Hello. I have installed fresh hardened system on partitionable software raid1 array. A can't to make system to check root filesystem at boot time. All my filesystems except /boot located inside of one partitionable raid array. /dev/md_d0p1 on / type ext3 (rw) /dev/md_d0p2 on /var type ext3 (rw) /dev/md_d0p5 on /media/local/media0 type ext3 (rw) /dev/sda1 on /boot type ext2 (rw,noatime) When I make "touch /forcefsck" and reboot the system. It check all filesystem except "/" :( Here is array config # grep -v '^#' /etc/mdadm.conf DEVICE /dev/sd[abcdef]5 ARRAY /dev/md_d0 level=raid1 num-devices=2 auto=mdp5 UUID=de49148c:b31b6b3e:3c4242ab:31fccc5a MAILADDR root # cat /proc/mdstat Personalities : [raid0] [raid1] [raid10] md_d0 : active raid1 sdc5[1] sda5[0] 488351744 blocks [2/2] [UU] unused devices: <none> Related software(as I think) sys-apps/baselayout-1.12.11.1 sys-fs/mdadm-2.6.4-r1 Reproducible: Always
please include the non-comment lines of your fstab?
# grep -v '^#' /etc/fstab LABEL=lina-/ / ext3 defaults 0 1 LABEL=lina-/boot /boot ext2 noauto,noatime 1 2 LABEL=lina-/var /var ext3 defaults 1 2 LABEL=lina-swp0 none swap sw 0 0 LABEL=lina-media0 /media/local/media0 ext3 defaults 0 2 /media/bind/home /home ext3 bind 0 0 /media/bind/srv /srv ext3 bind 0 0 mahoro:/usr/share/portage/distfiles /usr/share/portage/distfiles nfs rw,sync 0 0 proc /proc proc defaults 0 0 shm /dev/shm tmpfs nodev,nosuid,noexec 0 0
1. When you boot, after touching /forcefsck, or having forcefsck in your kernel parameters, do you see this message at all: "Checking root filesystem (full fsck forced)" 2. Do you see any other messages about the root filesystem during the boot runlevel? 3. Are you using an initrd at all? (genkernel or others) 4. When checkroot runs, does /dev/md_d0p1 exist?
(In reply to comment #3) > 1. When you boot, after touching /forcefsck, or having forcefsck in your kernel > parameters, do you see this message at all: > "Checking root filesystem (full fsck forced)" No, I don't see this message. I see following messages: - "Remounting root filesystem read/write" - Here is some module loaded, not remember wich one. :( - "Starting up RAID devices" - "Creating RAID device partitions" - "Checking all filesystems" - "A full fsck has been forced" After this going check of all filesystems, except root. > 2. Do you see any other messages about the root filesystem during the boot > runlevel? I see only message about remounting roorfs in rw mode. > 3. Are you using an initrd at all? (genkernel or others) Yes. I have used genkernel's initrd. But I have add overlay to it. # ls -lR ~/initrd-overlay /root/initrd-overlay: итого 12 drwxr-xr-x 2 root root 4096 Мар 3 05:50 bin drwxr-xr-x 2 root root 4096 Мар 2 22:40 etc drwxr-xr-x 2 root root 4096 Мар 23 13:32 sbin /root/initrd-overlay/bin: итого 0 lrwxrwxrwx 1 root root 13 Мар 3 05:50 mdadm -> ../sbin/mdadm /root/initrd-overlay/etc: итого 4 -rw-r--r-- 1 root root 2273 Мар 23 11:28 mdadm.conf /root/initrd-overlay/sbin: итого 728 -rwxr-xr-x 1 root root 191 Мар 23 13:32 mdadm -rwxr-xr-x 1 root root 733788 Мар 3 00:07 _mdadm _mdadm - it is a static linked mdadm. # cat ~/initrd-overlay/sbin/mdadm #!/bin/sh cfg='/etc/mdadm.conf' mdadm=/sbin/_mdadm IFS=' ' md_all=$(grep '^ARRAY' $cfg | sed -e 's/^ARRAY //' -e 's/[[:space:]].*$//') for md in $md_all ; do $mdadm --assemble $md done Original mdadm compiled into busybox does not create devices for partitions in raid array, it create only "main" array device. Or perhaps I don't found way how to make it do them. So I need to use such overlay and wrapper. > 4. When checkroot runs, does /dev/md_d0p1 exist? I think - no. I will check it, when I come home... in 8 hour. And what should I do, if it does not exist? Copy them from initrd's /dev?
the root filesystem is checked before all of the messages you just posted it should say: Checking root filesystem ... Remounting root filesystem read/write ...
(In reply to comment #5) > the root filesystem is checked before all of the messages you just posted > > it should say: > Checking root filesystem ... > Remounting root filesystem read/write ... > I knew it. I have read /etc/init.d/checkroot /lib/rcscripts/addons/raid-start etc... I will add something like "ls -l /dev/md*" into /etc/init.d/checkroot to see if there are required device files or they are missing. But what to do if they are missing?
it doesnt really matter if the device node exists ... you should have gotten some sort of output put like 'set +x' at the beginning of the start() func at then put 'set -x' before the 'remounting root filesystem read/write' part (so line ~70)
Created attachment 147153 [details] boot-log Here I made log you have asked. Is it show something useful?
yes, that is exactly what i wanted, thanks it would appear your root filesystem is being flagged as being unionfs-ized try doing this after booting: unionctl / --list echo $? and posting the output i imagine if you did `emerge -C unionfs` and rebooted, it'd check your root filesystem just fine ...
# unionctl / --list; echo $? 0 I am using unionfs... but i can unmerge unionfs-utils, unionctl belong to it, this is outdated and not using now utilites. And check is it help.
You are right. after unmerging unionfs-utils it begin work correct. Thanks.
but you arent using unionfs on your / are you ? ive never used unionfs at all myself, so i cant speak to the logistics ... all the code in baselayout was donated by someone else
No. I don't use unionfs on rootfs. I use it for user's "sandboxes". In official portage tree there is no unionfs-utils... it was my old copy in my local overlay. So this bug can be closed... Maybe should remove unionfs-utils related code from baselayout. Newer unionfs managets via "mount -o remount,... interface". I think that unionctl will not be used any more.
