Re: vge(4) driver not working at 1G speeds

On Wed, Apr 02, 2008 at 08:53:25PM +0900, To Stuart Fraser wrote:
 > On Wed, Apr 02, 2008 at 12:32:20PM +0100, Stuart Fraser wrote:
 >  > 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.
 > 

FYI: Support for IC Plus IP1001 PHY was committed to HEAD.

-- 
Regards,
Pyun YongHyeon