After installation, and the system is rebooted, if there's any traffic on the network whatsoever, the eepro100 driver will constantly display the following error messages: eth0: card reports no RX buffers. eth0: card reports no resources. eth0: card reports no RX buffers. eth0: card reports no resources. ... This is happening on the ProLiant DL 360 and ML 350, which use the Intel 82559 embedded NIC.
This defect is considered MUST-FIX for Winston Beta-5
FYI, on Jul 03, 2000, I contacted Andrey Savochkin regarding this problem with the eepro100 driver. He replied with the following message, but I haven't heard from him since: On Mon, Jul 03, 2000 at 06:27:31PM -0500, John Cagle wrote: > Andrey, > > Please let me know if there's any testing I can help you with. I've got > several different servers that use the eepro100 driver, and would be > happy to help you test it. I've got the hardware where I can reproduce the problem myself. I'll send you a note when I solve it to test the fix. Best regards Andrey V. Savochkin
Compaq Computers considers this defect a must-fix defect for Winston release (per Harry Heinisch on 21-Jul-2000).
This defect has been re-classified as MUST-FIX for Winston Gold-release
Would Red Hat consider including Intel's e100 driver to be used for the specific adapters it supports, and only use eepro100 for all other 8255x adapters?
From the experience in the field with the intel driver I think it would be at best a solution of dubious value. One other option is to use the 2.2.14 driver with that subspecies of card if we can identify it. Unfortunately Intel are not exactly forthcoming with eepro100 docs so it is very hard to debug this chip and handle its errata.
We will, as it turns out, be including the e100 driver, though I don't think it will be turned on by default; it will at least be there to try if the eepro100 driver fails you... Whether it will be a solution or not, we'll find out with more testing.
This is 100% repeatable with Pinstripe on the Compaq ProLiant ML330 server. Is there any way we could automatically use the e100 driver on specific PCI sub-system vendor & device IDs?
I don't think we can change that this late in the game, but I copied Bill as he can tell us for sure. Whether we would want to change still has to be determined, and I think we would need more analysis. Also, we have made a change to the eepro100 driver that may fix this. I'd like you to test the kernel that I put in your private directory on gribble. If you don't know how to get at gribble, let me know.
New kernel works great on the ML330! No errors, and no performance issues! GOOD WORK! I still need to test our other platforms that had this problem, but it's promising...
Great! Thanks for the report, and glad that it appears to fix the problem. Question: have you tried DHCP while doing this testing? The first test here DHCP didn't work but static config did work -- but we realized later that we might have run the DHCP server out of addresses, so if you have already tested that or get a chance to test that, let us know. Thanks!
FYI, I dropped our RC1 kernel in your private gribble dir