Attention is currently required from: arehbein.

Patch set 3:Code-Review -1

View Change

3 comments:

• File src/common/pcu_sock.c:

  • Patch Set #2, Line 1092: return -1;

    It doesn't change the control flow afaik (I looked at `osmo_wqueue_bfd_cb()`). […]

    osmo_wqueue_bfd_cb:
    		if (rc == -EBADF)
    			goto err_badfd;
    			
    You have to return -EBADF while handling the socket fd if you close/free it, otherwise osmo_wqueue_bfd_cb() continues using it and potentially calling the write_cb for a fd/mem which is no longer existing which can cause problems.

  • Patch Set #2, Line 1102: osmo_fd_write_disable(&state->upqueue.bfd);

    From my understanding/by looking at this code, it looks like `osmo_fd_write_disable()` is meant to p […]

    that's not to prevent concurrent write-access. It's to manage the main loop poll() to tell the kernel to trigger us when the socket is available for writing again.

  • Patch Set #2, Line 1181: osmo_wqueue_init(&state->upqueue, PCU_SOCK_QLENGTH_MAX_DEFAULT);

    No it's not possible, with the current code the current length of the wqueue is always incremented/t […]

    I'm fine with whatever but use WQUEUE instead of the QLENGTH stuff :)

To view, visit change 31533. To unsubscribe, or for help writing mail filters, visit settings.