Re: kern/48476: wm(4) transmit hang

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR kern/48476; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: kern/48476: wm(4) transmit hang
Date: Tue, 7 Jan 2014 17:54:50 +0000

 (not sent to gnats)
    ------
 
 From: Manuel Bouyer <bouyer%antioche.eu.org@localhost>
 To: kern-bug-people%NetBSD.org@localhost, gnats-admin%NetBSD.org@localhost, 
netbsd-bugs%NetBSD.org@localhost
 Subject: Re: kern/48476: wm(4) transmit hang
 Date: Tue, 7 Jan 2014 15:24:19 +0100
 
 On Tue, Dec 24, 2013 at 10:20:00AM +0000, bouyer%antioche.eu.org@localhost 
wrote:
 > >Description:
 >      This is on ftp.fr.netbsd.org, a Supermicro X7DBR motherboard.
 >      Its wm0 interface is connected to a gigabit switch and is identified
 >      as:
 > wm0 at pci4 dev 0 function 0: i80003 dual 1000baseT Ethernet (rev. 0x01)
 > wm0: interrupting at ioapic0 pin 18
 > wm0: PCI-Express bus
 > wm0: 65536 word (16 address bits) SPI EEPROM
 > wm0: Ethernet address 00:30:48:31:43:82
 > ikphy0 at wm0 phy 1: i82563 10/100/1000 media interface, rev. 2
 > ikphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 
 > 1000baseT-FDX
 > , auto
 > # ifconfig wm0
 > wm0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
 >         
 > capabilities=7ff80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx,TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx,TCP6CSUM_Rx,TCP6CSUM_Tx,UDP6CSUM_Rx,UDP6CSUM_Tx,TSO6>
 >         
 > enabled=7ff80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx,TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx,TCP6CSUM_Rx,TCP6CSUM_Tx,UDP6CSUM_Rx,UDP6CSUM_Tx,TSO6>
 >         address: 00:30:48:31:43:82
 >         media: Ethernet autoselect (1000baseT full-duplex)
 >         status: active
 > 
 >      under traffic, transmission stops. ping complains with
 >      "no buffer space available" and named with:
 > # Dec 24 11:01:07 antioche named[181]: client 80.10.200.35#32697 
 > (www.antioche.eu.org): error sending response: not enough free resources
 >      ifconfig wm0 doesn't show "OACTIVE flag".
 > 
 >      A "ifconfig wm0 down up" fixes it for a few minutes.
 > 
 >      Note that this system has been used for other tasks before,
 >      with older versions of NetBSD without problems.
 >      Also, other systems with different wm(4) models running
 >      kernels buidls from the same source shows no problems even
 >      under load.
 
 It looks like disabling tso4 and tso6 helps, but I don't know if TSO is the
 real cause of the problem, or if it only hides it by lowering the network
 load a bit.
 
 -- 
 Manuel Bouyer <bouyer%antioche.eu.org@localhost>
      NetBSD: 26 ans d'experience feront toujours la difference
 --