Suggested enhancements for pyb.mount()

[pythoncoder]

There is no obvious way for a user of a block device to format it. While it's trivial for anyone developing a block device, someone using a driver written by someone else might have problems. I'd suggest a force_format argument, defaulting False and only working if mkfs is True.

Secondly, and arguably a little specialised, but for developers of block devices it would be handy to have a way of registering object methods as callbacks when someone attempts to mount or unmount a device. The mount callback would be called at the start of a mount to initialise hardware, the umount one after a mount(None ...) was complete to de-initialise it. The unmount callback would look up the correct callback using the mountpoint.

A block device could include its own mount() method to achieve this (thereby losing a little generality) but perhaps it's food for thought?

[danicampora]

what about adding a new method

Code: Select all

pyb.mkfs()

or

Code: Select all

os.mkfs()

?

[Damien]

The semantics of pyb.mount(..., mkfs=True) are that it will make a new filesystem (ie format the device) only if one does not already exist. Maybe it should be changed to *always* make a new filesystem if mkfs=True. Would this satisfy the criteria of "format"?

[pythoncoder]

The only issue I can see with changing the semantics is that it might break (poorly written) existing code. But I think it's the neatest solution.

To expand on my point about callbacks. It's easy for the developer of a block device to detect that an attempt is being made to mount it: on any access check that the hardware is running, and if not, start it. However I can't see a clean way of detecting that it has been unmounted (or even determining where it is mounted). There are a number of reasons why you might want to detect an unmount: to conserve power, to save internal state such as flash block erase counts, or to issue an instruction to a processor local to the hardware.

The crude solution is for the driver to provide a shutdown function and rely on the user to call it - but it would be more elegant if the driver could detect the unmount and handle it. Incidentally this isn't a practical problem in my application; I'm merely floating a thought.