From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.1) Gecko/20031114 Description of problem: When booting redhats kernel, network is not working. No ping, arp, etc. With the vanilla 2.6.3 kernel everything work fine. Version-Release number of selected component (if applicable): kernel-smp-2.6.3-1.97 How reproducible: Always Steps to Reproduce: 1. Boot redhat kernel 2. ping to machine on same LAN 3. Actual Results: 6 packets transmitted, 0 received, +3 errors, 100% packet loss Expected Results: 3 packets transmitted, 3 received, 0% packet loss Additional info:
Created attachment 97943 [details] interrupts-rh Output of cat /proc/interrupts on rh kernel
Created attachment 97944 [details] interrupts-vanilla Output of cat /proc/interrupts on vanilla kernel
Created attachment 97945 [details] dmessg-rh dmessg on rh kernel
Created attachment 97946 [details] dmessg-vanilla dmessg on vanilla kernel
e100 got a complete rewrite in 2.6.3-bk1, looks like we regressed on at least some cards. Can you report this back upstream too please ?
Why does e100-3.x show irq = 169? Looks like we're not getting any interrupts with e100-3.x. What happens with e100-2.3.x against kernel-smp-2.6.3-1.97? Get e100- 2.3.38.tar.gz from http://sf.net/projects/e1000. You'll need kernel source for kernel-smp-2.6.3-1.97 installed to build the driver from tar ball.
Scott: one no longer can build drivers against the kernel-source rpm; all the needed headers are now part of the kernel-smp rpm proper.
Oh, ok, that's good. Are there any updates to ACPI between 2.6.3 and 2.6.3-1.97? That's the first thing I suspect when device isn't getting interrupts to driver. How about is we set 'acpi=off'?
Im out of office for one week about and cannt debug nonworking network driver over network. Thanks for patience. I will test acpi=off and SF driver after return.
Created attachment 98483 [details] boot-acpi-on
Created attachment 98484 [details] boot-acpi-off
Created attachment 98485 [details] boot-pci-noacpi
I tryied boot kernel-smp-2.6.3-1.118 (same for 2.6.3-2.1.253smp) with boot options acpion - not working acpioff - working pci=noacpi - working So it looks problem is not in driver itself, but in interrupt management and ACPI. Im going to usage pci=noacpi.
dmesg-rh and interrupts-rh is the failure and dmesg-vanilla and interrupts-vanilla (2.6.3) the e100 works? One big differences is CONFIG_PCI_USE_VECTOR, which is enabled in -rh and disabled on -vanilla. Perhaps you can disable it on -rh and enable it on -vanilla to see if the failure follows that? -vanilla has another problem though -- irq 18: nobody cared! on ide_intr() which may impact your ide disk... Looking at dmesg-vanilla, IRQ18 is set to LEVEL/LOW which perhaps is not what IDE expects... Also > ICH5: 100% native mode on irq 18 I've never seen 100% native mode before My (875) ICH5 system says: ICH5: not 100% native mode: will probe irqs later and IDE comes up on IRQ14 and IRQ15... curious, e100 is in the dmesg for acpi=off, but isn't present in dmesg for acpi=on or (working case) pci=noacpi. What model Motherboard is this -- does it have the latest BIOS? the output from acpidmp would be useful. thanks, -Len
Created attachment 98788 [details] 2.6.4-linus-config
Created attachment 98789 [details] 2.6.4-linux-dmesg
Created attachment 98790 [details] MSI865G-Neo2LS.acpidmp
This mainboard is MSI 865G Neo2-LS http://www.msi.com.tw/program/products/mainboard/mbd/pro_mbd_detail.php?UID=452 I compiled 2.6.4 vanilla kernel with CONFIG_PCI_USE_VECTOR=y and it is working well (except another warnings like APIC error on CPU0: 00(60) and some Call Trace)
Upgrade to kernel-smp-2.6.4-1.298.i686.rpm resolves this issue. Thank you for your hard work. Boys, you are the BEST!