Re: [RFC PATCH 3/4] Drivers: hv: vmbus: Introduce vmbus_sendpacket_getid()

From: Andrea Parri <parri.andrea@gmail.com>
To: "Michael Kelley (LINUX)" <mikelley@microsoft.com>
Cc: KY Srinivasan <kys@microsoft.com>,
	Haiyang Zhang <haiyangz@microsoft.com>,
	Stephen Hemminger <sthemmin@microsoft.com>,
	Wei Liu <wei.liu@kernel.org>, Dexuan Cui <decui@microsoft.com>,
	Wei Hu <weh@microsoft.com>,
	Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
	Rob Herring <robh@kernel.org>,
	Krzysztof Wilczynski <kw@linux.com>,
	Bjorn Helgaas <bhelgaas@google.com>,
	"linux-hyperv@vger.kernel.org" <linux-hyperv@vger.kernel.org>,
	"linux-pci@vger.kernel.org" <linux-pci@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH 3/4] Drivers: hv: vmbus: Introduce vmbus_sendpacket_getid()
Date: Fri, 1 Apr 2022 18:09:22 +0200	[thread overview]
Message-ID: <20220401160922.GB437893@anparri> (raw)
In-Reply-To: <PH0PR21MB3025F80C3F90284900C3D128D7E19@PH0PR21MB3025.namprd21.prod.outlook.com>

> > @@ -354,6 +354,8 @@ int hv_ringbuffer_write(struct vmbus_channel *channel,
> >  	}
> >  	desc = hv_get_ring_buffer(outring_info) + old_write;
> >  	desc->trans_id = (rqst_id == VMBUS_NO_RQSTOR) ? requestid : rqst_id;
> > +	if (trans_id)
> > +		*trans_id = desc->trans_id;
> 
> This line should *not* read the trans_id out of the ring buffer, since that
> memory is shared with the Hyper-V host and subject to being maliciously
> changed by the host.  Need to set *trans_id only from local variables, and
> somehow ensure the compiler doesn't generate code that reads the value
> from the ring buffer.  Maybe mark the desc->trans_id field as volatile, or cast
> it as such?  Or does WRITE_ONCE() work when setting it?

I'd stick to WRITE_ONCE() (with a comment).

Good catch!

Thanks,
  Andrea