Re: [Qemu-block] [PATCH for-2.9 25/30] block: add a model option for MTD devices

[Kevin Wolf]

Am 30.11.2016 um 16:09 hat Cédric Le Goater geschrieben:
> On 11/29/2016 07:08 PM, Kevin Wolf wrote:
> > Am 29.11.2016 um 18:30 hat Cédric Le Goater geschrieben:
> >> On 11/29/2016 04:44 PM, Cédric Le Goater wrote:
> >>> This could be used to define the flash model to use on some boards
> >>> definitions.
> >>
> >> As this patch was part of a larger set, I did not send the whole 
> >> set to qemu-block@ list. Could you please take a look at the proposal ? 
> > 
> > This is a device level option rather than a block backend one. We messed
> > up -drive in its early days by including some device options, but we
> > don't generally want to add to this.
> > 
> > The correct way would be to add a qdev property for this and specify it
> > with -device or -global.
> 
> OK. I see. I should use something like this on the command line :
> 
>       -drive file=flash-romulus-test,format=raw,if=mtd,id=bmc \
>       -device mx25l25635e,drive=bmc \
>       -drive file=flash-romulus-test2,format=raw,if=mtd,id=bmc2 \
>       -device mx25l25635e,drive=bmc2 \
>       -drive file=romulus.pnor,format=raw,if=mtd,id=pnor \
>       -device mx66l1g45g,drive=pnor 

I think you mean if=none instead of if=mtd everywhere, now that you're
using an explicit -device.

> and retrieve the attached m25p80 device from the drive with the 
> routine blk_get_attached_dev(). That is changing a bit the way
> the platform is initialized but it is cleaner as no default 
> devices are automatically added.

Right. For compatibility (and convenience) you can and probably should
still support if=mtd to create the device automatically, but then the
user just gets the defaults for the options that -drive doesn't allow to
specify.

Kevin