On 4/30/19 3:40 AM, Peter Maydell wrote: > On Mon, 29 Apr 2019 at 21:13, Richard Henderson > <richard.hender...@linaro.org> wrote: >> >> On 4/29/19 9:21 AM, Peter Maydell wrote: >>> This looks ok code-wise, but we'd need to hide it behind >>> a defaults-to-off x-something property if we wanted to >>> commit it before the kernel ABI is fixed. >> >> I'm not intending to change the user-level abi, only the >> internal abi within qemu, for handling of the elf notes. > > You're changing target_mmap(), which is used by the > guest mmap syscall, though, right?
Yes, but it's also used by elfload.c to map the executable. > I think that we shouldn't allow guest binaries written > to the ad-hoc TARGET_PROT_BTI ABI to work without an > explicit x-something command line argument to QEMU... The guest binary is not written to an ad-hoc abi. It is written to the finalized ELF note abi, which I am trying to implement with a *private* QEMU abi. r~
