Messages in this thread | | | Date | Wed, 4 Apr 2012 23:19:21 +0100 | From | Mark Brown <> | Subject | Re: [PATCH 5/6] regmap: add runtime PM calls to debugfs file IO |
| |
On Wed, Apr 04, 2012 at 03:48:32PM -0600, Stephen Warren wrote:
> It's quite probably that devices need to be active in order for their > registers to be read/written. In normal regmap usage by drivers, it's the > responsibility of the driver to assure this if needed. However, regmap > debugfs file handling is outside the driver's control, so we need to > explicitly ensure this.
This feels wrong, it's too invasive on the device. If we're in cache only mode (which we should be if the device is off and we have a cache) then I'd expect us to error out if we can't satisfy the read from cache rather than wake the device up which might be terribly expensive and cause register writes which change the device state we were trying to inspect, and if we have a cache then we might not need to interact with the hardware at all at which point the wakeup would've been needless. It's certainly not what I'd expect a diagnostic interface to be doing, I'd expect it to reflect the current state of the device.
Now, quite how the mmio bus deals with access to a suspended device is probably platform specific and certainly more fun than at least I2C (SPI has issues since it's hard to tell if the device is there or not and the MISO line might be undriven but at least you'll not lock up or anything). [unhandled content-type:application/pgp-signature] | |