coredump in lldpad with 500+ NICs

[huyizhen]

I have an environment with 500+ virtual NICs, and core dumps occur when I run lldpad on it.
When the function eloop_sock_table_dispatch calls FD_ISSET, the value of file descriptor table->table[i].sock exceeds 1024 (the kernel fds structure has only 1024 bits). As a result, the glibc determines that a buffer overflow occurs and aborts the process.
Have you ever had this problem? Other than recompiling the kernel and glibc magnifying the size of fds, is there any other way around this problem?
Change select to poll may solve this problem. Do you have any plans to change it?

[penguin359]

This is a bug and it would be best to move on to using poll()/epoll() in the future. In the meantime, I will patch this to check for potential overflow.

[huyizhen]

Thanks for your reply and pull request penguin359!
I recently tried to change select to poll. While ensuring that the original problem was resolved, I also do some tests, including lldpad -xxx, lldptool set-lldp xxx and memory leak test.
I also plan to test the NICs online/offline and the LLDP protocol interaction.
Welcome to review my PR #120 and provide feedback, i will also consider your pr to solve my problem as a temp plan.
Best regards!

@#120