bittorf:

Here the serial console output

(hangs forever)


bittorf:

a WDR4300 (some arch) works without this issue

john:

Hi Bastel Basti

image too big and decompressor overriding itself with the payload ?

John

bittorf:

Good idea: i will investigate tomorrow.

bittorf:

Bastelbasti reports back:

When building r17677 (head) with GCC10 instead of the default GCC11,
the image boots and works, although the resulting kernel
target-mips_24kc_musl/linux-ath79_generic/linux-5.4.150/vmlinuz
is much bigger: 6.25mb instead of 4.9mb with GCC11.

Other models or also affected, this problem is described e.g. here:
https://forum.openwrt.org/t/gcc11-images-are-not-working-on-cpe510v3

The next step can be a deeper look into the miscompiled
target/linux/ath79/image/lzma-loader/src/loader.c

bittorf:

After looking into the source: the loader.c is fine,
my bet: the problem is in ./target/linux/ath79/image/lzma-loader/src/LzmaDecode.c

UsernameIsAlreadyTakenByMe:

Hi,
I also had this issue by upgrading with the following image
[[https://downloads.openwrt.org/snapshots/targets/ath79/generic/openwrt-ath79-generic-tplink_tl-wr1043nd-v1-squashfs-sysupgrade.bin|tplink_tl-wr1043nd-v1-squashfs-sysupgrade.bin]]
69f36b0ea4631eafc24a5b9b2a71c1559005359677604b92f88bd1bbf522b512
5056.3 KB
Sun Oct 10 16:23:24 2021

I tried to upgrade from 21.02.0-rc3 and 19.07.8

bittorf:

fixed with r18106

toolchain/gcc: revert a faulty upstream commit that caused miscompiled code

This broke at least the OKLI loader, but possibly other things as well

Signed-off-by: Felix Fietkau <nbd@nbd.name>

Thanks a lot!