[arch-general] / mounted ro after update

[David C. Rankin]

On 09/18/2012 03:53 AM, P .NIKOLIC wrote:
> On Tue, 18 Sep 2012 03:47:26 -0400 (EDT)
> Jude DaShiell<jdashiel at shellworld.net>  wrote:
>
>> >Whenever does an install of archlinux, they also do a big update so
>> >it's safe to say I got nailed by this problem too.  I'm not going to
>> >dismiss out of hand the probability that util-linux is at fault, but
>> >when I tried the installs this past weekend I suspected mkinitcpio or
>> >perhaps syslinux-install_update might be at fault.  However if in
>> >this update process neither of those utilities were used, then both
>> >of them are cleared.  It seems when util-linux finishes running after
>> >install or update it fails to set the sticky bits on partitions and
>> >lesser components in the linux file system at least in ext4 which is
>> >what I used to try the installs this past weekend in line with the
>> >installation guide on the archlinux wiki.
> HUmmmm you got me wondering now that could well be  both partitions
> that have the problem are ext4  why i did not change them to my more
> normal XFS i dont know ..
>
> I may have to back a lot up and rebuild but this time i will let my
> normal hate of the entire EXT file system rule and go XFS never been
> let down there ..
>
> Pete .
>
>

Pete,

   I have run Arch on several filesystems and I've been lucky I guess. 
Currently on this box, I have ext3, ext4 and reiser (old SuSE 10.0 partition). 
This box has been running since mid-2009 and updates are usually weekly 
(sometimes I go a couple of weeks if I can't risk a break due to workload) I 
have not had any of the mount ro weirdness even after several multi-gigabyte 
updates. The current partitions I have are:

/dev/sdc5 on / type ext3 (rw,relatime,data=ordered)
/dev/sdc7 on /home type ext4 (rw,relatime,data=ordered)
/dev/sda2 on /mnt/pv type reiserfs (rw,relatime)
/dev/sdb2 on /mnt/win type fuseblk 
(rw,nosuid,nodev,noexec,relatime,user_id=0,group_id=0,default_permissions,allow_other,blksize=4096)

   I don't know what is doing it in your case, but it seems like we should be 
able to figure out where mount ro/rw logic for the resides (I picture 
something like the following buried somewhere):

   if [conditional]; then
     mount -o rw [whatever]
   else
     mount -o ro [whatever]
   fi

   I suspect this may be complicated by the fact that mounting (or remounting) 
takes place in several different places/processes during the boot. Anybody 
familiar with this off-hand or any idea where Pete might look to rule-in or 
rule-out the different parts of boot that could effect this? Sorry I don't 
have more, I just haven't had the need to dissect the boot mount process to 
that level before...

   I guess you are just lucky :)

-- 
David C. Rankin, J.D.,P.E.