On 11 August 2015 at 15:15, Peter Maydell <peter.mayd...@linaro.org> wrote: > This patchset updates the ancient pxa2xx_mmci device to something > resembling modern standards for devices. In particular it makes > it a proper sysbus device and switches to VMStateDescription structs. > > The major issue I have with this is in patch 1: > I wanted the device to have a property so its users can set > the BlockBackend* it should use. Unfortunately, DEFINE_PROP_DRIVE() > is no good here, because setting a drive property results in a > call to blk_attach_dev() which attaches the BlockBackend to this > device. That then means that the call in sd_init() to attach the > BlockBackend to the SD card object aborts. I needed a way to > have a drive property which didn't mean "and this device claims > the drive", and the best I could come up with was to use a > pointer property. Suggestions for better approaches welcome. > (The other SD controller devices are either also ancient non-QOM > devices, or use drive_get_next() in the init function...)
Ping! No opinions on this (or review of the patches)? thanks -- PMM