On a rx2600 machine, after a reboot, the EFI shows the MAC address as 'zero'. Power on/off of the node will bring the state back to normal. Here is a snapshot of the problem: After power off/power on: NO VOLUME LABEL [Acpi(HWP0002,100)/Pci(1|0)/Scsi(Pun0,Lun0)/HD(P Removable Media Boot [Acpi(HWP0002,0)/Pci(2|0)/Ata(Primary,Maste Load File [EFI Shell [Built-in]] Load File [Acpi(HWP0002,0)/Pci(3|0)/Mac(00306E4A7352)] -->Load File [Acpi(HWP0002,100)/Pci(2|0)/Mac(00306E4A7226)] Load File [Acpi(HWP0002,600)/Pci(1|0)/Mac(00306E4BA706)] Exit After performing a "reboot" from : NO VOLUME LABEL [Acpi(HWP0002,100)/Pci(1|0)/Scsi(Pun0,Lun0)/HD(P Removable Media Boot [Acpi(HWP0002,0)/Pci(2|0)/Ata(Primary,Maste Load File [EFI Shell [Built-in]] Load File [Acpi(HWP0002,0)/Pci(3|0)/Mac(00306E4A7352)] -->Load File [Acpi(HWP0002,100)/Pci(2|0)/Mac(000000000000)] Load File [Acpi(HWP0002,600)/Pci(1|0)/Mac(00306E4BA706)] Exit This problem is seen in RHEL4U4 which is kernel-2.6.9-42. The previous kernel kernel-2.6.9-34 did not have this problem. The broadcom tg3 driver had changed from 34 kernel to 42 kernel. Two patches had changed from 34 kernel to 42 kernel. The are --- linux-2.6.9-net-tg3-ethtool.patch and linux-2.6.9-net-tg3-update.patch The linux-2.6.9-net-tg3-ethtool.patch was the one that had introduced the problem in the tg3 driver. The problem was traced down to the function call tg3_set_power_state(tp, PCI_D3hot); in tg3_close() in the tg3.c file. When this was commented and the module built again, the MAC was visible.
Created attachment 152165 [details] Patch for RHEL5
Does this patch break WOL support?
Hi, I contacted Michael Chan (tg3 maintainer) and he confirmed my divination that the removal of the problematic line will break WoL support. Michael did some investigation and found that the problem is caused by EFI driver and the new version of it should fix this problem. Could you please provide me the version of EFI driver used on the RX2600 system? -- Ivan
If this is an EFI driver problem, then we should close it as NOTABUG.
The 235894 request same thing but for RHEL4. Michael Chan thinks that the problem is in EFI driver, but he requested additional information (EFI driver version & boot code firmware version - see 235894). I didn't receive any response so I'm still waiting.
According BZ235894 (the same issue but for RHEL4) this is not a problem of the tg3 driver but of the EFI driver (Doug Chapman was unable to reproduce this bug on his rx2600 with tg3 - EFI driver v3.0.7). So now I'm closing this issue as NOTABUG.