[eepro100] Ethernet slows and never recovers after heavy bus activity.
Fri Oct 4 12:14:25 2002
I'm running kernel version 2.4.18-rmk7 on an ARM 'Integrator'
development platform. The ethernet card connects to the main system
bus via a V3 PCI bridge chip (and is the only thing on the PCI part
of the bus).
The problem I have is that relatively short (10s of mSec ??) bursts
of intense activity on the main system bus cause the ethernet
performance to degrade significantly and never recover. (I'm not 100%
sure of the HW design, but I suspect that accesses via the PCI bridge
may be blocked during these bursts).
The first symptom is that ping response times from the ARM board
increase from approx 500 uSec to approx the period of the ping
requests. Even without any more bursts of intense activity on the
main bus, the system will stay in this state indefinitely.
If I a few generate more intense bursts of system bus activity, then
I start to see 'eth0: card reports no resources.' messages and some
eepro100.c:v1.09j-t 9/29/99 Donald Becker
eepro100.c: $Revision: 1.36 $ 2000/11/17 Modified by Andrey V.
w.sw.com.sg> and others
eth0: Intel Corp. 82557 [Ethernet Pro 100], 00:A0:C9:0D:FA:E7, IRQ
Receiver lock-up bug exists -- enabling work-around.
Board assembly 352509-003, Physical connectors present: RJ45
Primary interface chip DP83840 PHY #1.
DP83840 specific setup, setting register 23 to 8462.
General self-test: passed.
Serial sub-system self-test: passed.
Internal registers self-test: passed.
ROM checksum self-test: passed (0x49caa8d6).
Receiver lock-up workaround activated.
# cat /dev/net/dev:
Inter-| Receive |
face |bytes packets errs drop fifo frame compressed
multicast|bytes packets errs drop fifo colls carrier compressed
lo: 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
eth0: 2928375 4770 1 0 0 1 0 0
326190 2741 0 0 0 1014 0 0
Do you Yahoo!?
New DSL Internet Access from SBC & Yahoo!