On Fri, 13 Aug 2021 15:11:18 -0400
ori@eigenstate.org wrote:
> Second patch looks ok. First patch -- do we want to fail?
A computer should do what I tell it to do. -d is explicit.

> Also, why not get rid of the debug fd entirely, and either
> syslog() or fprint(2, ...)?
You make a good point. syslog() would be wrong since this is debug
info and not user info (like ssh logins).

Print to stderr is probably the best default for debug info, and can
also be redirected to a file. This obsoletes -f, doesn't care about
creating a debug log file, and rc fails if the file can't be written.

Attached patch is tested with the following:
srvfs -d man /sys/man
srvfs -d man /sys/man >[2]/tmp/dbg
auth/none srvfs -d man /sys/man >[2]/tmp/dbg

auth/none
srvfs -d man /sys/man >[2]/tmp/dbg  # correctly fails

Thanks,
Amavect