net: sockets: Add SOCK_RAW support for AF_INET/AF_INET6 sockets #88044

rlubos · 2025-04-02T11:49:42Z

Introduce support for raw IP sockets.

jukkar · 2025-04-02T17:14:49Z

Just wondering should we have a page in network documentation or doxygen comments in the code that describes how the various options are working (I am referring to the table in the issue of the PR). We mostly follow Linux but there are some deviations so perhaps these should be mentioned somewhere.

rlubos · 2025-04-02T17:20:28Z

Just wondering should we have a page in network documentation or doxygen comments in the code that describes how the various options are working (I am referring to the table in the issue of the PR). We mostly follow Linux but there are some deviations so perhaps these should be mentioned somewhere.

That's a good idea, we have a dedicated documentation page for sockets, perhaps it'd make most sense to add a table in there summarizing what socket types we do support.

jukkar

Some minor comments.
I am thinking about the tests, how can we verify that we have tested all combinations? Should we have a table in the test describing what we are testing and have a dedicated test function for each "cell" in the table, WDYT?

jukkar · 2025-04-03T10:16:39Z

subsys/net/ip/connection.c

+
+	raw_pkt = net_pkt_clone(pkt, CLONE_TIMEOUT);
+	if (!raw_pkt) {
+		return;


Cursor will be incorrect now. Perhaps have here goto to the end of the function as the cursor is restored there.

jukkar · 2025-04-03T10:19:52Z

subsys/net/ip/connection.c

@@ -757,6 +797,10 @@ enum net_verdict net_conn_input(struct net_pkt *pkt,
 				if (proto != ETH_P_ALL && proto != IPPROTO_RAW) {
 					continue; /* wrong protocol */
 				}
+			} else if (IS_ENABLED(CONFIG_NET_SOCKETS_INET_RAW) && raw_ip_pkt) {
+				if (conn->proto != 0) {


Should we do the check like this

if (conn->proto != IPPROTO_IP) {

to make it clear what we are checking against (value 0 is a bit vague).

jukkar · 2025-04-03T10:23:09Z

subsys/net/ip/connection.c

@@ -793,6 +837,15 @@ enum net_verdict net_conn_input(struct net_pkt *pkt,

 				continue; /* packet was consumed */
 			}
+		} else if (IS_ENABLED(CONFIG_NET_SOCKETS_INET_RAW) && raw_ip_pkt) {


Note that few lines above, the code is still checking IPPROTO_RAW for packet socket, we need to remove/change those checks from packet socket

Yes, do you want me to extend this PR with AF_PACKET/IPPROTO_RAW removal?

I think we need to remove the IPPROTO_RAW from AF_PACKET, that was actually the point of the original issue.

subsys/net/ip/net_context.c

jukkar · 2025-04-03T10:38:01Z

subsys/net/lib/sockets/sockets_inet.c

@@ -664,6 +680,11 @@ ssize_t zsock_sendmsg_ctx(struct net_context *ctx, const struct msghdr *msg,
 	k_timepoint_t buf_timeout, end;
 	int status;

+	if (net_context_get_type(ctx) == SOCK_RAW) {


Why do we prevent sendmsg for raw sockets?

Probably no good reason now that I think of it, I can drop this restriction and add some tests for sendmsg()/recvmsg().

jukkar · 2025-04-03T10:41:15Z

subsys/net/lib/sockets/sockets_inet.c

@@ -1479,6 +1503,11 @@ ssize_t zsock_recvmsg_ctx(struct net_context *ctx, struct msghdr *msg,
 	enum net_sock_type sock_type = net_context_get_type(ctx);
 	size_t i, max_len = 0;

+	if (net_context_get_type(ctx) == SOCK_RAW) {


Why prevent recvmsg for raw sockets?

rlubos · 2025-04-03T11:24:01Z

I am thinking about the tests, how can we verify that we have tested all combinations? Should we have a table in the test describing what we are testing and have a dedicated test function for each "cell" in the table, WDYT?

I can refer to the table from the issue, and extend the test suite a bit. We'll likely duplicate some tests, but that's probably not a big deal. And now I think we should also add tests verifying that all this still works if we enable AF_PACKET support.

rlubos · 2025-04-03T12:04:06Z

I have addressed the initial feedback, will work on extending the testsuite even more.

jukkar · 2025-04-03T12:42:11Z

subsys/net/ip/net_context.c

@@ -2711,6 +2851,14 @@ int net_context_send(struct net_context *context,
 		addrlen = 0;
 	}

+	if (dst_check) {
+		if (!(context->flags & NET_CONTEXT_REMOTE_ADDR_SET) ||
+		    !net_sin(&context->remote)->sin_port) {


Minor nit, could be written as
net_sin(&context->remote)->sin_port == 0
which is slightly easier to read

rlubos · 2025-04-03T14:18:35Z

In last push:

Dropped AF_PACKET/IPPROTO_RAW socket support.
Added table in sockets documentation with the summary of what is supported, let's see how it renders

jukkar

The documentation looks great. I think we need to mention in migration doc that the AF_PACKET will not accept IPPROTO_RAW as a protocol any more.

Register connection type along with family and protocol, so that it's possible to differentiate between connection listening for raw IP datagrams and TCP/UDP/other packets. Signed-off-by: Robert Lubos <[email protected]>

rlubos · 2025-04-07T14:22:35Z

Ok this should be more or less ready, changes since last week:

Refined the table in the socket documentation so that it renders better
Added migration guide entry about removed IPPROTO_RAW support from AF_PACKET
Added more test cases. I've also added summary in the test suite, based on table created by @jukkar, to make sure that all AF_INET/SOCK_RAW functionalities mentioned there are covered

rlubos · 2025-04-07T14:53:07Z

CI fixes

jukkar · 2025-04-09T08:33:16Z

@pdgendt please take a look if you have time

jukkar · 2025-04-09T08:34:28Z

cc: @go2sh please try if it solves your issues

pdgendt

Some minor comments

subsys/net/ip/ipv4.c

subsys/net/ip/ipv6.c

subsys/net/ip/net_context.c

Introduce changes in the networking stack which allow to create raw IP sockets, so that applications can send and receive raw IP datagrams. Signed-off-by: Robert Lubos <[email protected]>

Add test suite verifying AF_INET/AF_INET6 and SOCK_RAW sockets. Signed-off-by: Robert Lubos <[email protected]>

Add bullet point about raw IP sockets support. Signed-off-by: Robert Lubos <[email protected]>

Add a table in sockets documentation summarizing what socket types are being supported. Signed-off-by: Robert Lubos <[email protected]>

IPPROTO_RAW is not a valid protocol type for AF_PACKET sockets, which should only use IEEE 802.3 protocol numbers. Therefore remove support for this type of sockets. As an alternative, users can use AF_PACKET/SOCK_DGRAM or AF_INET(6)/SOCK_RAW, depending on the actual use case. Signed-off-by: Robert Lubos <[email protected]>

Add migration guide entry about dropped IPPROTO_RAW support from AF_PACKET sockets, with possible alternatives. Signed-off-by: Robert Lubos <[email protected]>

github-actions bot added area: Sockets Networking sockets area: Networking Release Notes To be mentioned in the release notes labels Apr 2, 2025

github-actions bot requested review from fabiobaltieri, jukkar, kartben, pdgendt, ssharks and tbursztyka April 2, 2025 11:50

github-actions bot assigned fabiobaltieri and kartben Apr 2, 2025

rlubos force-pushed the net/af-inet-raw-socket branch 2 times, most recently from b4d4adb to 88fffdd Compare April 2, 2025 14:02

jukkar mentioned this pull request Apr 2, 2025

net: ip: Do not use IPPROTO_RAW for AF_PACKET sockets #86885

Closed

jukkar assigned jukkar and rlubos and unassigned kartben and fabiobaltieri Apr 2, 2025

jukkar reviewed Apr 3, 2025

View reviewed changes

rlubos force-pushed the net/af-inet-raw-socket branch from 88fffdd to e96c469 Compare April 3, 2025 12:03

rlubos added the DNM This PR should not be merged (Do Not Merge) label Apr 3, 2025

jukkar reviewed Apr 3, 2025

View reviewed changes

rlubos force-pushed the net/af-inet-raw-socket branch from e96c469 to 6309400 Compare April 3, 2025 14:17

rlubos force-pushed the net/af-inet-raw-socket branch 2 times, most recently from 1479c42 to d617ec2 Compare April 4, 2025 09:18

rlubos force-pushed the net/af-inet-raw-socket branch from d617ec2 to 8f7173a Compare April 4, 2025 09:45

jukkar reviewed Apr 7, 2025

View reviewed changes

net: connection: Register connection type

9f9faa0

Register connection type along with family and protocol, so that it's possible to differentiate between connection listening for raw IP datagrams and TCP/UDP/other packets. Signed-off-by: Robert Lubos <[email protected]>

rlubos force-pushed the net/af-inet-raw-socket branch from 8f7173a to 2255341 Compare April 7, 2025 14:17

rlubos removed the DNM This PR should not be merged (Do Not Merge) label Apr 7, 2025

rlubos force-pushed the net/af-inet-raw-socket branch from 2255341 to d3f2108 Compare April 7, 2025 14:18

rlubos force-pushed the net/af-inet-raw-socket branch 2 times, most recently from b5f9065 to 615833b Compare April 7, 2025 14:52

jukkar previously approved these changes Apr 8, 2025

View reviewed changes

pdgendt previously approved these changes Apr 9, 2025

View reviewed changes

subsys/net/ip/ipv4.c Outdated Show resolved Hide resolved

subsys/net/ip/ipv6.c Outdated Show resolved Hide resolved

subsys/net/ip/net_context.c Outdated Show resolved Hide resolved

rlubos added 6 commits April 9, 2025 11:08

net: sockets: Add SOCK_RAW support for AF_INET/AF_INET6 sockets

91d1937

Introduce changes in the networking stack which allow to create raw IP sockets, so that applications can send and receive raw IP datagrams. Signed-off-by: Robert Lubos <[email protected]>

tests: net: socket: Add tests for AF_INET raw sockets

b7548ea

Add test suite verifying AF_INET/AF_INET6 and SOCK_RAW sockets. Signed-off-by: Robert Lubos <[email protected]>

doc: release-notes-4-2: Add entry for raw IP sockets

13d8e43

Add bullet point about raw IP sockets support. Signed-off-by: Robert Lubos <[email protected]>

doc: net: sockets: Add table summarizing supported socket types

633c358

Add a table in sockets documentation summarizing what socket types are being supported. Signed-off-by: Robert Lubos <[email protected]>

doc: migration-guide-4-2: Add entry about dropped IPPROTO_RAW support

847c208

Add migration guide entry about dropped IPPROTO_RAW support from AF_PACKET sockets, with possible alternatives. Signed-off-by: Robert Lubos <[email protected]>

rlubos dismissed stale reviews from pdgendt and jukkar via 847c208 April 9, 2025 09:08

rlubos force-pushed the net/af-inet-raw-socket branch from 615833b to 847c208 Compare April 9, 2025 09:08

pdgendt approved these changes Apr 9, 2025

View reviewed changes

rlubos requested a review from jukkar April 9, 2025 09:10

jukkar approved these changes Apr 9, 2025

View reviewed changes

jukkar mentioned this pull request Apr 9, 2025

net: sockets: Fix IPPROTO_RAW wrongly used with AF_PACKET #86869

Closed

kartben merged commit 52eb3e3 into zephyrproject-rtos:main Apr 9, 2025
25 checks passed

rlubos mentioned this pull request Apr 23, 2025

net: code improvement in process_data() for high throughput case #80289

Closed

rlubos deleted the net/af-inet-raw-socket branch April 24, 2025 13:13

net: sockets: Add SOCK_RAW support for AF_INET/AF_INET6 sockets #88044

net: sockets: Add SOCK_RAW support for AF_INET/AF_INET6 sockets #88044

Uh oh!

Conversation

rlubos commented Apr 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jukkar commented Apr 2, 2025

Uh oh!

rlubos commented Apr 2, 2025

Uh oh!

jukkar left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jukkar Apr 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rlubos commented Apr 3, 2025

Uh oh!

rlubos commented Apr 3, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rlubos commented Apr 3, 2025

Uh oh!

jukkar left a comment

Choose a reason for hiding this comment

Uh oh!

rlubos commented Apr 7, 2025

Uh oh!

rlubos commented Apr 7, 2025

Uh oh!

jukkar commented Apr 9, 2025

Uh oh!

jukkar commented Apr 9, 2025

Uh oh!

pdgendt left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

rlubos commented Apr 2, 2025 •

edited

Loading

jukkar Apr 3, 2025 •

edited

Loading