[openbeosnetteam] Re: "No more B_KEEP_LOADED kernel modules" patch

> David wrote:
> > Can you post a patch?
>
> Yes, but not before 8h from now, sorry.

OK, no rush...

> It's a big patch, as I've also touch libnet and some include files (
> small prototypes fix).
> From memory:

Philippe, I know you like the name net_stack_driver, but I realy don't.
Sorry. Also, I don't want to have a /dev/net/stack entry as it gives the
wrong impression - I mean it's NOT the stack, it's just a means of getting
at the stack...

I was going to rename driver/socket.c into socket_driver.c which I think is
more meaningful. ??

We need to discuss the settings, but I don't really see how usernames and
passwords relate to huge text files (etc/services and so on).

david

>
> - the /dev/net/socket is now /dev/net/stack, which is more meaning full
> (and, btw, only matter to libnet.so internal)
> - driver/socket.c was renamed into driver/net_stack_driver.c
> - driver/Makefile updated to build & install a net_stack_driver
> from now
> - libnet/net_structures.h was renamed into net_stack_driver.h, which I
find
> a more accurate name.
> - libnet/compat.c: I've add getpassword(), and have a *hacked*
implementation for gethostname() too, but I don't like it
> [In fact, it's time to think about where & how network settings (all of
them) would be stored. Some would follow the UNIXish way, for obvious reason
(/etc/services, /etc/hosts, /etc/resolv.conf maybe), but for hostname,
username and password settings
> it's becoming an issue. And this topic it's related to our future Network
preference app design...]; find_net_setting() & set_net_setting() is there
too, and I move closesocket() here, which it's where it belong /me think.
> At last, I'was blocking on how to implement uname()...
> - server/include/sys/socket.h: moving back to ssize_t send(), I was wrong
> about these protoypes, sorry.
>
> Did core.c stop_stack() works fully in kernel mode?
>
> -Philippe
>
>

Other related posts: