TDA4VE-Q1: Greely_Smart TDA4 SCI Issue

Hello,

May i know which SDK you are using? which HLOS A72 is running ?

COuld you share us the entire log?

Regards

Tarun Mukesh

Hello Tarun,

We are using greenhills for A72 core,and for the MCU R5 we using AUTOSAR os,and for the SCI server it works with the BL31,the Issue was found in GHS kernel booting.

Hello,

If AUTOSAR OS is running on MCU1_0 and SCI Server is running correctly as you confirm. what could be reason ? As far as i know TI doesn't have any GHS SDK to support.Without complete logs, i may not be very helpful here.

Regards

Tarun MUkesh

Hello Tarun,

this is in the very beginning of the GHS kernel boots,here is only Two log prints from BL31

NOTICE:  BL31: Built : 09:28:03, May  1 2024

ERROR:   GTC_CNTFID0 is 0! Assuming 200000000 Hz. Fix Bootloader

ERROR:   GTC_CNTFID0 is 0! Assuming 200000000 Hz. Fix Bootloader

and when the issue occures I open the CCS to connect the A72 core and found that the code was stuck in the 

and I go depth into the source code and I found that the failed was send by this function 

static Error J7Sci_IssueRequestSynchronous(
        J7SciDev * Dev,
        J7SciRequest Request,
        J7SciMessage * ToSend,
        J7SciMessage * ToReceive)
{
    J7SciMessage Dummy = { 0 };
    /* According to some definitions ti_sci.c in the TI linux repo the
     * RX timeout should be 1000msec to 10000msec depending on the target.
     */
    const Value TimeoutMax = RX_TIMEOUT_1MSEC * 1000;
    Error TheError = OperationTimedOut;

    Assert(Dev != NULL);
    Assert(ToSend != NULL);

    if (ToReceive == NULL)
        ToReceive = &Dummy;
    
    *&ToSend->msg.data.header = (struct SciHeader) {
        .type = Request.MessageType,
        .host = Dev->HostId,
        .seq = (Dev->MsgSequenceNo) % 255,
        .flags = Request.AdditionalFlags | (
                Request.PleaseAck? FieldValue(SCI_HDR_FLAG_ACK_ON_PROC, 1) : 0),
    };

    J7Mailbox_SendAsync(&Dev->MailboxDev, Dev->TxChanNo,
            (Address)&ToSend->msg.raw[0], sizeof(ToSend->msg.raw));

    Value Try = TimeoutMax;
    while (Try && J7Mailbox_MessagesAvailable(
            &Dev->MailboxDev, Dev->RxChanNo) == 0) {
        usDelay(10000);
        Try--;
    }

    if (Try)
        TheError = Success;

    if (TheError == Success) {
        J7Mailbox_DequeueMessage(&Dev->MailboxDev, Dev->RxChanNo,
                (Address)&ToReceive->msg.raw[0], sizeof(ToReceive->msg.raw));
    
        /* We should have gotten back an ACK with the same sequence number */
        Assert(ToReceive->msg.data.header.seq == ToSend->msg.data.header.seq);
    
        Dev->MsgSequenceNo += 1;
    }

    return TheError;
}

which call this function and check if there is replay send from MCU R5

Value J7Mailbox_MessagesAvailable(
        J7MailboxDev * Dev,
        Value Channel)
{
    Assert(Dev != NULL);
    Assert(Channel < Dev->ValidQueues);
    
    volatile uint32_t * Status =
        (void *)(Dev->StatusKernAddr + Channel * MBOX_RGN_SIZE);
    return ExtractField(*Status, MBOX_DBG_STS_MSG_CNT);
}

Hello,

Is this an sci client call J7Sci_IssueRequestSynchronous ? May i know how did you confirm SCI server running on MCU1_0?

Regards

Tarun Mukesh

Hello Tarun,

I think the SCI server was running successfully,because the before GHS kernel,there is a bl31 running,the sci client from the bl31 was running sucessfully.

and 

we add the g_dump_time_SCI array to save the times which the R5 core received ISR which send by the A72 core for Array index 0,and the task which haddle the SCI message times for array index 1,and the replay function for array index 2

and we add logs in the source bsp code in GHS kenel to count the send and success recv count to see if the message was missed in some where.

void J7Mailbox_SendAsync(
       J7MailboxDev * Dev,
       Value Channel,
       Address Data,
       Value Len)
{
    Assert(Dev != NULL);
    Assert(Data != 0);
    Assert(Channel < Dev->ValidQueues);
    static uint32_t send_count=0;
    if (Len > Dev->MaxMessageSize)
        Len = Dev->MaxMessageSize;
    
    console_printf("success send %d ",send_count++);

    /* This must be an TX Queue with some message slots available */
    Assert(J7Mailbox_IsTransmitChannel(Dev, Channel));
    Assert(J7Mailbox_MessagesAvailable(Dev, Channel) > 0);

Value J7Mailbox_MessagesAvailable(
        J7MailboxDev * Dev,
        Value Channel)
{
    Assert(Dev != NULL);
    Assert(Channel < Dev->ValidQueues);
    static uint32_t count_success=0;
    volatile uint32_t * Status =
        (volatile uint32_t*)(Dev->StatusKernAddr + Channel * MBOX_RGN_SIZE);
        
    if  (ExtractField(*Status, MBOX_DBG_STS_MSG_CNT)!=0)
    {
        console_printf("success recv %d ",count_success++);
    }

    return ExtractField(*Status, MBOX_DBG_STS_MSG_CNT);
}

and after the issue happend the log as bellows:

it seems the ghs kernel can not read the mail box cnt after the R5 replay the A72 core

Value J7Mailbox_MessagesAvailable(
        J7MailboxDev * Dev,
        Value Channel)
{
    Assert(Dev != NULL);
    Assert(Channel < Dev->ValidQueues);
    
    volatile uint32_t * Status =
        (void *)(Dev->StatusKernAddr + Channel * MBOX_RGN_SIZE);
    
    
    console_printf("J7Mailbox_MessagesAvailable Channel %d address %x status %x *status %x\r\n",Channel,Dev->StatusKernAddr,Status,*Status);

    return ExtractField(*Status, MBOX_DBG_STS_MSG_CNT);
}

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

J7Mailbox_MessagesAvailable Channel 11 address 32400000 status 3240b000 *status 40000000

is there any Precautions when we access the register 3240b000 mailbox status register ?

Hello,

yixiao han said:

I think the SCI server was running successfully,because the before GHS kernel,there is a bl31 running,the sci client from the bl31 was running sucessfully.

Can you share us the ATF logs ? WHere it is running fine. Please share me your boot flow as well.

If the SCI server is running fine on AUTOSAR MCU R5F , then instead of running GHS can you run the linux which TI provided and see everything is fine ?

since we don't have GHS related API's at our end , we cannot infer any what can be issue. TI doesn't provide any GHS SDK to look into. Mailbox registers can be accessed while using IPC.

Regards