Bug 10885
Summary: | CK804 Ethernet Controller (rev a3) failure | ||
---|---|---|---|
Product: | Drivers | Reporter: | Karen Shaeffer (shaeffer) |
Component: | Network | Assignee: | drivers_network (drivers_network) |
Status: | CLOSED OBSOLETE | ||
Severity: | blocking | CC: | aabdulla, alan, devzero, ljr, sebastian, shaeffer |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.23.12 | Subsystem: | |
Regression: | No | Bisected commit-id: |
Description
Karen Shaeffer
2008-06-07 14:43:34 UTC
One more detail. I personally reproduced the NIC mismatch and the NIC TX failure many times in the lab on 5 different Netra servers, using quite a few different kernel.org and RHEL kernels. It happened on every kernel I tested, with 32 and 64 bit compiles. And it was produced in a data center using the 2.6.22.10 kernel several times completely independent of me. For more details, please see the bug ID 10885. And I need to clarify that the last kernel I tested this on was actually linux-2.6.24-rc8-git6. I mistated in the bug that I observed this failure on the 2.6.25.4 kernel. That is inaccurate. I don't know, if it exist in the 2.6.25.4 kernel, because I never tested that kernel. My error. could you test if how that nic behaves with either "irqpoll" or "noapic acpi=off" ? (see http://bugzilla.kernel.org/show_bug.cgi?id=9015 - maybe related) Can you try the latest ethtool? I recall there was an issue with older ethtool that would not send down the correct settings to nic driver. For info, the problem appears also with last RHEL4 kernels also (2.6.9-78). And maybe before... forcedeth module versions 0.60 & 0.61 have the problem. Forcedeth module fails during big transfers after some seconds IF static configuration is set on switch (no autoneg, Full duplex) and autoneg is set on forcedeth card. In this cas, duplex can not be negotiate and eth falls back to 100 half duplex. In other cases, the transfers is done w/o problem. If transfer stalled, no more traffic can be done and network must be restarted. dmesg : ../.. forcedeth: Reverse Engineered nForce ethernet driver. Version 0.61. ACPI: PCI Interrupt 0000:00:14.0[A] -> GSI 21 (level, low) -> IRQ 201 PCI: Setting latency timer of device 0000:00:14.0 to 64 divert: allocating divert_blk for eth0 ../.. forcedeth 0000:00:14.0: ifname eth0, PHY OUI 0x1c1 @ 0, addr 00:19:db:44:b6:b8 forcedeth 0000:00:14.0: highdma pwrctl timirq gbit lnktim desc-v3 ../.. nv_stop_tx: TransmitterStatus remained busy<6>eth0: link down. nv_stop_tx: TransmitterStatus remained busy<6>eth0: link up. nv_stop_tx: TransmitterStatus remained busy ../.. As requested, "irqpoll" and "noapic acpi=off" options change nothing. HW config: NEC POWERMATE_VL360 C51MCP51 AMD3800+ 00:14.0 Bridge: nVidia Corporation MCP51 Ethernet Controller (rev a3) Behaviour with different HW: Problem does not appear with TG3 and E1000 modules. PS: restart network needs MACADDRESS field to be set to HWADDRESS in ifcfg-eth0 config file (RHEL^h^h^h^hLSB ;-)) to avoid reverse numbering address problem. Please re-open if seen on a modern (2.6.32+) kernel |