From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from zeniv.linux.org.uk ([195.92.253.2]:60510 "EHLO ZenIV.linux.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726144AbeIKBWb (ORCPT ); Mon, 10 Sep 2018 21:22:31 -0400 Date: Mon, 10 Sep 2018 21:26:37 +0100 From: Al Viro To: kbuild test robot Cc: kbuild-all@01.org, linux-fsdevel@vger.kernel.org Subject: Re: [vfs:work.termios 6/10] arch/arm/mach-sa1100/assabet.c:550:60: error: 'DOMAIN_IO' undeclared; did you mean 'DOMAIN_BUS_IPI'? Message-ID: <20180910202636.GM19965@ZenIV.linux.org.uk> References: <201809110224.2hcw3kvk%fengguang.wu@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <201809110224.2hcw3kvk%fengguang.wu@intel.com> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Tue, Sep 11, 2018 at 02:34:32AM +0800, kbuild test robot wrote: > tree: https://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs.git work.termios > head: a3318f219be23141d4784902d201e4d8e8d6d0d2 > commit: b62802628ae228744508363e28118427904bc338 [6/10] untangle asm/termios.h mess > config: arm-neponset_defconfig (attached as .config) > compiler: arm-linux-gnueabi-gcc (Debian 7.2.0-11) 7.2.0 > reproduce: Huh? > In file included from include/linux/const.h:4:0, > from include/linux/list.h:8, > from include/linux/module.h:9, > from arch/arm/mach-sa1100/assabet.c:14: > arch/arm/mach-sa1100/assabet.c: In function 'map_sa1100_gpio_regs': > >> arch/arm/mach-sa1100/assabet.c:550:60: error: 'DOMAIN_IO' undeclared (first use in this function); did you mean 'DOMAIN_BUS_IPI'? What in hell does that thing have to do with asm/termios.h? Especially on arm, where the only visible change in that commit is that include of asm/termios.h stopped pulling linux/uacess.h? Oh, dear... So we have this: linux/serial_core.h -> linux/tty.h -> [uapi] linux/termios.h -> [uapi] asm/termios.h -> (due to generic-y) asm-generic/termios.h -> linux/uaccess.h and [uapi] asm-generic/termios.h. with linux/uaccess.h -> asm/uaccess.h -> asm/domain.h being what used to bring DOMAIN_IO in. And that change has eliminated the side-trip into asm-generic/termios.h, replacing it with pulling uapi asm-generic/termios.h. With missing indirect include of asm/domain.h as the result. Note that *nothing* in linux/serial_core.h or the stuff pulled by it has any need of linux/uaccess.h. Wonderful, innit? As it were, prior to that commit any pull of asm/termios.h (realistically - linux/termios.h) used to imply the pull of linux/uaccess.h. Sigh... Let's see... Includes of linux/termios.h: arch/arm/mach-ep93xx/core.c:29:#include arch/arm/mach-integrator/integrator_ap.c:29:#include drivers/char/pcmcia/synclink_cs.c:68:#include drivers/misc/ibmasm/uart.c:25:#include drivers/net/slip/slhc.c:66:#include drivers/parisc/superio.c:73:#include drivers/s390/char/sclp_con.c:15:#include drivers/tty/isicom.c:119:#include drivers/tty/serial/icom.c:18:#include drivers/tty/serial/ifx6x60.c:28:#include drivers/tty/serial/serial-tegra.c:33:#include drivers/tty/serial/serial_mctrl_gpio.c:12:#include drivers/tty/synclink.c:93:#include drivers/tty/synclink_gt.c:69:#include drivers/tty/synclinkmp.c:66:#include drivers/tty/tty_baudrate.c:8:#include drivers/tty/tty_ioctl.c:11:#include include/linux/serdev.h:18:#include include/linux/tty.h:7:#include include/linux/tty_driver.h:246:#include net/appletalk/ddp.c:57:#include /* For TIOCOUTQ/INQ */ net/ax25/af_ax25.c:37:#include /* For TIOCINQ/OUTQ */ net/ax25/ax25_ip.c:28:#include /* For TIOCINQ/OUTQ */ net/decnet/dn_nsp_in.c:65:#include net/decnet/dn_nsp_out.c:56:#include net/ieee802154/socket.c:25:#include /* For TIOCOUTQ/INQ */ net/netlink/af_netlink.c:38:#include net/netrom/af_netrom.c:35:#include /* For TIOCINQ/OUTQ */ net/netrom/nr_route.c:30:#include /* For TIOCINQ/OUTQ */ net/qrtr/qrtr.c:17:#include /* For TIOCINQ/OUTQ */ net/rose/af_rose.c:39:#include net/rose/rose_route.c:30:#include /* For TIOCINQ/OUTQ */ net/unix/af_unix.c:97:#include net/x25/af_x25.c:56:#include /* For TIOCINQ/OUTQ */ Excluding the ones that have an explicit include of linux/uaccess.h and net/sock.h (it pulls linux/uaccess.h unconditionally) we are left with arch/arm/mach-ep93xx/core.c:29:#include arch/arm/mach-integrator/integrator_ap.c:29:#include drivers/misc/ibmasm/uart.c:25:#include drivers/parisc/superio.c:73:#include drivers/s390/char/sclp_con.c:15:#include drivers/tty/serial/ifx6x60.c:28:#include drivers/tty/serial/serial-tegra.c:33:#include drivers/tty/serial/serial_mctrl_gpio.c:12:#include drivers/tty/tty_baudrate.c:8:#include include/linux/serdev.h:18:#include include/linux/tty.h:7:#include include/linux/tty_driver.h:246:#include Hmm... serial_core.h pulls tty.h, which pulls tty_driver.h. And excluding those the pull either, we are left with arch/arm/mach-ep93xx/core.c:29:#include arch/arm/mach-integrator/integrator_ap.c:29:#include drivers/s390/char/sclp_con.c:15:#include include/linux/serdev.h:18:#include sclp_con.c pulls tty_driver.h via sclp_tty.h, so the minimal solution would be to shift the include of linux/uaccess.h into tty_driver.h, that pair of arm files and serdev.h ;-/ Odds are, that's a serious overkill, but... Sigh. I really, really hate chain includes ;-/