CCS/TDA2HV: Ethernet UDP/TCP

George Rodionov

Expert 1210 points

Part Number: TDA2HV
Other Parts Discussed in Thread: TDA2E, SYSBIOS

Tool/software: Code Composer Studio

Hello!

I'm trying to figure out some strange data transfer behavior. It's start work, then fall and now not work anyhow.

Preamble:

HW: Our custom board with TDA2HV.

SW: We don't use Vision SDK. Create project for CCS9 from nsp example. Memory: For memory configuration get file from SDK:

 c:\ti\SDK_VI\vision_sdk\sample_app\build\tda2xx\mem_segment_definition_bios.xs
 c:\ti\SDK_VI\vision_sdk\build\rtos\tda2xx\config_a15.bld

changed accordingly.

I start client and continiously send data.

Result:

The error arise 55 (No buffer space available) then it change on error 64 (Host is down). We try to increase pbm buffer size - it's not help.
I noticed that object PBMQ_free not allocated. Why it's heppend I don't know.
I called in send function _mmCheck and get next information.

When it's work by mm we get next information:

     14:48   (  5%)    15:96   ( 11%)     1:128  (  1%)     1:256  (  2%)  
     1:512  (  4%)     0:1536            0:12288         
     (61440/245760 mmAlloc: 41/0/9, mmBulk: 0/0/0)
     
     1 blocks alloced in 512 byte page
     14 blocks alloced in 48 byte page
     1 blocks alloced in 128 byte page
     15 blocks alloced in 96 byte page
     1 blocks alloced in 256 byte page

ProgrammeTDA2HV_a15_0_Ethernet.zip send ARP requests and get answers but transmition not start.
I check my send function on EVM TDA2E and it's work.

Can you help me understend why transmition not start?

over 5 years ago

0 George Rodionov over 5 years ago

Expert 1210 points

SYSBIOS: 6.46.06.00
CCS: 9.3
NDK: 2.24.02.31
NSP: nsp gmacsw 4_16_01_01
SDK: VISION SDK 3.08
XDS: xdctools 3.32.01.22
Device: TDA2HV

0 George Rodionov over 5 years ago in reply to George Rodionov

Expert 1210 points

I switch on additional debug information from NKD. Get next result:

[CortexA15_0] 
MAC Port 0 Address:
	00-35-ff-ad-15-f8
GMACSW has been started successfully
Registration of the GMACSW Successful
00002.901 KrnMsg: Msg=[New     ] A=[192.168.001.004] M=[255.255.255.255]
00004.200 KrnMsg: Msg=[New     ] A=[192.168.001.000] M=[255.255.255.000]
Network Added: If-1:192.168.1.4
Phy Status: Phy 1 found on MAC port 0
00005.501 KrnMsg: Msg=[New     ] A=[000.000.000.000] M=[000.000.000.000]
Negotiated connection: FullDuplex 1000 Mbs
Link Status: 1000Mb/s Full Duplex on PHY 1
After TaskCreate() RUN!!!
 ==================== UDP ==================== 
sudp: 1077308644
00009.801 KrnMsg: Msg=[New     ] A=[192.168.001.060] M=[255.255.255.255]
send unseccsessful, 2
ERROR task: 55
ERROR task: 55
ERROR task: 55
ERROR task: 55

ERROR task: 55
ERROR task: 55
00073.301 KrnMsg: Msg=[Down    ] A=[192.168.001.060] M=[255.255.255.255]
ERROR task: 64
ERROR task: 64
ERROR task: 64
ERROR task: 64
ERROR task: 64
ERROR task: 64
ERROR task: 64
ERROR task: 64
ERROR task: 64
00078.401 KrnMsg: Msg=[Up      ] A=[192.168.001.060] M=[255.255.255.255]
ERROR task: 55
ERROR task: 55

ERROR task: 55
00360.201 KrnMsg: Msg=[Expired ] A=[192.168.001.060] M=[255.255.255.255]
00360.802 KrnMsg: Msg=[New     ] A=[192.168.001.060] M=[255.255.255.255]
ERROR task: 55

ERROR task: 55
00367.901 KrnMsg: Msg=[Down    ] A=[192.168.001.060] M=[255.255.255.255]
ERROR task: 55

Geoge

0 Stanley Liu over 5 years ago in reply to George Rodionov

TI__Guru 50045 points

Which function returned error code 55 and which one returned 64?

0 George Rodionov over 5 years ago in reply to Stanley Liu

Expert 1210 points

This error arise in send or sento (I test work both) functions and get it error with fdError() function return.

Error 55 happend when run in udp.c:

if( !(pPkt = SockCreatePacket( hSock, (uint)size + UDPHDR_SIZE)) )
{
    udps.SndNoPacket++;
    error = ENOBUFS;
    goto UDP_EXIT;
}

error 64 returns function in the same file udp.c:

/* Send the packet */
    error=(int)IPTxPacket(pPkt,SockGetOptionFlags(hSock)&FLG_IPTX_SOSUPPORTED);

George

0 George Rodionov over 5 years ago in reply to George Rodionov

Expert 1210 points

I get the next information from structure ips and udps:

ips
Total	1328					/* total packets received */
Odropped	0					/* lost packets due to nobufs, etc. */
Badsum	0					/* checksum bad */
Badhlen	0					/* ip header length < data size */
Badlen	0					/* ip length < ip header length */
Badoptions	0					/* error in option processing */
Badvers	0					/* ip version != 4 */
Forward	0					/* packets forwarded */
Noproto	0					/* unknown or unsupported protocol */
Delivered	1328					 /* datagrams delivered to upper level */
Cantforward	1					 /* packets rcvd for unreachable dest */
CantforwardBA	0					 /* packets rcvd with illegal addressing */
Expired	0					/* Packets not forwards becaused expired */
Redirectsent	0					/* packets forwarded on same net */
Localout	2440713					/* total ip packets generated here */
Localnoroute	486					  /* Local packets discarded due to no route */
Reassembled	0					  /* total packets reassembled ok */
Fragments	0					   /* fragments received */
Fragdropped	0					  /* frags dropped (dups, out of space) */
Fragtimeout	0					 /* fragments timed out */
Fragmented	0					  /* datagrams successfully fragmented */
Ofragments	0					  /* output fragments created */
Cantfrag	0					/* don't fragment flag was set, etc. */
CacheHit	2441685					/* Route cache hit */
CacheMiss	3					   /* Route cache miss */
Filtered	0					/* packets filtered by firewall */

udps

RcvTotal	0
RcvShort	0
RcvBadLen	0
RcvBadSum	0
RcvFull	        0
RcvNoPort	0
RcvNoPortB	0
SndTotal	2441200
SndNoPacket	0

Should ARP packages be included in the information udps.RsvTotal?

George

0 George Rodionov over 5 years ago in reply to George Rodionov

Expert 1210 points

I can't find where located MAC address destination after ARP request send. If I set in plli->Status == LLI_STATUS_VALID and set MAC manualy in plli->MacAddr data start send:

George

0 George Rodionov over 5 years ago in reply to George Rodionov

Expert 1210 points

I think, that we solved this problem. The problem was in delay what need for work in signal CFG_RGMII0_RXC_IN.

To tune up delay wrote add function:

void netCloseHook()
{
    uint32_t regValue, delta, coarse, fine, siliconRev;
    static int fine_val = 3;
    static int coarse_val = 0;

    /* Global unlock for I/O Delay registers */
    CFG_IO_DELAY_LOCK = CFG_IO_DELAY_UNLOCK_KEY;

    /* Tweaks to RGMII0 Rc delay */
    CFG_RGMII0_RXC_IN = (CFG_IO_DELAY_ACCESS_PATTERN & ~CFG_IO_DELAY_LOCK_MASK);
    SYSCFG_PAD_RGMII0_RXC = (SYSCFG_PAD_RGMII0_RXC & ~0xF) | 0x0;
    delta       = (coarse_val << 5) + fine_val;     /* Delay value to add to calibrated value */
    regValue    = CFG_RGMII0_RXC_IN & ~0xFFFFFC00;
    coarse      = ((regValue >> 5) & 0x1F) + ((delta >> 5) & 0x1F);
    coarse      = (coarse > 0x1F) ? (0x1F) : (coarse);
    fine        = (regValue & 0x1F) + (delta & 0x1F);
    fine        = (fine > 0x1F) ? (0x1F) : (fine);
    regValue    = CFG_IO_DELAY_ACCESS_PATTERN | CFG_IO_DELAY_LOCK_MASK | ((coarse << 5) | (fine));
    CFG_RGMII0_RXC_IN = regValue;

    /* Global lock */
    CFG_IO_DELAY_LOCK = CFG_IO_DELAY_LOCK_KEY;

    if (fine_val >= 0x1F) {
        fine_val = 0;
        coarse_val++;
    }
    fine_val++;

    if (coarse_val >= 0x1F) {
        coarse_val = 0;
        BIOS_exit(0);
    }

    printf("Fine val: %d\n Coarse val: %d\n", fine_val, coarse_val);
    if (mCtrl.biosKILL) BIOS_exit(0); // Выход

    //DaemonFree( hHello );
}

In function send I add:

        if (status_fdError != 0){
            ErrorCounter++;
            outputData("ERROR task: %d\n", status_fdError); // C:\ti\SDK_VI\ti_components\networking\ndk_2_25_01_11\packages\ti\ndk\inc\serrno.h
            if (ErrorCounter > 3){
                printf("ERROR task: %d\n", status_fdError);
                     ErrorCounter = 0;
                     NC_NetStop(1);
                     fdCloseSession( TaskSelf() );
                     /* Close the socket. */
                     fdClose( sudp );
                     TaskExit();
                     //NDK_hookExit(TaskSelf());
            }
        }

George

0 George Rodionov over 5 years ago in reply to George Rodionov

Expert 1210 points

Please, do not close the theme, I will close it myself. Maybe I will can add something in coming weeks.

George

0 Misael Lopez Cruz over 5 years ago in reply to George Rodionov

TI__Genius 9295 points

George,

If IOdelay settings were missing on RX side, I assume you were observing some kind of errors, such as CRC. Anyway, great that the bug is solved.

You can create a new E2E ticket and mention that it's related to this one, so this ticket focuses on only one problem.

Processors

Processors forum

CCS/TDA2HV: Ethernet UDP/TCP