[dpdk-dev] [pktgen PATCH 0/6] fix compilation

Wiles, Keith keith.wiles at intel.com
Wed Jul 26 16:52:08 CEST 2017


> On Jul 26, 2017, at 12:39 AM, Thomas Monjalon <thomas at monjalon.net> wrote:
> 
> 26/07/2017 06:54, Wiles, Keith:
>> 
>>> On Jul 25, 2017, at 5:21 PM, Thomas Monjalon <thomas at monjalon.net> wrote:
>>> 
>>> Some errors were seen when trying to compile with gcc 7 or clang,
>>> for 32-bit, and with DPDK 17.08-rc2.
>>> 
>>> Thomas Monjalon (6):
>>> remove unused functions
>>> fix screen type saving
>>> fix compilation of printf based functions
>>> fix compilation with gcc 7
>>> fix 32-bit build
>>> fix build with dpdk-17.08-rc2
>>> 
>>> app/Makefile             |  3 ---
>>> app/cli-functions.c      |  2 +-
>>> app/lpktgenlib.c         | 12 ++----------
>>> app/pktgen-cmds.c        | 18 +++++++++---------
>>> app/pktgen-latency.c     | 26 ++++++++++++++------------
>>> app/pktgen-log.c         |  1 +
>>> app/pktgen-port-cfg.h    |  2 +-
>>> app/pktgen-stats.c       | 47 ++++++++++++++++++++++++++++-------------------
>>> lib/cli/cli.c            |  2 +-
>>> lib/cli/cli_cmds.c       |  6 +++---
>>> lib/cli/cli_common.h     |  1 +
>>> lib/cli/cli_map.c        |  2 +-
>>> lib/cli/cli_scrn.c       |  5 ++++-
>>> lib/cli/cli_scrn.h       |  1 +
>>> lib/common/l2p.c         | 34 ----------------------------------
>>> lib/common/mbuf.h        |  3 +++
>>> lib/common/port_config.c | 10 +---------
>>> 17 files changed, 71 insertions(+), 104 deletions(-)
>> 
>> Acked by: Keith Wiles <keith.wiles at intel.com>
>> 
>> I did apply these changes and in testing with i40e Pktgen now only sends a few thousand packets then stops sending. I am sure these patches are not the problem, but did you try out Pktgen to see if it worked with 17.08-rc2? Which version of Pktgen did you use?
> 
> I have not tested these patches except for compilation.
> I was just trying to enable testing for more people
> (with different compilers than yours).
> 
>> I tried with clang and gcc both behave the same.
>> 
>> I noticed a new message:
>> i40e_dev_start(): please call hierarchy_commit() before starting the port
>> i40e_dev_start(): please call hierarchy_commit() before starting the port
>> i40e_dev_start(): please call hierarchy_commit() before starting the port
>> i40e_dev_start(): please call hierarchy_commit() before starting the port
>> 
>> and I tried to add the call in pktgen, but it had no effect. To me the TM API should be defaulted to the original mode and just work unless someone needs to use the TM APIs. The hierarchy_commit routine could not need to be called in the normal case is my guess.
> 
> Interesting.
> Please start a new email thread for this bug, cc Wenzhuo and Cristian.
> Thanks
> 
> PS: I've noticed that you are the author of all the pktgen commits.
> Are you using git-am when applying patches from other contributors?

Yes, I do use git am, but most changes are based on bug reports and not patches. A few people have requested I do not show their name in the commit logs for some reason or privacy concerns.

In this case when I find the TX problem you will see your patches are applied. The previous patch for Intel DPDK, I had squashed it into changes I was making for all of the copyright changes. I should have stashed my changes and applied your patch instead, sorry.

Regards,
Keith



More information about the dev mailing list