Re: [PATCH v9] mac_oldworld: Allow loading binary ROM image

[Philippe Mathieu-Daudé]

On 10/17/20 6:31 PM, BALATON Zoltan via wrote:
> On Sat, 17 Oct 2020, Philippe Mathieu-Daudé wrote:
> > +Alistair for loader
> >
> > On 10/17/20 5:47 PM, BALATON Zoltan via wrote:
> > > The beige G3 Power Macintosh has a 4MB firmware ROM. Fix the size of
> > > the rom region and fall back to loading a binary image with -bios if
> > > loading ELF image failed. This allows testing emulation with a ROM
> > > image from real hardware as well as using an ELF OpenBIOS image.
> > >
> > > Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
> > > Reviewed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
> > > ---
> > > v9: Revert change from v8, back to the same as v7 rebased on latest
> > >
> > >   hw/ppc/mac_oldworld.c | 29 ++++++++++++++++++++---------
> > >   1 file changed, 20 insertions(+), 9 deletions(-)
> > >
> > > diff --git a/hw/ppc/mac_oldworld.c b/hw/ppc/mac_oldworld.c
> > > index 05e46ee6fe..0117ae17f5 100644
> > > --- a/hw/ppc/mac_oldworld.c
> > > +++ b/hw/ppc/mac_oldworld.c
> > > @@ -59,6 +59,8 @@
> > >   #define NDRV_VGA_FILENAME "qemu_vga.ndrv"
> > >     #define GRACKLE_BASE 0xfec00000
> > > +#define PROM_BASE 0xffc00000
> > > +#define PROM_SIZE (4 * MiB)
> > >     static void fw_cfg_boot_set(void *opaque, const char *boot_device,
> > >                               Error **errp)
> > > @@ -100,6 +102,7 @@ static void ppc_heathrow_init(MachineState *machine)
> > >       SysBusDevice *s;
> > >       DeviceState *dev, *pic_dev;
> > >       BusState *adb_bus;
> > > +    uint64_t bios_addr;
> > >       int bios_size;
> > >       unsigned int smp_cpus = machine->smp.cpus;
> > >       uint16_t ppc_boot_device;
> > > @@ -128,24 +131,32 @@ static void ppc_heathrow_init(MachineState 
> > > *machine)
> > >         memory_region_add_subregion(sysmem, 0, machine->ram);
> > >   -    /* allocate and load BIOS */
> > > -    memory_region_init_rom(bios, NULL, "ppc_heathrow.bios", BIOS_SIZE,
> > > +    /* allocate and load firmware ROM */
> > > +    memory_region_init_rom(bios, NULL, "ppc_heathrow.bios", PROM_SIZE,
> > >                              &error_fatal);
> > > +    memory_region_add_subregion(sysmem, PROM_BASE, bios);
> > >   -    if (bios_name == NULL)
> > > +    if (!bios_name) {
> > >           bios_name = PROM_FILENAME;
> > > +    }
> > >       filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
> > > -    memory_region_add_subregion(sysmem, PROM_ADDR, bios);
> > > -
> > > -    /* Load OpenBIOS (ELF) */
> > >       if (filename) {
> > > -        bios_size = load_elf(filename, NULL, 0, NULL, NULL, NULL, 
> > > NULL, NULL,
> > > -                             1, PPC_ELF_MACHINE, 0, 0);
> > > +        /* Load OpenBIOS (ELF) */
> > > +        bios_size = load_elf(filename, NULL, NULL, NULL, NULL, 
> > > &bios_addr,
> > > +                             NULL, NULL, 1, PPC_ELF_MACHINE, 0, 0);
> > > +        /* Unfortunately, load_elf sign-extends reading elf32 */
> >
> > Maybe this is what translate_fn() is for?
> >
> > uint64_t oldworld_phys(void *opaque, uint64_t addr)
> > {
> >    return addr & UINT32_MAX;
> > }
> >
> > Using as (untested):
> >
> >        bios_size = load_elf(filename, NULL, oldworld_phys, NULL,
> >                             NULL, &bios_addr, NULL,
> >                             NULL, 1, PPC_ELF_MACHINE, 0, 0);
>
> Please don't come up with any more great ideas for this patch unless you 
> also propose a replacement and test it. This one works and we could just 
> get this in as it is until the real problem with load_elf is fixed at 
> which point all this can be removed so no need to be more sophisticated 
> as the simple cast I have.

Zoltan, I'm not trying to block your patch to get merged,
I'm asking because I'm trying to understand how this API
is expected to be used.

> As you can see in the original discussion:
>
> http://patchwork.ozlabs.org/project/qemu-devel/patch/c69a791c7cad1246f3f34b3993dee4f549b75aa2.1593456926.git.balaton@eik.bme.hu/ 
>
>
> problem is really in include/hw/elf_ops.h this is just a work around for 
> that as I did not want to break anything I can't test so I'd rather fix 
> it up here and let you fix load_elf then drop this cast. But unless you 
> can do that before the freeze please don't hold up this patch any more.
>
> Regards,
> BALATON Zoltan
>
> > > +        bios_addr = (uint32_t)bios_addr;
> > > +
> > > +        if (bios_size <= 0) {
> > > +            /* or load binary ROM image */
> > > +            bios_size = load_image_targphys(filename, PROM_BASE, 
> > > PROM_SIZE);
> > > +            bios_addr = PROM_BASE;
> > > +        }
> > >           g_free(filename);
> > >       } else {
> > >           bios_size = -1;
> > >       }
> > > -    if (bios_size < 0 || bios_size > BIOS_SIZE) {
> > > +    if (bios_size < 0 || bios_addr - PROM_BASE + bios_size > 
> > > PROM_SIZE) {
> > >           error_report("could not load PowerPC bios '%s'", bios_name);
> > >           exit(1);
> > >       }