Bochs running ROM of a real BIOS

[sskras]

#148 inspired a thought: could Bochs run ROM of a real BIOS?
Just like PCem or 86Box does:

Some BIOS ROMs: https://github.com/86Box/roms/tree/master/machines

[stlintel]

Yes, at some day I was able to run and boot Win95 with real asus_p6np5.bin

[stlintel]

[stlintel]

But it dies a little later without being able to recognize Windows 98 boot disk

[stlintel]

Actually now booted it successfully after entering to the setup and going though HDD auto detection

[fysnet]

I too got Bochs to boot Win98 with a very similar Asus BIOS.
By default, the BIOS boots A,C and I have to (as you did) set the HDD Auto Detection.
Since Bochs doesn't save these settings on exit from the BIOS, you have to reset these every time you boot.

Therefore, I should use the the cmosimage: file=cmos.img, rtc_init=time0 option, so that the CMOS state is saved on exit.
However, if I don't have the cmos.img file created, Bochs panics. Maybe rather than panic, it should gracefully create the file and load the default CMOS values as if the file was never specified. This would allow you to save them on exit.
If not, an example cmos.img file should be provided, correct?

Does this sound like a reasonable addition (I am willing to add) or am I missing something?

[stlintel]

if you just press 'continue' after PANIC it works ?
Because Bochs knows how to act if there is no cmosimage option specified at all.
I believe if cmos.img file is not found it should act the same.
Eventually the file will be created and will be found.

[vruppert]

It's not only the CMOS RAM. A real BIOS writes configuration data into the flash ROM, so we would need to write back data to the ROM image. Using save/restore doesn't help, since it only saves the RAM contents.

[fysnet]

if you just press 'continue' after PANIC it works ?

No, it doesn't. I had tried this too.

Bochs/bochs/iodev/cmos.cc

Lines 202 to 214 in 9216d22

	fd = open(SIM->get_param_string(BXPN_CMOSIMAGE_PATH)->getptr(), O_RDONLY
	#ifdef O_BINARY
	| O_BINARY
	#endif
	);
	if (fd < 0) {
	BX_PANIC(("trying to open cmos image file '%s'",
	SIM->get_param_string(BXPN_CMOSIMAGE_PATH)->getptr()));
	}
	ret = fstat(fd, &stat_buf);
	if (ret) {
	BX_PANIC(("CMOS: could not fstat() image file."));
	}

If file not found, at line 208 it Panics. If I hit continue, at line 211 it tries to get the fstat of the opened file. Since it cannot, it should panic again at line 213. However, in Win10, the fstat function crashes Bochs instead of returning error.

It's not only the CMOS RAM. A real BIOS writes configuration data into the flash ROM

Indeed this is true, but I think we still need to do something about the CMOS binary file.

I think it would be a simple task to save/restore the BIOS/CMOS images as well as save the BIOS image on exit. A .bochsrc config flag could even be used to indicate to do this or not.

Just my thoughts on the matter.
Thanks

[vruppert]

I have now added the CMOS image option to the runtime menu. Now you can start without an image and then save the image on exit. In addition to this, we possibly need an "autosave" parameter to start with an image and not saving it on exit.

[vruppert]

Please have a look at pull request #162. It adds save/restore support for the ESCD space of the flash BIOS ROM.

[fysnet]

Thank you. I will have a look.

[vruppert]

Pull request #162 is already merged and I have now added the parameter 'flash_data' to the romimage option. Flash BIOS config data is loaded if file exists and saved on exit if modified.