Forgot to answer this in the first reply, sorry.
> Note the "XXX" comment in ptrace_abort_attach(). I am not sure a simple
> UTRACE_DETACH is always correct. We already did utrace_set_events(), we
> must not, say, lose a signal if it comes in between.
I don't follow what scenario you are concerned with here.
UTRACE_DETACH will never cause you to "lose" anything. Either you'll get
some callback you've enabled before the detach completes (before your
utrace_control call, or after or simultaneous with a -EINPROGRESS return
from it), or you won't. If you don't, the signal does its normal thing.
If you do, you do. The callback just needs to:
return utrace_signal_action(action) | UTRACE_DETACH;
or
return utrace_signal_action(action) | UTRACE_RESUME;
to have no effect on the signal.
Thanks,
Roland
