Realtek 8102EL

[Pyun YongHyeon]

On Tue, Jul 08, 2008 at 10:42:17AM +0400, Andrey V. Elsukov wrote:
 > Pyun YongHyeon wrote:
 > > > re0: <RealTek 8101E PCIe 10/100baseTX> port 0x2000-0x20ff mem 
 > > > 0x48200000-0x48200
 > > > fff,0x48000000-0x4800ffff irq 16 at device 0.0 on pci1
 > > > re0: Chip rev. 0x24800000
 > >                  ^^^^^^^^^^
 > 
 > Hi, Pyun
 > 
 > Did you look to the last vendors driver? It seems it has
 > different probe code and supports newest cards.
 > 
 > ftp://66.104.77.130/cn/nic/rtl_bsd_drv_v175.tgz
 > 

I already seen this vendor driver. Both rl(4) and re(4) in RELENG_7
use standard PCI probe code. However re(4) has a additional check
to verify the existence of supported chip revision. If we want to
support hardware assitance such as TSO, checksum offload, re(4)
should know exact chip revision as each revision requires different
workaround for its silicon bug.
The vedor driver for FreeBSD ignores all hardware assistance and it
works like dumb controller. Also the vendor driver has a lot of DSP
fixups and magic programming sequence which are really hard to
understand what/why they do that. I'd like to say to check Linux
driver released by the vendor which takes full advantage of
hardware assistance.
I tried to contact the vendor to get programming information for
newer controllers but they rejected to answer my specific
questions. :-(

 > -- 
 > WBR, Andrey V. Elsukov

-- 
Regards,
Pyun YongHyeon