Re: [PATCH v2 01/11] accel/kvm: Check MachineClass kvm_type() return value

[Philippe Mathieu-Daudé]

On 2/23/21 11:36 AM, Cornelia Huck wrote:
> On Tue, 23 Feb 2021 10:37:01 +1100
> David Gibson <david@gibson.dropbear.id.au> wrote:
> 
>> On Tue, Feb 23, 2021 at 10:33:55AM +1100, David Gibson wrote:
>>> On Mon, Feb 22, 2021 at 06:50:44PM +0100, Cornelia Huck wrote:  
>>>> On Mon, 22 Feb 2021 18:41:07 +0100
>>>> Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
>>>>   
>>>>> On 2/22/21 6:24 PM, Cornelia Huck wrote:  
>>>>>> On Fri, 19 Feb 2021 18:38:37 +0100
>>>>>> Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
>>>>>>     
>>>>>>> MachineClass::kvm_type() can return -1 on failure.
>>>>>>> Document it, and add a check in kvm_init().
>>>>>>>
>>>>>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>>>>>>> ---
>>>>>>>  include/hw/boards.h | 3 ++-
>>>>>>>  accel/kvm/kvm-all.c | 6 ++++++
>>>>>>>  2 files changed, 8 insertions(+), 1 deletion(-)
>>>>>>>
>>>>>>> diff --git a/include/hw/boards.h b/include/hw/boards.h
>>>>>>> index a46dfe5d1a6..68d3d10f6b0 100644
>>>>>>> --- a/include/hw/boards.h
>>>>>>> +++ b/include/hw/boards.h
>>>>>>> @@ -127,7 +127,8 @@ typedef struct {
>>>>>>>   *    implement and a stub device is required.
>>>>>>>   * @kvm_type:
>>>>>>>   *    Return the type of KVM corresponding to the kvm-type string 
>>>>>>> option or
>>>>>>> - *    computed based on other criteria such as the host kernel 
>>>>>>> capabilities.
>>>>>>> + *    computed based on other criteria such as the host kernel 
>>>>>>> capabilities
>>>>>>> + *    (which can't be negative), or -1 on error.
>>>>>>>   * @numa_mem_supported:
>>>>>>>   *    true if '--numa node.mem' option is supported and false otherwise
>>>>>>>   * @smp_parse:
>>>>>>> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
>>>>>>> index 84c943fcdb2..b069938d881 100644
>>>>>>> --- a/accel/kvm/kvm-all.c
>>>>>>> +++ b/accel/kvm/kvm-all.c
>>>>>>> @@ -2057,6 +2057,12 @@ static int kvm_init(MachineState *ms)
>>>>>>>                                                              "kvm-type",
>>>>>>>                                                              
>>>>>>> &error_abort);
>>>>>>>          type = mc->kvm_type(ms, kvm_type);
>>>>>>> +        if (type < 0) {
>>>>>>> +            ret = -EINVAL;
>>>>>>> +            fprintf(stderr, "Failed to detect kvm-type for machine 
>>>>>>> '%s'\n",
>>>>>>> +                    mc->name);
>>>>>>> +            goto err;
>>>>>>> +        }
>>>>>>>      }
>>>>>>>  
>>>>>>>      do {    
>>>>>>
>>>>>> No objection to this patch; but I'm wondering why some non-pseries
>>>>>> machines implement the kvm_type callback, when I see the kvm-type
>>>>>> property only for pseries? Am I holding my git grep wrong?    
>>>>>
>>>>> Can it be what David commented here?
>>>>> https://www.mail-archive.com/qemu-devel@nongnu.org/msg784508.html
>>>>>   
>>>>
>>>> Ok, I might be confused about the other ppc machines; but I'm wondering
>>>> about the kvm_type callback for mips and arm/virt. Maybe I'm just
>>>> confused by the whole mechanism?  
>>>
>>> For ppc at least, not sure about in general, pseries is the only
>>> machine type that can possibly work under more than one KVM flavour
>>> (HV or PR).  So, it's the only one where it's actually useful to be
>>> able to configure this.  
>>
>> Wait... I'm not sure that's true.  At least theoretically, some of the
>> Book3E platforms could work with either PR or the Book3E specific
>> KVM.  Not sure if KVM PR supports all the BookE instructions it would
>> need to in practice.
>>
>> Possibly pseries is just the platform where there's been enough people
>> interested in setting the KVM flavour so far.
> 
> If I'm not utterly confused by the code, it seems the pseries machines
> are the only ones where you can actually get to an invocation of
> ->kvm_type(): You need to have a 'kvm-type' machine property, and
> AFAICS only the pseries machine has that.

OMG you are right... This changed in commit f2ce39b4f06
("vl: make qemu_get_machine_opts static"):

@@ -2069,13 +2068,11 @@ static int kvm_init(MachineState *ms)
     }
     s->as = g_new0(struct KVMAs, s->nr_as);

-    kvm_type = qemu_opt_get(qemu_get_machine_opts(), "kvm-type");
-    if (mc->kvm_type) {
+    if (object_property_find(OBJECT(current_machine), "kvm-type")) {
+        g_autofree char *kvm_type =
object_property_get_str(OBJECT(current_machine),
+                                                            "kvm-type",
+                                                            &error_abort);
         type = mc->kvm_type(ms, kvm_type);
-    } else if (kvm_type) {
-        ret = -EINVAL;
-        fprintf(stderr, "Invalid argument kvm-type=%s\n", kvm_type);
-        goto err;
     }

Paolo, is that expected?

So these callbacks are dead code:
hw/arm/virt.c:2585:    mc->kvm_type = virt_kvm_type;
hw/mips/loongson3_virt.c:625:    mc->kvm_type = mips_kvm_type;
hw/ppc/mac_newworld.c:598:    mc->kvm_type = core99_kvm_type;
hw/ppc/mac_oldworld.c:447:    mc->kvm_type = heathrow_kvm_type;

> 
> (Or is something hiding behind some macro magic?)
>