On Wed, Sep 19, 2007 at 06:49:48PM +0100, Daniel P. Berrange wrote:
On Wed, Sep 19, 2007 at 02:51:42PM +0100, Richard W.M. Jones wrote:
>
> Are you sure they don't go over the wire? My reading of doRemoteOpen
> suggests that they do:
>
> static int
> doRemoteOpen (virConnectPtr conn, struct private_data *priv, const char
> *uri_str, int flags)
> {
> // bunch of code which reads 'flags' but never modifies it
>
> /* Finally we can call the remote side's open function. */
> remote_open_args args = { &name, flags };
>
> if (call (conn, priv, 1, REMOTE_PROC_OPEN,
> (xdrproc_t) xdr_remote_open_args, (char *) &args,
> (xdrproc_t) xdr_void, (char *) NULL) == -1)
> goto failed;
>
> and the 'call' function just serialises whatever is in the args array.
> But at this point flags could contain VIR_DRV_OPEN_REMOTE_* flags.
>
> Unless I'm reading this code wrong ...
Yep that's a bug - we're passing bogus data to the server. Fortunately its
not checking any flag except _RO, but I'll fix this & repost.
Here's a tweaked patch which explicitly only passes through the _RO flag
to the remote_open_args.
Dan.
--
|=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=|
|=- Perl modules:
http://search.cpan.org/~danberr/ -=|
|=- Projects:
http://freshmeat.net/~danielpb/ -=|
|=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|