Re: [Qemu-discuss] How do i tell QEMU that memory i've have mapped is code?

[Peter Fillmore]

Awesome!- that got it - i wasn’t aligned to the page so the emulator didn’t 
recognise it of course.

Reason behind it being a that address is that the SoC contains embedded boot 
code which is called by the firmware image -  since i don’t have the full image 
i needed to put some code to emulate a call to that address.

heres the code to copy a block of data into memory for others that may need it:

char iapcode[XXX] = {INSERT DATA HERE};                                         
        //array of data to copy
memory_region_init_ram(iap, NULL, "vivotech.iap", 0x2000,&error_abort);  //init 
iap memory (8k)
vmstate_register_ram_global(iap);                                               
                        //register with the system
memory_region_add_subregion(sysmem, 0x7FFFF000, iap);                   //add 
region to system memory at 0x7FFF F000, make sure your aligned to a page of 
memory
char *ramptr = (char *)memory_region_get_ram_ptr(iap);                          
//get a pointer to the memory
if(ramptr != NULL)
        memcpy(ramptr+0xFF0, iapcode, sizeof(iapcode));                         
//copy data into assigned memory

Thanks for the help again - saved me a ton of time.
-------------------------------------------------------
Peter Fillmore
Principle Consultant, CISSP





On 13 Jan 2015, at 10:48 am, Peter Maydell <address@hidden> wrote:

> On 12 January 2015 at 23:18, Peter Fillmore <address@hidden> wrote:
>> The setup code is <edited for brevity>:
>>        MemoryRegion *sysmem = get_system_memory();
>>        MemoryRegion *iap = g_new(MemoryRegion, 1);                           
>>           //create pointer for region
>>        memory_region_init_ram(iap, NULL, "vivotech.iap", 
>> 0x1000,&error_abort); //init 0x1000 bytes of ram
>>        vmstate_register_ram_global(iap);                                     
>>                                   //register with qemu
>>        memory_region_add_subregion(sysmem, 0x7FFFFFF0, iap);                 
>>   //add to the sysmem
> 
> I'm confused; this doesn't look wrong, but it also doesn't seem
> to have anything at all to do with the code fragment you quoted
> in your previous email as being how you set up the memory.
> 
> Also is it really 0x1000 bytes starting at 0x7ffffff0 ?? That
> sounds very weird -- usually memory is at some aligned address.
> In fact I think this is the problem: we require the whole page
> of the address space to be RAM to be able to execute out of it.
> We can't deal with executing from pages which are split between
> RAM and something else like the setup you have here. (Reading
> or writing from that sort of mixed-usage page is fine; it's just
> executing code that won't work.)
> 
> -- PMM

signature.asc
Description: Message signed with OpenPGP using GPGMail