Re: [PATCH 2/2] xen: cleanup unrealized flash devices

[Markus Armbruster]

Philippe Mathieu-Daudé <philmd@redhat.com> writes:

> On 6/30/20 5:44 PM, Paul Durrant wrote:
>>> -----Original Message-----
>>> From: Philippe Mathieu-Daudé <philmd@redhat.com>
>>> Sent: 30 June 2020 16:26
>>> To: Paul Durrant <paul@xen.org>; xen-devel@lists.xenproject.org; 
>>> qemu-devel@nongnu.org
>>> Cc: Eduardo Habkost <ehabkost@redhat.com>; Michael S. Tsirkin 
>>> <mst@redhat.com>; Paul Durrant
>>> <pdurrant@amazon.com>; Jason Andryuk <jandryuk@gmail.com>; Paolo Bonzini 
>>> <pbonzini@redhat.com>;
>>> Richard Henderson <rth@twiddle.net>
>>> Subject: Re: [PATCH 2/2] xen: cleanup unrealized flash devices
>>>
>>> On 6/24/20 2:18 PM, Paul Durrant wrote:
>>>> From: Paul Durrant <pdurrant@amazon.com>
>>>>
>>>> The generic pc_machine_initfn() calls pc_system_flash_create() which 
>>>> creates
>>>> 'system.flash0' and 'system.flash1' devices. These devices are then 
>>>> realized
>>>> by pc_system_flash_map() which is called from pc_system_firmware_init() 
>>>> which
>>>> itself is called via pc_memory_init(). The latter however is not called 
>>>> when
>>>> xen_enable() is true and hence the following assertion fails:
>>>>
>>>> qemu-system-i386: hw/core/qdev.c:439: qdev_assert_realized_properly:
>>>> Assertion `dev->realized' failed
>>>>
>>>> These flash devices are unneeded when using Xen so this patch avoids the
>>>> assertion by simply removing them using pc_system_flash_cleanup_unused().
>>>>
>>>> Reported-by: Jason Andryuk <jandryuk@gmail.com>
>>>> Fixes: ebc29e1beab0 ("pc: Support firmware configuration with -blockdev")
>>>> Signed-off-by: Paul Durrant <pdurrant@amazon.com>
>>>> Tested-by: Jason Andryuk <jandryuk@gmail.com>
>>>> ---
>>>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>>>> Cc: Richard Henderson <rth@twiddle.net>
>>>> Cc: Eduardo Habkost <ehabkost@redhat.com>
>>>> Cc: "Michael S. Tsirkin" <mst@redhat.com>
>>>> Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
>>>> ---
>>>>  hw/i386/pc_piix.c    | 9 ++++++---
>>>>  hw/i386/pc_sysfw.c   | 2 +-
>>>>  include/hw/i386/pc.h | 1 +
>>>>  3 files changed, 8 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
>>>> index 1497d0e4ae..977d40afb8 100644
>>>> --- a/hw/i386/pc_piix.c
>>>> +++ b/hw/i386/pc_piix.c
>>>> @@ -186,9 +186,12 @@ static void pc_init1(MachineState *machine,
>>>>      if (!xen_enabled()) {
>>>>          pc_memory_init(pcms, system_memory,
>>>>                         rom_memory, &ram_memory);
>>>> -    } else if (machine->kernel_filename != NULL) {
>>>> -        /* For xen HVM direct kernel boot, load linux here */
>>>> -        xen_load_linux(pcms);
>>>> +    } else {
>>>> +        pc_system_flash_cleanup_unused(pcms);
>>>
>>> TIL pc_system_flash_cleanup_unused().
>>>
>>> What about restricting at the source?
>>>
>> 
>> And leave the devices in place? They are not relevant for Xen, so why not 
>> clean up?
>
> No, I meant to not create them in the first place, instead of
> create+destroy.

Better.  Opinion, not demand :)

[...]