On 11/15/2018 12:59 PM, Applebee, Robert wrote:
My application is built with LWIP 1.4.1 running on a TI MCU and uses
raw TCP and raw UDP calls, no RTOS.
The TCP is used to send commands to the hardware and is always
acknowledged with an “ACK” message.
The UDP is used to send sensor data to the client every 10ms.
FWIW, I have been using LwIP 1.4.1 for over 5 years on a TI Tiva
platform with both TCP and UDP without any trouble.
I can send messages and receive “ACK” without error and I can enable
the UDP output without error but when I combine TCP and UDP my
application will eventually generate a hardware fault. Looking at the
stack it seems to be in the “plug_holes” function.
I am a first time user of LWIP. Any assistance would be appreciated.
This is my main loop that outputs the UDP:
// allocate _pbuf_ for UDP
p=*pbuf_alloc*(/PBUF_TRANSPORT/, 32, /PBUF_RAM/);
//
// Loop forever, processing the LED blinking. All the work is done in
// interrupt handlers.
//
*while*(1)
{
// get latest input
change = updateUdpPacket(udpOut);
/* can't send UDP _untill_ TCP connected */
*if*(clientIpAddr.addr!= 0) {
// send UPD if data delay expires or input changed state
*if*(change || g_100usTick == 0) {
// only output UDP if data delay is enabled (-1 is disabled)
*if*(data_stream_delay_ms >= 0) {
*if*(p) {
// Toggle the red LED.
MAP_GPIOPinWrite(GPIO_PORTP_BASE, GPIO_PIN_5,
(MAP_GPIOPinRead(GPIO_PORTP_BASE, GPIO_PIN_5) ^
GPIO_PIN_5));
*sprintf*(p->payload, "%s %04X %s\n", VM.PodId, g_usSeq, udpOut);
// increment the UDP number
g_usSeq++;
// send the UDP message
*udp_sendto*(UDPpcb, p, &clientIpAddr, UDP_PORT);
}
}
// reset the data delay counter
VM.DataStreamDelay= data_stream_delay_ms; // convert to float
_g_100usTick = VM.__DataStreamDelay__/ 0.1; __// set data delay
counter to 100us ticks_
}
}
}
How "stripped down" is this sample from your actual code?
From the looks of it, you're only ever allocating a single pbuf, then
constantly reusing it? If the stack hasn't completed the previous UDP
send by the time you try to use it again, that could break things?
Patrick Klos
Klos Technologies, Inc.
_______________________________________________
lwip-users mailing list
lwip-users@nongnu.org
https://lists.nongnu.org/mailman/listinfo/lwip-users
