Error 0x420c6095 on default selection

[duca]

Hi,

We have run into an issue we need some help with.
We have created a iPXE script that looks like this:

Code:

goto start

:start
menu iPXE boot menu
item --gap --           ------------------------- Operating systems ------------------------------
item vmlcus             Boot Vmware Esxi 6 installer
item --gap --           ------------------------- Advanced options -------------------------------
item config             Configure settings
item shell              Drop to iPXE shell
item reboot             Reboot computer
item --gap --           --------------------------------------------------------------------------
item exit               Exit iPXE and continue BIOS boot
choose --default vmlcus --timeout 5000 target && goto ${target}

:vmlcus
boot http://myserver.domain.com/esxi/mboot.efi ks=${testfile}

When we select the “vmlcus” menu item manually everything works perfect, but if we let the menu time out and it select the “vmlcus” menu item, we get this error:
http://i05652.dsv.com/esxi/mboot.efi... Error 0x420x6095 (http://ipxe.org/420c6095)
Could not boot image: Error 0x4206095 (http://ipxe.org/420c6095)

Why does it fail if the boot sequence a selected automatically by the time out?
Btw. We have read the 420c6095 article, but still it works if we select the item manually).[/quote]

[NiKiZe]

(2016-04-01 11:29)duca Wrote:  

Code:

menu iPXE boot menu
item vmlcus             Boot Vmware Esxi 6 installer
...
choose --default vmlcus --timeout 5000 target && goto ${target}

:vmlcus
boot http://myserver.domain.com/esxi/mboot.efi ks=${testfile}

if we let the menu time out and it select the “vmlcus” menu item, we get Error 0x4206095 (http://ipxe.org/420c6095)

If I read the source correctly you get that error due to the http server closing the connection.

I would add an debug exit in case of error so the issue can be debugged.

Code:

boot http://myserver.domain.com/esxi/mboot.efi ks=${testfile} || shell

ifstat would be interesting to see both before the error and also after the error have happened.

My current suspicion is that the network have somehow closed the connection.

[duca]

Hi,

Thanks for the quick reply!
If I add the "|| shell" and execute the ifstat, it shows:
net0: 00:90:fa..... 00.0 (closed)
net1: 00:90:fa..... 00.1 (closed)
net2: 00:90:fa:8d.... using NII on NII-PCI51:00.0 (open)
[Link:up, TX:18 TXE:1 RX:8192 RXE:875]
[TXE: 1 x "Error 0x2a086089 (http://ipxe.org/2a086089)"]
[RXE: 658 x "Operation not supported (http://ipxe.org/3c086083)"]
[RXE: 199 x "The socket is not connected (http://ipxe.org/440e6083)"]
net3: 00:90:fa..... 00.1 (closed)

Hope above helps.

[NiKiZe]

Could you try ipxe.efi instead of snponly.efi, just to see if it's maybe an efi firmware issue and not related to ipxe itself. (for example that the efi firmware closes the device after a certain time of inactivity)

Compiling snponly.efi with DEBUG=snpnet:3,nii:3 might reveal something about what is actually happening, In that case maybe use sleep 5000 instead to not have the menu interfere with the output.

Testing different timeouts to narrow down what timeout works and what does not might be helpful in future debugs.

[duca]

Hi,

Thanks once again for the reply!

We have tried to lower the timeout to 3s and now it is working!
I know that it is not a fix but a workaround, but it is good enough for me :o)

Thanks for all your help!

[NiKiZe]

(2016-04-04 08:21)duca Wrote:  We have tried to lower the timeout to 3s and now it is working!
I know that it is not a fix but a workaround, but it is good enough for me :o)

Glad to hear that.

Any chance you could still test my suggestions just to know if this is seen by anyone else or we might even be able to create a fix for it?
Would also be good to know which kind of hardware this is, both the nic and motherboard/efi firmware.

Thanks.

[duca]

Hi,

Of cause, I just didn't want to spend your time if I was the only one with that issue :o)

The server is a Lenovo X3850x6 and the NIC is a Emulex OCe14000 series 10Gbe (OEM).

I tried to compile as suggested (make bin-x86_64-efi/ipxe.efi DEBUG=snpnet:3,nii:3) and boot with the new ipxe.efi, but it just gave a lot debug info when booting, but the menu nor the mboot.efi seems to be loaded - did I do something wrong?

I have tried to find some information regarding ipxe.efi vs snponly.efi, but haven't been able to, do you know where I can find some?

[NiKiZe]

ok, lets go down on the amount of debug to start then:
make bin-x86_64-efi/snponly.efi DEBUG=snpnet,nii
(note using snponly.efi here not ipxe.efi, to minimize possible problem sources and only concentrate on the nii/snp issue)

Assume your menu was shown but got a lot of debug output that got the menu of screen, so maybe just put in a sleep 5 instead, hmm 5000 is milliseconds (my misstake earlier), so waiting 5 seconds fails, but waiting just 3 works? (could you please try your original ipxe with sleep instead of menu in the script?)

snponly.efi only has support for the nii/snp interface which is in the EFI spec i belive, you can see snponly as the efi equivalent of undionly.kpxe, while ipxe.efi have "all" the drivers that ipxe supports, including snp and nii to fallback on. For efi you will probably want ipxe.efi if size is not a limiting factor, the only other case where snponly.efi might be preferred instead of ipxe.efi is if ipxe.efi is causing hangs during boot, but in those cases please report since it clearly is a bug that we would like to get fixed.