On 1/31/19 11:15 PM, Paolo Bonzini wrote: > On 31/01/19 22:48, Philippe Mathieu-Daudé wrote: >> There is something I don't understand here: Does CONFIG_XEN in >> Kconfig.host take precedence over the target configs? I'm looking at >> these configs: >> >> if supported_xen_target $target; then >> echo "CONFIG_XEN=n" >> $config_target_mak >> if test "$xen_pci_passthrough" = yes; then >> echo "CONFIG_XEN_PCI_PASSTHROUGH=y" >> "$config_target_mak" >> fi >> fi > > You're right, Kconfig.host should contain CONFIG_XEN_BACKEND and not > CONFIG_XEN.
Now when I disable Xen, exec.o isn't rebuilt, so I get link errors: /usr/bin/ld: exec.o: in function `reclaim_ramblock': qemu/exec.c:2392: undefined reference to `xen_invalidate_map_cache_entry' /usr/bin/ld: exec.o: in function `qemu_ram_ptr_length': qemu/exec.c:2531: undefined reference to `xen_map_cache' /usr/bin/ld: exec.o: in function `qemu_map_ram_ptr': qemu/exec.c:2498: undefined reference to `xen_map_cache' /usr/bin/ld: exec.o: in function `qemu_ram_block_from_host': qemu/exec.c:2573: undefined reference to `xen_ram_addr_from_mapcache' /usr/bin/ld: exec.o: in function `address_space_unmap': qemu/exec.c:3699: undefined reference to `xen_invalidate_map_cache_entry' /usr/bin/ld: exec.o: in function `address_space_cache_destroy': qemu/exec.c:3791: undefined reference to `xen_invalidate_map_cache_entry' /usr/bin/ld: exec.o: in function `qemu_ram_ptr_length': qemu/exec.c:2528: undefined reference to `xen_map_cache' /usr/bin/ld: exec.o: in function `qemu_map_ram_ptr': qemu/exec.c:2495: undefined reference to `xen_map_cache' collect2: error: ld returned 1 exit status make[1]: *** [Makefile:204: qemu-system-i386] Error 1 Moving those function stubs into a real xen-stub.c file would be simpler from the buildsys PoV IMHO, but I also remember we prefer to avoid stubs, so not sure what's better here, neither what's wrong with make rules.
