[Qemu-devel] Fwd: Cosimulation

[Qemu-devel] Fwd: Cosimulation

[Alessandro Corradi]

[-- Attachment #1: Type: text/plain, Size: 1997 bytes --]

Hi Nicolas,
Thank you for reply! The "only" trouble is that I need to descrive it as a
pci device. Taking the parallel port as an example is  very usefull for
simple  hardware  as you  made.  But I need some info for write my device
into qemu, in particular the init function for pci device as the ne2000 (In
init function of network card it set pci_conf, should I need to set up? And
what arguments should I put in pci_register_io_region function?).
My emulated device for the moment need to run in i386 architeture so I have
not to be completeness about cpu target etc...
Anyway I appreciate if you can send me your written device so I can see if
it can help me!

Thanks

Ale

2006/8/22, Nicolas SAUZEDE-NONST <nicolas.sauzede-nonst@st.com>:
>
> Hi Alessandro,
>
> Concerning your cosimulation problem, I may help a little bit,
> because when you first posted on the qemu dev list about your sysc
> project,
> it motivated me to first learn about systemc (which I didn't know) and to
> start hacking qemu to connect them together.
>
> Well I must admit I'm pretty satisfied of the results,
> I managed to build some small sysc dumb devices (gates, counters,
> ram/buffers, fpu etc..)
> (with a FIR filter in mind indeed), defined some sort of sysc written
> "interface" that offers a bus+signals to sysc, while connecting to qemu
> via a socket.
> Within qemu, I created a parallel-like device (as you said on the ml
> recently) and so I can probe my device from within guest OS (dos, linux,
> etc..)
> the device is called remote.c (like a "remote" parallel-like device)
>
> If you want I can send you some rough version of my project to see how
> (ugly) I did
> and maybe we can share our experiences...
>
> Regards,
>
> NS.
>
> --
> Best regards,
> NS.
> ************************************
> Nicolas SAUZEDE -- Subco
> ST Microelectronics Grenoble
> 12, rue Jules Horowitz, BP 217
> 38019 Grenoble Cedex
> STB Tel : 041 5142
> EXT Tel : 0476585142
> ***********************************
>
>

[-- Attachment #2: Type: text/html, Size: 2379 bytes --]

[Qemu-devel] systemc trial

[Nicolas SAUZEDE-NONST]

[-- Attachment #1.1: Type: text/plain, Size: 3126 bytes --]

Hi Alessandro,

Ah OK, I see, your project seems to be way more ambitious than mine :-)
So my project may not help you as much as you could expect..
Maybe only the sysc/tcpip glue mechanism..

Anyway, attached is a tarball of my project, containing a patch against 
qemu-snapshot-2006-07-23_23 (I can provide it if it fails to patch 
against more recent versions) plus the directory related to sysc.

See sysc/README for more info..

Will be interested by your return.. and also by your PCI/sysc 
experiments also..
tell me if you manage to compile/run it (watch out the sysc signal trace 
log which is generated)

Can send me something from your project, just by curiosity ? :-)
I would be particularly interested by other sysc/qemu glue mechanism, 
and also to see if you managed to generate interrupts from sysc, and 
handled within qemu guest OS...

Good luck for your project,

NS.

ps: oops, I was not supposed to send my professional signature in the 
mail I sent you (fwd to the ML), well nevermind, I'll use another mail 
identity from now on.. :-)
pps: I also fwd it to the ML, in order for other interested people to 
have a look at it if anyone is interested..

Alessandro Corradi a écrit :
> Hi Nicolas,
> Thank you for reply! The "only" trouble is that I need to descrive it 
> as a pci device. Taking the parallel port as an example is  very 
> usefull for  simple  hardware  as you  made.  But I need some info for 
> write my device into qemu, in particular the init function for pci 
> device as the ne2000 (In init function of network card it set 
> pci_conf, should I need to set up? And what arguments should I put in 
> pci_register_io_region function?).
> My emulated device for the moment need to run in i386 architeture so I 
> have not to be completeness about cpu target etc...
> Anyway I appreciate if you can send me your written device so I can 
> see if it can help me!
>
> Thanks
>
> Ale
>
> 2006/8/22, Nicolas SAUZEDE:
>
>     Hi Alessandro,
>
>     Concerning your cosimulation problem, I may help a little bit,
>     because when you first posted on the qemu dev list about your sysc
>     project,
>     it motivated me to first learn about systemc (which I didn't know)
>     and to
>     start hacking qemu to connect them together.
>
>     Well I must admit I'm pretty satisfied of the results,
>     I managed to build some small sysc dumb devices (gates, counters,
>     ram/buffers, fpu etc..)
>     (with a FIR filter in mind indeed), defined some sort of sysc written
>     "interface" that offers a bus+signals to sysc, while connecting to
>     qemu
>     via a socket.
>     Within qemu, I created a parallel-like device (as you said on the ml
>     recently) and so I can probe my device from within guest OS (dos,
>     linux,
>     etc..)
>     the device is called remote.c (like a "remote" parallel-like device)
>
>     If you want I can send you some rough version of my project to see how
>     (ugly) I did
>     and maybe we can share our experiences...
>
>     Regards,
>
>     NS.
>
>

[-- Attachment #1.2: Type: text/html, Size: 3700 bytes --]

[-- Attachment #2: sysc0.tar.gz --]
[-- Type: application/gzip, Size: 11080 bytes --]

[Qemu-devel] Re: Cosimulation

[Alessandro Corradi]

[-- Attachment #1: Type: text/plain, Size: 2631 bytes --]

Hi nicolas,

Thank you very much for your reply. In attachment you find the sc dscription
of the cirrus logic, The problem is that it is implement with a different
version of SystemC built by my university. In fact the card description
isn't a problem for me.
This description concern the emulation of 8 register and the communication
is via socket trought a message struct. The open connection is made at init
function of the card (I think) and in read/write function
(..._ioport_read/write) there are the message send/receive via socket as you
write for your remote device.
The point is that I need to be recognize by qemu and more important
recognize by debian device driver of the cirrus. I think this will be
impossbile for me :) but I try!

There are some open point for me:

1) In read/write functions i need only to descrive the operation on the
register, right? As in parallel.c, for each register there is the ioctl call
in base of the register (status, data...) I think i only need to do the
same;

2) The init function is a full open point... I wrote than there is the
socket connection, but I don't know if I need to recognize it as a pci-card.
Can somebody help me?

3) Interrupt implementation: I need only to implement ..._update_irq right?
It seems no problem, but I want to be secure about it.

2006/8/22, Nicolas SAUZEDE-NONST <nicolas.sauzede-nonst@st.com>:
>
> Hi Alessandro,
>
> Concerning your cosimulation problem, I may help a little bit,
> because when you first posted on the qemu dev list about your sysc
> project,
> it motivated me to first learn about systemc (which I didn't know) and to
> start hacking qemu to connect them together.
>
> Well I must admit I'm pretty satisfied of the results,
> I managed to build some small sysc dumb devices (gates, counters,
> ram/buffers, fpu etc..)
> (with a FIR filter in mind indeed), defined some sort of sysc written
> "interface" that offers a bus+signals to sysc, while connecting to qemu
> via a socket.
> Within qemu, I created a parallel-like device (as you said on the ml
> recently) and so I can probe my device from within guest OS (dos, linux,
> etc..)
> the device is called remote.c (like a "remote" parallel-like device)
>
> If you want I can send you some rough version of my project to see how
> (ugly) I did
> and maybe we can share our experiences...
>
> Regards,
>
> NS.
>
> --
> Best regards,
> NS.
> ************************************
> Nicolas SAUZEDE -- Subco
> ST Microelectronics Grenoble
> 12, rue Jules Horowitz, BP 217
> 38019 Grenoble Cedex
> STB Tel : 041 5142
> EXT Tel : 0476585142
> ***********************************
>
>

[-- Attachment #2: Type: text/html, Size: 2990 bytes --]