[Qemu-devel] Re: SoftMMU address tracing

[Clemens Kolbitsch]

hi!
finally found out why it did not work ... the problem was kqemu after all.
after compiling qemu with kqemu-support disabled, the address 
translation/access worked as expected :-)

is there a good documentation/technical paper online that could help me 
understand how address handling is done using kqemu?

thanks!!


> hi!
> aaaaah... i'm going crazy... could someone help me please:
>
> i'm still (as previously posted on irc) trying to catch when the guest OS
> (or a program running in it) accesses a certain virtual address (e.g. reads
> from it).
>
> my guest code is something like:
>
> ...
> unsigned long *p = (unsigned long*)0xdeadbeef;
> unsigned long l = *p;
> ...
>
> now, i know it is compiled into something like this:
>
> ...
> movl $0xdeadbeef,0xfffffff8(%ebp)
> mov  0xfffffff8(%ebp),%eax
> mov  (%eax),%eax
> ...
>
> what will be translated by qemu into something like
>
> ...
> movl_A0_EAX
> ldl_kernel_T0_A0
> movl_EAX_T0
> ...
>
> I hope that that's correct up to here (i'm not quite sure about the
> ldl_kernel as it could also be ldl_user!!). but... as
> ldl_(kernel|user)_T0_A0 simply calls the ld-functions in softmmu_header i
> inserted a line into these functions that simply dumps out the currently
> accessed address (i.e. the parameter ptr) to stdout.
>
> i run qemu and grep for my "deadbeef" output, but it NEVER occurs!! how can
> that be?? I've been trying for hours now, could someone please help me
> here??
>
> don't know if that makes a difference but i've tried using -kernel-kqemu
> and also without kqemu.
>
> any help is greatly appreciated ... thanks!!