Re: rge(4) RTL8126 simple testing

On Mon, Apr 06, 2026 at 07:57:32PM -0400, Sean C. Farley wrote:
> 
> On Sun, 29 Mar 2026, Kevin Lo wrote:
> 
> > On Sat, Mar 28, 2026 at 03:12:56PM -0400, Sean C. Farley wrote:
> > > 
> > > On Fri, 27 Mar 2026, Kevin Lo wrote:
> > > 
> > > > On Wed, Mar 04, 2026 at 04:02:37PM -0500, Sean C. Farley wrote:
> > > > > 
> > > > > On Sat, 28 Feb 2026, Sean C. Farley wrote:
> 
> *snip*
> 
> > > > > I have spent awhile trying different things yet cannot get these to
> > > > > work at 2.5G.  The Realtek vendor driver cannot negotiate at 2.5G
> > > > > nor can Linux 6.18 (Arch Linux).
> > > > > 
> > > > > Important details:
> > > > > - FreeBSD dmesg chip rev: RTL8126_2 (0x64a00000)
> > > > 
> > > > Since all the RTL8126 chip revisions I have on hand are RTL8126_1
> > > > (0x64900000), I went ahead and bought the BrosTrend 5G network card you
> > > > recommended.
> > > > Purchasing directly from the BrosTrend official website wasn't an option
> > > > since they don't ship to Taiwan, so I ordered from Amazon [1] instead.
> > > > Unfortunately, the one I received turned out to be RTL8126_1 as well. :(
> > > > 
> > > > [1] https://www.amazon.com/dp/B0CYZ9M3LR
> > > 
> > > I take it that the RTL8126_1 cards you have can negotiate to 2.5G?  That
> > > gives me hope that these can do it too.
> > 
> > Regardless of the chipset, it should have no problem negotiating to 2.5G.
> > If there is an issue, it's definitely a bug.  I haven't encountered any
> > problems so far.
> > 
> > > That is too bad.  I guess that the warehouse for Amazon near you has older
> > > stock.  I appreciate you looking.  It is too bad that Realtek does not
> > > produce reference cards.
> > 
> > I want to know if the RTL8126_2 actually has a bug.
> > Anyway, I'll keep looking into it.
> 
> Thank you.  I will continue to look at it too although I am an amateur with
> hardware drivers.
> 
> Out of curiosity, are my cards actually revision 2?  In the Realtek driver,
> it looks like there are three revisions of the 8126, or is the first version
> something that was a test model?
> - 0x64800000:  re_type = MACFG_90, calls re_set_mac_mcu_8126a_1
>                re_sw_ram_code_ver = NIC_RAMCODE_VERSION_8126A_REV_A
> - 0x64900000:  re_type = MACFG_91, calls re_set_mac_mcu_8126a_2
>                re_sw_ram_code_ver = NIC_RAMCODE_VERSION_8126A_REV_B
> - 0x64A00000:  re_type = MACFG_92, calls re_set_mac_mcu_8126a_3
>                re_sw_ram_code_ver = NIC_RAMCODE_VERSION_8126A_REV_C
> 

Yes, the first revision (0x64800000) is a test model.

> Sean
> -- 
> [email protected]

	Kevin