Hi! On Mon, 2023-11-13 at 21:41:48 +0900, Simon Richter wrote: > On 13.11.23 19:08, Guillem Jover wrote: > > I was checking for dpkg arch definitions to cleanup and stumbled over > > the uclinux-any ones (added as part of #455501), where I noticed the > > µCLinux fork got merged into mainline Linux in 2.5.46, and then > > several of the no-MMU ports got removed from there more recently. > > I must admit that I haven't really used uClinux, or rather any MMUless > architecture in ages, and the company that I originally wrote this for (and > which did indeed run an internal Debian mirror with packages for a > big-endian MMUless ARM box running uClinux) has since shut down, so I > believe there are no users left. > > I'm not sure if it's entirely dead or just about to get a resurgence because > of FPGAs and FPGA SoCs with integrated hard CPUs that are gaining in > popularity. > > On the pure FPGA side, my expectation is that if it runs Linux, it will have > an MMU. > > The major commercial soft CPUs, Microblaze and NiosII, have MMU designs with > very little overhead (basically, a TLB and a processor mode for the TLB miss > handler to run in), so there is little incentive to run MMUless. > > RISC-V, ARM and SPARC implementations for FPGAs might be interesting to run > MMUless (because their MMU interface standards take a lot of resources to > implement), but I'm not aware of anyone actually doing that. > > On the SoC side, I think most available ICs contain an MMU, and those that > don't aren't really used with Linux (there are some MMUless Cortex-R cores > that can address enough memory, but I doubt that's a practical concern). > > So I think these can be removed.
Thanks for the background info! I'll then queue the patch removing these. Regards, Guillem
