Date: Fri, 6 Jul 2018 17:06:03 -0700 (PDT) From: Don Lewis <truckman@FreeBSD.org> To: Warner Losh <imp@bsdimp.com> Cc: "Rodney W. Grimes" <rgrimes@freebsd.org>, Hans Petter Selasky <hselasky@freebsd.org>, src-committers <src-committers@freebsd.org>, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r336025 - in head/sys: amd64/include i386/include Message-ID: <tkrat.0557727c4f01a481@FreeBSD.org> In-Reply-To: <CANCZdfpZpeCVSs1%2BMptwuN76SED-W4XyLCyQfrd4OpGwFk8Hrg@mail.gmail.com> References: <201807061013.w66ADgbJ087546@repo.freebsd.org> <201807061532.w66FWPEN052842@pdx.rh.CN85.dnsmgr.net> <CANCZdfpZpeCVSs1%2BMptwuN76SED-W4XyLCyQfrd4OpGwFk8Hrg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 6 Jul, Warner Losh wrote: > On Fri, Jul 6, 2018 at 9:32 AM, Rodney W. Grimes < > freebsd@pdx.rh.cn85.dnsmgr.net> wrote: > >> > Author: hselasky >> > Date: Fri Jul 6 10:13:42 2018 >> > New Revision: 336025 >> > URL: https://svnweb.freebsd.org/changeset/base/336025 >> > >> > Log: >> > Make sure kernel modules built by default are portable between UP and >> > SMP systems by extending defined(SMP) to include defined(KLD_MODULE). >> > >> > This is a regression issue after r335873 . >> > >> > Discussed with: mmacy@ >> > Sponsored by: Mellanox Technologies >> >> Though this fixes the issue, it also means that now when >> anyone intentionally builds a UP kernel with modules >> they are getting SMP support in the modules and I am >> not sure they would want that. I know I don't. >> > > > On UP systems, these additional opcodes are harmless. They take a few extra > cycles (since they lock an uncontested bus) and add a couple extra memory > barriers (which will be NOPs). On MP systems, atomics now work by default. > Had we not defaulted like this, all modules built outside of a kernel build > env would have broken atomics. Given that (a) the overwhelming majority > (99% or more) is SMP and (b) the MP code merely adds a few cycles to what's > already a not-too-expensive operation, this was the right choice. > > It simply doesn't matter for systems that are relevant to the project > today. While one could try to optimize this a little (for example, by > having SMP defined to be 0 or 1, say, and changing all the ifdef SMP to if > (defined(SMP) && SMP != 0)), it's likely not going to matter enough for > anybody to make the effort. UP on x86 is simply not relevant enough to > optimize for it. Even in VMs, people run SMP kernels typically even when > they just allocate one CPU to the VM. > > So while we still support the UP config, and we'll let people build > optimized kernels for x86, we've flipped the switch from pessimized for SMP > modules to pessimized for UP modules, which seems like quite the reasonable > trade-off. > > Were it practical to do so, I'd suggest de-orbiting UP on x86. However, > it's a lot of work for not much benefit and we'd need to invent much crazy > to get there. I would distinguish i386 from amd64 here. SMP is pretty rare and exotic in the i386 world. I do have one dual socket Pentium 3 machine here and even though I bought the parts for it used on eBay, it was still pretty pricey. That purchase was kind of a waste since it was shortly before the Athlon 64 X2 CPUs were released. I still have two viable 32-bit x86 machines here that get frequent usage. One runs 24x7 and has a Via C3 CPU. I started looking at migrating off this hardware. To get lower power consumption as well as ECC RAM I'd probably have to go with one of the Supermicro Atom boards. Those are pretty expensive, so I'd probably end up spending about half as much as what it cost to put together my fully-loaded Ryzen machine last summer. At that price, the payback time from the power savings is really long. This machine is mostly idle, so I really don't need more CPU power or RAM. The other machine is my Pentium-M laptop, which is mostly used for light browsing and as a vnc client when I'm on the road. Performance is acceptable for those uses. Both machines run stripped down UP kernels to avoid wasting RAM unnecessarily and to optimize CPU cycles on the laptop. A good reason for continuing UP support on x86 is to make it easy to test UP builds in the MI parts of the kernel so that we don't break things for the embedded architectures. Unfortunately "make universe" currently doesn't have any UP kernels, so I've managed to commit changes that break UP builds and not known it until I received reports of broken builds from other users.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?tkrat.0557727c4f01a481>