On Thu, Oct 22, 2015 at 08:37:19AM +0200, Peter Lieven wrote:
> Am 22.09.2015 um 08:13 schrieb Peter Lieven:
> >Am 25.06.2015 um 15:18 schrieb Stefan Hajnoczi:
> >>On Tue, Jun 23, 2015 at 10:12:15AM +0200, Peter Lieven wrote:
> >>>upcoming libnfs versions will support logging debug messages. Add
> >>>support for it in qemu through an URL parameter.
> >>>
> >>>Signed-off-by: Peter Lieven <p...@kamp.de>
> >>>---
> >>> block/nfs.c | 4 ++++
> >>> 1 file changed, 4 insertions(+)
> >>>
> >>>diff --git a/block/nfs.c b/block/nfs.c
> >>>index ca9e24e..f7388a3 100644
> >>>--- a/block/nfs.c
> >>>+++ b/block/nfs.c
> >>>@@ -329,6 +329,10 @@ static int64_t nfs_client_open(NFSClient *client,
> >>>const char *filename,
> >>> } else if (!strcmp(qp->p[i].name, "readahead")) {
> >>> nfs_set_readahead(client->context, val);
> >>> #endif
> >>>+#ifdef LIBNFS_FEATURE_DEBUG
> >>>+ } else if (!strcmp(qp->p[i].name, "debug")) {
> >>>+ nfs_set_debug(client->context, val);
> >>>+#endif
> >>> } else {
> >>> error_setg(errp, "Unknown NFS parameter name: %s",
> >>> qp->p[i].name);
> >>Untrusted users may be able to set these options since they are encoded
> >>in the URI. I'm imagining a hosting or cloud scenario like OpenStack.
> >>
> >>A verbose debug level spams stderr and could consume a lot of disk
> >>space.
> >>
> >>(The uid and gid options are probably okay since the NFS server cannot
> >>trust the uid/gid coming from QEMU anyway.)
> >>
> >>I think we can merge this patch for QEMU 2.4 but I'd like to have a
> >>discussion about the security risk of encoding libnfs options in the
> >>URI.
> >>
> >>CCed Eric Blake in case libvirt is affected.
> >>
> >>Has anyone thought about this and what are the rules?
> >
> >As I hadn't time to work further on the best way to add options for NFS (and
> >other
> >protocols), would it be feasible to allow passing debug as an URL parameter,
> >but
> >limit the maximum debug level to limit a possible security impact (flooding
> >logs)?
> >
> >If a higher debug level is needed it can be set via device specific options
> >as soon
> >there is a common scheme for them.
>
> Any objections?
If you are sure that ERROR and WARN levels (or similar) don't flood the
logs, then it sounds like a solution.
Stefan
