Re: kern/40345: nfe: watchdog timeout

[David Young <dyoung%pobox.com@localhost>]

The following reply was made to PR kern/40345; it has been noted by GNATS.

From: David Young <dyoung%pobox.com@localhost>
To: Roman Kraevskiy <rkraevskiy%gmail.com@localhost>
Cc: Christoph_Egger%gmx.de@localhost, kern-bug-people%netbsd.org@localhost,
        gnats-bugs%netbsd.org@localhost, netbsd-bugs%netbsd.org@localhost
Subject: Re: kern/40345: nfe: watchdog timeout
Date: Fri, 9 Jan 2009 11:25:00 -0600

 On Fri, Jan 09, 2009 at 05:24:09PM +0200, Roman Kraevskiy wrote:
 > >  Give this a try:
 > >
 > >  disable ciphy in userconf and check if ukphy attaches on nfe0.
 > >  If it does, test if you still have watchdog timeouts or not.
 > 
 > nfe0 at pci0 dev 10 function 0: vendor 0x10de product 0x0057 (rev. 0xa3)
 > nfe0: interrupting at ioapic0 pin 21
 > nfe0: Ethernet address 00:14:2a:7f:ba:ea
 > ukphy1 at nfe0 phy 17: Generic IEEE 802.3u media interface
 > ukphy1: OUI 0x00c08f, model 0x0001, rev. 3
 > ukphy1: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
 > 1000baseT-FDX, auto
 > 
 > Nothing changes: "nfe0: watchdog timeout" on network activity.
 > 
 > >  If you still have watchdog timeouts, then the problem is nfe0.
 > >  If there are no more watchdog timeouts, then the problem is ciphy0.
 
 nfe_intr() is bracketed by
 
         NFE_WRITE(sc, NFE_IRQ_MASK, 0);
 
 and
 
         NFE_WRITE(sc, NFE_IRQ_MASK, NFE_IRQ_WANTED);
 
 That should not be necessary.  It may be harmful.  I suggest moving
 those lines out of nfe_intr() and into nfe_init(), where they should
 bracket the nfe_intr() call.
 
 Dave
 
 -- 
 David Young             OJC Technologies
 dyoung%ojctech.com@localhost      Urbana, IL * (217) 278-3933