mirror of
https://github.com/freebsd/freebsd-src.git
synced 2026-10-28 11:25:21 +01:00
Andrew Gallatin
7e2a425f7e
Implement buf_ring/drbr deferred transmit in iflib. This is intended to allow the new simpler code path to replace mp_ring. This patch makes the simple_tx outperform mp_ring by a wide margin when CPU is the bottleneck (eg, cannot fill the NIC). See graphs at: https://people.freebsd.org/~gallatin/mpring_vs_simple_tx Note that the buf ring is used for contention, not capacity. Eg, it is used as a place for contending threads to put packets without waiting for a mutex. It is not designed to act as a software ring on top of the hardware descriptors provided by the underlying NIC driver. "stranded packets" are exceedingly rare due to the fact that if there is enough load to use the buf_ring, there will probably be more load coming that can be a drainer. Not scheduling a gtask to drain is intentional, and we really on the timer as a fallback. One thing I noticed while developing this patch is that a simple mutex with no deferral generally outperformed both mp_ring and drbr at high levels of contention for the same queue. This is inherent in a bounded MPSC queue where multiple producers are contending on claiming ring entries. So I came up with the idea of bounding the number of producers such that the deferral ring would devolve to a mutex when contention was high. Identifying the crossover point in a general way was hard. On different machines, the point between a mutex and a deferral ring was very different and also depended on the placement of the producers. I eventually realized that on the large AMD EPYC servers that I was testing on, the crossover point generally coincided with the work spilling into another CCX. So I developed an approach where we limit the number of producers by limiting simultanious producers to the same L3 AND by bounding the number of simultanious producers. This limit can be adjusted via the sysctl net.iflib.max_producers. The number of packets drained from the deferral ring is limited by net.iflib.simple_drain_quota. The intent is to process just enough packets in the gtaskq context so as to make space to allow threads to make progress. The task also uses a trylock so as to avoid blocking, waiting for a thread to drain. This change also enables ALTQ support for simple tx. Reviewed by: kbowling Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D58901
FreeBSD Kernel Source:
This directory contains the source files and build glue that make up the FreeBSD kernel and its modules, including both original and contributed software.
Kernel configuration files are located in the conf/ subdirectory of each
architecture. GENERIC is the configuration used in release builds. NOTES
contains documentation of all possible entries. LINT is a compile-only
configuration used to maximize build coverage and detect regressions.
Documentation:
Source code documentation is maintained in a set of man pages, under section 9.
These pages are located in share/man/man9, from the
top-level of the src tree. Consult intro(9)
for an overview of existing pages.
Some additional high-level documentation of the kernel is maintained in the Architecture Handbook.
Source Roadmap:
| Directory | Description |
|---|---|
| amd64 | AMD64 (64-bit x86) architecture support |
| arm | 32-bit ARM architecture support |
| arm64 | 64-bit ARM (AArch64) architecture support |
| bsm | Basic Security Module headers - audit(4) and bsm(3) |
| cam | Common Access Method storage subsystem - cam(4) and ctl(4) |
| cddl | CDDL-licensed optional sources such as DTrace |
| compat | Linux compatibility layer, FreeBSD 32-bit compatibility |
| conf | kernel build glue |
| contrib | 3rd-party imported software such as OpenZFS |
| crypto | crypto drivers |
| ddb | interactive kernel debugger - ddb(4) |
| dev | device drivers and other arch independent code |
| dts | FreeBSD-specific device tree sources |
| fs | most filesystems, excluding UFS, NFS, and ZFS |
| gdb | kernel remote GDB stub - gdb(4) |
| geom | GEOM framework - geom(4) |
| gnu | GPL-licensed sources |
| i386 | i386 (32-bit x86) architecture support |
| isa | PC ISA bus implementation |
| kern | main part of the kernel |
| kgssapi | kernel-space GSSAPI implementation |
| libkern | libc-like and other support functions for kernel use |
| modules | kernel module infrastructure |
| net | core networking code |
| net80211 | wireless networking (IEEE 802.11) - net80211(4) |
| netgraph | graph-based networking subsystem - netgraph(4) |
| netinet | IPv4 protocol implementation - inet(4) |
| netinet6 | IPv6 protocol implementation - inet6(4) |
| netipsec | IPsec protocol implementation - ipsec(4) |
| netlink | kernel network configuration protocol - netlink(4) |
| netpfil | packet filters - ipfw(4), pf(4), and ipfilter(4) |
| netsmb | Server Message Block protocol implementation |
| nfs | common code and headers for Network File System |
| nfsclient | NFS client implementation for mounting and stats |
| nfsserver | NFS server implementation for exporting local filesystems to network |
| nlm | Network Lock Management protocol implementation |
| ofed | OpenFabrics Enterprise Distribution implementation |
| opencrypto | OpenCrypto framework - crypto(7) |
| powerpc | PowerPC/POWER (32 and 64-bit) architecture support |
| riscv | 64-bit RISC-V architecture support |
| rpc | Open Network Computing Remote Procedure Call implementation - rpc(3) |
| security | security facilities - audit(4) and mac(4) |
| sys | kernel headers |
| teken | terminal emulation interface implementation - teken(3) |
| tests | kernel unit tests |
| tools | kernel build scripts and utilities |
| ufs | Unix File System - ffs(4) |
| vm | virtual memory system |
| x86 | code shared by AMD64 and i386 architectures |
| xdr | External Data Representation implementation - xdr(3) |
| xen | Xen hypervisor support - xen(4) |