Hi, all.
On 2015/01/09 4:53, Sverre Froyen wrote:
On 2015-01-07, at 15:31, Sverre Froyen <[email protected]> wrote:
On 2015-01-06, at 21:14, Masanobu SAITOH <[email protected]> wrote:
Hello, Sverre.
On 2015/01/06 5:33, Sverre Froyen wrote:
Hi,
I was performing some network maintenance and noticed that after disconnecting
and reconnecting bge/brgphy interfaces, the interfaces are stuck in a “no
carrier” state. I’m pretty sure this used not to be case but I have no idea how
long ago that was. It appears that the only way to revive a stuck interface it
to reboot the computer.
Here’s some data for the computer:
NetBSD 6.1_STABLE (VM) #12: Fri Jan 2 09:32:36 MST 2015
Dell Computer Corporation PowerEdge 2650
bge0 at pci1 dev 6 function 0: Broadcom BCM5701 Gigabit Ethernet
bge0: interrupting at ioapic1 pin 12
bge0: HW config 00000019, 00000000, 00000000, 00000000 00000000
bge0: ASIC BCM5701 B5 (0x105), Ethernet address 00:06:5b:fd:79:a6
brgphy0 at bge0 phy 1: BCM5701 1000BASE-T media interface, rev. 0
brgphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
1000baseT-FDX, auto
The interface is connected (twisted pair) to a 100Mb/s switch (also Dell).
The sources were updated 2 Jan.
ifconfig output (in the running state):
bge0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
capabilities=3f00<IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx,TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx>
enabled=3f00<IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx,TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx>
address: xx:xx:xx:xx:xx:xx
media: Ethernet autoselect (100baseTX
full-duplex,flowcontrol,rxpause,txpause)
status: active
inet xxx.xxx.xxx.xxx netmask 0xffffffc0 broadcast xxx.xxx.xxx.xxx
inet6 fe80::206:5bff:fefd:79a6%bge0 prefixlen 64 scopeid 0x1
I believe this also happens with bnx/brgphy interfaces (on amd64).
Any suggestions?
It might be my fault. If the problem is caused by brgphy, one of
a way to debug is to comment out brgphy entry from your kernel
configuration file to use ukphy. Could you try it?
Thanks.
Booting a kernel without brgphy restores the expected behavior (status returns
to active when cable reattached). Here’s the corresponding dmsg output:
bge0 at pci1 dev 6 function 0: Broadcom BCM5703X Gigabit Ethernet
bge0: interrupting at ioapic1 pin 12
bge0: HW config 00000115, 00000000, 00000000, 00000000 00000000
bge0: ASIC BCM5702/5703 A2 (0x1002), Ethernet address 00:0f:1f:6e:f3:2d
ukphy0 at bge0 phy 1: OUI 0x001018, model 0x0016, rev. 2
ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
1000baseT-FDX, auto
(I tested this on a another computer but I verified that its brgphy behavior
was the same as the one I used above.)
I also dug some old kernels out of my backups. It looks like a kernel from
2013-08-06 works OK and one from 2013-11-28 fails. Sources were cvs updated
from the netbsd-6 branch on or shortly before those dates.
Another datapoint:
The problem does not exist with NetBSD 7.0_BETA (GENERIC.201501071540Z) i386.
This change was cause by a problem that brgphy.c rev. 1.64 was not pulled up
and other changes after 1.64 were pulled up. I sent a pullup request to fix it.
Thanks Sverre Froyen who reported it and helped debugging with me.
--
-----------------------------------------------
SAITOH Masanobu ([email protected]
[email protected])
