Hi Oliver,
On Thursday, 3. September 2009 17:08:11 Oliver Mayer-Buschmann wrote:
Hi,
I've implemented my own rom service based on Genode::Rom_root and Genode::Rom_fs inside a server. Basically the service seems to work.
If I open a Rom_connection on a file that I've added to my Rom_fs before, I get a valid Dataspace_capability.
By using a Genode::Dataspace_client, I can see that the size and the base address of the Dataspace correspond to the values that I've passed to the Rom_module I've inserted into my Rom_fs before.
The problem is, that the data inside the Dataspace is 0. Here the code:
Dataspace_capability file_cap; Rom_connection rom(filename, unique_name); try { file_cap = rom.dataspace(); } catch (Rom_file_missing) { printf("Error: Could not access file \"%s\" from ROMservice.\n", filename); return 0; }
if(!file_cap.valid()) PWRN("file cap not valid!!"); void* addr = env()->rm_session()->attach(file_cap); Genode::Dataspace_client dsc(file_cap); PDBG("rom module filename: %s", filename); PDBG("rom module base address: 0x%lx", dsc.phys_addr()); PDBG("rom module size: 0x%zx", dsc.size()); PDBG("some data: 0x%x", *(int*)addr);.. and the corresponding output: rom module filename: pci_drv rom module base address: 0x3333000 rom module size: 0x26000 some data: 0x0
My understanding is, that the constructor of Rom_module needs a physical base address.
Well, 'Rom_module' and 'Rom_fs' normally are only core-internal and very simple datastructures filled by core, when it parses information provided by the bootloader (e.g.: boot modules). It's somehow up to you to use these abstractions or to write your own ones. The rom_session protocol doesn't depend on them. Moreover, I guess it's not useful to use them in your case, as long as you don't have all files in virtual memory of your filesystem already, but load them on demand.
Could you please outline in more details, what kind of filesystem you want to provide?
How can I copy data to a memory location, that is still available in another context?
I'm not sure what you mean by that. Do you mean how different address spaces share the same memory? This is done by the concept of a 'dataspace' - meaning a chunk of memory, that can be attached to different address spaces. The whole dataspace handling, paging etc. are all done by core. Nevertheless, you can setup your own dataspaces with the help of core and provide them to your clients, like a filesystem server would do. Does this answers your question?
I've tried out the following but this does not work:
// for now we copy data from core's to our local rom service Genode::Dataspace_capability file_cap; Rom_connection rom(filename); try { file_cap = rom.dataspace(); } catch (Rom_file_missing) { printf("Error: Could not access file \"%s\" from ROMservice.\n", filename); }
void* addr = env()->rm_session()->attach(file_cap); Genode::Dataspace_client dsc_source(file_cap); /* create new Rom_module */ Genode::Ram_dataspace *ram_dsc = new (env()->heap())Genode::Ram_dataspace(env()->ram_session(), dsc_source.size()); Genode::Dataspace_client dsc(ram_dsc->cap());
When returning the 'rom_connection' dataspace to the client, when it calls 'rom.dataspace()' (see the start of your mail), do you return: 'ram_dsc->cap()' (see above)? Otherwise, it will not work.
Genode::memcpy(ram_dsc->local_addr<void>(), addr,dsc_source.size());
Rom_module *r = new (&_sliced_heap) Rom_module(dsc.phys_addr(),dsc_source.size(), filename); PDBG("rom module filename: %s", filename); PDBG("rom module base address: 0x%lx", dsc_source.phys_addr()); PDBG("rom module size: 0x%zx", dsc_source.size()); PDBG("dataspace base address: 0x%lx", dsc.phys_addr()); PDBG("some data: 0x%x", *(int*)ram_dsc->local_addr<void>()); _rom_fs.insert(r); _rom_fs.print_fs();
The output in this context is ok: rom module filename: pci_drv rom module base address: 0x3140000 rom module size: 0x26000 dataspace base address: 0x3333000 some data: 0x464c457f
Is there an Allocator that I can use?
An allocator what for? In general there are different allocators defined in Genode.
I hope I could help you a bit, otherwise you should post some more details. Especially the whole 'rom_session_component' creation code would be helpful.
kind regards Stefan
Thx,
Oliver
--- Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Genode-main mailing list Genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main