2.6.8-rc2-bk13 and later: Read-only filesystem on USB

When I try to copy a file to my DataBar <www.data-bar.com> mounted at /mnt/cf under 2.6.8-rc-2-bk13 or 2.6.8-rc3, I get the error "cp: cannon stat `/mnt/cf/testfile': Permission denied". Same thing with a CF card in an USB adapter.

It worked under 2.6.8-rc2-bk12 and earlier.

Hilsen Harald.

Harald> When I try to copy a file to my DataBar <www.data-bar.com> Harald> mounted at /mnt/cf under 2.6.8-rc-2-bk13 or 2.6.8-rc3, I get Harald> the error "cp: cannon stat `/mnt/cf/testfile': Permission Harald> denied". Same thing with a CF card in an USB adapter.

There is a change in the FAT or VFAT filesystem. If you don't specify a default codepage in the mount command, the filesystem gets mounted
RO. I'm annoyed by this change since the error isn't propogated
properly back to user space. I ran into it this morning when trying to move images from a CF card to my disk.

Here's what I'm doing on a 2.6.8-rc3 system:

    # mount /dev/sdd1 /mnt/s30

Good, no errors.

    # mount
    /dev/sda2 on / type unknown (rw,errors=remount-ro)
    proc on /proc type proc (rw)
    sysfs on /sys type sysfs (rw)
    devpts on /dev/pts type devpts (rw,gid=5,mode=620)
    tmpfs on /dev/shm type tmpfs (rw)
    usbfs on /proc/bus/usb type usbfs (rw)
    /dev/sda5 on /var type ext3 (rw)
    /dev/sda1 on /boot type ext3 (rw)
    /dev/sda6 on /usr type ext3 (rw)
    /dev/mapper/data_vg-home_lv on /home type ext3 (rw)
    /dev/mapper/data_vg-local_lv on /local type ext3 (rw)
    /dev/sdd1 on /mnt/s30 type msdos (rw)

And it looks like it mounted properly too.

    # touch /mnt/s30/foo
    touch: cannot touch `/mnt/s30/foo': Read-only file system

BOOM! Fall down and smack my face. Ok, so now I try to mount it explicity as vfat, not the default msdos:

    # umount /mnt/s30
    # mount -t vfat /dev/sdd1 /mnt/s30
    # mount
    /dev/sda2 on / type unknown (rw,errors=remount-ro)
    proc on /proc type proc (rw)
    sysfs on /sys type sysfs (rw)
    devpts on /dev/pts type devpts (rw,gid=5,mode=620)
    tmpfs on /dev/shm type tmpfs (rw)
    usbfs on /proc/bus/usb type usbfs (rw)
    /dev/sda5 on /var type ext3 (rw)
    /dev/sda1 on /boot type ext3 (rw)
    /dev/sda6 on /usr type ext3 (rw)
    /dev/mapper/data_vg-home_lv on /home type ext3 (rw)
    /dev/mapper/data_vg-local_lv on /local type ext3 (rw)
    /dev/sdd1 on /mnt/s30 type vfat (rw)

    # touch /mnt/s30/foo
    touch: cannot touch `/mnt/s30/foo': Read-only file system

BOOM! Still broken, but still reporting as good in mount. WTF? Next attempt:

    # mount -t vfat -o codepage=437 /dev/sdd1 /mnt/s30
    # touch /mnt/s30/foo
    touch: cannot touch `/mnt/s30/foo': Read-only file system

Now WTF is happening?

    # mount
    /dev/sda2 on / type unknown (rw,errors=remount-ro)
    proc on /proc type proc (rw)
    sysfs on /sys type sysfs (rw)
    devpts on /dev/pts type devpts (rw,gid=5,mode=620)
    tmpfs on /dev/shm type tmpfs (rw)
    usbfs on /proc/bus/usb type usbfs (rw)
    /dev/sda5 on /var type ext3 (rw)
    /dev/sda1 on /boot type ext3 (rw)
    /dev/sda6 on /usr type ext3 (rw)
    /dev/mapper/data_vg-home_lv on /home type ext3 (rw)
    /dev/mapper/data_vg-local_lv on /local type ext3 (rw)
    /dev/sdd1 on /mnt/s30 type vfat (rw,codepage=437)

Stupid tools.... next attempt:

    # mount -o codepage=437 /dev/sdd1 /mnt/s30
    # mount
    /dev/sda2 on / type unknown (rw,errors=remount-ro)
    proc on /proc type proc (rw)
    sysfs on /sys type sysfs (rw)
    devpts on /dev/pts type devpts (rw,gid=5,mode=620)
    tmpfs on /dev/shm type tmpfs (rw)
    usbfs on /proc/bus/usb type usbfs (rw)
    /dev/sda5 on /var type ext3 (rw)
    /dev/sda1 on /boot type ext3 (rw)
    /dev/sda6 on /usr type ext3 (rw)
    /dev/mapper/data_vg-home_lv on /home type ext3 (rw)
    /dev/mapper/data_vg-local_lv on /local type ext3 (rw)
    /dev/sdd1 on /mnt/s30 type msdos (rw,codepage=437)
    # touch /mnt/s30/foo
    # ls -l /mnt/s30/foo
    -rwxr--r--  1 root root 0 Aug  6 10:27 /mnt/s30/foo

Finally, the damm thing is mounted RW and actually lets me write something to the goddamm thing.

What a pain in the ass. This change should be reverted until it's properly implemented to tell mount(8) that it's not mounted RW, but RO instead.

John
John Stoffel - Senior Unix Systems Administrator - Lucent Technologies

	 [email protected] - http://www.lucent.com - 978-952-7548

The kernel is already exporting the some information. cat /proc/mounts.

Yes, probably proper interface would be useful. But it's the another story.

OGAWA Hirofumi <[email protected]>