Re: [PATCH 2/7] ipc: fix trivial warning

[Jiri Kosina]

On Mon, 19 Oct 2009, Felipe Contreras wrote:

> >> >> ipc/msg.c: In function ?msgctl_down?:
> >> >> ipc/msg.c:415: warning: ?msqid64? may be used uninitialized in this function
> >> >>
> >> >> Signed-off-by: Felipe Contreras <felipe.contreras@xxxxxxxxx>
> >> >> ---
> >> >> Âipc/msg.c | Â Â2 +-
> >> >> Â1 files changed, 1 insertions(+), 1 deletions(-)
> >> >>
> >> >> diff --git a/ipc/msg.c b/ipc/msg.c
> >> >> index 2ceab7f..085bd58 100644
> >> >> --- a/ipc/msg.c
> >> >> +++ b/ipc/msg.c
> >> >> @@ -412,7 +412,7 @@ static int msgctl_down(struct ipc_namespace *ns, int msqid, int cmd,
> >> >> Â Â Â Â Â Â Â Â Â Â Âstruct msqid_ds __user *buf, int version)
> >> >> Â{
> >> >> Â Â Â struct kern_ipc_perm *ipcp;
> >> >> - Â Â struct msqid64_ds msqid64;
> >> >> + Â Â struct msqid64_ds uninitialized_var(msqid64);
> >> >> Â Â Â struct msg_queue *msq;
> >> >> Â Â Â int err;
[ ... snip ... ]
> > I have verified both with 4.1 and 4.3, and it doesn't emit this
> > false-positive warning, so there have been gcc versions getting this
> > right. Ergo gcc developers should rather fix this "regression" and revert
> > to the old behavior, methinks.
> 
> The other possibility is that the bug was in gcc 4.1/4.3, and now 4.4
> finds an actual problem in the code. I will try to dig deeper to see
> what's actually happening... at first glance I don't see who is
> initializing msqid64.

This statement of your makes me wonder why you have submitted the patch in 
the first place, as you are apparently not sure whether adding 
uninitialized_var() is a valid thing to do or not.

The gcc warning in this case is actually bogus, as msqid64 is touched only 
iff cmd == IPC_SET, and in such case, copy_msqid_from_user() initializes 
it properly.

-- 
Jiri Kosina
SUSE Labs, Novell Inc.