Do not try to cancel NULL thread (causing Segmentation fault)

[davidBar-On]

• Version of iperf3 (or development branch, such as master or
  3.1-STABLE) to which this pull request applies:

• Issues fixed (if any): none

• Brief description of code changes (suitable for use as a commit message):

Suggested Bug fix: thread id 0 is not regarded as illegal id (at least in Ubuntu Linux), and pthread_cancel() crashes (Segmentation fault) when trying to cancel a thread with id 0. That happens if streams creation was successful, but there was a failure before all threads are created. The problem happened to me at work and I can reproduce it in my private PC.

The suggested fix is to add an indication thread_created per stream to indicate if the thread was created for the stream, and try to cancel the stream's thread only if it was created.

[bmah888]

Thanks for the PR! This looks like a good change (set of changes). Just one comment, mostly FYI and not something that needs to be "fixed". Just need to do some testing on this...

[bmah888]

@davidBar-On I have just one more question...the code looks good according to my reading of it...I haven't been able to trigger the original bug to compare the original and fixed behavior. Any suggestion about how I can test this? Thanks!

[davidBar-On]

@bmah888, the way I tested it was by causing the client to fail after the streams creation, just before the threads creation. E.g., in iperf_client_api.c: adding goto cleanup_and_fail; just before:

                SLIST_FOREACH(sp, &test->streams, streams) {
                    if (pthread_create(&(sp->thr), &attr, &iperf_client_worker_run, sp) != 0) {
                        ...         

[bmah888]

@bmah888, the way I tested it was by causing the client to fail after the streams creation, just before the threads creation. E.g., in iperf_client_api.c: adding goto cleanup_and_fail; just before:

                SLIST_FOREACH(sp, &test->streams, streams) {
                    if (pthread_create(&(sp->thr), &attr, &iperf_client_worker_run, sp) != 0) {
                        ...         

Thanks! I'm still not able to reproduce the segmentation fault, but with the help of --debug I can see that the code in this PR is preventing iperf3 from trying to cancel non-existent threads, which is probably close enough.