Re: [dpdk-dev] [PATCH] virtio: fix packet corruption

["Tan, Jianfeng" <jianfeng.tan@intel.com>]

> -----Original Message-----
> From: Olivier Matz [mailto:olivier.matz@6wind.com]
> Sent: Tuesday, July 19, 2016 9:11 PM
> To: Tan, Jianfeng; dev@dpdk.org; Xie, Huawei; yuanhan.liu@linux.intel.com
> Subject: Re: [PATCH] virtio: fix packet corruption
> 
> Hi Jianfeng,
> 
> On 07/19/2016 03:03 PM, Tan, Jianfeng wrote:
> > Hi Oliver,
> >
> >> -----Original Message-----
> >> From: Olivier Matz [mailto:olivier.matz@6wind.com]
> >> Sent: Tuesday, July 19, 2016 8:32 PM
> >> To: dev@dpdk.org; Tan, Jianfeng; Xie, Huawei;
> yuanhan.liu@linux.intel.com
> >> Subject: [PATCH] virtio: fix packet corruption
> >>
> >> The support of virtio-user changed the way the mbuf dma address is
> >> retrieved, using a physical address in case of virtio-pci and a virtual
> >> address in case of virtio-user.
> >>
> >> This change introduced some possible memory corruption in packets,
> >> replacing:
> >>   m->buf_physaddr + RTE_PKTMBUF_HEADROOM
> >> by:
> >>   m->buf_physaddr + m->data_off     (through a macro)
> >>
> >> This patch fixes this issue, restoring the original behavior.
> >
> > Could you be more specific on why we cannot use m->data_off here?
> 
> There is no guarantee that m->data_off == RTE_PKTMBUF_HEADROOM here
> as
> virtqueue_enqueue_recv_refill() is called on a mbuf that is just
> allocated with rte_mbuf_raw_alloc(). An alternative would be to set
> data_off to RTE_PKTMBUF_HEADROOM, but as it's a fix and we are close to
> the release, I prefered to restore the initial behavior.

Oh yes, gotcha.

But if we do not set data_off properly, it's still buggy when others consume these mbufs, right?

Thanks,
Jianfeng

> 
> I did not include the test plan because it relies on patch that are not
> submitted yet (offload patches, they will be upstreamed very soon). It
> is a quite simple test case with testpmd.
> 
> Regards,
> Olivier