I checked the source file from 4.0_BETA and I noticed allow_netmask()
(netmask.c) had a bug.
(I did it a while ago, but I did not know how to report the fix.)
It works with any netmasks other than "0.0.0.0/0" ( "any", "all" ).
When it meets "0.0.0.0/0", it executes some strange code to avoid
shifting 32 bits,

After I modified the function as the following, the bug disappeared.
If you do not see the problem, it might already have been fixed.

Hiroshi Nakamoto
------------
int
allow_netmask(const char *netmaskarg, const char *addr)
{
struct in_addr a;
struct in_addr m;
const char *netmask;
magic_t *mp;
char maskaddr[NETMASK_BUFFER_SIZE];
char *cp;
int slash;
int i;

/* firstly check for any magic values in the netmask */
netmask = netmaskarg;
for (mp = magics ; mp->magic ; mp++) {
if (strcmp(netmask, mp->magic) == 0) {
netmask = mp->xform;
break;
}
}

/* find out if slash notation has been used */
(void) memset(&a, 0x0, sizeof(a));
if ((cp = strchr(netmask, '/')) == NULL) {
(void) strlcpy(maskaddr, netmask, sizeof(maskaddr));
slash = 32;
} else {
(void) strlcpy(maskaddr, netmask, MIN(sizeof(maskaddr), (int)(cp -
netmask) + 1));
slash = atoi(cp + 1);
}

/* canonicalise IPv4 address to dotted quad */
for (i = 0, cp = maskaddr ; *cp ; cp++) {
if (*cp == '.') {
i += 1;
}
}
for ( ; i < 3 ; i++) {
(void) strlcat(maskaddr, ".0", sizeof(maskaddr));
}

/* translate netmask to in_addr */
if (!inet_aton(maskaddr, &m)) {
(void) fprintf(stderr, "allow_netmask: can't interpret mask `%s' as an
IPv4 address\n", maskaddr);
return 0;
}

/* translate address to in_addr */
if (!inet_aton(addr, &a)) {
(void) fprintf(stderr, "allow_netmask: can't interpret address `%s' as
an IPv4 address\n", addr);
return 0;
}

#ifdef ALLOW_NETMASK_DEBUG
printf("addr %s %u, mask %s %u, slash %d\n", addr,
(ISCSI_HTONL(a.s_addr) >> (32 - slash)), maskaddr,
(ISCSI_HTONL(m.s_addr) >> (32 - slash)), slash);
#endif
if (slash == 0) {
return 1;
}

/* and return 1 if address is in netmask */
return ((ISCSI_HTONL(a.s_addr) >> (32 - slash)) ==
(ISCSI_HTONL(m.s_addr) >> (32 - slash))) ? 1 : 0;
}
---------------

> Synopsis: iscsi_target rejects Windows iscsi initiator.
>
> Responsible-Changed-From-To: bin-bug-people->agc
> Responsible-Changed-By: agc@netbsd.org
> Responsible-Changed-When: Sun, 26 Nov 2006 10:56:06 +0000
> Responsible-Changed-Why:
> agc looks after the NetBSD iSCSI target
>
>
> State-Changed-From-To: open->analyzed
> State-Changed-By: agc@netbsd.org
> State-Changed-When: Sun, 26 Nov 2006 10:56:06 +0000
> State-Changed-Why:
> I've analyzed the problem, and can't reproduce it in my set up.
>
> I have a target exported to 0/0, and made up the slides/snapshots of
> the target serving up storage to the Microsoft v2 initiator, running
> on a Windows XP laptop. Target is from 4.99.1 (when the snapshots
> were made).
>
> I haven't tried this on an amd64 box (ENOHARDWARE).
>
> Can you give me a few more details of the environments involved, please?
>
> Thanks.
>
>
>
>
>