On Tue, Nov 08, 2022 at 07:56:18AM +0000, Xia, Chenbo wrote:
> > -----Original Message-----
> > From: Pei, Andy <andy....@intel.com>
> > Sent: Tuesday, November 8, 2022 3:39 PM
> > To: Xia, Chenbo <chenbo....@intel.com>; Taekyung Kim
> > <kim.tae.ky...@navercorp.com>; dev@dpdk.org
> > Cc: sta...@dpdk.org; maxime.coque...@redhat.com; Wang, Xiao W
> > <xiao.w.w...@intel.com>
> > Subject: RE: [PATCH v3] vdpa/ifc: fix update_datapath error handling
> >
> > Hi
> >
> > See my reply inline.
> >
> > > -----Original Message-----
> > > From: Xia, Chenbo <chenbo....@intel.com>
> > > Sent: Tuesday, November 8, 2022 9:47 AM
> > > To: Taekyung Kim <kim.tae.ky...@navercorp.com>; dev@dpdk.org
> > > Cc: sta...@dpdk.org; maxime.coque...@redhat.com; Wang, Xiao W
> > > <xiao.w.w...@intel.com>
> > > Subject: RE: [PATCH v3] vdpa/ifc: fix update_datapath error handling
> > >
> > > > -----Original Message-----
> > > > From: Taekyung Kim <kim.tae.ky...@navercorp.com>
> > > > Sent: Monday, November 7, 2022 5:00 PM
> > > > To: dev@dpdk.org
> > > > Cc: sta...@dpdk.org; maxime.coque...@redhat.com; Xia, Chenbo
> > > > <chenbo....@intel.com>; Wang, Xiao W <xiao.w.w...@intel.com>;
> > > > kim.tae.ky...@navercorp.com
> > > > Subject: [PATCH v3] vdpa/ifc: fix update_datapath error handling
> > > >
> > > > Stop and return the error code when update_datapath fails.
> > > > update_datapath prepares resources for the vdpa device.
> > > > The driver should not perform any further actions if update_datapath
> > > > returns an error.
> > > >
> > > > Fixes: a3f8150eac6d ("net/ifcvf: add ifcvf vDPA driver")
> > > > Cc: sta...@dpdk.org
> > > >
> > > > Signed-off-by: Taekyung Kim <kim.tae.ky...@navercorp.com>
> > > > ---
> > > > v3:
> > > > * Fix coding style
> > > >
> > > > v2:
> > > > * Revert the prepared resources before returning an error
> > > > * Rebase to 22.11 rc2
> > > > * Add fixes and cc for backport
> > > >
> > > > ---
> > > > drivers/vdpa/ifc/ifcvf_vdpa.c | 26 ++++++++++++++++++++++----
> > > > 1 file changed, 22 insertions(+), 4 deletions(-)
> > > >
> > > > diff --git a/drivers/vdpa/ifc/ifcvf_vdpa.c
> > > > b/drivers/vdpa/ifc/ifcvf_vdpa.c index 8dfd49336e..0396d49122 100644
> > > > --- a/drivers/vdpa/ifc/ifcvf_vdpa.c
> > > > +++ b/drivers/vdpa/ifc/ifcvf_vdpa.c
> > > > @@ -1098,7 +1098,12 @@ ifcvf_dev_config(int vid)
> > > > internal = list->internal;
> > > > internal->vid = vid;
> > > > rte_atomic32_set(&internal->dev_attached, 1);
> > > > - update_datapath(internal);
> > > > + if (update_datapath(internal) < 0) {
> > > > + DRV_LOG(ERR, "failed to update datapath for vDPA device
> > > > %s",
> > > > + vdev->device->name);
> > > > + rte_atomic32_set(&internal->dev_attached, 0);
> > > > + return -1;
> > > > + }
> > > >
> > > > hw = &internal->hw;
> > > > for (i = 0; i < hw->nr_vring; i++) { @@ -1146,7 +1151,12 @@
> > > > ifcvf_dev_close(int vid)
> > > > internal->sw_fallback_running = false;
> > > > } else {
> > > > rte_atomic32_set(&internal->dev_attached, 0);
> > > > - update_datapath(internal);
> > > > + if (update_datapath(internal) < 0) {
> > > > + DRV_LOG(ERR, "failed to update datapath for vDPA
> > > > device %s",
> > > > + vdev->device->name);
> > > > + internal->configured = 0;
> > > > + return -1;
> > > > + }
> > > > }
> > > >
> > > > internal->configured = 0;
> > > > @@ -1752,7 +1762,14 @@ ifcvf_pci_probe(struct rte_pci_driver *pci_drv
> > > > __rte_unused,
> > > > }
> > > >
> > > > rte_atomic32_set(&internal->started, 1);
> > > > - update_datapath(internal);
> > > > + if (update_datapath(internal) < 0) {
> > > > + DRV_LOG(ERR, "failed to update datapath %s",
> > > > pci_dev->name);
> > > > + rte_atomic32_set(&internal->started, 0);
> > > > + pthread_mutex_lock(&internal_list_lock);
> > > > + TAILQ_REMOVE(&internal_list, list, next);
> > > > + pthread_mutex_unlock(&internal_list_lock);
> > > > + goto error;
> > > > + }
> > > >
> >
> > Is it necessary to unregister vdpa device?
>
> Good catch, yes it's needed.
>
> Kim, please add the unregistration.
>
> Thanks,
> Chenbo
Hi Andy and Chenbo,
Thanks for your comments.
I forgot to add `rte_vdpa_unregister_device(internal->vdev)`.
I will send a new patch soon.
By the way, it seems that deallocation for `ifcvf_vfio_setup(internal)`
is also ommitted in `ifcvf_pci_probe(...)`.
I will submit another commit to split `error:` into `error2:` and `error1:`,
which calls `rte_pci_unmap_device(...)` and `rte_vfio_container_destroy(...)`.
Thanks,
Taekyung
>
> >
> > > > rte_kvargs_free(kvlist);
> > > > return 0;
> > > > @@ -1781,7 +1798,8 @@ ifcvf_pci_remove(struct rte_pci_device *pci_dev)
> > > >
> > > > internal = list->internal;
> > > > rte_atomic32_set(&internal->started, 0);
> > > > - update_datapath(internal);
> > > > + if (update_datapath(internal) < 0)
> > > > + DRV_LOG(ERR, "failed to update datapath %s",
> > > > pci_dev->name);
> > > >
> > > > rte_pci_unmap_device(internal->pdev);
> > > > rte_vfio_container_destroy(internal->vfio_container_fd);
> > > > --
> > > > 2.34.1
> > >
> > > Reviewed-by: Chenbo Xia <chenbo....@intel.com>
