On Thu, Apr 27, 2017 at 17:46:16 +0200, Peter Krempa wrote:
On Thu, Apr 27, 2017 at 16:30:44 +0100, Daniel Berrange wrote:
> On Wed, Apr 26, 2017 at 07:52:44PM +0200, Peter Krempa wrote:
> > Format the string into the "curl" format so that it's accepted by
qemu.
> >
> > Partially resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=1140164
[snip]
> Your example cookie is rather tame, but I wonder if we should
> consider cookie values to be security sensitive data, and thus
> use the secrets mechanism. If we did this would also entail fixes
> to QEMU to let use its secrets mechanism too.
I thought briefly about the same before posting this, but I went through
anyways ...
>
> I'm just wary of re-introducing a bug like CVE-2015-5160 (rbd
> password information leak), via sensitive cookie values.
We could allow generic cookies passed on the command line
and then perhaps add a <cookie name="ble"
secure='yes'>value</cookie>
which will be passed via the secrets infrastructure.
Or better, treat them as secure by default and make users add
secure='no' explicitly so that they are aware of the way we pass it.