anyone with bmac network problem?

Hi,

I'm looking for testers on a possible workaround for bmac network problem.
As some of you might know, bmac tx dies abruptly when network environment
is very busy and never recovers.  A patch enclosed might alleviate the
situation, and I want to know if it works for anybody else.

This problem is hard to reproduce in that it takes time and you must have
busy traffic.  I can say I have not been able to reproduce the problem
when the patch is applied, whereas I could eventually reproduce the bmac
hang without the patch, but YMMV.

The patch should apply cleanly to reasonably recent 2.2.x source and with
offsets to 2.4.x.

Let me know if you see any (or no) changes or problems with this patch.


Takashi Oe

--- bmac.c.ORIG	Sun Dec 10 18:49:42 2000
+++ bmac.c	Thu Mar 15 12:38:00 2001

@@ -772,7 +772,13 @@
 		if (txintcount < 10) {
 			XXDEBUG(("bmac_txdma_xfer_stat=%#0x\n", stat));
 		}
-		if (!(stat & ACTIVE)) break;
+		if (!(stat & ACTIVE)) {
+			/*
+			 * status field might not have been filled by DBDMA
+			 */
+			if (cp == bus_to_virt(in_le32(&bp->tx_dma->cmdptr)))
+				break;
+		}

 		if (bp->tx_bufs[bp->tx_empty]) {
 			++bp->stats.tx_packets;

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Takashi Oe wrote:

[ ... ]

Cool, seems to fix it for me.  My box (RevB iMac) survived a full motif
build on a NFS filesystem without problems.  Looks like I can turn
of my network-card-monitor-cronjob now (ifconfig down/up used to bring
the box online again with the buggy driver, which was exactly what the
crobjob did if it wasn't able to ping the default gateway).

  Gerd

--
Get back there in front of the computer NOW. Christmas can wait.
	-- Linus "the Grinch" Torvalds,  24 Dec 2000 on linux-kernel

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Takashi Oe wrote:

I'm looking for testers on a possible workaround for bmac network problem.
As some of you might know, bmac tx dies abruptly when network environment
is very busy and never recovers.  A patch enclosed might alleviate the
situation, and I want to know if it works for anybody else.

This problem is hard to reproduce in that it takes time and you must have
busy traffic.  I can say I have not been able to reproduce the problem
when the patch is applied, whereas I could eventually reproduce the bmac
hang without the patch, but YMMV.

IT WORKS!!! My test is to run the security scanner nessus against several hosts
in parallel, and it *always* crashed the network. Now it works like a dream,
absolutely stable. Good work, indeed!!! When does it go into the standard
kernel?

Thanks, Albrecht.

--

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/