[RFC PATCH linux dev-4.7 4/6] misc: Add Aspeed BT IPMI BMC driver

Cédric Le Goater clg at kaod.org
Fri Sep 16 00:02:03 AEST 2016


On 09/15/2016 01:33 PM, Patrick Williams wrote:
> On Thu, Sep 15, 2016 at 01:22:50PM +0200, Cédric Le Goater wrote:
>> On 09/15/2016 12:51 PM, Patrick Williams wrote:
>>> On Thu, Sep 15, 2016 at 10:36:46AM +0200, Cédric Le Goater wrote:
>>>>>> I'm not sure there. When I open /dev/ttyUSB0 I'm talking to the USB0
>>>>>> serial device. When I open /dev/mtd0, I'm talking to the mtd0 device.
>>>>>> So when I open /dev/ipmi-bt-bmc, I would expect to talk to the BMC
>>>>>> device?
>>>>>
>>>>> Yes. I went a little too far in the global rename. /dev/ipmi-bt-bmc makes 
>>>>
>>>> I meant /dev/ipmi-bt-host ! :)
>>>
>>> Does it make sense to have these numbered?  I thought Brenden proposed
>>> an emulated bt-over-i2c.  We could also conceive of an SOC that had
>>> multiple LPC (+bt) engines to facilitate multiple nodes, or an off-board
>>> FPGA to do it.
>>>
>>
>> If we expect to have multiple btbridged process running, we will 
>> need multiple device nodes. How we should name them depends partly 
>> on the driver I think. 
>>
>> We could keep the prefix 'ipmi-bt' because it identifies the BT 
>> interface described in the IPMI specs and use node /dev/ipmi-bt-host 
>> for the original bt_bmc, (used be bt_host) driver. This one will be 
>> unique. 
>>
>> The others could use : 
>> 	
>> 		/dev/ipmi-bt-lpc
>> 		/dev/ipmi-bt-i2c-<bus>
> 
> I was thinking along the lines of /dev/ipmi-bt-hostN.  Maybe we do this
> through udev rules though?

yes. this look like the best approach. 

C.


More information about the openbmc mailing list