mbox series

[v4,0/4] ibacm: Replace ioctl with netlink and fix inablity to resurrect an interface

Message ID 20190201185049.239177-1-haakon.bugge@oracle.com (mailing list archive)
Headers show
Series ibacm: Replace ioctl with netlink and fix inablity to resurrect an interface | expand

Message

Haakon Bugge Feb. 1, 2019, 6:50 p.m. UTC
The first commit in this series re-factors the code which iterates
through the system interfaces. The code used the ioctl SIOCGIFCONF,
which only returns "running" interfaces. Restricting the enumeration
to only "running" interfaces poses several problems. The problems are
not solved by the first commit, but the third commit, "ibacm: Unable
to resurrect an interface".

As such, the first commit does not alter any functionality.

The second commit is small fix in a debug print.

The third commit is the new implementation of
https://patchwork.kernel.org/patch/10748357, which was NAKed. This
commit is now based on the netlink re-factoring from commit one.

The forth commit aligns the use of interface names vs. interface
aliases (label) with how the ip utility works.

v1 -> v2:
   * Added r-b for 2nd and 4th commit
   * Incorporated review comments from Jason and Ira
   * Minor nits fixed by the author

v2 -> v3:
   * Added Ira's r-b

v3 -> v4:
   * Fixed CMake checks for netlink (Jason)

Håkon Bugge (4):
  ibacm: Replace ioctl with netlink
  ibacm: Print correct pkey
  ibacm: Unable to resurrect an interface
  ibacm: Use alias name when possible

 CMakeLists.txt       |   5 +-
 ibacm/CMakeLists.txt |   2 +
 ibacm/src/acm.c      |   9 +-
 ibacm/src/acm_util.c | 244 +++++++++++++++++++++++--------------------
 ibacm/src/acm_util.h |   5 +-
 5 files changed, 151 insertions(+), 114 deletions(-)

--
2.20.1

Comments

Jason Gunthorpe Feb. 1, 2019, 7 p.m. UTC | #1
On Fri, Feb 01, 2019 at 07:50:45PM +0100, Håkon Bugge wrote:
> The first commit in this series re-factors the code which iterates
> through the system interfaces. The code used the ioctl SIOCGIFCONF,
> which only returns "running" interfaces. Restricting the enumeration
> to only "running" interfaces poses several problems. The problems are
> not solved by the first commit, but the third commit, "ibacm: Unable
> to resurrect an interface".
> 
> As such, the first commit does not alter any functionality.
> 
> The second commit is small fix in a debug print.
> 
> The third commit is the new implementation of
> https://patchwork.kernel.org/patch/10748357, which was NAKed. This
> commit is now based on the netlink re-factoring from commit one.
> 
> The forth commit aligns the use of interface names vs. interface
> aliases (label) with how the ip utility works.
> 
> v1 -> v2:
>    * Added r-b for 2nd and 4th commit
>    * Incorporated review comments from Jason and Ira
>    * Minor nits fixed by the author
> 
> v2 -> v3:
>    * Added Ira's r-b
> 
> v3 -> v4:
>    * Fixed CMake checks for netlink (Jason)
> 
> Håkon Bugge (4):
>   ibacm: Replace ioctl with netlink
>   ibacm: Print correct pkey
>   ibacm: Unable to resurrect an interface
>   ibacm: Use alias name when possible

You may as well send this as a github PR, lets see the CI results

Thanks
Jason
Haakon Bugge Feb. 1, 2019, 7:31 p.m. UTC | #2
> On 1 Feb 2019, at 20:00, Jason Gunthorpe <jgg@ziepe.ca> wrote:
> 
> On Fri, Feb 01, 2019 at 07:50:45PM +0100, Håkon Bugge wrote:
>> The first commit in this series re-factors the code which iterates
>> through the system interfaces. The code used the ioctl SIOCGIFCONF,
>> which only returns "running" interfaces. Restricting the enumeration
>> to only "running" interfaces poses several problems. The problems are
>> not solved by the first commit, but the third commit, "ibacm: Unable
>> to resurrect an interface".
>> 
>> As such, the first commit does not alter any functionality.
>> 
>> The second commit is small fix in a debug print.
>> 
>> The third commit is the new implementation of
>> https://patchwork.kernel.org/patch/10748357, which was NAKed. This
>> commit is now based on the netlink re-factoring from commit one.
>> 
>> The forth commit aligns the use of interface names vs. interface
>> aliases (label) with how the ip utility works.
>> 
>> v1 -> v2:
>>   * Added r-b for 2nd and 4th commit
>>   * Incorporated review comments from Jason and Ira
>>   * Minor nits fixed by the author
>> 
>> v2 -> v3:
>>   * Added Ira's r-b
>> 
>> v3 -> v4:
>>   * Fixed CMake checks for netlink (Jason)
>> 
>> Håkon Bugge (4):
>>  ibacm: Replace ioctl with netlink
>>  ibacm: Print correct pkey
>>  ibacm: Unable to resurrect an interface
>>  ibacm: Use alias name when possible
> 
> You may as well send this as a github PR, lets see the CI results

Sure, let me do that, but that's not tonight ;-)


Thxs, Håkon