Re: Call for Marvell/SysKonnect Yukon II Gigabit Ethernet testers.

On Mon, Oct 30, 2006 at 08:50:31AM +0100, O. Hartmann wrote:
 > Pyun YongHyeon wrote:
 > >On Sun, Oct 29, 2006 at 10:34:33PM +0100, Philip Paeps wrote:
 > > > The driver seems to work for me, but only when I put it in the kernel 
 > > proper.
 > > > If I try to load it as a module, it gives ENOMEM when trying 
 > > bus_dmamem_alloc()
 > > > for the jumbo buf (line 2177 of if_msk.c).
 > > > 
 > >
 > >If there is no cotiguous DMA memory requried by the driver you have
 > >to reboot to get it. Alternatively, you can put the driver in the
 > >kernel instead of using kernel module as you mentioned.
 > > 
 > > > Two things I'm curious about:
 > > > 
 > > > msk0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
 > > >         options=9a<TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM>
 > > >         inet6 fe80::211:11ff:fe59:1333%msk0 prefixlen 64 scopeid 0x1 
 > > >         inet 172.16.1.2 netmask 0xffffff00 broadcast 172.16.1.255
 > > >         inet6 2001:6f8:32f:10:211:11ff:fe59:1333 prefixlen 64 autoconf 
 > > >         ether 00:11:11:59:13:33
 > > >         media: Ethernet autoselect (100baseTX <full-duplex,flag0,flag1>)
 > > >         status: active
 > > > 
 > > > What are flag0 and flag1 in the media: line?
 > > > 
 > >
 > >AFAIK there is no official way to indicate negotiated Rx/Tx
 > >flow-control status in ifconfig(8). So I used flag0 to indicate
 > >Rx flow-control and flag1 to denote Tx flow-control. It just means
 > >the driver will flow-control(Rx, Tx) with its link-partner(usually
 > >switch).
 > >
 > > > The kernel also prints "mskc0: Uncorrectable PCI Express error" after 
 > > bringing
 > > > the interface up.  It doesn't seem to cause any harm, but it feels 
 > > fishy.  I
 > > > haven't looked at the source for this.
 > > > 
 > >
 > >You can safely ignore the error message unless the error shows
 > >endlessly.
 > >
 > 
 > Is this driver also gonna be used in STABLE/6.2?
 > 

ATM the driver is only for CURRENT. It needs more tests/feedbacks
from users and poor Rx performance issue should be fixed before
hitting src tree. Since RELENG_6 lacks some APIs used on this
driver it also need more changes to use it on RELENG_6.

-- 
Regards,
Pyun YongHyeon