Re: [Etherboot-users] Problems with Etherboot 4.6.0

["Robb Main [Genedyne]" <genedyne@aci.on.ca>]

Peter,

I'm not familiar with the card you are using, but I experienced problems
with Intel 82595 that might shed some light.  These chips "Auto-detected"
which interface (RJ45, AUI, or BNC (although we did not implement the BNC)),
but the autodetection algorithm had a definite pre-disposition for the AUI
port.  If the chip detected sufficient noise on the AUI, it would ignore the
RJ45 traffic, and switch over to the AUI port.  We fixed the "problem" by
disabling the auto-dectection mechanism, and forced the chip to use the
RJ45.

If your chip has a similar auto-detection mechanism, you might want to try a
similar fix.

Cheers,
R. Main.

----- Original Message -----
From: Peter Kunzmann <kunzmann@citron.de>
To: <netboot@baghira.han.de>
Sent: Saturday, August 12, 2000 10:51 AM
Subject: [Etherboot-users] Problems with Etherboot 4.6.0


>
> Hello UG,
>
> I use etherboot 4.6.0 and my problem is, that the system stalls
> sporadically while loading the kernel image. I have set the
> "SIZEINDICATOR" flag
> to see where it happens, but its always a different value. Sometimes
> after a few boots,
> sometimes over a thousand boots. I have programmed a loop where the
> system boots after every loading of the kernel (3 times per minute). I
> run this test over several weeks now. I use (and must use because of the
> BNC,
> AUI, RJ45 in one card) a Etherlink XL PCI card and
> a 650 MHz. Pentium III PC.
> First I thought there is a problem in the 3c90x driver because I found
> some bugs. But after removal of the "bug" the system continues stalling.
> My question is, has anybody else encountered such a problem and can
> someone help me to give me a hint where to search for a bug.
>
> Here a short description of the bug I found in the 3c90x driver:
>
>
> ----------------- snip ---------------------
>
>  /*** a3c90x_poll: exported routine that waits for a certain length of
> time
>  *** for a packet, and if it sees none, returns 0.  This routine should
>  *** copy the packet to nic->packet if it gets a packet and set the size
>  *** in nic->packetlen.  Return 1 if a packet was found.
>  ***/
> static int
> a3c90x_poll(struct nic *nic)
>     {
>     int i, errcode;
>
>     if (!(inw(INF_3C90X.IOAddr + regCommandIntStatus_w)&0x0010))
>     {
>     return 0;
>     }
>
>     /** we don't need to acknowledge rxComplete -- the upload engine
>      ** does it for us.
>      **/
>
>     /** Build the up-load descriptor **/
>     INF_3C90X.ReceiveUPD.UpNextPtr = 0;
>     INF_3C90X.ReceiveUPD.UpPktStatus = 0;
>     INF_3C90X.ReceiveUPD.DataAddr = virt_to_bus(nic->packet);
>     INF_3C90X.ReceiveUPD.DataLength = 1536 + (1<<31);
>
>     /** Submit the upload descriptor to the NIC **/
>     outl(virt_to_bus(&(INF_3C90X.ReceiveUPD)),
>          INF_3C90X.IOAddr + regUpListPtr_l);
>
>     /** Wait for upload completion (upComplete(15) or upError (14)) **/
>     for(i=0;i<40000;i++);
>     while((INF_3C90X.ReceiveUPD.UpPktStatus & ((1<<14) | (1<<15))) == 0)
>     for(i=0;i<40000;i++);
>
>     /** Check for Error (else we have good packet) **/
>     if (INF_3C90X.ReceiveUPD.UpPktStatus & (1<<14))
>  -------------------------snap-----------------------------------------
>
> I thaught the problem is the while loop:
>     for(i=0;i<40000;i++);
>     while((INF_3C90X.ReceiveUPD.UpPktStatus & ((1<<14) | (1<<15))) == 0)
>     for(i=0;i<40000;i++);
>
> First I wondered why there is a the timing loop before the while loop.
> Then I deleted the timing loop and the system stalls. The problem is
> that the 3Com card writes the changes to the memory but the memory area
> was not declared "volatile" therefore the optimizer makes a code where
> the location is never read. I added "volatile" to the structure and it
> works. But my system stalls nevertheless. There is a similar bug in the
> transmit routine.
>
>
> Thats the code I use instead of the above:
>
>     /** Wait for upload completion (upComplete(15) or upError (14)) **/
>     for(i=0; i<50000; i++)
>     {
>         Status = INF_3C90X.ReceiveUPD.UpPktStatus;
>         if((Status & ((1<<14) | (1<<15))) != 0)
>             break;
>     }
>     if(i >= 50000)
>             printf("a3c90x_poll: UpPktStatus timeout\n");       /* tell
> us there was a t
>
> Thanks in advance
>
> Peter
>
> _______________________________________________
> Etherboot-users mailing list
> Etherboot-users@lists.sourceforge.net
> http://lists.sourceforge.net/mailman/listinfo/etherboot-users
>
===========================================================================
> This Mail was sent to netboot mailing list by:
> Peter Kunzmann <kunzmann@citron.de>
> To get help about this list, send a mail with 'help' as the only string in
> it's body to majordomo@baghira.han.de. If you have problems with this
list,
> send a mail to netboot-owner@baghira.han.de.
>

===========================================================================
This Mail was sent to netboot mailing list by:
"Robb Main [Genedyne]" <genedyne@aci.on.ca>
To get help about this list, send a mail with 'help' as the only string in
it's body to majordomo@baghira.han.de. If you have problems with this list,
send a mail to netboot-owner@baghira.han.de.