[Qemu-devel] [Bug 1077838] Re: qemu-nbd -r -c taints device for subsequent usage, even after -d

[Launchpad Bug Tracker]

This bug was fixed in the package qemu-kvm - 1.0+noroms-0ubuntu14.6

---------------
qemu-kvm (1.0+noroms-0ubuntu14.6) precise-proposed; urgency=low

  * Fix qemu-kvm.upstart: just don't run in a container.  Otherwise we'll
    still try to load/unload kernel modules.  Also undo the || true after
    sysfs writes.  Since setting those is a part of configuring qemu-kvm
    on the host, failing when they fail makes sense.

qemu-kvm (1.0+noroms-0ubuntu14.5) precise-proposed; urgency=low

  * add udev to qemu-kvm Depends to ensure that postinst succeeds.
    (LP: #1080912)

qemu-kvm (1.0+noroms-0ubuntu14.4) precise-proposed; urgency=low

  [ Serge Hallyn ]
  * debian/qemu-kvm.postinst: use udevadm trigger to change /dev/kvm perms as
    recommended by Steve Langasek (LP: #1057024)
  * apply debian/patches/nbd-fixes-to-read-only-handling.patch from upstream to
    make read-write mount after read-only mount work.  (LP: #1077838)

  [ Robert Collins ]
  * Fix upstart job to succeed if ksm settings can't be altered in the same way
    other settings are handled. (LP: #1078530)
 -- Serge Hallyn <address@hidden>   Thu, 20 Dec 2012 12:34:52 -0600

-- 
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/1077838

Title:
  qemu-nbd -r -c taints device for subsequent usage, even after -d

Status in QEMU:
  In Progress
Status in “qemu-kvm” package in Ubuntu:
  Fix Released
Status in “qemu-kvm” source package in Precise:
  Fix Released
Status in “qemu-kvm” source package in Quantal:
  Fix Released

Bug description:
  Something about qemu-nbd -r -c /dev/nbd0 someimg leaves cruft behind -
  subsequent connections get marked readonly.

  This is on quantal, haven't checked precise or raring.

  To demonstrate:
  # use one image
  qemu-img create -f qcow2 /tmp/1.qcow2 100M
  sudo qemu-nbd -c /dev/nbd2 /tmp/1.qcow2
  sudo mkfs -t ext4 /dev/nbd2
  sudo qemu-nbd -d /dev/nbd2
  # use a second one on the same nbd device, shows that reuse works:
  qemu-img create -f qcow2 /tmp/2.qcow2 100M
  sudo qemu-nbd -c /dev/nbd2 /tmp/2.qcow2
  sudo mkfs -t ext4 /dev/nbd2
  sudo qemu-nbd -d /dev/nbd2
  # connect an image in read only mode
  sudo qemu-nbd -r -c /dev/nbd2 /tmp/2.qcow2
  sudo dumpe2fs /dev/nbd2 | head -n 3
  sudo qemu-nbd -d /dev/nbd2
  # now try to reuse in read-write mode again:
  qemu-img create -f qcow2 /tmp/3.qcow2 100M
  sudo qemu-nbd -c /dev/nbd2 /tmp/3.qcow2
  sudo mkfs -t ext4 /dev/nbd2
  # here it goes boom:
  mke2fs 1.42.5 (29-Jul-2012)
  /dev/nbd2: Operation not permitted while setting up superblock
  # still need to cleanup
  sudo qemu-nbd -d /dev/nbd2

  ===============================
  SRU Justification:
  1. Impact: mounting an nbd device as read-write after doing so read-only
  will cause the mount to erroneously (and quietly) be read-only.
  2. Development fix: have qemu-nbd set the device to read-write when asked,
  rather than only setting read-only.
  3. Stable fix: same as development fix
  4. Test case: See above
  5. Regression potential: The patch is localized to the handling of read-only
  flag in qemu-nbd, so any regression should not affect anything else.
  ===============================

To manage notifications about this bug go to:
https://bugs.launchpad.net/qemu/+bug/1077838/+subscriptions