sctp: implement memory accounting on tx path
authorXin Long <lucien.xin@gmail.com>
Mon, 15 Apr 2019 09:15:06 +0000 (17:15 +0800)
committerDavid S. Miller <davem@davemloft.net>
Mon, 15 Apr 2019 20:36:51 +0000 (13:36 -0700)
Now when sending packets, sk_mem_charge() and sk_mem_uncharge() have been
used to set sk_forward_alloc. We just need to call sk_wmem_schedule() to
check if the allocated should be raised, and call sk_mem_reclaim() to
check if the allocated should be reduced when it's under memory pressure.

If sk_wmem_schedule() returns false, which means no memory is allowed to
allocate, it will block and wait for memory to become available.

Note different from tcp, sctp wait_for_buf happens before allocating any
skb, so memory accounting check is done with the whole msg_len before it
too.

Reported-by: Matteo Croce <mcroce@redhat.com>
Tested-by: Matteo Croce <mcroce@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/sctp/socket.c

index 9874e60c9b0d00924042c1b377bc0c777edfc4cb..f66dca3b10556406ce6955eed7b6cb7128556036 100644 (file)
@@ -1913,7 +1913,10 @@ static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
        if (sctp_wspace(asoc) < (int)msg_len)
                sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
 
-       if (sctp_wspace(asoc) <= 0) {
+       if (sk_under_memory_pressure(sk))
+               sk_mem_reclaim(sk);
+
+       if (sctp_wspace(asoc) <= 0 || !sk_wmem_schedule(sk, msg_len)) {
                timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
                err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
                if (err)
@@ -8930,7 +8933,10 @@ static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
                        goto do_error;
                if (signal_pending(current))
                        goto do_interrupted;
-               if ((int)msg_len <= sctp_wspace(asoc))
+               if (sk_under_memory_pressure(sk))
+                       sk_mem_reclaim(sk);
+               if ((int)msg_len <= sctp_wspace(asoc) &&
+                   sk_wmem_schedule(sk, msg_len))
                        break;
 
                /* Let another process have a go.  Since we are going