May 2025 - kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

TCP high concurrency, RST connection cannot be released normally
by Oracle May 31, 2025

May 31, 2025

  Problem description: Abnormal message appears in high-concurrency short connection, the server sends rst to the client, the client connection is closed, but the server connection that sends rst is still in ESTABLISHED state? What is the reason? I captured some information through bpf and confirmed that the server sent rst and the client status changed to close. I tracked the failure return of tcp_v4_send_reset and tcp_rcv_state_process and found that TCP RST: ACK flag = 1, but the tcp state printed here is the listen state, which is not consistent with the current connection status viewed by netstat. (Not sure if some events are discarded), but the tcpdump packet capture analysis shows that the server is indeed rst and the connection is not released. [root@systest104 zx]# netstat -atpn|grep 45129 |grep ES tcp        0      0 10.13.8.104:45129       10.13.8.104:45052       ESTABLISHED 3942077/postgres:  #The connection still exists tcpdump: 02:58:06.494295 IP (tos 0x0, ttl 64, id 0, offset 0, flags [DF], proto TCP (6), length 60)     10.13.8.104.45129 > 10.13.8.104.45052: Flags [S.], cksum 0x2518 (incorrect -> 0x7ac8), seq 214961739, ack 2894898555, win 65535, options [mss 65495,sackOK,TS val 1670384838 ecr 1670384838,nop,wscale 11], length 0 02:58:06.496318 IP (tos 0x0, ttl 64, id 15240, offset 0, flags [DF], proto TCP (6), length 52)     10.13.8.104.45052 > 10.13.8.104.45129: Flags [.], cksum 0x2510 (incorrect -> 0xa39b), seq 65579, ack 39408264, win 32, options [nop,nop,TS val 1670384839 ecr 1670384838], length 0 02:58:06.496321 IP (tos 0x0, ttl 64, id 15241, offset 0, flags [DF], proto TCP (6), length 84)     10.13.8.104.45052 > 10.13.8.104.45129: Flags [P.], cksum 0x2530 (incorrect -> 0x2b50), seq 65579:65611, ack 39408264, win 32, options [nop,nop,TS val 1670384839 ecr 1670384838], length 32 02:58:06.497830 IP (tos 0x0, ttl 64, id 0, offset 0, flags [DF], proto TCP (6), length 40)     10.13.8.104.45129 > 10.13.8.104.45052: Flags [R], cksum 0x0f95 (correct), seq 214961740, win 0, length 0

2 1

Re: [ANNOUNCE] Call for Contributors: ASIOS – AI‑Native OS (Kernel Work)
by Vikram R May 24, 2025

May 24, 2025

Hi Venkatesh, Thanks for your interest in ASIOS — really excited to hear from you! What’s working so far: - Cross-arch kernel build system (x86_64 and ARM64) - Ubuntu HWE 6.11 base with AI-optimized configuration overlays - Automated builds with PREEMPT_RT, SCHED_EXT, and NUMA tuning - Multi-arch packaging, signing, and module verification pipeline Where we could use help: 1. Performance validation – Benchmark ASIOS kernel vs vanilla Ubuntu on AI workloads (Torch, JAX, etc.). 2. BPF sched_ext policies – Create AI-aware scheduling extensions (e.g. workload tagging, core isolation). 3. Low-level optimization – Enhance NUMA memory locality, improve GPU-thread mapping. You Can check out scripts and build instructions on: https://github.com/asi-os/asios-core If any of these areas match your expertise, feel free to dive in or open an issue/PR and you're very welcome to join our discord community: https://discord.gg/rWuU7cWU4E -- Best, Vikram

1 0

Re: [ANNOUNCE] Call for Contributors: ASIOS – AI‑Native OS (Kernel Work)
by Vikram R May 23, 2025

May 23, 2025

Hi Venkatesh, Thanks for your interest in ASIOS — really excited to hear from you! What’s working so far: - Cross-arch kernel build system (x86_64 and ARM64) - Ubuntu HWE 6.11 base with AI-optimized configuration overlays - Automated builds with PREEMPT_RT, SCHED_EXT, and NUMA tuning - Multi-arch packaging, signing, and module verification pipeline Where we could use help: 1. Performance validation – Benchmark ASIOS kernel vs vanilla Ubuntu on AI workloads (Torch, JAX, etc.). 2. BPF sched_ext policies – Create AI-aware scheduling extensions (e.g. workload tagging, core isolation). 3. Low-level optimization – Enhance NUMA memory locality, improve GPU-thread mapping. You can check out scrips and build instructions on: https://github.com/asi-os/asios-core If any of these areas match your expertise, feel free to dive in or open an issue/PR and you're very welcome to join our discord community: https://discord.gg/rWuU7cWU4E -- Best, Vikram

1 0

Re: [ANNOUNCE] Call for Contributors: ASIOS – AI‑Native OS (Kernel Work)
by Vikram R May 21, 2025

May 21, 2025

Hi Enrico, Thanks for the questions – please see my answers inline: > “What exactly do you mean by that?” By “AI-native” I mean ASIOS is designed from the ground up for AI workloads, not retrofitted. It embeds ML-driven optimizations (e.g. reinforcement-learning schedulers, Bayesian latency predictors), deterministic scheduling, native multi-accelerator support and built-in alignment/safety probes directly into the OS stack. > “Why an old Ubuntu?” We chose Ubuntu 24.04 LTS because it gives us a rock-solid, long-term-support base with an upstream PREEMPT_RT HWE kernel (6.11) that we can rebase regularly. It’s reproducible, widely deployed in enterprise and cloud environments, and lets us focus on AI-centric enhancements rather than reinventing the distribution. > “If you’re talking AI, you mean datacenter, right? I’d trim things down…” Our primary target today is bare-metal and VM-based AI clusters, but ASIOS also supports very slim builds (we strip out unneeded services) and future multi-distro releases (we’re planning Devuan, Yocto/PTXdist, RHEL/SUSE variants). > “Which symbols? ELF symbols or kconfig defines?” The overlay logic deals with Kconfig symbols (CONFIG_*). asios-config-overlay.sh skips any missing CONFIG options instead of failing, > Re: extra-deb tooling Great pointer on deb-pkg – we’ll evaluate integrating it into our .deb build pipeline to streamline packaging. I did run into some issues packaging cross-architecture kernels with dpkg-buildpackage. Please let me know if you have any other suggestions or questions! Cheers, — Vikram R KarLex AI, Inc.

1 0

[ANNOUNCE] Call for Contributors: ASIOS – AI‑Native OS (Kernel Work)
by Vikram R May 19, 2025

May 19, 2025

ASIOS is a new AI‑native OS (Ubuntu 24.04 LTS base) bringing deterministic scheduling, NUMA‑GPU optimizations, zero‑copy I/O, and eBPF telemetry into the Linux kernel. Key directions— — Deterministic CPU scheduling for reproducible AI runs — NUMA‑aware memory placement tuned for GPU DMA — Zero‑copy GPU I/O via GPUDirect RDMA/Storage — eBPF‑based telemetry hooks **Call for contributors:** scheduler/MM, GPU/accelerator integration, eBPF instrumentation. Project links— GitHub: <https://github.com/asi-os> Discord: <https://discord.gg/rWuU7cWU4E> Roadmap: <https://github.com/asi-os/asios-docs/blob/main/docs/ARCHITECTURE.md> Please reply on‑list with questions or join us on Discord to dive into the design. Vikram Karlex R KarLex AI, Inc. | <https://asios.ai> | <https://karlex.ai>

2 2

Re: [ANNOUNCE] Call for Contributors: ASIOS – AI‑Native OS (Kernel Work)
by Vikram R May 19, 2025

May 19, 2025

Sorry for the noise. On Tue, May 20, 2025 at 1:00 AM Grozdan <neutrino8(a)gmail.com> wrote: > > Please do not spam the list with your AI crap! > > On Mon, May 19, 2025 at 9:28 PM Vikram R <vikram(a)karlexai.com> wrote: >> >> ASIOS – Envisioned to Host and Sustain Cognition & Synthetic Intelligence >> The world’s first AI-native operating system. >> >> ASIOS is a new class of OS built on Ubuntu 24.04 LTS, designed from >> first principles for AI workloads. We’re creating the foundational >> infrastructure layer to support ever-more sophisticated >> intelligence—from today’s deep-learning models to tomorrow’s advanced >> cognitive and quantum-ready AI. >> >> Building on our May 3 thread, we’ve landed a resilient config-overlay >> and build script for ASIOS on both x86_64 and aarch64. Key artifacts: >> >> asios-config-overlay.sh: idempotent Kconfig overlay that tolerates >> missing symbols >> build-asios-kernel.sh: automated clone to config to compile workflow >> test-asios-config.sh: verifies .config flags post build >> Initial perf on Ubuntu 24.04 HWE 6.11: >> x86_64: mbw=7.3 GB/s, hackbench=53 ms, fio=1.6 GB/s >> arm64: mbw=27.4 GB/s, hackbench=4.8 ms, fio=3.2 GB/s >> >> GitHub (scripts and tests): https://github.com/asi-os/asios-core/ >> Discord: https://discord.gg/rWuU7cWU4E >> >> Areas for review and test: >> 1. Overlay logic: skipping versus failing on absent Kconfig symbols >> 2. Build-matrix logic: host-only versus cross-build >> 3. Test completeness: missing flags >> 4. Scripting style and robustness >> >> Any pointers or suggestions are highly appreciated before we push to a >> formal RFC series. >> >> — >> Vikram R >> Founder & CEO, KarLex AI, Inc. >> https://asios.ai >> >> >> On Sat, May 3, 2025 at 12:22 PM Vikram R <vikram(a)karlexai.com> wrote: >> > >> > ASIOS is a new AI‑native OS (Ubuntu 24.04 LTS base) bringing >> > deterministic scheduling, NUMA‑GPU optimizations, zero‑copy I/O, and >> > eBPF telemetry into the Linux kernel. >> > >> > Key directions— >> > — Deterministic CPU scheduling for reproducible AI runs >> > — NUMA‑aware memory placement tuned for GPU DMA >> > — Zero‑copy GPU I/O via GPUDirect RDMA/Storage >> > — eBPF‑based telemetry hooks >> > >> > **Call for contributors:** scheduler/MM, GPU/accelerator integration, >> > eBPF instrumentation. >> > >> > Project links— >> > GitHub: <https://github.com/asi-os> >> > Discord: <https://discord.gg/rWuU7cWU4E> >> > Roadmap: <https://github.com/asi-os/asios-docs/blob/main/docs/ARCHITECTURE.md> >> > >> > Please reply on‑list with questions or join us on Discord to dive into >> > the design. >> > >> > Vikram Karlex R >> > KarLex AI, Inc. | <https://asios.ai> | <https://karlex.ai> >> > > > -- > Yours truly

1 0

CONFIG_XXX vs CONFIG_HAVE_XXX ?
by Alexandre Ferrieux May 19, 2025

May 19, 2025

Hi, What is the rationale behind having, for some feature XXX, both configuration macros CONFIG_XXX and CONFIG_HAVE_XXX ? For example, I'd love to use ftrace's new life-saving feature "funcgraph-retval" (which instantly shows the root cause of an error returned from a fully graphed function). Unfortunately, on all recent Debians the config says: CONFIG_HAVE_FUNCTION_GRAPH_RETVAL=y # CONFIG_FUNCTION_GRAPH_RETVAL is not set As a result, the feature is not available. Okay, I'll just enable and recompile. However, I'm wondering why in the first place such a situation is possible: what is the intention there ? Either we want the feature, or we don't (e.g. because it may incur some CPU or memory overhead), but why sit halfway in between ? Thanks in advance, -Alex

3 4

how to begin contributing to the linux kernel
by Dileep Sankhla May 13, 2025

May 13, 2025

Hello everyone, 1. My background I am a professional software engineer with 1.5 years of experience. For 6-7 months, I worked on creating my own Unix-like kernel from scratch for my x86-64 laptop. I wrote the device drivers for NVMe over PCIe, a local APIC timer and interrupt controller (x2APIC mode), a PS/2 keyboard, the real-time clock, and terminal output. I developed my own UEFI bootloader and worked on an interrupt handling mechanism. I implemented a physical memory manager using the free stack data structure and an ext2 file system driver. I briefly worked on my own flat file system as an experiment. I also experimented with my own 64-bit paging for a while. That's it. I know a good amount of C (I think my skill in C is between beginner-intermediate level) and I have a year of experience using it. I also did Google Summer of Code with KDE in 2018. There I used C++. 2. My question How can I begin contributing to the linux kernel? I already have one contribution but that is about fixing a checkpatch.pl warning (see [0]). Now I think I want to begin fixing a beginner bug in a subsystem like "filesystem". I am interested in filesystems. I searched a lot about this over the internet but I can't figure out what I need to do and need to learn to begin contributing again. Thanks. Regards, Dileep [0]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?…

2 2

Documentation for Regmap API?
by Ali Nasrolahi May 3, 2025

May 3, 2025

Hi, I've been exploring the Regmap API and was wondering if there’s a dedicated documentation page that provides an overview or explanation of the API itself. So far, I’ve looked into the following: 1. https://docs.kernel.org/driver-api/index.html 2. The Documentation/ directory in both the mainline and linux-next trees. 3. The regmap tree mentioned in the MAINTAINERS file: --- REGISTER MAP ABSTRACTION M: Mark Brown <broonie(a)kernel.org> L: linux-kernel(a)vger.kernel.org S: Supported T: git git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap.git F: Documentation/devicetree/bindings/regmap/ F: drivers/base/regmap/ F: include/linux/regmap.h --- Despite checking these locations, I couldn’t find a dedicated document explaining the Regmap API in detail. 1. Is there any such documentation available that I might have missed? 2. If not, I’d be happy to contribute and help write one. Should I bring this up on the linux-doc mailing list for further guidance on how to proceed? Any pointers or suggestions would be greatly appreciated. Thanks a lot, Ali

3 3