Teach nut-scanner to handle several IP address ranges

[jimklimov]

Now the -m cidr and -s/-e option groups can be repeated on command line, and cause several IP address ranges to be scanned for SNMP, NetXML, IPMI and old NUT buses.

Preparation for #2244 (-m auto which would likely find several connected subnets on a typical server). That step would likely need something like:

• https://stackoverflow.com/questions/4139405/how-can-i-get-to-know-the-ip-address-for-interfaces-in-c
• https://dev.to/fmtweisszwerg/cc-how-to-get-all-interface-addresses-on-the-local-device-3pki
• https://stackoverflow.com/questions/2655544/solaris-programmatic-interface-to-ifconfig (if someone would want this on Solaris before 2011 - older than Solaris 11 or OpenSolaris/illumos)
• https://stackoverflow.com/questions/20469857/is-there-any-getifaddrs-alternative-on-hp-ux-solaris-or-aix

Also may need better support for CIDR netmasks (not bit counts), and larger masks for IPv6.

Review and practical testing in an environment with many devices to see would be welcome.

[jimklimov]

In a recent commit, added the code to getifaddrs() and report what we would pass further as IP address ranges, count the netmask bits, etc.

The final code would go beyond this PoC reporting of course; the primary goal of this push is to see how many platforms would complain about unknown methods and data types :)

Like hitting https://stackoverflow.com/questions/35551879/cast-from-sockaddr-to-sockaddr-in-increases-required-alignment ...

[AppVeyorBot]

❌ Build nut 2.8.2.1877-master failed (commit d54d893da2 by @jimklimov)

[jimklimov]

From messages seen so far, the code is more portable than I expected, at least across POSIX systems.

• One issue is about different alignment of sockaddr variants, as highlighted by clang builds:

nut-scanner.c:755:37: error: cast from 'struct sockaddr *' to 'struct sockaddr_in6 *' increases required alignment from 1 to 4 [-Werror,-Wcast-align]
                                                                                struct sockaddr_in6 *sm = (struct sockaddr_in6 *)ifa->ifa_netmask;
                                                                                                          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
nut-scanner.c:769:36: error: cast from 'struct sockaddr *' to 'struct sockaddr_in *' increases required alignment from 1 to 4 [-Werror,-Wcast-align]
                                                                                struct sockaddr_in *sa = (struct sockaddr_in *)ifa->ifa_addr;
                                                                                                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
nut-scanner.c:770:36: error: cast from 'struct sockaddr *' to 'struct sockaddr_in *' increases required alignment from 1 to 4 [-Werror,-Wcast-align]
                                                                                struct sockaddr_in *sm = (struct sockaddr_in *)ifa->ifa_netmask;
                                                                                                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3 errors generated.

• Another is about different integer word sizes for the flags variable and formatting string for the print-out (64-bit on illumos builds, so not an int-sized %x):

nut-scanner.c: In function 'main':
nut-scanner.c:767:77: error: format '%x' expects argument of type 'unsigned int', but argument 6 has type 'uint64_t' {aka 'long unsigned int'} [-Werror=format=]
  767 |           printf("Interface: %s\tAddress: %s\tMask: %s (len: %i)\tFlags: %08x", ifa->ifa_name, addr, mask, masklen, ifa->ifa_flags);
      |                                                                          ~~~^                                       ~~~~~~~~~~~~~~
      |                                                                             |                                          |
      |                                                                             unsigned int                               uint64_t {aka long unsigned int}
      |                                                                          %08lx

nut-scanner.c:780:77: error: format '%x' expects argument of type 'unsigned int', but argument 6 has type 'uint64_t' {aka 'long unsigned int'} [-Werror=format=]
  780 |           printf("Interface: %s\tAddress: %s\tMask: %s (len: %i)\tFlags: %08x", ifa->ifa_name, addr, mask, masklen, ifa->ifa_flags);
      |                                                                          ~~~^                                       ~~~~~~~~~~~~~~
      |                                                                             |                                          |
      |                                                                             unsigned int                               uint64_t {aka long unsigned int}
      |                                                                          %08lx

• And so far little substantial feedback from Windows:

nut-scanner.c:44:10: fatal error: arpa/inet.h: No such file or directory
   44 | #include <arpa/inet.h>
      |          ^~~~~~~~~~~~~
compilation terminated.

Build scenarios are still running, so maybe something else would crop up. But I can start addressing these already, to leave fewer raised alarms in the next iteration...

[jimklimov]

Log from a test run, about parsing different addresses (e.g. -s followed by another -s ends up as two individual addresses):

:; make -ks -j && ./tools/nut-scanner/nut-scanner -D -s 1.2.3.4 -s 1.1.1.1 -e 1.1.1.3 -m 10.94.56.0/24 -m 127.0.0.0/28 -O -m auto
...
   0.090346     [D5] add_ip_range: only start address was provided, setting end to same: 1.2.3.4
   0.090373     [D1] Recorded IP address range #1: [1.2.3.4 .. 1.2.3.4]
   0.090399     [D1] Recorded IP address range #2: [1.1.1.1 .. 1.1.1.3]
   0.090426     [D5] Processing CIDR net/mask: 10.94.56.0/24
   0.090454     [D5] nutscan_cidr_to_ip: parsed cidr=10.94.56.0/24 into first_ip=10.94.56.0 and mask=24
   0.090482     [D5] nutscan_cidr_to_ip: parsed mask value 24
   0.090529     [D5] Extracted IP address range from CIDR net/mask: 10.94.56.0 => 10.94.56.255
   0.090556     [D1] Recorded IP address range #3: [10.94.56.0 .. 10.94.56.255]
   0.090581     [D5] Processing CIDR net/mask: 127.0.0.0/28
   0.090589     [D5] nutscan_cidr_to_ip: parsed cidr=127.0.0.0/28 into first_ip=127.0.0.0 and mask=28
   0.090611     [D5] nutscan_cidr_to_ip: parsed mask value 28
   0.090638     [D5] Extracted IP address range from CIDR net/mask: 127.0.0.0 => 127.0.0.15
   0.090663     [D1] Recorded IP address range #4: [127.0.0.0 .. 127.0.0.15]
   0.090779     [D5] Discovering getifaddrs(): Interface: lo    Address: 255.0.0.0      Mask: 255.0.0.0 (len: 8)        Flags: 00010049 IFF_LOOPBACK IFF_UP IFF_RUNNING
   0.090808     [D5] Discovering getifaddrs(): Interface: eth0  Address: 255.255.240.0  Mask: 255.255.240.0 (len: 20)   Flags: 00011043 IFF_UP IFF_RUNNING IFF_BROADCAST(is assigned)
   0.090837     [D5] Discovering getifaddrs(): Interface: eth0  Address: 255.255.240.0  Mask: 255.255.240.0 (len: 20)   Flags: 00011043 IFF_UP IFF_RUNNING IFF_BROADCAST(is assigned)
   0.090866     [D5] Discovering getifaddrs(): Interface: lo    Address: ::1    Mask: ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff (len: 128)        Flags: 00010049 IFF_LOOPBACK IFF_UP IFF_RUNNING
   0.090982     [D5] Discovering getifaddrs(): Interface: eth0  Address: fe80::215:5dff:fe8b:d950%eth0  Mask: ffff:ffff:ffff:ffff:: (len: 64)   Flags: 00011043 IFF_UP IFF_RUNNING IFF_BROADCAST(is assigned)
...

[jimklimov]

Added -m auto logic for actually converting discovered subnets into CIDR and adding to IP ranges for scanning. Now the log looks like:

   0.082536     [D1] nutscan_init: failed to load the library for NUT Client library
   0.082553     [D5] add_ip_range: only start address was provided, setting end to same: 1.2.3.4
   0.082563     [D1] Recorded IP address range #1: [1.2.3.4 .. 1.2.3.4]
   0.082593     [D1] Recorded IP address range #2: [1.1.1.1 .. 1.1.1.3]
   0.082602     [D5] Processing CIDR net/mask: 10.94.56.0/24
   0.082616     [D5] nutscan_cidr_to_ip: parsed cidr=10.94.56.0/24 into first_ip=10.94.56.0 and mask=24
   0.082644     [D5] nutscan_cidr_to_ip: parsed mask value 24
   0.082673     [D5] Extracted IP address range from CIDR net/mask: 10.94.56.0 => 10.94.56.255
   0.082718     [D1] Recorded IP address range #3: [10.94.56.0 .. 10.94.56.255]
   0.082724     [D5] Processing CIDR net/mask: 127.0.0.0/28
   0.082733     [D5] nutscan_cidr_to_ip: parsed cidr=127.0.0.0/28 into first_ip=127.0.0.0 and mask=28
   0.082738     [D5] nutscan_cidr_to_ip: parsed mask value 28
   0.082746     [D5] Extracted IP address range from CIDR net/mask: 127.0.0.0 => 127.0.0.15
   0.082750     [D1] Recorded IP address range #4: [127.0.0.0 .. 127.0.0.15]
   0.082899     [D5] Discovering getifaddrs(): Interface: lo    Address: 127.0.0.1      Mask: 255.0.0.0 (len: 8)        Flags: 00010049 IFF_LOOPBACK IFF_UP IFF_RUNNING
   0.082938     [D5] Discovering getifaddrs(): Interface: eth0  Address: 172.28.79.185  Mask: 255.255.240.0 (len: 20)   Flags: 00011043 IFF_UP IFF_RUNNING IFF_BROADCAST(is assigned)
   0.082948     [D5] Processing CIDR net/mask: 172.28.79.185/20
   0.082957     [D5] nutscan_cidr_to_ip: parsed cidr=172.28.79.185/20 into first_ip=172.28.79.185 and mask=20
   0.082961     [D5] nutscan_cidr_to_ip: parsed mask value 20
   0.082967     [D5] Extracted IP address range from CIDR net/mask: 172.28.64.0 => 172.28.79.255
   0.082971     [D1] Recorded IP address range #5: [172.28.64.0 .. 172.28.79.255]
   0.082977     [D5] Discovering getifaddrs(): Interface: eth0  Address: 172.22.48.99   Mask: 255.255.240.0 (len: 20)   Flags: 00011043 IFF_UP IFF_RUNNING IFF_BROADCAST(is assigned)
   0.082981     [D5] Processing CIDR net/mask: 172.22.48.99/20
   0.082984     [D5] nutscan_cidr_to_ip: parsed cidr=172.22.48.99/20 into first_ip=172.22.48.99 and mask=20
   0.082987     [D5] nutscan_cidr_to_ip: parsed mask value 20
   0.083017     [D5] Extracted IP address range from CIDR net/mask: 172.22.48.0 => 172.22.63.255
   0.083023     [D1] Recorded IP address range #6: [172.22.48.0 .. 172.22.63.255]
   0.083037     [D5] Discovering getifaddrs(): Interface: lo    Address: ::1    Mask: ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff (len: 128)        Flags: 00010049 IFF_LOOPBACK IFF_UP IFF_RUNNING
   0.083121     [D5] Discovering getifaddrs(): Interface: eth0  Address: fe80::215:5dff:fe8b:d950%eth0  Mask: ffff:ffff:ffff:ffff:: (len: 64)   Flags: 00011043 IFF_UP IFF_RUNNING IFF_BROADCAST(is assigned)
   0.083188     [D5] Processing CIDR net/mask: fe80::215:5dff:fe8b:d950%eth0/64
   0.083195     [D5] nutscan_cidr_to_ip: parsed cidr=fe80::215:5dff:fe8b:d950%eth0/64 into first_ip=fe80::215:5dff:fe8b:d950%eth0 and mask=64
   0.083203     [D5] nutscan_cidr_to_ip: parsed mask value 64
   0.083211     [D5] nutscan_cidr_to_ip: getaddrinfo() failed for AF_INET (IPv4): -9
   0.083271     [D5] Extracted IP address range from CIDR net/mask: fe80:: => fe80::ffff:ffff:ffff:ffff
   0.083314     [D1] Recorded IP address range #7: [fe80:: .. fe80::ffff:ffff:ffff:ffff]

Need to chop off %interface from IPv6 addresses (or ignore it in nutscan_cidr_to_ip; probably move up from 32-bit mask assumption too).

[jimklimov]

Aha, not a big deal: [D5] nutscan_cidr_to_ip: getaddrinfo() failed for AF_INET (IPv4, will retry with IPv6): -9: Address family for hostname not supported

[jimklimov]

@aquette : link-local addresses (unicast ones like fe80::/10 or 169.254.0.0/16, or multicast ones) can be a problem by default, ordering millions or billions of scans. Do we want to impose some limits? Even a lazy org with 10.0.0.0/8 distributed over DHCP can be a problem :D

https://en.wikipedia.org/wiki/Link-local_address

Unfortunately, there does not seem to be a portable flag for those, e.g. https://man7.org/linux/man-pages/man7/netdevice.7.html -- any ideas?

[jimklimov]

Ultimately, separating the -m auto experiments from an already achieved and hopefully portable increment of ability to scan multiple IP ranges, and application of some lessons-learned (e.g. avoidance of alignment problems rather than going hush-hush and sweeping them under the rug. => the rest offloaded to #2513

[AppVeyorBot]

✅ Build nut 2.8.2.1878-master completed (commit 611e8ba4fa by @jimklimov)