[ipxe-devel] Realtek r8169 (10ec8168.rom) breaks download everytime at 47%

[Andrew Bobulsky]

Hello Michael, Joshua,

I wanted to chime in after I ran some tests a couple days ago.  I have
an x58 board with an i7-920 in it (a Gigabyte board, don't remember
the model off hand), and I modded the BIOS with CBROM32 to insert the
same iPXE ROM into it, and, after much breath holding and finger
crossing, it worked!

I did this some time back, and while I only ran a few tests to just
make sure it worked correctly, and it did.  The only caveat I found at
the time was that the Intel Matrix RAID BIOS seemed to conflict with
iPXE; that is to say that, when the Intel RAID BIOS was enabled, the
iPXE option ROM never loaded up, in spite of the fact that it was
configured to.  I tried two or three different versions of the Intel
ROM, and it didn't make any difference.  The conflict was always
present.  This did not happen with the Realtek PXE ROM, of course.

Coming back to the present, I decided to try out ESXi 5 on that system
booting from iSCSI.  The performance was terrible using the flashed-in
ROM.  It took about an hour to boot ESXi over 100Mbit.  I flashed the
stock ROM back into the BIOS, and went to a chainloaded ipxe.pxe to
see if that would work, but performance was identical.  I can't say
with any certainty which commit the ROM or the .pxe were built from,
but it was at least 6 months ago, and no more than a year ago.

I finally rebuilt undionly.kpxe from the latest git, and chainloaded
that from the Realtek ROM after flashing back to "stock" (in this
case, a BIOS modded just to add VT-d support), and performance was
excellent when booting from SAN.

I didn't try booting the lkrn build like Joshua did, but it seems
counterintuitive to me that it wouldn't have any performance problems
when the ROM and the .pxe build did, but I can try doing so if you
like.

Cheers,
Andrew Bobulsky

P.S. I remember the model of my board now, it's a Gigabyte x58
Extreme: http://www.gigabyte.com/products/product-page.aspx?pid=2957#ov

On Sun, Feb 12, 2012 at 8:32 AM, Michael Brown <mbrown at fensystems.co.uk> wrote:
> On Monday 23 Jan 2012 23:36:52 Joshua C. wrote:
>> Is there a way to make the bios fully initilize the system memory map
>> and start the rom image only if a full memory map is present? I know
>> that intel uses a ctrl+i to start their matrix storage rom, so there
>> should be a way to wait till a full memory map is present, shouldn't
>> it?
>
> Yes; just don't press Ctrl-B at the "Press Ctrl-B to configure iPXE..." prompt.
> Wait until the "Press Ctrl-B for the iPXE command line..." prompt and all will
> be well.
>
> Michael
> _______________________________________________
> ipxe-devel mailing list
> ipxe-devel at lists.ipxe.org
> https://lists.ipxe.org/mailman/listinfo/ipxe-devel