Re: [PATCH 0/8] aspeed: I2C fixes, -drive removal (first step)

[Cédric Le Goater]

On 2/15/23 11:45, Philippe Mathieu-Daudé wrote:
> On 14/2/23 18:18, Cédric Le Goater wrote:
> > Hello,
> >
> > This series starts with a first set of patches fixing I2C slave mode
> > in the Aspeed I2C controller, a test device and its associated test in
> > avocado.
> >
> > Follow some cleanups which allow the use of block devices instead of
> > drives. So that, instead of specifying :
> >
> >    -drive file=./flash-ast2600-evb,format=raw,if=mtd
> >    -drive file=./ast2600-evb.pnor,format=raw,if=mtd
> >    ...
> >
> > and guessing from the order which bus the device is attached to, we
> > can use :
> >
> >    -blockdev node-name=fmc0,driver=file,filename=./bmc.img
> >    -device mx66u51235f,bus=ssi.0,drive=fmc0
> >    -blockdev node-name=fmc1,driver=file,filename=./bmc-alt.img
> >    -device mx66u51235f,bus=ssi.0,drive=fmc1
> >    -blockdev node-name=pnor,driver=file,filename=./pnor
> >    -device mx66l1g45g,bus=ssi.1,drive=pnor
> >    ...
> >
> > It is not perfect, the CS index still depends on the order
>
> Quick thoughts here:
>
> TYPE_SSI_PERIPHERAL devices have one input SSI_GPIO_CS.
>
> TYPE_SSI_BUS could have a "cs-num" property (how many
> CS line associated with this bus) and create an array of
> #cs-num output SSI_GPIO_CS.
>
> TYPE_SSI_PERIPHERAL could have a "cs" (index) property;
> if set, upon ssi_peripheral_realize() when the device is
> plugged on the bus, the GPIO line is wired.

yes. I would like to check first the impact on migration compatibility.

Thanks,

C.

> So we could set the 'cs=' property from CLI:
>
>    -blockdev node-name=fmc0,driver=file,filename=./bmc.img
>    -device mx66u51235f,bus=ssi.0,cs=1,drive=fmc0
>    -blockdev node-name=fmc1,driver=file,filename=./bmc-alt.img
>    -device mx66u51235f,bus=ssi.0,cs=0,drive=fmc1
>
> > but it is
> > now possible to run a machine without -drive ...,if=mtd.
> >
> > This lacks the final patch enabling the '-nodefaults' option by not
> > creating the default devices if specified on the command line. It
> > needs some more evaluation of the possible undesired effects.
> > Thanks,
> >
> > C.