ext3 formatted usb mounted but not visible in journal

[ghunt]

build 711. I have 4 KINGSTON usb 1 gig sticks. All ext3's mount, but do not become visible in journal. Those which mount as vfat do become visible.

[Rmyers]

I noticed similar behavior when adding an ext2 formatted sd card a while back. With the sd I didn't consider it a bug, as an ext formatted sd could be considered an extension of the file system, rather than a pass around stick with journal entries on it.

Just noting that this might be considered proper behavior for a sd card. If it is decided that _any_ usb should show in the journal -- and I think that this can be argued as being sensible, thought should be given as to whether a non-fat sd should show as journal or as file system.

[mstone]

Jim, Tomeu, Rmyers: any thoughts on who should own this?

[tomeu]

Most probably an issue in the Journal or the DataStore.

Can we get logs as per http://wiki.laptop.org/go/Attaching_Sugar_Logs_to_Tickets ?

Thanks!

[paulwaite]

Logs showing non-mount of ext3 usb stick

[paulwaite]

Found that an ext3-formatted USB stick mounts fine on the machine onto /media, however in the Journal log there are permission errors, because the newly mounted filesystem is owned by root. Logs attached.

A vfat-formatted USB stick mounts fine and is read by the Journal.

[martin.langhoff]

Note that this would be trivial to fix just by setting the mount options to "uid=olpc".

[marco]

The problem is that hal doesn't allow us to pass an uid:

volume.mount.valid_options = {'ro', 'sync', 'dirsync', 'noatime', 'nodiratime', 'noexec', 'quiet', 'remount', 'exec', 'acl', 'user_xattr', 'data='} (string list)

This looks like a Fedora 9 bug, I get exactly the same permission issue on my stock GNOME/Fedora 9. So assigning it back to distro.

[martin.langhoff]

From the distro POV, what it's doing is correct.

Sugar wants something special however - that the disk gets mounted in a way that allows us to write to the root. Perhaps that's not a reasonable thing to do...?

From that POV, perhaps Sugar should see if it can write to the root of the disk, and if it cannot, do not write the hidden index file there (not write it at all, or place it /tmp )

[marco]

Oh, you are right. Ok, I think this will have to be fixed by the datastore rework we are planning for 9.1 then. The current implementation depends strongly on the existence of the index afaik (but Tomeu please chime in if I'm wrong).