2018-04-26 13:10 GMT-07:00 Matthias Schiffer <mschif...@universe-factory.net>: > On 04/26/2018 09:22 PM, Etienne Champetier wrote: >> Hi Matthias, >> >> 2018-04-26 11:43 GMT-07:00 LEDE Commits <lede-comm...@lists.infradead.org>: >>> neoraider pushed a commit to openwrt/openwrt.git, branch lede-17.01: >>> https://git.lede-project.org/eed9d40133fe6468cc253d0345c7d7332d2aaa7c >>> >>> commit eed9d40133fe6468cc253d0345c7d7332d2aaa7c >>> Author: Matthias Schiffer <mschif...@universe-factory.net> >>> AuthorDate: Thu Apr 26 20:04:48 2018 +0200 >>> >>> ar71xx: Ubiquiti Airmax M: add relocate-kernel to invalidate cache >>> >>> Some Ubiquiti U-boot versions, in particular the "U-Boot 1.1.4.2-s956 >>> (Jun 10 2015 - 10:54:50)" found with AirOS 5.6, do not correctly flush >>> the >>> caches for the whole kernel address range after decompressing the kernel >>> image, leading to hard to debug boot failures, depending on kernel >>> version >>> and configuration. >>> >>> As a workaround, prepend the relocate-kernels loader, which will >>> invalidate >>> the caches after moving the kernel to the correct load address. >> >> So if I understand correctly, >> with this commit I don't need to downgrade my bullet M to v.5.5.10 >> (U-Boot 1.1.4.2-s594 (Dec 5 2012 - 15:23:07)) >> and I can just downgrade to v.5.6.x (U-Boot 1.1.4.2-s956 (Jun 10 2015 >> - 10:54:50)) and maybe even the firsts 6.0.X (without the signed >> image) >> >> mtdparts differences between the 2 uboot version: >> https://community.ubnt.com/t5/airOS-SDK-Custom-Development/5-6-2-TFTP-ERROR/m-p/1310208/highlight/true#M2354 > > AFAIK, both AirOS 5.5 and 5.6 use the same flash layout. The mtdparts with > the EEPROM at 0x00730000 (also described in > https://wiki.openwrt.org/toh/ubiquiti/airmaxm ) is bogus and was not > directly caused by the newer Uboot or AirOS, but by installing "broken" > OpenWrt images (everything before [1]) from AirOS 5.6; the commit message > of [1] contains further details. Certain Uboot versions coming with 5.6 > also did not unlock the flash chip, leading to unwritable flash on OpenWrt > (fixed in [2]). > > The cache invalidation issue recently appeared after the update to kernel > 4.4.120 in lede-17.01, but I'm not sure if the OpenWrt/LEDE default kernel > configuration is affected. > > With these 3 fixes, installing OpenWrt from AirOS 5.6 should work fine; I > don't know about 6.0. It seems that I never backported [2] to lede-17.01 > because it requires 2 other patches on kernel 4.4, but we have a downstream > patch for that in Gluon [3].
Ok, so as of today, if I want to install 17.01, because [2] is not backported I need to flash 5.5.10 first (old uboot) If I install master 5.6.X should be fine I just bought a bullet M for testing / to play with, if you have any commands in mind that you want me to run on different AirOS versions, just tell me Thanks for the explanation Etienne > > Regards, > Matthias > > > [1] > https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=076d58d3440f382c536ea8874f58b0df23c263bc > [2] > https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=1abb5c6db11bd8a3c79b9e24ac4511985fe39ea4 > [3] > https://raw.githubusercontent.com/freifunk-gluon/gluon/563f351b78924bb12c2db49f63baa4db86c2c16b/patches/lede/0027-generic-mtd-backport-unlock-support-and-asserted-fixes-for-certain-Winbond-Spansion-flash-chips.patch > > > > >> >> Thanks >> Etienne >> > _______________________________________________ Lede-dev mailing list Lede-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/lede-dev
