Description of problem: On several systems IBM eServer BladeCenter LS20 -[885011Z]-, IBM BladeCenter HS20 -[884321Z]- Driver loads but link is never detected. Both types of systems have dual Broadcom Corporation NetXtreme BCM5704S Gigabit Ethernet (rev 10) network cards. Logs: http://rhts.lab.boston.redhat.com/cgi-bin/rhts/system.cgi?id=60 http://rhts.lab.boston.redhat.com/cgi-bin/rhts/system.cgi?id=76 http://rhts.lab.boston.redhat.com/cgi-bin/rhts/system.cgi?id=110 Version-Release number of selected component (if applicable): 2.6.20-12.el5rt How reproducible: Always Steps to Reproduce: 1. Install kernel on above systems Actual results: The systems installs fine with RHEL5 GA, Once the RT kernel boots they no longer see link. ********************************************************************** ACPI: PCI interrupt for device 0000:05:01.1 disabled ACPI: PCI interrupt for device 0000:05:01.0 disabled tg3.c:v3.72 (January 8, 2007) ACPI: PCI Interrupt 0000:05:01.0[A] -> GSI 77 (level, low) -> IRQ 77 eth0: Tigon3 [partno(BCM95704A41) rev 2100 PHY(serdes)] (PCIX:133MHz:64-bit) 1000Base-SX Ethernet 00:0d:60:5b:94:98 eth0: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[1] Split[0] WireSpeed[0] TSOcap[0] eth0: dma_rwctrl[769f4000] dma_mask[64-bit] ACPI: PCI Interrupt 0000:05:01.1[B] -> GSI 78 (level, low) -> IRQ 78 eth1: Tigon3 [partno(BCM95704A41) rev 2100 PHY(serdes)] (PCIX:133MHz:64-bit) 1000Base-SX Ethernet 00:0d:60:5b:94:99 eth1: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[0] Split[0] WireSpeed[0] TSOcap[1] eth1: dma_rwctrl[769f4000] dma_mask[64-bit] PM: Writing back config space on device 0000:05:01.0 at offset b (was 164814e4, writing 2e81014) PM: Writing back config space on device 0000:05:01.0 at offset 3 (was 804000, writing 804010) PM: Writing back config space on device 0000:05:01.0 at offset 2 (was 2000000, writing 2000010) PM: Writing back config space on device 0000:05:01.0 at offset 1 (was 2b00000, writing 2b00146) PM: Writing back config space on device 0000:05:01.0 at offset 0 (was 164914e4, writing 16a814e4) ADDRCONF(NETDEV_UP): eth0: link is not ready PM: Writing back config space on device 0000:05:01.1 at offset b (was 164814e4, writing 2e81014) PM: Writing back config space on device 0000:05:01.1 at offset 3 (was 804000, writing 804010) PM: Writing back config space on device 0000:05:01.1 at offset 2 (was 2000000, writing 2000010) PM: Writing back config space on device 0000:05:01.1 at offset 1 (was 2b00000, writing 2b00146) PM: Writing back config space on device 0000:05:01.1 at offset 0 (was 164914e4, writing 16a814e4) ADDRCONF(NETDEV_UP): eth1: link is not ready PM: Writing back config space on device 0000:05:01.0 at offset b (was 164814e4, writing 2e81014) PM: Writing back config space on device 0000:05:01.0 at offset 3 (was 804000, writing 804010) PM: Writing back config space on device 0000:05:01.0 at offset 2 (was 2000000, writing 2000010) PM: Writing back config space on device 0000:05:01.0 at offset 1 (was 2b00000, writing 2b00146) PM: Writing back config space on device 0000:05:01.0 at offset 0 (was 164914e4, writing 16a814e4) PM: Writing back config space on device 0000:05:01.1 at offset b (was 164814e4, writing 2e81014) PM: Writing back config space on device 0000:05:01.1 at offset 3 (was 804000, writing 804010) PM: Writing back config space on device 0000:05:01.1 at offset 2 (was 2000000, writing 2000010) PM: Writing back config space on device 0000:05:01.1 at offset 1 (was 2b00000, writing 2b00146) PM: Writing back config space on device 0000:05:01.1 at offset 0 (was 164914e4, writing 16a814e4) ********************************************************************** Expected results: System should work as expected. Additional info: This seems to be caused by the tg3 priority fix that was added to the build See thread mail on rhel-rt-kernel "Kernel 2.6.20-12.el5rt tg3 driver issues"
If you revert the patch then the tg3 link works? Sorry I had not tested with the RHEL kernel + tg3 as of yet as there is other unexplained behavior which seemed to mask tg3 issue all together. I had only sent the patch out for RFC and had only tested with 2.6.21-rt[something]. I will have a look on an LS20.
Keith, Yes, we reverted the tg3 patch and the link worked. The patch works fine on single tg3 systems (at least all the ones I've tried). Only seems to fail on dual/multi-NIC systems. Clark
Closing as duplicate of the original tg3 hack bug *** This bug has been marked as a duplicate of 228563 ***