[realtek] Oversize ethernet frame causes chip to hang?

[Lynn Winebarger]

On Tue, 15 May 2001, Donald Becker wrote:
> On Mon, 14 May 2001, Lynn Winebarger wrote:
> 
> >    I've been having problems with oversize ethernet frames, and noticed
> > this code in the driver:
> 
> What driver version are you using?
> 
   1.13 with a patch suggested on the list (included at the end).

> Yes.  The code immediately below resets the receiver:
> 
> 	/* Reset the receiver, based on RealTek recommendation. (Bug?) */
> 	tp->cur_rx = 0;
> 	outb(CmdTxEnb, ioaddr + ChipCmd);
> 	/* A.C.: Reset the multicast list. */
> 	set_rx_mode(dev);
> 	outb(CmdRxEnb | CmdTxEnb, ioaddr + ChipCmd);
> 	
> The reason for the comment is that the documentation implies that the
> chip should continue operation without a reset.
  
   I can report that the device stops responding (without going down), but
an ifdown/ifup resumes operation.  With 2 cards in the machine, 1 of cards
getting an oversize frame seems to cause the other card to experience a
memory squeeze (and then it appears to go down for all intents and
purposes).  We have about 10 machines with 2 of these cards in each,
though only 2 or three get heavily used.  
   Does this code go right after A.C's comment?
   Is there any way to track down what's putting in an oversize frame?

Lynn

diff -Naur linux-2.2.19/drivers/net/rtl8139.c
linux-2.2.19-mod/drivers/net/rtl8139.c
--- linux-2.2.19/drivers/net/rtl8139.c	Tue Jan  9 01:20:15 2001
+++ linux-2.2.19-mod/drivers/net/rtl8139.c	Sat Apr 14 01:56:09 2001
@@ -24,7 +24,7 @@
 
 /* These identify the driver base version and may not be removed. */
 static const char versionA[] =
-"rtl8139.c:v1.13 1/9/2001 Donald Becker, becker@scyld.com.\n";
+"rtl8139.c:v1.13a 1/9/2001 Donald Becker, becker@scyld.com.\n";
 static const char versionB[] =
 " http://www.scyld.com/network/rtl8139.html\n";
 
@@ -1186,6 +1186,23 @@
 				printk(" %2.2x", rx_ring[ring_offset +
i]);
 			printk(".\n");
 		}
+
+		/* E. Gill */
+		/* Note from BSD driver:
+		 * Here's a totally undocumented fact for you. When the
+		 * RealTek chip is in the process of copying a packet into
+		 * RAM for you, the length will be 0xfff0. If you spot a
+		 * packet header with this value, you need to stop. The
+		 * datasheet makes absolutely no mention of this and
+		 * RealTek should be shot for this.
+		 */
+		if (rx_size == 0xfff0)
+			{
+				if (debug > 1)
+					printk(KERN_DEBUG"%s: Tried
copying when chip was copying"
+						   , dev->name);
+				break;
+			}
 		if (rx_status &
(RxBadSymbol|RxRunt|RxTooLong|RxCRCErr|RxBadAlign)) {
 			if (debug > 1)
 				printk(KERN_DEBUG"%s: Ethernet frame had
errors,"