CODECOMPOSER: TMS570LC4357 with EMAC

Louis crz

Hi !

I use my TMS570 as HTTP server.

I have no difficult to run this server when I turn on the microcontroler.

But when I unplug and plugin again the Ethernet the server cannot be run again and I get stuck into a while loop in function : hdkif_rx_inthandler

while(hdkif_swizzle_data(curr_bd->flags_pktlen) & EMAC_BUF_DESC_SOP)

with

curr_bd: 0x00000000

next: 0xEA00DE24

bufptr: 3942645758

bufoff_len: 3942645758

flags_pktlen: 3942645758

pbuf: 0xEAFFFFFE

I have to turn off/on to have acces to the server.

I think there is garbage data when I plug the Ethernet but why the code cannot get over this ?

Here is the content when the server run normally :

curr_bd: 0xFC521004
next: 0x181052FC
bufptr: 3435201032
bufoff_len: 1375731712
flags_pktlen: 1375731904
pbuf: 0x0802C1BC

And when it get stuck

curr_bd: 0x00000000

next: 0xEA00DE27

bufptr: 3942645758

bufoff_len: 3942645758

flags_pktlen: 3942645758

pbuf: 0xEAFFFFFE

In my main I test if the connection is cut :

flag_Phy = 0

And then reset the server. But even with that I cannot return to a normal execution.

Have you any idea please ?

When the Ethernet is plug in this run and get stuck

1 month ago

+1 jagadish gundavarapu 1 month ago

TI__Guru* 80706 points

Hi Louis,

Could you please refer below two threads:

(+) TMS570LC4357: Ethernet controller (EMAC) and new undocumented race condition on receive - Arm-based microcontrollers forum - Arm-based microcontrollers - TI E2E support forums

(+) TMS570LC4357: The TMS570LC4357, when combined with the PHY DP83640, with the protocol stack being LWIP, can receive network packets and process them normally, but it cannot s - Arm-based microcontrollers forum - Arm-based microcontrollers - TI E2E support forums

Along with above threads, could you please also refer below given internal AI response:

The problem you're experiencing is a NULL buffer descriptor (BD) corruption when Ethernet is unplugged/replugged. The curr_bd = 0x00000000 clearly indicates the RX descriptor chain has been broken or not properly re-initialized.

The garbage values (0xEAFFFFFE, 3942645758) suggest uninitialized/corrupted DMA descriptors in the EMAC RX queue.

When link drops:

EMAC hardware may misalign the RX descriptor chain
The hdkif_rx_inthandler advances curr_bd past the end of the chain
No proper re-initialization of the EMAC/DMA occurs on link restore

--
Thanks & regards,
Jagadish.

0 Louis crz 28 days ago in reply to jagadish gundavarapu

Prodigy 105 points

Hi Jagadish,

Thank you for your answer. I solved my issue.

The way I handle a hardware disconnection (calling lwipInit and init HTTP server) cause the program to open multiple TCP port.

By this way it reach the maximum number of TCP connection and create a new instance that have not a proper memory space causing my TMS570 to freeze.

Thank you.

Best regards,

Louis