Re: Backtraces in mirage?

[Anil Madhavapeddy <anil@xxxxxxxxxx>]

Just to be clear, are you running separate OPAM switches for the different 
versions you are running?  3.12-generated code will not work on the 4.01 
runtime reliably.  Regarding the trace mode, that only works for OCaml code, 
and not for the actual C runtime (which it sounds like is your problem here).  
I'm afraid judicious printfs remain the best option, unless you setup a 
gdbserver environment for the VM.

What's actually involved in suspending a PV guest from a low-level perspective? 
Do you just need to reconstruct your page tables on resume?  If you have the 
3.12.1/Mirage/monolithic patches that used to work, I can eyeball them to see 
if anything has changed in the general area that might affect it on the latest 
tree.

-anil

On 12 Dec 2012, at 14:48, Jon Ludlam <jjl25@xxxxxxxxx> wrote:

> Well, I guess the large mem alloc thing is just the mismatch between the 4.01 
> runtime and the 3.12 generated code. Surprising it got that far :-) now I 
> need a page table debugger :-(
> 
> Sent from my iPad
> 
> On 12 Dec 2012, at 14:06, Jon Ludlam <jjl25@xxxxxxxxx> wrote:
> 
>> Hi Dave, Anil,
>> 
>> I'm trying to get suspend resume stuff working again, and am hitting some 
>> curious problems. In 4.01, I'm getting invalid page tables on suspend, and 
>> on 3.12.1, I'm getting a failure in alloc_pages much earlier than the 
>> suspend - in that case something's trying to allocate an enormous amount of 
>> memory. There used to be a trace/debug mode in the old monolithic repo - is 
>> there anything similar these days? And is there anything that's changed 
>> recently that might have caused any of this?
>> 
>> Cheers,
>> 
>> Jon
>> 
>> Sent from my iPad
>