I recently updated to udev udev-147 and during boot, udev startup failed with: Nov 24 20:07:54 sheldon udevd[12139]: error getting socket: Invalid argument Nov 24 20:07:54 sheldon udevd[12139]: error initializing netlink socket So I updated to 147-r1 - same problem. Backdating to current stable on alpha (146-r1) makes it work again, so it's probably not a setup/config problem. This might be alpha specific since I can't reproduce the error with my amd64 workstation. If this is a kernel config problem, it's non-obvious what I did wrong.
There's very little of your info, but this can be related to the use of SOCK_CLOEXEC. So, what's your glib/kernel/kernel headers version ?
(sorry, I was sure I had included emerge --info) $ uname -a Linux sheldon 2.6.31.3 #1 Sun Oct 11 15:37:38 CEST 2009 alpha EV68AL Nautilus GNU/Linux $ eix -Iec glib [I] dev-libs/glib (2.22.2(2)@11/22/09): The GLib library of C routines $ eix -Iec linux-headers [I] sys-kernel/linux-headers (2.6.30-r1@08/09/09): Linux system headers $ emerge --info Portage 2.1.7.6 (default/linux/alpha/10.0, gcc-4.3.4, glibc-2.10.1-r1, 2.6.31.3 alpha) ================================================================= System uname: Linux-2.6.31.3-alpha-EV68AL-with-gentoo-2.0.1 Timestamp of tree: Tue, 24 Nov 2009 18:45:01 +0000 distcc 3.1 alpha-unknown-linux-gnu [enabled] app-shells/bash: 4.0_p35 dev-java/java-config: 2.1.9-r1 dev-lang/python: 2.4.6, 2.5.4-r3, 2.6.4, 2.7_pre20091122, 3.1.1-r1 dev-python/pycrypto: 2.1.0_beta1 dev-util/cmake: 2.6.4-r3 sys-apps/baselayout: 2.0.1 sys-apps/openrc: 0.5.2-r2 sys-apps/sandbox: 2.2 sys-devel/autoconf: 2.13, 2.63-r1 sys-devel/automake: 1.10.2, 1.11 sys-devel/binutils: 2.20 sys-devel/gcc-config: 1.4.1 sys-devel/libtool: 2.2.6b virtual/os-headers: 2.6.30-r1 ACCEPT_KEYWORDS="alpha ~alpha" ACCEPT_LICENSE="* -@EULA" CBUILD="alpha-unknown-linux-gnu" CFLAGS="-mieee -pipe -O2 -mcpu=ev67" CHOST="alpha-unknown-linux-gnu" CONFIG_PROTECT="/etc" CONFIG_PROTECT_MASK="/etc/ca-certificates.conf /etc/env.d /etc/env.d/java/ /etc/eselect/postgresql /etc/fonts/fonts.conf /etc/gconf /etc/gentoo-release /etc/revdep-rebuild /etc/sandbox.d /etc/terminfo /etc/udev/rules.d" CXXFLAGS="-mieee -pipe -O2 -mcpu=ev67" DISTDIR="/usr/portage/distfiles" FEATURES="assume-digests distcc distlocks fixpackages news parallel-fetch protect-owned sandbox sfperms strict unmerge-logs unmerge-orphans usepkg userfetch" GENTOO_MIRRORS="http://gentoo.tiscali.nl/ http://mirrors.sec.informatik.tu-darmstadt.de/gentoo/" LDFLAGS="-Wl,-O1" MAKEOPTS="-j2" PKGDIR="/usr/portage/packages" PORTAGE_CONFIGROOT="/" PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress --force --whole-file --delete --stats --timeout=180 --exclude=/distfiles --exclude=/local --exclude=/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage/layman/python" SYNC="rsync://rsync5.de.gentoo.org/gentoo-portage" USE="X acl alpha alsa apache2 audiofile bash-completion berkdb bmp bzip2 calendar cdparanoia cdr cli cracklib crypt dio dri encode ethereal exif ffmpeg fftw firefox flac fortran ftp gdbm gpm iconv imlib2 jpeg kdeenablefinal libcaca lua mad matroska mmap mng modules moznocompose moznoirc moznomail mozsvg mpeg mudflap ncurses network-cron nls nptl nptlonly offensive ogg pam pcre pdflib perl png pnm ppds pppd python rar readline recode reflection session sharedmem sockets sox spl ssl svg sysfs szip tcpd tetex theora truetype unicode usb v4l v4l2 vcd vidix vim vim-pager vlm vorbis xcb xorg xosd xpm xvid zlib" ALSA_CARDS="ali5451 als4000 bt87x ca0106 cmipci emu10k1 ens1370 ens1371 es1938 es1968 fm801 hda-intel intel8x0 maestro3 trident usb-audio via82xx ymfpci" ALSA_PCM_PLUGINS="adpcm alaw asym copy dmix dshare dsnoop empty extplug file hooks iec958 ioplug ladspa lfloat linear meter mmap_emul mulaw multi null plug rate route share shm softvol" APACHE2_MODULES="actions alias auth_basic authn_alias authn_anon authn_dbm authn_default authn_file authz_dbm authz_default authz_groupfile authz_host authz_owner authz_user autoindex cache dav dav_fs dav_lock deflate dir disk_cache env expires ext_filter file_cache filter headers include info log_config logio mem_cache mime mime_magic negotiation rewrite setenvif speling status unique_id userdir usertrack vhost_alias" ELIBC="glibc" INPUT_DEVICES="keyboard mouse" KERNEL="linux" LCD_DEVICES="bayrad cfontz cfontz633 glk hd44780 lb216 lcdm001 mtxorb ncurses text" USERLAND="GNU" VIDEO_CARDS="vga glint mga nvidia vesa r128 " Unset: CPPFLAGS, CTARGET, EMERGE_DEFAULT_OPTS, FFLAGS, INSTALL_MASK, LANG, LC_ALL, LINGUAS, PORTAGE_COMPRESS, PORTAGE_COMPRESS_FLAGS, PORTAGE_RSYNC_EXTRA_OPTS
I've also noticed this during build: CCLD udev/udevadm udev/udev-watch.o: In function `udev_watch_init': (.text+0x8b8): warning: warning: inotify_init1 is not implemented and will always fail udev/udev-watch.o: In function `udev_watch_init': (.text+0x8b8): warning: warning: inotify_init1 is not implemented and will always fail Making all in libudev/docs
Right, it's arch problem: there are results if you search for it in kernel tree under arch/x86, but not in arch/alpha. Should this be considered udev upstream bug ?
I'd say so: either the udev guys rely on something that isn't available in every kernel or the kernel guys implemented only half of a new feature...
Also, I'll mask >=udev-147 on alpha until this is fixed.
Not this email exchange: http://lists.debian.org/debian-devel/2009/11/msg00466.html So apparently, future version might be fixed. I'll keep an eye on it.
udev-149 is in tree, but I doubt it fixes this bug. If you have a tested patch to be applied, please attach it here.
I just tested udev-149 (no patch) and it failed exactly the same way. So we have to either wait until someone implements that syscall or do it ourselves (my kernel-fu is way to weak to do that).
If the linked debian page is right, it could work to change the inotify_init1(IN_CLOEXEC) to code like this: inotify_fd = inotify_init1(IN_CLOEXEC); if (inotify_fd < 0) { /* better condition needed, maybe errno == ENOSYS */ /* fallback to inotify_init */ inotify_fd = inotify_init(); if (inotify_fd >= 0) util_set_fd_cloexec(inotify_fd); } if (inotify_fd < 0) err(udev, "inotify_init failed: %m\n");
The inotify_init1 syscall glue is in 2.6.33. I just tried a kernel with this in it, and patched linux-headers to have the correct defines in asm/unistd.h, but no luck. * Mounting /dev ... [ ok ] * Starting udevd ...udevd[488]: error getting socket: Invalid argument error initializing netlink socket udevd[488]: error initializing netlink socket [ !! ] No /sbin/udevd found running; none killed. * Mounting devpts at /dev/pts ... [ ok ] * Checking root filesystem ...fsck.ext3: No such file or directory while trying to open /dev/sda1 /dev/sda1: The superblock could not be read or does not describe a correct ext2 filesystem. If the device is valid and it really contains an ext2 filesystem (and not swap or ufs or something else), then the superblock is corrupt, and you might try running e2fsck with an alternate superblock: e2fsck -b 8193 <device> * Filesystem couldn't be fixed :( [ !! ]
Note that you don't have to install udev to see if it's broken: The linker warns about the absence of said syscall during build time (near the end, so it's easy to find).
Same for HPPA. Didn't notice that before. elmer ~ # /etc/init.d/udev restart * WARNING: you are stopping a sysinit service * Caching service dependencies ... [ ok ] * Stopping udevd ... * start-stop-daemon: no matching processes found [ ok ] * Starting udevd ... error initializing netlink socket * start-stop-daemon: failed to start `/sbin/udevd' [ !! ] * start-stop-daemon: no matching processes found * ERROR: udev failed to start
The problem is that the SOCK_CLOEXEC values in glibc didn't match the kernel. I've fixed it for Alpha and HPPA, and confirmed that udev does run on Alpha with a fixed glibc. Relevant commits are HPPA http://sourceware.org/git/?p=glibc-ports.git;a=commit;h=fa1b9bae63e87161783dafaf3187251901216bad http://sourceware.org/git/?p=glibc-ports.git;a=commit;h=f3ed03373affbd13ee8dbbeb933959381ae492bb Alpha http://sourceware.org/git/?p=glibc-ports.git;a=commit;h=0b37400ae41a175fa7f89bdf5411ff5d297e5ae4 http://sourceware.org/git/?p=glibc-ports.git;a=commit;h=b6c131119cff4a2a11f1766132f306464d9457da SOCK_NONBLOCK was also broken on Alpha. I don't remember if it was right on HPPA.
Created attachment 227781 [details, diff] 1. patch to create bits/socket.h on alpha
Created attachment 227783 [details, diff] 2. patch to add missing pieces to bits/socket.h on alpha
Created attachment 227785 [details, diff] 3. patch to fix SOCK_CLOEXEC on hppa
OK, so once glibc-2.12 (it contains these patches) is in the tree, we can have >=udev-147 depend on it, and we can remove the mask.
Ugh, I thought I'd reassigned this to toolchain a long time ago and that you guys were just ignoring us. Turns out it was my fault. :)
i thought we already added the fix for alpha hppa doesnt even build with glibc-2.11, so no regression there
Looks like you're right. I'm entirely out of it it seems. So >=udev-147 can be unmasked on alpha, and depend on >=glibc-2.11.1. Going to remove myself and alpha@ from CC before I do some damage. :)
added 1508_all_glibc-2.11-hppa-SOCK_CLOEXEC.patch to the glibc-2.11.2 patchset; not that it matters too much till Bug 301642 is fixed ...
Ugh. 6028_all_alpha-fix-SOCK_NONBLOCK.patch is wrong. I'm not sure what happened here, but it's got the wrong SOCK_CLOEXEC definition. Please kill it and add the first two patches I've attached to replace it.
this bug is for hppa only. we already have an alpha-specific bug 307651, so there is no value in re-opening multiple bugs.
