vge(4) driver not working at 1G speeds
Stuart Fraser
I'm trying to get a Via Velocity Gigabit Networking driver vge(4) to work at
1000baseTX on a Via EPIA SN18000 Motherboard.
It will always Auto negotiation and back-off to 100MB Full duplex, which
works
Perfectly (so far). When I try to manually force the card 1000baseTX the
interface
status reports 'no carrier' and I can get not connectivity. I have tried
Ubuntu live disk and it works seamlessly to 1GB, so I'm sure the
cabling/switch etc is fine and the issue is with the driver.
So I have tried 1000baseTX full and half duplex. I have tried to disable the
hardware handoff using the mediaopt commands -rxcsum and -txcsum. Further I
have tried to use the link0 operation as described in the man page but I
get:
nas# ifconfig vge0 mediaopt link0
ifconfig: SIOCSIFMEDIA (media): Device not configured
Further I have tried Release 7.0 and Stable 7 with the vge driver compiled
In to the kernel also as an external module loaded through boot/loader.conf
and lastly waiting until a full boot has completed and manually loading the
if_vge module.
The only thing I can see is that the chipset is actually a VT6130 and
perhaps the driver needs 'tweaked' or there is some sysctl somewhere to poke
it into life. As I say it works fine in 100MB mode and I can use 100M for a
while but I'm planning on using this as a NAS so 1G would be preferred.
Any help, suggestions, thoughts would be appreciated. Anyone got it working
or tried?
Should I report it as a bug?
Rgds
S.
_______________________________________________
freebsd-...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "freebsd-questi...@freebsd.org"
_______________________________________________
freebsd...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-curre...@freebsd.org"
Pyun YongHyeon
Would you show me verbosed boot message?
Old VIA VT6120/VT6122 controllers used to ciphy(4) but newer PCIe
Velocity gigabit controllers, VT6130/VT6132, might use other
variants of the PHY.
--
Regards,
Pyun YongHyeon
Stuart Fraser
Here is the dmesg section, it appears to be using UKPHY so as you say if the
older chips used a different physical layer perhaps we have a problem in
there.
-- dmesg --
pci3: <ACPI PCI bus> on pcib3
vge0: <VIA Networking Gigabit Ethernet> port 0xe800-0xe8ff mem
0xfeaffc00-0xfeaffcff irq 28 at device 0.0 on pci3
miibus0: <MII bus> on vge0
ukphy0: <Generic IEEE 802.3u media interface> PHY 22 on miibus0
ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
1000baseT-FDX, auto
vge0: using obsoleted if_watchdog interface
vge0: Ethernet address: 00:40:63:f3:7e:28
vge0: [ITHREAD]
rgds
Stu
Pyun YongHyeon
> Hi,
> older chips used a different physical layer perhaps we have a problem in
> there.
>
> -- dmesg --
> pci3: <ACPI PCI bus> on pcib3
> vge0: <VIA Networking Gigabit Ethernet> port 0xe800-0xe8ff mem
> 0xfeaffc00-0xfeaffcff irq 28 at device 0.0 on pci3
> miibus0: <MII bus> on vge0
> ukphy0: <Generic IEEE 802.3u media interface> PHY 22 on miibus0
This is the reason why I requested verbosed boot messages.
ukphy(4) prints OUI/model number of the PHY in verbosed boot mode.
Would you show me that information?
> ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
> 1000baseT-FDX, auto
> vge0: using obsoleted if_watchdog interface
> vge0: Ethernet address: 00:40:63:f3:7e:28
> vge0: [ITHREAD]
>
--
Stuart Fraser
pcib3: slot 0 INTA hardwired to IRQ 28
vge0: <VIA Networking Gigabit Ethernet> port 0xe800-0xe8ff mem
0xfeaffc00-0xfeaffcff irq 28 at device 0.0 on pci3
vge0: Reserved 0x100 bytes for rid 0x14 type 3 at 0xfeaffc00
miibus0: <MII bus> on vge0
ukphy0: <Generic IEEE 802.3u media interface> PHY 22 on miibus0
ukphy0: OUI 0x0090c3, model 0x0019, rev. 0
ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
1000baseT-FDX, auto
vge0: using obsoleted if_watchdog interface
vge0: bpf attached
vge0: Ethernet address: 00:40:63:f3:7e:28
ioapic1: routing intpin 4 (PCI IRQ 28) to vector 49
vge0: [MPSAFE]
vge0: [ITHREAD]
ta
S
-----Original Message-----
From: owner-free...@freebsd.org
[mailto:owner-free...@freebsd.org] On Behalf Of Pyun YongHyeon
Sent: 02 April 2008 09:17
To: Stuart Fraser
Cc: freebsd...@freebsd.org
Subject: Re: vge(4) driver not working at 1G speeds
Pyun YongHyeon
--zhXaljGHf11kAtnf
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
On Wed, Apr 02, 2008 at 09:30:40AM +0100, Stuart Fraser wrote:
> Oops, no coffee yet, here you go.
>
> pcib3: slot 0 INTA hardwired to IRQ 28
> vge0: <VIA Networking Gigabit Ethernet> port 0xe800-0xe8ff mem
> 0xfeaffc00-0xfeaffcff irq 28 at device 0.0 on pci3
> vge0: Reserved 0x100 bytes for rid 0x14 type 3 at 0xfeaffc00
> miibus0: <MII bus> on vge0
> ukphy0: <Generic IEEE 802.3u media interface> PHY 22 on miibus0
> ukphy0: OUI 0x0090c3, model 0x0019, rev. 0
> ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
> 1000baseT-FDX, auto
> vge0: using obsoleted if_watchdog interface
> vge0: bpf attached
> vge0: Ethernet address: 00:40:63:f3:7e:28
> ioapic1: routing intpin 4 (PCI IRQ 28) to vector 49
> vge0: [MPSAFE]
> vge0: [ITHREAD]
>
Thanks a lot! Try attached patch and let me know how it goes.
--
Regards,
Pyun YongHyeon
--zhXaljGHf11kAtnf
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="ip1000phy.patch"
--- sys/dev/mii/miidevs.orig 2008-03-07 06:42:48.000000000 +0900
+++ sys/dev/mii/miidevs 2008-04-02 17:48:19.000000000 +0900
@@ -51,6 +51,7 @@
oui ALTIMA 0x0010a9 Altima Communications
oui AMD 0x00001a Advanced Micro Devices
+oui ATHEROS 0x001374 Atheros Communications
oui BROADCOM 0x001018 Broadcom Corporation
oui CICADA 0x0003F1 Cicada Semiconductor
oui DAVICOM 0x00606e Davicom Semiconductor
@@ -112,6 +113,9 @@
model AMD 79c978 0x0039 Am79c978 HomePNA PHY
model xxAMD 79C873 0x0000 Am79C873/DM9101 10/100 media interface
+/* Atheros Communucations/Attansic PHYs. */
+model ATHEROS F1 0x0001 Atheros F1 10/100/1000 media interface
+
/* Broadcom Corp. PHYs. */
model BROADCOM 3C905B 0x0012 3c905B 10/100 internal PHY
model BROADCOM 3C905C 0x0017 3c905C 10/100 internal PHY
@@ -159,6 +163,7 @@
/* IC Plus Corp. PHYs */
model ICPLUS IP101 0x0005 IC Plus 10/100 PHY
model ICPLUS IP1000A 0x0008 IC Plus 10/100/1000 media interface
+model ICPLUS IP1001 0x0019 IC Plus IP1001 10/100/1000 media interface
/* Intel PHYs */
model xxINTEL I82553AB 0x0000 i83553 10/100 media interface
--- sys/dev/mii/ip1000phy.c.orig 2006-12-03 00:32:33.000000000 +0900
+++ sys/dev/mii/ip1000phy.c 2008-04-02 17:48:45.000000000 +0900
@@ -82,6 +82,7 @@
static const struct mii_phydesc ip1000phys[] = {
MII_PHY_DESC(ICPLUS, IP1000A),
+ MII_PHY_DESC(ICPLUS, IP1001),
MII_PHY_END
};
--zhXaljGHf11kAtnf
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
_______________________________________________
freebsd...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-curre...@freebsd.org"
--zhXaljGHf11kAtnf--
Stuart Fraser
The patch files didn't work at first but a did it by hand adding the
relevant lines. It compiled and installed fine.
So what we see is that things have changed but are not yet working, if not
actually worse. Here are my findings (all using the vge driver compiled into
kernel)
-first we fail if ifconfig media is set to auto, ifconfig reports media
Ethernet autoselect (none), status: active. No difference if LINK0 set.
-100BaseTX-FDX appears not to send, if I try to ping any other than the
local IP I 'ping: sendto: No buffer space available'. If I tcpdump on the
interface I see the traffic I would expect on the local subnet.
-1000BaseTX, still no carrier
The detailed dmesg snip shows that ip1000phy did connect rather than ukphy
pcib3: slot 0 INTA hardwired to IRQ 28
vge0: <VIA Networking Gigabit Ethernet> port 0xe800-0xe8ff mem
0xfeaffc00-0xfeaffcff irq 28 at device 0
.0 on pci3
vge0: Reserved 0x100 bytes for rid 0x14 type 3 at 0xfeaffc00
miibus0: <MII bus> on vge0
ip1000phy0: <IC Plus IP1001 10/100/1000 media interface> PHY 22 on miibus0
ip1000phy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseTX,
1000baseTX-FDX, auto
vge0: using obsoleted if_watchdog interface
vge0: bpf attached
vge0: Ethernet address: 00:40:63:f3:7e:28
ioapic1: routing intpin 4 (PCI IRQ 28) to vector 49
vge0: [MPSAFE]
vge0: [ITHREAD]
thanks again for the help, what next?
S
-----Original Message-----
From: owner-free...@freebsd.org
[mailto:owner-free...@freebsd.org] On Behalf Of Pyun YongHyeon
Sent: 02 April 2008 09:52
To: Stuart Fraser
Cc: freebsd...@freebsd.org
Subject: Re: vge(4) driver not working at 1G speeds
On Wed, Apr 02, 2008 at 09:30:40AM +0100, Stuart Fraser wrote:
> Oops, no coffee yet, here you go.
>
> pcib3: slot 0 INTA hardwired to IRQ 28 > vge0: <VIA Networking Gigabit
Ethernet> port 0xe800-0xe8ff mem > 0xfeaffc00-0xfeaffcff irq 28 at device
0.0 on pci3 > vge0: Reserved 0x100 bytes for rid 0x14 type 3 at 0xfeaffc00
> miibus0: <MII bus> on vge0 > ukphy0: <Generic IEEE 802.3u media
interface> PHY 22 on miibus0 > ukphy0: OUI 0x0090c3, model 0x0019, rev. 0
> ukphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, >
1000baseT-FDX, auto > vge0: using obsoleted if_watchdog interface > vge0:
bpf attached > vge0: Ethernet address: 00:40:63:f3:7e:28 > ioapic1:
routing intpin 4 (PCI IRQ 28) to vector 49 > vge0: [MPSAFE] > vge0:
[ITHREAD] >
Thanks a lot! Try attached patch and let me know how it goes.
--
Regards,
Pyun YongHyeon
Pyun YongHyeon
> Afraid, that didn't do it.
>
> The patch files didn't work at first but a did it by hand adding the
> relevant lines. It compiled and installed fine.
>
> So what we see is that things have changed but are not yet working, if not
> actually worse. Here are my findings (all using the vge driver compiled into
> kernel)
>
> -first we fail if ifconfig media is set to auto, ifconfig reports media
> Ethernet autoselect (none), status: active. No difference if LINK0 set.
>
> -100BaseTX-FDX appears not to send, if I try to ping any other than the
> local IP I 'ping: sendto: No buffer space available'. If I tcpdump on the
> interface I see the traffic I would expect on the local subnet.
Since vge(4) failed to detect correct link state you just see ENOBUFS.
>
> -1000BaseTX, still no carrier
>
This should be fixed in vge(4).
> The detailed dmesg snip shows that ip1000phy did connect rather than ukphy
>
> pcib3: slot 0 INTA hardwired to IRQ 28
> vge0: <VIA Networking Gigabit Ethernet> port 0xe800-0xe8ff mem
> 0xfeaffc00-0xfeaffcff irq 28 at device 0
> .0 on pci3
> vge0: Reserved 0x100 bytes for rid 0x14 type 3 at 0xfeaffc00
> miibus0: <MII bus> on vge0
> ip1000phy0: <IC Plus IP1001 10/100/1000 media interface> PHY 22 on miibus0
> ip1000phy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseTX,
> 1000baseTX-FDX, auto
> vge0: using obsoleted if_watchdog interface
> vge0: bpf attached
> vge0: Ethernet address: 00:40:63:f3:7e:28
> ioapic1: routing intpin 4 (PCI IRQ 28) to vector 49
> vge0: [MPSAFE]
> vge0: [ITHREAD]
>
>
> thanks again for the help, what next?
>
Maybe the next thing is try to fix bugs in vge(4). But I guess it
would take very long time as it needs more clean up.
I also have a old vge(4) hardware that exhibits the same issue you've
seen. I guess I wrote very minimal patch to make it detect link state
but can't find it. I'll let you know if I find that old patch in my box.
Josep Pujadas i Jubany
Its seems similar to my SOLVED problem with bge.
Please see at http://lists.freebsd.org/pipermail/freebsd-current/2008-
April/084628.html
Regards,
Josep Pujadas