will the setting achieved by this very ioctl() call survive after exec()ing into another binary (s6-svscan/stage 2) ?
Testing on s6-linux-init 1.0.4.0 shows that a SIGWINCH is sent to s6-svscan on kbrequest after s6-linux-init has performed the ioctl. So, the answer seems to be yes. Having device state changes not survive execve() would be a very unfriendly, un-Unixy design. I don't doubt there are some of those out there, but apparently this isn't one of them. -- Laurent
