"Michael S. Tsirkin" <m...@redhat.com> writes:
> On Mon, Jul 22, 2013 at 01:38:47PM +0200, Markus Armbruster wrote:
>> We set default boot order "cad" in every single machine definition
>> except "pseries" and "moxiesim", even though very few boards actually
>> care for boot order, and "cad" makes sense for even fewer.
>>
>> Machines that care:
>>
>> * pc and its variants
>>
>> Accept up to three letters 'a', 'b' (undocumented alias for 'a'),
>> 'c', 'd' and 'n'. Reject all others (fatal with -boot).
>>
>> * nseries (n800, n810)
>>
>> Check whether order starts with 'n'. Silently ignored otherwise.
>>
>> * prep, g3beige, mac99
>>
>> Extract the first character the machine understands (subset of
>> 'a'..'f'). Silently ignored otherwise.
>>
>> * spapr
>>
>> Accept an arbitrary string (vl.c restricts it to contain only
>> 'a'..'p', no duplicates).
>>
>> * sun4[mdc]
>>
>> Use the first character. Silently ignored otherwise.
>>
>> Strip characters these machines ignore from their default boot order.
>>
>> For all other machines, remove the unused default boot order
>> alltogether.
>>
>> Note that my rename of QEMUMachine member boot_order to
>> default_boot_order and QEMUMachineInitArgs member boot_device to
>> boot_order has a welcome side effect: it makes every use of boot
>> orders visible in this patch, for easy review.
>>
>> Signed-off-by: Markus Armbruster <arm...@redhat.com>
>> ---
[...]
>> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
>> index 9327ac1..3700bd5 100644
>> --- a/hw/i386/pc_piix.c
>> +++ b/hw/i386/pc_piix.c
>> @@ -216,7 +216,7 @@ static void pc_init1(QEMUMachineInitArgs *args,
>> }
>> }
>>
>> - pc_cmos_init(below_4g_mem_size, above_4g_mem_size, args->boot_device,
>> + pc_cmos_init(below_4g_mem_size, above_4g_mem_size, args->boot_order,
>> floppy, idebus[0], idebus[1], rtc_state);
>>
>> if (pci_enabled && usb_enabled(false)) {
>> @@ -324,7 +324,7 @@ static QEMUMachine pc_i440fx_machine_v1_6 = {
>> .hot_add_cpu = pc_hot_add_cpu,
>> .max_cpus = 255,
>> .is_default = 1,
>> - DEFAULT_MACHINE_OPTIONS,
>> + .default_boot_order = "cad",
>> };
>>
>> static QEMUMachine pc_i440fx_machine_v1_5 = {
>> @@ -337,7 +337,7 @@ static QEMUMachine pc_i440fx_machine_v1_5 = {
>> PC_COMPAT_1_5,
>> { /* end of list */ }
>> },
>> - DEFAULT_MACHINE_OPTIONS,
>> + .default_boot_order = "cad",
>> };
>>
>> static QEMUMachine pc_i440fx_machine_v1_4 = {
>
> So all PC machine types share this?
Correct, just like they share DEFAULT_MACHINE_OPTIONS before my patch.
Which is defined as
#define DEFAULT_MACHINE_OPTIONS \
.boot_order = "cad"
I.e. my patch merely peels off a layer of obfuscation :)
> Can we set this in some common code, somehow?
We don't have an inheritance notion for machine types.
vl.c uses machine->boot_order before calling one of its methods, so
monkey-patching .boot_order from a method won't do. Besides, that cure
looks much worse than the disease to me.
Can't think of anything else offhand.
[...]
