https://bugzilla.novell.com/show_bug.cgi?id=484306 Summary: x86_64 kernel panic upon 'ifconfig eth1 mtu 1500' after 'ifconfig eth1 mtu 5000' Classification: openSUSE Product: openSUSE 11.1 Version: Final Platform: HP OS/Version: openSUSE 11.1 Status: NEW Severity: Critical Priority: P5 - None Component: Kernel AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: rbsmith@sandia.gov QAContact: qa@suse.de Found By: --- Created an attachment (id=278820) --> (https://bugzilla.novell.com/attachment.cgi?id=278820) 2.6.27.7-9-default kernel trace User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.20) Gecko/20081217 Firefox/2.0.0.20 Enabling jumbo frames with 'ifconfig eth1 mtu 9000' results in a kernel crash within a few seconds. Using a smaller MTU allows continued operation, but the kernel crashes on subsequent ifconfig operations. For example, while ssh'd into the machine (on eth0) ifconfig eth1 mtu 5000 ping -q -Mdo -s 4000 192.168.0.1 # works fine ifconfig eth1 mtu 1500 # crashes quickly It doesn't seem to matter how much network activity occurs between the two ifconfig operations. Similarly the kernel will crash upon shutdown when the network is stopped. This behavior has been observed on multiple machines with both non-xen and xen x86_64 kernels: 2.6.27.7-9-default, 2.6.27.7-9-xen, 2.6.27.19-3.2-xen with both the stock bnx2 driver version 1.8.0, and the driver from Broadcom's web site, version 1.8.2b. This behavior was observed with minimal-server installs of openSuSE 11.1 during initial testing. Hardware: HP DL360 G5 with Broadcom NetXtreme II BCM5708 NICs (eth0 & eth1) The target machine being pinged is a JetStor SATA416iS iSCSI RAID connected via a dedicated NetGear GS724AT switch with jumbo frames enabled. Reproducible: Always Steps to Reproduce: 1. ifconfig eth1 9000 # then wait a second or so, or 2. ifconfig eth1 5000; ping -Mdo -q -s 4000 192.168.0.1 3. ifconfig eth1 1500; dmesg # crash occurs Actual Results: Machine becomes non-responsive. If the console is connected to vt10, then one observes a kernel trace. Usually, the machine will respond to sysrq events allowing a reboot. Other times a power cycle is required. Expected Results: Process the ifconfig request and remain operational. If jumbo frames have not been enabled, the OS seems quite stable. And while jumbo frames are enabled, we are able to perform heavy write operations to the iSCSI RAID (RAID 6) from an xen domU without incident. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.