netdev - Re: [PATCH RFC iproute2-next 1/2] rdma: add 'link add/delete' commands

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 28 Nov 2018 16:21:48 -0600
From:   Steve Wise <swise@...ngridcomputing.com>
To:     Jason Gunthorpe <jgg@...pe.ca>
Cc:     Leon Romanovsky <leon@...nel.org>, dsahern@...il.com,
        stephen@...workplumber.org, netdev@...r.kernel.org,
        linux-rdma@...r.kernel.org, BMT@...ich.ibm.com
Subject: Re: [PATCH RFC iproute2-next 1/2] rdma: add 'link add/delete'
 commands



On 11/28/2018 4:17 PM, Jason Gunthorpe wrote:
> On Wed, Nov 28, 2018 at 02:18:55PM -0600, Steve Wise wrote:
>>
>> On 11/28/2018 2:13 PM, Leon Romanovsky wrote:
>>> On Wed, Nov 28, 2018 at 02:07:29PM -0600, Steve Wise wrote:
>>>> On 11/28/2018 2:04 PM, Leon Romanovsky wrote:
>>>>> On Wed, Nov 28, 2018 at 01:08:05PM -0600, Steve Wise wrote:
>>>>>> On 11/28/2018 12:26 PM, Leon Romanovsky wrote:
>>>>>>> On Thu, Sep 13, 2018 at 10:19:21AM -0700, Steve Wise wrote:
>>>>>>>> Add new 'link' subcommand 'add' and 'delete' to allow binding a soft-rdma
>>>>>>>> device to a netdev interface.
>>>>>>>>
>>>>>>>> EG:
>>>>>>>>
>>>>>>>> rdma link add rxe_eth0 type rxe dev eth0
>>>>>>>> rdma link delete rxe_eth0
>>>>>>>>
>>>>>>>> Signed-off-by: Steve Wise <swise@...ngridcomputing.com>
>>>>>>>>  rdma/link.c  | 106 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>>>>>  rdma/rdma.h  |   1 +
>>>>>>>>  rdma/utils.c |   2 +-
>>>>>>>>  3 files changed, 108 insertions(+), 1 deletion(-)
>>>>>>>>
>>>>>>>> diff --git a/rdma/link.c b/rdma/link.c
>>>>>>>> index 7a6d4b7e356d..d4f76b0ce11f 100644
>>>>>>>> +++ b/rdma/link.c
>>>>>>>> @@ -14,6 +14,8 @@
>>>>>>>>  static int link_help(struct rd *rd)
>>>>>>>>  {
>>>>>>>>  	pr_out("Usage: %s link show [DEV/PORT_INDEX]\n", rd->filename);
>>>>>>>> +	pr_out("Usage: %s link add NAME type TYPE dev DEV\n", rd->filename);
>>>>>>> I suggest to rename "dev" to be "netdev", because we are using "dev" for
>>>>>>> ib devices.
>>>>>> Yea ok.
>>>>>>
>>>>>>>> +	pr_out("Usage: %s link delete NAME type TYPE\n", rd->filename);
>>>>>>> Why do you need "type" for "delete" command?
>>>>>> Because the type is used in the kernel to find the appropriate link
>>>>>> ops.  I could change the kernel side to search all types for the device
>>>>>> name to delete? 
>>>>> I would say, yes.
>>>>> It makes "delete" operation more natural.
>>>>>
>>>>> Thanks
>>>> Perhaps.
>>>>
>>>> Note: 'ip link delete' takes a type as well...
>>> According to man section, yes.
>>> According to various guides, no.
>>> https://docs.fedoraproject.org/en-US/Fedora/20/html/Networking_Guide/sec-Configure_802_1Q_VLAN_Tagging_ip_Commands.html
>>>
>>> Thanks
>> It does make sense to not require type.  The name must be unique so that
>> should be enough.  I'll have to respin the kernel side though...
> The delete_link really should be an operation on the ib_device, not
> the link_ops thing. 
>
> That directly prevents mis-matching function callbacks..
>
> Jason
Looking at the rtnetlink newlink/dellink, I see they cache the link_ops
ptr in the net_device struct.  So when the link is deleted, then
appropriate driver-specific dellink function can be called after finding
the device to be deleted.  Should I do something along these lines?  IE
add a struct rdma_link_ops pointer to struct ib_device.

Steve.