i vaguely remember that decades ago openBSD needed swap on, while free could be run without. if it was the same issue, than there should be a fix out there known by the mighty cvs.
On Sun 1. Feb 2026 at 11:54, Crystal Kolipe <[email protected]> wrote: > On Sun, Feb 01, 2026 at 07:29:36AM +0200, ????????? wrote: > > Could you please elaborate? > > Modern hardware has enough RAM to be able to theoretically run without > swap. > > Yes, and we have been doing this for many years on other architecures. > > However there must be a bug somewhere in the arm64 code that is mitigated > by > configuring swap - even a tiny swap partiiton of 32 Mb seems to mitigate > the > effects, on a machine with over 3 Gb of physical ram free. > > I spent a day or so looking at the code a few years ago, but found nothing > obvious. Since there was an obvious work-around, I moved on to other more > important things. > > > If you don't need to hibernate, disabling swap to save NVRAM sounds > > reasonable to me. > > Presumably by NVRAM you mean writes to flash memory, (which is technically > NVRAM, but NVRAM is generally understood to mean CMOS ram for the RTC, > etc). > > Yes, in principle you are correct, but in practice it's not going to make > much > of a difference, especially if the swap partition is very small. > > Remember that by default the swap partition is used for crash dumps as > well, > so most machines used for serious development are likely going to have > large a > large swap partition configured. > > So overall - if you configure your amd64 daily driver laptop to run without > swap to avoid wearing out the SSD, you'll _probably_ be OK. > > If you have an esoteric machine of an architecture that only a handful of > other people are using, then I would presume that they are mostly running > with swap, and if you don't configure swap then you are exercising code > paths > that are less well tested. >
