On Thu, 2021-05-20 at 00:13 -0400, Olivier Langlois wrote:
> I know that my test case isn't conclusive. It is a failed attempt to
> capture what my program is doing.
> 
> The priority of investigating my core dump issue has substantially
> dropped last week because I did solve my primary issue (A buffer leak
> in the provided buffers to io_uring during disconnection). My program
> did run for days but it did crash morning without any core dump
> again.
> It is a very frustrating situation because it would probably be a bug
> trivial to diagnostic and fix but without the core, the logs are
> opaque
> and they just don't give no clue about why the program did crash.
> 
> A key characteristic of my program, it is that it generates at least
> 1
> io-worker thread per io_uring instance.

As I get more familiar with io_uring source code, I have come to
realize that it is practically impossible to not end up with NO io-wq
threads. They are created in from io_uring_install_fd() which is called
for every instances created.

I'm a bit lazy for rebooting my desktop and I am still running 5.11.5
on it. I guess that with this kernel version, the io_uring threads
weren't threads belonging to the user process and arent showing with ps
by searching for a specific process LWPs.

I correctly see all the generated threads when I run the test program
on an up-to-date server (5.12.4).

I have rewritten the whole test program. It has now become an io_uring
multi connection http client (it could make a nice io_uring example
program, IMHO). Still no success in reproducing the problem with it.

So, I am giving up the idea of reproducing the problem with a showcase
program unless I have some clue about how to reproduce it.

However, I can reproduce it at will with my real program. So as Linus
has suggested, I'll investigate by searching where the PF_IO_WORKER is
used.

I'll keep the list updated if I discover something.

Greetings,