DCHP fail w/ Intel i210 lan port in Boradwell-EP platform

[dennis.huang]

Hi ,
We see a strange behavior w/ ipxe in Intel Broadwell -EP platform (PCH :8900).
We have 2 I210 lan w/ NCSI , one is 5:0:0(port 1) another one is 6:0:0(port 2).
We can get dhcp ip from dhcp server through port 1 , but port 2 can't get dhcp ip from dhcp server.
In Port 2 , ipxe will try to do dhcp discovery and cause timeout w/ debug mode .(error code :28086011 and 040ee119)
When I use the tftp server to get pxelinux.0 and boot w/o IPXE , both ports can download image and remote boot correctly.

(my dhcp server is Centos 6.6)
I also try to use other platform (Broadwell-DE soc) w/ 2 I210 lan w/ NCSI (8086:1533) , one is 8:0:0(port 1) another one is

9:0:0(port 2).
Both ports are workable for boot w/ ipxe (dhcp action can be finished).
I also find other Broadwell-EP board w/ I210 (PCH:C610),the issue still happen.
When I find a Broadwell-EP board w/ I350 (PCH:C610 ), I can boot correctly .
(PCIE bdf is 8:0:2 and 8:03)

May I have your suggestion or comment ? or someone also saw the same issue before .

Thanks!
B/R
Dennis

===I210 Port 1(Failed port)=========================================

CLIENT IP: 192.168.0.31 MASK: 255.255.255.0 DHCP IP: 192.168.0.202
PXE->EB: !PXE at 93B4:0070, entry point at 93B4:0106
UNDI code segment 93B4:5F20, data segment 8D7C:6380 (565-615kB)
UNDI device is PCI 06:00.0, type DIX+802.3
565kB free base memory after PXE unload
iPXE initialising devices...ok


iPXE 1.0.0+ (93665) -- Open Source Network Boot Firmware -- http://ipxe.org
Features: DNS HTTP iSCSI TFTP AoE ELF MBOOT PXE bzImage Menu PXEXT
net0: 74:fe:48:09:1d4 using undionly on

0000:06:00.0 (open) [Link:up, TX:0 TXE:1 RX:0 RXE:0]

[TXE: 1 x "Network unreachable (http://ipxe.org/28086011)"] DHCP 0x1e614 entering discovery state

Configuring (net0 74:fe:48:09:1d4)...DHCP 0x1e614 DHCPDISCOVER .DHCP

0x1e614 DHCPDISCOVER .DHCP 0x1e614 DHCPDISCOVER

....DHCP 0x1e614 DHCPDISCOVER ........ No

configuration methods succeeded (http://ipxe.org/040ee119) No more network devices


pci/
builtin/
smbios/
netX/
net0/
busid.......... 01:80:86:15:33
busloc......... 0x600
bustype........ PCI
chip........... undionly
ifname......... net0 mac............ 74:fe:48:09:1d4

mtu............ <not specified>

gateway........ <not specified> ip............. 192.168.0.31

netmask........ 255.255.255.0

dns............ <not specified> ...

===I210 Port 2(workable port)=========================================

CLIENT MAC ADDR: 74 FE 48 09 1D D6 GUID: 8804A5A0 8120 11E4 966F 0202A6D8CD24
CLIENT IP: 192.168.0.48 MASK: 255.255.255.0 DHCP IP: 192.168.0.202 PXE->EB: !PXE at 93B4:0070, entry point at

93B4:0106 UNDI code segment 93B4:5F20, data segment 8D7C:6380 (565-615kB)

UNDI device is PCI 05:00.0, type DIX+802.3 565kB free base memory after PXE unload

iPXE initialising devices...ok


iPXE> dhcp DHCP 0x1e674 entering discovery state

Configuring (net0 74:fe:48:09:1d6)...DHCP 0x1e674 DHCPDISCOVER .DHCP

0x1e674 DHCPDISCOVER DHCP 0x1e674 DHCPOFFER from 192.168.0.202:67 for

192.168.0.35 .DHCP 0x1e674 entering request state DHCP 0x1e674

DHCPREQUEST to 192.168.0.202:67 for 192.168.0.35 .DHCP 0x1e674 DHCPACK from 192.168.0.202:67 for

192.168.0.35 ok iPXE>

pci/
builtin/
smbios/
netX/
net0/
busid.......... 01:80:86:15:33
busloc......... 0x500
bustype........ PCI
chip........... undionly
ifname......... net0 mac............ 74:fe:48:09:1d6

mtu............ <not specified>

gateway........ <not specified> ip............. 192.168.0.35

netmask........ 255.255.255.0

dns............ <not specified>

[NiKiZe]

you are using undionly.kpxe correct?

Have you tried ipxe.pxe instead?

When using undionly driver only the port on a multiport nic that ipxe (undionly.kpxe) was chained via will be available (this supposedly is a limitation of the UNDI stack)

[dennis.huang]

(2017-08-03 13:29)NiKiZe Wrote:  you are using undionly.kpxe correct?

Have you tried ipxe.pxe instead?

When using undionly driver only the port on a multiport nic that ipxe (undionly.kpxe) was chained via will be available (this supposedly is a limitation of the UNDI stack)

Yes , I use the undionly.kpxe in original test.

I try to replace the Intel pxe rom to ipxe rom in BIOS rom file base on it is a on-board chipset today.
And I can boot successfully from 1st I210(5:0:0) ,but still stop on dhcp discovery from 2st I210(6:0:0).
It seems get the same result from ipxe rom.
And Sorry to make confuse about i210 port 1 and 2, we use 2 I210 x1 (1 port) on board lan as our management port.
1st port is 5:0:0,and 2nd port is 6:0:0. (that is why I said port 1 and port 2)

May I have your idea or comment about this issue?

Thanks!
B/R
Dennis
=============IPXE info w/ ipxe.rom in BIOS ========================

iPXE (http://ipxe.org) 05:00.0 2E00 PCI3.00 PnP PMM+21766000+21777000 C800
Press Ctrl-B to configure iPXE (PCI 05:00.0)...

iPXE (http://ipxe.org) 06:00.0 2E00 PCI3.00 PnP PMM 20A05000 20A16000 C900

===============2nd I210 port============================
0000:06:00.0 (open) [Link:up, TX:0 TXE:1 RX:0 RXE:0]

[TXE: 1 x "Network unreachable (http://ipxe.org/28086011)"] DHCP 0x1e614 entering discovery state

Configuring (net0 74:fe:48:09:1dBig Grin4)...DHCP 0x1e614 DHCPDISCOVER .DHCP

0x1e614 DHCPDISCOVER .DHCP 0x1e614 DHCPDISCOVER

....DHCP 0x1e614 DHCPDISCOVER ........ No

configuration methods succeeded (http://ipxe.org/040ee119) No more network devices

[NiKiZe]

Why are you trying to replace the ROM?
could you just boot ipxe.pxe, ipxe.lkrn or ipxe.iso and see that it works before trying to replace the existing ROM with intel.rom ?

[dennis.huang]

Because it is my first use ipxe to boot. From the original cognitive point of view, change pxe rom may be directly from ipxe.pxe boot.
Later found to be in tftp path replace undionly.kpxe to ipxe.pxe.
After I change to ipxe.pxe in my server site, both I210 ports is workable in my platform through ipxe.
Thank for your help.