RE: [PATCH net] hvsock: fix epollout hang from race condition

[Dexuan Cui <decui@microsoft.com>]

> From: linux-hyperv-owner@vger.kernel.org
> <linux-hyperv-owner@vger.kernel.org> On Behalf Of Dexuan Cui
> Sent: Friday, June 14, 2019 8:23 PM
> To: David Miller <davem@davemloft.net>; Sunil Muthuswamy
> <sunilmut@microsoft.com>
> Cc: KY Srinivasan <kys@microsoft.com>; Haiyang Zhang
> <haiyangz@microsoft.com>; Stephen Hemminger
> <sthemmin@microsoft.com>; sashal@kernel.org; Michael Kelley
> <mikelley@microsoft.com>; netdev@vger.kernel.org;
> linux-hyperv@vger.kernel.org; linux-kernel@vger.kernel.org
> Subject: RE: [PATCH net] hvsock: fix epollout hang from race condition
> 
> > From: linux-hyperv-owner@vger.kernel.org
> > <linux-hyperv-owner@vger.kernel.org> On Behalf Of David Miller
> > Sent: Friday, June 14, 2019 7:15 PM
> > To: Sunil Muthuswamy <sunilmut@microsoft.com>
> >
> > This adds lots of new warnings:
> >
> > net/vmw_vsock/hyperv_transport.c: In function ‘hvs_probe’:
> > net/vmw_vsock/hyperv_transport.c:205:20: warning: ‘vnew’ may be used
> > uninitialized in this function [-Wmaybe-uninitialized]
> >    remote->svm_port = host_ephemeral_port++;
> >    ~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~
> > net/vmw_vsock/hyperv_transport.c:332:21: note: ‘vnew’ was declared
> here
> >   struct vsock_sock *vnew;
> >                      ^~~~
> > net/vmw_vsock/hyperv_transport.c:406:22: warning: ‘hvs_new’ may be
> > used uninitialized in this function [-Wmaybe-uninitialized]
> >    hvs_new->vm_srv_id = *if_type;
> >    ~~~~~~~~~~~~~~~~~~~^~~~~~~~~~
> > net/vmw_vsock/hyperv_transport.c:333:23: note: ‘hvs_new’ was declared
> > here
> >   struct hvsock *hvs, *hvs_new;
> >                        ^~~~~~~
> 
> Hi David,
> These warnings are not introduced by this patch from Sunil.

Well, technically speaking, the warnings are caused by Suni's patch, but I think it should
be a bug of gcc (I'm using "gcc (Ubuntu 8.2.0-12ubuntu1) 8.2.0"). As you can see, the
line numbers given by gcc, i.e. line 205, line 406, are not related to the warnings.

Actually, the same warnings can repro with the below one-line patch on this commit of
today's net.git:
    9a33629ba6b2 ("hv_netvsc: Set probe mode to sync")

--- a/net/vmw_vsock/hyperv_transport.c
+++ b/net/vmw_vsock/hyperv_transport.c
@@ -403,6 +403,7 @@ static void hvs_open_connection(struct vmbus_channel *chan)

        set_per_channel_state(chan, conn_from_host ? new : sk);
        vmbus_set_chn_rescind_callback(chan, hvs_close_connection);
+       asm ("nop");

        if (conn_from_host) {
                new->sk_state = TCP_ESTABLISHED;

It looks a simple inline assembly code can confuse gcc. I'm not sure if I should
report a bug for gcc...

I posted a patch to suppress these bogus warnings just now. The Subject is:

[PATCH net] hv_sock: Suppress bogus "may be used uninitialized" warnings

Thanks,
-- Dexuan