to introduce a new function timerfd_settime() and timerfd_gettime() about nuttx

[xiaoxiang781216]

Ok, we support the final one(signalfd), great!

[patacongo]

I would prefer not. These are non-standard GLIBC functions. NuttX follows the standards defined at OpenGroup.org which does not include these interfaces.

There are exceptions to this in the code base now, but I don't think we should proliferate non-standard interfaces.

[patacongo]

The solution would also required timerfd_create(). https://linux.die.net/man/2/timerfd_settime

While I don't advocate including non-standard interfaces in the OS, I would not oppose the inclusion of these GLIBC/Linux interfaces if they were properly implemented in the nuttx/fs directory (with new interfaces to sched/timer, were configurable, and had proper system calls implemented in syscall/.

[edbek]

yes, of course, the timerfd_create () function is also required. I think it would be very useful for this project to have such a set of functions:

timerfd_create ();
thimerfd_settime ();
thimerfd_gettime ();

Thank you in advance if they appear !

[xiaoxiang781216]

We can add eventfd and signalfd to the list too, these three API set is the good addition to POSIX since they strengthen the core UNIX design principle: Everything is a file. And also simplify the signal and timer handling a lot especially if caller need deal with file/socket at the same time.

[edbek]

I agree !

[patacongo]

eventfd APIs introduced with PR #1479. PR #1479 does not address signalfd and timerfd interfaces, however.

[eerimoq]

Hello!

I'm interested in timerfd_create(), timerfd_settime() and timerfd_gettime() as well. Does anyone have an implementation ready? If not, I might try to implement it myself.

[xiaoxiang781216]

@eerimoq as far as I know nobody start the work, it will be great to add timerfd support! Actually, you can reference the recent addition of eventfd PR #1479 and #1479.

[pkarashchenko]

Trying to address this issue in scope of #5014

[xiaoxiang781216]

Nice!