kern/53776: wm(4) goes catatonic during large transfers

[Hauke Fath <hf%spg.tu-darmstadt.de@localhost>]

>Number:         53776
>Category:       kern
>Synopsis:       wm(4) goes catatonic during large transfers
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Dec 11 14:15:00 +0000 2018
>Originator:     Hauke Fath
>Release:        NetBSD 8.99.26
>Organization:
Technische Universitaet Darmstadt
>Environment:
	
	
System: NetBSD Gstoder 8.99.26 NetBSD 8.99.26 (GA-MA770-UD3-$Revision$) #0: Fri Nov 23 16:14:26 CET 2018 hf@Hochstuhl:/var/obj/netbsd-builds/developer/amd64/sys/arch/amd64/compile/GA-MA770-UD3 amd64
Architecture: x86_64
Machine: amd64
>Description:

	On an older AMD machine, an intel gbit ethernet interface

wm0 at pci2 dev 7 function 0: Intel i82541PI 1000BASE-T Ethernet (rev. 0x05)
wm0: interrupting at ioapic0 pin 21
wm0: 32-bit 33MHz PCI bus
wm0: 64 words (6 address bits) MicroWire EEPROM
wm0: Ethernet address 00:0e:0c:d8:3b:df
wm0: 0x220402<LOCK_EECD,IOH_VALID,ASF_FIRM,WOL>
igphy0 at wm0 phy 1: Intel IGP01E1000 Gigabit PHY, rev. 0
igphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 1000baseT-FDX, auto

	goes catatonic during large network transfers - typically pkg
	updates of firefox, rust, clang, or scp'ing an iso image.

	Since the machine uses YP and NFS, it becomes unusable, and
	the only way out is a hard reset. Log entries from dhcpcd(8)
	indicate that the interface is flapping:

/netbsd: [ 288847.9198854] wm0: device timeout (txfree 3968 txsfree 0 txnext 3031)
/netbsd: [ 288848.5300993] wm0: link state DOWN (was UP)
syslogd[342]: last message repeated 2 times
dhcpcd[303]: wm0: carrier lost

	Swapping out the NIC against another, known good intel card
	does not make a difference; the cards work fine in other
	(linux) machines.

	
>How-To-Repeat:

	Transfer multi-megabyte data over the above wm(4) NIC.

	
>Fix:
	Yes, please.

	

>Unformatted: