[LTP] [PATCH v2 1/3] syscalls/tgkill01: add new test

[Sumit Garg]

On Thu, 14 Mar 2019 at 17:53, Cyril Hrubis <chrubis@suse.cz> wrote:
>
> Hi!
> The test itself looks good, my only concern is actually that checkpoints
> cannot be used for the keeping the thread asleep during the test.
> However I can easily add one function to the futex library:
>
> TST_CHECKPOINT_SLEEP(id)
>
> That would cause the thread to wait on the checkpoint until:
>
> * we were woken up
> * we timeouted
>
> Which would basically loop on tst_checkpoint_wait() and retry in case of
> EINTR.
>

I am not sure how we would manage actual "msec_timeout" in case we get
EINTR and need to retry again as we may need to take care of elapsed
time till we receive asynchronous signal.

-Sumit

> Maybe it would be a good idea to retry on EINTR in the
> TST_CHECKPOINT_WAIT(), then we could easily use that one here as well.
> I'm not sure though if there are tests that depends on checkpoints being
> interrupted by signals though, I would have to check.
>
> For the second part we already have a function to wake all threads
> waiting on checkpoint, but we have to specify exact number of threads to
> wait for, which is there in order to avoid race coditions (i.e. thread
> was not sleeping yet at the time we tried to wake it). So we would have
> to count the number of threads we want to wake before the call to the
> TST_CHECKPOINT_WAKE2().
>
> --
> Cyril Hrubis
> chrubis@suse.cz