Re: [PATCH net-next v9 1/2] net: wwan: core: propagate modem control signals to port drivers

[Peter Hunt]

Thanks for the review, I'll send a v10.

[Medium] user memory accessed with ops_lock held: agreed. v10 copies
the argument in before taking ops_lock and copies results out after
dropping it, as the TTY core does, so ops_lock is never held across a
user-memory access and a stalled copy can't block port removal.

[Medium] every AT/QCDM ioctl waits on ops_lock: agreed for the ioctls
that don't touch the termios/TIOCM state. In v10 TCFLSH,
TIOCEXCL/TIOCNXCL/TIOCGEXCL and unknown commands, and so the FIONREAD
fallback, don't take ops_lock at all, and the rest return -ERESTARTSYS
when interrupted. The termios/TIOCM ioctls still take ops_lock, as
agreed with Loic for now, so on rpmsg ports they can wait behind a
blocking write.

[Low] TTY comparisons and the baud rate: v10 words the comments,
kernel-doc and commit message more precisely and says the baud rate is
not used. I've kept the existing termios defaults.

[Low] "on port removal": fixed, the comment and kernel-doc now say the
lines are dropped when a port is removed while still open, and that the
open-time raise only happens once ->start() has succeeded.