --io N start N workers spinning on sync() --io-ops N stop sync I/O after N io bogo operations --iomix N start N workers that have a mix of I/O operations --iomix-bytes N write N bytes per iomix worker (default is 1GB) --iomix-ops N stop iomix workers after N iomix bogo operations --ioport N start N workers exercising port I/O --ioport-ops N stop ioport workers after N port bogo operations --ioport-opts O option to select ioport access [ in | out | inout ] --ioprio N start N workers exercising set/get iopriority --ioprio-ops N stop after N io bogo iopriority operations --io-uring N start N workers that issue io-uring I/O requests --io-uring-entries N specify number if io-uring ring entries --io-uring-ops N stop after N bogo io-uring I/O requests --io-uring-rand enable randomized io-uring I/O request ordering --ipsec-mb N start N workers exercising the IPSec MB encoding --ipsec-mb-feature F specify CPU feature F --ipsec-mb-jobs N specify number of jobs to run per round (default 1) --ipsec-mb-method M specify crypto/integrity method --ipsec-mb-ops N stop after N ipsec bogo encoding operations --itimer N start N workers exercising interval timers --itimer-freq F set the itimer frequency, limited by jiffy clock rate --itimer-ops N stop after N interval timer bogo operations --itimer-rand enable random interval timer frequency --jpeg N start N workers that burn cycles with no-ops --jpeg-height N image height in pixels --jpeg-image type image type: one of brown, flat, gradient, noise, plasma or xstripes --jpeg-ops N stop after N jpeg bogo no-op operations --jpeg-quality Q compression quality 1 (low) .. 100 (high) --jpeg-width N image width in pixels --judy N start N workers that exercise a judy array search --judy-ops N stop after N judy array search bogo operations --judy-size N number of 32 bit integers to insert into judy array --kcmp N start N workers exercising kcmp --kcmp-ops N stop after N kcmp bogo operations --key N start N workers exercising key operations --key-ops N stop after N key bogo operations --kill N start N workers killing with SIGUSR1 --kill-ops N stop after N kill bogo operations --klog N start N workers exercising kernel syslog interface --klog-ops N stop after N klog bogo operations --kvm N start N workers exercising /dev/kvm --kvm-ops N stop after N kvm create/run/destroy operations --l1cache N start N CPU level 1 cache thrashing workers --l1cache-line-size N specify level 1 cache line size --l1cache-method M l1 cache thrashing method: forward, reverse, random --l1cache-mlock attempt to mlock memory --l1cache-sets N specify level 1 cache sets --l1cache-size N specify level 1 cache size --l1cache-ways N only fill specified number of cache ways --landlock N start N workers stressing landlock file operations --landlock-ops N stop after N landlock bogo operations --lease N start N workers holding and breaking a lease --lease-breakers N number of lease breaking workers to start --lease-ops N stop after N lease bogo operations --led N start N workers that read and set LED settings --led-ops N stop after N LED bogo operations --link N start N workers creating hard links --link-ops N stop after N link bogo operations --link-sync enable sync'ing after linking/unlinking --list N start N workers that exercise list structures --list-method M select list method: all, circleq, list, slist, slistt, stailq, tailq --list-ops N stop after N bogo list operations --list-size N N is the number of items in the list --llc-affinity N start N workers exercising low level cache over all CPUs --llc-affinity-mlock attempt to mlock pages into memory --llc-affinity-numa bind memory mappings to randomly selected NUMA nodes --llc-affinity-ops N stop after N low-level-cache bogo operations --loadavg N start N workers that create a large load average --loadavg-ops N stop load average workers after N bogo operations --loadavg-max N set upper limit on number of pthreads to create --locka N start N workers locking a file via advisory locks --locka-ops N stop after N locka bogo operations --lockbus N start N workers locking a memory increment --lockbus-nosplit disable split locks --lockbus-ops N stop after N lockbus bogo operations --lockf N start N workers locking a single file via lockf --lockf-nonblock don't block if lock cannot be obtained, re-try --lockf-ops N stop after N lockf bogo operations --lockmix N start N workers locking a file via flock, locka, lockf and ofd locks --lockmix-ops N stop after N lockmix bogo operations --lockofd N start N workers using open file description locking --lockofd-ops N stop after N lockofd bogo operations --logmath N start N workers exercising logarithmic math functions --logmath-ops N stop after N logmath bogo logarithmic math operations --logmath-method M select logarithmic math function to exercise --longjmp N start N workers exercising setjmp/longjmp --longjmp-ops N stop after N longjmp bogo operations --loop N start N workers exercising loopback devices --loop-ops N stop after N bogo loopback operations --lsearch N start N workers that exercise a linear search --lsearch-method M select lsearch method [ lsearch-libc | lsearch-nonlibc ] --lsearch-ops N stop after N linear search bogo operations --lsearch-size N number of 32 bit integers to lsearch --lsm N start N workers that exercise lsm kernel system calls --lsm-ops N stop after N lsm bogo operations --madvise N start N workers exercising madvise on memory --madvise-ops N stop after N bogo madvise operations --madvise-hwpoison enable hardware page poisoning (disabled by default) --malloc N start N workers exercising malloc/realloc/free --malloc-bytes N allocate up to N bytes per allocation --malloc-max N keep up to N allocations at a time --malloc-mlock attempt to mlock pages into memory --malloc-ops N stop after N malloc bogo operations --malloc-pthreads N number of pthreads to run concurrently --malloc-thresh N threshold where malloc uses mmap instead of sbrk --malloc-touch touch pages force pages to be populated --malloc-zerofree zero free'd memory --malloc-trim enable malloc trimming --matrix N start N workers exercising matrix operations --matrix-method M specify matrix stress method M, default is all --matrix-ops N stop after N matrix bogo operations --matrix-size N specify the size of the N x N matrix --matrix-yx matrix operation is y by x instead of x by y --matrix-3d N start N workers exercising 3D matrix operations --matrix-3d-method M specify 3D matrix stress method M, default is all --matrix-3d-ops N stop after N 3D matrix bogo operations --matrix-3d-size N specify the size of the N x N x N matrix --matrix-3d-zyx matrix operation is z by y by x instead of x by y by z --mcontend N start N workers that produce memory contention --mcontend-ops N stop memory contention workers after N bogo-ops --mcontend-numa bind memory mappings to randomly selected NUMA nodes --membarrier N start N workers performing membarrier system calls --membarrier-ops N stop after N membarrier bogo operations --memcpy N start N workers performing memory copies --memcpy-method M set memcpy method (M = all, libc, builtin, naive..) --memcpy-ops N stop after N memcpy bogo operations --memfd N start N workers allocating memory with memfd_create --memfd-bytes N allocate N bytes for each stress iteration --memfd-fds N number of memory fds to open per stressors --memfd-madvise add random madvise hints to memfd mapped pages --memfd-mlock attempt to mlock pages into memory --memfd-numa bind memory mappings to randomly selected NUMA nodes --memfd-ops N stop after N memfd bogo operations --memfd-zap-pte enable zap pte bug check (slow) --memhotplug N start N workers that exercise memory hotplug --memhotplug-mmap enable random memory mapping/unmapping --memhotplug-ops N stop after N memory hotplug operations --memrate N start N workers exercised memory read/writes --memrate-bytes N size of memory buffer being exercised --memrate-flush flush cache before each iteration --memrate-method M specify read/write memory exercising method --memrate-ops N stop after N memrate bogo operations --memrate-rd-mbs N read rate from buffer in megabytes per second --memrate-wr-mbs N write rate to buffer in megabytes per second --memthrash N start N workers thrashing a 16MB memory buffer --memthrash-method M specify memthrash method M, default is all --memthrash-ops N stop after N memthrash bogo operations --mergesort N start N workers merge sorting 32 bit random integers --mergesort-method M select sort method [ method-libc | method-nonlibc --mergesort-ops N stop after N merge sort bogo operations --mergesort-size N number of 32 bit integers to sort --metamix N start N workers that have a mix of file metadata operations --metamix-bytes N write N bytes per metamix file (default is 1MB, 16 files per instance) --metamix-ops N stop metamix workers after N metamix bogo operations --mincore N start N workers exercising mincore --mincore-ops N stop after N mincore bogo operations --mincore-random randomly select pages rather than linear scan --min-nanosleep N start N workers performing short sleeps --min-nanosleep-ops N stop after N bogo sleep operations --min-nanosleep-max N maximum nanosleep delay to be used --min-nanosleep-sched P select scheduler policy [ batch, deadline, idle, fifo, other, rr ] --misaligned N start N workers performing misaligned read/writes --misaligned-method M use misaligned memory read/write method --misaligned-ops N stop after N misaligned bogo operations --mknod N start N workers that exercise mknod --mknod-ops N stop after N mknod bogo operations --mlock N start N workers exercising mlock/munlock --mlock-ops N stop after N mlock bogo operations --mlockmany N start N workers exercising many mlock/munlock processes --mlockmany-ops N stop after N mlockmany bogo operations --mlockmany-procs N use N child processes to mlock regions --mmap N start N workers stressing mmap and munmap --mmap-async using asynchronous msyncs for file based mmap --mmap-bytes N mmap and munmap N bytes for each stress iteration --mmap-stressful enable most stressful mmap options (and slowest) --mmap-file mmap onto a file using synchronous msyncs --mmap-madvise enable random madvise on mmap'd region --mmap-mergeable where possible, flag mmap'd pages as mergeable --mmap-mlock attempt to mlock mmap'd pages --mmap-mmap2 use mmap2 instead of mmap (when available) --mmap-mprotect enable mmap mprotect stressing --mmap-numa bind memory mappings to randomly selected NUMA nodes --mmap-odirect enable O_DIRECT on file --mmap-ops N stop after N mmap bogo operations --mmap-osync enable O_SYNC on file --mmap-slow-munmap munmap pages inefficiently one at a time --mmap-write-check set check value in each page and perform sanity read check --mmapaddr N start N workers stressing mmap with random addresses --mmapaddr-mlock attempt to mlock pages into memory --mmapaddr-ops N stop after N mmapaddr bogo operations --mmapcow N start N workers stressing copy-on-write and munmaps --mmapcow-fork force more page copying by reguler process forking --mmapcow-free use madvise(MADV_FREE) on each page before unmapping --mmapcow-mlock lock copy-on-write page into memory before unmapping --mmapcow-numa bind memory mappings to randomly selected NUMA nodes --mmapcow-ops N stop after N mmapcow bogo operations --mmapfiles N start N workers stressing many mmaps and munmaps --mmapfiles-numa bind memory mappings to randomly selected NUMA nodes --mmapfiles-ops N stop after N mmapfiles bogo operations --mmapfiles-populate populate memory mappings --mmapfiles-shared enable shared mappings instead of private mappings --mmapfixed N start N workers stressing mmap with fixed mappings --mmapfixed-mlock attempt to mlock pages into memory --mmapfixed-numa bind memory mappings to randomly selected NUMA nodes --mmapfixed-ops N stop after N mmapfixed bogo operations --mmapfork N start N workers stressing many forked mmaps/munmaps --mmapfork-ops N stop after N mmapfork bogo operations --mmapfork-bytes N mmap and munmap N bytes by workers for each stress iteration --mmaphuge N start N workers stressing mmap with huge mappings --mmaphuge-file perform mappings on a temporary file --mmaphuge-mlock attempt to mlock pages into memory --mmaphuge-mmaps N select number of memory mappings per iteration --mmaphuge-numa bind memory mappings to randomly selected NUMA nodes --mmaphuge-ops N stop after N mmaphuge bogo operations --mmapmany N start N workers stressing many mmaps and munmaps --mmapmany-mlock attempt to mlock pages into memory --mmapmany-numa bind memory mappings to randomly selected NUMA nodes --mmapmany-ops N stop after N mmapmany bogo operations --mmaptorture N start N workers torturing page mappings --mmaptorture-bytes N size of file backed region to be memory mapped --mmaptorture-msync N percentage of pages to be msync'd (default 10%) --mmaptorture-ops N stop after N mmaptorture bogo operations --module N start N workers performing module requests --module-name F use the specified module name F to load. --module-no-unload skip unload of the module after module load --module-no-modver ignore symbol version hashes --module-no-vermag ignore kernel version magic --module-ops N stop after N module bogo operations --monte-carlo N start N workers performing monte-carlo computations --monte-carlo-ops N stop after N monte-carlo operations --monte-carlo-rand R select random number generator [ all | drand48 | getrandom | lcg | pcg32 | mwc32 | mwc64 | random | xorshift ] --monte-carlo-samples N specify number of samples for each computation --monte-carlo-method M select computation method [ pi | e | exp | sin | sqrt | squircle ] --mpfr N start N workers performing multi-precision floating point operations --mpfr-ops N stop after N multi-precision floating point operations --mpfr-precision N specific floating point precision as N bits --mprotect N start N workers exercising mprotect on memory --mprotect-ops N stop after N bogo mprotect operations --mq N start N workers passing messages using POSIX messages --mq-ops N stop mq workers after N bogo messages --mq-size N specify the size of the POSIX message queue --mremap N start N workers stressing mremap --mremap-bytes N mremap N bytes maximum for each stress iteration --mremap-mlock mlock remap pages, force pages to be unswappable --mremap-numa bind memory mappings to randomly selected NUMA nodes --mremap-ops N stop after N mremap bogo operations --mseal N start N workers exercising sealing of mmap'd memory --mseal-ops N stop mseal workers after N bogo mseal operations --msg N start N workers stressing System V messages --msg-ops N stop msg workers after N bogo messages --msg-types N enable N different message types --msg-bytes N set the message size 4..8192 --msync N start N workers syncing mmap'd data with msync --msync-bytes N size of file and memory mapped region to msync --msync-ops N stop msync workers after N bogo msyncs --msyncmany N start N workers stressing msync on many mapped pages --msyncmany-ops N stop after N msyncmany bogo operations --mtx N start N workers exercising ISO C mutex operations --mtx-ops N stop after N ISO C mutex bogo operations --mtx-procs N select the number of concurrent processes --munmap N start N workers stressing munmap --munmap-ops N stop after N munmap bogo operations --mutex N start N workers exercising mutex operations --mutex-affinity change CPU affinity randomly across locks --mutex-ops N stop after N mutex bogo operations --mutex-procs N select the number of concurrent processes --nanosleep N start N workers performing short sleeps --nanosleep-ops N stop after N bogo sleep operations --nanosleep-threads N number of threads to run concurrently (default 8) --nanosleep-method M select nanosleep sleep time method [ all | cstate | random ] --netdev N start N workers exercising netdevice ioctls --netdev-ops N stop netdev workers after N bogo operations --netlink-proc N start N workers exercising netlink process events --netlink-proc-ops N stop netlink-proc workers after N bogo events --netlink-task N start N workers exercising netlink tasks events --netlink-task-ops N stop netlink-task workers after N bogo events --nice N start N workers that randomly re-adjust nice levels --nice-ops N stop after N nice bogo operations --nop N start N workers that burn cycles with no-ops --nop-instr INSTR specify nop instruction to use --nop-ops N stop after N nop bogo no-op operations --null N exercising /dev/null with write, ioctl, lseek, fcntl, fallocate and fdatasync --null-ops N stop after N /dev/null bogo operations --null-write just exercise /dev/null with writing --numa N start N workers stressing NUMA interfaces --numa-bytes N size of memory region to be exercised --numa-ops N stop after N NUMA bogo operations --numa-shuffle-addr shuffle page addresses to move to numa nodes --numa-shuffle-node shuffle numa nodes on numa pages moves --oom-pipe N start N workers exercising large pipes --oom-pipe-ops N stop after N oom-pipe bogo operations --opcode N start N workers exercising random opcodes --opcode-method M set opcode stress method (M = random, inc, mixed, text) --opcode-ops N stop after N opcode bogo operations -o N, --open N start N workers exercising open/close --open-fd open files in /proc/$pid/fd --open-max N specficify maximum number of files to open --open-ops N stop after N open/close bogo operations --pagemove N start N workers that shuffle move pages --pagemove-bytes N size of mmap'd region to exercise page moving in bytes --pagemove-mlock attempt to mlock pages into memory --pagemove-numa bind memory mappings to randomly selected NUMA nodes --pagemove-ops N stop after N page move bogo operations --pageswap N start N workers that swap pages out and in --pageswap-ops N stop after N page swap bogo operations --pci N start N workers that read and mmap PCI regions --pci-dev name specify the pci device 'xxxx:xx:xx.x' to exercise --pci-ops N stop after N PCI bogo operations --personality N start N workers that change their personality --personality-ops N stop after N bogo personality calls --peterson N start N workers that exercise Peterson's algorithm --peterson-ops N stop after N peterson mutex bogo operations --physpage N start N workers performing physical page lookup --physpage-ops N stop after N physical page bogo operations --physpage-mtrr enable modification of MTRR types on physical page --physmmap N start N workers performing /dev/mem physical page mmaps/munmaps --physmmap-ops N stop after N /dev/mem physical page mmap/munmap bogo operations --physmmap-read read data from mapping --pidfd N start N workers exercising pidfd system call --pidfd-ops N stop after N pidfd bogo operations --ping-sock N start N workers that exercises a ping socket --ping-sock-ops N stop after N ping sendto messages -p N, --pipe N start N workers exercising pipe I/O --pipe-data-size N set pipe size of each pipe write to N bytes --pipe-ops N stop after N pipe I/O bogo operations --pipe-size N set pipe size to N bytes --pipe-vmsplice use vmsplice for pipe data transfer -p N, --pipeherd N start N multi-process workers exercising pipes I/O --pipeherd-ops N stop after N pipeherd I/O bogo operations --pipeherd-yield force processes to yield after each write --pkey N start N workers exercising pkey_mprotect --pkey-ops N stop after N bogo pkey_mprotect bogo operations --plugin N start N workers exercising random plugins --plugin-method M set plugin stress method --plugin-ops N stop after N plugin bogo operations --plugin-so file specify plugin shared object file -P N, --poll N start N workers exercising zero timeout polling --poll-fds N use N file descriptors --poll-ops N stop after N poll bogo operations --poll-random-us N use random poll delays of 0..N-1 microseconds --powmath N start N workers exercising power math functions --powmath-ops N stop after N powmath bogo power math operations --powmath-method M select power math function to exercise --prctl N start N workers exercising prctl system call --prctls-ops N stop prctl workers after N bogo prctl operations --prefetch N start N workers exercising memory prefetching --prefetch-l3-size N specify the L3 cache size of the CPU --prefetch-method M specify the prefetch method --prefetch-ops N stop after N bogo prefetching operations --prime N start N workers that find prime numbers --prime-ops N stop after N prime operations --prime-method M method of searching for next prime [ factorial | inc | pwr2 | pwr10 ] --prime-progress show prime progress every 60 seconds (just first stressor instance) --prime-start N value N from where to start computing primes --prio-inv start N workers exercising priority inversion lock operations --prio-inv-ops N stop after N priority inversion lock bogo operations --prio-inv-policy P select scheduler policy [ batch, idle, fifo, other, rr ] --prio-inv-type T pthread priority type [ inherit | none | protect ] --priv-instr N start N workers exercising privileged instruction --priv-instr-ops N stop after N bogo instruction operations --procfs N start N workers reading portions of /proc --procfs-ops N stop procfs workers after N bogo read operations -d N, --pseek N start N workers spinning on seek/write/seek/read --pseek-rand perform random seeks rather than fixed seeks --pseek-io-size N set the default write/read I/O size to N bytes --pthread N start N workers that create multiple threads --pthread-max P create P threads at a time by each worker --pthread-ops N stop pthread workers after N bogo threads created --ptrace N start N workers that trace a child using ptrace --ptrace-ops N stop ptrace workers after N system calls are traced --ptr-chase N start N workers that chase pointers around many nodes --ptr-chase-ops N stop after N bogo pointer chase operations --ptr-chase-pages N N is the number of pages for nodes of pointers --pty N start N workers that exercise pseudoterminals --pty-max N attempt to open a maximum of N ptys --pty-ops N stop pty workers after N pty bogo operations -Q N, --qsort N start N workers qsorting 32 bit random integers --qsort-method M select qsort method [ qsort-libc | qsort_bm ] --qsort-ops N stop after N qsort bogo operations --qsort-size N number of 32 bit integers to sort --quota N start N workers exercising quotactl commands --quota-ops N stop after N quotactl bogo operations --race-sched N start N workers that race cpu affinity --race-sched-ops N stop after N bogo race operations --race-sched-method M method M: all, rand, next, prev, yoyo, randinc --radixsort N start N workers radix sorting random strings --radixsort-method M select sort method [ radixsort-libc | radixsort-nonlibc] --radixsort-ops N stop after N radixsort bogo operations --radixsort-size N number of strings to sort --ramfs N start N workers exercising ramfs mounts --ramfs-size N set the ramfs size in bytes, e.g. 2M is 2MB --ramfs-fill attempt to fill ramfs --ramfs-ops N stop after N bogo ramfs mount operations --randlist N start N workers that exercise random ordered list --randlist-compact reduce mmap and malloc overheads --randlist-items N number of items in the random ordered list --randlist-ops N stop after N randlist bogo no-op operations --randlist-size N size of data in each item in the list --rawdev N start N workers that read a raw device --rawdev-method M specify the rawdev read method to use --rawdev-ops N stop after N rawdev read operations --rawpkt N start N workers exercising raw packets --rawpkt-ops N stop after N raw packet bogo operations --rawpkt-port P use raw packet ports P to P + number of workers - 1 --rawpkt-rxring N setup raw packets with RX ring with N number of blocks, this selects TPACKET_V3 --rawsock N start N workers performing raw socket send/receives --rawsock-ops N stop after N raw socket bogo operations --rawsock-port P use socket P to P + number of workers - 1 --rawudp N start N workers exercising raw UDP socket I/O --rawudp-if I use network interface I, e.g. lo, eth0, etc. --rawudp-ops N stop after N raw socket UDP bogo operations --rawudp-port P use raw socket ports P to P + number of workers - 1 --rdrand N start N workers exercising rdrand (x86 only) --rdrand-ops N stop after N rdrand bogo operations --rdrand-seed use rdseed instead of rdrand --readahead N start N workers exercising file readahead --readahead-bytes N size of file to readahead on (default is 1GB) --readahead-ops N stop after N readahead bogo operations --reboot N start N workers that exercise bad reboot calls --reboot-ops N stop after N bogo reboot operations --regex N start N workers exercise POSIX regular expressions --regex-ops N stop after N regular expression operations --regs N start N workers exercising CPU generic registers --regs-ops N stop after N x 1000 rounds of register shuffling --remap N start N workers exercising page remappings --remap-mlock attempt to mlock pages into memory --remap-ops N stop after N remapping bogo operations --remap-pages N specify N pages to remap (N must be power of 2) -R, --rename N start N workers exercising file renames --rename-ops N stop after N rename bogo operations --resched N start N workers that spawn renicing child processes --resched-ops N stop after N nice bogo nice'd yield operations --resources N start N workers consuming system resources --resources-mlock attempt to mlock pages into memory --resources-ops N stop after N resource bogo operations --revio N start N workers performing reverse I/O --revio-bytes N specify file size (default is 1GB) --revio-ops N stop after N revio bogo operations --revio-opts list specify list of various stressor options --ring-pipe N start N workers exercising a ring of pipes --ring-pipe-num N number of pipes to use --ring-pipe-ops N stop after N ring pipe I/O bogo operations --ring-pipe-size N size of data to be written and read in bytes --ring-pipe-splice use splice instead of read+write --rmap N start N workers that stress reverse mappings --rmap-ops N stop after N rmap bogo operations --rmap N start N workers that stress reverse mappings --rmap-ops N stop after N rmap bogo operations --rotate N start N workers performing rotate ops --rotate-method M select rotate method M --rotate-ops N stop after N rotate bogo operations --rseq N start N workers that exercise restartable sequences --rseq-ops N stop after N bogo restartable sequence operations --rtc N start N workers that exercise the RTC interfaces --rtc-ops N stop after N RTC bogo operations --schedmix N start N workers that exercise a mix of scheduling loads --schedmix-ops N stop after N schedmix bogo operations --schedmix-procs N select number of schedmix child processes 1..64 --schedpolicy N start N workers that exercise scheduling policy --schedpolicy-ops N stop after N scheduling policy bogo operations --schedpolicy-rand select scheduling policy randomly --sctp N start N workers performing SCTP send/receives --sctp-domain D specify sctp domain, default is ipv4 --sctp-if I use network interface I, e.g. lo, eth0, etc. --sctp-ops N stop after N SCTP bogo operations --sctp-port P use SCTP ports P to P + number of workers - 1 --sctp-sched S specify sctp scheduler --seal N start N workers performing fcntl SEAL commands --seal-ops N stop after N SEAL bogo operations --seccomp N start N workers performing seccomp call filtering --seccomp-ops N stop after N seccomp bogo operations --secretmem N start N workers that use secretmem mappings --secretmem-ops N stop after N secretmem bogo operations --seek N start N workers performing random seek r/w IO --seek-ops N stop after N seek bogo operations --seek-punch punch random holes in file to stress extents --seek-size N length of file to do random I/O upon --sem N start N workers doing semaphore operations --sem-ops N stop after N semaphore bogo operations --sem-procs N number of processes to start per worker --sem-shared share the semaphore across all semaphore stressors --sem-sysv N start N workers doing System V semaphore operations --sem-sysv-ops N stop after N System V sem bogo operations --sem-sysv-procs N number of processes to start per worker --sem-sysv-setall exercise semctl SETALL (will alter the processes' semaphore set) --sendfile N start N workers exercising sendfile --sendfile-ops N stop after N bogo sendfile operations --sendfile-size N size of data to be sent with sendfile -f N, --session N start N workers that exercise new sessions --session-ops N stop after N session bogo operations --set N start N workers exercising the set*() system calls --set-ops N stop after N set bogo operations --shellsort N start N workers shell sorting 32 bit random integers --shellsort-ops N stop after N shell sort bogo operations --shellsort-size N number of 32 bit integers to sort --shm N start N workers that exercise POSIX shared memory --shm-bytes N allocate/free N bytes of POSIX shared memory --shm-mlock attempt to mlock pages into memory --shm-objs N allocate N POSIX shared memory objects per iteration --shm-ops N stop after N POSIX shared memory bogo operations --shm-sysv N start N workers that exercise System V shared memory --shm-sysv-bytes N allocate and free N bytes of shared memory per loop --shm-sysv-mlock attempt to mlock pages into memory --shm-sysv-ops N stop after N shared memory bogo operations --shm-sysv-segs N allocate N shared memory segments per iteration --sigabrt N start N workers generating SIGABRT signals --sigabrt-ops N stop after N bogo SIGABRT operations --sigbus N start N workers generating bus faults --sigbus-ops N stop after N bogo bus faults --sigchld N start N workers that handle SIGCHLD --sigchld-ops N stop after N bogo SIGCHLD signals --sigfd N start N workers reading signals via signalfd reads --sigfd-ops N stop after N bogo signalfd reads --sigfpe N start N workers generating floating point math faults --sigfpe-ops N stop after N bogo floating point math faults --sighup N start N workers generating SIGHUP signals --sighup-ops N stop after N bogo SIGHUP operations --sigio N start N workers that exercise SIGIO signals --sigio-ops N stop after N bogo sigio signals --sigill N start N workers generating SIGILL signals --sigill-ops N stop after N SIGILL signals --signal N start N workers that exercise signal --signal-ops N stop after N bogo signals --signest N start N workers generating nested signals --signest-ops N stop after N bogo nested signals --sigpending N start N workers exercising sigpending --sigpending-ops N stop after N sigpending bogo operations --sigpipe N start N workers exercising SIGPIPE --sigpipe-ops N stop after N SIGPIPE bogo operations --sigq N start N workers sending sigqueue signals --sigq-ops N stop after N sigqueue bogo operations --sigrt N start N workers sending real time signals --sigrt-ops N stop after N real time signal bogo operations --sigsegv N start N workers generating segmentation faults --sigsegv-ops N stop after N bogo segmentation faults --sigsuspend N start N workers exercising sigsuspend --sigsuspend-ops N stop after N bogo sigsuspend wakes --sigtrap N start N workers generating SIGTRAP signals --sigtrap-ops N stop after N bogo SIGTRAP operations --sigurg N start N workers exercising SIGURG on MSB_OOB socket sends --sigurg-ops N stop after N SIGURG signals --sigvtalrm N start N workers exercising SIGVTALRM signals --sigvtalrm-ops N stop after N SIGVTALRM signals --sigxcpu N start N workers that exercise SIGXCPU signals --sigxcpu-ops N stop after N bogo SIGXCPU signals --sigxfsz N start N workers that exercise SIGXFSZ signals --sigxfsz-ops N stop after N bogo SIGXFSZ signals --skiplist N start N workers that exercise a skiplist search --skiplist-ops N stop after N skiplist search bogo operations --skiplist-size N number of 32 bit integers to add to skiplist --sleep N start N workers performing various duration sleeps --sleep-max P create P threads at a time by each worker --sleep-ops N stop after N bogo sleep operations --smi N start N workers that trigger SMIs --smi-ops N stop after N SMIs have been triggered -S N, --sock N start N workers exercising socket I/O --sock-domain D specify socket domain, default is ipv4 --sock-if I use network interface I, e.g. lo, eth0, etc. --sock-msgs N number of messages to send per connection --sock-nodelay disable Nagle algorithm, send data immediately --sock-ops N stop after N socket bogo operations --sock-opts option socket options [send|sendmsg|sendmmsg] --sock-port P use socket ports P to P + number of workers - 1 --sock-protocol use socket protocol P, default is tcp, can be mptcp --sock-type T socket type (stream, seqpacket) --sock-zerocopy enable zero copy sends --sockabuse N start N workers abusing socket I/O --sockabuse-ops N stop after N socket abusing bogo operations --sockabuse-port P use socket ports P to P + number of workers - 1 --sockdiag N start N workers exercising sockdiag netlink --sockdiag-ops N stop sockdiag workers after N bogo messages --sockfd N start N workers sending file descriptors over sockets --sockfd-ops N stop after N sockfd bogo operations --sockfd-port P use socket fd ports P to P + number of workers - 1 --sockfd-reuse reuse file descriptors between sender and receiver --sockmany N start N workers exercising many socket connections --sockmany-if I use network interface I, e.g. lo, eth0, etc. --sockmany-ops N stop after N sockmany bogo operations --sockmany-port use socket ports P to P + number of workers - 1 --sockpair N start N workers exercising socket pair I/O activity --sockpair-ops N stop after N socket pair bogo operations --softlockup N start N workers that cause softlockups --softlockup-ops N stop after N softlockup bogo operations --sparsematrix N start N workers that exercise a sparse matrix --sparsematrix-items N N is the number of items in the spare matrix --sparsematrix-method M select storage method: all, hash, hashjudy, judy, list, mmap, qhash, rb, splay --sparsematrix-ops N stop after N bogo sparse matrix operations --sparsematrix-size N M is the width and height X x Y of the matrix --spawn N start N workers spawning stress-ng using posix_spawn --spawn-ops N stop after N spawn bogo operations --spinmem start N workers exercising shared memory spin write/read operations --spinmem-affinity use CPU affinity (specific CPUS can be defined by --taskset option) --spinmem-method select method of write/reads, default is 32bit --spinmem-numa move pages to randomly chosen NUMA nodes --spinmem-ops stop after N bogo shared memory spin write/read operations --splice N start N workers reading/writing using splice --splice-bytes N number of bytes to transfer per splice call --splice-ops N stop after N bogo splice operations --stack N start N workers generating stack overflows --stack-fill fill stack, touches all new pages --stack-mlock mlock stack, force pages to be unswappable --stack-ops N stop after N bogo stack overflows --stack-pageout use madvise to try to swap out stack --stack-unmap unmap a page in the stack on each iteration --stackmmap N start N workers exercising a filebacked stack --stackmmap-ops N stop after N bogo stackmmap operations --statmount N start N workers exercising statmount and listmount --statmount-ops N stop after N bogo statmount and listmount operations --str N start N workers exercising lib C string functions --str-method func specify the string function to stress --str-ops N stop after N bogo string operations --stream N start N workers exercising memory bandwidth --stream-index N specify number of indices into the data (0..3) --stream-l3-size N specify the L3 cache size of the CPU --stream-madvise M specify mmap'd stream buffer madvise advice --stream-mlock attempt to mlock pages into memory --stream-prefetch use prefetching (where available) --stream-ops N stop after N bogo stream operations --swap N start N workers exercising swapon/swapoff --swap-ops N stop after N swapon/swapoff operations --swap-self attempt to swap stressors pages out -s N, --switch N start N workers doing rapid context switches --switch-freq N set frequency of context switches --switch-method M mq | pipe | sem-sysv --switch-ops N stop after N context switch bogo operations --symlink N start N workers creating symbolic links --symlink-ops N stop after N symbolic link bogo operations --symlink-sync enable sync'ing after symlinking/unsymlinking --sync-file N start N workers exercise sync_file_range --sync-file-bytes N size of file to be sync'd --sync-file-ops N stop after N sync_file_range bogo operations --syncload N start N workers that synchronize load spikes --syncload-msbusy M maximum busy duration in milliseconds --syncload-mssleep M maximum sleep duration in milliseconds --syncload-ops N stop after N syncload bogo operations --sysbadaddr N start N workers that pass bad addresses to syscalls --sysbadaddr-ops N stop after N sysbadaddr bogo syscalls --syscall N start N workers that exercise a wide range of system calls --syscall-method M select method of selecting system calls to exercise --syscall-ops N stop after N syscall bogo operations --syscall-top N display fastest top N system calls --sysfs N start N workers reading files from /sys --sysfs-ops N stop after sysfs bogo operations --sysinfo N start N workers reading system information --sysinfo-ops N stop after sysinfo bogo operations --sysinval N start N workers that pass invalid args to syscalls --sysinval-ops N stop after N sysinval bogo syscalls --tee N start N workers exercising the tee system call --tee-ops N stop after N tee bogo operations -T N, --timer N start N workers producing timer events --timer-freq F run timer(s) at F Hz, range 1 to 1000000000 --timer-ops N stop after N timer bogo events --timer-rand enable random timer frequency --timerfd N start N workers producing timerfd events --timerfd-fds N number of timerfd file descriptors to open --timerfd-freq F run timer(s) at F Hz, range 1 to 1000000000 --timerfd-ops N stop after N timerfd bogo events --timerfd-rand enable random timerfd frequency --time-warp N start N workers checking for timer/clock warping --time-warp-ops N stop workers after N bogo timer/clock reads --tlb-shootdown N start N workers that force TLB shootdowns --tlb-shootdown-ops N stop after N TLB shootdown bogo ops --tmpfs N start N workers mmap'ing a file on tmpfs --tmpfs-mmap-async using asynchronous msyncs for tmpfs file based mmap --tmpfs-mmap-file mmap onto a tmpfs file using synchronous msyncs --tmpfs-ops N stop after N tmpfs bogo ops --touch N start N stressors that touch and remove files --touch-method M specify method to touch tile file, open | create --touch-ops N stop after N touch bogo operations --touch-opts list touch open options all, direct, dsync, excl, noatime, sync, trunc --tree N start N workers that exercise tree structures --tree-method M select tree method: all,avl,binary,btree,rb,splay --tree-ops N stop after N bogo tree operations --tree-size N N is the number of items in the tree --trig N start N workers exercising trigonometric functions --trig-ops N stop after N trig bogo trigonometric operations --trig-method M select trigonometric function to exercise --tsc N start N workers reading the time stamp counter --tsc-ops N stop after N TSC bogo operations --tsc-lfence add lfence after TSC reads for serialization (x86 only) --tsc-rdscp use rdtscp instead of rdtsc, disables tsc-lfence (x86 only) --tsearch N start N workers that exercise a tree search --tsearch-ops N stop after N tree search bogo operations --tsearch-size N number of 32 bit integers to tsearch --tun N start N workers exercising tun interface --tun-ops N stop after N tun bogo operations --tun-tap use TAP interface instead of TUN --udp N start N workers performing UDP send/receives --udp-domain D specify domain, default is ipv4 --udp-gro enable UDP-GRO --udp-if I use network interface I, e.g. lo, eth0, etc. --udp-lite use the UDP-Lite (RFC 3828) protocol --udp-ops N stop after N udp bogo operations --udp-port P use ports P to P + number of workers - 1 --udp-flood N start N workers that performs a UDP flood attack --udp-flood-domain D specify domain, default is ipv4 --udp-flood-if I use network interface I, e.g. lo, eth0, etc. --udp-flood-ops N stop after N udp flood bogo operations --umask N start N workers exercising umask, file create/stat/close/unlink --umask-ops N stop after N umask and file operations --umount N start N workers exercising umount races --umount-ops N stop after N bogo umount operations --unlink N start N unlink exercising stressors --unlink-ops N stop after N unlink exercising bogo operations --unshare N start N workers exercising resource unsharing --unshare-ops N stop after N bogo unshare operations --uprobe N start N workers that generate uprobe events --uprobe-ops N stop after N uprobe events -u N, --urandom N start N workers reading /dev/urandom --urandom-ops N stop after N urandom bogo read operations --userfaultfd N start N page faulting workers with userspace handling --userfaultfd-bytes N size of mmap'd region to fault on --userfaultfd-ops N stop after N page faults have been handled --usersyscall N start N workers exercising a userspace system call handler --usersyscall-ops N stop after N successful SIGSYS system calls --utime N start N workers updating file timestamps --utime-fsync force utime meta data sync to the file system --utime-ops N stop after N utime bogo operations --vdso N start N workers exercising functions in the VDSO --vdso-func F use just vDSO function F --vdso-ops N stop after N vDSO function calls --veccmp N start N workers performing integer vector comparison ops --veccmp-ops N stop after N integer vector comparison bogo operations --vecfp N start N workers performing vector math ops --vecfp-ops N stop after N vector math bogo operations --vecmath N start N workers performing vector math ops --vecmath-ops N stop after N vector math bogo operations --vecshuf N start N workers performing vector shuffle ops --vecshuf-method M select vector shuffling method --vecshuf-ops N stop after N vector shuffle bogo operations --vecwide N start N workers performing vector math ops --vecwide-ops N stop after N vector math bogo operations --verity N start N workers exercising file verity ioctls --verity-ops N stop after N file verity bogo operations --vfork N start N workers spinning on vfork() and exit() --vfork-ops N stop after N vfork bogo operations --vfork-max P create P vforked processes per iteration, default is 1 --vforkmany N start N workers spawning many vfork children --vforkmany-ops N stop after spawning N vfork children --vforkmany-vm enable extra virtual memory pressure --vforkmany-vm-bytes set the default vm mmap'd size, default 64MB -m N, --vm N start N workers spinning on anonymous mmap --vm-bytes N allocate N bytes per vm worker (default 256MB) --vm-flush cache flush data after write --vm-hang N sleep N seconds before freeing memory --vm-keep redirty memory instead of reallocating --vm-locked lock the pages of the mapped region into memory --vm-madvise M specify mmap'd vm buffer madvise advice --vm-method M specify stress vm method M, default is all --vm-numa bind memory mappings to randomly selected NUMA nodes --vm-ops N stop after N vm bogo operations --vm-populate populate (prefault) page tables for a mapping --vma N start N workers that exercise kernel VMA structures --vma-ops N stop N workers after N mmap VMA operations --vm-addr N start N vm address exercising workers --vm-addr-method M select method to exercise vm addresses --vm-addr-mlock attempt to mlock pages into memory --vm-addr-numa bind memory mappings to randomly selected NUMA nodes --vm-addr-ops N stop after N vm address bogo operations --vm-rw N start N vm read/write process_vm* copy workers --vm-rw-bytes N transfer N bytes of memory per bogo operation --vm-rw-ops N stop after N vm process_vm* copy bogo operations --vm-segv N start N workers that unmap their address space --vm-segv-ops N stop after N vm-segv unmap'd SEGV faults --vm-splice N start N workers reading/writing using vmsplice --vm-splice-bytes N number of bytes to transfer per vmsplice call --vm-splice-ops N stop after N bogo splice operations --vnni N start N workers performing vector neural network ops --vnni-intrinsic use x86 intrinsic vnni methods, disable generic methods --vnni-method M specify specific vnni methods to exercise --vnni-ops N stop after N vnni bogo operations --wait N start N workers waiting on child being stop/resumed --wait-ops N stop after N bogo wait operations --waitcpu N start N workers exercising wait/pause/nop instructions --waitcpu-ops N stop after N wait/pause/nop bogo operations --watchdog N start N workers that exercise /dev/watchdog --watchdog-ops N stop after N bogo watchdog operations --wcs N start N workers on lib C wide char string functions --wcs-method func specify the wide character string function to stress --wcs-ops N stop after N bogo wide character string operations --workload N start N workers that exercise a mix of scheduling loads --workload-dist type workload distribution type [random1, random2, random3, cluster] --workload-load P percentage load P per workload time slice --workload-ops N stop after N workload bogo operations --workload-quanta-us N max duration of each quanta work item in microseconds --workload-sched P select scheduler policy [idle, fifo, rr, other, batch, deadline] --workload-slice-us N duration of workload time load in microseconds --workload-threads N number of workload threads workers to use, default is 0 (disabled) --workload-method M select a workload method, default is all --x86cpuid N start N workers exercising the x86 cpuid instruction --x86cpuid-ops N stop after N cpuid bogo operations --x86syscall N start N workers exercising functions using syscall --x86syscall-func F use just syscall function F --x86syscall-ops N stop after N syscall function calls --xattr N start N workers stressing file extended attributes --xattr-ops N stop after N bogo xattr operations -y N, --yield N start N workers doing sched_yield() calls --yield-ops N stop after N bogo yield operations --yield-procs N specify number of yield processes per stressor --yield-sched P select scheduler policy [idle, fifo, rr, other, batch, deadline] --zero N start N workers exercising /dev/zero with read, mmap, ioctl, lseek --zero-read just exercise /dev/zero with reading --zero-ops N stop after N /dev/zero bogo read operations --zlib N start N workers compressing data with zlib --zlib-level L specify zlib compression level 0=fast, 9=best --zlib-mem-level L specify zlib compression state memory usage 1=minimum, 9=maximum --zlib-method M specify zlib random data generation method M --zlib-ops N stop after N zlib bogo compression operations --zlib-strategy S specify zlib strategy 0=default, 1=filtered, 2=huffman only, 3=rle, 4=fixed --zlib-stream-bytes S specify the number of bytes to deflate until the current stream will be closed --zlib-window-bits W specify zlib window bits -8-(-15) | 8-15 | 24-31 | 40-47 --zombie N start N workers that rapidly create and reap zombies --zombie-max N set upper limit of N zombies per worker --zombie-ops N stop after N bogo zombie fork operations Example: stress-ng --cpu 8 --iomix 4 --vm 2 --vm-bytes 128M --fork 4 --timeout 10s Note: sizes can be suffixed with B, K, M, G and times with s, m, h, d, y
比如执行最后给出的这个 Example 命令：
/data/local/tmp # ./stress-ng --cpu 8 --iomix 4 --vm 2 --vm-bytes 128M --fork 4 --timeout 10s stress-ng: info: [16659] setting to a 10 secs run per stressor stress-ng: info: [16659] dispatching hogs: 8 cpu, 4 iomix, 2 vm, 4 fork stress-ng: info: [16668] iomix: using 256M file system space per stressor instance (total 1G of 55.78G available file system space) stress-ng: info: [16672] vm: using 64M per stressor instance (total 128M of 1.64G available memory) stress-ng: info: [16659] skipped: 0 stress-ng: info: [16659] passed: 18: cpu (8) iomix (4) vm (2) fork (4) stress-ng: info: [16659] failed: 0 stress-ng: info: [16659] metrics untrustworthy: 0 stress-ng: info: [16659] successful run completed in 10.58 secs
posted on 2026-08-04 17:09 Hello-World3 阅读(13) 评论(0) 收藏 举报
浙公网安备 33010602011771号