[realtek] CNF401 CardBus card : realtek_cb reboots

[Donald Becker]

On Sun, 16 Jun 2002, The Black Unicorn wrote:

> I bought a CNET SinglePoint CNF401 Cardbus card a few days ago. It came
> with a modified rtl8139.c with cardbus support.

What was the base version?

> eth0: Abnormal interrupt, status 00000011
> eth0: Abnormal interrupt, status 00000021
> And after a few of those, the system reboots.

That should never happen!

> Some digging into the code tells me that
> 0011 = RxOverflow (0x10) & RxOK (0x01)
> 0021 = RxUnderrun (0x20) & RxOK (0x01)
> The page on scyld.com says
> 0x0010 : Rx Buffer overflow, set when the Rx ring overflows due
> to high interrupt latency
> 0x0020 : Rx packet underrun: this should never occur

I'm curious about the PCI parameters that could cause a Rx underrun.
There must be a real problem with the CardBus-PCI bridge bandwidth, or
the driver parameters were changed.

> Jun 16 01:41:01 orbit kernel: eth0: interrupt  status=0x0000 new
> intstat=0x0000.Jun 16 01:41:01 orbit kernel: eth0: exiting interrupt,
> intr_status=0x0000.

You have the debug level set really high to see these!
Logging so many messages puts a real load on the disk I/O.

Note that the current driver allows selecting messages by type.

> The next couple of thousand lines (about 8000 actually) in syslog are
> unreadable and seem to be dumped binary code (raw packets ??)

The fix 16 bytes of the packet header... don't turn the debug level up
so high for tracking down transfer errors.

> Does anyone have any idea of what could cause this ? Or, as an
> alternative, could anyone tell me where to find a 2.4 driver for
> realtek_cb and everything that goes with it ?

http://www.scyld.com/network/updates.html

-- 
Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993