[i2c] [PATCH] I2C && DVB: fix recursive locking lockped warning in i2c_transfer()

[Jiri Kosina]

Hi Jean,

On Wed, 18 Oct 2006, Jean Delvare wrote:

> OK, I've now read this documentation file, things are now somewhat 
> clearer in my head. I had heard about the "lock class" theory before but 
> didn't remember all the details. I understand the good idea behind this 
> theory.

Thanks for looking at it.

> I just took a look at the usb code to see how they have been dealing 
> with it. They do have a structure member to store the lock level 
> (usb_device.level), it's included unconditionally. In the USB case, the 
> nested locking is only needed to handle power management, while for I2C 
> it is needed to even use the devices, but this shouldn't make much 
> difference. So it would probably be OK to do the same in the i2c 
> subsystem, which actually means taking your patch or something very 
> similar.

I agree, thanks.

> OK, after some more thought, I understand your point and take back my
> peremptory comment above ;)
> I'd like to have 2 different patches:
> 1* One adding the "functionality" to i2c-core. I propose to name the
> extra structure member "level" and to make it an u8, as was done in the
> usb subsystem, for consistency. This patch would go in my i2c tree.

Thanks, will send you a patch probably later today.

> 2* One changing the affected DVB drivers to make use of it. This patch 
> would go in the v4l-dvb tree. I'd let the driver and media/video 
> subsystem maintainers deal with it.

The only callpath which needs fixing I am currently aware of is the one I 
have provided patch for in the previous mail. So I will fix this one, and 
provide it to you in a separate patch.

Thanks,

-- 
Jiri Kosina