On Mon, 20 Apr 2020 22:16:09 -0500 bijan tabatabai <bijan311@xxxxxxxxx> wrote: > > Perhaps using the new pidfd interface may work here. > > > > https://lwn.net/Articles/794707/ > > > > It was added in 5.4 (which is still rather new), but this would be a > > new trace-cmd feature as well. > > > > -- Steve > > > This is a stupid question, but what would be the best way to use the > pidfd interface and also keep backwards compatibility with older > kernel versions? > Or is that not really a concern? Not a stupid question at all. The way to check is to call the system call and see if it works. If it doesn't simply fall back to the original behavior (to have the user do the Ctrl^C). A system call should error (errno) with -ENOSYS, if the kernel does not support the system call. As I stated, the pidfd is a new feature, but so is stopping trace-cmd when the pids end. This wouldn't be a regression. It is just an enhancement if the kernel supports it, which there's lots of other cases in trace-cmd that does the same thing (add features depending on support of the kernel). -- Steve
|