On 02/07/2014 08:33 AM, Daniel P. Berrange wrote:
> Rewrite lxcDomainAttachDeviceHostdevStorageLive function
> to use the virProcessRunInMountNamespace helper. This avoids
> risk of a malicious guest replacing /dev with a absolute
> symlink, tricking the driver into changing the host OS
> filesystem.
>
> Signed-off-by: Daniel P. Berrange <berrange@xxxxxxxxxx>
> ---
> src/lxc/lxc_driver.c | 64 ++++++++++++++--------------------------------------
> 1 file changed, 17 insertions(+), 47 deletions(-)
>
> + if (virCgroupAllowDevicePath(priv->cgroup,
> + def->source.caps.u.storage.block,
> + VIR_CGROUP_DEVICE_RWM) != 0) {
> virReportError(VIR_ERR_INTERNAL_ERROR,
> _("cannot allow device %s for domain %s"),
> def->source.caps.u.storage.block, vm->def->name);
Here, I think we're okay setting up cgroup by name - because it is the
host-visible name, not the guest-visible name, and cgroup just stat()s
the host name to use the major:minor anyways. But since we already have
sb.st_rdev, why not avoid a second stat() and just tell cgroup to change
by major:minor in the first place?
> goto cleanup;
> }
>
> + if (lxcDomainAttachDeviceMknod(driver,
> + 0700 | S_IFBLK,
> + sb.st_rdev,
> + vm,
> + dev,
> + def->source.caps.u.storage.block) < 0) {
> + if (virCgroupDenyDevicePath(priv->cgroup,
> + def->source.caps.u.storage.block,
> + VIR_CGROUP_DEVICE_RWM) != 0)
> + VIR_WARN("cannot deny device %s for domain %s",
> + def->source.caps.u.storage.block, vm->def->name);
> + goto cleanup;
Same problem as in 10/14 about no audit of cleanup attempts. And again,
something probably worth fixing in its own patch, so for this one:
ACK.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
Attachment:
signature.asc
Description: OpenPGP digital signature
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list