| Message ID | 1538117388-11945-1-git-send-email-yamada.masahiro@socionext.com |
|---|---|
| State | New |
| Headers | show |
| Series | samples: run headers_install for host arch instead of target arch | expand |
On Fri, Sep 28, 2018 at 8:51 AM Masahiro Yamada <yamada.masahiro@socionext.com> wrote: > > Some samples search headers in $(objtree)/usr/include, which is made > available by "make headers_install". It is not kernel-space code but > host programs that need this header search path. > > Commit 3fca1700c4c3 ("kbuild: make samples really depend on > headers_install") is wrong because it installs headers of the target > architecture. Besides, UML fails to build with CONFIG_SAMPLES=y > because UML does not support headers_install. > > Invoke "make headers_install" for the _host_ architecture in the > prepare stage. Introduce CONFIG_HOST_HEADERS_INSTALL so that this > happens only for samples that need it. > > I also removed 'Documentation/: headers_install', which I guess is > stale code. > > Fixes: 3fca1700c4c3 ("kbuild: make samples really depend on headers_install") > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> > Cc: David Howells <dhowells@redhat.com> I don't understand this one. Why would we want to build the samples for the host architecture rather than the target architecture? I would think that the bug is that we try to build them using HOSTCC, and they end up being unusable on the target as a consequence. Arnd
2018年9月29日(土) 4:38 Arnd Bergmann <arnd@arndb.de>: > > On Fri, Sep 28, 2018 at 8:51 AM Masahiro Yamada > <yamada.masahiro@socionext.com> wrote: > > > > Some samples search headers in $(objtree)/usr/include, which is made > > available by "make headers_install". It is not kernel-space code but > > host programs that need this header search path. > > > > Commit 3fca1700c4c3 ("kbuild: make samples really depend on > > headers_install") is wrong because it installs headers of the target > > architecture. Besides, UML fails to build with CONFIG_SAMPLES=y > > because UML does not support headers_install. > > > > Invoke "make headers_install" for the _host_ architecture in the > > prepare stage. Introduce CONFIG_HOST_HEADERS_INSTALL so that this > > happens only for samples that need it. > > > > I also removed 'Documentation/: headers_install', which I guess is > > stale code. > > > > Fixes: 3fca1700c4c3 ("kbuild: make samples really depend on headers_install") > > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> > > Cc: David Howells <dhowells@redhat.com> > > I don't understand this one. Why would we want to build the samples for > the host architecture rather than the target architecture? > > I would think that the bug is that we try to build them using > HOSTCC, and they end up being unusable on the target as > a consequence. I just read Makefiles under samples/, and all that require -I$(objtree)/usr/include are listed in hostprogs-y, so they are built for host architectures. If this is different from our intent, we need to fix this first. -- Best Regards Masahiro Yamada
diff --git a/Makefile b/Makefile index 0c90c43..f04564c 100644 --- a/Makefile +++ b/Makefile @@ -1029,10 +1029,9 @@ ifdef CONFIG_GDB_SCRIPTS endif +$(call if_changed,link-vmlinux) -# Build samples along the rest of the kernel. This needs headers_install. +# Build samples along the rest of the kernel. ifdef CONFIG_SAMPLES vmlinux-dirs += samples -samples: headers_install endif # The actual objects are generated when descending, @@ -1098,6 +1097,9 @@ archprepare: archheaders archscripts prepare1 scripts_basic prepare0: archprepare gcc-plugins $(Q)$(MAKE) $(build)=. +ifdef CONFIG_HOST_HEADERS_INSTALL + $(Q)$(MAKE) -f $(srctree)/Makefile ARCH=$(SUBARCH) headers_install +endif # All the preparing.. prepare: prepare0 prepare-objtool @@ -1682,8 +1684,6 @@ endif $(cmd_crmodverdir) $(Q)$(MAKE) KBUILD_MODULES=$(if $(CONFIG_MODULES),1) \ $(build)=$(build-dir) -# Make sure the latest headers are built for Documentation -Documentation/ samples/: headers_install %/: prepare scripts FORCE $(cmd_crmodverdir) $(Q)$(MAKE) KBUILD_MODULES=$(if $(CONFIG_MODULES),1) \ diff --git a/samples/Kconfig b/samples/Kconfig index bd133ef..fcd6b7c 100644 --- a/samples/Kconfig +++ b/samples/Kconfig @@ -1,3 +1,6 @@ +config HOST_HEADERS_INSTALL + bool + menuconfig SAMPLES bool "Sample kernel code" help @@ -95,6 +98,7 @@ config SAMPLE_CONFIGFS config SAMPLE_CONNECTOR tristate "Build connector sample -- loadable modules only" depends on CONNECTOR && m + select HOST_HEADERS_INSTALL help When enabled, this builds both a sample kernel module for the connector interface and a user space tool to communicate @@ -104,6 +108,7 @@ config SAMPLE_CONNECTOR config SAMPLE_SECCOMP tristate "Build seccomp sample code -- loadable modules only" depends on SECCOMP_FILTER && m + select HOST_HEADERS_INSTALL help Build samples of seccomp filters using various methods of BPF filter construction. @@ -149,6 +154,7 @@ config SAMPLE_VFIO_MDEV_MBOCHS config SAMPLE_STATX bool "Build example extended-stat using code" depends on BROKEN + select HOST_HEADERS_INSTALL help Build example userspace program to use the new extended-stat syscall.
Some samples search headers in $(objtree)/usr/include, which is made available by "make headers_install". It is not kernel-space code but host programs that need this header search path. Commit 3fca1700c4c3 ("kbuild: make samples really depend on headers_install") is wrong because it installs headers of the target architecture. Besides, UML fails to build with CONFIG_SAMPLES=y because UML does not support headers_install. Invoke "make headers_install" for the _host_ architecture in the prepare stage. Introduce CONFIG_HOST_HEADERS_INSTALL so that this happens only for samples that need it. I also removed 'Documentation/: headers_install', which I guess is stale code. Fixes: 3fca1700c4c3 ("kbuild: make samples really depend on headers_install") Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> Cc: David Howells <dhowells@redhat.com> --- Makefile | 8 ++++---- samples/Kconfig | 6 ++++++ 2 files changed, 10 insertions(+), 4 deletions(-) -- 2.7.4