Hi again, Jason, On Wed, 10 Jun 2020 at 11:09, Jason A. Donenfeld <[email protected]> wrote: > > Eventually I can probably get this building and testing and find some > hardware for this and such. But if you'd like things to move faster, > trying to reproduce the issue in the qemu test suite will result in a > quicker fix.
After a bit more digging [1], I believe I've narrowed it down. CONFIG_THUMB2_AVOID_R_ARM_THM_JUMP11=y is required in order to avoid the emission of R_ARM_THM_JUMP11 relocations in the WireGuard module. I'm now wondering why the compat modules haven't exhibited the same problem (maybe it was just a fluke), but since this kconfig option effectively implies -fno-optimize-sibling-calls [2], it's quite a hefty hammer. Is this something that can be solved in the WireGuard build itself? Thanks in advance, Rui [1] https://github.com/openwrt/openwrt/pull/3079#issuecomment-645297337 [2] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm/Makefile?h=linux-5.4.y#n125 _______________________________________________ openwrt-devel mailing list [email protected] https://lists.openwrt.org/mailman/listinfo/openwrt-devel
