Re: [PATCH rdma-next 3/3] RDMA/hns: Modify hns RoCE device's name

[Jason Gunthorpe <jgg@ziepe.ca>]

On Sat, Nov 24, 2018 at 09:01:19PM +0800, Wei Hu (Xavier) wrote:
> 
> 
> On 2018/11/24 4:39, Jason Gunthorpe wrote:
> > On Fri, Nov 23, 2018 at 11:14:25PM +0800, Wei Hu (Xavier) wrote:
> >> This patch modifies the name of hns RoCE device's name in order
> >> to ensure that the name is consistent before and after reset.
> >>
> >> Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
> >>  drivers/infiniband/hw/hns/hns_roce_device.h | 1 +
> >>  drivers/infiniband/hw/hns/hns_roce_hw_v2.c  | 3 +++
> >>  drivers/infiniband/hw/hns/hns_roce_main.c   | 4 +++-
> >>  3 files changed, 7 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
> >> index 259977b..a8cfe76 100644
> >> +++ b/drivers/infiniband/hw/hns/hns_roce_device.h
> >> @@ -954,6 +954,7 @@ struct hns_roce_dev {
> >>  	struct pci_dev		*pci_dev;
> >>  	struct device		*dev;
> >>  	struct hns_roce_uar     priv_uar;
> >> +	char			name[IB_DEVICE_NAME_MAX];
> >>  	const char		*irq_names[HNS_ROCE_MAX_IRQ_NUM];
> >>  	spinlock_t		sm_lock;
> >>  	spinlock_t		bt_cmd_lock;
> >> diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> >> index 1d639a0..678c7ec 100644
> >> +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
> >> @@ -6110,6 +6110,9 @@ static int hns_roce_hw_v2_get_cfg(struct hns_roce_dev *hr_dev,
> >>  		hr_dev->irq[i] = pci_irq_vector(handle->pdev,
> >>  						i + handle->rinfo.base_vector);
> >>  
> >> +	snprintf(hr_dev->name, IB_DEVICE_NAME_MAX, "hns%s",
> >> +		 handle->rinfo.netdev->name);
> > Why is this making up its own driver name? How is this avoiding
> > colliding with an existing name?
> >
> > This is very dangerous since we now have device renaming, the driver
> > could fail to load with no recovery.
> Hi, Jason
>        
>     The NIC driver notifies the RoCE driver to perform reset related
> processing by calling the .reset_notify() interface registered by the
> RoCE driver. If the RoCE reset processing fails, .reset_notify()
> returns non-zero, and then hns NIC driver will reschedule the
> reset task again.
>    
>     The current hardware version in hip08 SoC cannot support
> after reset process the application still communicates with the
> resources like QP requested before reset. In RoCE reset process,
> we will release the resources through ib_unregister_device, after
> the hardware reset is completed, driver will re-execute
> ib_register_device.
>    
>     Currently, we find that the ib_device's name after reset
> and the one before reset may be different. We can specify the
> device name to solve this problem.

No, now you just have unsolved races.

If you want to reset like this then you will need to do some kind of
revision to the IB core code to not loose the name assigned to the
device and not hacks like this.

Jason