Re: pxeboot does not work on Sun X4540

Danny Braniss wrote:
>> Attila Nagy wrote:
>>     
>>> Hello,
>>>
>>> I'm trying to boot FreeBSD 8-CURRENT on a Sun X4540 (2xquad core 
>>> Opteron, 64 GB memory, 48 SATA disks (in fact 47, the 48th is a write 
>>> optimized SSD, because this box is a 7210 OpenStorage "appliance", 
>>> running OpenSolaris) on LSI controllers, nvidia NIC) from our boot 
>>> server without success.
>>>
>>> The PXE on the NIC downloads pxeboot with TFTP, but then gets stuck at 
>>> loading the kernel:
>>> http://people.fsn.hu/~bra/freebsd/20090107-freebsd-x4540/Screenshot-6.png
>>>
>>> pxeboot is configured to load the kernel via NFS (the default) and 
>>> according to the tcpdump output, the client (X4540) issues an NFS 
>>> GETPORT call and gets the response from the NFS server and then 
>>> nothing else happens.
>>>
>>> I've tried with LOADER_TFTP_SUPPORT=YES and the result is the same, 
>>> the only difference is that the machine stops sending packets after 
>>> the first TFTP read request (for /boot/boot.4th.split), which doesn't 
>>> exist, so the server sends the answer and then nothing comes back from 
>>> the X4540 box.
>>>
>>> It seems that it gets to send exactly one packet, then freezes.
>>>
>>> What else should be done to debug this problem?
>>>
>>> Thanks,
>>>       
>> BTW, the last FreeBSD 8-CURRENT snapshot from december boots fine (at 
>> least I can get to sysinstall, I haven't yet tried installation, because 
>> I wouldn't like to touch the boot drives, which house the OpenSolaris 
>> image -that's why I want to boot from the net).
>>
>>     
>
> older pxeboot - pre interrupt handling fixes - work ok, so you can use that 
> instead.
>   
I thought that I did testing with an older version, but as it turned out 
I was wrong.

Confirmed, older versions work fine.

The next problem is that a freshly compiled kernel (HEAD from today) 
panics with this:
http://people.fsn.hu/~bra/freebsd/20090107-freebsd-x4540/Screenshot-55.png

"BIOS smap did not include a basemem segment!"
which is true, according to the above verbose output and taking a quick 
look at the code. (in the loop there is only a base=0x0 and len=0x0 entry)
I'm not an expert of this topic (too), for me these numbers look pretty 
strange.

Maybe Peter can help?

Thanks,