[SCore-users] IRQ

[Shinji Sumimoto]

Hi.

Are you using sis900.o for PM/Ethernet?

If so, could you add the follwing patch and re-build your sis900.o
modules? Some NIC devices cause dead lock on PM/Ethernet because of
their spin-lock use in device driver.

Index: sis900.c
===================================================================
RCS file: /develop/cvsroot/score-src/kernel/linux2.4.18/drivers/net/sis900.c,v
retrieving revision 1.2
diff -u -r1.2 sis900.c
--- sis900.c	18 Feb 2002 11:40:11 -0000	1.2
+++ sis900.c	21 Dec 2002 06:34:42 -0000
@@ -1584,7 +1584,9 @@
 			skb = sis_priv->rx_skbuff[entry];
 			skb_put(skb, rx_size);
 			skb->protocol = eth_type_trans(skb, net_dev);
+			spin_unlock (&sis_priv->lock);
 			netif_rx(skb);
+			spin_lock (&sis_priv->lock);
 
 			/* some network statistics */
 			if ((rx_status & BCAST) == MCAST)
===================================================================

Shinji.

From: Marcin Krol <mykrol at cyf-kr.edu.pl>
Subject: [SCore-users] IRQ
Date: Fri, 20 Dec 2002 14:22:29 +0100 (MET)
Message-ID: <Pine.HPX.4.44.0212201346480.18989-100000 at kinga.cyf-kr.edu.pl>

mykrol> Thank you for the PM/ethernet answer, the problem was that EIT didn't
mykrol> change the grub.conf file and after reboot the server machine startes with
mykrol> the old kernel.
mykrol> Unfortunately I have another problem - when I run all mpic++ applications
mykrol> except for hello.cc (found in examples directory), either in single-, or
mykrol> multi-user environment, the server machine hungs and I have to switch the
mykrol> power off to reboot it. hello.cc works, but it takes about 1 min to
mykrol> execyte it. A compute node writes out a message:
mykrol> spurious 8259A interrupt in IRQ7
mykrol> I checked /proc/interrupts on the server and the compute nodes and they
mykrol> look like:
mykrol> Server:
mykrol> 
mykrol> marcin at node6 marcin]$ more /proc/interrupts
mykrol> 	   CPU0
mykrol>      0:     104979          XT-PIC  timer
mykrol>      1:        440          XT-PIC  keyboard
mykrol>      2:          0          XT-PIC  cascade
mykrol>      4:       4554          XT-PIC  serial
mykrol>      8:          1          XT-PIC  rtc
mykrol>      10:       2051          XT-PIC  usb-uhci, eth0
mykrol>      11:          0          XT-PIC  usb-uhci
mykrol>      12:          0          XT-PIC  usb-uhci, via82cxxx
mykrol>      14:      12039          XT-PIC  ide0
mykrol>      NMI:          0
mykrol> 	 LOC:     104941
mykrol> 	 ERR:         14
mykrol> 	 MIS:          0
mykrol> [marcin at node6 marcin]$
mykrol> 
mykrol> and the compute node
mykrol>            CPU0
mykrol> 	     0:      57890          XT-PIC  timer
mykrol> 	       1:       2307          XT-PIC  keyboard
mykrol> 		 2:          0          XT-PIC  cascade
mykrol> 		   5:          0          XT-PIC  usb-ohci
mykrol> 		     8:          1          XT-PIC  rtc
mykrol> 		      12:        389          XT-PIC  usb-ohci, eth0, PS/2 Mouse
mykrol> 		       14:      19309          XT-PIC  ide0
mykrol> 		       NMI:          0
mykrol> 		       LOC:      57853
mykrol> 		       ERR:          1
mykrol> 		       MIS:          0
mykrol> 
mykrol> 
mykrol> How can I find and fix any overlapping?
mykrol> Thank you in advance for your help
mykrol> Marcin
mykrol> 
mykrol> 
mykrol> _______________________________________________
mykrol> SCore-users mailing list
mykrol> SCore-users at pccluster.org
mykrol> http://www.pccluster.org/mailman/listinfo/score-users
mykrol> 
-----
Shinji Sumimoto    E-Mail: s-sumi at bd6.so-net.ne.jp