On Fri, Oct 22, 2010 at 4:50 PM, Artyom Tarasenko atar4qemu@gmail.com wrote:
On Fri, Oct 22, 2010 at 6:38 PM, Tarl Neustädter tarl-b2@tarl.net wrote:
On 2010-10-22 5:28 AM, Artyom Tarasenko wrote:
On Fri, Oct 15, 2010 at 7:57 PM, Blue Swirlblauwirbel@gmail.com wrote:
Actually, while mc146818rtc is linked in, it's not used (there are some machines that have similar device, so that's why it still hangs around). m48t59 is on ebus.
Btw, can you explain how the physical address of an ebus device can be calculated? Looks like they are relative to the ebus beginning and then to PCI, but what constants define the offsets?
In real Sun systems, offsets onto EBus are hard-coded into the circuitry. E.g., a 16550 chip will be at offset x003f8. Somewhere in the upper bits (I think where I have the "x") will be a select value which is decoded to provide a select pin for the particular chip.
In answer to your basic question, the EBus address values are provided to firmware by the hardware designers.
Actually I meant "in qemu", but asked on a wrong mailing list, sorry for the noise. (Most of the developers are present in both lists anyway).
In QEMU, EBus provides an isa_mmio region, so all ISA devices with PIO show up as MMIO in the EBUS PCI region 0 . For example, the first 16550 serial port has a hard-coded PIO base of 0x3f8, so it can be accessed at 0x1fe020003f8.
Though actually also APB/PBM happens to route PCI PIO to this region, so maybe EBus has no effect ;-)
Is Netra-T1 OK?
Wanted to check whether the current ebus mappings match Netra-T1 or U5, but haven't succeeded.
I would generally suggest steering clear of the Netra systems - they are Telco modifications of more basic systems. Going with a netra may cause you to deal with additional stuff you don't care about.
Thanks for the hint! This can save us some time & troubles. Just out of curiosity - do you have an example for the additional stuff`?
These devices are not in U5: 'SUNW,lom', 'adc', 'beeper', 'display7seg', 'gpio', 'i2c'.