[net-next,v3,0/2] Add IP_LOCAL_PORT_RANGE socket option

Message

Jakub Sitnicki Jan. 20, 2023, 11:53 a.m. UTC

This patch set is a follow up to the "How to share IPv4 addresses by
partitioning the port space" talk given at LPC 2022 [1].

Please see patch #1 for the motivation & the use case description.
Patch #2 adds tests exercising the new option in various scenarios.

Documentation
-------------

Proposed update to the ip(7) man-page:

       IP_LOCAL_PORT_RANGE (since Linux X.Y)
              Set or get the per-socket default local port range. This
              option  can  be used to clamp down the global local port
              range, defined by the ip_local_port_range  /proc  inter‐
              face described below, for a given socket.

              The option takes an uint32_t value with the high 16 bits
              set to the upper range bound, and the low 16 bits set to
              the lower range bound. Range bounds are inclusive.

              The lower bound has to be less than the upper bound when
              both bounds are not zero. Otherwise, setting the  option
              fails with EINVAL.

              If  either  bound  is  outside  of the global local port
              range, or is zero, then that bound has no effect.

              To reset the setting, pass zero as both  the  upper  and
              the lower bound.

Interaction with SELinux bind() hook
------------------------------------

SELinux bind() hook - selinux_socket_bind() - performs a permission check
if the requested local port number lies outside of the netns ephemeral port
range.

The proposed socket option cannot be used change the ephemeral port range
to extend beyond the per-netns port range, as set by
net.ipv4.ip_local_port_range.

Hence, there is no interaction with SELinux, AFAICT.
	      
Changelog:
---------

v2 -> v3:
v2: https://lore.kernel.org/r/[e-mail]

 * Describe interaction considerations with SELinux.
 * Code changes called out in individual patches.

v1 -> v2:
v1: https://lore.kernel.org/netdev/[e-mail]/

 * Fix the corner case when the per-socket range doesn't overlap with the
   per-netns range. Fallback correctly to the per-netns range. (Kuniyuki)

 * selftests: Instead of iterating over socket families (ip4, ip6) and types
   (tcp, udp), generate tests for each combo from a template. This keeps the
   code indentation level down and makes tests more granular.

 * Rewrite man-page prose:
   - explain how to unset the option,
   - document when EINVAL is returned.

RFC -> v1
RFC: https://lore.kernel.org/netdev/[e-mail]/

 * Allow either the high bound or the low bound, or both, to be zero
 * Add getsockopt support
 * Add selftests

Links:
------

[1]: https://lpc.events/event/16/contributions/1349/

To: [e-mail]
Cc: "David S. Miller" <[e-mail]>
Cc: Eric Dumazet <[e-mail]>
Cc: Jakub Kicinski <[e-mail]>
Cc: Paolo Abeni <[e-mail]>
Cc: Kuniyuki Iwashima <[e-mail]>
Cc: [e-mail]
Cc: Paul Moore <[e-mail]>
Cc: Stephen Smalley <[e-mail]>
Cc: Eric Paris <[e-mail]>
Cc: [e-mail]
Signed-off-by: Jakub Sitnicki <[e-mail]>

---
Jakub Sitnicki (2):
      inet: Add IP_LOCAL_PORT_RANGE socket option
      selftests/net: Cover the IP_LOCAL_PORT_RANGE socket option

 include/net/inet_sock.h                            |   4 +
 include/net/ip.h                                   |   3 +-
 include/uapi/linux/in.h                            |   1 +
 net/ipv4/inet_connection_sock.c                    |  25 +-
 net/ipv4/inet_hashtables.c                         |   2 +-
 net/ipv4/ip_sockglue.c                             |  18 +
 net/ipv4/udp.c                                     |   2 +-
 net/sctp/socket.c                                  |   2 +-
 tools/testing/selftests/net/Makefile               |   2 +
 tools/testing/selftests/net/ip_local_port_range.c  | 447 +++++++++++++++++++++
 tools/testing/selftests/net/ip_local_port_range.sh |   5 +
 11 files changed, 505 insertions(+), 6 deletions(-)
---
base-commit: 147c50ac3a4ea4f5ddbcf064e1adcf3aa7e6aa11
change-id: 20221221-sockopt-port-range-e142de700f4d

Best regards,

Comments

Neal Cardwell Jan. 20, 2023, 4:44 p.m. UTC | #1

)On Fri, Jan 20, 2023 at 6:53 AM Jakub Sitnicki <[e-mail]> wrote:
>
> This patch set is a follow up to the "How to share IPv4 addresses by
> partitioning the port space" talk given at LPC 2022 [1].
>
> Please see patch #1 for the motivation & the use case description.
> Patch #2 adds tests exercising the new option in various scenarios.
>
> Documentation
> -------------
>
> Proposed update to the ip(7) man-page:
>
>        IP_LOCAL_PORT_RANGE (since Linux X.Y)
>               Set or get the per-socket default local port range. This
>               option  can  be used to clamp down the global local port
>               range, defined by the ip_local_port_range  /proc  inter‐
>               face described below, for a given socket.
>
>               The option takes an uint32_t value with the high 16 bits
>               set to the upper range bound, and the low 16 bits set to
>               the lower range bound. Range bounds are inclusive.

IMHO it would be nice for this text to document whether the port
numbers are in host order or network order, and perhaps whether "high"
and "low" here refer to host or network order. Key parts of the
sockets API express port numbers in network order, but this new API
seems to express port numbers in host order, so it seem worth (a)
deciding carefully, and (b) documenting explicitly in the man page
text (here in the cover letter) and commit message for the patch
(patch #1).

thanks,
neal

Jakub Sitnicki Jan. 21, 2023, 12:45 p.m. UTC | #2

On Fri, Jan 20, 2023 at 11:44 AM -05, Neal Cardwell wrote:
> )On Fri, Jan 20, 2023 at 6:53 AM Jakub Sitnicki <[e-mail]> wrote:
>>
>> This patch set is a follow up to the "How to share IPv4 addresses by
>> partitioning the port space" talk given at LPC 2022 [1].
>>
>> Please see patch #1 for the motivation & the use case description.
>> Patch #2 adds tests exercising the new option in various scenarios.
>>
>> Documentation
>> -------------
>>
>> Proposed update to the ip(7) man-page:
>>
>>        IP_LOCAL_PORT_RANGE (since Linux X.Y)
>>               Set or get the per-socket default local port range. This
>>               option  can  be used to clamp down the global local port
>>               range, defined by the ip_local_port_range  /proc  inter‐
>>               face described below, for a given socket.
>>
>>               The option takes an uint32_t value with the high 16 bits
>>               set to the upper range bound, and the low 16 bits set to
>>               the lower range bound. Range bounds are inclusive.
>
> IMHO it would be nice for this text to document whether the port
> numbers are in host order or network order, and perhaps whether "high"
> and "low" here refer to host or network order. Key parts of the
> sockets API express port numbers in network order, but this new API
> seems to express port numbers in host order, so it seem worth (a)
> deciding carefully, and (b) documenting explicitly in the man page
> text (here in the cover letter) and commit message for the patch
> (patch #1).

Good point. Thanks for feedback.

I will expand the description for the man page and the patch #1.

Personally I don't see any upside to using the network byte order here.
With host byte order, users don't need to do anything else but pack the
two u16 values.