On Sat, Jul 03, 1999 at 01:50:41PM +0200, Carsten Jacobi wrote:
> Hi there!
>
> I have a little problem in our dorm network. We have over 400 appartments in
....
> Now we thought it wouldn't be a problem to set a unique network on all
> computers and our dorm server, but Linux is refusing this netmask.
> After some search in the kernel source (2.2.9) I found this in
> a header file; include/linux/inetdevice.h line 90:
>
> extern __inline__ int bad_mask(u32 mask, u32 addr)
> {
> if (addr & (mask = ~mask))
> return 1;
> mask = ntohl(mask);
> if (mask & (mask+1))
> /* ^^^^^^^^^^^^^^^^^ */
> return 1;
> return 0;
> }
>
> Who said that a netmask must begin with a closed set of "1"-bits!?!
RFC 1519; Classless Inter-Domain Routing (CIDR)
RFC 1518; An Architecture for IP Address Allocation with CIDR
RFC 1517; Applicability Statement for the Implementation of CIDR
RFC 1817; CIDR and Classful Routing
Of course it does not say anything about INTRA-domain
routing issues..
> How shall we set our network in Linux? Indeed, we can set a second
> route, but the nice solution is the interrupted netmask.
> Arpwatch complains all the time when he finds packets on the bus
> that don't fit into the device's network too, though a second route
> is set that includes this network. We would appreciate it, if this
> condition is removed from the kernel again, or at least if one can
> switch it off in the kernel configuration. This "if condition" makes
> it even senseless to save the netmask in a 32-bit word; 5 bit
> indicating how many "1" bits are set were enough too!
> If this netmask corrupts the Linux system, than why?
Full binary masks are easier to use internally, than
e.g. mask width counters.
> Carsten Jacobi
> E-Mail: [EMAIL PROTECTED]
/Matti Aarnio <[EMAIL PROTECTED]>
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
