Load gen segmentation fault fix

[dennisafa]

We didn't call nf_setup before onvm_nflib_run, so segfault chaos ensued. See #164

Summary:

Added nf_setup before the onvm_nflib_run call banished the segfault. Test it by running manager with ./go.sh 0,1,2,3 3 0xF0 -a 0x7f000000000 -s stdout and load_generator with ./go.sh 1 -d 1
Should get this:

Usage:

This PR includes
Resolves issues	X
Breaking API changes
Internal API changes
Usability improvements
Bug fixes
New functionality
New NF/onvm_mgr args
Changes to starting NFs
Dependency updates
Web stats updates

Merging notes:

• Dependencies: None

TODO before merging :

• PR is ready for review

Test Plan:

Run the manager and NF as described above

Review:

@kevindweb tiny change, just need a sanity check

[onvm]

In response to PR creation

CI Message

Your results will arrive shortly

[onvm]

In response to PR creation

CI Message

Run successful see results:
✔️ PR submitted to develop branch
✔️ Pktgen performance check passed
✔️ Speed Test performance check passed
❌ Linter Failed (please fix style errors)

[Results from nimbnode17]

• Median TX pps for Pktgen: 10746272
  Performance rating - 107.46% (compared to 10000000 average)

• Median TX pps for Speed Tester: 40098629
  Performance rating - 100.25% (compared to 40000000 average)

Linter Output

examples/load_generator/load_generator.c:368: Line ends in whitespace. Consider deleting these extra spaces. [whitespace/end_of_line] [4]
examples/load_generator/load_generator.c:369: Line ends in whitespace. Consider deleting these extra spaces. [whitespace/end_of_line] [4]
Total errors found: 2

[kevindweb]

@onvm will you play nice?

[kevindweb]

@onvm maybe with the config?

[onvm]

@onvm maybe with the config?

CI Message

Your results will arrive shortly

[kevindweb]

@onvm again

[onvm]

@onvm again

CI Message

Your results will arrive shortly

[kevindweb]

@onvm ubuntu 18! nimbnode1 let's go!

[onvm]

@onvm ubuntu 18! nimbnode1 let's go!

CI Message

Your results will arrive shortly

[onvm]

@onvm ubuntu 18! nimbnode1 let's go!

CI Message

Error: ERROR: Failed to copy ONVM files to nimbnode17

[onvm]

pasta

CI Message

Your results will arrive shortly

[onvm]

pasta 2

CI Message

Your results will arrive shortly

[onvm]

pasta 2

CI Message

Error: ERROR: Failed to fetch results from nimbnode17

[kevindweb]

@onvm please grab the files this time

[onvm]

@onvm please grab the files this time

CI Message

Your results will arrive shortly

[onvm]

@onvm please grab the files this time

CI Message

Error: ERROR: Script failed on nimbnode17

[kevindweb]

@onvm can you be kind

[onvm]

@onvm can you be kind

CI Message

Your results will arrive shortly

[onvm]

@onvm can you be kind

CI Message

Run successful see results:
✔️ PR submitted to develop branch
✔️ Pktgen performance check passed
✔️ Speed Test performance check passed
✔️ Linter passed

[Results from nimbnode17]

• Median TX pps for Pktgen: 10746272
  Performance rating - 107.46% (compared to 10000000 average)

• Median TX pps for Speed Tester: 40125988
  Performance rating - 100.31% (compared to 40000000 average)

[kevindweb]

@dennisafa Thanks for fixing this! Here's a screenshot of my runs. I ran load_generator, and to test more than CI, ran 2 talking speed_testers in addition. I instantiated and destroyed them in random orders which worked fine, so cleanup worked. The fix seems to work well and makes sense that we just forgot the context setup from a few months ago.

[koolzz]

@dennisafa good find, this seems easily detectable if we set some flags during init process. Can you create a task for that/submit a pr if you have time.

[dennisafa]

thanks @kevindweb and @koolzz! yes, I will look into modifying init to check for this.

[Yongeverhuang]

May i depict a problem , in ./go.sh 0,1,2,3 3 0xF0 -s stdout the load-generator run well, but i test in share core **./go.sh 0,1,2,3 3 0xF0 -s stdout -c ** ,the load-generator not run . In OpenNetVM V 19.07.

[dennisafa]

@Yongeverhuang Thanks for your comment. The only way I could replicate this issue was by having no active igb_uio devices loaded. If an active port is not detected, then mempools are not properly set up leading to that issue. Please verify that the manager + load_generator look like this when active:

As a side note, a warning message in the NF should pop up indicating that no port is present.

[Yongeverhuang]

@dennisafa Thanks for help . Let me make a full description . In no share core way, run ./go.sh 0,1,2,3 0x1 0xF0 -s stdout in the manager ，the load_generator work well, like this

But then ,run in share mode , it doesn't work ,amd no warning message pop up .Like this , the load_generator not run unless ctrl -c stop it ,the screen print out that.

It seems looks like usual when run other NF when in share core mode. Such speed-tester .

[dennisafa]

Thank you for the screenshots. What branch of openNetVM are you running? Also, what command are you running to initiate the manager in shared core mode?
It also seems like you are using the -t argument. This will shut down the NF after a period of time.

[Yongeverhuang]

@dennisafa
Thanks！ In my test ,i use the latest OpenNetVM.
In no share mode :
manager: ./go.sh 0,1,2,3 0x1 0xF0 -s stdout
load_generator : ./start_nf.sh load_generator 1 -d 1 -t 400000
In share mode
manager: ./go.sh 0,1,2,3 0x1 0xF0 -s stdout -c
load_generator : ./start_nf.sh load_generator 1 -d 1 -t 400000

-t argument just define how many pkts sent ,it will not shut down all the time in no share mode , unless using ctrl -c to stop it .
But ,in share mode ,the load_generator could not run , no thing to pop out like this:

In manager ,the NF has generator:

Anyway，i also use the commands : manager with ./go.sh 0,1,2,3 0x1 0xF0 -a 0x7f000000000 -s stdout -c and load_generator with ./go.sh 1 -d 1.It performances not run too.

[dennisafa]

Thanks! What branch are you using? We recently fixed the load_generator network function, but have not yet integrated the changes into the latest version of ONVM (master).

[Yongeverhuang]

Thanks ! I have solved the problem. Now the load_generator network function works well.