Date: Mon, 13 Feb 2023 01:43:08 +0300 From: Dmitry Chagin <dchagin@heemeyer.club> To: Antoine Brodin <antoine@freebsd.org> Cc: Dmitry Chagin <dchagin@freebsd.org>, src-committers@freebsd.org, dev-commits-src-all@freebsd.org, dev-commits-src-main@freebsd.org Subject: Re: git: c21b080f3dc2 - main - cpuset: Fix sched_[g|s]etaffinity() for better compatibility with Linux. Message-ID: <Y%2BlrfNUilsS9xzMN@heemeyer.club> In-Reply-To: <CAALwa8=rpBRG6wBrDqUmzz95A2GR%2BVxQGQGWOEHcG%2BQ08CZYHg@mail.gmail.com> References: <202301291318.30TDIq4Q008231@gitrepo.freebsd.org> <CAALwa8kqtTa0ocrC_pT4uPSC-duBCQNfT4KUFH2Y3COsmnf32A@mail.gmail.com> <Y%2BjCO8IEb%2BG4zsQd@heemeyer.club> <Y%2BjJu2WP8nQGOCOx@heemeyer.club> <CAALwa8=hHMVF1W-g54eUH-EGKH87Yj70rarwKDiT=fKXQ2_S_Q@mail.gmail.com> <Y%2BliCaNstuAOSpyt@heemeyer.club> <CAALwa8=rpBRG6wBrDqUmzz95A2GR%2BVxQGQGWOEHcG%2BQ08CZYHg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Feb 12, 2023 at 10:23:12PM +0000, Antoine Brodin wrote: > On Sun, Feb 12, 2023 at 10:02 PM Dmitry Chagin <dchagin@heemeyer.club> wrote: > > On Sun, Feb 12, 2023 at 07:58:07PM +0000, Antoine Brodin wrote: > > > On Sun, Feb 12, 2023 at 11:13 AM Dmitry Chagin <dchagin@heemeyer.club> wrote: > > > > > > > > On Sun, Feb 12, 2023 at 01:40:59PM +0300, Dmitry Chagin wrote: > > > > > On Sun, Feb 12, 2023 at 09:47:15AM +0000, Antoine Brodin wrote: > > > > > > On Sun, Jan 29, 2023 at 1:18 PM Dmitry Chagin <dchagin@freebsd.org> wrote: > > > > > > > > > > > > > > The branch main has been updated by dchagin: > > > > > > > > > > > > > > URL: https://cgit.FreeBSD.org/src/commit/?id=c21b080f3dc2f5e91ada608d4385b7ed6538ba9b > > > > > > > > > > > > > > commit c21b080f3dc2f5e91ada608d4385b7ed6538ba9b > > > > > > > Author: Dmitry Chagin <dchagin@FreeBSD.org> > > > > > > > AuthorDate: 2023-01-29 13:17:33 +0000 > > > > > > > Commit: Dmitry Chagin <dchagin@FreeBSD.org> > > > > > > > CommitDate: 2023-01-29 13:17:33 +0000 > > > > > > > > > > > > > > cpuset: Fix sched_[g|s]etaffinity() for better compatibility with Linux. > > > > > > > > > > > > > > Under Linux to sched_[g|s]etaffinity() functions the value returned from a call > > > > > > > to gettid(2) (thread id) can be passed in the argument pid. Specifying pid as 0 > > > > > > > will set the attribute for the calling thread, and passing the value returned > > > > > > > from a call to getpid(2) (process id) will set the attribute for the main thread > > > > > > > of the thread group. > > > > > > > > > > > > > > Native cpuset(2) family of system calls has "which" argument to determine how > > > > > > > the value of id argument is interpreted, i.e., CPU_WHICH_TID is used to pass > > > > > > > a thread id and CPU_WHICH_PID - to pass a process id. > > > > > > > > > > > > > > For now native sched_[g|s]etaffinity() implementation is wrong as uses "which" > > > > > > > CPU_WHICH_PID to pass both (process and thread id) to the kernel. To fix this > > > > > > > adding a new "which" CPU_WHICH_TIDPID intended to handle both id's. > > > > > > > > > > > > > > Reviewed by: kib > > > > > > > Differential Revision: https://reviews.freebsd.org/D38209 > > > > > > > MFC after: 1 week > > > > > > > --- > > > > > > > lib/libc/gen/sched_getaffinity.c | 2 +- > > > > > > > lib/libc/gen/sched_setaffinity.c | 2 +- > > > > > > > lib/libc/sys/cpuset.2 | 4 +++- > > > > > > > sys/kern/kern_cpuset.c | 42 +++++++++++++++++++++++++++++++++++++++- > > > > > > > sys/sys/cpuset.h | 1 + > > > > > > > 5 files changed, 47 insertions(+), 4 deletions(-) > > > > > > > > > > > > > > > > > > Hello, > > > > > > > > > > > > This broke math/py-numpy on main and stable/13 > > > > > > > > > > > > https://pkg-status.freebsd.org/gohan03/data/main-amd64-default-baseline/p52768539d46f_scd3193f640/logs/errors/py39-numpy-1.24.1,1.log > > > > > > https://pkg-status.freebsd.org/gohan02/data/13stable-amd64-quarterly-baseline/841610d9bfc6/logs/errors/py39-numpy-1.23.5_1,1.log > > > > > > > > > > > > File "/wrkdirs/usr/ports/math/py-numpy/work-py39/numpy-1.23.5/numpy/distutils/misc_util.py", > > > > > > line 93, in get_num_build_jobs > > > > > > cpu_count = len(os.sched_getaffinity(0)) > > > > > > OverflowError: could not allocate a large enough CPU set > > > > > > > > > > hi, Antoine, > > > > > > > > > > on stable/13 > > > > > dchagin@heemeyer:~ % uname -a > > > > > FreeBSD heemeyer.club 13.2-STABLE FreeBSD 13.2-STABLE #268 stable/13-n254471-51485f81b01a-dirty: Sat Feb 11 12:16:21 MSK 2023 dchagin@heemeyer.club:/usr/home/dchagin/tmp/obj/usr/home/dchagin/freebsd/amd64.amd64/sys/YOY amd64 > > > > > dchagin@heemeyer:~ % python3.9 > > > > > Python 3.9.16 (main, Dec 18 2022, 01:15:32) > > > > > [Clang 13.0.0 (git@github.com:llvm/llvm-project.git llvmorg-13.0.0-0-gd7b669b3a on freebsd13 > > > > > Type "help", "copyright", "credits" or "license" for more information. > > > > > >>> import os > > > > > >>> len(os.sched_getaffinity(0)) > > > > > 1 > > > > > >>> os.sched_getaffinity(0) > > > > > {0} > > > > > >>> exit() > > > > > dchagin@heemeyer:~ % > > > > > > > > > > on current: > > > > > dchagin@mordor:~/freebsd % uname -a > > > > > FreeBSD mordor.heemeyer.club 14.0-CURRENT FreeBSD 14.0-CURRENT #458 futex-n260850-52456667491d-dirty: Sat Feb 11 14:24:46 MSK 2023 dchagin@mordor.heemeyer.club:/home/dchagin/tmp/obj/home/dchagin/freebsd/amd64.amd64/sys/YOY amd64 > > > > > dchagin@mordor:~/freebsd % python3.9 > > > > > Python 3.9.16 (main, Dec 18 2022, 01:24:21) > > > > > [Clang 14.0.5 (https://github.com/llvm/llvm-project.git llvmorg-14.0.5-0-gc1238 on freebsd14 > > > > > Type "help", "copyright", "credits" or "license" for more information. > > > > > >>> import os > > > > > >>> len(os.sched_getaffinity(0)) > > > > > 16 > > > > > >>> os.sched_getaffinity(0) > > > > > {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15} > > > > > >>> exit() > > > > > dchagin@mordor:~/freebsd % > > > > > > > > > > > > > > > could you please show out of os.sched_getaffinity(0) ? > > > > > > > > and os.cpu_count() > > > > also sysctl kern.smp > > > > > > Here is what I have in the head jail: > > > > > > >>> len(os.sched_getaffinity(0)) > > > Traceback (most recent call last): > > > File "<stdin>", line 1, in <module> > > > OverflowError: could not allocate a large enough CPU set > > > >>> os.sched_getaffinity(0) > > > Traceback (most recent call last): > > > File "<stdin>", line 1, in <module> > > > OverflowError: could not allocate a large enough CPU set > > > >>> os.cpu_count() > > > 64 > > > >>> > > > root@mainamd64PR269425-default:~ # sysctl kern.smp > > > kern.smp.forward_signal_enabled: 1 > > > kern.smp.topology: 0 > > > kern.smp.cores: 32 > > > kern.smp.threads_per_core: 2 > > > kern.smp.cpus: 64 > > > kern.smp.disabled: 0 > > > kern.smp.active: 1 > > > kern.smp.maxcpus: 256 > > > kern.smp.maxid: 63 > > > > > > thanks, I've implemented python os_sched_getaffinity_impl and could not > > get a similar result. Is it posiible that you run old wolrd > > (pre 3e11d3f6)? > > This is with a new world, and a kernel that is a few weeks old. > > Antoine ok, so I need some help from you, I've put test program to https://people.freebsd.org/~dchagin/bsd_sched_getaffinity.tar please build and run like ktrace -di ./getaffinity, then kdump -HAR -m 64 > affinity.log and send log to me please. I can't reproduce problem in any my environment, I'll have hw with 88 cores only at the end of week. Thank you
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Y%2BlrfNUilsS9xzMN>