Re: [Qemu-devel] [PATCH v3 4/4] bootdevice: update boot_order in MachineState

[Gonglei <arei.gonglei@huawei.com>]

On 2015/2/12 18:21, Markus Armbruster wrote:
> <arei.gonglei@huawei.com> writes:
> 
>> From: Dinar Valeev <dvaleev@suse.com>
>>
>> on sPAPR we need to update boot_order in MachineState in case it
>> got changed on reset.
>>
>> Signed-off-by: Dinar Valeev <dvaleev@suse.com>
>> Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>> Signed-off-by: Gonglei <arei.gonglei@huawei.com>
>> ---
>>  bootdevice.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
>> diff --git a/bootdevice.c b/bootdevice.c
>> index d3d4277..ac586b1 100644
>> --- a/bootdevice.c
>> +++ b/bootdevice.c
>> @@ -26,6 +26,7 @@
>>  #include "qapi/visitor.h"
>>  #include "qemu/error-report.h"
>>  #include "hw/hw.h"
>> +#include "hw/boards.h"
>>  
>>  typedef struct FWBootEntry FWBootEntry;
>>  
>> @@ -50,6 +51,7 @@ void qemu_register_boot_set(QEMUBootSetHandler *func, void *opaque)
>>  void qemu_boot_set(const char *boot_order, Error **errp)
>>  {
>>      Error *local_err = NULL;
>> +    MachineState *machine = MACHINE(qdev_get_machine());
>>  
>>      validate_bootdevices(boot_order, &local_err);
>>      if (local_err) {
>> @@ -57,6 +59,8 @@ void qemu_boot_set(const char *boot_order, Error **errp)
>>          return;
>>      }
>>  
>> +    machine->boot_order = boot_order;
>> +
>>      if (boot_set_handler) {
>>          boot_set_handler(boot_set_opaque, boot_order, errp);
>>      }
> 
> Looks harmless enough, but I'm afraid I don't quite understand why we
> need this.  Can you explain it to me real slow?  :)
> 
Alex pointed that there is two reset logic for boot order, one is machine struct
(machine->order) and another is set_boot_handler by parameter. But the
qemu_boot_set() only support to use set_boot_handler to set boot order.
So, I decide to remove the check for boot_set_handler if NULL or not (PATCH 1),
and change the value of machine->boot_order in qemu_boot_set().

But I ignored this change cause a regression for HMP command set_boot. To avoid
the regression, we only support one way to set boot_order, that is set_boot_handler.
Fortunately the patches for supporting boot 'once' argument on sPAPR on the flight. We
have the opportunity to drop this change and avoid the regression.

Regards,
-Gonglei