[arch-releng] [PATCH] Mount tmpfs filesystems with mode=0755

[Thomas Bächler]

Am 09.07.2010 17:11, schrieb Gerardo Exequiel Pozzi:
>> I don't think this is the case here. You are not mounting an aufs branch
>> inside another one, but you are moving an AUFS branch into the AUFS
>> mount point itself. To me, it seems like AUFS and the Linux VFS should
>> be able to handle that. I haven't tested it though.
>>
>> What the bug report stated is different: You nested two branches into
>> each other - if AUFS uses the whole directory as a branch instead of
>> just the mount point, I can understand why that would fail.
>>
> Seems to work moving all to /new_root/somedir... I have a dude why are
> interested in doing that?

It is simply cleaner and more transparent. Having unaccessible mounts
below the visible root hides part of the system from the end user, which
is something I never like to do.

> (nothing left except /new_root , hehe)
> 
> https://gist.github.com/468957/7284f470b968064a8342060e64d3784f6658a42c
> 
> ------------------------------------------------------------
> (copytoram=n)
> tmpfs on /new_root/aufs/rw_branch type tmpfs
> (rw,relatime,size=383696k,mode=755)
> none on /new_root type aufs (rw,relatime,si=ddcbbff9)
> /dev/disk/by-label/ARCH_201007 on /new_root/bootmnt type udf
> (ro,relatime,utf8)
> /dev/loop0 on /new_root/aufs/ro_branch/overlay type squashfs (ro,relatime)
> /dev/loop1 on /new_root/aufs/ro_branch/root-image type squashfs
> (ro,relatime)
> 
> (on shutdown)
> umount: /dev/loop1 busy - remounted read-only
> umount: /dev/loop0 busy - remounted read-only
> umount: /dev/disk/by-label/ARCH_201007 busy - remounted read-only
> umount: /aufs/rw_branch: device is busy.
> 
> ------------------------------------------------------------
> (copytoram=y)
> tmpfs on /new_root/aufs/rw_branch type tmpfs
> (rw,relatime,size=383696k,mode=755)
> none on /new_root type aufs (rw,relatime,si=88cacedb)
> tmpfs on /new_root/copyonram type tmpfs (rw,relatime,size=383696k,mode=755)
> /dev/loop0 on /new_root/aufs/ro_branch/overlay type squashfs (ro,relatime)
> /dev/loop1 on /new_root/aufs/ro_branch/root-image type squashfs
> (ro,relatime)
> 
> (on shutdown)
> umount: /dev/loop1 busy - remounted read-only
> umount: /dev/loop0 busy - remounted read-only
> umount: /copyonram: device is busy.
> umount: /aufs/rw_branch: device is busy.
> ------------------------------------------------------------

Just as I suspected: It works fine and IMO looks very nice (you can
investigate /bootmnt/ and /aufs/r?_branch if you want to see what's
going on).

Two problems:
1) umounting fails ... I need to work on the umount code in rc.shutdown
anyway, so there may be a solution for that.
2) /aufs/rw_branch/ is writable! Writing to it will definitely cause
trouble with aufs - which would be a reason not to do this.

Thanks for your tests, I am not sure anymore whether this is a good idea
though.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 262 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.archlinux.org/pipermail/arch-releng/attachments/20100709/596a25c8/attachment-0001.bin>