Add IPv4 and IPv6 multicast implementation for UDPSocket #5196

kegilbert · 2017-09-25T23:31:49Z

Continuation of initial pull request that only added IPv4 support. Extensive rework largely took place in lwip_stack.c to add IPv4/IPv6 logic.

Tested by enabling IPv6 in mbed_lib and mbed_app.json using the following test script on two K64Fs hooked up to a common router, one acting as a broadcast node and the other as a receiver.

#include "mbed.h"
#include "EthernetInterface.h"

const char* MCAST_GRP = "FF02::1";
//const char* MCAST_GRP = "229.0.0.1";
const int MCAST_PORT = 5007;

#ifdef SEND
int main() {
    DigitalOut led(LED2);
    EthernetInterface eth;
    eth.connect();
    
    UDPSocket sock;
    sock.open(&eth);
    
    SocketAddress multicast_group;
    multicast_group.set_ip_address(MCAST_GRP);
    multicast_group.set_port(MCAST_PORT);
    
    char out_buffer[] = "very important data";
    while (true) {
        printf("Multicast to group: %s\r\n", MCAST_GRP);
        sock.sendto(multicast_group, out_buffer, sizeof(out_buffer));
        led = !led;
        Thread::wait(1000);
    }
}
#else
DigitalOut led(LED3);

void beat(void) {
    led = !led;
}

int main() {
    printf("Running recv...\r\n");
    Ticker heart;
    heart.attach(&beat, 0.25f);
    EthernetInterface eth;
    eth.connect();
    
    UDPSocket server;
    server.open(&eth);
    server.bind(MCAST_PORT);
    int ret = server.join_multicast_group(MCAST_GRP);
    if (ret != 0) {
        printf("Error joining the multicast group, error: %d\r\n", ret);
        while (true) {}
    }
   
    SocketAddress client;
    char buffer[256];
    while (true) {
        printf("\r\nWait for packet...\r\n");
        int n = server.recvfrom(&client, buffer, sizeof(buffer));
        printf("Packet from \"%s\": %s\r\n", client.get_ip_address(), buffer);
    }
}
#endif

@kjbracey-arm Thoughts on this implementation?

@theotherjimmy @geky

tommikas · 2017-09-26T05:57:28Z

Failing to compile with IAR:

23:49:50 [K64F_6LOWPAN IAR] [Warning] lwip_stack.c@298,18: [Pe111]: statement is unreachable
23:49:50 [K64F_6LOWPAN IAR] [Warning] lwip_stack.c@663,18: [Pe174]: expression has no effect
23:49:50 [K64F_6LOWPAN IAR] [Warning] lwip_stack.c@691,37: [Pe174]: expression has no effect
23:49:50 [K64F_6LOWPAN IAR] [Error] lwip_stack.c@1134,25: [Pe144]: a value of type "void const *" cannot be used to initialize an entity of type "nsapi_ip_mreq_t *"
23:49:50 [K64F_6LOWPAN IAR] [Warning] lwip_stack.c@1158,0: [Pe069]: integer conversion resulted in truncation
```

kjbracey

Looks pretty good now. My only remaining concern is: what happens when the socket is closed?

In a conventional stack, one of the main reasons for making this a socket option, as opposed to some non-socket API, is to track ownership of the joins, giving a clean-up guarantee.

When the application closes the socket, or the application dies, the joins made on that socket are cancelled.

So, the question is - do we define that as part of NSAPI, in which case this LWIP adaptation layer will need to keep its own socket membership lists for cleanup on close, as lwIP doesn't?

Or do we specify that behaviour is undefined if a socket is closed with group joins in force, and the application is required to leave before closing?

My preference would be to define cleanup - the undefined behaviour we already have between stacks causes enough problems, and here we have it in our power to implement it in the adaptation layer.

On the other hand - how much do applications come and go on mbed systems - is it worth devoting RAM+ROM space to clean-up that may never happen?

(Nanostack does implement this as a true socket option internally, so does its own clean-up on close)

kjbracey · 2017-09-26T06:03:10Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                return NSAPI_ERROR_UNSUPPORTED;
+            }
+            err_t igmp_err;
+            nsapi_ip_mreq_t *imr = optval;


Missing const on imr.

kjbracey · 2017-09-26T06:07:43Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+            }
+            #endif
+
+            return igmp_err;


Need to do error conversion from lwip to NSAPI here. I guess you could then initialise igmp_err to ERR_VAL to get NSAPI_ERROR_PARAMETER - no err_t currently maps to NSAPI_ERROR_UNSUPPORTED.

Seems reasonable - I kind of like distinguishing "I don't know this option" from "I don't like what you specified for this option". I'd argue all existing size checks returning UNSUPOPRTED should be PARAMETER too.

kjbracey · 2017-09-26T06:19:28Z

features/netsocket/UDPSocket.h

+     * @param address   Multicast group IP address
+     * @return          Negative error code on failure
+     */
+    int join_multicast_group(const SocketAddress &address);


Wondering if this should be in Socket. This is falling into a trap much of the rest of the class does in that stuff ends up in UDPSocket only because it's "not TCP". Other types of socket like raw or ICMP will want this API, and it's calling an IP-level socket option.

(And Socket is currently "IP socket", rather than generic socket)

Moved it over to Socket. Added an overload in TCPSocket to return an error for TCPSocket objects, let me know if you'd rather remove that altogether or had a better implementation.

I don't think it is particularly important to block it on a TCP socket - it's weird but harmless. On some stacks it does work - it does make you join the group regardless. The TCP socket wouldn't process those packets itself, as it never processes multicasts, but other sockets and pings would start responding.

But if you do want to block it, it logically needs to be in setsockopt, looking at s->conn->type like the TCP-specific options do.

I guess blocking it might help consistency - if some stacks block it maybe it's good if all stacks do.

kjbracey · 2017-09-26T06:27:23Z

That FF02::1 test isn't very good - that's "all hosts", so any IPv6 node should already be a member of it without you adding it.

May be worth checking you're seeing the correct IGMP/MLD messages being sent.

kegilbert · 2017-09-26T20:08:36Z

Updated the branch to fix the IAR issue and move the implementation to the Socket class. Also successfully retested on IPv6 with FF05::208 as an unassigned address.

kjbracey · 2017-09-27T10:10:04Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                return NSAPI_ERROR_PARAMETER;
+            }
+            err_t igmp_err;
+            const nsapi_ip_mreq_t *imr = (const nsapi_ip_mreq_t*) optval;


Unnecessary C-style cast is unhelpful - would have hidden the const-correctness error previously reported.

kjbracey · 2017-09-27T10:10:50Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                ip_addr_set_any(IP_IS_V6(&if_addr), &if_addr);
+            }
+
+            igmp_err = NSAPI_ERROR_UNSUPPORTED;


This needs to be an lwIP error value, as you're going to convert it.

kegilbert · 2017-09-28T00:19:30Z

Addressed @kjbracey-arm's comments and added some basic tracking of joined multicast addresses in the Socket class. On Socket close, all joined multicast addresses are left before closing to create a defined behavior.

kjbracey · 2017-09-28T11:31:45Z

features/netsocket/Socket.cpp

@@ -83,11 +94,39 @@ int Socket::modify_multicast_group(const SocketAddress &address, nsapi_socket_op

 int Socket::join_multicast_group(const SocketAddress &address)
 {
+    if(_multicast_addr_count == SOCKET_MAX_MULTICAST_MEMBERSHIPS) {


This is the wrong layer - it's in the "friendly veneer" rather than the core socket option. It means direct use of the socket options won't track.

kjbracey · 2017-09-28T11:36:07Z

features/netsocket/Socket.cpp

+    while((_multicast_addr_registry & (0x0001 << index))) { index++; }
+
+    // Create copy of socketaddress to prevent dead pointers when reading back during leave group
+    _socket_multicast_register[index] = SocketAddress(address);


It would be a copy-assignment anyway with just = address, wouldn't it?

Regardless, I think you would need to be storing the actual multicast request structures, not just the socket address, as you want to track memberships per-interface.

You might want to check for existing matches - add group normally refuses if you've already joined on the same socket. And cancel the entry if the underlying join fails?

kjbracey · 2017-09-28T11:40:53Z

features/netsocket/Socket.h

@@ -225,6 +228,9 @@ class Socket {
    mbed::Callback<void()> _event;
    mbed::Callback<void()> _callback;
    rtos::Mutex _lock;
+    uint32_t _multicast_addr_count;
+    uint32_t _multicast_addr_registry;
+    SocketAddress  _socket_multicast_register[SOCKET_MAX_MULTICAST_MEMBERSHIPS];


An array feels a little memory inefficient, especially if it was actually the multicast request with both group and interface address. Most sockets won't be doing any joins at all.

Maybe have these fields in a malloced sublock you allocate on first join? Although I know some disfavour dynamic allocation.

Agree, I wasn't sure what the stance on dynamic memory allocation was compared to static allocation in mbed generally so went with this option. Completely open to changing it.

kegilbert · 2017-10-02T20:13:56Z

Moved the socket membership logging to the lwip_stack layer. Switched to dynamic memory allocation on first multicast address add to avoid extra memory on every Socket object that does not use multicast.
@kjbracey-arm

kjbracey

I think you need to update your array first, then back out in the case of an error. In the case of an add error, you just clear your bit, the add didn't stick.

In the case of a remove error, you just remove anyway. I think if you get an error on leave, that probably means you somehow weren't a member anyway.

You should precheck if the address is in the array whether or not its an add or remove. Prechecking for add avoids pointless duplicates, and checking for remove stops you removing another socket's membership.

Something like

    index = find_multicast_member()
    if (add) {
        if (index != -1) {
             return error
        }
        copy to table, set bit
        do lwip join
        if lwip call failed, clear bit
    } else {
       if (index == -1) {
           return error;
       }
       clear bit regardless
       do lwip leave
    }

The error on the add should be the equivalent of EADDRINUSE, but doesn't seem to exist in nsapi_types.h. Maybe it should be added. It's the sort of thing we might want to return from bind...

kjbracey · 2017-10-04T11:24:29Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+            .imr_multiaddr = s->multicast_memberships[index].imr_multiaddr,
+            .imr_interface = s->multicast_memberships[index].imr_interface
+        };
+        mbed_lwip_setsockopt(0, s, 0, NSAPI_DROP_MEMBERSHIP, &address, sizeof(address));


You could just pass &s->multicast_memberships[index], no need for the temporary.

kjbracey · 2017-10-04T11:25:19Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                if(igmp_err == ERR_OK) {
+                    if(!s->multicast_memberships) {
+                        // First multicast join on this socket, allocate space for membership tracking
+                        s->multicast_memberships = malloc(sizeof(nsapi_ip_mreq_t) * LWIP_SOCKET_MAX_MEMBERSHIPS);


Need to check for memory allocation failure here.

kjbracey · 2017-10-04T11:25:50Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                    next_free_multicast_member(s->multicast_memberships_registery, index);
+
+                    s->multicast_memberships[index].imr_multiaddr = imr->imr_multiaddr;
+                    s->multicast_memberships[index].imr_interface = imr->imr_interface;


s->multicast_memberships[index] = *imr; would do.

kjbracey · 2017-10-04T11:31:09Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                        next_registered_multicast_member(s->multicast_memberships_registery, index);
+
+                        bool address_matched = true;
+                        for(int i = 0; i < NSAPI_IP_BYTES; i++) {


I'd be a bit less literal about this; rather than a manual loop, address_matched is just

memcmp(&s->multicast_memberships[index].imr_multiaddr, &imr->imr_multiaddr, sizeof(nsapi_addr_t)) == 0 && memcmp(&s->multicast_memberships[index].imr_interface, &imr->imr_interface, sizeof(nsapi_addr_t)) == 0

You could just memcmp the whole structure, if you assumed no padding, which I think is probably fine.

kjbracey · 2017-10-04T11:34:07Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                        // First multicast join on this socket, allocate space for membership tracking
+                        s->multicast_memberships = malloc(sizeof(nsapi_ip_mreq_t) * LWIP_SOCKET_MAX_MEMBERSHIPS);
+                    } else if(s->multicast_memberships_count == LWIP_SOCKET_MAX_MEMBERSHIPS) {
+                        return ERR_BUF;


If failing here, you've already joined, and the reference counting is borked. Maybe you should check for space first and fill it in.

Then clear the bit if the actual call to lwIP fails.

kegilbert · 2017-10-05T00:19:09Z

Reordered the registry updates and added in the additional allocation checks along with some minor cleanup.

kjbracey

I think this is basically there. Just style nits now.

kjbracey · 2017-10-05T08:01:59Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                    return NSAPI_ERROR_NO_ADDRESS;
+                }
+
+                clear_multicast_member_registry_bit(s->multicast_memberships_registery, find_multicast_member(s, imr));


member_pair_index - don't need to find again.

kjbracey · 2017-10-05T08:04:27Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                    // First multicast join on this socket, allocate space for membership tracking
+                    s->multicast_memberships = malloc(sizeof(nsapi_ip_mreq_t) * LWIP_SOCKET_MAX_MEMBERSHIPS);
+                    if(!s->multicast_memberships) {
+                        return ERR_BUF;


You're mixing error codes types again - you should be returning NSAPI errors

kjbracey · 2017-10-05T08:04:49Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+            igmp_err = ERR_USE; // Maps to NSAPI_ERROR_UNSUPPORTED
+            int32_t member_pair_index = find_multicast_member(s, imr);
+
+            if(optname == NSAPI_ADD_MEMBERSHIP) {


Style - if (. Applies in a lot of the most recent change

kjbracey · 2017-10-05T08:05:45Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+    // Track multicast addresses subscribed to by this socket
+    nsapi_ip_mreq_t *multicast_memberships;
+    uint32_t         multicast_memberships_count;
+    uint32_t         multicast_memberships_registery;


Spelling - registry

kjbracey · 2017-10-05T08:06:14Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+    nsapi_ip_mreq_t *multicast_memberships;
+    uint32_t         multicast_memberships_count;
+    uint32_t         multicast_memberships_registery;
+} lwip_arena[MEMP_NUM_NETCONN] = {0};


={0} is meaningless. Static data is always default-initialised to zero.

kjbracey · 2017-10-05T08:07:18Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

 static void mbed_lwip_arena_dealloc(struct lwip_socket *s)
 {
    s->in_use = false;
+
+    while(s->multicast_memberships_count > 0) {


style - while (

Tidy still needed

kjbracey · 2017-10-05T08:07:37Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

@@ -1080,6 +1108,24 @@ static nsapi_size_or_error_t mbed_lwip_socket_recvfrom(nsapi_stack_t *stack, nsa
    return recv;
 }

+int32_t find_multicast_member(struct lwip_socket *s, const nsapi_ip_mreq_t *imr) {


Should be static. s could be const.

kjbracey · 2017-10-05T08:13:04Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+    #define LWIP_SOCKET_MAX_MEMBERSHIPS 4
+#endif
+
+#define next_registered_multicast_member(registry, index)       while(!(registry & (0x0001 << index))) { index++; }


Magic macros seem a bit tricksy, and not gaining much here.

Maybe avoid confusion by making them static inline functions that return an updated index, rather than modifying index by reference?

We assume C99 or later, so static inline is portable (and works the same as in C++03).

kjbracey · 2017-10-05T08:14:35Z

As it's nearly there I'd like more review from @geky, @c1728p9, @SeppoTakalo and @mikaleppanen.

mikaleppanen · 2017-10-05T12:28:09Z

Looks good to me.

kegilbert · 2017-10-09T23:11:19Z

Updated style and moved from macros to static inline functions

mbed-ci · 2017-10-09T23:31:46Z

Build : SUCCESS

Build number : 31
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5196/

Triggering tests

/test mbed-os

mbed-ci · 2017-10-10T01:10:23Z

Test : FAILURE

Build number : 11
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/5196/11

kegilbert · 2017-10-10T03:49:33Z

/morph test

theotherjimmy · 2017-10-10T03:58:22Z

@kegilbert That was not a morph test. :D

theotherjimmy · 2017-10-10T04:01:23Z

/test mbed-os

mbed-ci · 2017-10-10T07:11:43Z

Test : FAILURE

Build number : 14
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/5196/14

kjbracey

Nearly, nearly there. Should be good to go into 5.7, I think.

kjbracey · 2017-10-10T06:43:20Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+                }
+
+                member_pair_index = 0;
+                member_pair_index = next_free_multicast_member(s, member_pair_index);


simplify to member_pair_index = next_free_multicast_member(s, 0).

kjbracey · 2017-10-10T06:45:24Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

@@ -84,6 +116,18 @@ static struct lwip_socket *mbed_lwip_arena_alloc(void)
 static void mbed_lwip_arena_dealloc(struct lwip_socket *s)
 {
    s->in_use = false;
+
+    while(s->multicast_memberships_count > 0) {
+        static uint32_t index = 0;


This static is broken - will fail on second dealloc - make it non-static, declared & initialised outside the loop.

kjbracey · 2017-10-10T06:46:09Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

 static void mbed_lwip_arena_dealloc(struct lwip_socket *s)
 {
    s->in_use = false;
+
+    while(s->multicast_memberships_count > 0) {


Tidy still needed

kjbracey · 2017-10-10T06:55:46Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+
+                s->multicast_memberships[member_pair_index] = *imr;
+
+                set_multicast_member_registry_bit(s, member_pair_index);


I guess you could reduce code size by deferring the member table update to line 1255, making the condition "if err == OK", rather than doing it now then backing out. As long as you've already chosen the member index and know there's room.

But while pondering that, it occurs to me that there's no multithreading protection here. igmp_joingroup is not one of the thread-safe APIs in lwIP.

Now, calls to your API are protected by a mutex on your Socket, so you don't need to worry about your own membership list protection. But you will need a sys_arch_protect() around just the lwIP calls, for both join and leave, to protect its per-interface list.

kjbracey · 2017-10-10T07:00:50Z

features/netsocket/Socket.cpp

+    mreq.imr_multiaddr = address.get_addr();
+    mreq.imr_interface = nsapi_addr_t();   // Default address, NSAPI_UNSPEC
+
+    return this->setsockopt(0, socketopt, &mreq, sizeof(mreq));


First parameter here should be NSAPI_SOCKET

kjbracey · 2017-10-10T07:08:23Z

features/FEATURE_LWIP/lwip-interface/lwip_stack.c

+        static uint32_t index = 0;
+        index = next_registered_multicast_member(s, index);
+
+        mbed_lwip_setsockopt(0, s, 0, NSAPI_DROP_MEMBERSHIP, &s->multicast_memberships[index],


Third parameter should technically be NSAPI_SOCKET - at some point mbed_lwip_setsockopt should enforce that.

First parameter should be written as NULL.

mbed-bot · 2017-10-10T08:39:14Z

Result: FAILURE

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 1559

Example Build failed!

mbed-ci · 2017-10-10T13:32:53Z

Test : SUCCESS

Build number : 17
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/5196/17

mbed-ci · 2017-10-11T09:12:37Z

Test : SUCCESS

Build number : 29
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/5196/29

kegilbert · 2017-10-18T19:43:54Z

This still has the Needs Review tag, should I add someone else to the review list?

geky · 2017-10-18T20:19:56Z

I think this is good to go then, @theotherjimmy, @0xc0170, should this be "needs: preceding PR"?

Unfortunately this will have to wait until we get the network tests working again: #5335

0xc0170 · 2017-10-19T15:38:08Z

I think this is good to go then, @theotherjimmy, @0xc0170, should this be "needs: preceding PR"?

I am not certain now, @kegilbert tests are green, so all should be good now?

geky · 2017-10-20T14:08:46Z

Although this needs #5335 before all tests are ran

0xc0170 · 2017-11-09T10:49:37Z

Although this needs #5335 before all tests are ran

@kegilbert Can you help @yennster with the 5335 as this depends on it? Or anything else can be done for this PR to make it ready for intergration?

kegilbert · 2017-11-09T13:21:28Z

I can take a look at it when I get back next week from GGC, don't have access to a Nucleo board at the moment.

kegilbert · 2017-11-16T20:16:54Z

#5335 was merged, rekicking off CI

kegilbert · 2017-11-16T20:17:02Z

/morph build

mbed-ci · 2017-11-16T20:49:55Z

Build : SUCCESS

Build number : 537
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5196/

Triggering tests

/morph test
/morph uvisor-test
/morph export-build

mbed-ci · 2017-11-16T22:05:53Z

Exporter Build : SUCCESS

Build number : 153
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/5196/

mbed-ci · 2017-11-16T23:08:06Z

Test : SUCCESS

Build number : 348
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/5196/348

kegilbert · 2017-11-16T23:12:24Z

@0xc0170 @geky Tests passed with the new network test patches.

geky · 2017-11-16T23:40:07Z

LGTM

kegilbert mentioned this pull request Sep 25, 2017

Add multicast implementation for UDPSocket #4820

Closed

kjbracey reviewed Sep 26, 2017

View reviewed changes

kjbracey reviewed Sep 27, 2017

View reviewed changes

kjbracey reviewed Sep 28, 2017

View reviewed changes

0xc0170 added the needs: work label Sep 28, 2017

kegilbert force-pushed the multicast-ipv6-2 branch from da69e58 to efe29d4 Compare October 2, 2017 21:04

kjbracey reviewed Oct 4, 2017

View reviewed changes

kjbracey reviewed Oct 5, 2017

View reviewed changes

theotherjimmy added needs: review and removed needs: work labels Oct 5, 2017

kegilbert force-pushed the multicast-ipv6-2 branch 2 times, most recently from 46e02df to 0c48e8b Compare October 9, 2017 23:10

kjbracey reviewed Oct 10, 2017

View reviewed changes

kjbracey approved these changes Oct 11, 2017

View reviewed changes

geky approved these changes Oct 18, 2017

View reviewed changes

geky added needs: CI needs: review needs: preceding PR and removed needs: review needs: CI labels Oct 18, 2017

geky added the needs: CI label Oct 18, 2017

geky added ready for merge and removed needs: CI needs: preceding PR labels Nov 16, 2017

theotherjimmy added the release-version: 5.7.0-rc1 label Nov 20, 2017

theotherjimmy merged commit 41dcd25 into ARMmbed:master Nov 20, 2017

theotherjimmy removed the ready for merge label Nov 20, 2017

kjbracey mentioned this pull request Jan 24, 2018

Not possible to set socket options through socket API #4034

Closed


		s->multicast_memberships[member_pair_index] = *imr;

		set_multicast_member_registry_bit(s, member_pair_index);

Add IPv4 and IPv6 multicast implementation for UDPSocket #5196

Add IPv4 and IPv6 multicast implementation for UDPSocket #5196

Uh oh!

Conversation

kegilbert commented Sep 25, 2017

Uh oh!

tommikas commented Sep 26, 2017

Uh oh!

kjbracey 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!

kjbracey Sep 27, 2017 • 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!

kjbracey commented Sep 26, 2017

Uh oh!

kegilbert commented Sep 26, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kegilbert commented Sep 28, 2017

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!

kegilbert commented Oct 2, 2017

Uh oh!

kjbracey left a comment • 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!

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!

kegilbert commented Oct 5, 2017

Uh oh!

kjbracey 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!

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!

kjbracey Oct 5, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

kjbracey Sep 27, 2017 •

edited

Loading

kjbracey left a comment •

edited

Loading

kjbracey Oct 5, 2017 •

edited

Loading