Hi all.
I'll use in my application socket. I'll use UPD connection and TCP
connection for receiving packets.
I have limited buffers in my application, because i have few RAM available.
I saw that setting some lwip options, i can limit packet queueing, pbuf used
(i don't know how.. i have to try) and limit memory available for lwip
stack.
My question is:
if my application tasks cannot consume all packet received from the network,
because has all buffers full, what does it happen? lwip will crash my
application? or it will drop udp packet and kill tcp connections?
My emac driver send packet to tcp_ip thread if present.... i could drop
packet in the low level side, but driver need to know when drop packet and
when pass packet to tcpip thread...
can i know this information in some way???
in my ethernetif implementation (low_level_input), if packet is present, i
do this:
....
/* We allocate a pbuf chain of pbufs from the pool. */
p = pbuf_alloc(PBUF_RAW, len, PBUF_POOL);
if (p != NULL) {
....
} else {
// Drop packet because no free pbuf available
endReadFrame(0);
LINK_STATS_INC(link.memerr);
LINK_STATS_INC(link.drop);
}
...
I suppose (please CONFIRM):
if lwip has no memory free DOESN'T crash application
BUT pbuf_alloc will reurn NULL until pbufs will free.
SO, application CAN CONSUME packets slower than network produces them,
because low_level_input will drop some packets.
I saw in lwip code that calls to sys_mbox_trypost avoid lwip blocks if
mailboxes are full, and free pbufs (SO... drop packets... IS IT RIGHT???)
again.... if lwip has no memory free DOESN'T crash application ???
Someone can discuss with me about this issues?
thanks,
Piero
_______________________________________________
lwip-users mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/lwip-users
