Bug 336 - Unable to exit DPDK application when running as separate thread.
Summary: Unable to exit DPDK application when running as separate thread.
Status: CONFIRMED
Alias: None
Product: DPDK
Classification: Unclassified
Component: other (show other bugs)
Version: 17.11
Hardware: All All
: Normal normal
Target Milestone: ---
Assignee: ullasdg
URL:
Depends on:
Blocks:
 
Reported: 2019-08-05 09:53 CEST by ullasdg
Modified: 2019-08-14 16:34 CEST (History)
3 users (show)



Attachments

Description ullasdg 2019-08-05 09:53:22 CEST
We have written application where EAL initialization and resource allocation done in main program. Then we are launching/calling a function below code(Reference 1).Launched Function is responsible for creating thread and perform corresponding job,but we see that after thread completes it's job we are not able to exit from application it gets hanged. 
And we are seeing newly created thread unable to access lcore value as accessed by main thread.
using below function 
rte_lcore_id()

 Below is command used to run application.

#./<executable file> -l 1-2 -d /usr/local/lib/librte_pmd_e1000.so.1.1 -w 0000:04:00.1 -w 0000:04:00.2

Also whatever statement/code written after below code is not executed in main thread.

Reference 1: Code to launch function in main program. 
________________________________________

rte_eal_mp_remote_launch(<function_name>,NULL, CALL_MASTER);
RTE_LCORE_FOREACH_SLAVE(lcore) {
      if(rte_eal_wait_lcore(lcore) < 0 ) {
                     return -1;
                   }
 }


can you please guide us what is the correct procedure to run DPDK application as separate thread/Tell us what is the mistake we are doing?.

Thanks & Regards,
ullas
Comment 1 Ajit Khaparde 2019-08-05 17:58:42 CEST
Bruce, Is this something you can look at? Can you reassign otherwise? Thanks
Comment 2 Anatoly Burakov 2019-08-12 11:36:15 CEST
i think CALL_MASTER will launch the function on the master core, rather than on one of the available slave cores. Replacing that with SKIP_MASTER should do the trick.
Comment 3 Ajit Khaparde 2019-08-12 16:00:53 CEST
Ullas, Can you try Anatoly's recommendation? Thanks
Comment 4 Anatoly Burakov 2019-08-14 16:32:37 CEST
Further investigation by Oleksandr showed that this is indeed an igb_uio issue. I'm reassigning this to Ferruh.
Comment 5 Anatoly Burakov 2019-08-14 16:34:24 CEST
Oops, wrong bug :(

Note You need to log in before you can comment on or make changes to this bug.