Re: [lwip-users] LwIP PCBs reset

[address@hidden]

Am 13.01.2020 um 09:27 schrieb Amena El Homsi:
> Hi, thanks for your reply, please check my answers below.
>
> On Thu, Jan 9, 2020 at 6:35 PM goldsimon <address@hidden
> <mailto:address@hidden>> wrote:
>
>
>
>     Amena El Homsi wrote:
>      >Hello,
>      >
>      >After disconnecting from an AP I have to reset the PCBs pools. If
>     I did
>      >not
>      >do that and I reconnect to the AP and If I tried to bind a socket to a
>      >port
>      >that I already used in the previous connection, binding fails (port in
>      >use).
>
>     That should not be required. Why don't you just close all
>     connections instead of such a hack? You might have SO_REUSE or
>     something like that...
>
> Sometimes the AP disassociates the STA, it doesn't make sense to go over
> the connections and reset them (it is useless to send tcp fin segment or
> dhcp release packet).
> That's why we need a way to reset everything in lwip. Currently I am
> calling lwip_init() everytime I associate to an AP and after that I set
> the netif and start network connections.
> what's wrong on this?

Calling lwip_init() won't work. It's not meant to. You cannot just call
an init function at runtime and expect it to work.

>      >
>      >To reset the PCBs I have to call memp_init() with MEMP_MEM_INIT set to
>      >1,
>      >right?
>
>     No, no, no! Hack! Don't do that. And just so that everyone reading
>     your post knows this is wrong, I *had* to answer this :-)
>
>   why is MEMP_MEM_INIT added to opt.h if we should not use it?!
> If I did not set to 1, PCBs will not be reset if we disconnected from
> the AP.
>
> So there is no way to reset everything, I have to go over the active
> connections and memset the PCB structures, without calling lwip_init()?

Ehrm, not MEMP_MEM_INIT=1 is wrong, but calling memp_init() after the
init phase is wrong. (BTW, with lwip_init(), calling memp_init()
yourself is always wrong as it is called for you)

>      >
>      >When I call memp_init(), even if I did not set MEMP_MEM_INIT to 1,
>     DHCP
>      >fails.
>
>     Ok, that's strange, but your problem lies somewhere else. Just
>     "randomly" wiping some memory (memp_init) might make your problems
>     go away for now, but others will arise!
>
>   This is happening because dhcp_dec_pcb_refcount() is not called unless
> dhcp_stop() is called.
> After disconnecting, I am calling netif_remove(), I thought it is enough
> because it doesn't make sense to send DHCP release packet if we already
> disassociated from the AP.
> Anyway, the problem is solved by calling dhcp_stop() and then
> netif_remove().

So the actual bug is that netif_remove() doesn't call dhcp_stop() if
DHCP is running? That might be worth a bug report...

I'm not sure I get the rest of your problem though.

Regads,
Simon

>      >
>      >According to wireshark capture, we send DHCP discover packet, the AP
>      >replies with the Offer, however we didn't enter dhcp_handle_offer()
>      >function.
>      >
>      >Do you have any idea what is happening?
>
>     No.
>
>     Regards,
>     Simon
>
>     _______________________________________________
>     lwip-users mailing list
>     address@hidden <mailto:address@hidden>
>     https://lists.nongnu.org/mailman/listinfo/lwip-users
>
>
>
> --
>
> Amena El-Homsi
> Computer & Communication Engineer
> Dipl. Eng,  M.S.
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/lwip-users