From mboxrd@z Thu Jan 1 00:00:00 1970 From: arnd@arndb.de (Arnd Bergmann) Date: Tue, 01 Jul 2014 13:35:49 +0200 Subject: [GIT PULL] nommu fixes for 3.17-rc1 In-Reply-To: <20140630170738.GB30740@arm.com> References: <20140630144355.GH14781@pengutronix.de> <20140630170738.GB30740@arm.com> Message-ID: <4831043.FLJmCthlNt@wuerfel> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Monday 30 June 2014 18:07:38 Will Deacon wrote: > On Mon, Jun 30, 2014 at 06:00:37PM +0100, Linus Walleij wrote: > > However if it is a good testbase for people synthesizing these old non-MMU > > ARM cores still and using Linux with them, then it would make sense to > > get that into shape and supported I guess? Especially now that the > > Integrator machine is soon-ish even multiplatform (famous last words). > > > > So we need some input here ... (adding Greg Ungerer to Cc) FWIW, qemu supports integrator with arm946e in principle. There are a few bugs that I found recently with the qemu model running recent Linux kernels, but they are all trivially fixable. I haven't tried running !MMU though, only ARM926 and ARM1026 with MMU, as I haven't managed to find or build a user space image. > To be honest, I'm finding it more and more difficult to care about uclinux > outside of Cortex-M. Could we simply decide to support only v7m and rip out > the code for other !MMU targets, or are there actually users of those things > with mainline kernels? As I said here earlier, at91x40 is the only pre-v7-m platform we support upstream that doesn't have an MMU, and we are going to remove it at the point when mach-at91 moves to multiplatform. I suppose uClinux is the place for arm7tdmi these days, I just downloaded their latest release and found support for these additional platforms that were never upstreamed: * S5C7375 (arm940t) * S3C2500 (arm940t) * S3C3410 (arm7tdmi) * S3C4510B (arm7tdmi) * S3C44B0 (arm7tdmi) * P2001 (arm9tdmi) * LPC22XX (arm7tdmi) * LPC28XX (arm7tdmi) The last one of these is the only one that actually builds, all the other ones have been broken for many years as far as I can tell and were just kept in the tree for reference, or because nobody bothered to remove them. In contrast, the latest uclinux-2.4 kernel supports 23 (!) platforms that I assume all work to some degree. This supports the position that people using armv4t-nommu are either going to stay on ancient kernels or moving on to other CPUs later. There are a few other semi-recent kernel trees outside of uClinux-dist: * at91x40 (to be dropped, as mentioned above) * lapis ml67xxxx (also mentioned before) * nuvoton nuc700 (abandoned at 3.1?) https://github.com/rhuitl/linux/commits/nuc700 * multiple cortex-m on 2.6.33 (actively maintained by emcraft) https://github.com/EmcraftSystems/linux-emcraft Greg, do you have any other input? Do you think anybody is still interested in upstream !MMU Linux on ARMv4T or ARMv5? Arnd