Re: [PATCH] runqemu: Add palmetto target

[Andrew Jeffery <andrew@aj.id.au>]

[-- Attachment #1: Type: text/plain, Size: 2661 bytes --]

On Thu, 2016-03-17 at 11:28 -0500, Patrick Williams wrote:
> On Thu, Mar 17, 2016 at 03:22:48PM +1030, Andrew Jeffery wrote:
> > diff --git a/yocto-poky/scripts/runqemu b/yocto-poky/scripts/runqemu
> > index 23cf5be..f09da50 100755
> > --- a/yocto-poky/scripts/runqemu
> > +++ b/yocto-poky/scripts/runqemu
> > @@ -110,7 +110,7 @@ while true; do
> >      arg=${1}
> >      case "$arg" in
> >          "qemux86" | "qemux86-64" | "qemuarm" | "qemuarm64" | "qemumips" | "qemumipsel" | \
> > -        "qemumips64" | "qemush4"  | "qemuppc" | "qemumicroblaze" | "qemuzynq")
> > +        "qemumips64" | "qemush4"  | "qemuppc" | "qemumicroblaze" | "qemuzynq" | "palmetto")
> 
> It seems like there should be a way to extend the machines in runqemu
> without modifying the script.  Maybe an environment variable like
> "palmetto_SUPPORTS_QEMU=1" that we can set in one of the bitbake machine
> files for each machine?

Yeah reading the script it did seem odd that there doesn't appear to be
a way to add qemu machine support.

> 
> Might want to bounce this problem off of #yocto on IRC.

Will do.

> 
> > +PALMETTO_DEFAULT_KERNEL=cuImage
> > +PALMETTO_DEFAULT_FSTYPE=cpio.gz
> 
> Also ideally have these come from a bitbake file.

So I was just following the convention here - but I'm taking your
comment was tied to the "it should be easier" issue above? I'll have to
see what my priorities are wrt reorganising the upstream scripts.

> 
> > +        "palmetto")
> > +            mem_size=512
> > +            ;;
> 
> > +if [ "$MACHINE" = "palmetto" ]; then
> > +    QEMU=qemu-system-arm
> > +    MACHINE_SUBTYPE=palmetto-bmc
> > +    export QEMU_AUDIO_DRV="none"
> > +    KERNCMDLINE="console=ttyS4"
> > +    QEMUOPTIONS="$QEMU_NETWORK_CMD -M ${MACHINE_SUBTYPE} -initrd $ROOTFS -no-reboot -nographic"
> > +fi
> > +
> 
> Would be nice if this stuff was also in a bitbake file of sorts.  Would
> really be nice if we could leverage the machine classes (palmetto is-a
> openpower) to reduce the redundancy.

I'm not enough of a bitbake master to know how easily that can be
achieved, but it sounds nice in theory. The script as it stands just
tests against the relevant 'qemu*' names and goes from there - the code
I've added just follows that same pattern. Eg, from the script:

    ...
    if [ "$MACHINE" = "qemuarm" -o "$MACHINE" = "qemuarmv6" -o "$MACHINE" = "qemuarmv7" ]; then
        QEMU=qemu-system-arm
        ...
    fi
    ...
    if [ "$MACHINE" = "qemumips" -o "$MACHINE" = "qemumipsel" -o "$MACHINE" = "qemumips64" ]; then
        case "$MACHINE" in
        ...
    fi
    ...

Andrew

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]